Datasets:

J0nasW

/

paperswithcode

like 1

a66f31b0-79e1-4a4d-bc78-6891a7727b67

distilling-the-knowledge-of-romanian-berts

2112.12650

https://arxiv.org/abs/2112.12650v3

https://arxiv.org/pdf/2112.12650v3.pdf

Distilling the Knowledge of Romanian BERTs Using Multiple Teachers

Running large-scale pre-trained language models in computationally constrained environments remains a challenging problem yet to be addressed, while transfer learning from these models has become prevalent in Natural Language Processing tasks. Several solutions, including knowledge distillation, network quantization, or network pruning have been previously proposed; however, these approaches focus mostly on the English language, thus widening the gap when considering low-resource languages. In this work, we introduce three light and fast versions of distilled BERT models for the Romanian language: Distil-BERT-base-ro, Distil-RoBERT-base, and DistilMulti-BERT-base-ro. The first two models resulted from the individual distillation of knowledge from two base versions of Romanian BERTs available in literature, while the last one was obtained by distilling their ensemble. To our knowledge, this is the first attempt to create publicly available Romanian distilled BERT models, which were thoroughly evaluated on five tasks: part-of-speech tagging, named entity recognition, sentiment analysis, semantic textual similarity, and dialect identification. Our experimental results argue that the three distilled models offer performance comparable to their teachers, while being twice as fast on a GPU and ~35% smaller. In addition, we further test the similarity between the predictions of our students versus their teachers by measuring their label and probability loyalty, together with regression loyalty - a new metric introduced in this work.

['Dan Tufiş', 'Vasile Păiş', 'Traian Rebedea', 'Mihai Dascălu', 'Dumitru-Clementin Cercel', 'Darius Catrina', 'Andrei-Marius Avram']

2021-12-23

https://aclanthology.org/2022.lrec-1.39

https://aclanthology.org/2022.lrec-1.39.pdf

lrec-2022-6

['dialect-identification']

['natural-language-processing']

[-1.51037171e-01 1.17232211e-01 -2.42825300e-01 -6.13986790e-01 -6.38244867e-01 -5.29608607e-01 4.21231359e-01 5.35686135e-01 -9.67905939e-01 7.82554924e-01 -7.64392614e-02 -3.74812514e-01 -1.70069560e-01 -6.92285895e-01 -4.23764378e-01 -4.49520022e-01 1.75667524e-01 9.79585767e-01 2.81419933e-01 -5.00592411e-01 1.02706440e-01 6.73270166e-01 -1.66282582e+00 2.64062047e-01 1.13661981e+00 9.01293576e-01 1.58146515e-01 5.45798838e-01 -3.32841456e-01 9.88616526e-01 -5.94303250e-01 -1.06168163e+00 -6.98773097e-03 5.36656240e-03 -9.41406786e-01 -5.90849340e-01 5.60339630e-01 8.84197578e-02 -8.96278024e-02 9.24524188e-01 7.27637291e-01 4.03836280e-01 5.44712543e-01 -9.12823141e-01 -6.01492941e-01 1.16422772e+00 -1.30611271e-01 1.78966615e-02 2.20534831e-01 -2.78221875e-01 1.11592662e+00 -8.16647172e-01 5.87800145e-01 1.30547416e+00 9.87824976e-01 5.62471688e-01 -8.76614749e-01 -9.02625322e-01 -4.07971591e-02 4.26314175e-01 -1.73705602e+00 -3.69891822e-01 4.70203787e-01 -3.24286878e-01 1.15109932e+00 1.69228613e-01 5.17310858e-01 7.27998912e-01 -2.57495165e-01 9.84218478e-01 1.37972426e+00 -6.30554676e-01 1.61385149e-01 6.82114244e-01 6.31812960e-02 7.42939711e-01 4.92041111e-02 -3.06295276e-01 -5.68632483e-01 -1.09930746e-01 2.17344806e-01 -3.09517175e-01 -1.30792456e-02 -2.21024409e-01 -8.19005072e-01 9.94914234e-01 2.90925980e-01 5.90182245e-01 -6.25488162e-02 -3.10526520e-01 5.39316654e-01 2.88445055e-01 6.61665499e-01 6.63522363e-01 -8.85676622e-01 -4.95620459e-01 -1.11851704e+00 8.31300318e-02 1.22971284e+00 1.00210595e+00 7.16937304e-01 9.45572183e-02 2.88873650e-02 1.18134522e+00 2.26551499e-02 4.77831423e-01 8.67383063e-01 -4.94980395e-01 3.19222093e-01 5.36006391e-01 -3.53227198e-01 -7.30482161e-01 -4.80136991e-01 -2.93979973e-01 -6.96369350e-01 -2.78062880e-01 3.39361876e-01 -3.17607641e-01 -6.13059998e-01 1.49910152e+00 3.80190104e-01 2.83089072e-01 1.10712379e-01 4.25006300e-01 1.23921609e+00 3.52245450e-01 9.83994305e-02 7.53456727e-03 1.23867071e+00 -1.23142946e+00 -4.59955186e-01 -5.92047758e-02 9.24087405e-01 -1.04200065e+00 7.78420985e-01 6.27542377e-01 -9.82520044e-01 -5.44429481e-01 -8.44596267e-01 3.43119347e-04 -7.50816524e-01 2.33978674e-01 8.15048635e-01 1.06625307e+00 -1.15939772e+00 7.54951298e-01 -5.95483422e-01 -4.25283551e-01 2.34297216e-01 6.22455060e-01 -3.72589856e-01 -3.13033834e-02 -1.23032451e+00 1.25079906e+00 5.70477486e-01 -1.33106396e-01 -4.63924378e-01 -5.34950614e-01 -7.36368775e-01 2.42051184e-01 2.18508929e-01 -2.19198912e-01 1.50496042e+00 -8.67703855e-01 -1.97133315e+00 1.17085087e+00 -2.06034426e-02 -5.91841698e-01 3.39833021e-01 -1.45934790e-01 -5.31398833e-01 -7.03697056e-02 -1.23420998e-01 7.06516623e-01 4.02461350e-01 -8.57866645e-01 -7.29280174e-01 -2.81102806e-01 1.31724223e-01 2.67440528e-01 -6.47822976e-01 2.16736913e-01 -4.30431336e-01 -5.51058531e-01 -2.18246982e-01 -1.02601647e+00 -3.73924285e-01 -6.20610833e-01 -8.87215212e-02 -6.42995238e-01 2.84428686e-01 -6.34513080e-01 1.37614918e+00 -1.99315429e+00 -1.38207063e-01 2.45444551e-01 -9.48082432e-02 8.32594037e-01 -1.97285414e-01 5.16325176e-01 -1.16494961e-01 1.47446379e-01 6.45791441e-02 -4.78166908e-01 -1.62102226e-02 2.84470409e-01 -8.96033198e-02 1.77932262e-01 7.14622997e-03 5.93774021e-01 -1.03605258e+00 -7.17299163e-01 5.29291779e-02 5.30951977e-01 -6.89018428e-01 9.88247246e-02 1.44114599e-01 2.06314906e-01 -1.17244624e-01 6.27524614e-01 6.71217680e-01 2.31891975e-01 2.61887580e-01 1.38973845e-02 -2.38173872e-01 7.22361803e-01 -1.17897594e+00 1.52140236e+00 -1.03856361e+00 5.15573859e-01 2.48719379e-01 -1.28968394e+00 1.19434786e+00 3.68116677e-01 4.08037454e-01 -5.46178639e-01 3.25580627e-01 5.71705878e-01 1.31751895e-01 -2.05034479e-01 9.53515768e-01 -2.24348351e-01 -6.27569556e-02 1.81835726e-01 4.51411635e-01 -6.07219934e-01 4.90165859e-01 3.11049111e-02 7.90274858e-01 6.72368100e-03 2.73865372e-01 -1.05942823e-01 7.51542330e-01 -1.48515508e-01 4.75234449e-01 7.06843138e-01 -3.20050389e-01 3.54508877e-01 2.17247322e-01 -1.89332142e-01 -5.60352862e-01 -9.31376338e-01 -3.18291038e-01 1.49860942e+00 -2.36495405e-01 -5.65618336e-01 -7.44640768e-01 -9.36151266e-01 -1.46546990e-01 8.90215039e-01 -1.27586633e-01 -9.07512978e-02 -6.93669021e-01 -4.53079909e-01 9.98125970e-01 4.18814421e-01 3.39279175e-01 -1.04314756e+00 -2.34264180e-01 2.59545535e-01 -5.30806892e-02 -1.21756363e+00 -8.10519904e-02 4.81949121e-01 -6.34022772e-01 -7.88045168e-01 -5.17238319e-01 -1.19526219e+00 3.03498924e-01 -1.40279576e-01 1.43066943e+00 -1.63199678e-02 -2.15575129e-01 1.92930609e-01 -3.74362230e-01 -6.61018133e-01 -4.18535233e-01 4.81490880e-01 2.84587473e-01 -2.55013615e-01 7.29333222e-01 -4.42750543e-01 -1.44075558e-01 2.04668373e-01 -6.79082811e-01 -1.17241345e-01 6.04563296e-01 8.72495413e-01 4.11703974e-01 4.36017243e-03 5.16966224e-01 -1.05552232e+00 4.55863565e-01 -2.84059912e-01 -4.93876964e-01 4.00478303e-01 -6.93039596e-01 3.24556492e-02 7.92428374e-01 -5.25677860e-01 -1.14199531e+00 1.82795495e-01 -8.79550397e-01 -3.29921573e-01 -3.50423872e-01 4.25920725e-01 7.91045353e-02 -1.23881496e-01 4.90776271e-01 -2.95220185e-02 -4.52022940e-01 -6.70638084e-01 7.26592839e-01 1.15947533e+00 3.85847986e-01 -6.91927433e-01 3.43104601e-01 2.28171155e-01 -4.08473283e-01 -1.03849757e+00 -1.10118651e+00 -9.40015674e-01 -8.60907733e-01 1.71201840e-01 7.07552314e-01 -9.36435103e-01 -8.34055841e-01 4.33334738e-01 -1.08891749e+00 -1.56734079e-01 -4.31672096e-01 6.80505514e-01 -3.43484372e-01 1.58849403e-01 -7.15991199e-01 -6.40340507e-01 -4.51589733e-01 -1.13218546e+00 8.28579485e-01 2.41399005e-01 -1.67725757e-01 -1.24028957e+00 1.23468637e-01 6.00202143e-01 5.48052311e-01 -4.40693080e-01 8.21907997e-01 -1.36731744e+00 2.35070720e-01 -7.34867230e-02 -5.70915081e-02 6.88411295e-01 -2.00410455e-01 -2.97098421e-02 -1.13047326e+00 -2.75246173e-01 -8.25943872e-02 -6.43806279e-01 6.81260765e-01 -1.25214206e-02 1.07879698e+00 -2.24533960e-01 -1.15754932e-01 4.38571036e-01 1.13402939e+00 1.12877160e-01 1.88777819e-01 3.84758681e-01 7.20650017e-01 6.92415893e-01 5.51881850e-01 3.03584546e-01 5.75237513e-01 6.26933575e-01 1.24481604e-01 -2.29522455e-02 -1.11915894e-01 -2.40175948e-01 3.87028426e-01 1.59557831e+00 7.57695083e-03 -3.11271735e-02 -1.12972319e+00 6.90831542e-01 -1.46268034e+00 -7.32410192e-01 -3.40637751e-02 2.17726445e+00 1.12307310e+00 1.31363854e-01 -9.04690921e-02 -8.03346410e-02 7.00564086e-01 -1.13053985e-01 -1.85919374e-01 -1.02972889e+00 -9.96852815e-02 9.29431379e-01 5.31289876e-01 4.37764257e-01 -1.09222960e+00 1.27988291e+00 5.79518032e+00 1.39956713e+00 -1.39329898e+00 2.89273709e-01 6.25600457e-01 -8.49830359e-02 1.76745337e-02 -6.88161096e-03 -1.38826513e+00 1.34222895e-01 1.40041339e+00 3.49650672e-03 2.47769043e-01 1.12162805e+00 -2.13176623e-01 1.04832426e-02 -1.01175141e+00 9.66330707e-01 3.15416485e-01 -8.76399934e-01 -1.23886883e-01 -3.88172626e-01 9.81629789e-01 4.74539846e-01 -1.10088609e-01 1.00468898e+00 6.57897413e-01 -1.01630867e+00 5.70312619e-01 2.07516238e-01 5.55732965e-01 -9.88909960e-01 1.00850475e+00 4.44439530e-01 -1.14579809e+00 8.28546658e-02 -4.11854565e-01 -2.25611497e-02 -8.29741582e-02 6.71533644e-01 -1.35875797e+00 5.83169281e-01 8.10229361e-01 5.48633337e-01 -8.00584555e-01 9.60825443e-01 -3.70296657e-01 1.07179308e+00 -3.51558775e-01 -4.51110095e-01 2.98085839e-01 -2.21622556e-01 2.66362071e-01 1.54997659e+00 2.49929041e-01 -2.88151503e-01 3.09295505e-01 3.83475631e-01 -1.79312944e-01 6.04923785e-01 -3.51668417e-01 -1.31452516e-01 4.42380518e-01 1.56184757e+00 -6.49321079e-01 -6.18624270e-01 -5.10927975e-01 8.18209767e-01 6.74942553e-01 -1.68636322e-01 -6.76772654e-01 -7.32924402e-01 5.31878412e-01 -9.29190218e-02 3.41557294e-01 -2.70110872e-02 -1.94503143e-01 -1.20805514e+00 -4.66055959e-01 -9.42895412e-01 4.81457144e-01 -5.07901430e-01 -1.44272852e+00 7.00396776e-01 -1.87753644e-02 -8.65541995e-01 -3.57136607e-01 -8.33488584e-01 -2.59155661e-01 7.38755524e-01 -1.49256814e+00 -9.58263695e-01 1.34957105e-01 5.59340894e-01 5.15014112e-01 -2.96067536e-01 1.03765893e+00 6.27887189e-01 -5.85509479e-01 7.97354519e-01 5.25467157e-01 1.26317918e-01 9.15280402e-01 -1.35168886e+00 1.99187294e-01 3.10149997e-01 4.34491545e-01 6.09958053e-01 3.29243600e-01 -2.38474548e-01 -9.01532888e-01 -8.77198160e-01 1.45666122e+00 -2.00890064e-01 9.00982916e-01 -3.93860281e-01 -9.04341042e-01 4.01730299e-01 2.68360317e-01 -2.44357347e-01 8.78433645e-01 4.12713766e-01 -3.14916074e-01 -6.43533766e-02 -1.20449638e+00 5.23762822e-01 7.04036891e-01 -5.31861782e-01 -6.76718175e-01 4.09185618e-01 6.01356328e-01 -4.36980128e-01 -1.12137926e+00 3.15366864e-01 3.95781577e-01 -1.08355820e+00 8.85808587e-01 -3.96509230e-01 1.81226030e-01 2.41911560e-01 1.29406035e-01 -1.43110299e+00 9.45686996e-02 -3.86974961e-01 2.40180165e-01 1.74449587e+00 3.86338502e-01 -6.03687763e-01 8.76495004e-01 2.74792016e-01 -2.36496717e-01 -9.73769307e-01 -8.86993706e-01 -8.33853066e-01 4.13753539e-01 -5.95411956e-01 6.28494859e-01 1.30339921e+00 3.71950008e-02 4.75691795e-01 -4.66836952e-02 -2.90717989e-01 -8.23624060e-02 5.49082831e-02 6.40452266e-01 -1.30012560e+00 -4.13202822e-01 -6.44105673e-01 -3.52228135e-01 -1.09499466e+00 5.61628640e-01 -1.22747242e+00 1.16862826e-01 -1.41173780e+00 -1.18382247e-02 -8.60583603e-01 -1.82683349e-01 7.30917156e-01 4.44852449e-02 3.35787952e-01 1.59344926e-01 -1.86989874e-01 -5.99546254e-01 5.59274971e-01 8.88731539e-01 -1.08355314e-01 -1.48972034e-01 1.85188860e-01 -3.77107561e-01 1.20953286e+00 7.86445439e-01 -5.66859901e-01 -1.40052691e-01 -2.66732335e-01 1.82211071e-01 -2.63449758e-01 -3.24233204e-01 -1.05880868e+00 2.31368139e-01 4.47012596e-02 -6.08860627e-02 -5.62024534e-01 3.95365775e-01 -6.75739348e-01 -4.57783192e-01 3.50482553e-01 -2.14671463e-01 1.25094891e-01 2.76302636e-01 -1.20507397e-01 -5.19820392e-01 -8.17208648e-01 9.23492849e-01 -1.37355864e-01 -9.00460839e-01 -9.50281229e-03 -4.13568497e-01 3.55901927e-01 8.39080155e-01 1.22108966e-01 -5.50863892e-02 -2.08255157e-01 -6.01935208e-01 1.19537469e-02 -3.71177979e-02 5.40764093e-01 2.30108455e-01 -9.37842727e-01 -7.57679343e-01 1.26796409e-01 -3.60860042e-02 5.77329881e-02 8.36071968e-02 8.66178036e-01 -8.37354839e-01 7.54792809e-01 5.81873208e-03 -4.50856775e-01 -1.55851567e+00 4.10829276e-01 4.37849425e-02 -7.78158009e-01 4.19035628e-02 1.18424976e+00 -1.54827431e-01 -1.34664369e+00 2.83992678e-01 -3.17036629e-01 -5.06409526e-01 3.51861864e-01 1.97831109e-01 5.81869125e-01 5.93892336e-01 -9.67463553e-01 -2.65092909e-01 4.81250256e-01 -1.77796528e-01 5.22949640e-03 1.24979532e+00 1.43602625e-01 -2.65405804e-01 3.50243479e-01 1.09429824e+00 3.71726513e-01 -2.89690375e-01 -2.63173878e-01 2.78069347e-01 -5.18800318e-02 -9.64579508e-02 -8.09621513e-01 -8.57270777e-01 1.14750826e+00 5.62810779e-01 -1.72810759e-02 1.08087170e+00 -3.64851393e-02 7.75773823e-01 7.70386279e-01 4.37285215e-01 -1.25817001e+00 -1.55815467e-01 1.08815145e+00 3.08967441e-01 -1.08253849e+00 -7.48568028e-02 -4.16229516e-01 -6.40259683e-01 9.45766091e-01 9.02670264e-01 2.43753120e-01 6.43882036e-01 2.05809265e-01 1.19368389e-01 1.19309992e-01 -5.85688055e-01 -4.68606919e-01 3.48530948e-01 4.79249269e-01 8.76957595e-01 2.21219361e-01 -5.35244286e-01 7.16529191e-01 -8.58728170e-01 -2.04268292e-01 2.50692368e-01 9.36966240e-01 -4.11040366e-01 -1.42911184e+00 -3.26465130e-01 2.94787169e-01 -7.42071927e-01 -5.20218551e-01 -5.04106700e-01 7.99201429e-01 4.61773008e-01 1.03157115e+00 -6.44799247e-02 -3.40627372e-01 4.23337966e-01 2.41567090e-01 3.45583409e-01 -8.88926387e-01 -1.19917655e+00 -3.33725572e-01 2.36493617e-01 -1.01037569e-01 -3.53530765e-01 -4.42821383e-01 -1.37348795e+00 -2.93529212e-01 -7.59004414e-01 6.20202303e-01 9.72667754e-01 8.93974721e-01 7.73738399e-02 3.25509280e-01 3.98362845e-01 -7.51026928e-01 -8.50223422e-01 -1.13506722e+00 -6.25783026e-01 2.44105786e-01 -2.70220101e-01 -4.69512731e-01 -2.41116107e-01 -2.60621041e-01]

[10.233713150024414, 9.578152656555176]

8e24ebfd-512e-4f92-87a6-7f0670961234

progressive-attention-networks-for-visual

1606.02393

http://arxiv.org/abs/1606.02393v5

http://arxiv.org/pdf/1606.02393v5.pdf

Progressive Attention Networks for Visual Attribute Prediction

We propose a novel attention model that can accurately attends to target objects of various scales and shapes in images. The model is trained to gradually suppress irrelevant regions in an input image via a progressive attentive process over multiple layers of a convolutional neural network. The attentive process in each layer determines whether to pass or block features at certain spatial locations for use in the subsequent layers. The proposed progressive attention mechanism works well especially when combined with hard attention. We further employ local contexts to incorporate neighborhood features of each location and estimate a better attention probability map. The experiments on synthetic and real datasets show that the proposed attention networks outperform traditional attention methods in visual attribute prediction tasks.

['Xiaohui Shen', 'Zhe Lin', 'Scott Cohen', 'Bohyung Han', 'Paul Hongsuck Seo']

2016-06-08

['hard-attention']

['methodology']

[ 4.10510778e-01 -8.61264914e-02 -1.85724825e-01 -3.70050728e-01 -2.74842143e-01 4.60364763e-03 3.85152817e-01 3.54935914e-01 -5.18892705e-01 5.75629115e-01 2.78271735e-01 1.07825510e-01 -7.68196583e-02 -7.97742248e-01 -8.11170757e-01 -5.71188748e-01 -1.99428247e-03 3.43396574e-01 6.81870282e-01 1.20934583e-01 5.56246161e-01 6.89361930e-01 -1.48998129e+00 3.66315871e-01 7.13936985e-01 1.12057304e+00 6.97807670e-01 7.02786684e-01 -1.21931639e-02 1.22754288e+00 -4.47125286e-01 -1.77003518e-01 1.07189812e-01 -1.89705670e-01 -8.99600446e-01 4.14119884e-02 5.85787714e-01 -3.13444227e-01 -2.41402179e-01 1.15225041e+00 2.54004419e-01 4.58343208e-01 7.39828050e-01 -8.75237942e-01 -1.54261518e+00 2.54733145e-01 -9.25987124e-01 1.04083037e+00 -1.00449078e-01 7.72580281e-02 9.39269364e-01 -1.16788113e+00 6.37870356e-02 1.30783522e+00 3.38367671e-01 2.66938895e-01 -1.11903489e+00 -4.90562558e-01 6.89042151e-01 7.18357861e-01 -1.53802526e+00 -2.22786605e-01 7.71079004e-01 -2.72110909e-01 1.02009678e+00 1.57663822e-01 3.92621577e-01 6.59023881e-01 6.50318980e-01 9.36221719e-01 7.41395891e-01 -3.21063876e-01 1.03025645e-01 4.20215167e-02 2.42372423e-01 5.82542896e-01 1.90207779e-01 -1.77592412e-01 -3.77124459e-01 4.71250527e-03 8.60531509e-01 3.66540462e-01 -4.52876091e-04 -1.39679506e-01 -1.00983346e+00 7.61250198e-01 1.29049706e+00 1.42714739e-01 -1.01254833e+00 3.14817697e-01 8.23216066e-02 -1.43535301e-01 5.77204704e-01 3.81308824e-01 -3.05852175e-01 5.08267879e-01 -6.21361852e-01 -1.04525506e-01 -7.21759647e-02 1.01297319e+00 9.04442489e-01 -3.50432545e-02 -7.85287321e-01 8.85929465e-01 1.50217608e-01 2.15547055e-01 4.12226707e-01 -7.04489589e-01 3.86583388e-01 5.98287225e-01 1.61628291e-01 -1.20393431e+00 -1.81638360e-01 -5.78598678e-01 -9.16968524e-01 4.25097972e-01 5.97565286e-02 5.73326312e-02 -1.42811179e+00 1.46178448e+00 1.82029322e-01 3.29851627e-01 -2.36415967e-01 1.08952606e+00 6.93860173e-01 8.08064282e-01 8.73183548e-01 7.16064647e-02 1.35531425e+00 -1.23890233e+00 -5.93949795e-01 -8.55605423e-01 -2.20200971e-01 -8.46452713e-01 1.10282326e+00 1.64622575e-01 -1.23396385e+00 -1.09773362e+00 -9.27853763e-01 -3.51793557e-01 -5.01782298e-01 2.21369848e-01 4.99095351e-01 -2.55661011e-02 -1.01167476e+00 4.25697684e-01 -5.72678387e-01 -3.46503198e-01 1.04912245e+00 4.75300282e-01 -1.15479581e-01 -1.93181708e-01 -1.08316839e+00 1.05556345e+00 2.89088488e-01 2.33126462e-01 -1.03662753e+00 -3.93739522e-01 -8.91852677e-01 5.83192527e-01 -5.29613905e-02 -6.47128701e-01 1.03844059e+00 -1.60830474e+00 -9.85688746e-01 6.95640922e-01 -4.40334439e-01 -4.80054080e-01 3.73574309e-02 -6.63996100e-01 -2.05701098e-01 2.21071556e-01 1.42584100e-01 1.00535715e+00 1.19860244e+00 -1.02961636e+00 -9.22226489e-01 -2.18604565e-01 4.57427874e-02 3.92025560e-01 -3.25751573e-01 1.97751313e-01 -5.56893766e-01 -9.77004766e-01 5.37635833e-02 -5.62362075e-01 -4.21148658e-01 3.11493948e-02 -3.34201008e-01 -2.75487840e-01 1.01693153e+00 -4.89487380e-01 1.00512266e+00 -1.95957470e+00 1.36829689e-01 -4.60871756e-02 2.04950646e-01 3.30634862e-01 -3.41787547e-01 -1.10267170e-01 -2.55957484e-01 8.13388750e-02 5.41217905e-03 -1.40046254e-01 -4.49661732e-01 7.69888312e-02 -3.92833471e-01 3.45070660e-01 6.74010694e-01 1.05183959e+00 -8.15778911e-01 -6.77336931e-01 4.34644043e-01 6.19083226e-01 -5.13092279e-01 3.05982798e-01 -9.12756249e-02 2.86917329e-01 -6.63719296e-01 6.60487294e-01 8.34328294e-01 -4.46007073e-01 -4.12783772e-01 -6.60734251e-02 -4.78197634e-03 1.32323608e-01 -7.71472752e-01 1.25863492e+00 -3.40975463e-01 8.00306976e-01 -2.03410521e-01 -8.66947651e-01 9.64434743e-01 -1.06962711e-01 -1.26899719e-01 -7.56141365e-01 1.25264540e-01 -4.50549901e-01 1.73432171e-01 -5.90036333e-01 5.54867446e-01 5.60227968e-02 9.01088566e-02 2.68082589e-01 9.52160805e-02 3.76283973e-01 -3.19284171e-01 5.30883186e-02 7.73711026e-01 8.15394223e-02 4.83820289e-01 -4.82085884e-01 8.18790734e-01 -2.07053542e-01 6.37204051e-01 1.08930397e+00 -5.95307827e-01 6.60573483e-01 2.03398153e-01 -9.51971650e-01 -1.02986860e+00 -8.62222493e-01 3.62490974e-02 1.91061783e+00 3.93140048e-01 1.61160260e-01 -5.43624222e-01 -9.19311762e-01 -2.98372805e-02 6.09936714e-01 -1.25180948e+00 -2.96808243e-01 -5.37099838e-01 -4.44773018e-01 -1.81490213e-01 1.06407487e+00 7.02164829e-01 -1.71223450e+00 -6.51832104e-01 1.63622049e-03 1.45245567e-01 -5.90048552e-01 -5.96809506e-01 3.70483786e-01 -6.64969623e-01 -7.86276281e-01 -7.88522303e-01 -1.18532848e+00 1.04951108e+00 3.56455058e-01 1.05858374e+00 2.55856067e-01 -3.07129979e-01 -1.02769576e-01 -3.21252286e-01 -6.00652933e-01 1.73145667e-01 3.65740269e-01 -2.79630333e-01 3.87890667e-01 7.10857511e-01 -2.90429413e-01 -1.00898826e+00 3.68636459e-01 -5.82368016e-01 -1.00499205e-01 1.02032447e+00 8.73724043e-01 7.39092946e-01 -2.74533272e-01 5.68060517e-01 -8.61875534e-01 4.76864368e-01 -6.03511512e-01 -4.78999794e-01 3.62534702e-01 -1.19389609e-01 3.16625923e-01 7.01464951e-01 -5.31015754e-01 -1.18057919e+00 2.35292748e-01 1.93199515e-02 -5.13333321e-01 -3.17125052e-01 1.64471388e-01 -9.78400111e-02 -9.71443802e-02 4.81745362e-01 2.89342791e-01 -5.00073791e-01 -4.01911408e-01 2.90713459e-01 4.89502132e-01 4.92479235e-01 -2.20636770e-01 4.91812527e-01 3.13371688e-01 -1.88823193e-01 -5.20671189e-01 -1.21234083e+00 -2.34981701e-01 -9.97216523e-01 -7.79646710e-02 1.01312363e+00 -6.68056667e-01 -6.52890444e-01 2.92817295e-01 -9.93336141e-01 -2.19988376e-01 -2.26281270e-01 4.12896007e-01 -3.65366578e-01 -5.99896722e-02 -5.77560782e-01 -7.60693312e-01 -3.25355798e-01 -1.03201413e+00 9.43893969e-01 6.58261299e-01 -1.69003144e-01 -8.93170238e-01 -2.47839749e-01 -9.48316231e-02 5.86450458e-01 -1.32784739e-01 8.03515851e-01 -5.13986111e-01 -8.16989958e-01 -1.22058786e-01 -6.90224588e-01 2.32288837e-01 3.04478556e-01 4.70651388e-02 -1.15560865e+00 -1.80469289e-01 -1.97146714e-01 -3.64357799e-01 1.40266693e+00 7.69617677e-01 1.77571237e+00 -3.29232424e-01 -5.50132990e-01 6.21131718e-01 1.27183807e+00 2.93987930e-01 9.02313888e-01 3.69166136e-01 6.64052844e-01 3.81303787e-01 8.21217120e-01 2.00673580e-01 -3.37431170e-02 2.73354888e-01 7.78132021e-01 -4.68086421e-01 -1.14462033e-01 -1.38957396e-01 -2.36924648e-01 1.98103875e-01 -2.16592446e-01 -1.83975950e-01 -8.04440737e-01 1.00000155e+00 -1.74394107e+00 -1.10849166e+00 1.27916247e-01 2.07835793e+00 4.37508970e-01 3.64501208e-01 -2.41462588e-01 -2.93591470e-01 1.11056244e+00 3.10518593e-01 -7.15774536e-01 -6.24074936e-01 -3.10604926e-02 3.12072188e-01 5.40707231e-01 5.80941796e-01 -1.40797150e+00 1.12852728e+00 7.14639282e+00 5.94461799e-01 -1.11034894e+00 1.47617999e-02 1.21707869e+00 -2.74279658e-02 1.27314225e-01 -1.99673995e-01 -6.79260612e-01 3.90522331e-01 4.33563173e-01 -2.20399164e-02 1.52122021e-01 9.39147949e-01 1.81399286e-02 -2.65141100e-01 -7.65880287e-01 6.42888188e-01 1.81402713e-01 -1.23991489e+00 4.41285342e-01 -3.83818716e-01 7.58817732e-01 2.21169945e-02 5.60548425e-01 3.36820692e-01 4.55617160e-01 -1.37739134e+00 5.84536850e-01 9.38810349e-01 4.65283185e-01 -1.07542181e+00 6.95591748e-01 2.00428799e-01 -1.20867097e+00 -4.41795111e-01 -8.93483818e-01 -3.76396447e-01 -3.35284293e-01 -3.27795222e-02 -8.09380174e-01 -3.30033936e-02 9.23808753e-01 6.94405854e-01 -8.95651400e-01 1.24559927e+00 -5.18012345e-01 4.40982699e-01 1.28581196e-01 -4.79028113e-02 4.59878623e-01 2.62133121e-01 2.64036328e-01 1.25122011e+00 8.93430784e-02 3.57226223e-01 -1.50753960e-01 8.96600008e-01 -1.66129097e-01 2.26475988e-02 -4.50983375e-01 5.87250292e-01 4.43411767e-01 1.31406784e+00 -7.95417726e-01 -5.06933808e-01 -5.14056742e-01 1.32163870e+00 8.77979994e-01 6.07984781e-01 -9.14046884e-01 -6.88319504e-01 6.15414500e-01 1.36893257e-01 8.96000445e-01 1.64541468e-01 -3.72223914e-01 -5.78493178e-01 -3.33878577e-01 -2.20148727e-01 7.59653568e-01 -1.30724728e+00 -1.31228507e+00 9.46015894e-01 -3.19957256e-01 -9.95109200e-01 3.45436245e-01 -4.83839631e-01 -1.00741422e+00 1.40116537e+00 -1.73228478e+00 -1.18618274e+00 -5.77221334e-01 7.47367442e-01 9.52931225e-01 6.54528383e-03 6.27674699e-01 -1.03530223e-02 -6.22372687e-01 4.28000778e-01 -9.37998742e-02 9.13874358e-02 6.99117601e-01 -1.26820946e+00 5.27773082e-01 8.60091805e-01 -7.42233023e-02 7.19473422e-01 5.01297355e-01 -6.27948940e-01 -4.48168069e-01 -1.45205581e+00 8.35462451e-01 -3.38220805e-01 2.15474308e-01 -1.90551415e-01 -1.23372567e+00 9.96208310e-01 7.31251121e-01 5.07746041e-01 2.89513409e-01 5.08165918e-02 -1.50490791e-01 -2.20968783e-01 -1.08149958e+00 2.83970386e-01 7.32292116e-01 -3.62614363e-01 -8.50146830e-01 8.16505253e-02 6.49911463e-01 -2.65629470e-01 -3.70239824e-01 3.29236150e-01 4.02252644e-01 -6.96269572e-01 1.13601863e+00 -7.98044026e-01 3.85691226e-01 -3.94471437e-01 1.34202957e-01 -1.26396430e+00 -1.33094800e+00 -3.22454154e-01 -5.87283401e-03 1.05693853e+00 3.86441469e-01 -2.63310164e-01 7.00482428e-01 6.44419074e-01 -5.23471199e-02 -5.78672588e-01 -5.68786025e-01 -2.72260189e-01 -2.28493586e-02 4.40305509e-02 6.57668829e-01 6.71582758e-01 -3.07092309e-01 5.08359551e-01 -3.44528466e-01 4.40697461e-01 4.86097008e-01 1.39378414e-01 1.71542913e-01 -1.10140896e+00 2.00199008e-01 -3.86763722e-01 -4.88617390e-01 -1.17051578e+00 -5.00358380e-02 -4.44141716e-01 9.60146636e-02 -1.59517527e+00 5.91209769e-01 -3.26326787e-01 -1.03771842e+00 6.35399997e-01 -9.08304334e-01 6.49151742e-01 8.09379742e-02 2.63432503e-01 -8.97593200e-01 5.78034043e-01 1.22919118e+00 -3.73747766e-01 -1.15952380e-01 3.90286855e-02 -8.08058083e-01 8.92004311e-01 8.11100602e-01 -4.74212885e-01 -3.07119429e-01 -6.74777150e-01 -2.45974302e-01 -5.35224557e-01 4.84848022e-01 -1.18214810e+00 3.79890800e-01 -3.90633345e-01 1.20314026e+00 -7.87026584e-01 1.34763092e-01 -8.29777718e-01 -3.08492392e-01 3.10434669e-01 -6.89590931e-01 1.25989363e-01 5.45422375e-01 5.54995775e-01 -2.08019719e-01 -1.19862743e-01 1.02490366e+00 -9.03427228e-02 -1.11069024e+00 4.33207750e-01 -3.54058206e-01 -1.77364647e-01 1.36565435e+00 -1.61479637e-01 -1.94963530e-01 -3.52122366e-01 -1.19064176e+00 2.55759001e-01 2.47571826e-01 5.12970448e-01 9.47302461e-01 -1.48245108e+00 -6.88153505e-01 3.73124212e-01 1.98135585e-01 -6.98673353e-02 4.53567982e-01 4.65859324e-01 -4.97063488e-01 4.52233344e-01 -6.98364019e-01 -5.10087967e-01 -1.18455637e+00 1.21281815e+00 3.98415595e-01 2.30593467e-03 -3.56406599e-01 1.30703807e+00 8.57653856e-01 3.51003647e-01 3.20229292e-01 -4.36245292e-01 -6.74134552e-01 -2.95002937e-01 8.65043759e-01 8.54669958e-02 -2.00077251e-01 -7.91495264e-01 -4.44733709e-01 6.79212391e-01 -4.27796990e-01 3.78955990e-01 1.44319630e+00 -3.57677132e-01 6.51301742e-02 1.80137992e-01 9.15142000e-01 -2.14181781e-01 -1.83608115e+00 -3.82490128e-01 -2.16550350e-01 -7.09540844e-01 2.99226910e-01 -7.80230582e-01 -1.27366209e+00 1.05142260e+00 6.38068914e-01 1.96890786e-01 1.42095685e+00 1.34514332e-01 3.29105288e-01 6.15030043e-02 -2.09938765e-01 -9.93897617e-01 2.71836907e-01 4.39683616e-01 1.24835813e+00 -1.29009771e+00 1.35124519e-01 -1.82574630e-01 -9.00023520e-01 8.81877363e-01 1.15800607e+00 -5.49199164e-01 5.73180497e-01 1.11456998e-01 -1.31298780e-01 -2.65804410e-01 -8.54306400e-01 -4.99331236e-01 4.89442617e-01 7.38280058e-01 3.77671301e-01 -3.11099589e-01 2.30576769e-01 3.23711097e-01 4.94229615e-01 -1.57059655e-01 7.74254650e-02 8.11619759e-01 -8.88180971e-01 -4.90117252e-01 -5.46826243e-01 3.89662266e-01 -5.91483653e-01 -2.75068969e-01 -3.93880516e-01 6.44216955e-01 4.12916481e-01 5.26791692e-01 5.81452310e-01 -1.78990632e-01 2.41221979e-01 -1.34128541e-01 2.37808108e-01 -6.63488030e-01 -5.88275969e-01 -1.03116930e-01 -5.38118243e-01 -4.33021784e-01 -3.89165401e-01 -5.89071393e-01 -8.98560584e-01 -9.98297557e-02 -1.77079752e-01 -5.74733093e-02 -6.10778295e-02 6.13251746e-01 4.91889715e-01 8.61878932e-01 5.36649466e-01 -9.51741576e-01 -9.91603732e-03 -1.22844303e+00 -3.21374536e-01 4.19505447e-01 7.02889204e-01 -7.96185195e-01 -7.81984441e-03 2.48689055e-01]

[9.789026260375977, 1.8759835958480835]

dee8173a-0922-419b-9a4a-d522bb979b7c

umcc_dlsi-a-probabilistic-automata-for-aspect

https://aclanthology.org/S14-2129

https://aclanthology.org/S14-2129.pdf

UMCC\_DLSI: A Probabilistic Automata for Aspect Based Sentiment Analysis

['Rafael Mu{\\~n}oz', "Andr{\\'e}s Montoyo", "David Tom{\\'a}s", "Yoan Guti{\\'e}rrez", 'o', 'Yenier Casta{\\~n}eda', 'Elvis Crego', 'Arm Collazo', 'Jorge L. Garcia']

2014-08-01

semeval-2014-8

['subjectivity-analysis']

['natural-language-processing']

[-8.63703638e-02 1.71006292e-01 -6.22772932e-01 -4.08054382e-01 -8.41685571e-03 -9.08429027e-01 6.55310392e-01 -6.53472245e-01 -2.85945535e-01 1.06888819e+00 -4.63127941e-02 -1.01159286e+00 -3.91567826e-01 -9.63214397e-01 -4.95059669e-01 -6.31337762e-01 -9.79754329e-01 7.25764990e-01 3.30370307e-01 -6.93831444e-01 7.03166842e-01 7.88774848e-01 -1.68942046e+00 7.18545914e-01 7.04417467e-01 8.52217197e-01 2.49141872e-01 1.14950800e+00 -1.95044339e-01 1.55633950e+00 -7.48382092e-01 -5.46825826e-01 3.13719302e-01 -1.23176083e-01 -7.22945035e-01 -1.01074085e-01 9.28529128e-02 -8.59008506e-02 -2.09758401e-01 9.22211111e-01 5.37373662e-01 4.49454933e-02 1.08379531e+00 -1.42548037e+00 -5.91619551e-01 6.10313773e-01 -4.01565880e-02 1.21627934e-01 1.03678203e+00 -5.39447069e-01 1.19919395e+00 -1.13026452e+00 7.20913768e-01 1.26888943e+00 8.66221786e-01 5.44149756e-01 -1.22286928e+00 -1.94712028e-01 -3.26822817e-01 -9.51717794e-02 -1.46558487e+00 -3.25250506e-01 4.25783843e-02 -2.08119690e-01 1.66093647e+00 1.26596653e+00 1.20609856e+00 1.01401424e+00 1.26658809e+00 8.34431887e-01 1.04267764e+00 -5.13792276e-01 3.35295945e-01 3.66983831e-01 1.54683650e-01 6.33519173e-01 8.40953708e-01 5.26628852e-01 -7.06372619e-01 -9.13127720e-01 9.33553874e-01 -2.94925272e-01 1.71355158e-01 -5.05680561e-01 -9.05919552e-01 6.91228509e-01 1.78732842e-01 3.83959889e-01 -1.39880210e-01 9.89067405e-02 1.26390755e-01 5.30987144e-01 -2.58292928e-02 6.47037446e-01 -9.11868811e-01 -1.33165747e-01 -8.71728659e-01 5.10332465e-01 1.25398111e+00 1.52653182e+00 1.24482810e-01 2.94908643e-01 -9.34252143e-02 3.17179203e-01 8.92314315e-01 1.01808000e+00 4.28362608e-01 -1.36146402e+00 -6.87414408e-02 1.72361732e-01 5.01781464e-01 -8.52631688e-01 -6.33224547e-01 -9.64177120e-03 -8.93263519e-01 4.49267089e-01 3.49161088e-01 4.57367361e-01 -8.02827001e-01 5.07305264e-01 4.33481112e-02 -2.34125629e-01 4.53833073e-01 5.55570945e-02 4.99930978e-01 3.76208365e-01 -1.34477139e-01 -5.73289394e-01 1.06082785e+00 -1.36716676e+00 -1.35299087e+00 2.33215362e-01 9.05734658e-01 -1.07320261e+00 4.35900748e-01 5.33875942e-01 -1.55548143e+00 -1.37560293e-01 -1.08699942e+00 1.78573877e-01 -7.27255583e-01 -3.14239264e-01 8.57801437e-01 1.43120694e+00 -1.60129595e+00 9.73287821e-01 -4.91727620e-01 6.59165755e-02 1.20568443e-02 8.24621081e-01 -2.64718989e-03 4.62812334e-01 -1.33193445e+00 1.08501506e+00 2.22979754e-01 -1.21242590e-01 -1.65216476e-01 -2.13068098e-01 -8.23704481e-01 -5.63443303e-01 -4.78693932e-01 -5.29636025e-01 1.44139910e+00 -2.59346128e-01 -1.65295815e+00 9.71794367e-01 -1.42069459e-01 -1.97814897e-01 6.14786744e-01 -1.28011424e-02 -8.31891418e-01 2.42498964e-01 -1.89849049e-01 5.76383233e-01 9.28263724e-01 -1.35132408e+00 -7.59897232e-01 -1.67359829e-01 -1.23336017e-01 2.66287565e-01 -1.25510961e-01 1.89734384e-01 2.11616129e-01 -1.12999000e-01 3.27147305e-01 -7.20919967e-01 -2.53068686e-01 -5.32041907e-01 -1.46512717e-01 -7.10518599e-01 7.70373225e-01 -4.51523662e-01 1.83705616e+00 -1.67618537e+00 -2.06720144e-01 3.98590982e-01 3.57815564e-01 -1.24705513e-03 2.40583986e-01 1.08380008e+00 -2.76906848e-01 7.32199550e-01 4.11965609e-01 -1.10722095e-01 2.24991128e-01 4.85861301e-01 -4.16602850e-01 3.05609167e-01 -1.29282743e-01 1.15307164e+00 -1.16605783e+00 -5.23096442e-01 4.11106765e-01 9.42391157e-02 -4.58732933e-01 4.79237735e-01 2.99364805e-01 1.47170946e-01 -3.56553018e-01 1.39399457e+00 1.15709066e+00 -1.31984919e-01 1.45911396e-01 5.30878425e-01 -3.79135728e-01 3.55090618e-01 -6.96863770e-01 1.05554795e+00 7.08333924e-02 5.00173986e-01 1.02364108e-01 -7.94621468e-01 3.33247900e-01 8.61540735e-01 4.37155962e-01 -9.67555881e-01 -2.26398129e-02 6.92409754e-01 1.12803578e-01 -6.07703328e-01 7.58228302e-01 3.94563079e-02 -3.64872098e-01 6.40070081e-01 -2.37588286e-01 -6.59476995e-01 9.64643434e-02 3.08100313e-01 6.22585893e-01 -5.18246442e-02 5.62923312e-01 -1.05613089e+00 7.86340594e-01 -1.82965681e-01 -1.81299388e-01 1.03415680e+00 -3.09923887e-01 3.19085121e-01 2.99841821e-01 -6.65102363e-01 -6.45341039e-01 -1.12307119e+00 -4.89381433e-01 1.30636716e+00 3.24267983e-01 -4.39044595e-01 -9.54439282e-01 -2.49762803e-01 1.77620783e-01 6.89606130e-01 -5.90509653e-01 3.84124845e-01 -5.03739953e-01 -8.32535863e-01 7.39044368e-01 3.45434904e-01 -5.07752821e-02 -1.33414865e+00 -6.58416986e-01 1.25490099e-01 -2.20292807e-01 -6.63697243e-01 -6.23428151e-02 4.48765576e-01 -1.35989368e+00 -5.18594682e-01 -6.66252747e-02 -8.20914626e-01 5.87345481e-01 2.46782884e-01 1.27047324e+00 5.39230824e-01 -2.31483161e-01 4.26904231e-01 -1.21292919e-01 -4.95818377e-01 -4.59671497e-01 -8.00336525e-02 5.28869390e-01 -5.87835789e-01 5.19427478e-01 -2.50617653e-01 -7.29350567e-01 5.37953973e-01 -6.88540697e-01 1.62748516e-01 1.79803044e-01 1.04410267e+00 1.35816500e-01 -9.34035778e-02 1.22507080e-01 -6.38007045e-01 8.72274399e-01 -1.69219792e-01 -3.78732830e-01 5.77745810e-02 -6.77108407e-01 -3.74140263e-01 3.21430594e-01 -3.25342178e-01 -1.01981449e+00 -4.87835288e-01 -9.82677937e-02 2.45538145e-01 1.11353043e-02 -1.46784872e-01 6.47139177e-02 -5.24923325e-01 8.02199244e-01 9.25758183e-02 1.99174434e-02 -6.80815242e-03 3.01039815e-01 7.09525108e-01 -6.82967342e-03 -6.68678164e-01 8.44880998e-01 4.91470337e-01 7.98524171e-02 -9.57177758e-01 -1.52186140e-01 -2.60129690e-01 -9.51962709e-01 -6.54426932e-01 6.56643391e-01 -6.78531289e-01 -9.10833478e-01 3.91110867e-01 -9.38691139e-01 -3.38627815e-01 -3.91645581e-01 4.25431967e-01 -1.01278400e+00 2.75717527e-02 -3.90154392e-01 -1.27895141e+00 -5.10977268e-01 -1.02017939e+00 9.43384409e-01 5.30070923e-02 -5.10597289e-01 -1.26927447e+00 5.87685481e-02 2.71537274e-01 1.81734428e-01 -1.73075795e-01 6.90226793e-01 -2.38256708e-01 -4.24233019e-01 -1.53791070e-01 2.34436691e-02 -1.39755070e-01 1.70832314e-02 4.95917559e-01 -9.81751978e-01 -5.31145096e-01 6.65065646e-02 -1.92070693e-01 -1.08835101e-01 6.52520418e-01 5.91872573e-01 -2.29931593e-01 -8.56000841e-01 5.40386558e-01 1.38545322e+00 3.85070026e-01 5.32770038e-01 7.28214979e-01 1.41836226e-01 5.53460240e-01 9.17806149e-01 4.63203549e-01 1.30579369e-02 3.28798652e-01 2.40537539e-01 1.49327129e-01 1.11720070e-01 -1.54819340e-01 3.77893507e-01 1.16112018e+00 -8.18235934e-01 -2.69281328e-01 -5.07867396e-01 4.42987174e-01 -1.72482407e+00 -1.40330648e+00 -4.32368398e-01 6.90478683e-01 6.25676990e-01 1.56016424e-01 -1.48347050e-01 3.35214496e-01 4.99015123e-01 -2.03574806e-01 -1.19133167e-01 -1.06291151e+00 -1.43546045e-01 3.15233678e-01 7.37729073e-01 1.00061214e+00 -7.20721722e-01 1.03317809e+00 1.29781246e+01 1.02230716e+00 2.21112028e-01 1.03134915e-01 5.16071796e-01 3.48020852e-01 -4.36954498e-01 -4.56139445e-02 -1.04416132e+00 2.72933897e-02 1.38140702e+00 -4.30666685e-01 6.85999811e-01 5.44219851e-01 3.44648361e-01 -4.23268199e-01 -1.26188684e+00 5.26221812e-01 9.73738134e-02 -1.40886843e+00 -2.83300440e-04 6.85225725e-01 7.73699820e-01 -5.08050561e-01 6.22419357e-01 3.24184299e-01 6.09259963e-01 -1.14389277e+00 8.60300779e-01 2.53660440e-01 1.03040910e+00 -6.05088234e-01 5.67372203e-01 1.68872893e-01 -1.14389896e+00 -2.20873043e-01 -8.77727985e-01 -1.00755692e+00 3.93533185e-02 -1.81779593e-01 -4.29956943e-01 3.48861217e-01 9.58353162e-01 2.99398601e-01 -3.93658698e-01 9.95779395e-01 -4.78694476e-02 1.04875881e-02 -2.71853864e-01 -4.48467314e-01 4.83122796e-01 -3.54241252e-01 4.66730654e-01 1.00164843e+00 2.48499006e-01 3.51035744e-01 -9.84472036e-02 4.01770771e-01 5.45058846e-01 3.29446048e-02 -1.19659424e+00 -1.78908288e-01 2.83276141e-01 9.16795909e-01 -4.83487815e-01 -4.22520459e-01 -2.00212970e-01 8.62069130e-01 -3.55488248e-02 5.01107454e-01 -6.11489356e-01 -4.35615242e-01 9.72222984e-01 -1.27327025e-01 -1.14700586e-01 -3.48497719e-01 -6.23769283e-01 -7.30352640e-01 -5.89872956e-01 -4.54965204e-01 5.93606755e-02 -5.53365827e-01 -1.39813089e+00 5.79277515e-01 -2.27688253e-02 -1.40553558e+00 -6.99901402e-01 -1.27676582e+00 -4.76714373e-01 4.92853165e-01 -1.11898029e+00 -1.10984349e+00 2.50124663e-01 4.52870727e-01 1.64141744e-01 -5.34416080e-01 1.39563632e+00 3.57715860e-02 1.00637585e-01 9.24474537e-01 6.69434488e-01 -7.37814724e-01 5.56605101e-01 -1.27867436e+00 5.68737745e-01 -1.37897313e-01 -4.31265175e-01 9.05828118e-01 6.28349900e-01 -5.39804697e-01 -1.41196322e+00 -3.66917729e-01 1.08350635e+00 -9.83769417e-01 6.55218959e-01 -3.86345625e-01 4.23767231e-02 7.88592756e-01 7.15902448e-01 -6.18741751e-01 8.21781039e-01 -1.83753878e-01 1.80774391e-01 5.75296998e-01 -1.39248300e+00 6.12354755e-01 1.66275799e+00 -4.63594139e-01 -6.25784039e-01 7.60327101e-01 8.13696027e-01 -6.94087505e-01 -1.30082703e+00 3.34633321e-01 8.65424156e-01 -8.75409484e-01 1.61978090e+00 -1.32660246e+00 -4.43697497e-02 2.81152606e-01 -2.61993498e-01 -9.32519078e-01 -5.96193194e-01 -1.23518765e+00 -5.33532679e-01 -5.83747849e-02 5.96577883e-01 -1.13057327e+00 3.42365682e-01 8.74560475e-01 -2.82833427e-01 -6.42737269e-01 -1.06996536e+00 -1.32016802e+00 -3.35779637e-02 -1.45572275e-01 4.90409225e-01 7.63798356e-01 6.83744550e-01 1.09839931e-01 -6.36873543e-02 -1.00294888e-01 5.33176839e-01 9.55312885e-03 4.41501856e-01 -1.34294486e+00 3.86843324e-01 -5.75816095e-01 -3.07655483e-01 -9.45992947e-01 -8.85957032e-02 -8.12076271e-01 -6.53862000e-01 -1.28511906e+00 -8.32044985e-03 -1.91056758e-01 -1.11109078e-01 -1.65725678e-01 3.67937148e-01 2.13746816e-01 1.20859891e-02 1.03788137e-01 -3.71160030e-01 6.18435517e-02 1.29639816e+00 6.91750320e-05 -1.62315920e-01 4.85058486e-01 -4.81304944e-01 7.84440815e-01 8.58408585e-02 -2.99253196e-01 -6.78878546e-01 6.11881316e-02 6.69384480e-01 4.61409837e-02 3.24159935e-02 -7.42885649e-01 5.37211418e-01 -3.75702560e-01 4.78586555e-01 -1.32223868e+00 1.30741090e-01 -9.61415648e-01 6.95283338e-02 9.40189242e-01 2.74610907e-01 1.20424610e-02 8.66204947e-02 5.39500564e-02 -1.44871444e-01 -5.70943117e-01 9.21121240e-01 -4.07591403e-01 -4.92852688e-01 -5.20386267e-03 -1.03226590e+00 8.97834301e-02 9.92593169e-01 -7.84614205e-01 -3.59281451e-01 -4.20183957e-01 -8.29068601e-01 -1.95836127e-02 6.50830388e-01 3.03609259e-02 7.30431557e-01 -1.51530886e+00 -2.30721906e-01 7.18729138e-01 -3.22939813e-01 -3.74400020e-01 -1.70157343e-01 6.58265352e-01 -1.32361674e+00 1.02442718e+00 -5.41665435e-01 -4.55340147e-01 -1.14228773e+00 4.78126436e-01 4.28307921e-01 -2.41845414e-01 -2.15481281e-01 1.11954463e+00 2.71224789e-02 -8.23025763e-01 1.85185194e-01 -9.21545625e-02 -7.54407048e-01 3.55081353e-03 6.88606799e-01 1.05194807e+00 -2.91290224e-01 -6.04341030e-01 -4.56784427e-01 6.65885091e-01 2.32151806e-01 -2.87484169e-01 9.17833567e-01 -2.19243199e-01 -9.89108324e-01 4.28274393e-01 8.38715494e-01 -1.36269778e-01 -3.69319022e-02 4.16855574e-01 1.25943512e-01 -8.02164078e-01 -4.32406247e-01 -3.55811834e-01 -1.85641110e-01 5.54822803e-01 5.34874737e-01 8.99602413e-01 8.57008278e-01 -3.02566767e-01 8.18335712e-01 9.66778398e-01 5.72402716e-01 -1.68019545e+00 -2.13140488e-01 6.89524531e-01 9.12339568e-01 -9.28350806e-01 5.44190466e-01 -7.27165341e-01 -4.14997995e-01 1.32979155e+00 4.68304873e-01 -1.55325383e-01 1.27306652e+00 5.74917436e-01 1.14069022e-02 -3.70670199e-01 -9.44949508e-01 1.12705544e-01 3.75366658e-01 1.11147714e+00 5.06513238e-01 5.07374525e-01 -9.66500878e-01 3.21953118e-01 -7.28706717e-01 -2.34555230e-01 4.90474731e-01 1.41972518e+00 -6.43810987e-01 -1.20391917e+00 -7.23931909e-01 4.61561680e-01 -5.49773455e-01 -1.16372630e-01 -5.28106689e-01 8.46754074e-01 -5.76629937e-02 1.49448860e+00 -1.97535474e-03 -5.03491640e-01 4.11356747e-01 1.41089618e-01 7.21762300e-01 -1.23501487e-01 -9.04846430e-01 4.15413082e-01 3.84890139e-01 -1.23056793e+00 -8.58632207e-01 -1.05834293e+00 -1.40667629e+00 -1.19437599e+00 -5.12782812e-01 1.89310342e-01 3.83317530e-01 3.90289724e-01 -2.06836104e-01 2.85260603e-02 9.80917513e-01 -1.07949340e+00 -3.90341938e-01 -9.39418614e-01 -1.00026262e+00 -7.84516707e-02 2.89751232e-01 -8.00943017e-01 -7.83523321e-01 2.83909619e-01]

[-7.356940746307373, 3.6964635848999023]

f4c57060-ab7e-4ce8-9817-af2619329253

event-based-star-tracking-via-multiresolution

1906.07866

https://arxiv.org/abs/1906.07866v2

https://arxiv.org/pdf/1906.07866v2.pdf

Event-based Star Tracking via Multiresolution Progressive Hough Transforms

Star trackers are state-of-the-art attitude estimation devices which function by recognising and tracking star patterns. Most commercial star trackers use conventional optical sensors. A recent alternative is to use event sensors, which could enable more energy efficient and faster star trackers. However, this demands new algorithms that can efficiently cope with high-speed asynchronous data, and are feasible on resource-constrained computing platforms. To this end, we propose an event-based processing approach for star tracking. Our technique operates on the event stream from a star field, by using multiresolution Hough Transforms to time-progressively integrate event data and produce accurate relative rotations. Optimisation via rotation averaging is then used to fuse the relative rotations and jointly refine the absolute orientations. Our technique is designed to be feasible for asynchronous operation on standard hardware. Moreover, compared to state-of-the-art event-based motion estimation schemes, our technique is much more efficient and accurate.

['Tat-Jun Chin', 'Samya Bagchi']

2019-06-19

['event-based-motion-estimation']

['computer-vision']

[-1.06222205e-01 -4.39810187e-01 -9.68073234e-02 8.81406069e-02 -1.76666066e-01 -7.19516754e-01 7.71147549e-01 2.41089180e-01 -5.33896983e-01 2.74785191e-01 -2.68935710e-01 -1.09663308e-01 5.58986962e-02 -7.52408803e-01 -4.12934780e-01 -7.02317536e-01 -1.96561575e-01 4.16173846e-01 9.01763618e-01 -6.11035265e-02 1.45963773e-01 1.05885088e+00 -1.71634233e+00 -6.34739637e-01 6.99077785e-01 1.02725565e+00 1.37115493e-01 1.08038497e+00 3.16947192e-01 6.75182819e-01 -6.93863392e-01 -1.50552362e-01 3.52384210e-01 -4.29905325e-01 -7.02187717e-02 2.93893009e-01 5.06342471e-01 -2.40025118e-01 -3.95009130e-01 9.64803219e-01 4.52360034e-01 1.44740090e-01 -1.38303740e-02 -8.80393207e-01 3.45538706e-01 -2.54473478e-01 -4.35075998e-01 3.35862964e-01 5.25069475e-01 2.60970831e-01 5.24324834e-01 -8.91655445e-01 5.73579729e-01 7.26546288e-01 7.82110631e-01 1.00325935e-01 -1.03229475e+00 -3.24266851e-01 -2.32772902e-01 3.05485576e-01 -1.41060400e+00 -7.04940975e-01 6.78749263e-01 -1.80918500e-01 1.06478798e+00 3.67983341e-01 1.26046121e+00 3.59307319e-01 3.17714125e-01 5.03229678e-01 8.41323555e-01 -4.42656428e-01 3.23134869e-01 -4.23921198e-01 -3.98963451e-01 6.01708770e-01 6.24967396e-01 4.25675541e-01 -8.67452919e-01 -1.17165595e-01 1.08732355e+00 1.13295011e-01 -4.06583160e-01 -7.20553398e-01 -1.81558549e+00 6.24102354e-01 2.06872478e-01 1.71728805e-01 -6.29530132e-01 3.35182518e-01 1.38390452e-01 8.86208788e-02 3.93837780e-01 4.24432755e-01 -1.58959597e-01 -4.74094182e-01 -1.33821809e+00 3.04124862e-01 9.02380168e-01 7.09511042e-01 5.13222277e-01 3.80441785e-01 1.75835386e-01 -3.10961101e-02 7.23326445e-01 7.34442711e-01 5.20425975e-01 -7.65863419e-01 -9.36023891e-02 3.50817829e-01 3.13375831e-01 -7.24220812e-01 -5.96415043e-01 -4.91078734e-01 -5.65425277e-01 6.30260766e-01 6.31393135e-01 6.47126511e-02 -7.28408813e-01 1.01454031e+00 8.43027115e-01 6.39915168e-01 1.74698472e-01 1.28607833e+00 4.18549448e-01 5.54015279e-01 -4.72048640e-01 -6.02139533e-01 1.86104679e+00 -7.34874725e-01 -8.93949807e-01 -2.91901916e-01 4.83606577e-01 -1.15075600e+00 1.26797810e-01 5.48301756e-01 -9.84189987e-01 -4.94758040e-01 -1.53071010e+00 1.04819559e-01 -4.99575250e-02 2.05313876e-01 7.68404186e-01 7.39864826e-01 -1.05012655e+00 3.67021799e-01 -1.52494383e+00 -4.61434931e-01 -4.61409569e-01 4.84407425e-01 -1.05540706e-02 5.44763803e-01 -7.36314237e-01 1.14161444e+00 5.01586735e-01 -1.77571662e-02 -5.77460825e-01 -3.59244734e-01 -8.63957405e-01 -2.00030163e-01 4.76090759e-01 -6.73156023e-01 1.68675792e+00 -6.52226150e-01 -2.16570926e+00 5.95204175e-01 -2.24947929e-01 -9.48787868e-01 -3.16065177e-02 -2.58396834e-01 -6.36065125e-01 3.01535338e-01 -5.70644438e-01 2.87461311e-01 1.29420996e+00 -5.04503250e-01 -7.60580540e-01 -5.44128902e-02 -4.25980181e-01 4.36789304e-01 -7.59037510e-02 2.63113797e-01 -4.92367595e-01 -5.98311722e-01 4.86852735e-01 -9.17246521e-01 -3.34317118e-01 2.40638986e-01 5.19557655e-01 -1.50485680e-01 1.14023960e+00 -4.56313372e-01 1.29061341e+00 -2.07644272e+00 1.27095535e-01 -1.53145967e-02 2.22270384e-01 4.98544842e-01 6.53614759e-01 1.27304852e-01 2.30237320e-01 -1.12807941e+00 1.37801304e-01 -2.55520016e-01 -2.83356339e-01 3.42101485e-01 -3.67114276e-01 1.04280245e+00 1.73344668e-02 4.44515586e-01 -9.26266491e-01 -4.18463945e-01 8.23737204e-01 5.10923564e-01 -1.86075732e-01 1.62784472e-01 -1.79664731e-01 4.97113496e-01 -2.59855181e-01 6.15803421e-01 4.58266258e-01 -8.32119361e-02 2.05582783e-01 -3.61939430e-01 -8.23273361e-01 4.80759501e-01 -1.50428522e+00 1.88124597e+00 -1.31482318e-01 7.85601795e-01 3.60425234e-01 -7.06164837e-01 1.17038429e+00 4.73590463e-01 5.21287620e-01 -3.65595192e-01 2.57444978e-01 4.09227371e-01 -2.60308176e-01 1.05817318e-01 1.11870086e+00 1.61729708e-01 1.53129520e-02 8.46290216e-02 7.57143870e-02 -5.25160730e-01 3.72727841e-01 1.99863706e-02 1.03071189e+00 5.92045367e-01 7.71094739e-01 -1.53658584e-01 4.54922169e-01 1.84035733e-01 6.73556149e-01 1.64482474e-01 -1.12875104e-01 2.88164794e-01 -3.67944330e-01 -7.20097005e-01 -8.56417000e-01 -9.09445703e-01 -2.33533438e-02 4.66763586e-01 4.90489006e-01 -7.43629038e-01 -3.36849302e-01 -2.77710676e-01 -1.09607160e-01 2.63219416e-01 3.74367386e-01 9.93263796e-02 -7.99889088e-01 -7.98120737e-01 1.75927266e-01 4.42276120e-01 3.33011121e-01 -3.53380203e-01 -1.71327269e+00 6.56871557e-01 1.17399216e-01 -1.13686216e+00 -4.01900172e-01 2.87217468e-01 -1.25911224e+00 -8.67727995e-01 -4.41548496e-01 -4.42455053e-01 6.02198541e-01 6.50572300e-01 1.17780602e+00 -1.98731244e-01 -1.62394494e-01 4.65871185e-01 -4.63029295e-01 -4.58182305e-01 3.29442620e-02 -4.53074664e-01 6.20569825e-01 2.25690290e-01 4.06776518e-01 -3.38142663e-01 -6.05959594e-01 4.06124115e-01 -7.32468426e-01 9.95280370e-02 5.66672266e-01 5.59364498e-01 7.72779524e-01 6.57217354e-02 -4.38785225e-01 2.72906303e-01 -8.30059424e-02 3.67757934e-03 -1.62706125e+00 -1.56421661e-02 -5.77440917e-01 3.57402235e-01 4.60836321e-01 -5.88599026e-01 -8.52465332e-01 7.30782092e-01 2.40644470e-01 -6.24826729e-01 1.76757753e-01 1.89661011e-01 2.76067525e-01 -5.80989838e-01 5.44011295e-01 1.73583105e-01 2.01367978e-02 -4.01083559e-01 3.85913104e-01 4.81420428e-01 1.21321714e+00 -2.62082964e-01 1.05524170e+00 7.89953887e-01 4.47805643e-01 -1.09442365e+00 -2.49727845e-01 -8.64626408e-01 -5.30431509e-01 -5.20154536e-01 8.03554535e-01 -1.19792259e+00 -8.16964388e-01 6.59439743e-01 -1.22016335e+00 1.67801812e-01 -3.98064643e-01 1.20216143e+00 -4.38995481e-01 6.38413489e-01 -2.57109016e-01 -1.02602911e+00 -3.50428790e-01 -9.39198792e-01 1.16739798e+00 8.61907065e-01 -2.10339069e-01 -7.79141605e-01 5.01030922e-01 -3.28389078e-01 3.19035679e-01 2.57055789e-01 -6.72417939e-01 -2.11831257e-01 -1.03933334e+00 -3.05427670e-01 1.79667354e-01 -3.91269952e-01 -7.69273490e-02 1.22238241e-01 -5.70998132e-01 -8.06776226e-01 3.25925082e-01 2.03153551e-01 5.17327130e-01 3.82444531e-01 5.04561186e-01 -4.76611815e-02 -4.80438471e-01 9.79476273e-01 1.28990042e+00 2.51427025e-01 4.66088146e-01 3.88429612e-01 4.57292318e-01 -5.59569634e-02 1.07835245e+00 6.73678100e-01 3.24039340e-01 1.29945934e+00 4.35894817e-01 2.33710874e-02 -1.44729361e-01 1.38673382e-02 7.56351233e-01 1.19787490e+00 -3.65247637e-01 -2.49249693e-02 -5.57870567e-01 6.85874403e-01 -1.89087963e+00 -8.41197312e-01 -4.44024473e-01 2.60987759e+00 6.00513458e-01 1.63560465e-01 1.28005147e-01 2.18456432e-01 3.97361428e-01 3.63155335e-01 -3.03176135e-01 -2.63761282e-01 2.01010960e-03 4.17155951e-01 1.26926303e+00 2.21211314e-01 -1.11255538e+00 7.23863602e-01 6.01116514e+00 4.26609665e-01 -1.19180524e+00 4.79089133e-02 -5.14743865e-01 -3.19743752e-01 1.76756844e-01 4.38483953e-01 -1.14783537e+00 3.76085073e-01 1.31275237e+00 -1.74585506e-01 3.47710967e-01 8.14824998e-01 7.75712505e-02 -4.78563935e-01 -7.71151245e-01 1.26301754e+00 2.40259871e-01 -1.14906800e+00 -7.20233142e-01 9.27540958e-02 3.42117637e-01 1.86473325e-01 -5.39708912e-01 -2.15558246e-01 4.65895027e-01 -2.30556086e-01 8.16767395e-01 6.01470947e-01 6.63020611e-01 -6.43495798e-01 4.41401392e-01 3.01856369e-01 -1.89912128e+00 3.57138515e-01 -4.35999751e-01 -4.06405449e-01 5.65280020e-01 8.76686275e-01 -9.52998340e-01 9.15961683e-01 8.54182720e-01 5.90547621e-01 -5.34787118e-01 1.39227211e+00 -2.96249628e-01 4.05521184e-01 -1.02903402e+00 -2.00518250e-01 -1.03833787e-01 -3.15024167e-01 8.75171840e-01 8.17025721e-01 9.18181896e-01 1.00974761e-01 3.41502577e-01 -1.09790005e-02 4.22674984e-01 -3.31423789e-01 -5.42123079e-01 1.93758064e-03 5.56459844e-01 1.49828231e+00 -8.58903408e-01 -6.09016418e-01 -7.50954330e-01 9.69696820e-01 -1.05243646e-01 -2.11182907e-01 -8.45053673e-01 -5.04673541e-01 7.25225866e-01 -1.46407597e-02 6.34369195e-01 -1.03758311e+00 -1.79118663e-02 -1.50749147e+00 -8.90322030e-03 -8.12592208e-01 4.53254670e-01 -7.80049264e-01 -5.15155554e-01 3.98026556e-01 -6.30949065e-02 -1.84752464e+00 -6.52219594e-01 -6.17628396e-01 -4.66343969e-01 5.42499483e-01 -1.31381738e+00 -9.43235099e-01 -5.88239253e-01 5.43603539e-01 4.11694199e-01 -7.60290325e-02 6.18636847e-01 5.52228875e-02 -9.15936604e-02 -1.22848554e-02 6.71419278e-02 -2.90259421e-01 8.73464108e-01 -1.20126379e+00 7.69799769e-01 1.70997965e+00 6.22983873e-01 2.12308109e-01 1.09443998e+00 -7.32393444e-01 -2.18092155e+00 -6.21210277e-01 8.28209400e-01 -4.06494528e-01 8.36882591e-01 -6.24221377e-02 -6.04942620e-01 5.27707756e-01 2.53780991e-01 4.66171503e-01 1.81213334e-01 -3.80555272e-01 8.90191495e-02 -3.46341938e-01 -7.56656170e-01 3.73636872e-01 8.21770608e-01 -3.88792962e-01 -7.83613920e-01 1.77431226e-01 3.94000441e-01 -1.06973553e+00 -9.56564248e-01 5.84419370e-01 4.09610420e-01 -7.81800985e-01 1.04541123e+00 3.54939848e-01 -8.70733261e-01 -1.43387020e+00 2.41182759e-01 -1.16288745e+00 -3.95368665e-01 -1.29001200e+00 -1.02191257e+00 7.81365097e-01 -3.80234450e-01 -7.48198032e-01 7.36207783e-01 -3.95042188e-02 -2.87477404e-01 6.34563267e-02 -1.26767743e+00 -9.12081420e-01 -1.10446060e+00 -1.24860562e-01 5.30343831e-01 7.39871502e-01 7.03283250e-02 9.24144909e-02 -4.68871742e-01 5.82147777e-01 1.03430009e+00 3.42801481e-01 1.01889682e+00 -1.40456355e+00 -5.65283477e-01 -6.34295121e-02 -8.34684789e-01 -1.49019301e+00 -5.12737572e-01 -3.94157410e-01 1.98413447e-01 -9.79576290e-01 -5.86434305e-01 1.33990183e-01 1.65890291e-01 8.80368724e-02 -7.01724067e-02 4.28831965e-01 8.25015083e-03 3.93785477e-01 -7.28353202e-01 3.22461367e-01 7.36428440e-01 3.05821955e-01 -1.71818927e-01 -4.40455265e-02 2.21183941e-01 7.58540332e-01 5.59965730e-01 -2.52400726e-01 5.50929867e-02 -2.96604186e-01 2.42581710e-01 1.18845455e-01 4.47752088e-01 -1.56561089e+00 9.88717556e-01 1.43817469e-01 5.35539806e-01 -1.16895807e+00 4.85600799e-01 -9.86537158e-01 6.17745697e-01 7.61338711e-01 8.92209113e-01 5.10405242e-01 1.39673933e-01 4.47810024e-01 -3.82721186e-01 -1.68248359e-02 7.40607083e-01 6.91070855e-02 -8.93040419e-01 1.35803998e-01 -6.83737099e-01 -5.32401979e-01 1.24006832e+00 -3.90419573e-01 -1.68210089e-01 -2.01765314e-01 -1.71151832e-01 5.72321676e-02 9.62287426e-01 1.54380798e-01 3.53497177e-01 -1.43953454e+00 -3.79409283e-01 4.95062917e-01 6.41518235e-02 1.94550022e-01 -2.40399554e-01 1.05566061e+00 -1.03045976e+00 6.22581244e-01 -2.08008289e-02 -1.02001750e+00 -1.47917712e+00 5.28659642e-01 1.90642104e-01 -2.49438763e-01 -6.41463578e-01 4.89540249e-01 -5.33685863e-01 2.06373215e-01 -4.08462249e-02 -5.24402738e-01 -5.07026725e-02 4.38681059e-02 6.76048994e-01 3.83432239e-01 2.99262255e-01 -6.54685318e-01 -5.24293542e-01 8.34644377e-01 3.45300078e-01 -3.54378194e-01 8.42354596e-01 -6.43306673e-02 3.80892195e-02 2.85729263e-02 2.74767250e-01 2.59162724e-01 -1.47833633e+00 -2.24827304e-01 2.62395293e-01 -5.93548536e-01 4.44350868e-01 -2.25790292e-02 -7.60296881e-01 3.55256647e-01 7.68024385e-01 2.59673655e-01 1.49153233e+00 -2.57167876e-01 8.83858204e-01 2.68929809e-01 6.92832768e-01 -1.04528725e+00 -3.29194665e-01 4.35873419e-01 3.07938904e-01 -8.96940231e-01 5.92231333e-01 -3.60885620e-01 2.39183637e-03 1.28351438e+00 3.73483509e-01 -1.23194687e-01 1.45190552e-01 6.25515878e-01 -1.59058943e-02 -1.65080950e-01 -7.32071936e-01 -5.68347514e-01 3.49383384e-01 3.92864585e-01 2.73500532e-01 3.66210155e-02 -6.85973883e-01 -2.30839118e-01 -1.35557964e-01 -8.37864280e-02 4.42256212e-01 1.24574339e+00 -6.20317101e-01 -1.33368397e+00 -1.23028028e+00 1.60297134e-03 -2.52590626e-01 1.86096966e-01 -9.71014276e-02 4.81089950e-01 -1.60302982e-01 8.34404230e-01 4.30201322e-01 -2.76862174e-01 3.36898476e-01 -9.23064575e-02 7.32523382e-01 -2.06152752e-01 -5.90899169e-01 3.59261066e-01 2.37122357e-01 -8.08544576e-01 -7.86919594e-01 -1.01287031e+00 -1.04287684e+00 -3.14056009e-01 -4.70448762e-01 2.14092419e-01 1.00705969e+00 6.44629776e-01 3.30335170e-01 3.66252065e-01 5.67490935e-01 -1.15702415e+00 -3.86030912e-01 -6.85072958e-01 -2.14636728e-01 -9.57916602e-02 3.28774065e-01 -8.14814031e-01 -3.26286286e-01 2.02767625e-01]

[7.661462306976318, -1.7986226081848145]

ad098d8b-d329-42bc-aa67-6e01efbf59e1

learning-the-string-partial-order

2305.15057

https://arxiv.org/abs/2305.15057v1

https://arxiv.org/pdf/2305.15057v1.pdf

Learning the String Partial Order

We show that most structured prediction problems can be solved in linear time and space by considering them as partial orderings of the tokens in the input string. Our method computes real numbers for each token in an input string and sorts the tokens accordingly, resulting in as few as 2 total orders of the tokens in the string. Each total order possesses a set of edges oriented from smaller to greater tokens. The intersection of total orders results in a partial order over the set of input tokens, which is then decoded into a directed graph representing the desired structure. Experiments show that our method achieves 95.4 LAS and 96.9 UAS by using an intersection of 2 total orders, 95.7 LAS and 97.1 UAS with 4 on the English Penn Treebank dependency parsing benchmark. Our method is also the first linear-complexity coreference resolution model and achieves 79.2 F1 on the English OntoNotes benchmark, which is comparable with state of the art.

['Ryan Cotterell', 'Mrinmaya Sachan', 'Afra Amini', 'Tianyu Liu']

2023-05-24

['dependency-parsing', 'coreference-resolution']

['natural-language-processing', 'natural-language-processing']

[ 4.08523947e-01 5.41143715e-01 -4.93101865e-01 -3.87083620e-01 -1.04980135e+00 -1.16086924e+00 3.56307044e-03 4.59430784e-01 -5.74667752e-01 7.69618809e-01 3.39783043e-01 -5.05825996e-01 1.33033274e-02 -9.17307377e-01 -8.97292078e-01 -4.58119631e-01 -1.81944668e-01 1.04415452e+00 6.44556761e-01 -1.57279581e-01 1.73250660e-01 3.07955146e-01 -1.35360777e+00 6.75066411e-01 5.95939934e-01 8.18866253e-01 1.49525404e-01 7.72561014e-01 -5.76285422e-01 3.92949939e-01 -6.67319000e-01 -7.04740047e-01 3.68461609e-01 -2.06829205e-01 -1.31611001e+00 -3.54869753e-01 4.73535091e-01 -1.43819451e-01 -2.50297993e-01 1.31553185e+00 1.58718273e-01 7.07308436e-06 1.63887471e-01 -1.09055269e+00 -3.29875767e-01 1.48077524e+00 -3.92559022e-01 1.17520571e-01 7.03710854e-01 -4.65679914e-01 1.95242417e+00 -4.38614875e-01 8.28699648e-01 1.24497402e+00 3.76951933e-01 5.77567279e-01 -1.18629766e+00 -7.17350721e-01 1.58447251e-01 5.80211878e-02 -8.20988476e-01 -2.64126450e-01 3.16129178e-01 1.09824061e-01 1.37310350e+00 5.03638804e-01 6.58585370e-01 1.52004227e-01 1.70531701e-02 7.22505391e-01 9.59485054e-01 -8.02889347e-01 9.20728445e-02 -2.24302053e-01 8.09739292e-01 8.12134743e-01 4.65568691e-01 -1.04928963e-01 -6.65402234e-01 -2.25676671e-01 3.14978033e-01 -4.65100855e-01 -1.01506710e-01 -6.50384203e-02 -1.02205443e+00 6.84925318e-01 1.85289383e-01 6.59170374e-02 7.00717941e-02 1.50075600e-01 4.24115419e-01 3.31286758e-01 -2.20274881e-01 4.17647690e-01 -6.91815674e-01 -2.34113336e-01 -5.05633771e-01 2.41135716e-01 1.25579035e+00 1.21190894e+00 7.16034412e-01 -4.60315019e-01 2.63250947e-01 8.61562312e-01 6.83461651e-02 4.02705342e-01 1.69578695e-03 -1.35498297e+00 9.57228661e-01 6.85432136e-01 8.32648352e-02 -6.17820978e-01 -5.22674501e-01 -1.16198314e-02 -2.03815222e-01 -1.09398760e-01 9.12026823e-01 2.89947670e-02 -4.91622537e-01 1.93386710e+00 1.82451129e-01 -2.66215771e-01 2.66241431e-01 6.90797508e-01 6.16989672e-01 9.21052039e-01 8.07159021e-02 -3.89450103e-01 2.02491951e+00 -7.61654675e-01 -5.14004827e-01 -3.91398340e-01 9.58100319e-01 -8.20001364e-01 8.11994493e-01 5.12936652e-01 -1.47309077e+00 -7.69274905e-02 -9.97546971e-01 -2.27169618e-01 -1.55982748e-02 -2.09263742e-01 6.12413108e-01 6.04509473e-01 -9.04607773e-01 7.12039053e-01 -8.60722363e-01 7.13162422e-02 -1.23797424e-01 7.73216546e-01 -4.34102386e-01 -4.62919660e-02 -1.17706728e+00 6.39131129e-01 6.48242056e-01 -1.28001317e-01 8.53695441e-03 -7.34009445e-01 -8.77085209e-01 4.18055981e-01 3.91029000e-01 -1.66222274e-01 1.62781656e+00 -6.31716073e-01 -1.16573036e+00 1.14061964e+00 -4.58961964e-01 -5.72066784e-01 1.14859425e-01 -1.47609815e-01 -2.75430173e-01 1.69870809e-01 1.45108953e-01 6.84032440e-01 -2.62526125e-01 -1.02009583e+00 -1.27254343e+00 -3.34140688e-01 3.88852835e-01 2.73528129e-01 -1.65911764e-01 3.99822235e-01 -6.92282259e-01 -2.38919735e-01 5.10681331e-01 -1.25423467e+00 -1.87819675e-01 -4.16375637e-01 -6.37616456e-01 -5.46103120e-01 6.60454929e-02 -5.33302248e-01 1.42482805e+00 -2.04022217e+00 7.72449225e-02 1.49652496e-01 7.04258457e-02 4.07889672e-02 -1.92616925e-01 4.30299759e-01 -6.74866289e-02 1.61418885e-01 -3.32747281e-01 -8.64680633e-02 3.22047472e-01 4.81492758e-01 -3.58250141e-01 2.57514507e-01 -1.65341631e-01 5.51858366e-01 -1.05271852e+00 -5.95288992e-01 -3.92711967e-01 -1.51661009e-01 -9.21371639e-01 9.72970948e-02 -3.94474179e-01 -3.30020100e-01 -3.49552155e-01 2.99689621e-01 7.32769668e-01 1.10457055e-01 8.74250591e-01 -7.18313307e-02 -2.94936210e-01 1.05779028e+00 -1.30087256e+00 1.51863182e+00 -1.93147898e-01 4.35117632e-01 -5.72506003e-02 -7.47193992e-01 9.61942673e-01 1.87611267e-01 4.60792661e-01 -6.24685585e-01 -1.32901669e-01 4.32403117e-01 2.18161032e-01 -1.04133129e-01 7.72190988e-01 -1.41209021e-01 -7.64164448e-01 4.78589147e-01 -1.53164223e-01 1.74745813e-01 6.13739312e-01 4.72199559e-01 1.32542741e+00 2.88195908e-03 1.90167025e-01 -3.31314415e-01 3.94934177e-01 2.12904021e-01 1.03842616e+00 5.76466143e-01 -9.74837765e-02 3.14907670e-01 1.22357047e+00 -5.73125958e-01 -9.14794981e-01 -1.14296722e+00 3.50056253e-02 1.21916997e+00 7.35532865e-02 -8.58480453e-01 -8.00067127e-01 -7.17186928e-01 -1.53469801e-01 7.12719858e-01 -1.60108477e-01 9.27945971e-02 -1.23313057e+00 -4.49705869e-01 6.31087244e-01 7.85542965e-01 2.14669064e-01 -1.27122498e+00 -8.52620840e-01 4.60177869e-01 -4.78610486e-01 -1.41561258e+00 -6.62329257e-01 7.21072197e-01 -9.15165722e-01 -1.25526559e+00 1.07241713e-01 -1.43270433e+00 6.32043839e-01 -5.87346368e-02 1.21819341e+00 1.55825615e-02 9.24373046e-02 -2.41354898e-01 -3.48142266e-01 -2.16237664e-01 -4.35596973e-01 2.33531430e-01 -1.69732347e-01 -5.14976978e-01 6.34941638e-01 -3.17354232e-01 -2.65141100e-01 6.58744872e-02 -5.24765313e-01 -9.72429290e-02 3.46772760e-01 8.90403748e-01 7.38776147e-01 -1.00349642e-01 1.18718348e-01 -1.40109575e+00 1.45720854e-01 -2.16213167e-01 -9.69622850e-01 4.28647131e-01 -2.82182932e-01 5.00837386e-01 8.38104427e-01 -2.73021221e-01 -9.58251655e-01 5.37291348e-01 -1.05343191e-02 2.12416694e-01 -1.41054749e-01 4.85657662e-01 -3.03102106e-01 4.99309093e-01 3.32533807e-01 -6.24582916e-02 -4.23834413e-01 -5.32410741e-01 3.04404110e-01 3.69499058e-01 9.87170637e-01 -9.45033371e-01 4.08635259e-01 -2.27572378e-02 2.53728479e-01 -4.74116683e-01 -8.49674225e-01 -4.80740339e-01 -7.53647506e-01 2.50312418e-01 6.84328258e-01 -5.27420044e-01 -9.81443942e-01 1.66493803e-01 -1.37159169e+00 -1.94035813e-01 -2.48689100e-01 3.08442086e-01 -4.01445031e-01 4.97508407e-01 -1.13908970e+00 -6.51551306e-01 -4.03453618e-01 -1.00480843e+00 6.65284574e-01 1.31402671e-01 -5.22128880e-01 -4.82793570e-01 1.24162734e-01 3.39394182e-01 -3.60645205e-01 -1.81840599e-01 1.50726736e+00 -1.09791076e+00 -5.63480854e-01 -1.34097531e-01 -1.38539627e-01 -1.44295394e-03 -2.39655018e-01 -1.90629840e-01 -3.47246408e-01 2.65807807e-02 -5.22563517e-01 -1.22563474e-01 6.82802379e-01 6.55214712e-02 6.42892659e-01 -3.88736099e-01 -5.09215891e-01 3.91412348e-01 1.34937537e+00 6.42679751e-01 5.27944803e-01 5.42669535e-01 4.53058153e-01 7.54055500e-01 7.33017862e-01 2.56252259e-01 4.16328192e-01 5.31964481e-01 2.30923221e-01 2.49525651e-01 8.18079188e-02 -4.44810152e-01 4.28979963e-01 8.58565032e-01 -3.90463769e-02 -1.32871732e-01 -9.43720043e-01 6.19484603e-01 -1.63901901e+00 -8.77538502e-01 -4.24081802e-01 2.31813359e+00 1.21721113e+00 4.52760100e-01 1.47663534e-01 3.66345078e-01 8.31102550e-01 -2.53154691e-02 -1.72287628e-01 -1.09096086e+00 1.36865735e-01 7.03925669e-01 8.26569259e-01 6.58095121e-01 -8.37306976e-01 1.27714217e+00 6.54124212e+00 2.78587252e-01 -5.21384895e-01 -3.34651321e-01 2.17900112e-01 -1.26183763e-01 -3.94692570e-01 3.19377273e-01 -1.21037352e+00 2.70385832e-01 1.13032901e+00 -2.03329623e-01 4.39654708e-01 6.01485789e-01 -3.90407234e-01 4.51526940e-02 -1.43843424e+00 5.29864252e-01 -3.32422882e-01 -1.09315979e+00 -5.13460934e-02 1.84059888e-03 4.02949572e-01 -1.99436307e-01 -2.19334126e-01 2.23010704e-01 6.92422450e-01 -6.93969429e-01 9.32276249e-01 -2.37163186e-01 7.08114445e-01 -1.03383863e+00 6.01721883e-01 2.40857080e-01 -1.46011305e+00 -1.43986642e-01 -4.93772358e-01 -1.01889946e-01 2.70798296e-01 1.26206279e-01 -7.45395720e-01 2.96658307e-01 6.72567427e-01 2.10565746e-01 1.86802804e-01 9.45686817e-01 -5.13880312e-01 7.29935467e-01 -4.90693241e-01 -3.13589394e-01 2.11883381e-01 -1.99267074e-01 5.32618344e-01 1.49958611e+00 1.26167402e-01 7.96325862e-01 1.46055341e-01 2.16656029e-01 -4.29232448e-01 9.54567418e-02 -1.68176100e-01 9.77224112e-02 1.09368658e+00 9.59141552e-01 -8.33889008e-01 -4.89510924e-01 -4.65728998e-01 7.16409862e-01 6.65278435e-01 -9.28959623e-02 -7.22865403e-01 -7.33283699e-01 7.54881859e-01 -2.76638508e-01 4.66836780e-01 -7.79667869e-02 -3.03685397e-01 -8.46685350e-01 2.12034389e-01 -9.63602304e-01 1.10352218e+00 -2.87800163e-01 -8.22987735e-01 7.34777689e-01 6.72882842e-03 -9.07358408e-01 -9.48038101e-02 -7.92515337e-01 -4.79993254e-01 6.34748459e-01 -1.28606856e+00 -5.54166973e-01 2.64044344e-01 2.69433439e-01 4.24716383e-01 1.98558703e-01 1.28103042e+00 1.03636362e-01 -4.64212567e-01 7.77018249e-01 -2.24987507e-01 3.97097558e-01 4.95004028e-01 -1.57771456e+00 6.80922985e-01 8.57425451e-01 2.43219063e-01 6.04358137e-01 6.11079574e-01 -4.17630047e-01 -1.62469006e+00 -5.25665462e-01 1.80836070e+00 -1.32260308e-01 6.81210041e-01 -3.02616656e-01 -9.11370456e-01 1.11404920e+00 2.79806584e-01 -1.68522060e-01 7.61784017e-01 3.11527252e-01 -8.27107131e-01 -1.19470775e-01 -9.68988836e-01 5.74968398e-01 1.37125432e+00 -2.55404115e-01 -1.01383162e+00 1.02805473e-01 8.68526936e-01 -9.10887420e-01 -9.44727123e-01 1.48024276e-01 6.88212097e-01 -6.83461189e-01 7.73137629e-01 -8.37868989e-01 4.58719730e-01 -3.11286002e-02 -3.83430868e-01 -1.06955421e+00 -5.12629330e-01 -6.64334893e-01 1.01938307e-01 1.20147657e+00 9.17803824e-01 -5.49681485e-01 1.12234533e+00 9.11866009e-01 -3.14986557e-01 -6.59939706e-01 -9.74200249e-01 -4.63158697e-01 2.02117607e-01 -3.66915464e-01 7.65990555e-01 6.87477529e-01 7.79913604e-01 4.02082115e-01 -2.78891344e-02 3.10940832e-01 7.19097912e-01 6.56079829e-01 2.51217961e-01 -1.33735311e+00 -3.09422374e-01 -4.41255957e-01 -1.07435085e-01 -1.30024230e+00 3.60788107e-01 -1.28150165e+00 7.86503702e-02 -1.50427389e+00 2.75152594e-01 -7.76758850e-01 -3.80243450e-01 8.81410599e-01 3.92392538e-02 -9.42805782e-02 4.25097734e-01 1.01379253e-01 -6.05092466e-01 -1.88066140e-01 8.32421124e-01 -1.02901518e-01 -2.66694397e-01 -1.37811795e-01 -7.29098797e-01 6.92599773e-01 8.73423874e-01 -8.52094829e-01 1.94048047e-01 -5.35670459e-01 2.29979679e-01 5.93257844e-01 -3.47702175e-01 -4.78621036e-01 4.36492473e-01 -1.65662810e-01 -4.17380482e-02 -7.48737156e-01 2.38630220e-01 -6.91318333e-01 -6.25267401e-02 7.42510140e-01 -6.89059079e-01 3.40349913e-01 3.25562924e-01 2.28013992e-01 -1.20619036e-01 -4.55358654e-01 6.12537205e-01 -1.28957763e-01 -7.18654633e-01 -8.73361155e-02 -2.07528710e-01 2.49416828e-01 7.44134247e-01 -3.77878435e-02 -3.95894051e-01 1.04154326e-01 -8.29098701e-01 3.87115508e-01 1.23792849e-01 3.22369158e-01 3.90323162e-01 -1.11185741e+00 -5.89804947e-01 -4.33356948e-02 -2.94335783e-01 1.01614214e-01 -1.53612718e-01 3.25030833e-01 -6.86099827e-01 5.93395054e-01 -2.99778461e-01 -2.06930757e-01 -1.77564180e+00 3.77926022e-01 -1.99606325e-02 -6.61442697e-01 -6.42559826e-01 8.66059780e-01 -7.58819059e-02 -4.07049656e-01 4.34525162e-01 -5.70525110e-01 -2.19824120e-01 5.69341183e-02 5.38386166e-01 3.62350076e-01 -8.65563229e-02 -5.47123134e-01 -5.93610048e-01 6.04491055e-01 -2.55271167e-01 -3.29871565e-01 1.37571049e+00 4.22462374e-01 -6.02339327e-01 4.01665010e-02 1.24189818e+00 3.90626281e-01 -8.02646816e-01 -3.33231300e-01 6.71958745e-01 -2.88130850e-01 -4.91979033e-01 -7.03238010e-01 -9.61735010e-01 4.04263645e-01 -1.10193022e-01 2.45505571e-01 1.07475221e+00 3.35863590e-01 1.18344486e+00 6.88133240e-01 6.74103737e-01 -8.45034301e-01 -3.01772922e-01 7.88945615e-01 3.18788260e-01 -5.01184165e-01 -2.16348529e-01 -9.23435509e-01 -6.04177296e-01 1.28684318e+00 4.40634370e-01 -2.56557196e-01 1.92456007e-01 8.61913383e-01 -1.18044734e-01 1.76570892e-01 -1.02510571e+00 -1.94704942e-02 -7.04887956e-02 1.97019890e-01 5.57708561e-01 3.51843864e-01 -7.09568679e-01 1.08813977e+00 -7.78387547e-01 -3.32716763e-01 5.52724838e-01 1.05606306e+00 -6.99991226e-01 -1.75994027e+00 -4.34847742e-01 1.19963266e-01 -8.22575390e-01 -2.41861179e-01 -4.06544685e-01 6.59295917e-01 -1.32492974e-01 9.08368289e-01 3.65978569e-01 -2.93000460e-01 7.28447914e-01 3.98845524e-01 7.00536728e-01 -6.94131672e-01 -8.36083233e-01 -8.99074450e-02 7.22620487e-01 -4.42259938e-01 9.48587880e-02 -9.53718960e-01 -2.20590138e+00 -4.47750300e-01 -2.84113050e-01 5.08247375e-01 3.32585216e-01 6.30999744e-01 1.88396171e-01 1.99641466e-01 5.36372244e-01 -1.63345158e-01 -5.76507330e-01 -5.33679783e-01 -4.77725893e-01 3.69980246e-01 4.13094275e-02 -1.40037119e-01 -2.64154285e-01 7.63533711e-02]

[10.346930503845215, 9.538091659545898]

99535788-1e7d-419a-aa57-807b790127ef

freevc-towards-high-quality-text-free-one

2210.15418

https://arxiv.org/abs/2210.15418v1

https://arxiv.org/pdf/2210.15418v1.pdf

FreeVC: Towards High-Quality Text-Free One-Shot Voice Conversion

Voice conversion (VC) can be achieved by first extracting source content information and target speaker information, and then reconstructing waveform with these information. However, current approaches normally either extract dirty content information with speaker information leaked in, or demand a large amount of annotated data for training. Besides, the quality of reconstructed waveform can be degraded by the mismatch between conversion model and vocoder. In this paper, we adopt the end-to-end framework of VITS for high-quality waveform reconstruction, and propose strategies for clean content information extraction without text annotation. We disentangle content information by imposing an information bottleneck to WavLM features, and propose the spectrogram-resize based data augmentation to improve the purity of extracted content information. Experimental results show that the proposed method outperforms the latest VC models trained with annotated data and has greater robustness.

['Li Xiao', 'Weiping tu', 'Jingyi Li']

2022-10-27

['text-annotation']

['natural-language-processing']

[ 2.64689356e-01 -3.96428704e-02 -4.32146899e-03 -6.28654212e-02 -1.46944356e+00 -8.21652591e-01 1.99059799e-01 -1.91795945e-01 -2.80472320e-02 6.00765169e-01 8.06534886e-01 -1.88117623e-01 3.04153174e-01 -2.70700485e-01 -3.88815045e-01 -6.59212172e-01 3.20786923e-01 -1.40644848e-01 -5.64485714e-02 -1.49376225e-02 -5.58147915e-02 1.51470602e-01 -1.41431272e+00 2.30101034e-01 1.08971763e+00 1.07734644e+00 4.46746737e-01 9.49390352e-01 -3.77197623e-01 6.45807207e-01 -1.04016232e+00 -4.39667046e-01 1.54590920e-01 -8.75971258e-01 -4.34788406e-01 -3.49619389e-02 1.05615482e-01 -2.34701917e-01 -4.35804158e-01 1.03923559e+00 1.00845659e+00 7.16267303e-02 3.92610937e-01 -9.85700190e-01 -2.86614567e-01 7.87992239e-01 -2.15497166e-01 1.90660790e-01 3.23107690e-01 4.31214944e-02 9.92967129e-01 -1.15701187e+00 4.84250456e-01 9.43704665e-01 6.64589882e-01 5.86219847e-01 -1.18028414e+00 -8.24081600e-01 -2.24570259e-01 3.88041198e-01 -1.19399691e+00 -1.45538139e+00 1.24324572e+00 -2.21714422e-01 5.96971035e-01 7.09575593e-01 4.39120710e-01 1.31080413e+00 -3.16692352e-01 8.39067876e-01 8.17950070e-01 -5.75813174e-01 -3.76029387e-02 2.71963149e-01 -1.61503375e-01 1.78827196e-01 -1.35775656e-01 2.51053393e-01 -9.68564153e-01 -5.26828878e-02 3.09843987e-01 -5.83416760e-01 -8.68099749e-01 2.26794660e-01 -9.43203390e-01 4.31368500e-01 -2.96149284e-01 2.09650084e-01 -2.52358407e-01 -2.28954315e-01 5.24019957e-01 4.20599252e-01 3.99383545e-01 3.32914829e-01 -3.18656176e-01 -3.69419128e-01 -1.21432471e+00 -2.42138356e-01 7.95162439e-01 1.16860127e+00 2.28250861e-01 9.80352998e-01 -4.59515005e-01 1.26560986e+00 2.34429494e-01 8.74124229e-01 6.98970735e-01 -8.25399816e-01 7.78900862e-01 -3.40508670e-01 1.43805355e-01 -6.63060009e-01 -4.01538610e-02 -6.10163808e-01 -7.36844599e-01 -1.57304138e-01 4.51995339e-03 -4.04704124e-01 -6.92477226e-01 1.64397275e+00 1.72257230e-01 2.68270910e-01 3.30385566e-01 8.63379180e-01 9.41726148e-01 9.40742850e-01 -3.67273539e-01 -5.64661145e-01 1.09955966e+00 -1.10614216e+00 -1.70733821e+00 -1.68017328e-01 -6.58950433e-02 -1.13206160e+00 1.18485439e+00 7.01001167e-01 -1.17356372e+00 -7.99991012e-01 -1.22784019e+00 -1.02005035e-01 1.27530739e-01 3.42227072e-01 -1.86260030e-01 9.76598322e-01 -7.27803588e-01 5.21596134e-01 -5.40417612e-01 3.47986341e-01 2.24368293e-02 4.95098568e-02 -2.66157806e-01 7.48108029e-02 -1.21699405e+00 6.61642015e-01 2.25584343e-01 2.15199620e-01 -9.45069492e-01 -8.22326124e-01 -9.41198587e-01 1.19344756e-01 2.38952577e-01 -3.04084122e-02 1.38110363e+00 -5.40692449e-01 -2.04357958e+00 -7.43561089e-02 -2.78412074e-01 -2.68763483e-01 3.78471106e-01 -3.65892947e-01 -1.17975795e+00 2.97926009e-01 -2.59520143e-01 2.18418799e-02 1.43971324e+00 -1.32457197e+00 -5.27754545e-01 9.67251509e-02 -7.24384904e-01 4.44217995e-02 -5.61964035e-01 4.87350635e-02 -4.54581410e-01 -8.74493003e-01 6.50129765e-02 -3.28440875e-01 2.82449216e-01 -4.40047473e-01 -7.06594944e-01 4.34461385e-01 8.61734986e-01 -1.44460618e+00 1.48834968e+00 -2.38005710e+00 -2.15814617e-02 8.44501778e-02 -6.97026551e-02 6.29408300e-01 -4.32689220e-01 5.60363472e-01 4.05289307e-02 1.78555548e-01 -2.37422407e-01 -6.81799591e-01 9.35754105e-02 -6.84061125e-02 -8.45215738e-01 3.99685800e-01 4.19278204e-01 5.09611189e-01 -6.84754550e-01 -5.80839574e-01 1.62326053e-01 7.33804584e-01 -5.94658256e-01 6.41005874e-01 2.25321084e-01 7.82156646e-01 -1.90478582e-02 3.51616919e-01 7.60226250e-01 5.66599488e-01 1.39794409e-01 -4.22125846e-01 -1.80758923e-01 9.11246538e-01 -1.39456606e+00 1.58448362e+00 -6.71700597e-01 7.89854884e-01 6.04341924e-01 -5.92236519e-01 9.48628783e-01 8.86007249e-01 2.72705823e-01 -7.03232527e-01 1.56087339e-01 1.67134017e-01 1.01112816e-02 -6.22009099e-01 4.65604216e-01 -2.87649602e-01 1.17972016e-01 1.80785000e-01 4.11894917e-01 -5.49846828e-01 -2.81423867e-01 -1.97482049e-01 8.04031193e-01 1.69213545e-02 1.36678908e-02 2.12863788e-01 4.84578192e-01 -4.12521064e-01 8.36063623e-01 3.04185897e-01 -1.01279438e-01 8.62062156e-01 1.43654093e-01 6.61148846e-01 -1.09073138e+00 -1.01808143e+00 -1.16822653e-01 8.82236421e-01 -1.65486783e-01 -5.82462788e-01 -9.11132276e-01 -1.74993947e-01 -4.06035572e-01 1.11665773e+00 9.36636999e-02 -3.71332198e-01 -7.51295388e-01 -2.90871829e-01 1.19513130e+00 3.43788266e-01 1.86425924e-01 -8.74359131e-01 1.49825275e-01 5.88718235e-01 -8.75197768e-01 -1.27361596e+00 -9.38005149e-01 1.70443386e-01 -8.05995345e-01 -5.39702475e-01 -3.36452365e-01 -5.35402596e-01 2.16793910e-01 2.86040008e-01 6.79904938e-01 -2.20853180e-01 8.91224518e-02 9.23921391e-02 -6.11338854e-01 -4.14113969e-01 -7.54638851e-01 -2.62521044e-03 2.95325249e-01 3.56385350e-01 -3.29138577e-01 -7.20602989e-01 -3.31631631e-01 1.86752632e-01 -8.63792002e-01 2.55606454e-02 5.47949672e-01 9.10011053e-01 4.48786974e-01 2.52825230e-01 1.06215715e+00 -5.30734003e-01 7.29213297e-01 -2.91442990e-01 -3.48979115e-01 -1.32603317e-01 -6.42806411e-01 1.04107838e-02 1.03560364e+00 -5.94420612e-01 -1.51958978e+00 -7.66563341e-02 -7.04858184e-01 -5.80543995e-01 8.37760046e-02 3.99128944e-01 -8.40583265e-01 3.16967487e-01 5.40422380e-01 5.41540205e-01 -1.87449947e-01 -1.03077698e+00 5.83487868e-01 1.22693253e+00 9.88543868e-01 -3.88324201e-01 1.10732567e+00 6.46557286e-02 -6.42058372e-01 -1.01435971e+00 -6.16513848e-01 -4.44493145e-01 -4.97236878e-01 -6.27610534e-02 3.84346038e-01 -9.73462582e-01 -3.06932479e-01 3.81145746e-01 -1.21951425e+00 -3.16392966e-02 -5.04540980e-01 8.22219133e-01 -3.30287486e-01 5.51241100e-01 -6.89114809e-01 -1.14400184e+00 -5.72762907e-01 -9.23052430e-01 7.62122750e-01 -1.00066051e-01 -4.14989032e-02 -5.64995408e-01 1.38744205e-01 4.37602371e-01 6.20265424e-01 -2.09729552e-01 7.33503699e-01 -6.90224588e-01 -2.28629351e-01 -2.68993080e-01 1.77018970e-01 7.71232784e-01 3.09133649e-01 -2.34083533e-01 -1.68128324e+00 -2.75184631e-01 4.48810399e-01 -7.85123035e-02 6.35197222e-01 1.30315363e-01 9.34739232e-01 -8.29353690e-01 9.89432558e-02 7.02679276e-01 1.10546196e+00 1.66499764e-01 7.30861723e-01 -5.19429147e-01 6.86618865e-01 5.70228338e-01 5.86695969e-01 3.88832986e-01 6.92475364e-02 4.70461100e-01 1.04695983e-01 2.43833400e-02 -9.00835156e-01 -5.34509599e-01 7.22202301e-01 2.13942337e+00 1.64346188e-01 -4.60855782e-01 -3.03982764e-01 6.58639908e-01 -1.12577987e+00 -9.66636419e-01 -2.32399926e-01 2.13861799e+00 1.43991315e+00 9.21180472e-02 -5.53173525e-03 6.51935697e-01 8.29903364e-01 2.66064584e-01 -4.99776095e-01 -1.86323985e-01 -4.50374857e-02 3.39978814e-01 2.55368471e-01 6.96322620e-01 -5.52392125e-01 7.26306498e-01 6.47966957e+00 1.16307867e+00 -1.23819506e+00 5.21235108e-01 -1.11949198e-01 -2.86259174e-01 -6.38043344e-01 -2.26461634e-01 -6.62388623e-01 3.97254229e-01 1.32054150e+00 -9.50085744e-02 7.47651994e-01 5.02655089e-01 4.28714305e-01 4.15549576e-01 -9.37277138e-01 1.01230109e+00 1.44874632e-01 -8.14197302e-01 -2.68843621e-01 -1.61928132e-01 3.95882487e-01 -3.32931072e-01 5.91684170e-02 3.14967126e-01 -3.48230824e-02 -7.57783473e-01 1.23737288e+00 4.68436390e-01 1.37783694e+00 -6.25410497e-01 3.89948100e-01 4.54628706e-01 -1.43624127e+00 -6.95860684e-02 -2.17627198e-01 3.10128599e-01 1.88252494e-01 6.45995796e-01 -9.92063224e-01 7.61815071e-01 2.46384487e-01 1.36385038e-01 -1.46786764e-01 1.24157238e+00 -3.83713067e-01 1.27219307e+00 -1.08528383e-01 2.52475262e-01 -4.33999807e-01 -6.67963773e-02 9.03961778e-01 1.45092165e+00 4.19270903e-01 -4.32263762e-02 -1.96550995e-01 7.89490640e-01 -2.53148675e-01 2.09611908e-01 -4.00807232e-01 -3.13107282e-01 9.38642979e-01 1.20991230e+00 -1.12856338e-02 -2.64766663e-02 -3.05966347e-01 1.01184690e+00 -8.52125064e-02 5.39706469e-01 -9.14490998e-01 -7.72882164e-01 4.89948064e-01 -2.91231960e-01 3.66379142e-01 -1.71734273e-01 -2.10092559e-01 -1.15261090e+00 9.76816192e-02 -9.19286132e-01 -1.57233909e-01 -7.30411470e-01 -1.02993739e+00 7.82338321e-01 -5.07692397e-01 -1.59253120e+00 -3.62745732e-01 -1.33830473e-01 -6.57843292e-01 9.57400203e-01 -1.80031049e+00 -9.67410684e-01 7.96082318e-02 5.84586740e-01 9.09878373e-01 -1.43052846e-01 9.34447110e-01 6.43388152e-01 -5.33758223e-01 8.65408897e-01 3.30676049e-01 1.36479616e-01 8.04341674e-01 -1.10382581e+00 4.14852679e-01 1.03733981e+00 2.60112107e-01 3.41378659e-01 8.13891768e-01 -4.77935433e-01 -1.65770209e+00 -1.28748047e+00 7.49084115e-01 -1.68080047e-01 3.51511031e-01 -6.47361457e-01 -9.52094138e-01 3.39108080e-01 3.62767965e-01 -1.39331028e-01 1.01166403e+00 -1.60136074e-01 -4.57261294e-01 -3.77970129e-01 -1.07988369e+00 4.24765915e-01 8.05953979e-01 -9.35767233e-01 -8.21366072e-01 -1.51093379e-01 1.21562910e+00 -3.78974736e-01 -8.01931500e-01 5.53455502e-02 4.46565509e-01 -4.96231526e-01 8.81546557e-01 -3.72838676e-01 6.91869557e-02 -3.42520416e-01 -3.89070034e-01 -1.73268569e+00 -1.50112167e-01 -1.19555950e+00 -2.56553203e-01 2.04664159e+00 6.28829718e-01 -2.55985290e-01 8.47487375e-02 2.13101014e-01 -6.52907014e-01 5.41770235e-02 -8.58512282e-01 -9.37051833e-01 -1.69445634e-01 -9.13436353e-01 7.89211333e-01 7.85651743e-01 2.99442708e-01 4.55682546e-01 -8.91321957e-01 4.03131098e-01 6.11150086e-01 3.24182026e-02 6.08956277e-01 -1.08173048e+00 -2.48262361e-01 -1.67816654e-01 3.50317031e-01 -8.04714561e-01 4.70529273e-02 -7.68749654e-01 5.62716901e-01 -1.33538556e+00 -3.70996684e-01 -1.94083765e-01 -2.30339363e-01 1.78110614e-01 -2.00953022e-01 1.72075287e-01 3.72404665e-01 1.50767386e-01 -8.66111740e-02 8.96046281e-01 1.24959397e+00 -3.25884558e-02 -3.85518283e-01 1.41957894e-01 -6.73878610e-01 6.76885664e-01 7.40578413e-01 -6.50303483e-01 -5.35663784e-01 -2.80066341e-01 -2.66906738e-01 6.22622490e-01 4.46734205e-02 -1.06297719e+00 2.33302712e-01 2.44574845e-02 3.01175356e-01 -6.93718910e-01 8.04571092e-01 -9.35011864e-01 1.11089952e-01 3.22646946e-02 -3.65851700e-01 -5.66975057e-01 3.06564629e-01 5.38265646e-01 -3.60648870e-01 -3.72207671e-01 7.94367790e-01 2.92127818e-01 -7.23163113e-02 6.33992851e-02 -5.32564104e-01 2.79908389e-01 1.17742069e-01 1.24615900e-01 -2.53948569e-01 -5.89900672e-01 -5.52028596e-01 -2.24763781e-01 -1.03196867e-01 3.16824406e-01 7.41435230e-01 -1.56663406e+00 -8.25404167e-01 4.28461641e-01 -1.68850005e-01 -3.15355510e-01 3.65227371e-01 7.16784894e-01 1.48383260e-01 2.75187790e-01 5.16650528e-02 -1.33960247e-01 -1.00021946e+00 4.89856184e-01 1.11106485e-01 1.08249895e-01 -7.48409331e-01 5.54596126e-01 -2.83789098e-01 -3.48258764e-01 5.04768074e-01 -2.41013199e-01 -3.04708511e-01 2.55423099e-01 7.34709561e-01 5.46367228e-01 4.30662960e-01 -7.08491504e-01 7.17041781e-03 3.16917539e-01 3.09848964e-01 -7.56375134e-01 1.19496346e+00 -5.37523389e-01 2.04627275e-01 4.70170677e-01 1.25445235e+00 9.54354405e-01 -1.12727368e+00 -3.72341335e-01 -2.32365847e-01 -5.47373414e-01 3.18093777e-01 -9.09527779e-01 -1.14251649e+00 1.19139898e+00 3.94457400e-01 2.84916550e-01 1.39762163e+00 -4.71151471e-01 1.28966808e+00 3.42817903e-02 7.97894225e-02 -1.24184632e+00 1.40726313e-01 2.78815269e-01 1.17270601e+00 -7.71164358e-01 -4.00053054e-01 -4.81934220e-01 -7.68124044e-01 9.46087062e-01 2.22253442e-01 3.03748280e-01 5.17230928e-01 6.46534562e-01 2.68793225e-01 3.40029478e-01 -6.29652143e-01 -2.71410733e-01 5.00160038e-01 8.85623574e-01 1.03626683e-01 -1.03934474e-01 3.13636810e-02 1.28228462e+00 -5.86442709e-01 -3.60535413e-01 5.50297141e-01 4.52128887e-01 -5.63148439e-01 -1.22873902e+00 -7.83516645e-01 2.17132345e-01 -7.49890506e-01 -4.48924005e-01 -2.71961659e-01 2.22713202e-01 -1.30319491e-01 1.62212694e+00 -3.46025288e-01 -6.22614622e-01 5.45679152e-01 5.87994397e-01 1.18868805e-01 -3.86266530e-01 -4.01122332e-01 8.34593654e-01 4.43619758e-01 -1.54163644e-01 3.46385650e-02 -6.58532619e-01 -1.22980666e+00 -6.15066811e-02 -8.76946270e-01 2.81017423e-01 8.93619776e-01 7.78055310e-01 3.44856024e-01 1.00591421e+00 1.13598561e+00 -6.79536402e-01 -7.54591405e-01 -1.31201673e+00 -6.71095371e-01 8.49643052e-02 9.05535758e-01 -2.01656163e-01 -4.78904605e-01 5.54679811e-01]

[15.001486778259277, 6.201258659362793]

1096e5d5-52de-41da-9681-d3987b39b683

casteism-in-india-but-not-racism-a-study-of

https://aclanthology.org/2022.lateraisse-1.1

https://aclanthology.org/2022.lateraisse-1.1.pdf

Casteism in India, but Not Racism - a Study of Bias in Word Embeddings of Indian Languages

In this paper, we studied the gender bias in monolingual word embeddings of two Indian languages Hindi and Tamil. Tamil is one of the classical languages of India from the Dravidian language family. In Indian society and culture, instead of racism, a similar type of discrimination called casteism is against the subgroup of peoples representing lower class or Dalits. The word embeddings measurement to evaluate bias using the WEAT score reveals that the embeddings are biased with gender and casteism which is in line with the common stereotypical human biases.

['Aravindan Chandrabose', 'Aman Chandra Kumar', 'Pranav Tiwari', 'Senthil Kumar B']

lateraisse-lrec-2022-6

['culture']

['speech']

[-6.72634363e-01 5.00879213e-02 -3.01756233e-01 -4.76556331e-01 4.96365279e-01 -6.71894252e-01 1.16071594e+00 3.71138871e-01 -1.04993916e+00 7.27135360e-01 8.29647601e-01 -5.81967473e-01 9.65592861e-02 -9.65949237e-01 -4.33037952e-02 -7.34578848e-01 2.28995711e-01 7.63961434e-01 -3.86610746e-01 -8.96435916e-01 8.46596301e-01 3.07353348e-01 -9.44117010e-01 -4.06393498e-01 9.82692182e-01 1.01671904e-01 -4.63301927e-01 3.76873940e-01 -5.96106611e-02 6.16958380e-01 -3.19542289e-01 -7.57251620e-01 2.07388237e-01 -4.17738199e-01 -6.07313931e-01 -5.04685044e-01 8.37531984e-01 -3.99197847e-01 -5.52820206e-01 1.47769403e+00 7.39301443e-01 -3.82784367e-01 1.02344239e+00 -8.92295897e-01 -1.37190843e+00 8.78209472e-01 -8.00426602e-01 3.02797079e-01 1.45937636e-01 -3.91866803e-01 9.72162306e-01 -1.24093676e+00 8.28676999e-01 1.86932218e+00 6.13206327e-01 6.16428733e-01 -1.19695604e+00 -9.98292267e-01 9.24320966e-02 1.12863027e-01 -1.71429884e+00 -1.80698425e-01 7.17333078e-01 -8.15071881e-01 -7.03917071e-02 3.80021870e-01 7.22522914e-01 1.13786817e+00 6.29312932e-01 2.84669399e-01 1.71086586e+00 -1.66868806e-01 6.56337738e-02 7.09702492e-01 3.88113797e-01 4.25561994e-01 8.34050179e-01 1.31747022e-01 -2.44102180e-01 -3.93266946e-01 5.81653953e-01 1.03857614e-01 2.41659939e-01 -3.14482450e-01 -1.33912933e+00 1.50924599e+00 3.12326521e-01 3.85210574e-01 -3.24425497e-03 3.75123285e-02 6.12703919e-01 2.35266179e-01 5.56693673e-01 3.95138025e-01 -1.50227651e-01 -9.73630399e-02 -8.09173703e-01 5.66292644e-01 3.88992101e-01 2.95774430e-01 5.04278719e-01 3.83019120e-01 5.70254177e-02 1.13125491e+00 5.52984715e-01 1.01611435e+00 5.72142720e-01 -6.76625848e-01 5.31273782e-02 6.45597517e-01 -1.43947765e-01 -1.28431773e+00 -2.68232167e-01 -2.01265052e-01 -6.81660652e-01 5.16158998e-01 6.09705508e-01 -1.11402594e-01 -8.23388457e-01 1.82098520e+00 1.08457908e-01 -8.58795941e-01 -2.03252956e-01 1.07132685e+00 5.35154164e-01 3.17115843e-01 3.05175394e-01 1.55203283e-01 1.55098689e+00 -2.70254076e-01 -6.35831237e-01 -1.63514480e-01 5.18881917e-01 -7.88881004e-01 1.08845413e+00 4.06846665e-02 -5.31753719e-01 -4.22145069e-01 -9.96477365e-01 6.40865862e-02 -6.77803516e-01 -3.47194880e-01 4.79975730e-01 1.47667682e+00 -8.25998724e-01 4.11047488e-01 -2.63163865e-01 -7.57562637e-01 5.32205284e-01 6.53894693e-02 -4.30793285e-01 -4.61634584e-02 -1.24774277e+00 1.22566247e+00 1.61542073e-01 -2.62086242e-01 -6.35217130e-01 -7.38470972e-01 -8.31257641e-01 -7.24084079e-01 -4.60641891e-01 1.19286127e-01 4.10761505e-01 -1.05525947e+00 -7.57030308e-01 1.50730050e+00 1.54079035e-01 -7.46995509e-02 6.57795310e-01 -3.65960538e-01 -9.49736416e-01 -4.39497799e-01 3.36744070e-01 5.70598125e-01 4.88637179e-01 -1.35566044e+00 -3.27362388e-01 -8.94115508e-01 8.35074559e-02 -2.36106515e-02 -2.79875785e-01 5.47263384e-01 5.67455590e-01 -9.23082829e-01 6.09530732e-02 -1.18398011e+00 -1.81204289e-01 -4.80810523e-01 -3.91186386e-01 -1.41240597e-01 6.67578459e-01 -9.33389127e-01 1.52934802e+00 -2.24013638e+00 -1.41310319e-01 4.42327857e-01 5.51492333e-01 -4.84828949e-02 2.60636330e-01 7.44351149e-01 -1.88452885e-01 3.27018410e-01 -1.24152094e-01 6.30662858e-01 4.76904333e-01 5.33735394e-01 -3.76196623e-01 9.90636587e-01 -3.10033299e-02 5.17205596e-01 -1.20960772e+00 -5.99529505e-01 1.26142707e-02 3.11670214e-01 -8.32121670e-01 -4.42108274e-01 7.80618846e-01 5.01385808e-01 -1.60504039e-02 6.89294577e-01 1.14151680e+00 6.84002280e-01 4.87523824e-01 1.43323317e-01 -5.70981979e-01 2.40136057e-01 -7.35681713e-01 8.96034002e-01 2.32599564e-02 7.78971314e-01 -2.27824688e-01 -6.06403470e-01 9.23268318e-01 -1.92188203e-01 -2.01158002e-01 -9.90136027e-01 4.89138931e-01 6.97077751e-01 7.85892546e-01 1.42305065e-02 1.01122248e+00 -3.75864446e-01 -6.11213446e-01 4.96936828e-01 -3.52755077e-02 -1.21169843e-01 2.49167025e-01 1.61327317e-01 1.26226604e-01 -9.70901772e-02 4.42549258e-01 -1.34463310e+00 4.01541620e-01 -3.30388874e-01 7.60729373e-01 4.59982604e-01 -5.40047526e-01 6.37489140e-01 7.14947522e-01 -7.39154041e-01 -1.35362506e+00 -1.67018521e+00 -6.18093729e-01 1.40801275e+00 -9.19261854e-03 -1.38785616e-02 -5.35532534e-01 -5.28018713e-01 2.43225291e-01 1.03915203e+00 -9.60790575e-01 -1.83393434e-01 -6.80453300e-01 -9.88660038e-01 7.64066994e-01 2.65711993e-01 2.92663991e-01 -6.66317999e-01 -6.76140070e-01 -4.34739292e-01 4.04478580e-01 -4.70527977e-01 -5.35717726e-01 -7.45828226e-02 -5.53727686e-01 -8.48456740e-01 -1.05605221e+00 -7.69028902e-01 6.97478592e-01 -4.59273815e-01 1.10893404e+00 -9.55897942e-02 -7.20543638e-02 -1.40350759e-01 -1.33404076e-01 -6.47081852e-01 -5.66769063e-01 -2.13337719e-01 6.67689919e-01 -3.06953818e-01 8.94503415e-01 -4.00569230e-01 -7.08729267e-01 1.32824779e-01 -5.79981148e-01 -3.97036552e-01 2.01649711e-01 3.98745805e-01 -2.30330259e-01 -4.68494982e-01 2.12504461e-01 -1.19285703e+00 5.27524412e-01 -5.83060026e-01 -7.18504116e-02 -3.23767543e-01 -9.02748883e-01 -1.35628760e-01 2.76464045e-01 -3.90878528e-01 -7.78866291e-01 -8.68326485e-01 -1.30780473e-01 4.96499389e-01 1.12666033e-01 1.70447633e-01 1.58691958e-01 1.18522644e-01 5.59590280e-01 5.45412712e-02 -2.62930959e-01 -3.11563432e-01 3.29640001e-01 9.53648865e-01 3.34626108e-01 -6.48856640e-01 8.17060411e-01 5.12050807e-01 -4.02708143e-01 -1.18042707e+00 -2.44534060e-01 1.72216311e-01 -8.01883996e-01 -3.72023135e-01 1.16950560e+00 -9.03183579e-01 -4.55626070e-01 4.39356983e-01 -9.05163288e-01 1.27654985e-01 -1.82901666e-01 6.95948541e-01 1.25973806e-01 1.90986946e-01 -5.78684807e-01 -9.54654038e-01 -2.89032273e-02 -7.53921270e-01 5.77083111e-01 8.99062455e-02 -9.09286916e-01 -1.21636856e+00 6.10043049e-01 -8.33218023e-02 3.84313524e-01 3.93374234e-01 1.38609278e+00 -9.54173982e-01 5.84427416e-01 -3.04300059e-03 -4.16529328e-01 4.01818663e-01 1.05877519e-01 8.82169828e-02 -7.48860300e-01 -3.70896578e-01 -1.64797768e-01 -2.00885504e-01 5.96352220e-01 1.59995109e-02 3.38189512e-01 -1.45642549e-01 1.29905283e-01 8.67843479e-02 1.61364079e+00 5.31928957e-01 5.57394147e-01 4.14381653e-01 8.37583482e-01 6.32462859e-01 3.08580935e-01 3.84981275e-01 5.72659075e-01 2.39391595e-01 5.43813966e-02 6.52036676e-03 8.50717798e-02 -4.49326962e-01 5.50288618e-01 1.30629063e+00 -4.27715510e-01 5.51336288e-01 -1.36958826e+00 9.78936493e-01 -1.18142414e+00 -8.67000341e-01 -3.19357723e-01 2.27978230e+00 7.62815535e-01 2.32618991e-02 2.44618729e-01 2.53943890e-01 7.01614618e-01 5.38669109e-01 -2.57603936e-02 -1.31958961e+00 -2.02012897e-01 4.20495957e-01 9.98834252e-01 6.70363009e-01 -7.86719561e-01 8.64406347e-01 7.26266575e+00 5.12598217e-01 -1.21144319e+00 1.91767782e-01 5.00283897e-01 1.63740307e-01 -6.19241357e-01 -1.16906390e-01 -5.10933042e-01 5.86113572e-01 8.58357012e-01 -4.96150136e-01 1.19217731e-01 4.27996427e-01 -4.02642824e-02 1.10043630e-01 -8.68320048e-01 8.15756381e-01 1.36609092e-01 -5.56008637e-01 1.06922686e-01 5.20255506e-01 6.77257359e-01 1.06359996e-01 6.13339603e-01 4.40233797e-01 7.03366876e-01 -1.23257196e+00 1.12963390e+00 2.51176558e-03 7.49266267e-01 -1.09056675e+00 7.78818369e-01 -1.97442695e-01 -3.81468236e-01 -2.41753295e-01 -7.21751392e-01 -6.75883174e-01 -1.61486432e-01 3.38999420e-01 -2.25345850e-01 1.20638553e-02 6.49746001e-01 3.20342153e-01 -9.02571440e-01 2.02097520e-01 -5.14559299e-02 6.34582996e-01 2.47902319e-01 4.16582674e-02 5.16281724e-01 -7.47510672e-01 4.68980014e-01 1.32814801e+00 3.23309720e-01 -5.55723548e-01 -2.13355988e-01 5.34242570e-01 2.36350819e-01 3.48993957e-01 -8.52279663e-01 -3.70756477e-01 3.64338964e-01 1.06225204e+00 -8.06627750e-01 -2.26347908e-01 -4.15411770e-01 7.65771925e-01 -5.69272749e-02 1.44218057e-01 -8.19752216e-01 -3.82934242e-01 1.18538439e+00 3.53040665e-01 -2.95967638e-01 -4.14572895e-01 -4.80693489e-01 -8.71860802e-01 -7.27828205e-01 -9.14498627e-01 1.94324508e-01 4.18730415e-02 -1.28889430e+00 2.37083763e-01 -4.41383906e-02 -8.16056252e-01 2.34015703e-01 -9.32134628e-01 -4.21889871e-01 7.55766392e-01 -8.28577340e-01 -1.06301987e+00 4.37377244e-01 3.45784605e-01 -5.26111349e-02 -5.24084330e-01 9.70613956e-01 5.82057476e-01 -4.52437967e-01 6.16703808e-01 2.36357048e-01 5.43800116e-01 8.96909893e-01 -1.48442936e+00 3.63934606e-01 3.66538197e-01 -2.14857846e-01 1.00734103e+00 1.08629525e+00 -6.08717680e-01 -8.66555452e-01 -7.30426967e-01 1.45058382e+00 -6.66466892e-01 9.22348976e-01 -6.14762425e-01 -2.72907078e-01 6.83524370e-01 7.09120035e-01 -4.77249205e-01 1.04646468e+00 3.50861937e-01 -8.53620708e-01 8.25012103e-02 -1.10903060e+00 1.01888919e+00 9.43206012e-01 -7.58823156e-01 -8.78666222e-01 -2.64032930e-02 3.70817512e-01 2.24109307e-01 -7.27003932e-01 -2.46052444e-02 1.14894378e+00 -1.04892945e+00 6.88125014e-01 -6.05108202e-01 7.19392180e-01 -2.02936810e-02 -8.31815898e-01 -1.43151808e+00 -6.67469561e-01 3.88270505e-02 7.84704447e-01 1.32095253e+00 2.14879170e-01 -9.14958000e-01 4.56200719e-01 1.09362332e-02 3.00626963e-01 -1.33683234e-01 -1.07603407e+00 -6.11400545e-01 1.04354227e+00 3.65946591e-02 4.39952493e-01 1.54217482e+00 -1.21521972e-01 4.31903064e-01 -4.28726166e-01 -2.66912520e-01 7.12450385e-01 -2.52240509e-01 5.64996898e-01 -1.39666438e+00 4.86608446e-01 -6.56066120e-01 -7.27499127e-01 8.50713179e-02 1.42321423e-01 -1.15530527e+00 -4.67676103e-01 -1.24065471e+00 3.21082562e-01 -5.17148711e-02 -4.72240090e-01 -3.14461648e-01 2.88791470e-02 7.02764332e-01 2.93649495e-01 -3.09300162e-02 -5.41797690e-02 1.93006024e-01 1.03420556e+00 -1.60177514e-01 2.17769131e-01 -6.70509517e-01 -1.27259815e+00 1.13272786e+00 8.06969106e-01 -5.72321475e-01 -1.13231570e-01 -4.92901951e-01 6.28502071e-01 -6.80305481e-01 1.22401170e-01 -8.87365639e-01 -5.19216299e-01 -4.57448214e-01 5.64452469e-01 -3.02918464e-01 5.68641163e-02 -6.66328490e-01 -1.43934246e-02 1.11532664e+00 -7.85465091e-02 4.58502054e-01 -8.75477493e-02 -1.51377156e-01 -7.51379952e-02 2.35875487e-01 1.11092758e+00 -1.09039769e-01 -6.27859175e-01 3.16948444e-02 -8.92748356e-01 3.78631860e-01 7.91991115e-01 -3.40489060e-01 -3.82543027e-01 -4.44259942e-02 -1.08533829e-01 -8.44635814e-02 7.54130483e-01 7.18513548e-01 4.69727844e-01 -1.86514831e+00 -1.26142526e+00 1.11845046e-01 4.05232310e-01 -1.35039663e+00 -6.63007721e-02 8.45400095e-01 -9.00022447e-01 1.03714548e-01 -9.41365659e-01 -4.57228795e-02 -1.07388461e+00 2.93211311e-01 -8.28155428e-02 2.91796982e-01 -1.92548886e-01 5.04971504e-01 6.29257798e-01 -1.26439893e+00 -5.21902621e-01 2.08342314e-01 -5.79648077e-01 6.23620868e-01 3.71942133e-01 8.95361900e-01 -5.81966579e-01 -1.33918869e+00 -8.85779500e-01 5.66828728e-01 -9.56986398e-02 -3.43367338e-01 9.10953701e-01 -2.31965333e-02 -8.41272235e-01 9.78286982e-01 1.07443047e+00 9.03946996e-01 -1.72891408e-01 4.40361500e-01 2.11042613e-01 -7.66266525e-01 -3.70079368e-01 -6.08853936e-01 -6.48923039e-01 8.49439740e-01 1.02544224e+00 2.36051194e-02 2.19323203e-01 -3.26067239e-01 4.16704029e-01 -9.14171487e-02 4.17627960e-01 -1.67332506e+00 -4.75036711e-01 9.29706097e-01 8.73084843e-01 -8.71166587e-01 -3.21913362e-02 3.09488982e-01 -8.88199449e-01 6.66241050e-01 4.63377208e-01 -7.27659762e-01 9.20868576e-01 -1.10923402e-01 6.65490627e-01 -1.43874526e-01 -8.45196620e-02 1.22650433e-02 1.00058295e-01 7.71747530e-01 1.25490832e+00 6.11092031e-01 -1.37642443e+00 4.20451701e-01 -9.03961420e-01 -4.74197119e-01 7.75379241e-01 5.11291742e-01 -3.79821241e-01 -9.12342131e-01 -7.30624795e-01 3.57891619e-01 -7.25015938e-01 -1.02326423e-02 -6.18544459e-01 1.15589702e+00 7.32256830e-01 6.74414217e-01 4.80929166e-01 -5.12100577e-01 2.28033707e-01 2.48008147e-01 5.19771755e-01 -2.56063402e-01 -6.62030041e-01 -2.88447559e-01 1.72476336e-01 -1.23789981e-01 -1.18713647e-01 -3.97738487e-01 -1.04267466e+00 -1.07897437e+00 3.98334771e-01 1.82887062e-01 5.74038208e-01 5.57319880e-01 -4.69488919e-01 6.80002496e-02 4.02367115e-01 -4.88739491e-01 -2.74048835e-01 -9.88163710e-01 -1.26138890e+00 5.97640336e-01 8.39524791e-02 -6.24801397e-01 -2.26599216e-01 -6.32681787e-01]

[9.352226257324219, 10.191417694091797]

55bb423c-6afc-4fd3-9d28-8406a88ccda6

amplitude-independent-machine-learning-for

2305.14062

https://arxiv.org/abs/2305.14062v1

https://arxiv.org/pdf/2305.14062v1.pdf

Amplitude-Independent Machine Learning for PPG through Visibility Graphs and Transfer Learning

Photoplethysmography (PPG) signals are omnipresent in wearable devices, as they measure blood volume variations using LED technology. These signals provide insight into the body's circulatory system and can be employed to extract various bio-features, such as heart rate and vascular ageing. Although several algorithms have been proposed for this purpose, many exhibit limitations, including heavy reliance on human calibration, high signal quality requirements, and a lack of generalization. In this paper, we introduce a PPG signal processing framework that integrates graph theory and computer vision algorithms, which is invariant to affine transformations, offers rapid computation speed, and exhibits robust generalization across tasks and datasets.

['Danilo P. Mandic', 'Harry J. Davies', 'Yuyang Miao']

2023-05-23

['photoplethysmography-ppg']

['medical']

[ 1.31635964e-01 -2.42810994e-01 -3.74533534e-02 -3.59167278e-01 -2.44425200e-02 -5.88325024e-01 1.24399088e-01 4.63606454e-02 -2.69561559e-01 7.90120482e-01 2.21820876e-01 1.04512339e-02 6.69142976e-02 -3.39039356e-01 1.86998427e-01 -7.58139551e-01 -1.91553786e-01 -2.53459781e-01 -4.00148071e-02 1.41306028e-01 1.35789230e-01 4.95131940e-01 -8.74707699e-01 -4.43812519e-01 9.25878346e-01 1.23521686e+00 -6.56869292e-01 8.41272056e-01 2.31519014e-01 2.27758542e-01 -7.08851993e-01 -4.43891734e-01 3.63485277e-01 -7.72035837e-01 -4.65796664e-02 -1.54399559e-01 4.71724093e-01 -1.60121039e-01 -7.76283085e-01 7.80162275e-01 1.00131464e+00 -2.06076920e-01 4.55171704e-01 -1.27986050e+00 -3.80859822e-01 -7.39817023e-02 -5.06008029e-01 6.23838663e-01 5.46561956e-01 4.16691065e-01 4.30455357e-01 -6.76699936e-01 1.66631520e-01 7.39744186e-01 9.92653728e-01 5.79637229e-01 -1.47664821e+00 -5.41926622e-01 -6.79157376e-01 1.77117854e-01 -1.24911332e+00 -7.22530127e-01 1.13566470e+00 -3.26030016e-01 5.65852523e-01 5.48474073e-01 1.11564422e+00 8.61838281e-01 7.25526392e-01 2.82596678e-01 1.61830902e+00 -2.62879163e-01 1.64940551e-01 2.67437696e-01 6.86608776e-02 5.46508014e-01 5.83080113e-01 5.91445230e-02 -8.71509194e-01 -3.94428402e-01 8.97127390e-01 -6.58058329e-03 -8.13442767e-01 -4.43261862e-01 -9.19662476e-01 5.27340233e-01 -2.01910902e-02 3.16196740e-01 -5.39507389e-01 8.72389004e-02 5.78157365e-01 3.92810464e-01 2.98105657e-01 4.97190058e-01 2.51295678e-02 -3.11196715e-01 -8.15739155e-01 -1.49974540e-01 1.20398903e+00 5.41811824e-01 3.05533022e-01 3.22486043e-01 -2.31617048e-01 5.38295567e-01 4.02144432e-01 7.83116758e-01 5.51542819e-01 -9.88417804e-01 1.95798069e-01 4.84814733e-01 -1.34345874e-01 -1.44535398e+00 -8.98333013e-01 -2.19596848e-01 -1.02613246e+00 -5.58648631e-02 4.13271397e-01 -3.59663159e-01 -5.05535126e-01 1.43765557e+00 3.60252112e-01 3.52854908e-01 -2.06082001e-01 1.18328404e+00 1.12190759e+00 8.37716237e-02 2.57311970e-01 -4.60101008e-01 1.47003114e+00 -1.59126103e-01 -9.08379257e-01 -2.84812808e-01 5.79317734e-02 -5.17830849e-01 8.17710161e-01 3.94856364e-01 -9.63695586e-01 -5.12475550e-01 -1.18846548e+00 4.12817001e-02 -2.28989974e-01 -2.22567692e-01 6.59096897e-01 1.53325605e+00 -1.01213241e+00 6.23831391e-01 -8.16626966e-01 -4.25675690e-01 3.09546947e-01 1.87937185e-01 -2.12312862e-01 1.65163968e-02 -1.19640326e+00 9.61282253e-01 -1.27931193e-01 3.98120910e-01 -5.71264401e-02 -9.54128563e-01 -9.26794708e-01 1.34662380e-02 -9.16640759e-02 -8.37121308e-01 4.90077376e-01 -3.41951132e-01 -2.03575492e+00 7.72262692e-01 -1.92528486e-01 -3.13078940e-01 4.14302796e-01 9.92968529e-02 -7.90953457e-01 6.35029614e-01 -5.23336530e-01 -7.63917789e-02 1.02288520e+00 -5.08367836e-01 3.85769308e-01 -7.58177936e-01 -6.56384408e-01 2.11146668e-01 -4.17758167e-01 -1.24888413e-01 -1.49192020e-01 -4.13131267e-01 4.00847822e-01 -7.75242984e-01 -5.73358871e-02 1.89114273e-01 -8.48793238e-02 3.13025028e-01 5.72040141e-01 -7.20459759e-01 1.16127098e+00 -2.03918719e+00 -2.12597296e-01 5.68658888e-01 6.35899901e-01 4.60929036e-01 2.31507704e-01 2.86821336e-01 2.40860999e-01 5.25341649e-03 -1.09858312e-01 2.09347196e-02 -1.79975525e-01 -1.25431840e-03 -4.92559886e-03 1.05914164e+00 -2.06748068e-01 1.26501143e+00 -7.27428317e-01 -4.37997997e-01 7.29739487e-01 5.68248272e-01 1.63888156e-01 7.49021843e-02 9.21385050e-01 5.80086827e-01 -1.75225109e-01 7.87457705e-01 5.66968560e-01 3.21143717e-02 2.82831430e-01 -5.90450585e-01 1.65986374e-01 1.63296089e-01 -1.02382207e+00 1.37510061e+00 -1.17583565e-01 8.92421782e-01 1.88911378e-01 -8.10951948e-01 1.40791667e+00 5.73561013e-01 7.12735116e-01 -8.76924276e-01 3.47535610e-01 1.38679501e-02 1.61402956e-01 -8.16117167e-01 -1.44302403e-03 -5.31893790e-01 3.05277556e-01 4.23262149e-01 -3.83874960e-02 -3.77512515e-01 2.40316037e-02 -6.24266304e-02 9.41119075e-01 -8.27482194e-02 5.80932438e-01 -4.72522944e-01 6.50520086e-01 -4.76002038e-01 6.07598066e-01 4.59232301e-01 -7.02700436e-01 6.24403179e-01 4.82932925e-01 -5.53788722e-01 -7.24310994e-01 -1.00968945e+00 -3.55618268e-01 2.76266485e-01 3.28011006e-01 -3.08010191e-01 -4.01055574e-01 -6.52915612e-02 3.56179804e-01 1.68098092e-01 -5.28705835e-01 -4.25958037e-01 -2.88853317e-01 -8.87604594e-01 7.50598729e-01 6.33999169e-01 4.47583109e-01 -7.78975964e-01 -9.79100645e-01 4.16928321e-01 -1.36214018e-01 -1.24926305e+00 -4.77083772e-01 -5.45663536e-01 -1.30941904e+00 -1.31894946e+00 -7.49096394e-01 -1.15213260e-01 4.58925039e-01 1.28021583e-01 1.17186439e+00 -3.01681936e-01 -9.84865665e-01 9.74923015e-01 9.10397917e-02 -6.28817499e-01 -4.70312387e-02 -4.38053161e-01 3.04078549e-01 3.82664651e-01 7.05392361e-01 -7.96054840e-01 -1.17365491e+00 4.12266880e-01 -1.46751195e-01 -6.36320651e-01 4.84168261e-01 3.64441335e-01 4.57673818e-01 -4.33994412e-01 6.68200791e-01 -8.15369308e-01 1.04184842e+00 2.78421678e-02 -3.48111093e-01 5.71993999e-02 -9.46783364e-01 -6.78156734e-01 6.07446730e-01 -2.31889322e-01 -8.18487227e-01 -2.40991667e-01 4.47984874e-01 -3.74020785e-01 -7.10836500e-02 2.75896877e-01 2.54973501e-01 -6.63758636e-01 1.01904130e+00 4.27897751e-01 5.62189162e-01 -6.84584454e-02 1.36789933e-01 4.98101175e-01 8.55048716e-01 -3.75398934e-01 5.58976889e-01 4.33728248e-01 4.98918444e-01 -1.29172838e+00 -7.64865056e-02 -6.37673318e-01 -7.54283428e-01 -6.09518588e-01 4.38626796e-01 -7.60152876e-01 -1.07267296e+00 6.15214944e-01 -5.65769911e-01 1.61492765e-01 -1.76416352e-01 8.80864143e-01 -3.97928864e-01 8.22411656e-01 -5.86424768e-01 -1.00883746e+00 -9.68663216e-01 -3.59271884e-01 2.54561007e-01 6.06647730e-01 -4.63353932e-01 -1.32920516e+00 1.25933871e-01 1.66037157e-01 7.94141531e-01 9.18456912e-01 4.45817858e-01 2.29555145e-02 3.55220735e-02 -7.88128436e-01 -1.68222860e-02 3.36763561e-01 3.93524408e-01 -1.92182913e-01 -1.03867590e+00 -4.06047195e-01 2.59215951e-01 -5.41477185e-03 3.50051731e-01 8.60944271e-01 9.00672495e-01 -2.20240690e-02 -2.34361783e-01 7.92514145e-01 1.18751633e+00 2.14046896e-01 9.50533628e-01 -3.05388629e-01 5.37146449e-01 6.42796338e-01 -2.43516862e-02 4.69418019e-01 2.55108356e-01 1.73746556e-01 -1.17848225e-01 -4.59059715e-01 -1.14155255e-01 1.10823661e-01 5.96931167e-02 7.13704407e-01 -5.47081769e-01 3.31092328e-01 -5.54326594e-01 1.25611216e-01 -1.37504208e+00 -7.22053111e-01 -5.38812339e-01 2.38130498e+00 4.64573413e-01 -1.60212725e-01 5.08089781e-01 1.07342549e-01 6.29520953e-01 1.14118733e-01 -8.43884766e-01 -4.45198894e-01 -4.70946962e-03 4.25874054e-01 6.82698190e-01 3.44824791e-02 -6.93073869e-01 3.16479266e-01 8.23417473e+00 -4.29452449e-01 -1.33055294e+00 -1.62977993e-01 3.81362677e-01 1.20896250e-02 9.25316960e-02 -2.96323419e-01 -1.30563065e-01 5.21342516e-01 9.19871449e-01 -6.85741186e-01 3.02437901e-01 3.92522812e-01 4.40887004e-01 -1.87002048e-01 -9.56527352e-01 1.44130993e+00 3.82778317e-01 -7.99843669e-01 -7.05083191e-01 4.73559983e-02 4.97443452e-02 -2.74299264e-01 -9.09077376e-02 -2.22735628e-01 -7.11821735e-01 -8.13171744e-01 -2.93451548e-01 7.16639698e-01 9.27109361e-01 -3.21880490e-01 5.33186078e-01 -3.56150091e-01 -1.05056560e+00 2.08478063e-01 -4.99610633e-01 -2.80884326e-01 3.38420458e-02 8.64404380e-01 -6.95997477e-01 3.87664944e-01 5.19954681e-01 7.29291558e-01 -4.44887817e-01 1.45551312e+00 -3.15190017e-01 6.76692009e-01 -3.28802109e-01 -2.41220996e-01 -5.40340602e-01 -5.71399808e-01 5.49845815e-01 1.00759125e+00 3.34824413e-01 4.49972302e-01 -3.45407985e-02 9.57790673e-01 1.82234317e-01 2.91445524e-01 -7.61209786e-01 9.81899574e-02 4.16756421e-01 1.59481251e+00 -6.44345045e-01 -2.62352735e-01 -7.10644424e-01 6.41412377e-01 -5.26585400e-01 5.55532038e-01 -5.30593514e-01 -8.39133024e-01 6.93003893e-01 2.47999877e-01 -5.19006789e-01 -3.93941581e-01 -8.12536836e-01 -1.26123512e+00 6.91483989e-02 -5.24088800e-01 5.82642734e-01 -5.54258525e-01 -1.25603497e+00 2.72076577e-01 -3.67908835e-01 -1.25415874e+00 -1.04883358e-01 -5.72100461e-01 -7.86254346e-01 1.24675536e+00 -1.50798166e+00 -6.55561149e-01 -7.60618865e-01 8.41785610e-01 -1.61438614e-01 1.58935692e-02 7.94928670e-01 2.37419695e-01 -5.53616762e-01 5.93300641e-01 -4.49769139e-01 3.75563391e-02 8.03327680e-01 -1.32412946e+00 1.89145014e-01 9.00857627e-01 -2.95701027e-01 6.71185970e-01 5.67794263e-01 -3.15903604e-01 -1.93138313e+00 -5.62477589e-01 5.80974042e-01 -3.27114910e-01 3.64531785e-01 -1.58748463e-01 -8.27688575e-01 1.38206854e-01 -2.66414545e-02 4.21472192e-01 1.10782766e+00 7.28896782e-02 -1.09256327e-01 -6.33506179e-01 -1.40147710e+00 3.75188291e-01 5.96700490e-01 -5.12055874e-01 -6.56063318e-01 -3.98605503e-03 -3.16250056e-01 -6.27013266e-01 -1.53921640e+00 1.64336145e-01 9.47013259e-01 -8.58993411e-01 8.66425335e-01 -2.16033384e-01 -2.98689812e-01 -5.08705527e-02 5.96270263e-01 -1.25962889e+00 -4.86645609e-01 -1.17317414e+00 -2.56434530e-01 1.01584685e+00 3.59319039e-02 -1.41788697e+00 7.55432487e-01 1.45436084e+00 -3.21538723e-03 -3.36551398e-01 -9.43143129e-01 -7.80388057e-01 -4.29261267e-01 -5.33407554e-02 2.91745573e-01 9.59823072e-01 8.11070740e-01 2.64453351e-01 -7.42461264e-01 -8.74381214e-02 9.62861121e-01 1.65803805e-01 7.03698814e-01 -1.52190483e+00 -6.64515197e-02 -2.77962625e-01 -7.24818766e-01 -4.12699372e-01 -3.53928417e-01 -6.64104342e-01 -2.87172198e-01 -1.32131350e+00 -1.60695359e-01 3.41920406e-02 -6.38417006e-01 4.64849979e-01 -1.56608522e-01 7.83697963e-01 -7.15124235e-02 1.02422737e-01 1.60769999e-01 1.71101123e-01 1.09078658e+00 2.51923501e-01 -8.09758484e-01 3.73458043e-02 -8.48072052e-01 3.44710171e-01 9.81533945e-01 -2.19815969e-01 -4.28794801e-01 2.00280651e-01 -1.10413052e-01 2.26996303e-01 2.78098464e-01 -1.08602750e+00 2.30431795e-01 2.46257529e-01 9.74349856e-01 9.73071903e-02 3.36425453e-01 -5.33504009e-01 1.73549786e-01 8.58020782e-01 1.23235874e-01 -5.86577505e-03 1.83954909e-01 4.09962654e-01 -1.77526101e-01 2.32938409e-01 8.74678731e-01 -1.80105731e-01 -3.65628958e-01 3.89883965e-01 -3.04962426e-01 1.95271552e-01 8.18449736e-01 -7.11144030e-01 -3.21093917e-01 -4.91471022e-01 -4.94034946e-01 1.67236298e-01 1.62772030e-01 1.89713597e-01 7.83269942e-01 -1.20292485e+00 -7.06862390e-01 5.07453918e-01 6.85620978e-02 -7.27894902e-01 2.98312604e-01 1.51046777e+00 -5.35164952e-01 2.96722054e-01 -5.36408603e-01 -6.80480003e-01 -1.23279727e+00 3.61170053e-01 5.79253256e-01 4.61932153e-01 -1.12396336e+00 2.95657873e-01 -2.46012345e-01 2.57394850e-01 -1.69781640e-01 5.14803976e-02 -2.32160494e-01 -8.32109246e-03 6.56876862e-01 7.30972528e-01 -9.97238699e-03 -2.96543181e-01 -4.60100085e-01 8.88864040e-01 5.64881027e-01 1.96330890e-01 8.92136991e-01 -4.07631218e-01 -1.88992560e-01 3.95957708e-01 8.46582055e-01 -1.92937538e-01 -9.27301109e-01 -1.45177826e-01 -1.48220420e-01 -9.51290786e-01 1.34143427e-01 -6.99670613e-01 -1.17639339e+00 9.33215559e-01 7.89483190e-01 5.42266369e-01 1.38458323e+00 -5.13878822e-01 6.55868173e-01 1.42025715e-02 2.46466860e-01 -1.12954664e+00 -1.35320216e-01 -2.85397410e-01 6.72031224e-01 -7.53285646e-01 4.81129289e-01 -6.44579947e-01 -6.24071896e-01 1.37571144e+00 4.20793384e-01 -9.08048823e-02 7.09155023e-01 1.97647631e-01 4.91758883e-01 -1.95019677e-01 -1.63932830e-01 -3.76688018e-02 6.18624926e-01 1.03915787e+00 5.55774629e-01 6.27259761e-02 -9.56815004e-01 7.91180879e-02 -6.01910912e-02 3.11459810e-01 6.16521001e-01 5.84742010e-01 -2.81029046e-01 -6.34667277e-01 -2.64223516e-01 6.76952422e-01 -6.91071928e-01 2.51090527e-01 -2.95649141e-01 6.72432482e-01 -5.37470162e-01 1.08360505e+00 -3.03230107e-01 -1.00678317e-01 6.56643450e-01 4.01158154e-01 7.28635192e-01 -1.28257513e-01 -2.90363997e-01 3.05181772e-01 -3.01137362e-02 -7.50296175e-01 -5.58531940e-01 -6.24991179e-01 -6.96598887e-01 -2.06656963e-01 -1.12884670e-01 -6.20797984e-02 8.09988081e-01 6.46537066e-01 6.49977863e-01 2.01267868e-01 6.90638423e-01 -5.60470521e-01 -2.47377858e-01 -9.99153316e-01 -1.23257446e+00 5.53299606e-01 1.81034505e-01 -4.37380880e-01 -4.33143109e-01 -2.57823300e-02]

[13.935588836669922, 2.906017541885376]

b197dfc3-9f9e-47d0-8311-4cfd3fef29b6

introducing-randomized-high-order-fuzzy

2201.02158

https://arxiv.org/abs/2201.02158v2

https://arxiv.org/pdf/2201.02158v2.pdf

Introducing Randomized High Order Fuzzy Cognitive Maps as Reservoir Computing Models: A Case Study in Solar Energy and Load Forecasting

Fuzzy Cognitive Maps (FCMs) have emerged as an interpretable signed weighted digraph method consisting of nodes (concepts) and weights which represent the dependencies among the concepts. Although FCMs have attained considerable achievements in various time series prediction applications, designing an FCM model with time-efficient training method is still an open challenge. Thus, this paper introduces a novel univariate time series forecasting technique, which is composed of a group of randomized high order FCM models labeled R-HFCM. The novelty of the proposed R-HFCM model is relevant to merging the concepts of FCM and Echo State Network (ESN) as an efficient and particular family of Reservoir Computing (RC) models, where the least squares algorithm is applied to train the model. From another perspective, the structure of R-HFCM consists of the input layer, reservoir layer, and output layer in which only the output layer is trainable while the weights of each sub-reservoir components are selected randomly and keep constant during the training process. As case studies, this model considers solar energy forecasting with public data for Brazilian solar stations as well as Malaysia dataset, which includes hourly electric load and temperature data of the power supply company of the city of Johor in Malaysia. The experiment also includes the effect of the map size, activation function, the presence of bias and the size of the reservoir on the accuracy of R-HFCM method. The obtained results confirm the outperformance of the proposed R-HFCM model in comparison to the other methods. This study provides evidence that FCM can be a new way to implement a reservoir of dynamics in time series modelling.

['Frederico Gadelha Guimarães', 'Petrônio Cândido de Lima Silva', 'Omid Orang']

2022-01-06

['time-series-prediction', 'univariate-time-series-forecasting']

['time-series', 'time-series']

[-2.39185914e-02 -2.78209150e-01 2.55251676e-01 -1.02420554e-01 4.40614372e-01 -5.10064900e-01 8.09077263e-01 4.94461991e-02 -1.07160650e-01 7.54443645e-01 -3.15539911e-02 -4.64131385e-01 -7.81676054e-01 -1.18738449e+00 -4.97636110e-01 -9.29684699e-01 -4.79436725e-01 4.15759206e-01 2.34318331e-01 -6.19414985e-01 6.63121045e-01 5.66835761e-01 -1.88911736e+00 8.84456411e-02 1.25531483e+00 1.08710980e+00 6.75429106e-01 4.07448292e-01 -1.24360524e-01 9.12995517e-01 -4.88185465e-01 3.68339390e-01 2.11852938e-01 -2.93878913e-01 -2.02864140e-01 -2.93450743e-01 -4.33101386e-01 -2.72081532e-02 -1.35876834e-01 8.35266292e-01 3.67735058e-01 6.39174938e-01 6.63523138e-01 -1.41182077e+00 -5.97270131e-01 5.43700218e-01 -1.25939012e-01 4.46447968e-01 -1.91724479e-01 -2.28496045e-01 2.14355677e-01 -8.83968055e-01 1.29594594e-01 8.63287687e-01 5.66051602e-01 -6.05876856e-02 -8.80799353e-01 -6.93970501e-01 2.56856531e-02 4.10783827e-01 -1.39996445e+00 -7.44285434e-02 8.93812895e-01 -6.47912860e-01 1.28670275e+00 1.80903897e-01 7.72409618e-01 1.04217567e-01 4.93664801e-01 -3.54360789e-02 1.46595931e+00 -6.25973344e-01 5.37250817e-01 3.90397757e-01 2.28195101e-01 3.32740724e-01 2.91822702e-01 4.44855899e-01 -3.74362588e-01 -1.68649126e-02 5.75470328e-01 1.93293914e-02 -1.05562359e-01 4.01937738e-02 -7.14396119e-01 8.68490934e-01 5.93475878e-01 8.47111940e-01 -6.70795977e-01 -1.54152498e-01 8.17891732e-02 4.16446537e-01 5.00895739e-01 1.86593190e-01 -1.47617459e-01 3.55947584e-01 -1.20150912e+00 -6.79697469e-02 7.56130040e-01 4.15326655e-01 4.73752409e-01 7.11339712e-01 7.93144926e-02 4.47109550e-01 3.94957393e-01 7.00284898e-01 7.91238844e-01 -5.45428693e-01 2.26782769e-01 8.64345908e-01 1.16206400e-01 -1.33829355e+00 -4.29560870e-01 -6.94935679e-01 -9.33373034e-01 2.70903975e-01 -2.48259336e-01 -1.54688001e-01 -7.96799719e-01 1.51027620e+00 -1.57745183e-01 6.78223848e-01 2.70960063e-01 6.39244378e-01 3.75827610e-01 1.22681499e+00 -4.72726859e-02 -4.32070762e-01 7.53004849e-01 -6.51179314e-01 -7.41108298e-01 9.64497179e-02 9.06339586e-02 -1.96942255e-01 4.11420494e-01 1.82811692e-01 -9.25632298e-01 -6.33183062e-01 -1.28039587e+00 5.80316424e-01 -1.04850471e+00 2.16904387e-01 5.37140191e-01 4.16454792e-01 -1.34968138e+00 8.08559716e-01 -6.78807676e-01 -2.82313943e-01 -1.42013028e-01 4.88929480e-01 6.36453880e-03 2.23121926e-01 -1.55233312e+00 1.47034895e+00 7.57403970e-01 7.71129608e-01 -5.04777670e-01 -4.34566081e-01 -4.88871604e-01 2.88729608e-01 -2.56881028e-01 -2.12335959e-01 4.79412138e-01 -9.27554607e-01 -1.40032411e+00 6.92030415e-02 -1.56180292e-01 -5.64496458e-01 2.97821850e-01 4.48922306e-01 -9.29702044e-01 -5.86822368e-02 -9.71237719e-02 2.42136251e-02 5.70230722e-01 -1.16478229e+00 -7.12282717e-01 -1.98615700e-01 -3.77883077e-01 1.42585129e-01 -4.02031839e-01 -3.80686164e-01 4.46838945e-01 -2.20753700e-01 1.49730131e-01 -6.31368279e-01 -1.25071853e-01 -6.26784801e-01 2.31776312e-01 -2.33855039e-01 8.67110372e-01 -9.00245488e-01 1.62440085e+00 -1.87467921e+00 6.58693761e-02 8.80749285e-01 -2.62194276e-01 1.73871070e-01 1.16672449e-01 7.79736876e-01 -9.49966535e-02 -1.49866717e-03 -4.87412244e-01 3.61178905e-01 -7.58118033e-02 3.39529723e-01 -2.73671448e-01 2.10233554e-01 8.42427760e-02 6.35223746e-01 -7.95893788e-01 -9.40816253e-02 6.19475305e-01 6.44300342e-01 1.59974277e-01 -1.61798805e-01 -9.74316150e-02 1.71587586e-01 -2.71740228e-01 3.11498821e-01 7.52941310e-01 -2.34470412e-01 3.00730586e-01 -6.71503460e-03 -8.33033562e-01 -1.42470628e-01 -1.31875479e+00 8.63154233e-01 -3.24957788e-01 6.35588765e-01 -2.22343206e-01 -1.24781382e+00 1.30547476e+00 3.78364503e-01 4.68171269e-01 -8.95640194e-01 8.05089921e-02 5.22931337e-01 1.60875455e-01 -4.32253480e-01 5.90589821e-01 -2.87022322e-01 4.52273846e-01 2.03014821e-01 -1.36062294e-01 3.37801456e-01 4.53670770e-01 -2.61487424e-01 4.77817267e-01 3.11461315e-02 5.76217696e-02 -7.10707009e-01 9.28304255e-01 1.04703799e-01 4.11797911e-01 3.53967607e-01 1.92465082e-01 -5.49072772e-02 6.62301481e-02 -2.53505588e-01 -8.78252447e-01 -7.68732667e-01 -2.43783444e-01 6.69922531e-01 2.08325014e-01 3.20611537e-01 -1.56926185e-01 2.86852866e-01 2.33586431e-01 1.24741220e+00 -7.03731298e-01 -1.89012811e-01 -5.10729373e-01 -1.15457726e+00 1.60211161e-01 3.58561248e-01 6.35496974e-01 -1.16465926e+00 -8.59873116e-01 4.88260508e-01 4.84014675e-02 -4.15618569e-01 2.04114288e-01 3.52057159e-01 -1.20730257e+00 -8.35450709e-01 -5.12604892e-01 -8.93923402e-01 7.54576385e-01 2.52181262e-01 9.43005800e-01 1.79496959e-01 1.86995432e-01 1.11238342e-02 -1.84777796e-01 -6.43627882e-01 -3.96000832e-01 -7.71125481e-02 -7.57931173e-02 5.23402691e-02 3.04071426e-01 -8.77696991e-01 -4.67494309e-01 1.96811855e-01 -8.27380419e-01 -4.53420216e-03 4.18898612e-01 6.34409070e-01 2.79086500e-01 8.79253507e-01 1.04311872e+00 -5.78709006e-01 9.66747940e-01 -1.01618850e+00 -9.06291783e-01 6.04821563e-01 -1.32349527e+00 -5.71553186e-02 4.84800607e-01 -1.77109510e-01 -1.17236662e+00 -1.70721292e-01 4.21597421e-01 -3.49613249e-01 4.67960268e-01 1.06903696e+00 3.19416225e-01 -1.21382758e-01 3.78477424e-01 7.10707307e-01 1.73890516e-01 -3.03978950e-01 -4.85982709e-02 5.93865573e-01 4.74125981e-01 -2.97033876e-01 8.61677945e-01 9.53295082e-02 3.69602710e-01 -7.66122699e-01 1.22773528e-01 -2.72138655e-01 -4.97142524e-01 -7.83928454e-01 5.17082155e-01 -7.80526519e-01 -7.64577925e-01 4.80232507e-01 -8.35441589e-01 -1.65129453e-01 -6.45749196e-02 5.60385287e-01 -1.36217639e-01 -8.89723673e-02 -3.89398187e-01 -1.37950873e+00 -3.80796313e-01 -5.39355338e-01 1.77095175e-01 4.72835183e-01 3.47374976e-01 -1.27274358e+00 -3.39208287e-03 -1.51250497e-01 1.06502891e+00 5.56608796e-01 1.15682435e+00 -3.70918930e-01 -4.59655434e-01 -1.31885573e-01 1.70577750e-01 4.77649927e-01 -6.62029088e-02 3.77346724e-02 -6.90128446e-01 -3.60016137e-01 2.22740293e-01 3.09241414e-01 6.04019403e-01 4.76868153e-01 4.22083974e-01 -2.47183263e-01 -1.99880928e-01 4.16228622e-02 2.03664422e+00 9.56322670e-01 7.79041886e-01 5.42568684e-01 9.66888890e-02 4.99952734e-01 3.21068585e-01 3.72509599e-01 4.98762190e-01 7.86425695e-02 3.29676151e-01 1.47566348e-01 3.98446769e-01 -1.18056275e-01 4.95063454e-01 1.22594428e+00 -5.21205187e-01 -2.73805588e-01 -8.16830158e-01 4.63305026e-01 -1.87485039e+00 -1.34003520e+00 -2.24258557e-01 2.21262598e+00 1.72672972e-01 8.37736800e-02 -1.04470134e-01 6.14087284e-01 8.89813304e-01 -3.68575715e-02 -4.78498727e-01 -4.08077091e-01 -3.97698134e-01 2.97959030e-01 5.49023509e-01 5.82940757e-01 -6.03574634e-01 2.56005615e-01 5.35850096e+00 2.81830937e-01 -1.47578025e+00 9.20886397e-02 1.97828338e-01 2.18572572e-01 -4.21721995e-01 4.88938466e-02 -2.99319744e-01 9.99871016e-01 1.40502977e+00 -3.42660874e-01 8.42500985e-01 4.31587577e-01 5.41772187e-01 -3.66129696e-01 -3.58925551e-01 4.91924226e-01 -6.19347394e-02 -1.25616753e+00 -6.30371124e-02 -1.42356202e-01 9.00028884e-01 1.73054097e-04 -1.85671136e-01 3.35761130e-01 -9.19607058e-02 -9.72080708e-01 7.61725247e-01 1.17040038e+00 2.52254218e-01 -7.08401144e-01 1.03427529e+00 5.41102767e-01 -1.40046012e+00 -5.00663936e-01 -3.75549197e-01 -3.76534522e-01 7.31634870e-02 4.82649088e-01 -4.29929405e-01 8.92877698e-01 6.37331605e-01 4.52235192e-01 -3.59542429e-01 9.16334569e-01 3.14188689e-01 7.16735482e-01 -5.19309819e-01 -2.17804834e-01 1.05007477e-01 -5.69005847e-01 1.89774171e-01 9.02083397e-01 8.15567553e-01 3.25225174e-01 -1.83125451e-01 8.63126457e-01 4.82262582e-01 2.81149410e-02 -5.35576642e-01 -1.58311442e-01 8.74241531e-01 9.72965419e-01 -8.94151628e-01 -3.20115030e-01 -2.40966663e-01 3.15072328e-01 -1.69014931e-01 4.56051618e-01 -5.66427410e-01 -3.26790750e-01 -3.57694179e-02 3.27265412e-01 2.95483708e-01 -1.06140323e-01 -3.27976435e-01 -7.93658853e-01 6.19544983e-02 -2.41864696e-01 2.74186403e-01 -9.02733326e-01 -1.28564548e+00 7.72742927e-01 1.69452220e-01 -1.30582917e+00 -3.50643069e-01 -5.13499677e-01 -9.11720932e-01 1.40982056e+00 -1.79945648e+00 -7.71168709e-01 -5.02348900e-01 5.11608660e-01 1.11338474e-01 -3.85867804e-01 8.06272209e-01 3.94381404e-01 -4.68898773e-01 -1.12651929e-01 6.67314291e-01 -3.53772253e-01 -1.87582642e-01 -1.03817427e+00 -2.66927123e-01 8.93644869e-01 -4.94338900e-01 4.96594489e-01 6.47629321e-01 -8.64410222e-01 -1.05340528e+00 -7.82127082e-01 1.17417550e+00 7.64981061e-02 4.34781849e-01 1.06141055e-02 -1.04578316e+00 2.98937321e-01 2.95239568e-01 -2.43298247e-01 2.75814086e-01 -4.34964061e-01 2.90359199e-01 -3.92766118e-01 -1.30041432e+00 6.16145730e-02 2.93364555e-01 -3.51069927e-01 -4.80741143e-01 3.12415399e-02 -1.13799155e-03 -8.17710627e-03 -1.04832137e+00 4.62473989e-01 5.14625967e-01 -9.63107586e-01 7.62802660e-01 -1.08518936e-01 2.03148544e-01 -7.38859892e-01 -2.00723603e-01 -1.40236306e+00 -5.99558592e-01 -1.58310711e-01 -3.66906941e-01 1.01887619e+00 4.00986224e-01 -1.17455733e+00 4.35261518e-01 6.20930910e-01 -1.32519215e-01 -5.50988615e-01 -9.04716492e-01 -7.24426568e-01 -1.03224047e-01 5.40027842e-02 8.41984987e-01 1.05981421e+00 1.02614097e-01 -9.14874207e-03 -1.99992597e-01 1.89769194e-01 5.10834932e-01 1.76518112e-01 -1.62581787e-01 -1.65565598e+00 -1.55685591e-02 -4.00660247e-01 -2.66350597e-01 -5.84161207e-02 -1.23355955e-01 -9.99444008e-01 -1.80326983e-01 -1.78968143e+00 -1.97463617e-01 -5.83460569e-01 -8.61731529e-01 2.15458423e-01 1.33431926e-01 -2.29113162e-01 3.97892505e-01 5.94045401e-01 4.34015095e-01 4.71645087e-01 8.39607120e-01 1.58473719e-02 -3.88659954e-01 2.96408385e-01 -2.66181141e-01 2.07694128e-01 9.94215429e-01 -2.41421044e-01 -1.01107395e+00 -1.45679444e-01 1.97440192e-01 5.10655820e-01 2.95939475e-01 -1.18107247e+00 5.97641170e-01 -2.15651110e-01 6.43346250e-01 -7.85782576e-01 1.33446202e-01 -1.18982422e+00 9.49941337e-01 7.12639153e-01 -2.45236047e-03 6.59515738e-01 1.88004494e-01 3.75803322e-01 -3.73432994e-01 -2.66618967e-01 3.68246883e-01 -2.20496476e-01 -9.63107347e-01 -2.02614814e-01 -3.37360948e-01 -6.80508316e-01 1.09938872e+00 -6.13572538e-01 -4.91710156e-01 -2.58547574e-01 -5.71302772e-01 4.08071011e-01 -8.04596171e-02 3.46924305e-01 4.13322389e-01 -1.12515783e+00 -6.78010702e-01 3.34264517e-01 -4.33012187e-01 -3.58270913e-01 3.91700894e-01 7.42059588e-01 -4.94221002e-01 6.47758126e-01 -5.61336219e-01 -3.11818480e-01 -4.20103133e-01 3.67572546e-01 5.36866486e-01 -1.22285888e-01 -3.00584614e-01 1.66409671e-01 -5.65495253e-01 -3.85300130e-01 2.63928645e-03 -1.72563776e-01 -7.14740574e-01 4.17035729e-01 1.06011949e-01 7.88330793e-01 2.83849299e-01 -6.53097749e-01 -3.50576818e-01 4.17423666e-01 6.62186623e-01 -1.40087098e-01 1.66181803e+00 -1.46447688e-01 -4.87311572e-01 7.43276834e-01 7.80921221e-01 -5.96119881e-01 -8.82207930e-01 6.06405102e-02 2.53273308e-01 2.57969219e-02 2.23376304e-01 -1.19186187e+00 -1.06715333e+00 5.46609163e-01 9.88322020e-01 3.63614023e-01 1.49218643e+00 -8.00430477e-01 2.15770856e-01 1.08934134e-01 3.76233071e-01 -1.23780096e+00 -5.41243374e-01 4.24035609e-01 8.95996392e-01 -7.50336766e-01 -1.37525633e-01 6.56663999e-02 -3.37165594e-01 1.23713124e+00 4.00736839e-01 -2.93702662e-01 1.17576659e+00 2.42771447e-01 -2.77287781e-01 -7.74591938e-02 -7.24462926e-01 3.36228162e-02 1.90484256e-01 2.51448989e-01 2.30412781e-01 3.18378121e-01 -6.28127217e-01 6.16064191e-01 -1.41943857e-01 5.59540749e-01 4.28301215e-01 1.14565539e+00 -6.14958763e-01 -6.41176641e-01 -4.08353657e-01 6.61136746e-01 7.93090016e-02 1.14627331e-02 3.05582792e-01 9.19941962e-01 2.55207658e-01 1.05915368e+00 2.33025178e-01 -5.76140404e-01 4.14420098e-01 3.46382141e-01 6.40451349e-03 7.64417872e-02 -8.43171477e-01 -3.08670312e-01 -2.38215417e-01 2.39085615e-01 -5.36544144e-01 -5.48021376e-01 -1.41846645e+00 -5.63881934e-01 -4.04180348e-01 6.58013344e-01 1.08523309e+00 9.24612761e-01 2.03215256e-01 6.15901649e-01 8.93910646e-01 -8.52989376e-01 -5.33525407e-01 -1.18658280e+00 -7.83797681e-01 -5.38145155e-02 3.30863893e-02 -7.67049313e-01 -5.77072680e-01 -1.13673002e-01]

[5.949615001678467, 3.022780656814575]

cc38add8-889f-47fe-9e79-befcc4dd3d4c

designing-fairness-in-autonomous-peer-to-peer

2302.04771

https://arxiv.org/abs/2302.04771v1

https://arxiv.org/pdf/2302.04771v1.pdf

Designing Fairness in Autonomous Peer-to-peer Energy Trading

Several autonomous energy management and peer-to-peer trading mechanisms for future energy markets have been recently proposed based on optimization and game theory. In this paper, we study the impact of trading prices on the outcome of these market designs for energy-hub networks. We prove that, for a generic choice of trading prices, autonomous peer-to-peer trading is always network-wide beneficial but not necessarily individually beneficial for each hub. Therefore, we leverage hierarchical game theory to formalize the problem of designing locally-beneficial and network-wide fair peer-to-peer trading prices. Then, we propose a scalable and privacy-preserving price-mediation algorithm that provably converges to a profile of such prices. Numerical simulations on a 3-hub network show that the proposed algorithm can indeed incentivize active participation of energy hubs in autonomous peer-to-peer trading schemes.

['Florian Dörfler', 'John Lygeros', 'Philipp Heer', 'Giuseppe Belgioioso', 'Andrew Irvine', 'Varsha Behrunani']

2023-02-09

['energy-management']

['time-series']

[-6.94983780e-01 4.53906029e-01 -3.45394254e-01 -4.57782345e-03 -3.97438347e-01 -9.51935530e-01 3.30358237e-01 2.94538438e-02 -2.21309841e-01 1.02636445e+00 -1.98144212e-01 -1.97033018e-01 -5.58260679e-01 -1.39353991e+00 -4.63007152e-01 -8.03522110e-01 -4.98019964e-01 5.44479430e-01 1.33406386e-01 8.86054430e-03 -1.01481058e-01 8.73920619e-02 -1.01090860e+00 -8.31727982e-01 8.02553415e-01 9.81003344e-01 -1.71571821e-01 6.33366048e-01 3.32298487e-01 4.64935958e-01 -3.32328141e-01 -3.38040650e-01 9.51763690e-01 -5.23063123e-01 -4.30374593e-01 -2.40524679e-01 -9.20286238e-01 -4.68827724e-01 4.86903451e-03 1.16276026e+00 3.65940541e-01 5.42942658e-02 3.68009776e-01 -2.12600732e+00 -5.96679628e-01 1.07647634e+00 -6.07151449e-01 -2.98113227e-01 5.20903431e-02 8.68032202e-02 1.32611668e+00 -2.18598977e-01 6.16274714e-01 4.21564937e-01 2.99538106e-01 7.13674963e-01 -1.21029019e+00 -9.80498374e-01 -2.90787101e-01 -2.75093466e-01 -1.22022450e+00 -1.83573440e-01 9.19387817e-01 2.59078354e-01 8.06604624e-01 7.25752950e-01 1.24533951e+00 -4.39834893e-02 1.93628937e-01 2.26359949e-01 1.26955402e+00 -1.03690341e-01 9.49751437e-01 -9.06695984e-03 -5.76905191e-01 -1.47837456e-02 5.48745394e-01 1.56296551e-01 -4.83069301e-01 -8.41680884e-01 8.92775476e-01 -2.86295921e-01 -1.03270046e-01 -1.04813528e+00 -6.96010768e-01 9.03064191e-01 5.57731032e-01 1.00599959e-01 -8.06547582e-01 6.89376116e-01 7.17136115e-02 4.91931438e-01 4.32055175e-01 2.81392246e-01 -5.15066087e-01 -1.68949098e-03 -9.14802492e-01 3.09420466e-01 1.34754229e+00 9.11070764e-01 5.96612513e-01 -1.74160495e-01 2.48811133e-02 1.22943573e-01 6.16610765e-01 5.65529227e-01 -1.99864686e-01 -1.58230352e+00 -1.76064029e-01 3.77688229e-01 7.49122977e-01 -6.35062099e-01 -3.02220017e-01 2.87278861e-01 -9.77265716e-01 3.92431438e-01 1.06322713e-01 -7.61887491e-01 7.19403476e-02 1.92464900e+00 6.41280651e-01 5.36084063e-02 3.05428673e-02 7.99675882e-01 1.19499400e-01 6.14702404e-01 1.20629311e-01 -9.46855605e-01 1.19468999e+00 -6.95155859e-01 -6.31259322e-01 5.30921996e-01 1.79264218e-01 -1.43070698e-01 3.06921303e-01 -3.32147151e-01 -1.61474311e+00 6.61839247e-01 -6.73309505e-01 5.26212931e-01 -3.56646568e-01 -7.69870996e-01 2.13274419e-01 1.07065392e+00 -1.80514348e+00 5.47011375e-01 -5.41385591e-01 -5.68309605e-01 3.04196656e-01 8.19140494e-01 2.33261809e-01 8.07122648e-01 -9.90135789e-01 4.65708226e-01 1.90864548e-01 -1.88382015e-01 -9.06400084e-01 -8.35453212e-01 -2.39688292e-01 4.99212831e-01 7.98782408e-01 -1.14731979e+00 1.39019012e+00 -5.72953582e-01 -1.73486781e+00 5.71698904e-01 2.46496812e-01 -8.65855157e-01 8.72713506e-01 6.82879269e-01 -4.63301875e-02 1.54501140e-01 2.53703266e-01 6.58633649e-01 3.26521397e-01 -1.19477260e+00 -7.30712533e-01 -2.00144704e-02 3.69443148e-01 4.71686184e-01 -3.19846898e-01 2.03773022e-01 3.12730163e-01 -1.10360578e-01 -4.03099239e-01 -7.74897754e-01 -6.42402411e-01 3.29648465e-01 -3.82715017e-01 -5.94520926e-01 7.63017118e-01 1.68455541e-01 5.06805241e-01 -1.75046051e+00 -3.61679047e-01 7.44766235e-01 5.39761901e-01 -5.59534132e-01 2.50883132e-01 6.02466226e-01 5.87680399e-01 6.42591476e-01 -6.36078417e-02 -1.65580973e-01 7.34570444e-01 3.71481210e-01 2.63524413e-01 5.03018558e-01 -4.78126258e-01 1.07212865e+00 -9.09420609e-01 -1.10211007e-01 2.25098088e-01 -1.78023756e-01 -6.59624338e-01 1.46951914e-01 -3.21181983e-01 3.96788090e-01 -7.54585624e-01 3.80246431e-01 7.20894873e-01 -3.18891406e-01 6.19750202e-01 4.60621864e-01 -3.27159286e-01 -2.64739066e-01 -1.04967642e+00 9.43285882e-01 -7.50238672e-02 5.05954362e-02 1.18365765e+00 -5.62449992e-01 4.69037741e-01 5.55810928e-01 1.19064820e+00 -4.26454544e-01 3.79842669e-01 5.26560724e-01 -2.62584776e-01 4.26252574e-01 5.53323805e-01 -3.34277898e-01 -4.07247275e-01 9.87026811e-01 -4.94427085e-01 -7.55476877e-02 -1.11826628e-01 2.99946874e-01 1.10093164e+00 -5.14894485e-01 5.54599226e-01 -9.81050134e-01 4.11366642e-01 -2.06422210e-02 7.42045343e-01 6.74582601e-01 -8.62308681e-01 -3.94585967e-01 5.79219878e-01 1.67236939e-01 -1.07935011e+00 -9.72018957e-01 5.59849322e-01 8.24009478e-01 6.66895270e-01 -3.00110251e-01 -9.81810629e-01 -4.67294574e-01 2.25472674e-01 6.66164041e-01 -2.89456785e-01 2.43300453e-01 -2.97735278e-02 -4.33895290e-01 -3.98081262e-03 1.49471998e-01 6.05328023e-01 -9.42579865e-01 -9.69138682e-01 3.69759828e-01 6.87764212e-02 -7.29156077e-01 -9.31471407e-01 9.16056335e-02 -4.16026145e-01 -1.30000353e+00 -4.59409148e-01 -3.30501437e-01 8.69376779e-01 4.54640210e-01 1.02219212e+00 1.71143144e-01 3.45219553e-01 1.03280842e+00 5.89186400e-02 -3.71561736e-01 -3.93224120e-01 9.26018432e-02 1.88706681e-01 -1.11510798e-01 -4.08038199e-02 -9.59292591e-01 -1.28589034e+00 5.40389776e-01 -6.57318830e-01 -1.25736848e-01 3.80528048e-02 2.73966998e-01 8.47652435e-01 4.42539603e-01 1.05795276e+00 -5.12428224e-01 1.15122521e+00 -6.54437959e-01 -1.36010051e+00 3.25034410e-01 -1.12494445e+00 -3.51067036e-01 8.33625019e-01 1.77053407e-01 -6.45830572e-01 1.44658461e-01 6.03049636e-01 -2.65590191e-01 3.54078293e-01 -9.43045616e-02 -5.54173648e-01 -6.36976182e-01 -3.41777176e-01 7.84285367e-03 -4.55903733e-04 3.24118771e-02 1.05107617e+00 3.07202697e-01 2.13794231e-01 -7.40551889e-01 1.25577760e+00 6.02984190e-01 2.24851131e-01 -2.63629079e-01 5.88504612e-01 -1.43717989e-01 -2.48250421e-02 -4.66016948e-01 8.72751772e-01 -1.01715207e+00 -1.64676130e+00 4.16048318e-01 -6.24025941e-01 -3.73562694e-01 -9.57343876e-01 -1.72778428e-01 -8.26090991e-01 1.24123573e-01 -3.58317345e-01 -1.33566737e+00 -7.41140008e-01 -7.57488966e-01 2.31918097e-01 7.90241599e-01 3.08885891e-02 -9.05082703e-01 1.20652981e-01 1.15791701e-01 1.06256616e+00 1.95171550e-01 4.04500216e-01 -6.70496106e-01 -1.30552137e+00 1.76715359e-01 -4.11950834e-02 -3.21794629e-01 1.15399234e-01 -8.17135274e-02 -5.19012451e-01 -6.16250396e-01 1.84647720e-02 -8.84960815e-02 -1.32692084e-01 5.77389717e-01 6.98236644e-01 -9.58443046e-01 -3.87883574e-01 1.48234740e-01 1.79857063e+00 3.41717243e-01 2.92805135e-01 2.37255052e-01 2.14424267e-01 5.39960861e-01 2.61966497e-01 8.00199866e-01 9.89710987e-01 1.94001973e-01 7.94880927e-01 1.98723450e-01 8.27374876e-01 -3.86130244e-01 1.83320910e-01 4.53327626e-01 1.71859607e-01 -2.26521388e-01 -1.11279480e-01 9.11786079e-01 -1.89252961e+00 -6.75950468e-01 1.11180618e-01 2.21145248e+00 5.24321795e-01 -1.47796616e-01 7.48915434e-01 -3.17471415e-01 7.08456933e-01 8.32210779e-02 -9.83517408e-01 -5.69490969e-01 -8.12040865e-02 3.37300040e-02 1.37110281e+00 2.54591256e-01 -4.76583838e-01 5.27098656e-01 6.40691423e+00 8.59856367e-01 -4.92980272e-01 5.81938088e-01 8.72489870e-01 -3.56268138e-01 -9.14845407e-01 2.76693612e-01 6.47969320e-02 3.91851693e-01 1.32192791e+00 -1.44164610e+00 1.03337014e+00 8.65635812e-01 8.48332822e-01 -2.93930918e-01 -8.42096150e-01 5.37136793e-01 -1.05351186e+00 -1.64817846e+00 -6.91833258e-01 6.99477613e-01 1.03976238e+00 3.30225676e-02 -3.81450057e-01 -2.49486312e-01 1.23008049e+00 -6.01909637e-01 7.33890891e-01 -1.50389401e-02 2.04983711e-01 -1.24133384e+00 1.75209478e-01 2.39147052e-01 -1.69323337e+00 -2.62363106e-01 -1.89661369e-01 1.46853980e-02 3.77715230e-01 3.80993187e-01 -1.25103239e-02 7.85319269e-01 8.88264716e-01 -1.77518887e-04 1.40862986e-01 1.05197608e+00 -1.80742100e-01 3.01992774e-01 -8.65832746e-01 -2.85229474e-01 6.56830966e-02 -6.84987843e-01 5.75144410e-01 1.22109860e-01 4.68664527e-01 5.66935122e-01 1.58225760e-01 1.38739693e+00 -9.10918295e-01 1.00306697e-01 -6.45374238e-01 1.48538090e-02 1.17596889e+00 1.63739955e+00 -1.08412755e+00 -8.52828249e-02 -2.86618680e-01 9.13743258e-01 -3.56836110e-01 1.85211703e-01 -6.23451412e-01 -1.06136128e-01 1.09539080e+00 -1.00357421e-02 4.95703518e-01 1.12869889e-01 -5.16987383e-01 -6.43716276e-01 -8.69830176e-02 -1.33287296e-01 5.90331495e-01 -6.81493580e-01 -1.48564613e+00 -3.86316478e-01 -2.62126535e-01 -6.74735069e-01 -8.16665590e-02 3.35998267e-01 -1.55378902e+00 5.18707335e-01 -1.58137715e+00 -7.07150519e-01 4.09762204e-01 7.61810482e-01 -2.50007033e-01 1.92111015e-01 6.34292603e-01 3.07651274e-02 -2.28195801e-01 3.67696077e-01 6.28876925e-01 -4.38697226e-02 2.35685408e-02 -1.46142244e+00 2.02896342e-01 9.11221147e-01 -1.82261482e-01 1.16991997e-01 5.65328598e-01 -5.19617856e-01 -1.84364200e+00 -1.00850034e+00 5.31244218e-01 -1.05982877e-01 1.09241760e+00 -4.39211071e-01 -3.31585616e-01 2.50190824e-01 1.00637043e+00 7.47458637e-02 5.44927359e-01 -6.55484617e-01 3.88029575e-01 -4.86095130e-01 -2.16771030e+00 5.93340933e-01 1.00020707e+00 -5.51189303e-01 3.77139002e-01 2.34987155e-01 8.63983154e-01 2.30345502e-02 -1.02303457e+00 -7.06583336e-02 2.59264857e-01 -7.35252917e-01 6.13122463e-01 3.83658037e-02 -5.15683115e-01 -3.94521356e-01 -9.24566686e-02 -1.30056918e+00 -5.27358055e-03 -1.57647371e+00 7.97036942e-03 1.40009892e+00 1.97128475e-01 -1.26714373e+00 1.21449959e+00 1.30362833e+00 5.61289728e-01 -3.82228255e-01 -1.88740742e+00 -9.46098328e-01 6.89094007e-01 1.79191932e-01 1.09645414e+00 6.92013979e-01 4.14865226e-01 -2.19738230e-01 -3.37659210e-01 3.49355966e-01 1.40630400e+00 2.34621778e-01 4.40448940e-01 -9.63893652e-01 -4.04086746e-02 -6.26399636e-01 1.99080944e-01 -5.18052936e-01 6.04745895e-02 -6.45974696e-01 2.06459478e-01 -1.54977942e+00 2.44399890e-01 -6.55220449e-01 -4.59940225e-01 3.74952525e-01 4.67465997e-01 1.46476412e-02 7.57630587e-01 2.73260146e-01 -9.28887367e-01 7.53807068e-01 9.84308958e-01 -8.09018463e-02 -2.18536004e-01 7.67839095e-03 -1.11238396e+00 -1.47956172e-02 1.11067724e+00 -6.09268904e-01 -6.93121314e-01 5.19789398e-01 2.30729088e-01 6.11135304e-01 4.15023714e-01 -3.53815556e-01 7.82270074e-01 -6.47734880e-01 -4.64108646e-01 -4.86078858e-01 -1.15308493e-01 -1.36440289e+00 1.00651848e+00 8.17038774e-01 -5.60206659e-02 -4.51937951e-02 -4.77850914e-01 1.07632303e+00 4.22313958e-01 2.72128105e-01 4.97968107e-01 -2.23019704e-01 -2.95080662e-01 5.92681825e-01 -7.03371346e-01 9.02591348e-02 1.91350102e+00 -1.07538961e-01 -3.64239126e-01 -8.96159708e-01 -4.87041354e-01 1.31722724e+00 8.65345895e-01 1.95957571e-01 2.96579599e-02 -1.32766569e+00 -5.94390273e-01 -2.59023786e-01 -4.44498807e-01 -2.64734417e-01 -1.14630505e-01 7.38251925e-01 -7.32186995e-03 5.45200557e-02 -1.06545277e-01 -9.64865610e-02 -9.10146654e-01 3.22889477e-01 8.61158907e-01 -2.60723025e-01 -1.40417114e-01 3.31101060e-01 -1.02483280e-01 -5.94364941e-01 -6.39093816e-02 -1.53553978e-01 5.32062173e-01 -1.17278099e-01 -2.12834269e-01 6.74432397e-01 -6.93786442e-01 -3.75483632e-01 -3.90587717e-01 3.73430163e-01 6.74558699e-01 -4.07462806e-01 1.25581157e+00 -7.27486372e-01 -6.07186496e-01 -1.61913067e-01 5.89569807e-01 -1.42416015e-01 -1.12766790e+00 2.50065356e-01 6.51065409e-02 -3.70419860e-01 -1.07668921e-01 -5.95154703e-01 -1.78841984e+00 -1.67845458e-01 4.50061351e-01 1.03653419e+00 1.26202130e+00 -9.56117213e-02 7.80043006e-01 9.82217267e-02 1.20174789e+00 -1.41051710e+00 -7.43938804e-01 -4.14388210e-01 1.43282771e-01 -8.42167914e-01 -3.22881669e-01 -4.29101765e-01 -2.39486694e-01 5.93099356e-01 4.08526778e-01 -1.48312792e-01 1.20733368e+00 2.56988734e-01 -3.52822870e-01 -1.96124420e-01 -9.20571685e-01 -5.16771227e-02 -6.86762571e-01 7.85207033e-01 -5.12604475e-01 1.00967014e+00 -6.19067311e-01 6.79194987e-01 4.96808179e-02 3.23977083e-01 1.01523960e+00 1.05804539e+00 -3.70539099e-01 -1.49785471e+00 -9.98338163e-02 3.51121724e-01 -3.81901830e-01 2.78714448e-01 -7.08559990e-01 5.64807117e-01 -1.82282105e-01 1.44367719e+00 -2.61007436e-02 9.58527401e-02 1.90917626e-01 -1.62248850e-01 -4.76536117e-02 4.06159535e-02 -1.01196206e+00 1.77302957e-01 -1.70065593e-02 -6.93216324e-01 -7.23948419e-01 -5.53017497e-01 -1.41026199e+00 -1.03267622e+00 -3.81427318e-01 8.21165621e-01 3.81173909e-01 4.68585938e-01 5.68956971e-01 5.85846044e-02 1.31779647e+00 -3.50321084e-01 -6.99356019e-01 1.54964980e-02 -1.26624537e+00 -2.65395641e-01 4.22827564e-02 -1.24013238e-01 -6.89849615e-01 -7.03193903e-01]

[5.5230255126953125, 2.5851080417633057]

8e2decd0-58c4-4b14-b5ec-5135bfe0b048

breast-density-classification-with-deep

1711.03674

http://arxiv.org/abs/1711.03674v1

http://arxiv.org/pdf/1711.03674v1.pdf

Breast density classification with deep convolutional neural networks

Breast density classification is an essential part of breast cancer screening. Although a lot of prior work considered this problem as a task for learning algorithms, to our knowledge, all of them used small and not clinically realistic data both for training and evaluation of their models. In this work, we explore the limits of this task with a data set coming from over 200,000 breast cancer screening exams. We use this data to train and evaluate a strong convolutional neural network classifier. In a reader study, we find that our model can perform this task comparably to a human expert.

['Nan Wu', 'Kyunghyun Cho', 'Krzysztof J. Geras', 'Stacey Wolfson', 'Linda Moy', 'Eric Kim', 'S. Gene Kim', 'Jingyi Su', 'Yiqiu Shen']

2017-11-10

['breast-density-classification']

['medical']

[ 1.76468223e-01 5.69886088e-01 -4.90285605e-01 -6.71794176e-01 -8.10020983e-01 -3.10989112e-01 5.02741456e-01 4.94244248e-01 -7.05200255e-01 7.34453797e-01 1.02294097e-02 -9.20695007e-01 1.74099430e-01 -8.90503466e-01 -8.27018321e-01 -3.42220187e-01 -2.20129013e-01 7.52081096e-01 6.62338018e-01 1.04443058e-01 -1.41410723e-01 3.93804431e-01 -1.05095410e+00 5.33978462e-01 7.26760983e-01 1.07766140e+00 -1.00268513e-01 8.89478147e-01 1.26793340e-01 1.07624698e+00 -4.49660808e-01 -5.78579724e-01 -8.62463415e-02 -2.58608192e-01 -1.15787983e+00 -2.41013825e-01 4.41892803e-01 -4.46091890e-01 -3.31569314e-01 8.77037883e-01 5.98259747e-01 -4.57376063e-01 8.88578832e-01 -5.97835422e-01 -3.87712389e-01 5.69122732e-01 -4.03935403e-01 7.53678501e-01 7.96861351e-02 -1.38847053e-01 5.29140115e-01 -5.91374278e-01 4.89268124e-01 8.01567018e-01 1.13027740e+00 7.34591842e-01 -8.96832943e-01 -5.49869299e-01 -4.74187642e-01 -4.99408282e-02 -1.12514973e+00 -5.27021766e-01 3.37896943e-01 -5.01481771e-01 6.28811300e-01 1.36590764e-01 8.48867953e-01 1.13015795e+00 3.86186659e-01 9.67376709e-01 1.05139756e+00 -6.18050396e-01 4.86261666e-01 6.24404848e-01 3.79988670e-01 7.85576940e-01 6.52527630e-01 1.61646545e-01 1.36908870e-02 -4.14801985e-01 7.97021985e-01 -3.70569497e-01 -2.55351625e-02 -3.17697525e-01 -7.16192305e-01 1.05709636e+00 4.33159858e-01 4.06100333e-01 -8.20805728e-02 1.42467558e-01 4.55801129e-01 1.51782826e-01 6.75187409e-01 3.12736720e-01 -2.48360500e-01 -1.63418755e-01 -1.07729995e+00 1.18126661e-01 1.12271380e+00 4.79092389e-01 2.00048342e-01 -5.44892490e-01 -8.91822726e-02 7.50936687e-01 1.98607609e-01 1.08869255e-01 5.66853762e-01 -5.01638532e-01 1.37393147e-01 6.04505718e-01 -1.61912128e-01 -8.96036327e-01 -7.59779036e-01 -2.57077873e-01 -1.05905366e+00 1.56267837e-01 7.83672929e-01 -4.54137415e-01 -1.03023934e+00 1.22621286e+00 3.01714033e-01 2.64346421e-01 -6.59509972e-02 6.65810168e-01 1.08320725e+00 1.61252379e-01 2.17288882e-01 1.15001857e-01 1.26825917e+00 -4.88787681e-01 -4.66284215e-01 -2.98911929e-01 1.01963449e+00 -5.28756082e-01 5.07356405e-01 4.41389173e-01 -1.11279047e+00 -3.63465935e-01 -1.01150537e+00 -7.70321861e-02 -2.91439801e-01 2.32114464e-01 1.30128503e+00 1.03959739e+00 -1.00840747e+00 6.78144038e-01 -1.39569438e+00 -6.62728906e-01 8.95776212e-01 3.03211898e-01 -4.30150926e-01 -2.69946903e-01 -1.14037120e+00 1.19068611e+00 3.92408669e-01 -2.48448737e-02 -8.62091303e-01 -6.30166769e-01 -9.93288159e-01 -2.31996834e-01 1.25417382e-01 -5.48408329e-01 1.87306607e+00 -7.15248108e-01 -1.08630288e+00 1.16109550e+00 2.40178569e-03 -8.87770534e-01 6.02818847e-01 1.10498503e-01 -2.65960127e-01 -1.28665213e-02 -1.83662280e-01 5.77918887e-01 2.74425417e-01 -9.00438488e-01 -7.88273871e-01 -2.52054751e-01 1.90086709e-03 -1.87385067e-01 -4.20115441e-01 -1.12729706e-02 -5.60575008e-01 -3.00817847e-01 -8.73664171e-02 -8.71744037e-01 -5.47893345e-01 1.82483211e-01 -1.08331569e-01 -1.98963031e-01 6.23319268e-01 -5.45973420e-01 1.14584661e+00 -1.87047672e+00 -2.61930078e-01 1.47024050e-01 2.68563658e-01 4.96217281e-01 3.25594723e-01 -9.43192318e-02 -2.39497144e-02 3.42698127e-01 -1.59715876e-01 -1.26158297e-01 -3.75707924e-01 3.15680832e-01 3.29396844e-01 7.01159000e-01 4.53456104e-01 1.12726700e+00 -9.24543023e-01 -8.38482797e-01 1.42306060e-01 6.17415369e-01 -5.54814219e-01 1.90457731e-01 1.26462102e-01 -5.40565420e-03 -5.64661860e-01 8.24716985e-01 5.50721526e-01 -5.90569079e-01 3.33381265e-01 -1.53768733e-01 3.07936430e-01 4.97950390e-02 -7.93873250e-01 1.32962239e+00 -3.70933324e-01 7.24250972e-01 -5.14767542e-02 -1.25465810e+00 4.83609021e-01 3.09633285e-01 5.08696377e-01 -5.82679331e-01 3.59508753e-01 4.44482088e-01 5.69315851e-01 -5.38034976e-01 1.56703517e-01 -2.97394961e-01 5.80944940e-02 1.61296710e-01 2.56584555e-01 -1.57816291e-01 2.88924009e-01 2.17510462e-02 1.50927758e+00 -2.84719259e-01 6.19468272e-01 -5.87668240e-01 2.67737001e-01 2.71569431e-01 2.13581383e-01 8.26171458e-01 -3.40039223e-01 6.66860819e-01 7.42714703e-01 -8.24575245e-01 -9.55775201e-01 -7.84564853e-01 -8.69156957e-01 6.33224249e-01 -2.17349619e-01 -1.26541644e-01 -7.74883687e-01 -8.26965868e-01 7.76572749e-02 2.77202964e-01 -1.14983523e+00 -1.35625392e-01 -5.05208611e-01 -1.41050899e+00 6.64556682e-01 8.88461888e-01 3.13254237e-01 -9.95553613e-01 -6.89087808e-01 1.84247583e-01 2.76629865e-01 -8.34244967e-01 -9.49273258e-03 5.13696134e-01 -8.87090743e-01 -1.43835211e+00 -1.02850175e+00 -8.58278692e-01 5.85554838e-01 -2.47745514e-01 1.59033656e+00 1.50821716e-01 -6.89118207e-01 1.23597063e-01 -3.15733403e-01 -1.03549862e+00 -9.08488572e-01 4.38981235e-01 -4.37541366e-01 -5.62344015e-01 8.88823271e-01 -1.23950489e-01 -6.61247790e-01 1.31704703e-01 -9.40157950e-01 -7.63722137e-02 1.01338565e+00 9.36613262e-01 4.63675827e-01 2.10774504e-02 3.17436904e-01 -1.27441633e+00 7.10338473e-01 -8.00522268e-01 -4.84427631e-01 4.13048491e-02 -3.84875953e-01 -2.15943962e-01 2.13430971e-01 -4.07566607e-01 -7.60186434e-01 1.91227853e-01 -5.12991846e-01 8.48428532e-02 -2.47098863e-01 7.26882100e-01 3.22135270e-01 -1.71570435e-01 1.07602787e+00 -6.99794814e-02 5.29220849e-02 -6.09720051e-01 -2.73383975e-01 8.36994171e-01 2.97622859e-01 -2.57946312e-01 4.34713691e-01 3.97402018e-01 1.97401986e-01 -7.67616093e-01 -1.11937082e+00 -5.33619165e-01 -7.01070070e-01 3.79451029e-02 7.82969952e-01 -8.51931036e-01 -4.14202780e-01 4.15239602e-01 -8.26373219e-01 -5.21451592e-01 -2.91035831e-01 5.17560840e-01 -1.38618007e-01 1.69267744e-01 -8.05701971e-01 -7.38169968e-01 -1.94721475e-01 -1.01928568e+00 1.13292146e+00 1.20132364e-01 -2.27813691e-01 -1.38997722e+00 3.85807753e-01 9.31380019e-02 5.93554735e-01 4.91458565e-01 5.79485774e-01 -7.11982369e-01 4.31961529e-02 -7.15101421e-01 -4.03746694e-01 4.11554426e-01 2.23380476e-01 -6.55822828e-02 -1.12240183e+00 -2.72328585e-01 -1.07184179e-01 -7.86476791e-01 1.28641272e+00 7.67454863e-01 1.61444831e+00 1.53080314e-01 -8.85365069e-01 5.33387899e-01 1.21777844e+00 -4.40981798e-02 7.87194669e-01 2.64690518e-01 4.39716518e-01 3.79247725e-01 4.02368933e-01 3.55809219e-02 3.49892795e-01 2.92503476e-01 3.78052980e-01 -8.03682387e-01 -4.61837500e-02 8.25131778e-04 -2.62600243e-01 2.75549501e-01 -1.28140211e-01 -1.62755400e-01 -1.56523657e+00 6.21582389e-01 -1.49998343e+00 -6.79226041e-01 -8.36172104e-02 1.86096179e+00 1.06475031e+00 4.21062917e-01 1.10022701e-01 1.39259607e-01 3.44872147e-01 -2.60273337e-01 -3.73800993e-01 -2.31084391e-01 2.87520051e-01 4.22367215e-01 6.36873305e-01 7.47074410e-02 -1.53996885e+00 3.80323410e-01 7.97039700e+00 7.42831409e-01 -1.11000395e+00 -3.50972340e-02 1.45964229e+00 4.57296930e-02 1.08581729e-01 -3.79403144e-01 -6.73714757e-01 2.37730488e-01 1.26064086e+00 1.35730073e-01 -3.00804853e-01 1.07893336e+00 -2.48829708e-01 -3.53384882e-01 -1.46044159e+00 5.04579663e-01 7.41764754e-02 -1.32570672e+00 -3.42129499e-01 1.10868007e-01 6.30746365e-01 1.53792128e-01 -9.75399651e-03 4.90169078e-01 2.98489124e-01 -1.69735885e+00 1.09611973e-01 4.09109682e-01 7.96353459e-01 -5.52455246e-01 1.40227306e+00 4.13817912e-01 -6.93718493e-01 2.16153428e-01 -4.14673716e-01 1.46127746e-01 -2.76304483e-01 8.19454312e-01 -1.16333055e+00 2.27977052e-01 9.24013078e-01 4.95417148e-01 -1.01838517e+00 1.37403154e+00 4.26698565e-01 9.90814626e-01 -3.82309258e-01 -1.79943919e-01 1.86332092e-01 4.70319688e-01 -3.23362738e-01 1.55961537e+00 3.46852481e-01 -1.41039342e-01 -1.09969480e-02 4.75556612e-01 -1.50089860e-01 5.19204587e-02 -6.15803003e-01 -1.55341774e-01 -8.54157433e-02 1.43014264e+00 -7.36625612e-01 -2.61889070e-01 -5.97289622e-01 4.32602525e-01 4.37180430e-01 -1.20369934e-01 -7.52427876e-01 -1.34095609e-01 1.33579671e-01 5.19810796e-01 -4.56918404e-02 2.60349035e-01 -1.61993667e-01 -8.89876127e-01 -1.68327078e-01 -6.92014754e-01 4.65219378e-01 -1.90164864e-01 -1.50730896e+00 4.10260767e-01 1.47915572e-01 -9.99309182e-01 -4.66451198e-01 -1.15427136e+00 -6.94923043e-01 7.38893330e-01 -1.66883802e+00 -9.86681044e-01 -4.23304588e-01 1.26777828e-01 1.29007697e-01 -2.60417223e-01 9.62168455e-01 4.73917872e-01 -3.67726415e-01 6.00371301e-01 1.09444223e-01 5.80588222e-01 6.30507529e-01 -1.49736309e+00 3.89208615e-01 2.98590958e-01 -1.48742452e-01 4.02701885e-01 4.46645468e-01 -5.30451417e-01 -1.11807334e+00 -1.19466829e+00 5.75046957e-01 -6.06773794e-01 5.86198986e-01 -9.92882177e-02 -1.06432450e+00 7.30684698e-01 1.02468044e-01 6.88080907e-01 9.78471041e-01 2.44061992e-01 2.91123271e-01 2.59836435e-01 -1.23528802e+00 2.27499396e-01 7.10060656e-01 4.29657958e-02 -3.16491544e-01 4.70486015e-01 1.76407292e-01 -9.32334661e-01 -1.07085729e+00 9.20868099e-01 5.77004611e-01 -1.05897570e+00 7.70735502e-01 -6.40010297e-01 8.67083132e-01 3.35402399e-01 2.33084679e-01 -1.17573500e+00 -1.44236743e-01 2.60629300e-02 -3.12531143e-01 7.40891218e-01 6.44619644e-01 -2.17786103e-01 1.26574314e+00 4.16447610e-01 6.29027039e-02 -1.22578323e+00 -9.73499835e-01 -4.10666883e-01 7.67566502e-01 -3.43291104e-01 2.23407581e-01 8.06569517e-01 -2.67125517e-02 3.33436653e-02 -1.13983126e-02 -1.36287153e-01 4.28058267e-01 -3.82761270e-01 5.27713001e-01 -1.29584050e+00 -4.46758211e-01 -4.30024326e-01 -5.69200039e-01 -4.92833376e-01 -8.53863209e-02 -9.94747281e-01 -1.25390351e-01 -1.67000425e+00 6.46387935e-01 -5.89911997e-01 -2.71928251e-01 4.29535568e-01 -2.36705348e-01 5.33358753e-01 -4.85736787e-01 -1.62464708e-01 -4.12028491e-01 -1.11585118e-01 1.32806706e+00 -2.21602648e-01 1.07544191e-01 3.29802424e-01 -8.32527518e-01 8.51234853e-01 8.85957241e-01 -4.68980730e-01 -3.70430738e-01 -2.06702173e-01 2.61203405e-02 7.75947571e-02 4.38471794e-01 -1.14655221e+00 -5.94012067e-02 -1.17758289e-01 1.01854241e+00 -3.77292544e-01 7.21347108e-02 -6.80318892e-01 -5.35844341e-02 7.56511807e-01 -3.40795636e-01 -4.50314701e-01 5.94497800e-01 4.37535405e-01 -2.46144190e-01 -5.18483698e-01 8.20210814e-01 -4.53978837e-01 -4.73730743e-01 2.65185267e-01 -3.06623578e-01 5.97005412e-02 1.15330529e+00 -1.00294247e-01 -2.41895318e-01 -1.19909957e-01 -5.86149395e-01 1.49745956e-01 2.29290977e-01 6.58897683e-02 3.76204908e-01 -1.30199206e+00 -8.62284780e-01 7.09432736e-02 2.11268529e-01 4.08720255e-01 -8.94207656e-02 8.65023911e-01 -8.39300275e-01 5.94619334e-01 -6.00177981e-02 -8.22142899e-01 -1.23032033e+00 3.38851929e-01 7.61532366e-01 -5.83374083e-01 -5.10524631e-01 8.65006506e-01 -8.44520703e-03 -4.17945981e-01 3.44834596e-01 -5.27671218e-01 -4.49994594e-01 -1.62529916e-01 7.83305705e-01 -5.61428927e-02 2.77047664e-01 -1.42570227e-01 -3.53324354e-01 2.61975918e-02 -3.03377777e-01 2.66528457e-01 1.30028498e+00 5.48144937e-01 1.61417961e-01 4.41603899e-01 1.18547058e+00 -3.99124324e-01 -8.47281933e-01 -8.08466747e-02 1.10237025e-01 -3.08262259e-01 2.41337746e-01 -7.49724746e-01 -9.64745998e-01 8.79239440e-01 1.00041986e+00 4.08217728e-01 9.72620428e-01 2.14974046e-01 3.03217709e-01 5.55003881e-01 1.40489131e-01 -9.74056482e-01 5.66192307e-02 2.38690600e-01 5.39388239e-01 -2.01488638e+00 1.06203191e-01 -3.96898270e-01 -4.74562526e-01 1.11111712e+00 8.13287795e-01 -1.78321108e-01 1.00542629e+00 6.12802267e-01 1.06491074e-01 -2.83055037e-01 -6.30939960e-01 3.01450379e-02 5.54313004e-01 8.50258529e-01 1.15275621e+00 -1.35678828e-01 -3.78161132e-01 5.60600400e-01 -1.53081238e-01 6.79753721e-01 5.51341236e-01 1.10825109e+00 -5.30975282e-01 -1.08261907e+00 -1.31787822e-01 1.04648244e+00 -9.43429768e-01 5.55020720e-02 -3.51038337e-01 1.21824980e+00 7.65953735e-02 6.07702851e-01 1.92249104e-01 -3.37503105e-02 2.35283762e-01 -1.58084780e-01 5.45234621e-01 -9.01652634e-01 -4.23657596e-01 -2.57576883e-01 2.63629615e-01 -2.73633599e-01 -4.43324208e-01 -4.50802088e-01 -9.91544843e-01 -2.24267364e-01 -2.00265899e-01 -3.64478044e-02 4.61366415e-01 7.67868042e-01 -1.99731573e-01 8.10893774e-01 1.21901028e-01 -6.12894535e-01 -6.92350745e-01 -1.31885219e+00 -7.55983174e-01 4.80579764e-01 3.04356009e-01 -4.21137303e-01 -1.40702769e-01 -3.37864906e-02]

[15.202239036560059, -2.4722073078155518]

7e1fe4e1-c142-45f4-862f-5ac9f9621f9b

bengali-text-document-categorization-based-on

https://www.sciencedirect.com/science/article/pii/S0957417421008174

https://www.sciencedirect.com/science/article/pii/S0957417421008174

Bengali text document categorization based on very deep convolution neural network

In recent years, the amount of digital text contents or documents in the Bengali language has increased enormously on online platforms due to the effortless access of the Internet via electronic gadgets. As a result, an enormous amount of unstructured data is created that demands much time and effort to organize, search or manipulate. To manage such a massive number of documents effectively, an intelligent text document classification system is proposed in this paper. Intelligent classification of text document in a resource-constrained language (like Bengali) is challenging due to unavailability of linguistic resources, intelligent NLP tools, and larger text corpora. Moreover, Bengali texts are available in two morphological variants (i.e., Sadhu-bhasha and Cholito-bhasha) making the classification task more complicated. The proposed intelligent text classification model comprises GloVe embedding and Very Deep Convolution Neural Network (VDCNN) classifier. Due to the unavailability of standard corpus, this work develops a large Embedding Corpus (EC) containing 969,000 unlabelled texts and Bengali Text Classification Corpus (BDTC) containing 156,207 labelled documents arranged into 13 categories. Moreover, this work proposes the Embedding Parameters Identification (EPI) Algorithm, which selects the best embedding parameters for low-resource languages (including Bengali). Evaluation of 165 embedding models with intrinsic evaluators (semantic & syntactic similarity measures) shows that the GloVe model is more suitable (regarding Spearman & Pearson correlation) than other embeddings (Word2Vec, FastText, m-BERT) in Bengali text. Experimental results on the test dataset confirm that the proposed GloVe+VDCNN model outperformed (achieving the highest 96.96% accuracy) the other classification models and existing methods to perform the Bengali text classification task.

['Iqbal H.Sarker', 'NazmulSiddiqueb', 'Mohammed MoshiulHoquea', 'Md. RajibHossaina']

2021-07-02

expert-systems-with-applications-2021-7

['document-classification']

['natural-language-processing']

[-3.84148091e-01 -2.07274809e-01 2.23166943e-01 -5.45755215e-02 -2.03785107e-01 -7.63289809e-01 8.01696301e-01 4.51729059e-01 -8.58032048e-01 4.70493019e-01 2.86546171e-01 -4.96328235e-01 -3.53883594e-01 -1.13778234e+00 1.55010998e-01 -6.41938388e-01 1.20935217e-01 6.85469985e-01 -8.12883154e-02 -5.97627401e-01 4.74699676e-01 5.24903238e-01 -1.40969110e+00 1.39558583e-01 8.85167241e-01 8.48386586e-01 5.15327215e-01 9.23039198e-01 -4.17160779e-01 4.89988804e-01 -6.94717586e-01 -6.57526016e-01 7.65397623e-02 -1.31672129e-01 -9.15991724e-01 -2.29924783e-01 -1.13007255e-01 -2.57099688e-01 -6.68859184e-01 7.82696247e-01 8.67953598e-01 3.48086298e-01 6.74370706e-01 -8.60540867e-01 -1.27346623e+00 6.55709624e-01 -1.52823731e-01 2.63564020e-01 -1.34226665e-01 -4.79522735e-01 9.91107225e-01 -1.20391798e+00 4.66252178e-01 9.83100712e-01 4.97727245e-01 2.20901906e-01 -5.73519588e-01 -3.62842560e-01 -4.04395372e-01 4.97496933e-01 -1.45186067e+00 -7.18062520e-02 6.94682419e-01 -2.95710981e-01 1.22254467e+00 4.95528340e-01 3.46834987e-01 6.70213997e-01 2.39819348e-01 5.31621814e-01 8.28978479e-01 -8.41283083e-01 8.30054581e-02 7.91959882e-01 4.08102006e-01 4.26403999e-01 3.22073370e-01 -4.10675555e-01 6.61330819e-02 -1.23768754e-01 1.01069815e-01 1.05830684e-01 -1.18496612e-01 1.66745931e-01 -8.89031053e-01 1.10994279e+00 9.29813385e-02 7.76960611e-01 -2.24940136e-01 -4.55110729e-01 9.57840145e-01 4.46184188e-01 4.99787897e-01 2.79988497e-01 -5.24651766e-01 -2.87295401e-01 -5.70715427e-01 -2.66653728e-02 8.25047255e-01 8.88638675e-01 4.43587750e-01 3.38311136e-01 1.97121546e-01 1.32921207e+00 4.24188554e-01 4.09973532e-01 1.18109846e+00 1.03539564e-01 7.62432456e-01 8.61061931e-01 -2.56159753e-01 -1.52191329e+00 -3.51581901e-01 -1.67881534e-01 -9.52002048e-01 -2.64924228e-01 2.15301253e-02 -1.85159788e-01 -6.81944728e-01 8.75877976e-01 2.70173848e-01 -6.73721910e-01 4.94214147e-01 5.62358856e-01 1.06915033e+00 1.27428210e+00 8.92337039e-02 -1.07837081e-01 1.56991243e+00 -8.23823273e-01 -9.35327530e-01 1.57107860e-01 8.49387705e-01 -1.02136278e+00 1.00906074e+00 2.71723717e-01 -6.31789148e-01 -5.78646839e-01 -9.85899687e-01 -2.37041131e-01 -1.15634167e+00 4.09331650e-01 5.84193826e-01 1.02437985e+00 -7.43524134e-01 2.29915053e-01 -4.66161549e-01 -5.54086149e-01 1.40442714e-01 5.60754716e-01 -7.48774409e-01 -1.76505834e-01 -1.43202531e+00 8.62460196e-01 9.78266597e-01 2.61781693e-01 -1.32317647e-01 -1.30200371e-01 -9.15955961e-01 -1.85325388e-02 -2.04723239e-01 1.87535971e-01 4.24477637e-01 -6.87588871e-01 -1.20094800e+00 6.78455889e-01 3.00946504e-01 -1.41874105e-01 3.66239786e-01 4.16731462e-03 -8.40915203e-01 1.01423115e-01 -2.18355700e-01 2.01893017e-01 3.75847876e-01 -8.59247923e-01 -7.34868884e-01 -6.01100862e-01 -1.34829342e-01 3.83422613e-01 -1.27246130e+00 2.89009809e-01 -3.79589200e-01 -8.01102936e-01 9.22088847e-02 -7.26330101e-01 2.65041351e-01 -6.73153818e-01 -5.41888289e-02 -5.44453561e-01 1.42515469e+00 -1.14594591e+00 1.60749125e+00 -2.02057242e+00 -1.80616856e-01 1.16719604e-01 1.23200320e-01 6.66924119e-01 4.88420650e-02 8.54466856e-01 -1.03605697e-02 3.19948226e-01 -3.45920883e-02 1.14602655e-01 2.09153444e-01 3.61967981e-01 1.03685847e-02 4.58727151e-01 -8.71610194e-02 7.92161405e-01 -5.01105607e-01 -7.38708735e-01 4.06093925e-01 7.61467636e-01 -2.26886123e-01 1.47484401e-02 4.49243516e-01 -2.94903606e-01 -4.43945944e-01 8.49509716e-01 6.82797432e-01 2.69999206e-01 1.40703365e-01 -2.92798519e-01 -2.62974530e-01 5.25176525e-03 -1.24970770e+00 8.96993458e-01 -8.12082827e-01 9.54670846e-01 -2.74244219e-01 -1.17905664e+00 1.20258141e+00 6.00051165e-01 2.80914962e-01 -5.39747357e-01 5.89240789e-01 2.49921173e-01 1.50473535e-01 -8.85304570e-01 9.80918109e-01 9.07972157e-02 -1.34117380e-01 5.43846786e-01 2.62526572e-01 1.49450466e-01 2.99079210e-01 6.02170713e-02 9.25798953e-01 -3.98297906e-01 2.82555223e-01 -1.65603817e-01 8.13692391e-01 -1.86318457e-02 2.92325020e-02 5.14718294e-01 -2.95079589e-01 3.32338959e-01 5.45113571e-02 -5.60816646e-01 -1.40130460e+00 -3.75854820e-01 -5.58688939e-01 1.11358154e+00 -2.01875076e-01 -3.96179289e-01 -4.19023156e-01 -6.61575913e-01 -1.66880697e-01 6.09668493e-01 -5.26452720e-01 1.18097365e-02 -6.10124946e-01 -1.04020417e+00 7.43582070e-01 3.93043339e-01 8.90210152e-01 -1.31947148e+00 -2.10070670e-01 3.77599359e-01 -7.13968053e-02 -8.55091870e-01 -6.40657470e-02 2.20085859e-01 -7.09510207e-01 -6.89416885e-01 -6.02653563e-01 -1.25666988e+00 5.12555718e-01 3.74131240e-02 5.84078550e-01 -2.52640806e-02 -3.05808008e-01 1.71954527e-01 -9.38027322e-01 -4.69694942e-01 -4.90376055e-01 2.00071141e-01 6.68374300e-02 -1.11177683e-01 9.26378608e-01 -2.87679466e-03 -1.17806956e-01 7.27222860e-02 -1.28159273e+00 -3.47614914e-01 4.32566702e-01 1.08187640e+00 1.11720063e-01 8.53630066e-01 6.72487676e-01 -7.38886893e-01 1.04862559e+00 -5.98872781e-01 -4.40662354e-01 3.11220616e-01 -7.47452796e-01 -3.05477232e-01 9.30517614e-01 -4.92022961e-01 -9.12264645e-01 -5.28304219e-01 -3.94693017e-01 1.79645479e-01 -1.34161353e-01 7.47727215e-01 -2.48727426e-01 2.93291844e-02 3.64487827e-01 5.01163542e-01 -2.92449772e-01 -4.03816134e-01 1.48821354e-01 1.75943029e+00 2.85674669e-02 -2.52014607e-01 6.91032708e-01 1.03775740e-01 -4.73327756e-01 -1.22291946e+00 -1.01892784e-01 -5.00970185e-01 -8.61612201e-01 -2.68156128e-03 9.89969254e-01 -6.11379504e-01 -5.93064010e-01 5.08374810e-01 -1.19360471e+00 3.62064540e-01 8.04654062e-02 8.38775218e-01 1.40030116e-01 5.11990488e-01 -7.08736479e-01 -9.59366262e-01 -6.65919304e-01 -1.06063175e+00 4.49739784e-01 2.02970374e-02 -1.44231319e-01 -1.34005117e+00 -5.37406318e-02 5.71844935e-01 5.09301960e-01 -9.92941409e-02 1.28031516e+00 -1.19211507e+00 1.91276640e-01 -7.21364796e-01 -3.33623201e-01 7.88138926e-01 3.48379731e-01 1.95448939e-02 -9.02879834e-01 -1.76202089e-01 1.17931552e-01 -2.83521444e-01 4.76029694e-01 -7.64070153e-02 9.97603536e-01 -5.25688231e-01 1.14132926e-01 2.60302126e-01 1.57981884e+00 6.69651270e-01 6.59859896e-01 8.84832203e-01 7.89209366e-01 6.96317792e-01 5.24349749e-01 5.23014426e-01 4.00787264e-01 3.28302622e-01 1.53439507e-01 2.86937684e-01 2.82697499e-01 3.18670012e-02 1.74756423e-01 1.75863016e+00 -1.63987488e-01 -7.31329262e-01 -1.03164721e+00 6.11860812e-01 -1.36761689e+00 -8.49893093e-01 -1.92261368e-01 1.85418200e+00 8.75752807e-01 -1.72929913e-01 -2.91233867e-01 7.57809758e-01 6.89285040e-01 -2.58623194e-02 -2.62623746e-02 -1.02049267e+00 -2.15542763e-01 3.97591472e-01 4.54072297e-01 2.85193235e-01 -1.16948843e+00 8.80055428e-01 4.74947166e+00 1.01589525e+00 -1.08307004e+00 1.17820404e-01 3.74467820e-01 4.57815289e-01 -4.61260863e-02 -4.02727008e-01 -9.13849354e-01 6.50931835e-01 1.19123566e+00 -1.35064945e-01 1.39020339e-01 8.63294482e-01 1.90748880e-03 3.38762730e-01 -5.34939468e-01 1.01707304e+00 3.94453436e-01 -1.31610942e+00 2.38797590e-01 5.56731001e-02 5.14354408e-01 -5.73060364e-02 -1.15664322e-02 5.62606156e-01 1.18767977e-01 -1.00837517e+00 3.18679929e-01 -8.87271240e-02 7.09727764e-01 -1.20830095e+00 1.37271595e+00 3.59626263e-01 -9.73966897e-01 -2.73591340e-01 -8.87853086e-01 1.38815537e-01 -3.70012730e-01 1.06209569e-01 -9.09964263e-01 5.34979999e-01 7.37874448e-01 3.04506302e-01 -6.32712483e-01 5.38483679e-01 1.94856495e-01 7.26639986e-01 -2.15608701e-01 -5.76839924e-01 5.56723535e-01 -2.25213110e-01 2.79823989e-01 1.36290812e+00 3.36090624e-01 2.23940417e-01 -3.09274107e-01 1.68357730e-01 -1.08841904e-01 8.97186935e-01 -5.58595240e-01 -5.07853329e-01 4.37339574e-01 1.38852537e+00 -8.53735268e-01 -3.06096464e-01 -4.33041662e-01 1.02323401e+00 1.20960966e-01 1.69947781e-02 -5.66151321e-01 -1.19775665e+00 1.56675711e-01 -2.57945031e-01 1.90954059e-01 -3.72765034e-01 -4.50910330e-02 -9.99694169e-01 -1.14402287e-01 -9.80094016e-01 4.06007618e-01 -4.47609454e-01 -1.17556632e+00 1.01224744e+00 -1.98695585e-01 -1.09914613e+00 -4.15615290e-02 -9.85471547e-01 -3.18382233e-01 8.78394008e-01 -1.12851405e+00 -1.16462719e+00 -1.64159283e-01 5.25514126e-01 8.73100877e-01 -7.06008136e-01 1.15041399e+00 6.74638450e-01 -5.34801424e-01 7.57416129e-01 8.28769743e-01 7.14546800e-01 3.79953414e-01 -1.18694341e+00 4.89694476e-02 5.68983138e-01 1.72439918e-01 5.89408934e-01 2.49082208e-01 -4.66602921e-01 -1.61240828e+00 -9.04633224e-01 1.29805064e+00 -3.89722556e-01 8.57584000e-01 -5.14043093e-01 -9.55267787e-01 3.86018306e-01 3.66927177e-01 -1.68282047e-01 1.09029126e+00 -2.72287041e-01 -4.60150279e-02 -2.16171984e-02 -1.19862854e+00 5.47990739e-01 1.06470734e-01 -5.34763455e-01 -7.17695534e-01 3.82391036e-01 5.82160294e-01 9.71724987e-02 -1.06538963e+00 1.57394484e-02 5.87007344e-01 -4.00499374e-01 6.52059853e-01 -7.02076495e-01 3.97322088e-01 -1.54328465e-01 -4.27482277e-01 -1.15375447e+00 -1.59277380e-01 -1.01289697e-01 5.54804951e-02 1.55207992e+00 2.43983299e-01 -8.22285056e-01 5.46052694e-01 5.76486588e-01 1.72699858e-02 -6.10378921e-01 -7.72821307e-01 -5.27094901e-01 1.66020736e-01 -4.45555955e-01 4.34415072e-01 1.50582051e+00 3.23547572e-02 2.75619835e-01 -1.46618903e-01 8.13882202e-02 2.05417365e-01 -3.30462724e-01 3.85667920e-01 -1.14000762e+00 -3.95666100e-02 -1.81100368e-01 -1.00717640e+00 -4.45012033e-01 1.41346291e-01 -1.16225362e+00 -6.19856000e-01 -1.57738686e+00 -1.08272910e-01 -6.86768413e-01 -2.16963083e-01 5.80513775e-01 4.16986682e-02 3.84164244e-01 6.21417947e-02 2.42627218e-01 -2.89070178e-02 5.45222759e-01 9.47327852e-01 -3.54270101e-01 -1.74451917e-01 -2.28980601e-01 -5.26626825e-01 4.31029052e-01 8.66911650e-01 -3.83371472e-01 -3.34952980e-01 -5.88576078e-01 3.70929718e-01 -2.75277346e-01 -2.24310547e-01 -7.33695209e-01 1.08300775e-01 1.32962272e-01 6.04114592e-01 -8.28522623e-01 4.89900820e-02 -1.01341784e+00 -1.94078147e-01 2.52341509e-01 -2.69583434e-01 3.02793950e-01 2.58118242e-01 1.28866404e-01 -4.42977101e-01 -7.82927573e-01 5.73318541e-01 9.55617279e-02 -7.87749827e-01 9.58720371e-02 -6.07863545e-01 -2.66828269e-01 9.56648588e-01 -5.92391312e-01 -2.43760839e-01 -3.98730673e-02 -5.77730119e-01 -1.57597605e-02 9.14952718e-03 6.84416294e-01 7.04108953e-01 -1.27094793e+00 -8.63921762e-01 3.57993662e-01 1.97832927e-01 -1.98737532e-01 1.85644612e-01 3.33245993e-01 -1.08279335e+00 5.50433636e-01 -1.74809828e-01 -5.68776624e-03 -1.36145210e+00 2.62785345e-01 -1.57847628e-01 -2.23224938e-01 -4.57451671e-01 6.27218246e-01 -2.94363976e-01 -7.89576530e-01 7.72462338e-02 1.47929177e-01 -8.17503989e-01 4.58228439e-01 5.47688186e-01 6.89080417e-01 4.33591485e-01 -1.04134119e+00 -8.10306445e-02 2.82457620e-01 -3.74048680e-01 1.00950763e-01 1.50885510e+00 -1.51341468e-01 -5.01333535e-01 3.36486161e-01 1.59689963e+00 1.64752483e-01 -1.02217332e-01 8.24295580e-02 -6.76407712e-03 -4.34931606e-01 3.93462390e-01 -6.45394683e-01 -8.13815355e-01 9.48192060e-01 8.71852815e-01 5.45459628e-01 8.72807741e-01 -3.48851770e-01 8.18675637e-01 8.08290422e-01 4.09581652e-03 -1.60058296e+00 -2.12154493e-01 7.51567781e-01 6.30756080e-01 -1.27156305e+00 -4.74912971e-02 1.87954605e-01 -6.39012873e-01 1.51788461e+00 3.80224466e-01 2.04513133e-01 9.17468965e-01 1.07181907e-01 1.82988971e-01 -9.35395658e-02 -3.79668564e-01 7.38949701e-02 1.86263263e-01 6.13824308e-01 6.25818431e-01 2.17132941e-02 -7.49469101e-01 7.02023029e-01 -5.81609130e-01 -6.58626914e-01 5.79973519e-01 1.01145279e+00 -5.16152263e-01 -1.02403581e+00 -5.17204523e-01 5.52280664e-01 -8.02698851e-01 -2.07283601e-01 -1.71186641e-01 1.04693115e+00 1.87900022e-01 1.13811612e+00 1.61081731e-01 -3.53314131e-01 3.73966321e-02 1.66502789e-01 1.02530885e-02 -4.77493286e-01 -6.98644638e-01 -6.66746497e-02 -6.72660861e-03 4.75372195e-01 -1.71432540e-01 -1.41480729e-01 -1.27909613e+00 -4.95725870e-01 -8.45186830e-01 4.28925872e-01 1.30165839e+00 7.87673473e-01 -3.39231454e-02 2.00025499e-01 8.03877950e-01 -3.61462176e-01 -3.78425360e-01 -1.19372439e+00 -8.58319640e-01 2.74971545e-01 -2.33177602e-01 -3.33053142e-01 -4.31013048e-01 7.74786994e-02]

[10.238934516906738, 8.743565559387207]

57062724-7a8e-444e-9a46-358d878f07f1

ecsp-a-new-task-for-emotion-cause-span-pair

2003.03507

https://arxiv.org/abs/2003.03507v1

https://arxiv.org/pdf/2003.03507v1.pdf

ECSP: A New Task for Emotion-Cause Span-Pair Extraction and Classification

Emotion cause analysis such as emotion cause extraction (ECE) and emotion-cause pair extraction (ECPE) have gradually attracted the attention of many researchers. However, there are still two shortcomings in the existing research: 1) In most cases, emotion expression and cause are not the whole clause, but the span in the clause, so extracting the clause-pair rather than the span-pair greatly limits its applications in real-world scenarios; 2) It is not enough to extract the emotion expression clause without identifying the emotion categories, the presence of emotion clause does not necessarily convey emotional information explicitly due to different possible causes. In this paper, we propose a new task: Emotion-Cause Span-Pair extraction and classification (ECSP), which aims to extract the potential span-pair of emotion and corresponding causes in a document, and make emotion classification for each pair. In the new ECSP task, ECE and ECPE can be regarded as two special cases at the clause-level. We propose a span-based extract-then-classify (ETC) model, where emotion and cause are directly extracted and paired from the document under the supervision of target span boundaries, and corresponding categories are then classified using their pair representations and localized context. Experiments show that our proposed ETC model outperforms the SOTA model of ECE and ECPE task respectively and gets a fair-enough results on ECSP task.

['Hongliang Bi', 'Pengyuan Liu']

2020-03-07

['emotion-cause-pair-extraction', 'emotion-cause-extraction']

['natural-language-processing', 'natural-language-processing']

[ 3.41993392e-01 -9.35725793e-02 -1.95022285e-01 -5.43438673e-01 -5.82080007e-01 -5.72976708e-01 4.49927241e-01 2.44810939e-01 -4.64819185e-02 7.45758414e-01 3.49780619e-01 1.69085041e-02 -1.89097911e-01 -5.60958982e-01 -1.73621058e-01 -7.02702463e-01 4.94664274e-02 8.96001533e-02 -8.90850425e-02 -3.18782866e-01 1.96382344e-01 1.19158395e-01 -1.79527080e+00 6.56736851e-01 1.12473273e+00 1.42558420e+00 -1.44286737e-01 3.06029260e-01 -6.93385363e-01 1.14341450e+00 -7.27150857e-01 -2.95084655e-01 -2.46928126e-01 -7.26183236e-01 -9.05482113e-01 4.36147824e-02 -2.74763465e-01 1.39700651e-01 5.13974845e-01 9.70740676e-01 3.93987477e-01 1.73271000e-02 8.72219861e-01 -1.82733560e+00 -3.24422896e-01 6.12926662e-01 -8.45376730e-01 -2.99260928e-03 6.11069620e-01 -5.87462962e-01 1.19792616e+00 -7.82859683e-01 4.91557360e-01 1.33182657e+00 3.89488906e-01 4.25764114e-01 -5.72212994e-01 -8.22065175e-01 6.10721231e-01 5.32532752e-01 -1.30275130e+00 -2.16955245e-02 1.13493431e+00 -1.22236475e-01 1.02385974e+00 4.54786211e-01 7.78570473e-01 9.22477782e-01 1.90811515e-01 1.13310802e+00 1.43474722e+00 -3.55107665e-01 2.47750387e-01 1.46749467e-01 4.29930091e-01 3.39428306e-01 -4.46708202e-01 -3.44432384e-01 -3.00824910e-01 -1.74192533e-01 2.00933814e-01 -6.91553652e-02 -5.66762686e-01 2.12959215e-01 -9.74988222e-01 8.04994822e-01 3.86543572e-01 5.55517972e-01 -5.01645386e-01 -2.19330654e-01 7.64779389e-01 3.93982232e-01 5.75582325e-01 2.53222585e-01 -7.81605840e-01 -3.88041347e-01 -6.28631055e-01 4.45013523e-01 8.81294608e-01 9.95023787e-01 6.23169482e-01 -3.72396260e-01 7.89158940e-02 8.74328792e-01 -9.06423032e-02 2.35212401e-01 5.48400104e-01 -2.33533502e-01 5.08040369e-01 9.05218363e-01 -4.84240800e-02 -1.51595390e+00 -4.47224945e-01 -3.45491976e-01 -7.43237257e-01 -3.93501282e-01 -2.06338242e-01 -4.60344881e-01 -5.07104218e-01 1.89612222e+00 5.72946727e-01 2.94166386e-01 3.33088309e-01 9.76284623e-01 1.14751565e+00 1.07524657e+00 2.27402061e-01 -7.72210062e-01 1.62463295e+00 -8.35616767e-01 -1.37220347e+00 -3.74150991e-01 7.12493658e-01 -7.99289346e-01 9.21825945e-01 6.27212167e-01 -5.89185834e-01 -2.13870928e-01 -9.98849690e-01 -5.85811399e-02 -5.84240198e-01 3.36172730e-01 8.14916790e-01 1.09978110e-01 -3.19571078e-01 1.00770280e-01 -1.40437692e-01 -1.89481750e-01 6.67762533e-02 2.29681388e-01 -3.74648839e-01 -4.60908525e-02 -1.58256853e+00 8.02064836e-01 6.25528157e-01 2.52986014e-01 -6.02629893e-02 -5.62212050e-01 -9.76881385e-01 1.82619795e-01 6.82728827e-01 -1.76252037e-01 1.01727664e+00 -1.49990320e+00 -9.25689816e-01 6.49331272e-01 -4.17408824e-01 1.58536017e-01 -8.14821497e-02 -3.37282866e-01 -9.80658531e-01 -1.65012863e-03 6.28333241e-02 5.00277281e-01 4.89573449e-01 -1.38814223e+00 -9.60898042e-01 -3.68360728e-01 -2.75767416e-01 4.73643482e-01 -3.80377680e-01 4.69676107e-01 -3.92044187e-01 -5.67127049e-01 2.12829664e-01 -5.84931850e-01 1.78118035e-01 -4.21614677e-01 -6.34780765e-01 -7.90427387e-01 1.37546504e+00 -7.54063249e-01 1.59363830e+00 -2.30938649e+00 8.17004740e-02 5.80926910e-02 1.33352652e-01 -1.50966272e-01 -3.02852225e-02 3.63245279e-01 -6.12939060e-01 2.41351873e-01 -2.64851123e-01 2.60459315e-02 9.95365679e-02 3.56257945e-01 -6.18779719e-01 -8.03643540e-02 6.26482427e-01 4.90273237e-01 -8.78136754e-01 -9.36720371e-01 -1.82171717e-01 2.05669105e-01 -3.15812707e-01 4.25090939e-01 -1.46522880e-01 -4.39091139e-02 -6.82547629e-01 7.24148571e-01 8.02803218e-01 2.44766161e-01 9.88341942e-02 -4.37016249e-01 2.37318408e-02 1.91961750e-01 -1.37264836e+00 9.77145016e-01 -4.71731246e-01 2.90764630e-01 8.70049521e-02 -9.69227731e-01 1.17337799e+00 5.38896799e-01 4.89216447e-01 -5.58933496e-01 4.48536158e-01 2.60239363e-01 8.21388513e-02 -8.39859724e-01 3.62528652e-01 -5.78197360e-01 -6.13594413e-01 1.86788499e-01 -1.36237323e-01 -1.23399608e-01 3.02382767e-01 3.27899083e-02 9.78884220e-01 -1.05938271e-01 6.07995212e-01 -1.20107241e-01 7.60127187e-01 5.84563799e-02 1.14752936e+00 -1.27933294e-01 -3.86585116e-01 3.24744523e-01 1.00484157e+00 -2.36410558e-01 -3.92316252e-01 -4.97984767e-01 -4.20042053e-02 8.32984388e-01 4.41467196e-01 -6.67801917e-01 -5.47096729e-01 -1.24043024e+00 -3.29885095e-01 7.40582585e-01 -6.28862798e-01 -2.19694778e-01 -5.26038945e-01 -7.80542731e-01 3.01104546e-01 5.73284030e-01 7.09845901e-01 -1.24068189e+00 -3.69346589e-01 2.96766490e-01 -7.43184686e-01 -1.11497724e+00 -2.34618708e-01 5.58127403e-01 -2.86035329e-01 -1.18140852e+00 -2.51023203e-01 -1.04088056e+00 5.34504056e-01 -1.10871978e-01 1.04360497e+00 3.87007743e-02 -1.63598001e-01 -1.23300023e-01 -7.63940573e-01 -7.45893478e-01 1.17044695e-01 -8.83434564e-02 -2.03568593e-01 2.52205074e-01 7.61373222e-01 -5.13099134e-01 -2.28916958e-01 1.84117883e-01 -7.24262655e-01 1.58008009e-01 5.81624866e-01 7.33578265e-01 6.40918851e-01 5.90887308e-01 8.63500059e-01 -8.68181288e-01 9.60280120e-01 -7.40948081e-01 1.13473870e-01 2.27787063e-01 -5.22035122e-01 -3.06948453e-01 8.09755325e-01 -4.26436394e-01 -1.28746033e+00 -8.51026326e-02 -2.22517520e-01 -3.66253704e-01 -3.76312494e-01 8.91297996e-01 -6.73926830e-01 6.16324544e-01 -2.11242475e-02 1.97928205e-01 -3.93869698e-01 -2.38782793e-01 1.69409081e-01 8.98247659e-01 3.93840194e-01 -5.69231868e-01 3.36028993e-01 1.54116133e-04 -1.41812712e-01 -3.02079707e-01 -9.72503126e-01 -4.47722971e-01 -3.84798914e-01 -2.71618217e-01 9.71837878e-01 -8.17574561e-01 -5.95461249e-01 1.66148096e-01 -1.45027709e+00 3.22417289e-01 1.08676367e-01 3.09772491e-01 -1.29373416e-01 1.84144840e-01 -5.94710886e-01 -9.62640226e-01 -4.24818665e-01 -9.31392372e-01 1.19929123e+00 4.42321539e-01 -5.07284760e-01 -7.38888383e-01 -2.33579069e-01 -8.55348036e-02 -7.07759485e-02 7.78723478e-01 1.33956146e+00 -8.52292538e-01 2.10106790e-01 -2.90546268e-01 -2.90160865e-01 1.47620514e-01 3.08269024e-01 2.27414459e-01 -7.47984111e-01 2.75312066e-01 3.07793528e-01 -5.89070976e-01 4.96930718e-01 3.69880125e-02 1.11696756e+00 -6.42507493e-01 -4.08482790e-01 2.39764482e-01 1.58682454e+00 6.03465736e-01 5.14447808e-01 5.94805442e-02 5.16499996e-01 1.07281327e+00 1.15936518e+00 4.92858678e-01 3.37244302e-01 5.17236173e-01 4.27264899e-01 -3.18664998e-01 2.84460992e-01 -4.84945588e-02 3.97933334e-01 1.07671607e+00 1.67434998e-02 -3.34316432e-01 -6.69492126e-01 4.93668407e-01 -1.86128736e+00 -9.88677144e-01 -5.40702283e-01 1.42958450e+00 1.12475562e+00 -7.07165375e-02 -2.30279956e-02 6.19338036e-01 8.66687596e-01 1.53834090e-01 -2.43494034e-01 -8.24400783e-01 -1.76984489e-01 -9.71165895e-02 -4.33353275e-01 1.62446663e-01 -1.15176904e+00 6.30718112e-01 4.92221832e+00 1.21248055e+00 -1.17647660e+00 -9.83215030e-03 7.57890165e-01 1.01941630e-01 -2.21681595e-01 6.72604889e-02 -5.85100174e-01 6.20062113e-01 2.59097487e-01 -6.48859143e-02 1.47331312e-01 1.07211220e+00 4.95747700e-02 -3.19198109e-02 -1.15373659e+00 1.18498516e+00 2.64040202e-01 -5.50370038e-01 -5.51419742e-02 -1.66592464e-01 5.11434197e-01 -7.94988990e-01 -3.02829146e-01 7.11685181e-01 -3.34689021e-01 -8.81021023e-01 6.54268622e-01 2.66282409e-01 6.29736364e-01 -1.25560915e+00 1.23781133e+00 2.91283011e-01 -1.55735707e+00 -1.47779495e-01 -2.37113595e-01 -2.36882403e-01 1.07357606e-01 9.31895196e-01 -3.98674369e-01 8.45182061e-01 7.24548697e-01 8.34058523e-01 -3.64807725e-01 6.40036643e-01 -4.91569936e-01 6.84892476e-01 -2.49430999e-01 -4.56965148e-01 2.30058327e-01 -1.08413763e-01 4.67565268e-01 1.44140208e+00 3.10190052e-01 5.71172357e-01 2.33211294e-01 9.44600999e-01 6.60655648e-02 5.61075389e-01 -2.01971993e-01 1.36224464e-01 7.24142134e-01 1.66804600e+00 -9.00359571e-01 -3.97785187e-01 -8.48736092e-02 9.55173254e-01 2.46154293e-01 1.52096182e-01 -1.07834125e+00 -1.05898201e+00 3.32688093e-01 -5.53792834e-01 2.00848743e-01 5.30001104e-01 -2.41536245e-01 -1.06860101e+00 4.41786468e-01 -9.72979784e-01 6.04874253e-01 -9.14878190e-01 -1.61068571e+00 8.36152315e-01 -1.62895825e-02 -1.33223855e+00 -2.66035140e-01 -5.07469594e-01 -1.13735247e+00 6.21831119e-01 -1.51768517e+00 -1.16437006e+00 -4.59778696e-01 6.04756415e-01 4.49249238e-01 3.32767695e-01 8.25748861e-01 2.24960223e-01 -8.24327767e-01 4.21933889e-01 -4.36486185e-01 2.02324271e-01 6.29356623e-01 -1.35769844e+00 -7.64846444e-01 6.37715578e-01 -2.39615723e-01 5.37707031e-01 8.44921410e-01 -7.03355908e-01 -1.05527544e+00 -1.04249299e+00 1.38422489e+00 -2.51380298e-02 3.88041794e-01 -4.88917589e-01 -7.56188810e-01 4.59700346e-01 4.86776233e-01 -3.92547362e-02 6.90882027e-01 4.63778496e-01 -2.21184477e-01 -2.17947856e-01 -1.18343496e+00 5.29838622e-01 7.00475454e-01 -2.83668727e-01 -9.26234484e-01 2.07939044e-01 7.44630516e-01 -1.86145112e-01 -7.01011121e-01 5.23514748e-01 2.05780566e-01 -9.43005323e-01 4.13052887e-01 -4.25472647e-01 1.06755817e+00 -3.90952080e-01 -8.96495208e-02 -1.26933110e+00 3.53962407e-02 -2.85643935e-01 5.97520843e-02 2.19739795e+00 3.19115400e-01 -3.05322438e-01 2.61476338e-01 5.53138316e-01 -1.69867679e-01 -1.27844989e+00 -9.34530854e-01 -5.86090922e-01 -1.20180137e-01 -6.12004161e-01 1.04129720e+00 1.32720006e+00 5.66414893e-01 8.14022183e-01 -1.75941408e-01 -5.65728769e-02 -1.06767155e-02 7.10980177e-01 2.42950231e-01 -1.01063085e+00 -1.92958519e-01 -4.71873790e-01 -1.26912475e-01 -6.07409477e-01 2.78073370e-01 -7.29863048e-01 3.55942667e-01 -1.49427998e+00 3.26314658e-01 -4.10043091e-01 -2.99788237e-01 7.80522525e-01 -6.56815171e-01 -2.13837877e-01 1.59307495e-01 5.08281738e-02 -5.75018048e-01 7.91191339e-01 1.02267969e+00 -2.00807862e-02 -1.23605557e-01 -3.00083190e-01 -7.81535983e-01 8.77001524e-01 5.14338195e-01 -4.60815430e-01 -5.31326473e-01 1.10741302e-01 3.99929464e-01 1.24044694e-01 1.21845424e-01 -7.14068413e-01 -7.68651813e-02 -3.55861366e-01 3.88103247e-01 -7.67875671e-01 9.55285802e-02 -1.07423306e+00 -3.12609494e-01 1.34872228e-01 -2.23413110e-01 2.24115163e-01 7.84854218e-02 3.20796102e-01 -7.73340106e-01 -1.92779675e-01 3.18584204e-01 1.31851971e-01 -9.19387996e-01 -6.33675605e-02 -3.58531922e-01 6.20725304e-02 1.16270638e+00 1.42296227e-02 -2.94316739e-01 -3.29288870e-01 -4.59198117e-01 4.08901244e-01 -1.10117234e-01 3.99520457e-01 6.25228345e-01 -1.66604173e+00 -5.07613957e-01 -2.53836304e-01 3.05016935e-01 1.57396078e-01 3.77416104e-01 8.69483292e-01 1.30803436e-01 2.91768741e-02 2.83997841e-02 -1.34825841e-01 -1.54992175e+00 8.03784370e-01 1.23750225e-01 -6.09378040e-01 -2.06632003e-01 8.08947086e-01 3.43916923e-01 -3.66724074e-01 7.27874264e-02 -4.35701311e-01 -8.21215868e-01 3.92063469e-01 4.28636909e-01 -5.88485785e-02 -1.14476770e-01 -5.85701287e-01 -5.59433043e-01 5.29343009e-01 -6.39302805e-02 6.47473112e-02 1.17814696e+00 -5.85610904e-02 -6.03567898e-01 4.98077184e-01 1.40935028e+00 1.66258320e-01 -5.28097391e-01 7.31470659e-02 4.88919616e-02 -6.69659972e-02 6.43120259e-02 -1.08990991e+00 -1.14581239e+00 9.11431074e-01 2.20686868e-01 2.84783721e-01 1.77677560e+00 4.93657626e-02 1.05260956e+00 -6.81284741e-02 8.82770717e-02 -1.25852334e+00 -5.02751209e-02 5.24167776e-01 1.23056090e+00 -8.86274815e-01 -1.36897326e-01 -1.01358712e+00 -8.23765218e-01 1.16442895e+00 1.03022540e+00 2.24929396e-02 5.61862826e-01 4.66143131e-01 -2.64765322e-02 -4.05924231e-01 -9.83614564e-01 -1.74737886e-01 3.34236890e-01 3.65952253e-01 6.33908093e-01 -1.83562431e-02 -8.16655576e-01 1.69563210e+00 -3.92394781e-01 -1.56579852e-01 1.52827799e-01 1.15259862e+00 -2.91928232e-01 -9.63515520e-01 -4.07003611e-01 4.38169450e-01 -5.24788737e-01 -8.65697265e-02 -8.75212193e-01 8.37142110e-01 8.84788811e-01 1.24768066e+00 2.36078296e-02 -6.78975999e-01 3.29545289e-01 3.04514945e-01 3.41489874e-02 -4.52382267e-01 -6.32532418e-01 3.10732394e-01 4.63232785e-01 -4.77280051e-01 -5.14937401e-01 -5.82277775e-01 -1.75593972e+00 -4.33525220e-02 -6.31601632e-01 5.80408216e-01 3.17603916e-01 1.37032199e+00 3.01159710e-01 7.50446975e-01 8.91162992e-01 -3.09805602e-01 -9.96137559e-02 -1.01790941e+00 -7.96335280e-01 7.99246013e-01 4.94581275e-02 -6.79180980e-01 -4.09629911e-01 8.23676661e-02]

[12.629487991333008, 6.211156845092773]

6955afdc-25a9-4865-9021-32a3947bc1d9

battery-and-hydrogen-energy-storage-control

2208.12779

https://arxiv.org/abs/2208.12779v1

https://arxiv.org/pdf/2208.12779v1.pdf

Battery and Hydrogen Energy Storage Control in a Smart Energy Network with Flexible Energy Demand using Deep Reinforcement Learning

Smart energy networks provide for an effective means to accommodate high penetrations of variable renewable energy sources like solar and wind, which are key for deep decarbonisation of energy production. However, given the variability of the renewables as well as the energy demand, it is imperative to develop effective control and energy storage schemes to manage the variable energy generation and achieve desired system economics and environmental goals. In this paper, we introduce a hybrid energy storage system composed of battery and hydrogen energy storage to handle the uncertainties related to electricity prices, renewable energy production and consumption. We aim to improve renewable energy utilisation and minimise energy costs and carbon emissions while ensuring energy reliability and stability within the network. To achieve this, we propose a multi-agent deep deterministic policy gradient approach, which is a deep reinforcement learning-based control strategy to optimise the scheduling of the hybrid energy storage system and energy demand in real-time. The proposed approach is model-free and does not require explicit knowledge and rigorous mathematical models of the smart energy network environment. Simulation results based on real-world data show that: (i) integration and optimised operation of the hybrid energy storage system and energy demand reduces carbon emissions by 78.69%, improves cost savings by 23.5% and renewable energy utilisation by over 13.2% compared to other baseline models and (ii) the proposed algorithm outperforms the state-of-the-art self-learning algorithms like deep-Q network.

['Jun Cao', 'Zhong Fan', 'Cephas Samende']

2022-08-26

['self-learning']

['natural-language-processing']

[-4.30022180e-01 4.00065444e-02 -2.18355179e-01 3.08094501e-01 -7.18588382e-02 -5.50747216e-01 6.53207242e-01 7.75884986e-02 -1.19601026e-01 1.58684337e+00 -3.39393280e-02 -2.92131424e-01 -5.31767249e-01 -1.17803121e+00 -5.14441729e-01 -1.08030748e+00 -1.84949115e-01 5.46241283e-01 -3.14143270e-01 -3.66123587e-01 1.31312400e-01 4.80224937e-01 -1.60848904e+00 -8.16952586e-01 1.22628486e+00 1.34697270e+00 5.21902084e-01 6.29147351e-01 2.65976638e-01 1.90053388e-01 -4.91256237e-01 5.45774817e-01 4.02575135e-01 -3.66044074e-01 -2.55629152e-01 -5.20174980e-01 -8.26678276e-01 -2.94309676e-01 -8.24290607e-03 1.14136481e+00 9.76448417e-01 6.08646274e-01 6.13684714e-01 -1.22952032e+00 -5.78116179e-01 3.54735076e-01 -2.88540393e-01 3.48917514e-01 -1.96561590e-01 3.78971100e-01 4.71520394e-01 -2.16155916e-01 -6.71649948e-02 8.36131752e-01 1.08767457e-01 5.32079339e-01 -8.20978761e-01 -5.25285900e-01 -2.03891262e-01 5.02858698e-01 -1.06855929e+00 -2.22317994e-01 5.75355470e-01 -1.59063824e-02 1.65697384e+00 3.05059701e-01 1.36251664e+00 4.59522039e-01 4.94409084e-01 2.73926198e-01 1.20866537e+00 -4.37943965e-01 6.97334528e-01 -1.65264666e-01 -6.58692539e-01 1.50658548e-01 4.74839211e-01 7.51347363e-01 1.15855165e-01 3.75936478e-01 2.30152041e-01 -2.26787135e-01 7.81960320e-03 -3.01740050e-01 -6.25684917e-01 7.28662789e-01 7.26092279e-01 4.23070282e-01 -8.14649343e-01 2.51736462e-01 2.33965844e-01 9.02767759e-03 1.65789291e-01 4.89456922e-01 -5.32115638e-01 -2.44528383e-01 -9.29152369e-01 -1.03458814e-01 7.81040251e-01 5.40252805e-01 3.40697974e-01 1.06994605e+00 1.43311143e-01 4.21402097e-01 5.77614486e-01 1.05126905e+00 9.85124826e-01 -1.08265603e+00 -6.12335131e-02 4.27797496e-01 5.51400423e-01 -4.51726526e-01 -5.84459782e-01 -4.34389681e-01 -1.28078306e+00 6.05150282e-01 -3.50825727e-01 -7.12369323e-01 -8.85909677e-01 1.42648971e+00 3.58047456e-01 -6.79722577e-02 2.55175889e-01 6.73916578e-01 -3.55121642e-02 1.13566017e+00 1.81174390e-02 -7.97384441e-01 9.90913212e-01 -7.46851265e-01 -9.71067011e-01 1.40206307e-01 1.05839320e-01 -1.31291643e-01 1.78410292e-01 -9.83619224e-03 -1.64196253e+00 -4.21854615e-01 -1.33299220e+00 6.11379266e-01 -1.10799718e+00 -2.15341210e-01 -4.89971787e-03 5.67540407e-01 -1.23881102e+00 6.14772439e-01 -9.08457518e-01 -1.66510463e-01 1.38784558e-01 5.57618797e-01 4.70207602e-01 5.54229975e-01 -1.59584081e+00 1.72208929e+00 1.17896545e+00 4.49916571e-01 -8.70184183e-01 -6.70065224e-01 -7.06747413e-01 4.14014518e-01 4.53600585e-01 -8.29116642e-01 1.17913246e+00 -6.04890406e-01 -2.10975909e+00 -3.76562685e-01 1.85494781e-01 -9.16205883e-01 3.91427070e-01 1.32099077e-01 -5.48282385e-01 -4.45656618e-03 -5.42160511e-01 1.33768111e-01 4.83914733e-01 -9.78598893e-01 -9.05141532e-01 -1.44516096e-01 -5.26780009e-01 5.41653991e-01 -4.20593679e-01 -5.14005244e-01 7.82518029e-01 -1.33317500e-01 -9.79416251e-01 -8.66489887e-01 -3.06400746e-01 -6.40505731e-01 -5.39471619e-02 -6.52968287e-01 1.19782138e+00 -7.58810818e-01 9.43998039e-01 -1.40488160e+00 4.32007849e-01 4.38131064e-01 -4.05743867e-01 7.71908820e-01 3.34022731e-01 6.94647729e-01 -1.03408135e-01 2.43155047e-01 -1.25037134e-01 -7.22315395e-03 3.88401836e-01 7.45734692e-01 1.11562081e-01 8.61366838e-02 -7.23640770e-02 1.09538972e+00 -1.12346864e+00 1.91587768e-02 9.50436950e-01 4.80999619e-01 3.54447842e-01 2.57258266e-01 -6.69855297e-01 2.71033853e-01 -3.33567619e-01 4.06855136e-01 4.20013040e-01 3.14360484e-02 2.90142596e-01 -6.07530661e-02 -5.67968011e-01 -2.24546820e-01 -1.32487142e+00 1.21301866e+00 -9.16667521e-01 2.20939815e-01 4.69583362e-01 -1.66477466e+00 7.31501758e-01 2.55607277e-01 7.58571267e-01 -1.51461923e+00 4.52545248e-02 6.30421758e-01 1.03458436e-02 -3.20778340e-01 4.01860029e-01 -2.53335655e-01 4.56894606e-01 2.56309927e-01 -1.14308782e-01 -2.59663522e-01 4.28945869e-01 -2.91281104e-01 6.85026109e-01 -5.23920469e-02 5.00448585e-01 -6.36830628e-01 7.83754230e-01 -6.02836430e-01 5.44735730e-01 1.02060348e-01 -1.24707334e-01 -5.65100074e-01 -2.20628932e-01 -1.22542247e-01 -1.25486541e+00 -9.40052986e-01 8.95979479e-02 5.69827795e-01 1.25376731e-01 4.63448584e-01 -2.60511786e-01 -1.17748380e-01 2.92573810e-01 1.35511613e+00 -3.42710286e-01 -2.56778896e-01 -6.28043175e-01 -1.04353237e+00 -4.87620413e-01 3.83070946e-01 7.75266409e-01 -9.59168315e-01 -1.10762036e+00 4.63964194e-01 3.12712580e-01 -6.64147615e-01 1.88006030e-03 5.52415490e-01 -4.17410403e-01 -1.04486406e+00 -6.68495655e-01 -2.84429163e-01 4.50409353e-01 -3.34566087e-01 1.10830212e+00 6.03681989e-02 -1.35780588e-01 3.86402875e-01 -7.31702074e-02 -6.04068637e-01 -4.71718282e-01 1.81568742e-01 3.91214758e-01 -4.68334019e-01 -2.74956077e-01 -6.35079384e-01 -1.11575794e+00 -9.17846989e-03 -7.43872702e-01 -4.61802864e-03 6.33904815e-01 8.80509317e-01 6.30898714e-01 1.00071502e+00 1.15006626e+00 -8.55323300e-02 8.47510278e-01 -8.68925095e-01 -1.14812839e+00 4.47983772e-01 -1.61372399e+00 1.38623580e-01 1.06811869e+00 1.78651810e-01 -1.04490972e+00 -2.65704636e-02 2.15205565e-01 -1.96782589e-01 1.51896104e-01 3.74223858e-01 -8.26720893e-02 -1.04980534e-02 -3.24799836e-01 6.19982839e-01 7.86497295e-02 -2.45590940e-01 3.26051861e-01 4.54443038e-01 4.87403005e-01 -3.59713852e-01 1.03812182e+00 -1.12981290e-01 6.85737610e-01 -5.47723591e-01 -1.22148089e-01 -7.05325529e-02 -2.36627847e-01 -4.25500751e-01 6.73753858e-01 -7.70213544e-01 -1.12157214e+00 4.43399221e-01 -6.42135024e-01 -5.67893207e-01 -5.17758489e-01 1.75550714e-01 -6.44508362e-01 1.37043774e-01 1.04079477e-01 -1.35087705e+00 -1.00523663e+00 -8.10546577e-01 2.89445937e-01 9.02333915e-01 3.60081881e-01 -1.33124971e+00 2.95356512e-01 -2.52830774e-01 1.13092959e+00 7.36943781e-01 8.30010712e-01 -9.20947120e-02 -4.47205067e-01 1.78187519e-01 9.94214490e-02 7.20503330e-01 3.06110501e-01 -1.44096408e-02 -3.05881828e-01 -9.02249992e-01 -2.70057082e-01 -2.62171060e-01 2.36394882e-01 4.32741821e-01 1.08634245e+00 -7.01570988e-01 -2.20659256e-01 2.61552840e-01 2.20168018e+00 7.78431892e-01 4.93625641e-01 4.73988444e-01 1.64827675e-01 1.00625657e-01 2.80832976e-01 7.17959285e-01 6.09614134e-01 4.62250173e-01 1.18492162e+00 1.94665790e-01 2.00599432e-01 2.44388599e-02 2.69565374e-01 8.77906859e-01 -1.11953646e-01 -4.25830215e-01 -6.23203993e-01 8.22530031e-01 -2.02081943e+00 -1.04151261e+00 1.99236676e-01 2.20443368e+00 7.83832073e-01 -2.05679443e-02 2.21477240e-01 1.93363339e-01 4.96173441e-01 1.09949037e-01 -1.06759846e+00 -8.79503310e-01 -2.27046669e-01 2.56995350e-01 9.60223854e-01 3.73400986e-01 -3.65919173e-01 2.76957482e-01 5.12157297e+00 6.12985611e-01 -1.09072495e+00 5.55976257e-02 4.77698743e-01 -3.98269385e-01 -7.77559206e-02 -3.89135629e-01 -4.10239130e-01 1.35794437e+00 1.84203577e+00 -1.02835107e+00 1.20076227e+00 3.84348214e-01 9.93492305e-01 -5.15790105e-01 -6.34434223e-01 3.82969767e-01 -5.11246085e-01 -1.40101445e+00 -5.27350962e-01 2.95226723e-01 1.20688367e+00 4.60547388e-01 -3.62004876e-01 3.38814855e-01 7.36775875e-01 -8.39509189e-01 5.57507515e-01 1.04874039e+00 2.47457236e-01 -1.29545236e+00 9.98610854e-01 5.23674428e-01 -1.38004303e+00 -7.13280976e-01 -1.40930668e-01 -1.76580951e-01 4.64097381e-01 6.08314991e-01 -3.68741125e-01 9.06470001e-01 9.07329857e-01 3.83638412e-01 -1.23257808e-01 9.93885696e-01 -1.69574022e-01 3.26119393e-01 -7.88887143e-01 -5.80272973e-01 2.55932510e-01 -2.99756497e-01 1.63836554e-01 6.70596242e-01 5.11383593e-01 1.62289038e-01 1.24249987e-01 6.47958457e-01 -1.55195162e-01 -4.00935531e-01 -5.02160490e-01 -1.03358783e-01 9.11169529e-01 1.45784235e+00 -4.05816972e-01 -3.78893763e-01 -5.28262109e-02 3.85045111e-01 -2.55993277e-01 3.27530742e-01 -6.06799483e-01 -3.93814355e-01 5.77874541e-01 -3.50192904e-01 4.78678167e-01 -3.56726825e-01 -1.98447898e-01 -5.65725148e-01 -8.52055997e-02 -1.98957935e-01 1.09030791e-01 -6.82839990e-01 -1.24298894e+00 -1.14845410e-01 9.96271055e-03 -7.27300227e-01 -1.02476990e+00 -4.54456508e-01 -9.76250947e-01 1.11185622e+00 -2.44146109e+00 -6.96500361e-01 -1.25456735e-01 4.49843049e-01 4.96916264e-01 -3.56946588e-01 8.48796070e-01 1.75263837e-01 -9.87852335e-01 -1.27000660e-01 1.27382934e+00 -4.47408468e-01 -2.10079387e-01 -1.82502627e+00 -1.75459191e-01 9.53294277e-01 -6.76679254e-01 -1.23862803e-01 7.36253142e-01 -5.11925757e-01 -1.90344763e+00 -1.07152104e+00 1.67405516e-01 4.79471117e-01 8.88778746e-01 1.89855129e-01 -6.48390472e-01 -1.70359593e-02 1.25971901e+00 -6.26221532e-03 3.37228894e-01 -7.46765256e-01 5.63354135e-01 -4.61227357e-01 -1.36365783e+00 1.70792654e-01 3.82678866e-01 -2.84081455e-02 -1.09093711e-01 3.37509304e-01 2.03928009e-01 -2.07637757e-01 -1.20759988e+00 5.49206614e-01 4.91938651e-01 -4.19213623e-01 6.60580814e-01 -2.94786334e-01 -8.61330852e-02 -2.31063187e-01 -6.01386763e-02 -2.18132067e+00 -1.82115272e-01 -8.22262943e-01 -1.30279934e+00 1.31162429e+00 1.64496019e-01 -9.25782919e-01 4.36479032e-01 5.55515349e-01 -1.43996939e-01 -9.57848072e-01 -1.62069654e+00 -1.00508630e+00 2.98296928e-01 3.94338667e-01 7.98205376e-01 7.98893452e-01 -3.47292796e-02 1.07446369e-02 -8.49754885e-02 3.49070758e-01 9.19758081e-01 2.16352288e-02 7.23452345e-02 -8.82079959e-01 2.27024347e-01 -7.55328953e-01 9.53132510e-02 1.14412094e-02 3.23517889e-01 -4.45200354e-01 -4.14115489e-02 -2.28889585e+00 -3.53791058e-01 -9.05571803e-02 -8.37093472e-01 5.03161132e-01 1.29083604e-01 -2.05087900e-01 3.22237879e-01 -1.67769343e-01 -3.29348803e-01 1.38747239e+00 8.60063076e-01 -1.80579767e-01 5.38693257e-02 1.21133365e-01 -3.29776257e-01 3.33450586e-01 1.33351648e+00 1.62830621e-01 -6.58130050e-01 -2.46863827e-01 3.08768690e-01 2.18211487e-01 1.82767183e-01 -1.10165977e+00 3.58595610e-01 -5.30594349e-01 4.62774545e-01 -5.91600776e-01 -8.89286846e-02 -1.34862185e+00 5.64908564e-01 1.13185143e+00 2.08724886e-01 3.59550327e-01 1.25347987e-01 7.02935278e-01 1.53118536e-01 -3.46334800e-02 9.97488856e-01 -2.10594252e-01 -1.00485325e+00 3.57271321e-02 -3.99594277e-01 -3.30865711e-01 1.60192275e+00 -7.34309480e-02 -7.43007362e-01 -2.90251702e-01 -5.03559947e-01 1.27199233e+00 1.92583442e-01 5.87563217e-01 2.98158467e-01 -1.19432902e+00 -4.56111848e-01 -2.01918080e-01 -8.50215435e-01 -8.49065259e-02 7.55041987e-02 3.11956972e-01 -2.14268327e-01 5.08450687e-01 -4.98833925e-01 -8.48990381e-02 -4.69512194e-01 4.75490808e-01 1.03063452e+00 -5.02332211e-01 -1.28659084e-01 1.05615355e-01 -1.01828206e+00 -4.18593436e-02 -6.27933592e-02 -1.66709810e-01 -6.56695589e-02 2.62312353e-01 1.61266416e-01 9.74579155e-01 2.24383846e-01 -2.48816997e-01 -2.81077027e-01 3.30809563e-01 7.02149093e-01 2.99071014e-01 1.66995335e+00 -3.87104630e-01 -1.47509173e-01 3.12273890e-01 9.76099610e-01 -7.16378987e-01 -1.28341544e+00 3.35531533e-01 7.61630163e-02 2.90999822e-02 6.95020199e-01 -1.63075769e+00 -1.33364666e+00 4.16383475e-01 1.00555253e+00 1.05329335e+00 1.46597791e+00 -8.05317760e-01 8.57007742e-01 2.20556691e-01 2.96157926e-01 -1.81995845e+00 -3.29128414e-01 3.46128315e-01 6.87997639e-01 -1.01392603e+00 1.93875656e-01 6.95479155e-01 -1.23447282e-02 1.05285954e+00 6.48274541e-01 -1.10208608e-01 7.20943570e-01 1.46138608e-01 -6.13892555e-01 2.80241847e-01 -8.06780934e-01 -2.73443788e-01 -1.41827285e-01 5.79012334e-01 -1.11680739e-01 3.77950579e-01 -4.74437118e-01 -6.69562072e-02 4.71007600e-02 1.41280860e-01 3.01298261e-01 9.16183949e-01 -5.94735682e-01 -9.47452724e-01 -2.10726321e-01 5.24896801e-01 -2.13485360e-01 3.77673358e-01 2.40260512e-01 6.36557519e-01 3.03154647e-01 8.40568960e-01 6.70838654e-02 2.51636416e-01 4.41348463e-01 5.18263727e-02 1.21686876e-01 -4.03698310e-02 -3.04386556e-01 -1.48809657e-01 -4.20173928e-02 -1.96816698e-01 -5.07309675e-01 -6.25707746e-01 -1.54216194e+00 -4.31337088e-01 -5.38726926e-01 5.10641992e-01 1.43253016e+00 1.22013450e+00 5.70658982e-01 9.43068266e-01 1.33123004e+00 -1.13662255e+00 -1.08418798e+00 -8.48776042e-01 -7.49699950e-01 -4.56738293e-01 3.72895092e-01 -7.35097170e-01 -5.01964450e-01 -5.51375031e-01]

[5.601675033569336, 2.5356552600860596]

d4a5dd9a-01cc-429a-b808-c9aef91df17b

mining-fine-grained-semantics-via-graph

2201.06885

https://arxiv.org/abs/2201.06885v2

https://arxiv.org/pdf/2201.06885v2.pdf

Evidence-aware Fake News Detection with Graph Neural Networks

The prevalence and perniciousness of fake news has been a critical issue on the Internet, which stimulates the development of automatic fake news detection in turn. In this paper, we focus on the evidence-based fake news detection, where several evidences are utilized to probe the veracity of news (i.e., a claim). Most previous methods first employ sequential models to embed the semantic information and then capture the claim-evidence interaction based on different attention mechanisms. Despite their effectiveness, they still suffer from two main weaknesses. Firstly, due to the inherent drawbacks of sequential models, they fail to integrate the relevant information that is scattered far apart in evidences for veracity checking. Secondly, they neglect much redundant information contained in evidences that may be useless or even harmful. To solve these problems, we propose a unified Graph-based sEmantic sTructure mining framework, namely GET in short. Specifically, different from the existing work that treats claims and evidences as sequences, we model them as graph-structured data and capture the long-distance semantic dependency among dispersed relevant snippets via neighborhood propagation. After obtaining contextual semantic information, our model reduces information redundancy by performing graph structure learning. Finally, the fine-grained semantic representations are fed into the downstream claim-evidence interaction module for predictions. Comprehensive experiments have demonstrated the superiority of GET over the state-of-the-arts.

['Liang Wang', 'Shu Wu', 'Qiang Liu', 'Junfei Wu', 'Weizhi Xu']

2022-01-18

['graph-structure-learning']

['graphs']

[-2.12408118e-02 1.44783244e-01 -9.49166775e-01 9.52813402e-02 -4.14516807e-01 -1.84009701e-01 5.62862694e-01 6.55740142e-01 8.39345083e-02 7.09148765e-01 5.28357506e-01 -3.69480312e-01 -3.18997175e-01 -1.05115855e+00 -7.57916451e-01 -1.98672861e-01 2.38128558e-01 2.58991361e-01 7.68713951e-01 -5.78622520e-01 7.18193352e-01 -7.22599030e-02 -1.15745270e+00 3.93307149e-01 1.18194509e+00 9.45486426e-01 -1.54914439e-01 -2.84694493e-01 -5.18050790e-01 1.16938937e+00 -7.64593720e-01 -8.29155624e-01 -2.29539111e-01 -7.29158282e-01 -6.07750773e-01 2.55755126e-01 -1.79229245e-01 9.46205407e-02 -7.62419283e-01 1.55380869e+00 1.88905504e-02 -4.15337890e-01 4.07914400e-01 -1.41825771e+00 -1.04712677e+00 8.36067498e-01 -8.04948270e-01 5.13714194e-01 3.09376061e-01 -2.25239456e-01 1.21636045e+00 -7.67022252e-01 8.17786515e-01 1.08800876e+00 7.27333188e-01 -1.38070425e-02 -6.19891584e-01 -5.79568088e-01 3.04489315e-01 7.54945576e-01 -1.15487874e+00 -2.49717906e-01 1.26587343e+00 -3.08625966e-01 5.22381663e-01 1.68680638e-01 9.17102635e-01 1.27409375e+00 1.76324338e-01 9.60027337e-01 1.09568369e+00 -2.61383921e-01 1.11409545e-01 1.02499127e-01 3.40763479e-01 8.98834467e-01 8.64013195e-01 -1.86027721e-01 -5.05720794e-01 -4.20476228e-01 3.50687057e-01 1.29145831e-01 -3.73585165e-01 -6.77588955e-02 -1.00590765e+00 1.23449004e+00 6.11510098e-01 5.52371144e-01 -5.57595730e-01 -1.38630196e-01 6.16538882e-01 3.50050181e-01 7.41584480e-01 1.57225281e-01 -2.86872745e-01 1.92856178e-01 -8.23814809e-01 1.86988235e-01 7.25848675e-01 8.58076990e-01 4.24683779e-01 -1.95577785e-01 -8.56135339e-02 7.15172470e-01 4.40839022e-01 1.63813904e-02 5.12479067e-01 -2.65884846e-01 9.88722920e-01 1.14938450e+00 4.00594808e-03 -2.15245891e+00 -2.19339401e-01 -7.34330893e-01 -7.04872906e-01 -6.10586762e-01 7.88337458e-03 2.04135746e-01 -6.42677605e-01 1.21449614e+00 3.94161880e-01 6.27550900e-01 -9.41732377e-02 9.56887543e-01 9.95900810e-01 6.09874964e-01 1.93050317e-02 -5.73208451e-01 1.41134238e+00 -1.18020380e+00 -1.11416423e+00 -2.82471806e-01 6.10999763e-01 -7.26156950e-01 8.45437646e-01 1.96372956e-01 -6.68496788e-01 4.97269109e-02 -1.19361448e+00 2.07405791e-01 -4.73235905e-01 -3.48784298e-01 6.33616745e-01 4.67222810e-01 -3.24569702e-01 4.48910683e-01 -3.55476856e-01 3.50885466e-02 8.76950800e-01 -2.81164289e-01 3.88886221e-02 -3.56208086e-01 -1.85822141e+00 7.84641087e-01 6.33341551e-01 -7.52487853e-02 -3.76063675e-01 -2.85246760e-01 -7.22311139e-01 5.40757261e-04 1.25450480e+00 -5.43870330e-01 6.35619760e-01 -8.33011031e-01 -1.01451409e+00 4.81263071e-01 -4.55204137e-02 -4.52647120e-01 5.89628577e-01 2.71518707e-01 -1.10856271e+00 3.17098200e-01 4.74052191e-01 -3.16072077e-01 8.06145489e-01 -1.41216743e+00 -5.00665843e-01 -4.82076108e-01 2.06753612e-02 6.62633181e-02 -5.08048236e-01 8.53253454e-02 -7.07188785e-01 -1.10767341e+00 6.13618374e-01 -4.12479520e-01 -6.92733899e-02 -5.83780587e-01 -7.83958316e-01 -2.78503507e-01 1.07502997e+00 -7.68170893e-01 1.76604211e+00 -1.59785807e+00 -1.00462295e-01 3.22104096e-01 5.74218214e-01 2.93519795e-01 1.02321595e-01 6.39706492e-01 2.91343927e-01 4.32819843e-01 -1.88840538e-01 1.30727246e-01 -3.14009994e-01 1.21529363e-01 -3.79113078e-01 6.70460820e-01 -6.21591508e-03 1.14077663e+00 -1.13390207e+00 -8.57700050e-01 -2.28602752e-01 1.65250495e-01 -2.52713054e-01 -4.24953789e-01 -4.15922642e-01 3.65500987e-01 -1.03849721e+00 9.65840757e-01 6.46268785e-01 -7.56741881e-01 2.57381737e-01 -3.24266762e-01 2.33921692e-01 6.18223548e-01 -7.55008280e-01 1.30842710e+00 -4.01865095e-02 6.01817854e-02 -1.54231861e-01 -1.53532839e+00 8.60622883e-01 1.78740859e-01 5.65365016e-01 -9.73572433e-01 1.90470994e-01 3.90371710e-01 -1.57691792e-01 -7.55534708e-01 3.93421501e-01 -3.56864840e-01 -1.92837402e-01 4.15621191e-01 -3.43296200e-01 2.41965398e-01 1.02386682e-03 5.11098325e-01 1.11663294e+00 -2.89549917e-01 3.76630604e-01 7.79648125e-02 5.68831563e-01 4.89001632e-01 6.82290018e-01 5.04801571e-01 -1.57909155e-01 9.53199193e-02 8.40294421e-01 -1.30146787e-01 -8.10867667e-01 -5.30084133e-01 1.48736775e-01 4.69924271e-01 9.25449193e-01 -6.47106886e-01 -2.60113686e-01 -1.29220092e+00 1.50725305e-01 8.09141994e-01 -5.51969707e-01 -3.41174662e-01 -4.19233322e-01 -8.22220027e-01 4.15838003e-01 4.09696177e-02 7.03223228e-01 -9.12884653e-01 1.89671531e-01 4.85899121e-01 -7.97638595e-01 -1.12148380e+00 -3.22491914e-01 -6.38827443e-01 -8.48509669e-01 -1.38359106e+00 -1.00682415e-01 -4.86150920e-01 3.99221689e-01 8.10323775e-01 9.87855852e-01 6.42554581e-01 1.97061703e-01 -1.94421694e-01 -6.88913763e-01 -3.33215415e-01 -4.31714088e-01 -1.97332591e-01 -2.42929488e-01 1.56037122e-01 4.74735826e-01 -4.37853456e-01 -6.14068806e-01 4.96050656e-01 -9.22815025e-01 -8.84068534e-02 5.94446898e-01 9.04109478e-01 5.55484474e-01 6.91783965e-01 1.10607994e+00 -1.21734798e+00 1.07051551e+00 -1.27995801e+00 -4.19212580e-01 3.68997753e-01 -1.00346160e+00 -3.25978845e-01 6.31345332e-01 -2.31254771e-01 -8.89130056e-01 -7.05082953e-01 1.24886408e-01 -3.52256298e-01 4.41524595e-01 1.21955431e+00 -1.23467356e-01 1.36307672e-01 4.09897685e-01 3.17917049e-01 1.21025499e-02 -1.95582077e-01 3.06424409e-01 6.87910318e-01 1.06601538e-02 -2.43291005e-01 7.33659327e-01 6.06560051e-01 -1.12400390e-01 -4.15663749e-01 -1.35865283e+00 -5.92215598e-01 4.78618257e-02 -1.87630385e-01 4.38890874e-01 -7.40355015e-01 -6.41039729e-01 2.37654433e-01 -1.17127955e+00 5.99261880e-01 1.80882916e-01 3.94081175e-01 -2.18105093e-01 8.57902169e-01 -8.21652412e-01 -7.11148858e-01 2.14929003e-02 -8.65533829e-01 5.90337873e-01 -1.27393410e-01 2.04875499e-01 -1.04515207e+00 -2.18127787e-01 9.63754535e-01 3.70410942e-02 2.21679851e-01 8.82940650e-01 -8.82947564e-01 -5.93884885e-01 -5.14920235e-01 -6.25843763e-01 1.24386735e-02 2.47926369e-01 -4.58093792e-01 -2.73607194e-01 1.28513083e-01 3.37939322e-01 -3.32717389e-01 1.03258145e+00 1.30181551e-01 1.05410540e+00 -1.10764539e+00 -5.24146080e-01 -6.83527216e-02 1.32577693e+00 -1.19866073e-01 5.96502721e-01 6.13882720e-01 7.41601288e-01 7.21755147e-01 9.40317333e-01 4.29957837e-01 6.98347986e-01 5.91122866e-01 6.92895472e-01 1.80479839e-01 8.13149102e-03 -7.77245283e-01 4.73950654e-02 1.03577793e+00 -1.66298240e-01 -5.63309014e-01 -6.66234255e-01 4.96686697e-01 -2.18425536e+00 -1.34294760e+00 -6.66879714e-01 1.71202230e+00 6.22724712e-01 4.81097698e-01 2.69727856e-01 2.54160136e-01 1.14510906e+00 4.88502145e-01 -3.59037310e-01 1.81132793e-01 -4.23066139e-01 -4.08712029e-01 7.02372432e-01 2.34361276e-01 -8.43646765e-01 1.09840393e+00 4.97071552e+00 1.19574320e+00 -7.50118136e-01 5.08493900e-01 4.75767195e-01 3.47707987e-01 -7.86403179e-01 2.31079534e-01 -3.67563576e-01 1.02283549e+00 4.84205395e-01 -6.03337772e-02 1.44956738e-01 6.14237487e-01 3.52079391e-01 1.05454788e-01 -1.28168434e-01 7.46660888e-01 3.50218862e-01 -1.89266193e+00 1.98878288e-01 2.07768127e-01 7.83091366e-01 -1.60570592e-01 -1.58618674e-01 1.86012909e-01 2.76360512e-01 -6.62289739e-01 8.99937928e-01 4.80433643e-01 2.38135442e-01 -7.59042382e-01 9.90909040e-01 6.82646096e-01 -8.61830711e-01 -2.60025948e-01 -4.67446059e-01 3.57997492e-02 4.03682530e-01 1.05000305e+00 -5.91546178e-01 1.07783580e+00 2.49138534e-01 1.01949489e+00 -3.07153404e-01 8.88299882e-01 -5.99985778e-01 8.56097162e-01 7.72628486e-02 -3.34335059e-01 5.12953281e-01 -3.43163699e-01 6.66267633e-01 9.14939940e-01 3.88355911e-01 1.62955925e-01 3.59046966e-01 7.64750063e-01 -4.51840758e-01 3.07174653e-01 -6.76670194e-01 -3.14939201e-01 6.56206906e-01 8.39267790e-01 -8.51182401e-01 -5.41323841e-01 -6.82332277e-01 8.67193520e-01 2.77925104e-01 1.04538389e-01 -1.07122409e+00 -1.42762244e-01 -4.20337580e-02 4.30544436e-01 3.73430490e-01 -3.76159581e-03 -4.34453100e-01 -1.45166886e+00 3.17220062e-01 -8.14090073e-01 6.90932572e-01 -4.61024672e-01 -1.79346144e+00 3.64692777e-01 -3.09318960e-01 -1.43270290e+00 4.00858670e-01 -3.13413665e-02 -6.53927743e-01 3.66919160e-01 -1.71156490e+00 -1.29291975e+00 -7.84913916e-03 6.32191837e-01 4.68965381e-01 -1.38926506e-01 1.83790118e-01 3.52171421e-01 -5.35125434e-01 2.77299404e-01 -1.75131842e-01 1.86492354e-01 2.59877533e-01 -5.12584269e-01 5.61817773e-02 8.25132608e-01 2.69921750e-01 4.83133525e-01 6.92561090e-01 -1.41614878e+00 -1.21986127e+00 -1.00337315e+00 1.05893540e+00 -3.19923759e-01 1.31302500e+00 -4.15121838e-02 -1.06833076e+00 4.63750929e-01 -1.04013257e-01 1.83978677e-01 4.96473461e-01 1.02454692e-01 -6.69892073e-01 3.32178891e-01 -1.15275395e+00 5.27227581e-01 1.39096189e+00 -3.23743552e-01 -1.08162034e+00 6.20901227e-01 9.32914078e-01 8.09133723e-02 -4.76626754e-01 3.38561743e-01 1.03098273e-01 -1.02445602e+00 8.64281058e-01 -8.37355316e-01 7.62329519e-01 -3.23952615e-01 1.13458738e-01 -1.17505813e+00 -3.07922482e-01 -2.02720508e-01 -5.79525828e-01 1.04567206e+00 4.72161353e-01 -7.71397948e-01 8.36538553e-01 -3.11344981e-01 -1.63367942e-01 -8.19642425e-01 -9.41470146e-01 -9.11532164e-01 -4.23807710e-01 -4.13484961e-01 6.58191919e-01 1.62968194e+00 4.07415122e-01 6.54325068e-01 -5.45651495e-01 2.64242738e-01 6.97975338e-01 5.73260546e-01 2.22378999e-01 -1.37611926e+00 -1.57187000e-01 -6.31552160e-01 -4.96822894e-01 -9.53531027e-01 1.41128212e-01 -1.00125754e+00 -2.80016690e-01 -1.81330299e+00 2.63665199e-01 -5.77093065e-01 -3.56386125e-01 2.35284597e-01 -3.41614693e-01 1.06798723e-01 -1.75047264e-01 6.90602303e-01 -9.08350766e-01 8.11448574e-01 1.51274812e+00 -4.12953049e-01 -3.64866741e-02 8.31347853e-02 -1.06231654e+00 9.05293465e-01 7.60710120e-01 -8.17227304e-01 -5.93385518e-01 -1.41959459e-01 5.54956138e-01 4.39822793e-01 5.20761132e-01 -4.31849748e-01 3.37843090e-01 -4.34951186e-01 -1.01500168e-01 -6.40315533e-01 1.57144532e-01 -6.79074168e-01 -1.92276090e-01 5.35804987e-01 -7.41360039e-02 -7.35068172e-02 -3.38910729e-01 1.42799330e+00 -5.23960590e-01 -2.47075722e-01 4.01603550e-01 -2.97912180e-01 -5.97200096e-01 4.55416083e-01 -9.43401307e-02 2.71931589e-01 9.30991888e-01 -2.19996870e-01 -8.45666826e-01 -4.91761774e-01 -5.43856800e-01 6.60297200e-02 2.46806204e-01 5.08534968e-01 8.25801075e-01 -1.39269841e+00 -9.07842398e-01 -3.58375728e-01 5.14395416e-01 -4.79647189e-01 4.17004645e-01 1.18754900e+00 -1.15222342e-01 1.78975388e-01 2.35250115e-01 -2.08731685e-02 -9.05618846e-01 1.17524457e+00 -2.72384167e-01 -4.50148284e-01 -7.34228730e-01 4.25387412e-01 -4.22750503e-01 6.36629686e-02 -3.07393551e-01 2.57456303e-01 -5.33906221e-01 1.87867567e-01 2.84941167e-01 3.75306696e-01 1.01818413e-01 -7.73632288e-01 -3.79468322e-01 1.81912780e-01 1.85347795e-02 1.88491702e-01 1.49177754e+00 -2.67368138e-01 -4.62893248e-01 2.36354187e-01 9.30306792e-01 6.07681334e-01 -5.54775000e-01 -4.83352602e-01 4.43790346e-01 -8.69641185e-01 1.67812005e-01 -7.27148116e-01 -1.22574961e+00 3.64077777e-01 -3.68706167e-01 7.97226787e-01 6.91633284e-01 3.72798979e-01 9.94229972e-01 7.07009360e-02 4.47448671e-01 -1.19989789e+00 2.65924990e-01 2.56445348e-01 8.90975118e-01 -1.16142738e+00 2.46911705e-01 -1.12354505e+00 -7.53550231e-01 7.87591636e-01 3.81032556e-01 -1.29435003e-01 6.11255050e-01 -2.31811911e-01 -2.97868401e-01 -9.05722201e-01 -3.06755960e-01 -1.20171279e-01 3.13690841e-01 2.54512906e-01 -5.50523847e-02 2.38848627e-02 -9.91238296e-01 8.85675490e-01 1.37159750e-01 -4.56277020e-02 6.26481175e-01 7.15618610e-01 -5.94439626e-01 -8.68351758e-01 -2.92920738e-01 6.14688277e-01 -5.75718820e-01 -2.43844494e-01 -5.02533317e-01 7.27984548e-01 2.33758137e-01 1.28810668e+00 -5.93318403e-01 -5.96367538e-01 2.90647566e-01 -2.70001531e-01 5.29641882e-02 -4.14732933e-01 -2.06058756e-01 -4.62197419e-03 4.23838139e-01 -4.59097147e-01 -5.54378271e-01 -3.64925474e-01 -1.18329275e+00 -3.93609673e-01 -7.26765931e-01 4.55920160e-01 5.71069598e-01 1.15955973e+00 5.01102149e-01 5.16112089e-01 6.54851675e-01 1.88833058e-01 -5.48191369e-01 -8.77693295e-01 -6.03021979e-01 5.06554723e-01 5.02379723e-02 -1.03632641e+00 -4.44877267e-01 -4.82929200e-01]

[8.133291244506836, 10.212803840637207]

1c6d40fc-f1ed-4806-ab2f-8522021fcfc7

multi-output-learning-for-camera

http://openaccess.thecvf.com/content_cvpr_2014/html/Guzman-Rivera_Multi-Output_Learning_for_2014_CVPR_paper.html

http://openaccess.thecvf.com/content_cvpr_2014/papers/Guzman-Rivera_Multi-Output_Learning_for_2014_CVPR_paper.pdf

Multi-Output Learning for Camera Relocalization

We address the problem of estimating the pose of a cam- era relative to a known 3D scene from a single RGB-D frame. We formulate this problem as inversion of the generative rendering procedure, i.e., we want to find the camera pose corresponding to a rendering of the 3D scene model that is most similar with the observed input. This is a non-convex optimization problem with many local optima. We propose a hybrid discriminative-generative learning architecture that consists of: (i) a set of M predictors which generate M camera pose hypotheses; and (ii) a 'selector' or 'aggregator' that infers the best pose from the multiple pose hypotheses based on a similarity function. We are interested in predictors that not only produce good hypotheses but also hypotheses that are different from each other. Thus, we propose and study methods for learning 'marginally relevant' predictors, and compare their performance when used with different selection procedures. We evaluate our method on a recently released 3D reconstruction dataset with challenging camera poses, and scene variability. Experiments show that our method learns to make multiple predictions that are marginally relevant and can effectively select an accurate prediction. Furthermore, our method outperforms the state-of-the-art discriminative approach for camera relocalization.

['Andrew Fitzgibbon', 'Ben Glocker', 'Abner Guzman-Rivera', 'Toby Sharp', 'Shahram Izadi', 'Jamie Shotton', 'Pushmeet Kohli']

2014-06-01

cvpr-2014-6

['camera-relocalization']

['computer-vision']

[ 4.16101843e-01 -5.87185025e-02 3.62712033e-02 -5.70459604e-01 -1.32919061e+00 -5.55998087e-01 5.73811114e-01 -2.74389893e-01 -2.11589172e-01 2.65396923e-01 2.14378461e-01 4.94503081e-02 -7.33421941e-04 -5.21163583e-01 -1.11335516e+00 -7.99880505e-01 2.89343625e-01 1.08602500e+00 3.04592878e-01 9.01987031e-02 3.25710446e-01 5.74719489e-01 -1.65181386e+00 2.65066296e-01 5.78879833e-01 1.06488538e+00 6.93932414e-01 7.59311378e-01 2.82409102e-01 6.44407749e-01 -5.05757213e-01 -2.11295888e-01 3.81138355e-01 -5.47413647e-01 -6.26810372e-01 6.14204943e-01 7.08479941e-01 -1.35138959e-01 1.52029410e-01 8.79405856e-01 4.65301752e-01 1.44596115e-01 8.35695148e-01 -1.02832556e+00 -1.51401624e-01 -6.31832182e-02 -5.65677643e-01 -1.37113601e-01 8.26304257e-01 -1.16513506e-01 9.13087726e-01 -1.04531825e+00 8.08217943e-01 1.35905230e+00 4.57949579e-01 4.46321726e-01 -1.62311804e+00 -2.75053084e-01 2.82191068e-01 1.06154628e-01 -1.66512096e+00 -4.33480263e-01 9.64992285e-01 -6.07001960e-01 8.43221903e-01 2.77071625e-01 7.00137258e-01 1.21707177e+00 2.58241475e-01 6.44491315e-01 1.14757025e+00 -3.91680241e-01 3.85803491e-01 1.96513608e-01 -3.65237474e-01 7.12596774e-01 -1.31395862e-01 -6.53377622e-02 -7.51869977e-01 -1.94988683e-01 7.51776218e-01 -2.25735173e-01 -4.15572315e-01 -9.26286042e-01 -1.18272209e+00 8.41459036e-01 2.74812281e-01 -2.95371801e-01 -4.28586483e-01 1.10803343e-01 -3.20246011e-01 -6.44636601e-02 4.47640449e-01 5.48943460e-01 -4.78122562e-01 2.59469002e-02 -6.41591251e-01 3.09257776e-01 9.23052609e-01 1.07494426e+00 1.10580075e+00 -2.17495784e-01 2.40468845e-01 6.95128918e-01 5.66613555e-01 5.27756870e-01 1.41088471e-01 -1.03304827e+00 4.66239691e-01 4.35575724e-01 7.90363401e-02 -9.16955352e-01 -2.51722217e-01 -3.06917369e-01 -5.96806705e-01 2.27822497e-01 2.04473697e-02 3.74765657e-02 -8.88292134e-01 1.63275135e+00 4.87952739e-01 3.96273077e-01 -9.32896510e-02 1.13645017e+00 6.05362713e-01 7.85285950e-01 -3.40673774e-01 -2.83556700e-01 8.53359461e-01 -7.20268607e-01 -1.03344828e-01 -6.17873251e-01 1.86070532e-01 -9.55791175e-01 8.71574342e-01 4.53595310e-01 -9.17630970e-01 -7.03829229e-01 -8.50221515e-01 -7.36818910e-02 7.58038163e-02 4.13590163e-01 2.49318525e-01 3.01465034e-01 -1.14230072e+00 5.12900352e-01 -8.50828111e-01 -3.55407119e-01 -6.85966611e-02 4.91388768e-01 -4.22767848e-01 -2.37025321e-02 -3.58234733e-01 1.16554821e+00 4.55142915e-01 4.86631952e-02 -1.18779111e+00 -3.67797852e-01 -8.78964961e-01 -3.34303856e-01 5.93578398e-01 -1.02999544e+00 1.05745173e+00 -9.61092412e-01 -1.60613561e+00 1.04634786e+00 -4.31887269e-01 -4.53278311e-02 3.99479002e-01 -4.77523863e-01 -6.18782779e-03 -1.07908258e-02 1.49084508e-01 7.73289502e-01 1.03972006e+00 -1.83402383e+00 -6.79111540e-01 -3.94740075e-01 -1.34093696e-02 6.44444585e-01 3.92007262e-01 -1.61986768e-01 -9.85962927e-01 -2.32769683e-01 7.54172742e-01 -1.37013686e+00 -4.35949445e-01 -1.33489415e-01 -6.41547203e-01 9.98730659e-02 6.81709826e-01 -5.39273262e-01 7.33660042e-01 -1.89163959e+00 9.05509949e-01 4.09704924e-01 1.38631508e-01 -2.59813339e-01 6.53242022e-02 2.47593567e-01 -3.20578516e-02 -2.23816797e-01 -4.96598072e-02 -9.12529707e-01 -9.06863436e-02 3.91908973e-01 -3.73656511e-01 6.44088566e-01 1.02243640e-01 4.65676636e-01 -6.43942773e-01 -5.34056425e-01 5.56585133e-01 5.51517606e-01 -7.24025667e-01 7.28047907e-01 -3.64339978e-01 7.67450929e-01 -4.27954793e-01 5.61240077e-01 5.45681238e-01 -3.23327690e-01 2.82003790e-01 -3.97083640e-01 -4.04570214e-02 2.79283524e-01 -1.51017213e+00 1.90287030e+00 -3.61351639e-01 3.61325383e-01 -2.23023131e-01 -6.96348071e-01 1.23785532e+00 -1.52741104e-01 3.57794195e-01 -1.83855399e-01 1.99647903e-01 1.29433036e-01 -4.71266061e-01 -3.91689628e-01 4.12386835e-01 -1.86841488e-01 -1.08829834e-01 1.81147382e-01 2.27782249e-01 -5.80811858e-01 -2.54391491e-01 -5.66107035e-02 8.50479066e-01 6.71069384e-01 4.28065211e-01 -4.04345393e-02 3.45452458e-01 -2.33321577e-01 6.35078490e-01 6.43138289e-01 2.74502665e-01 1.00214028e+00 4.46867794e-01 -4.90612507e-01 -1.03605020e+00 -1.13687897e+00 1.40580833e-01 7.87636101e-01 6.90449893e-01 -5.45420766e-01 -4.70127821e-01 -5.93591452e-01 -6.25801533e-02 7.50275970e-01 -5.84459662e-01 -6.13042749e-02 -6.53309286e-01 -5.88766456e-01 -3.31061810e-01 3.59505296e-01 1.02243669e-01 -5.71261346e-01 -8.16508591e-01 -5.84403500e-02 -3.18236917e-01 -1.18251204e+00 -4.23685193e-01 4.75678086e-01 -9.17734087e-01 -9.77320552e-01 -2.40733996e-01 -6.31365538e-01 8.66341174e-01 3.11222583e-01 1.42635357e+00 -3.72220755e-01 1.46943107e-01 5.14316559e-01 -3.00432801e-01 -3.04921061e-01 -5.10623991e-01 3.47944833e-02 1.01962633e-01 1.94115072e-01 8.44370648e-02 -4.10674870e-01 -3.86576653e-01 5.90172172e-01 -6.21573567e-01 4.70448494e-01 7.64708996e-01 5.21810949e-01 1.31637919e+00 -2.16673568e-01 -2.46683255e-01 -7.58919656e-01 1.46461219e-01 -3.41458052e-01 -7.76894569e-01 3.11102062e-01 -4.17152375e-01 4.24601972e-01 5.02740681e-01 -4.93723541e-01 -9.80489492e-01 8.98363590e-01 -6.71771094e-02 -9.17635798e-01 -3.10972273e-01 2.36226618e-01 -5.35561025e-01 -1.90052792e-01 6.31335914e-01 2.32055247e-01 -2.90722132e-01 -6.93061233e-01 2.91121483e-01 1.94134578e-01 6.02136612e-01 -7.71651566e-01 9.92482066e-01 2.66321033e-01 2.52954066e-01 -7.13508368e-01 -1.03327620e+00 -5.64689279e-01 -1.05119169e+00 -4.87470090e-01 9.35222268e-01 -1.05271232e+00 -5.51154137e-01 1.61601469e-01 -1.35877681e+00 -1.71502784e-01 -1.67925745e-01 6.64168596e-01 -9.08724666e-01 1.30715638e-01 -9.32148769e-02 -6.68539107e-01 1.72110900e-01 -1.48747861e+00 1.81308413e+00 1.58526525e-01 -2.19620511e-01 -7.14398682e-01 4.92878765e-01 3.51273537e-01 -1.87377051e-01 4.94044602e-01 6.01507068e-01 -2.63726085e-01 -1.08627987e+00 -1.98610470e-01 1.21863209e-01 2.21383795e-01 -1.43235296e-01 1.77510846e-02 -1.18505156e+00 -3.92024010e-01 7.65911788e-02 -2.05015451e-01 7.19275951e-01 4.54548568e-01 1.04639757e+00 -2.54941791e-01 -4.18447703e-01 1.14635897e+00 1.59937000e+00 2.62342632e-01 4.98155117e-01 2.72767484e-01 9.42431271e-01 3.76233101e-01 6.64094031e-01 4.80780512e-01 5.31983733e-01 1.11777568e+00 7.90725172e-01 7.87049830e-02 1.89710557e-01 -6.43618405e-01 3.76054168e-01 7.51397371e-01 -2.49873206e-01 -3.22339505e-01 -8.93227696e-01 9.57520753e-02 -1.98360527e+00 -6.16325378e-01 7.66065270e-02 2.40622830e+00 4.07750815e-01 1.58891138e-02 1.98334046e-02 -2.81473815e-01 5.97444654e-01 2.41296127e-01 -5.48254251e-01 -8.90624672e-02 -8.03041533e-02 7.79142007e-02 3.54859442e-01 5.93355775e-01 -1.31098437e+00 8.28084052e-01 6.13583517e+00 4.30347651e-01 -9.59952056e-01 -1.01828322e-01 5.28883755e-01 1.22673111e-02 -3.79026085e-01 3.54775995e-01 -1.16276455e+00 2.40915850e-01 6.27743721e-01 1.38361052e-01 4.79143798e-01 1.02263546e+00 -9.10088122e-02 -3.05572152e-01 -1.42841387e+00 1.28420377e+00 6.45087123e-01 -1.20199406e+00 1.61115274e-01 2.15028241e-01 8.70500147e-01 2.97109894e-02 -7.55427629e-02 -9.13091153e-02 3.02201957e-01 -8.61132562e-01 1.13063002e+00 8.15292418e-01 4.54479128e-01 -6.35748625e-01 5.39143562e-01 5.54235756e-01 -1.19851947e+00 9.39326510e-02 -3.96853328e-01 2.75606692e-01 2.31122077e-01 4.58847255e-01 -1.01743066e+00 4.33604360e-01 6.82975233e-01 8.73298168e-01 -6.63621783e-01 1.19709921e+00 -5.62085569e-01 2.13678434e-01 -4.24270064e-01 6.52293265e-02 -3.01052667e-02 -4.27610844e-01 7.23143578e-01 9.75622296e-01 5.89337230e-01 1.37942716e-01 5.45988798e-01 7.66578257e-01 3.53878140e-02 -1.44223142e-02 -7.66809106e-01 6.54452085e-01 3.67656887e-01 1.00438416e+00 -6.94592178e-01 -1.04149900e-01 -1.14613593e-01 1.18659151e+00 3.99725646e-01 2.27659494e-01 -8.61908913e-01 4.11505252e-01 4.79076922e-01 -2.58175265e-02 3.91087919e-01 -3.66603583e-01 3.80151197e-02 -1.28547251e+00 7.98336416e-02 -7.28313744e-01 3.28563362e-01 -1.19037449e+00 -1.06951952e+00 8.31227899e-01 1.77145004e-01 -1.62948465e+00 -5.26046216e-01 -5.04308105e-01 -3.45742822e-01 8.83604646e-01 -1.18163121e+00 -1.21926856e+00 -5.27177513e-01 5.72244108e-01 6.42801285e-01 1.13454619e-02 8.32520425e-01 -1.33801907e-01 -3.54659319e-01 5.80816679e-02 -2.05454342e-02 -3.99151862e-01 6.26087546e-01 -1.38884115e+00 1.17364667e-01 8.67769837e-01 5.33386171e-01 1.94500431e-01 7.90408194e-01 -5.36061764e-01 -1.64920139e+00 -1.13622034e+00 7.75719225e-01 -9.21628952e-01 1.66970611e-01 -5.06997645e-01 -6.57720745e-01 9.01409626e-01 -2.22616315e-01 -2.35749334e-01 4.49134439e-01 1.80200294e-01 -1.49936989e-01 -2.16093659e-01 -7.64138639e-01 3.54437351e-01 9.75622654e-01 -5.53283632e-01 -4.68195468e-01 4.11814839e-01 4.55897421e-01 -9.27149415e-01 -7.36288905e-01 1.97987840e-01 4.01815623e-01 -1.09395766e+00 1.28455281e+00 -2.84486890e-01 4.37223375e-01 -5.04393458e-01 -5.86906135e-01 -1.35663950e+00 -3.81126076e-01 -4.79099363e-01 -5.88636547e-02 9.19045150e-01 3.04597229e-01 -1.89807713e-01 8.08164179e-01 5.68223536e-01 -2.71572649e-01 -8.05669785e-01 -9.28366303e-01 -6.33505523e-01 -4.95231718e-01 -4.47686940e-01 3.41263860e-01 5.83353639e-01 -7.25003123e-01 7.57276952e-01 -6.85411751e-01 6.61280334e-01 7.91824102e-01 5.41175127e-01 1.23570025e+00 -1.14151108e+00 -6.84537470e-01 -1.06649898e-01 -5.86088896e-01 -1.41502059e+00 2.62772560e-01 -6.20747447e-01 5.27680993e-01 -1.31327665e+00 4.24107850e-01 -3.61771107e-01 2.18002677e-01 2.84763038e-01 -2.10250586e-01 -9.18049831e-03 3.28030765e-01 5.23526192e-01 -7.90763974e-01 5.74513495e-01 1.03716350e+00 -1.09558748e-02 -3.98525834e-01 2.70072639e-01 -4.81023312e-01 9.32995617e-01 3.60642105e-01 -5.41811287e-01 -4.24248695e-01 -5.56063592e-01 3.20431024e-01 3.50164205e-01 5.49712718e-01 -9.90584791e-01 1.70871004e-01 -2.53296256e-01 6.91977441e-01 -8.17141056e-01 8.62516403e-01 -8.87874067e-01 6.41741216e-01 7.97181055e-02 -2.00069174e-01 2.17991635e-01 -7.26057440e-02 6.06535733e-01 -8.02297145e-02 -1.42243043e-01 7.70418644e-01 -1.66677684e-01 -1.02511895e+00 3.49906296e-01 3.38265151e-02 -1.60734206e-01 9.74314511e-01 -2.74294436e-01 1.09729655e-02 -5.26180029e-01 -7.24440038e-01 -1.17673412e-01 7.86544025e-01 4.35732394e-01 8.59646320e-01 -1.31041598e+00 -6.49690449e-01 2.69770354e-01 2.03430235e-01 3.54432851e-01 -1.77917808e-01 5.27530849e-01 -4.46538717e-01 2.37774327e-01 9.38087627e-02 -1.16911411e+00 -1.32726526e+00 4.45467174e-01 3.70724469e-01 -1.17340967e-01 -4.36659068e-01 1.04420269e+00 3.34886461e-01 -4.83010232e-01 3.69857550e-02 -2.72831857e-01 1.96388625e-02 -8.45903605e-02 3.15483101e-02 1.78138092e-01 4.23577949e-02 -1.10630965e+00 -5.21527946e-01 1.02892995e+00 3.64923090e-01 -1.35986179e-01 1.36898959e+00 -1.63326576e-01 6.40445650e-02 5.42005002e-01 1.42756557e+00 -1.42277047e-01 -1.52110744e+00 -2.23176390e-01 -2.51479745e-01 -9.58987594e-01 5.79366907e-02 -6.27932847e-01 -9.19260204e-01 6.25908375e-01 6.55143201e-01 -1.36935577e-01 1.34642386e+00 5.66559315e-01 1.88326731e-01 3.51289570e-01 6.97144330e-01 -8.13203573e-01 2.48954445e-01 4.00956392e-01 1.07678604e+00 -1.27953541e+00 2.64012337e-01 -4.01519179e-01 -7.54080236e-01 1.05507827e+00 6.72644794e-01 -1.60037890e-01 4.48563397e-01 1.46172017e-01 -2.00872682e-02 -3.04215550e-01 -8.26646030e-01 -2.45699748e-01 6.79134190e-01 5.29476404e-01 6.97791278e-02 1.63464054e-01 4.19779241e-01 1.04235329e-01 -3.88188511e-01 -3.98144156e-01 2.59282261e-01 7.25360572e-01 -3.58824611e-01 -1.00587261e+00 -6.58552468e-01 3.03398758e-01 2.83559859e-01 2.77679145e-01 -6.11233771e-01 6.76693678e-01 3.07374835e-01 5.69707513e-01 3.28739062e-02 -7.99861073e-01 5.22194624e-01 -1.05200104e-01 6.50416434e-01 -7.46053338e-01 -4.30535525e-02 6.01703823e-01 1.31041463e-02 -9.61951494e-01 -5.70828021e-01 -9.74328160e-01 -6.88278377e-01 4.79784468e-03 -4.19811368e-01 -9.06038508e-02 7.53495634e-01 9.50513005e-01 1.79668933e-01 1.77746460e-01 1.01630664e+00 -1.44331515e+00 -3.52114528e-01 -5.51393628e-01 -4.51475829e-01 3.29188675e-01 2.65967727e-01 -8.15019906e-01 -3.21548522e-01 4.38816547e-01]

[8.143527030944824, -2.643231153488159]

d1669b2f-475a-4496-be98-512e76fd55a0

s-2-me-spatial-spectral-mutual-teaching-and

2306.00451

https://arxiv.org/abs/2306.00451v1

https://arxiv.org/pdf/2306.00451v1.pdf

S$^2$ME: Spatial-Spectral Mutual Teaching and Ensemble Learning for Scribble-supervised Polyp Segmentation

Fully-supervised polyp segmentation has accomplished significant triumphs over the years in advancing the early diagnosis of colorectal cancer. However, label-efficient solutions from weak supervision like scribbles are rarely explored yet primarily meaningful and demanding in medical practice due to the expensiveness and scarcity of densely-annotated polyp data. Besides, various deployment issues, including data shifts and corruption, put forward further requests for model generalization and robustness. To address these concerns, we design a framework of Spatial-Spectral Dual-branch Mutual Teaching and Entropy-guided Pseudo Label Ensemble Learning (S$^2$ME). Concretely, for the first time in weakly-supervised medical image segmentation, we promote the dual-branch co-teaching framework by leveraging the intrinsic complementarity of features extracted from the spatial and spectral domains and encouraging cross-space consistency through collaborative optimization. Furthermore, to produce reliable mixed pseudo labels, which enhance the effectiveness of ensemble learning, we introduce a novel adaptive pixel-wise fusion technique based on the entropy guidance from the spatial and spectral branches. Our strategy efficiently mitigates the deleterious effects of uncertainty and noise present in pseudo labels and surpasses previous alternatives in terms of efficacy. Ultimately, we formulate a holistic optimization objective to learn from the hybrid supervision of scribbles and pseudo labels. Extensive experiments and evaluation on four public datasets demonstrate the superiority of our method regarding in-distribution accuracy, out-of-distribution generalization, and robustness, highlighting its promising clinical significance. Our code is available at https://github.com/lofrienger/S2ME.

['Hongliang Ren', 'Mobarakol Islam', 'Yang Zhang', 'Mengya Xu', 'An Wang']

2023-06-01

['pseudo-label']

['miscellaneous']

[ 4.08239216e-01 1.43802091e-01 -5.06780803e-01 -3.27786565e-01 -1.27542710e+00 -4.67478663e-01 2.35545158e-01 4.35661435e-01 -2.85371512e-01 7.05175102e-01 2.58568943e-01 -3.19005311e-01 -5.81717908e-01 -4.25153762e-01 -5.30786097e-01 -1.14834523e+00 6.44517783e-03 -3.19200940e-02 -1.01268910e-01 8.90257116e-03 -1.90883651e-01 -4.06652912e-02 -1.12295580e+00 2.14362711e-01 1.63377678e+00 9.44930494e-01 3.02543521e-01 2.05526158e-01 2.38658682e-01 4.59306210e-01 -2.83567514e-02 -2.91787684e-01 1.92572579e-01 -4.27095950e-01 -5.38314939e-01 2.81755865e-01 4.99470122e-02 -2.10336000e-02 -2.88224407e-02 1.15165710e+00 7.08901405e-01 9.41400323e-03 5.31203091e-01 -7.44699657e-01 -4.48637754e-01 4.94399309e-01 -7.87160873e-01 5.08065708e-02 6.59775510e-02 3.17339987e-01 1.01662076e+00 -7.72156060e-01 2.78143078e-01 4.07827049e-01 8.70341301e-01 2.82752216e-01 -1.29006779e+00 -5.34518242e-01 3.38022441e-01 -2.13431910e-01 -1.28790891e+00 -1.50175214e-01 7.07332194e-01 -4.15359885e-01 3.03937763e-01 4.85862166e-01 6.81910574e-01 9.62593079e-01 4.99063693e-02 1.06805658e+00 1.21863091e+00 -4.29821163e-01 1.74814984e-02 3.66493911e-01 -5.33011891e-02 1.01588356e+00 7.06278831e-02 2.15610176e-01 -4.63505059e-01 -2.32666880e-01 3.87759954e-01 9.98482183e-02 -5.19126654e-01 -5.36634505e-01 -1.34475982e+00 6.76310360e-01 7.06766307e-01 2.57829636e-01 -1.81843355e-01 -1.40166521e-01 3.56794715e-01 -1.81064233e-01 6.84393346e-01 4.13476944e-01 -3.15182745e-01 1.56222358e-01 -1.08132410e+00 -2.41106197e-01 4.19809103e-01 6.77656949e-01 3.98131132e-01 -4.06772494e-01 -9.76797640e-02 8.71296883e-01 4.50732589e-01 3.98803651e-01 4.80975986e-01 -6.91513956e-01 1.89861014e-01 6.56130731e-01 -2.13200852e-01 -8.16961944e-01 -6.24499977e-01 -1.13639164e+00 -9.87524569e-01 -1.77350029e-01 3.97955418e-01 -1.23997323e-01 -8.59712303e-01 1.59740078e+00 6.41235709e-01 3.60647291e-01 -8.94103125e-02 9.71705794e-01 8.08120966e-01 1.02157325e-01 3.08142304e-01 -3.44956160e-01 1.45768940e+00 -9.23831403e-01 -5.27527928e-01 -5.59237041e-02 9.27731872e-01 -5.04718125e-01 8.45761001e-01 2.29701385e-01 -8.35642695e-01 -3.47889572e-01 -9.46037769e-01 1.58581689e-01 6.01896970e-03 4.45336282e-01 6.66325212e-01 7.33511627e-01 -7.37872303e-01 4.07784402e-01 -9.87484813e-01 1.27262324e-01 7.44309783e-01 5.29878438e-01 -1.64242923e-01 -2.03840479e-01 -1.08033824e+00 5.27055144e-01 5.06731033e-01 2.12599799e-01 -5.64797044e-01 -1.08248031e+00 -8.25880766e-01 -2.98622429e-01 6.07280433e-01 -8.02720249e-01 9.15290892e-01 -9.18828130e-01 -1.23817480e+00 7.81624079e-01 9.48650017e-02 -3.26877564e-01 6.26863956e-01 9.93167832e-02 -2.67628759e-01 2.87640274e-01 2.23464921e-01 7.70256162e-01 6.14270449e-01 -1.19747913e+00 -7.87194908e-01 -4.93501008e-01 -2.00521216e-01 4.81637716e-01 -5.48794031e-01 -6.32932723e-01 -2.80978531e-01 -8.89470518e-01 2.99870640e-01 -1.11751926e+00 -5.30897677e-01 1.55322375e-02 -3.49795073e-01 1.15764402e-01 3.11023265e-01 -7.78718650e-01 1.20605159e+00 -2.46416903e+00 1.77497864e-01 4.26945478e-01 4.26920831e-01 1.20786659e-01 1.73250601e-01 -7.49141946e-02 -5.62629625e-02 1.12319298e-01 -6.68850124e-01 -1.78597704e-01 -1.38375044e-01 2.89846379e-02 1.70229554e-01 7.03151286e-01 2.13079050e-01 9.49579775e-01 -1.27041674e+00 -7.33003616e-01 2.96581089e-01 5.05261898e-01 -6.20672762e-01 -8.75804275e-02 -2.60957330e-01 1.09742928e+00 -5.54127276e-01 9.00788903e-01 4.85332936e-01 -7.34149754e-01 3.23619843e-01 -3.79489034e-01 3.08623035e-02 6.96018189e-02 -7.36711502e-01 2.00280309e+00 -4.51848447e-01 9.80518609e-02 8.61197263e-02 -1.04745591e+00 5.59107721e-01 3.73814732e-01 9.69999969e-01 -6.19145572e-01 6.98803440e-02 6.56624019e-01 6.33748323e-02 -5.13464212e-01 3.78870480e-02 -1.69574782e-01 3.85604836e-02 9.40557271e-02 -5.57062775e-02 -1.39906749e-01 7.20091313e-02 9.42715779e-02 7.99496114e-01 2.34874696e-01 3.11024666e-01 -4.10506576e-01 5.11669755e-01 2.35304460e-02 6.47534847e-01 5.26330531e-01 -4.24136132e-01 4.61146355e-01 2.29400367e-01 1.04955822e-01 -4.78631914e-01 -9.23389316e-01 -5.85136414e-01 8.87640655e-01 4.50710297e-01 -1.84866175e-01 -4.39101994e-01 -9.71747577e-01 1.25412485e-02 3.69075686e-01 -5.71620524e-01 -1.47665232e-01 -4.22814876e-01 -1.35806394e+00 3.67175221e-01 4.09630805e-01 4.40237731e-01 -4.00346279e-01 -4.56611127e-01 2.24622756e-01 -4.25598919e-01 -9.43896055e-01 -5.16467631e-01 4.15522426e-01 -8.96662652e-01 -1.00726068e+00 -9.85877573e-01 -7.32969582e-01 8.33917737e-01 2.62348235e-01 8.12405586e-01 5.70048802e-02 -4.52917695e-01 3.80721152e-01 -4.20748800e-01 -3.25872034e-01 -3.03483635e-01 1.27081379e-01 -2.75851041e-01 8.47598463e-02 4.50363159e-02 -5.20379364e-01 -9.47778642e-01 3.51466268e-01 -7.30790734e-01 1.79414779e-01 8.28258991e-01 1.27467787e+00 8.59839976e-01 1.27194852e-01 6.17327213e-01 -9.28243816e-01 1.91239014e-01 -6.54689670e-01 -4.27548289e-01 3.17428410e-01 -7.71820426e-01 -1.35358395e-02 3.02400649e-01 -1.89795688e-01 -1.14483368e+00 2.75479585e-01 -5.42174317e-02 -8.28144997e-02 6.43483624e-02 8.54764223e-01 1.85216174e-01 -2.64956892e-01 7.86226511e-01 9.56622884e-02 2.96863019e-01 -8.47656280e-02 3.07687908e-01 5.74929118e-01 3.33457589e-01 -5.43957174e-01 4.12816972e-01 7.79142737e-01 -5.78842387e-02 -5.76229751e-01 -1.12661159e+00 -7.73916245e-01 -2.84824759e-01 -2.20498785e-01 8.01544964e-01 -1.14766681e+00 -4.29474324e-01 3.98086846e-01 -4.24716413e-01 -1.88683793e-01 -3.98032695e-01 6.79423511e-01 -3.14891547e-01 5.39076388e-01 -5.51017404e-01 -5.93630433e-01 -3.30506086e-01 -1.50823200e+00 1.15846419e+00 2.85335064e-01 2.99160872e-02 -1.14368486e+00 -7.68784583e-02 7.29103148e-01 2.29284361e-01 4.29572314e-01 6.83542550e-01 -5.68315506e-01 -6.14532292e-01 -1.12479612e-01 -2.95879334e-01 2.01971710e-01 3.87263834e-01 -3.28648597e-01 -9.51424658e-01 -4.08726841e-01 4.21969593e-02 -2.82575339e-01 1.09857738e+00 7.30927706e-01 1.26744044e+00 1.69490069e-01 -5.80137610e-01 8.04970801e-01 1.18310523e+00 -8.69536698e-02 -6.63628355e-02 1.56533197e-01 6.48757279e-01 6.57957077e-01 7.02849448e-01 6.26688361e-01 4.16394532e-01 4.97017086e-01 4.12231922e-01 -4.43234712e-01 -2.03005284e-01 -1.89399585e-01 5.08255884e-02 8.38887334e-01 7.86721110e-02 -1.60264432e-01 -8.94288063e-01 5.44340789e-01 -1.66531730e+00 -4.88136649e-01 -3.91909992e-03 2.07439351e+00 1.01974130e+00 -8.13971460e-02 1.10280206e-02 1.29497826e-01 5.51491916e-01 8.21265802e-02 -6.13479495e-01 4.59991455e-01 -1.25468388e-01 -1.29561469e-01 6.26485467e-01 2.91753918e-01 -1.41219020e+00 2.45013893e-01 4.95142460e+00 1.10774136e+00 -1.08023691e+00 3.26483905e-01 1.11596024e+00 -1.26918942e-01 -5.51759422e-01 -1.58402100e-01 -6.35097086e-01 6.00944340e-01 5.46844602e-01 2.22344816e-01 2.08585680e-01 5.05196691e-01 1.50335774e-01 -2.67216146e-01 -8.69846463e-01 7.95755029e-01 -1.73527412e-02 -1.23645556e+00 -5.54377019e-01 2.35036612e-01 1.02165520e+00 2.31144920e-01 4.06034946e-01 1.54890744e-02 1.39452487e-01 -8.49362254e-01 4.22952294e-01 3.36425304e-01 9.17090595e-01 -4.41538095e-01 8.32832575e-01 3.99003804e-01 -9.79047120e-01 -1.66682407e-01 3.12806457e-01 5.92859030e-01 2.04934195e-01 1.10317695e+00 -8.21423173e-01 9.47238684e-01 4.52315837e-01 9.95566070e-01 -4.65331107e-01 9.86415148e-01 -7.76904225e-02 5.80095828e-01 -4.84109193e-01 1.17817707e-01 2.01188162e-01 -2.50361383e-01 5.26213706e-01 1.12575245e+00 2.82894135e-01 2.45740533e-01 3.77930254e-01 6.57653809e-01 1.09206542e-01 2.24472433e-01 -9.71581340e-02 -7.69472793e-02 3.88228059e-01 1.28097606e+00 -7.94639349e-01 -8.04490149e-02 -4.00266856e-01 5.81433415e-01 -2.45734793e-03 2.56946802e-01 -9.79352236e-01 1.72525734e-01 -7.85804689e-02 1.10124955e-02 8.69402215e-02 2.09783465e-01 -5.50691843e-01 -1.16109228e+00 1.98212832e-01 -1.01484275e+00 6.43396735e-01 -1.46943748e-01 -1.26453233e+00 4.33110565e-01 -9.38002765e-02 -1.45003569e+00 2.09750220e-01 -4.26671028e-01 -1.30851775e-01 6.18487358e-01 -1.76679432e+00 -1.28588307e+00 -4.10637110e-01 2.89096445e-01 3.05935413e-01 9.76284407e-03 7.56319761e-01 5.82310140e-01 -6.78534627e-01 8.38222563e-01 3.53644639e-01 -1.48904100e-01 7.16791809e-01 -1.36323202e+00 -4.16612059e-01 5.11419177e-01 -7.81586990e-02 2.40676343e-01 3.92866254e-01 -5.30959666e-01 -9.94174898e-01 -1.07311261e+00 3.18190277e-01 -2.11581737e-01 6.27635539e-01 -6.19634707e-03 -7.77455151e-01 2.97980964e-01 -1.83653727e-01 3.19704920e-01 1.04631948e+00 2.11366545e-02 -1.30214347e-02 -9.05581117e-02 -1.13090312e+00 4.41909075e-01 7.82283127e-01 -3.67609471e-01 -7.15072230e-02 7.14041173e-01 5.68209946e-01 -6.80484176e-01 -1.21702480e+00 9.70291257e-01 3.83853376e-01 -9.27060962e-01 9.46900308e-01 5.13748340e-02 2.68619776e-01 -3.08938682e-01 -1.87186003e-01 -1.18821812e+00 -1.61672249e-01 -3.91910106e-01 4.26628888e-02 9.58203793e-01 6.33396149e-01 -7.49941945e-01 8.63581240e-01 4.55986768e-01 -5.03222466e-01 -1.32122684e+00 -8.72117400e-01 -4.07888621e-01 -5.55306338e-02 -2.27238938e-01 2.34893173e-01 1.01547861e+00 2.14640856e-01 -1.55080348e-01 -1.93163738e-01 3.49324077e-01 6.65140629e-01 1.08623587e-01 1.72923774e-01 -8.90750706e-01 -7.16113627e-01 -5.20839751e-01 -1.52007192e-01 -1.02399027e+00 -1.20548502e-01 -1.24810505e+00 1.14367679e-01 -1.20154989e+00 3.61317396e-01 -9.87390339e-01 -5.87886751e-01 4.25017446e-01 -5.46485066e-01 2.93610334e-01 -8.07340145e-02 3.31159115e-01 -6.73041523e-01 5.93011618e-01 1.39385664e+00 -2.20360070e-01 -2.17086166e-01 1.63428679e-01 -7.92225063e-01 6.55319571e-01 6.03285432e-01 -3.95469725e-01 -4.85008180e-01 -1.96946055e-01 1.52535394e-01 1.37327194e-01 4.77111518e-01 -8.34595501e-01 2.15407200e-02 -3.01143937e-02 2.71056771e-01 -4.18610245e-01 1.44220367e-01 -8.04595292e-01 1.76316336e-01 5.36852479e-01 -3.85961860e-01 -5.46837151e-01 4.36795428e-02 9.04224217e-01 -3.51801932e-01 -1.54626161e-01 8.16872001e-01 -1.24487147e-01 -3.68146926e-01 3.28598917e-01 1.62574306e-01 1.11671492e-01 1.15869904e+00 -1.40019745e-01 -1.44004479e-01 -4.53375420e-03 -8.07978034e-01 6.61684275e-01 2.78012037e-01 1.09619433e-02 2.86652386e-01 -9.24141765e-01 -7.40265548e-01 2.09812775e-01 1.43732086e-01 3.97964239e-01 6.63875043e-01 1.69007170e+00 -2.52858222e-01 3.31985861e-01 4.28357333e-01 -9.91526127e-01 -8.65750790e-01 3.56390476e-01 5.54908931e-01 -5.48853040e-01 -5.28046429e-01 1.10019970e+00 3.82691741e-01 -6.04019225e-01 1.27786756e-01 -1.30537167e-01 9.21053588e-02 2.31672712e-02 1.05533898e-01 2.67676383e-01 2.73299128e-01 -4.20848191e-01 -3.78231168e-01 3.73173952e-01 -2.09699303e-01 1.58978581e-01 1.12711751e+00 -1.61706924e-01 3.38234380e-02 1.66701928e-01 1.01871371e+00 8.20590854e-02 -1.43946362e+00 -4.06185031e-01 -2.64638644e-02 -3.29479128e-01 3.25043261e-01 -1.02944255e+00 -1.13734102e+00 5.55810392e-01 8.87590170e-01 -1.54987331e-02 1.34659410e+00 -1.02263568e-02 8.39301169e-01 -2.12205976e-01 1.04922898e-01 -9.82584774e-01 8.22461322e-02 -1.57156840e-01 4.35985714e-01 -1.62974048e+00 1.83437571e-01 -7.71007240e-01 -6.79716110e-01 6.24117792e-01 2.65352160e-01 1.98765725e-01 7.77252853e-01 1.62166342e-01 1.75568238e-01 -2.19985873e-01 -4.17787194e-01 -9.89024788e-02 4.73777592e-01 2.93999493e-01 5.44657350e-01 2.73977280e-01 -3.36384654e-01 5.53865194e-01 2.11982250e-01 -5.24481758e-02 -7.88329840e-02 9.07248199e-01 -2.50271440e-01 -1.08861804e+00 -2.10486904e-01 6.47401035e-01 -5.87624252e-01 -2.40212664e-01 2.46859968e-01 6.56929970e-01 3.30899984e-01 7.57778049e-01 -4.25851524e-01 -1.58042118e-01 5.38554266e-02 -2.07002759e-01 2.78136760e-01 -4.46300030e-01 -5.94678164e-01 5.37447095e-01 -1.54286966e-01 -2.80920863e-01 -7.28689730e-01 -7.46487975e-01 -1.22037256e+00 3.20493996e-01 -8.13044131e-01 1.99666917e-01 5.96007884e-01 9.80837822e-01 4.30982798e-01 7.13815987e-01 7.97847986e-01 -5.69953144e-01 -8.71217251e-01 -6.01194203e-01 -4.47538108e-01 2.17795894e-01 5.06862402e-01 -6.36200666e-01 -5.43676019e-01 -7.48643130e-02]

[14.726409912109375, -2.0860626697540283]

e471a7d8-c28e-4dc6-a8d2-47d627b509cc

worst-case-performance-of-greedy-policies-in

2204.04773

https://arxiv.org/abs/2204.04773v2

https://arxiv.org/pdf/2204.04773v2.pdf

Worst-case Performance of Greedy Policies in Bandits with Imperfect Context Observations

Contextual bandits are canonical models for sequential decision-making under uncertainty in environments with time-varying components. In this setting, the expected reward of each bandit arm consists of the inner product of an unknown parameter with the context vector of that arm. The classical bandit settings heavily rely on assuming that the contexts are fully observed, while study of the richer model of imperfectly observed contextual bandits is immature. This work considers Greedy reinforcement learning policies that take actions as if the current estimates of the parameter and of the unobserved contexts coincide with the corresponding true values. We establish that the non-asymptotic worst-case regret grows poly-logarithmically with the time horizon and the failure probability, while it scales linearly with the number of arms. Numerical analysis showcasing the above efficiency of Greedy policies is also provided.

['Mohamad Kazem Shirani Faradonbeh', 'Hongju Park']

2022-04-10

['decision-making-under-uncertainty', 'decision-making-under-uncertainty']

['medical', 'reasoning']

[ 1.63880199e-01 2.71851987e-01 -9.22106326e-01 -1.84724092e-01 -9.13252473e-01 -7.93442130e-01 3.47882897e-01 1.98843047e-01 -5.67485869e-01 1.21243227e+00 2.35314533e-01 -6.25832260e-01 -5.04417539e-01 -6.92126393e-01 -1.16138291e+00 -1.06027496e+00 -2.61677027e-01 7.70056307e-01 -1.40678898e-01 2.98465550e-01 1.83082759e-01 1.15422495e-01 -9.47493553e-01 6.05348423e-02 4.91287082e-01 1.43255889e+00 5.27243502e-02 1.00429595e+00 -4.56191488e-02 8.52853537e-01 -4.57092434e-01 -4.31879401e-01 5.86854815e-01 -4.14965093e-01 -4.84842718e-01 3.51736248e-01 -1.83912441e-01 -8.89423192e-01 -6.13249719e-01 1.11394668e+00 4.76875789e-02 9.75326598e-02 3.05592448e-01 -1.00129366e+00 -3.90798509e-01 1.02630258e+00 -5.91332316e-01 3.43168020e-01 1.07139789e-01 1.16906889e-01 1.18414879e+00 1.80485860e-01 3.39637399e-01 1.11217737e+00 1.88656390e-01 3.07622343e-01 -1.24407578e+00 -4.25509840e-01 6.88797653e-01 4.53952044e-01 -6.34855568e-01 -2.49111310e-01 3.13648760e-01 -1.76107138e-01 5.90543628e-01 2.10423127e-01 8.46116960e-01 1.02790153e+00 6.92217499e-02 1.12416363e+00 1.18891048e+00 -6.65091515e-01 9.01889622e-01 1.02299089e-02 1.76161945e-01 3.54054689e-01 4.99159694e-01 7.16787755e-01 -4.00567502e-01 -3.45288545e-01 7.71622539e-01 5.46817482e-01 2.90596783e-02 -4.21585500e-01 -7.81017423e-01 7.97384381e-01 9.75871682e-02 -3.31416018e-02 -9.43922102e-01 7.31017113e-01 -1.59047861e-02 7.09361136e-01 3.01582247e-01 2.72493586e-02 -2.97511250e-01 -2.70567030e-01 -7.25573957e-01 3.40926141e-01 7.71093130e-01 1.22178137e+00 5.92102766e-01 -1.40290027e-02 -6.61745012e-01 2.67913669e-01 -7.68019781e-02 8.01442146e-01 1.56807765e-01 -1.19265521e+00 9.72211719e-01 -6.06390312e-02 1.27348304e+00 -2.55039126e-01 -1.84198454e-01 -6.54678464e-01 -1.62623003e-01 4.88607166e-03 7.80316830e-01 -6.01815879e-01 -9.30645525e-01 1.74316764e+00 4.04617786e-01 1.28696576e-01 9.31542218e-02 7.05104291e-01 -5.09769738e-01 4.41051543e-01 -1.25023842e-01 -7.01910555e-01 9.08601522e-01 -8.04604530e-01 -9.06141877e-01 -5.99591851e-01 1.88110694e-01 -3.92309010e-01 3.51376921e-01 4.71254766e-01 -1.24739158e+00 1.97018296e-01 -9.72316861e-01 6.87249899e-01 2.71859378e-01 -2.00733885e-01 7.46508837e-01 8.61794651e-01 -5.76794982e-01 5.18628538e-01 -8.66758227e-01 2.22947896e-01 4.64394838e-01 1.42708331e-01 1.63822874e-01 -2.42932826e-01 -6.09182119e-01 5.51499963e-01 4.74014521e-01 1.54773161e-01 -1.47426128e+00 -3.74442041e-01 -1.21113196e-01 1.82145238e-01 1.21111190e+00 -4.60105717e-01 1.95489001e+00 -1.18122005e+00 -1.45732987e+00 6.63303137e-02 -1.83526561e-01 -1.02587497e+00 9.34709013e-01 -3.39748323e-01 7.35951513e-02 7.35124648e-02 5.87907992e-02 -3.15137804e-01 9.91418242e-01 -8.14199865e-01 -1.19719303e+00 -6.71065748e-01 5.89045525e-01 3.70216608e-01 2.13157862e-01 -6.76368177e-01 5.98407388e-02 -3.21741372e-01 8.86874869e-02 -1.18377364e+00 -6.30008340e-01 -2.62871206e-01 -4.91723031e-01 1.40651554e-01 5.04661873e-02 -3.02625060e-01 9.20504987e-01 -2.08486819e+00 -3.12501669e-01 2.53466338e-01 -3.26976299e-01 -3.61949652e-01 1.72279745e-01 6.70394778e-01 2.09886134e-01 -8.25166553e-02 1.18160307e-01 -1.05465375e-01 2.98916131e-01 3.87524545e-01 -7.65133858e-01 7.67993808e-01 -5.49456418e-01 6.69463933e-01 -8.25813651e-01 6.97960407e-02 3.63135524e-02 -6.09779119e-01 -3.53713661e-01 1.32092506e-01 -7.30612695e-01 2.16095045e-01 -9.79802012e-01 3.39209616e-01 3.95370424e-01 -4.35245454e-01 6.37578845e-01 6.34263933e-01 9.50597078e-02 1.80755451e-01 -1.41389990e+00 1.15603292e+00 -3.96022946e-01 -6.82774633e-02 1.14634097e-01 -1.12192905e+00 2.23144237e-02 5.28020561e-01 3.34770888e-01 -6.79634213e-01 8.16027895e-02 1.00298718e-01 -1.70891836e-01 -3.96168411e-01 2.24256352e-01 -4.81534392e-01 -8.54590237e-02 7.03771710e-01 -1.53402790e-01 2.95215160e-01 8.53675529e-02 4.84247273e-03 1.12337005e+00 -5.34314997e-02 5.29629886e-01 1.71088293e-01 -2.32980892e-01 1.66849457e-02 6.32489562e-01 1.64356172e+00 -2.25359052e-01 -6.72301352e-02 8.44878495e-01 -4.25759703e-01 -1.00823808e+00 -8.84763122e-01 9.63547900e-02 1.16771626e+00 1.56508073e-01 3.23069245e-01 -5.81920266e-01 -5.05437553e-01 4.30144757e-01 1.04351246e+00 -1.02709067e+00 3.62648487e-01 1.03457849e-02 -7.01089084e-01 -2.53100544e-01 4.32659566e-01 3.39984059e-01 -7.38363206e-01 -7.74065137e-01 7.53351808e-01 8.08773786e-02 -1.13139546e+00 -3.50247949e-01 3.48326474e-01 -1.00421703e+00 -1.14040363e+00 -5.03264308e-01 9.19001475e-02 4.11262274e-01 3.04220259e-01 6.05967164e-01 -3.34088355e-01 2.21995011e-01 7.78513253e-01 -2.62076914e-01 -7.30619073e-01 1.84555575e-02 -4.10659879e-01 -2.35082254e-01 2.66932398e-01 2.74082851e-02 -2.94826150e-01 -9.47630346e-01 4.09396328e-02 -7.92947531e-01 -9.38948914e-02 4.65484709e-01 9.74875867e-01 7.18940914e-01 -3.79095674e-02 4.62682068e-01 -1.03897429e+00 3.71959537e-01 -6.11812353e-01 -1.03801835e+00 6.12904787e-01 -4.18119758e-01 1.82187825e-01 3.10606152e-01 -5.73283434e-01 -1.38088799e+00 -2.26297770e-02 5.42488217e-01 -1.53012648e-01 -1.86431613e-02 3.38490754e-01 3.23829174e-01 4.69868541e-01 3.04097265e-01 1.63478389e-01 -3.29942763e-01 -3.46210748e-01 3.03674549e-01 4.99456823e-01 2.78347492e-01 -9.59721804e-01 2.12552890e-01 6.69747770e-01 2.05611676e-01 -3.50904614e-01 -1.30026317e+00 -2.36829430e-01 3.00297588e-01 2.76944377e-02 3.05229038e-01 -9.77101386e-01 -1.00790083e+00 1.19546324e-01 -7.01435804e-01 -6.19998395e-01 -7.13182628e-01 7.49115348e-01 -1.24335372e+00 1.38805164e-02 -3.97267967e-01 -1.76369011e+00 1.21224567e-01 -9.62591052e-01 6.32139325e-01 2.55613625e-01 3.60857546e-01 -6.77872956e-01 -1.37515470e-01 3.83448929e-01 2.26063102e-01 4.27573025e-02 8.58854592e-01 -6.51754260e-01 -1.00684631e+00 -4.80302066e-01 1.35691226e-01 -5.10458611e-02 -1.78272799e-01 -5.64377487e-01 -7.23341882e-01 -3.07714224e-01 1.34638861e-01 -1.71939924e-01 4.90015179e-01 1.05373621e+00 1.08714104e+00 -9.25997734e-01 -3.09967309e-01 3.41249630e-02 1.60320675e+00 8.17967296e-01 2.53904939e-01 4.68775213e-01 -1.63516864e-01 2.66592950e-01 8.85398746e-01 1.17778039e+00 -2.05068681e-02 4.47151750e-01 7.74758995e-01 6.21588051e-01 7.73514926e-01 -4.69987810e-01 1.40917778e-01 -4.29537743e-01 2.03466266e-02 -3.04963231e-01 -3.63866746e-01 8.68304729e-01 -2.33498502e+00 -1.12942481e+00 4.57692415e-01 2.68032289e+00 6.21813536e-01 3.04211736e-01 3.41611743e-01 -4.38559577e-02 7.40838051e-01 -1.05673686e-01 -1.28956509e+00 -2.84587979e-01 5.72371893e-02 -9.57624912e-02 1.20636785e+00 7.12384284e-01 -7.30615318e-01 5.75074375e-01 6.71380186e+00 7.14186132e-01 -5.53240716e-01 2.92303592e-01 9.65057909e-01 -7.36194074e-01 -1.22438341e-01 1.94833845e-01 -7.44572997e-01 6.17677152e-01 1.03728366e+00 -3.11106563e-01 9.05607283e-01 9.81982648e-01 1.41670778e-01 -6.30166292e-01 -1.26843417e+00 7.31924295e-01 -9.18992043e-01 -1.27082658e+00 -6.25102162e-01 3.56098562e-01 1.01117420e+00 4.75065522e-02 2.25003123e-01 2.08861530e-01 1.18742049e+00 -5.86603045e-01 1.18354380e+00 5.84830999e-01 4.71704155e-01 -9.54947293e-01 6.07137501e-01 6.35268927e-01 -4.49718118e-01 -1.01347435e+00 -2.75163233e-01 -4.63504046e-01 1.20633394e-02 5.24523973e-01 -9.66892362e-01 3.21565062e-01 4.51921612e-01 1.66595623e-01 3.22983146e-01 1.39472175e+00 -3.78476024e-01 8.72585952e-01 -6.22969925e-01 -2.16074154e-01 6.35330319e-01 -3.91397804e-01 4.31736678e-01 6.00301325e-01 3.07743549e-01 2.45982662e-01 3.53266835e-01 2.64874518e-01 8.09705630e-02 -8.57562199e-02 -2.83936977e-01 -1.89499676e-01 7.29843020e-01 4.44669664e-01 -4.03391868e-01 -6.20392799e-01 -2.49965042e-01 5.76512575e-01 2.51070261e-01 7.32969820e-01 -5.63160717e-01 2.90286601e-01 6.72835827e-01 -3.08506578e-01 8.51489663e-01 4.60681915e-02 -4.50437278e-01 -7.15966880e-01 2.86929816e-01 -6.06934547e-01 6.42746806e-01 -3.06536943e-01 -1.19606805e+00 -1.64497092e-01 6.45454004e-02 -7.62907207e-01 -5.26571989e-01 -2.02978998e-01 -2.73889393e-01 6.23558223e-01 -1.37133753e+00 -5.21950245e-01 6.35511100e-01 4.98590708e-01 4.16326582e-01 2.58107223e-02 5.96764803e-01 -6.22792661e-01 -6.33324564e-01 2.28166133e-01 1.03220654e+00 -3.34101170e-01 -1.68564618e-02 -1.31137025e+00 -3.55425999e-02 6.77865207e-01 -9.96307507e-02 2.86452502e-01 1.23113978e+00 -5.02701819e-01 -1.55814648e+00 -7.10054934e-01 1.31109044e-01 2.20223628e-02 9.65653419e-01 9.28969756e-02 -3.08055222e-01 1.12557364e+00 1.34146318e-01 2.18115449e-02 5.04582167e-01 3.15231591e-01 -1.92290127e-01 -3.50885391e-01 -1.22278416e+00 4.75734353e-01 8.89060974e-01 -3.65268558e-01 -1.99855387e-01 6.04460061e-01 6.20531201e-01 -4.97709334e-01 -5.58659792e-01 -1.04664788e-01 7.11049438e-01 -9.53822255e-01 5.75529516e-01 -1.06326699e+00 8.36920273e-03 2.58401453e-01 -4.22336996e-01 -1.31621158e+00 -1.67625651e-01 -9.93912458e-01 -3.61630708e-01 6.16946399e-01 4.22036976e-01 -7.17387617e-01 1.12209475e+00 8.57348740e-01 4.37608212e-01 -5.14962852e-01 -1.59920299e+00 -7.79490471e-01 -3.24159227e-02 -5.38575232e-01 8.33152950e-01 3.13599020e-01 1.51959389e-01 -9.87285301e-02 -7.07299590e-01 2.60008663e-01 8.64646077e-01 5.60976326e-01 5.09228945e-01 -6.54127002e-01 -9.33239162e-01 -1.03382841e-01 1.20950103e-01 -1.32569027e+00 -5.71060181e-02 4.51192679e-03 2.15131477e-01 -1.08821595e+00 3.78399491e-01 -5.26759744e-01 -5.69395602e-01 3.13314021e-01 -8.18218291e-02 -6.84280872e-01 2.74592102e-01 -6.64596334e-02 -8.82265508e-01 3.82362545e-01 1.26471686e+00 -1.93827674e-01 -1.88308567e-01 8.53919029e-01 -8.24479103e-01 3.76488745e-01 7.56140709e-01 -6.66493177e-01 -5.73329329e-01 -2.96784639e-01 3.07274103e-01 1.15823400e+00 9.96331945e-02 -4.61178482e-01 1.08307347e-01 -9.28578138e-01 1.60993859e-01 -7.03082860e-01 4.33650792e-01 -1.21395600e+00 3.12712461e-01 5.52585423e-01 -8.37853968e-01 -1.43072575e-01 -1.77410454e-01 1.65930247e+00 2.48650104e-01 -5.87440372e-01 4.37064320e-01 -4.34316039e-01 -7.30065107e-02 3.47320795e-01 -5.09407163e-01 -2.59266049e-02 1.22277629e+00 5.38676009e-02 -1.77733421e-01 -1.02328849e+00 -1.00998485e+00 2.84650952e-01 1.30598038e-01 -1.30264863e-01 2.06442371e-01 -1.00228381e+00 -3.57154340e-01 -2.58135527e-01 -3.00412565e-01 -1.83477238e-01 5.73815286e-01 5.33739507e-01 2.10765123e-01 6.63658738e-01 8.04815888e-02 -1.94361851e-01 -8.65977764e-01 8.95749331e-01 3.99833530e-01 -5.35906613e-01 -1.92819700e-01 4.92576748e-01 1.39424294e-01 4.15647745e-01 6.17703259e-01 -2.78869748e-01 3.40505272e-01 -1.02430969e-01 8.82768929e-01 5.10014951e-01 -9.97506455e-03 2.20072493e-01 2.83030480e-01 -2.01752365e-01 -3.65290105e-01 -5.72246253e-01 1.38265884e+00 -4.11627561e-01 3.37147593e-01 5.48190951e-01 5.85593700e-01 -3.55545312e-01 -1.84971392e+00 -9.24561679e-01 2.33447611e-01 -7.66238391e-01 1.89320967e-01 -9.46590722e-01 -7.33545542e-01 4.59793925e-01 6.93952501e-01 6.20171428e-01 8.12921822e-01 -1.22212850e-01 1.49277568e-01 4.54586506e-01 9.52593148e-01 -1.22034395e+00 -3.06357086e-01 2.15597972e-01 6.00899398e-01 -9.53864098e-01 -1.73409116e-02 7.25611150e-02 -7.18127728e-01 7.24353433e-01 1.44129735e-03 -2.09327355e-01 5.15929163e-01 -3.88704869e-03 -4.23472136e-01 2.82467216e-01 -1.12525856e+00 -3.60945612e-01 -2.65623540e-01 4.70107615e-01 -2.79917419e-01 8.06107879e-01 -2.46202260e-01 5.38771331e-01 -6.16185542e-04 2.86551595e-01 6.43953025e-01 1.18354821e+00 -7.23433316e-01 -8.80757689e-01 -7.27335930e-01 6.11427426e-01 -1.03219032e+00 1.52860671e-01 1.27215356e-01 3.36172879e-01 -4.79157239e-01 1.19839942e+00 9.78205819e-03 2.95514911e-01 2.80335009e-01 2.06862539e-02 8.16434741e-01 -4.06454414e-01 -2.34497059e-02 2.98302501e-01 2.52593100e-01 -5.61209023e-01 -1.78764313e-01 -9.05927122e-01 -6.47414327e-01 -2.83277154e-01 -2.46922523e-01 2.82355279e-01 7.08320618e-01 1.37252426e+00 4.37589996e-02 7.13109598e-02 8.68881762e-01 -4.58114058e-01 -1.55514598e+00 -8.76363218e-01 -9.07857358e-01 3.95538397e-02 9.29227293e-01 -6.03469908e-01 -2.47247607e-01 -4.00703698e-01]

[4.465785503387451, 3.194587469100952]

e8d3e276-f7a2-48c9-b319-ef7759e37de7

using-a-frustratingly-easy-domain-and-tagset

https://aclanthology.org/2021.bsnlp-1.12

https://aclanthology.org/2021.bsnlp-1.12.pdf

Using a Frustratingly Easy Domain and Tagset Adaptation for Creating Slavic Named Entity Recognition Systems

We present a collection of Named Entity Recognition (NER) systems for six Slavic languages: Bulgarian, Czech, Polish, Slovenian, Russian and Ukrainian. These NER systems have been trained using different BERT models and a Frustratingly Easy Domain Adaptation (FEDA). FEDA allow us creating NER systems using multiple datasets without having to worry about whether the tagset (e.g. Location, Event, Miscellaneous, Time) in the source and target domains match, while increasing the amount of data available for training. Moreover, we boosted the prediction on named entities by marking uppercase words and predicting masked words. Participating in the 3rd Shared Task on SlavNER, our NER systems reached a strict match micro F-score of up to 0.908. The results demonstrate good generalization, even in named entities with weak regularity, such as book titles, or entities that were never seen during the training.

['Antoine Doucet', 'Jose G. Moreno', 'Luis Adrián Cabrera-Diego']

eacl-bsnlp-2021-4

['miscellaneous']

['miscellaneous']

[-5.77038944e-01 5.81886480e-03 1.29445434e-01 -3.60395044e-01 -5.31372607e-01 -1.18489909e+00 9.23816264e-01 4.33793932e-01 -1.02513134e+00 1.15524054e+00 3.80586565e-01 -4.12869155e-01 -7.01691210e-03 -9.00630355e-01 -3.77295971e-01 -1.96931884e-01 5.36223780e-03 7.22489595e-01 3.13607812e-01 -2.63885766e-01 1.02486014e-01 8.19144547e-01 -1.16671479e+00 1.35318443e-01 7.78085351e-01 4.38352764e-01 1.75708845e-01 4.65298086e-01 -4.17977750e-01 6.74725473e-01 -8.84337127e-01 -8.72927666e-01 1.96188852e-01 -6.04741368e-03 -1.11725652e+00 -5.67868769e-01 3.09018761e-01 3.31374794e-01 -3.79623264e-01 9.06300008e-01 5.35258472e-01 2.13793218e-01 6.57272041e-01 -7.93412328e-01 -6.72805607e-01 8.26595008e-01 8.20188224e-03 2.65110254e-01 2.93190390e-01 -1.09473005e-01 1.04689300e+00 -1.03196728e+00 1.26936853e+00 6.50992990e-01 9.98756886e-01 7.19550371e-01 -9.41733539e-01 -7.07425475e-01 -1.78295687e-01 3.12152989e-02 -1.62151909e+00 -4.28944379e-01 -1.53730094e-01 -4.35007542e-01 1.16359985e+00 2.13417172e-01 8.88589323e-02 1.07119453e+00 -5.15948124e-02 3.44180405e-01 9.96324658e-01 -5.41382551e-01 2.87655145e-01 5.52547753e-01 1.07046612e-01 4.01043057e-01 3.89675677e-01 -5.21369129e-02 -1.48198634e-01 -3.04325998e-01 5.19481361e-01 -5.29068932e-02 -2.41428427e-03 4.07627933e-02 -1.30975413e+00 4.87622052e-01 1.86410025e-02 9.89005864e-01 -5.82499802e-01 -5.78447104e-01 5.48870385e-01 2.23590612e-01 1.31252453e-01 8.72589409e-01 -1.29108691e+00 -2.91554004e-01 -1.16805756e+00 -1.03196755e-01 1.21600676e+00 1.25916660e+00 7.04567373e-01 -9.02645290e-02 -1.56468377e-01 9.52533662e-01 -1.02897562e-01 4.04705405e-01 7.51275003e-01 -3.13585848e-01 5.06835759e-01 5.99283218e-01 4.92639244e-01 -5.05568385e-01 -6.73233509e-01 -1.57944947e-01 -5.90265989e-01 -1.24767207e-01 6.94192767e-01 -7.26918936e-01 -1.23202503e+00 1.66255128e+00 2.31847644e-01 1.04989089e-01 4.38870221e-01 4.66318607e-01 9.57140923e-01 6.01621389e-01 6.64174020e-01 8.33462030e-02 1.61410904e+00 -6.65807664e-01 -6.16401613e-01 -3.09074167e-02 8.80511403e-01 -9.28900361e-01 5.15098870e-01 1.01072937e-01 -7.90837228e-01 -5.40618956e-01 -4.93640363e-01 2.45219916e-02 -1.15455437e+00 3.35086077e-01 5.82080901e-01 8.36018384e-01 -7.93791950e-01 8.19086909e-01 -5.63505650e-01 -6.42373562e-01 -6.67310283e-02 3.01883042e-01 -8.89675021e-01 9.81296003e-02 -1.64798427e+00 1.14193821e+00 8.41789484e-01 -2.64537722e-01 -4.64092910e-01 -8.51531982e-01 -8.36342752e-01 8.78536403e-02 4.89839613e-02 -2.52690941e-01 9.90276039e-01 -4.79346335e-01 -1.10127616e+00 1.28883779e+00 -3.82744707e-02 -4.94943947e-01 4.07889068e-01 -1.76325873e-01 -1.36020339e+00 -3.46073896e-01 6.04834855e-01 3.51387560e-01 5.40099554e-02 -7.33870566e-01 -9.73613799e-01 -2.02228859e-01 -1.58538342e-01 -2.59313524e-01 -1.46130532e-01 5.18503010e-01 -6.85909167e-02 -6.17522895e-01 -2.51156300e-01 -6.25352621e-01 -4.05384511e-01 -9.67934310e-01 -4.65066582e-01 -4.42931592e-01 2.65576869e-01 -9.30143535e-01 1.36905050e+00 -2.32989717e+00 -4.01409656e-01 2.78433383e-01 -6.98731020e-02 4.61614937e-01 -2.25775316e-02 8.91055405e-01 -5.08964121e-01 5.07315516e-01 1.43655732e-01 1.28736019e-01 1.33600920e-01 1.53161705e-01 -3.56782824e-01 3.01398814e-01 3.34200293e-01 6.47695482e-01 -1.02777791e+00 -5.29792488e-01 3.49673280e-03 1.68458447e-01 -2.51715519e-02 8.08826908e-02 2.52148807e-01 1.38244703e-01 -2.26902619e-01 4.65391338e-01 5.84326267e-01 -3.64632010e-02 2.90553510e-01 1.70667157e-01 -6.89903677e-01 7.54026413e-01 -1.50189614e+00 1.40382934e+00 -6.68132365e-01 4.68824565e-01 -1.01655565e-01 -4.27670032e-01 9.96884227e-01 6.72479391e-01 3.61013561e-01 -6.10685587e-01 -1.22585811e-01 4.04185504e-01 -2.26837829e-01 -3.67163032e-01 9.15848553e-01 -2.37299412e-01 -4.55606371e-01 1.20214686e-01 6.24064624e-01 3.23531419e-01 4.69548166e-01 9.74076837e-02 1.23897862e+00 -6.96388111e-02 5.91074824e-01 -2.64973253e-01 4.60159391e-01 3.13029259e-01 9.17696118e-01 6.49904847e-01 -2.63286054e-01 5.33900082e-01 3.28266591e-01 -3.12696368e-01 -1.37222767e+00 -9.60190475e-01 -4.64726835e-01 1.15091002e+00 -3.42655689e-01 -5.77124238e-01 -5.36533833e-01 -1.00480783e+00 -2.89964765e-01 9.23215806e-01 -4.34323460e-01 2.49051630e-01 -7.30230927e-01 -3.06176692e-01 1.24684596e+00 4.24642116e-01 3.98009866e-01 -1.32629550e+00 -8.12315792e-02 4.50817734e-01 -9.37486514e-02 -1.22418225e+00 -4.32086289e-01 6.63790345e-01 -5.76874375e-01 -7.63975620e-01 -9.51635599e-01 -1.05425918e+00 2.99272388e-01 -4.39521551e-01 1.34932780e+00 -4.03666705e-01 -1.51326075e-01 1.08193822e-01 -3.51519406e-01 -4.56438214e-01 -3.00738841e-01 5.01076818e-01 1.28483176e-01 -5.17296195e-01 7.56566882e-01 -3.22310388e-01 -1.59761906e-01 4.89923626e-01 -7.53918469e-01 -6.86118066e-01 5.95337987e-01 8.28550816e-01 2.56228983e-01 -1.48409174e-03 6.64076805e-01 -1.29321766e+00 2.31756866e-01 -7.51990736e-01 -4.96321112e-01 4.26460654e-01 -5.48889697e-01 2.69847736e-02 8.52723122e-01 -3.77882153e-01 -1.25575054e+00 2.68835872e-01 -4.33995754e-01 2.25193769e-01 -9.68590796e-01 3.40832204e-01 -3.69280308e-01 4.24000233e-01 8.69312704e-01 2.48494633e-02 -9.88895535e-01 -7.31730521e-01 4.29764986e-01 7.95478106e-01 6.85678840e-01 -4.42010850e-01 7.35674500e-01 4.90495563e-02 -4.24990296e-01 -8.97199512e-01 -5.45769036e-01 -1.01874804e+00 -8.76499474e-01 2.56848454e-01 9.40081716e-01 -1.05839205e+00 -2.34437153e-01 3.03427994e-01 -1.28458345e+00 2.70598214e-02 -4.70565915e-01 6.93293214e-01 -8.45113918e-02 2.72578090e-01 -7.45250762e-01 -6.92197740e-01 -1.63997054e-01 -2.88041234e-01 5.32009423e-01 6.12110198e-01 -5.72028756e-01 -1.25209439e+00 4.16281641e-01 -3.91390920e-02 1.35070905e-01 -3.33760702e-03 8.37992072e-01 -1.74791980e+00 1.67684928e-01 -4.48800713e-01 -1.95651457e-01 1.46205738e-01 3.69983204e-02 -1.20944090e-01 -8.38800788e-01 1.23351581e-01 -3.86959970e-01 1.36814803e-01 5.87522209e-01 -2.29562953e-01 3.78029406e-01 -2.74246573e-01 -4.88537669e-01 2.78638482e-01 1.31165445e+00 4.30516034e-01 8.30645561e-01 6.29552126e-01 3.34161729e-01 5.54059684e-01 5.33403575e-01 3.27048481e-01 1.99122995e-01 4.11191642e-01 -2.95961171e-01 -1.15295529e-01 -8.52973983e-02 -4.40598905e-01 2.96771675e-01 4.86618251e-01 -2.20338836e-01 -2.32054561e-01 -1.28107941e+00 1.15344059e+00 -1.40459859e+00 -1.18328655e+00 -4.03426915e-01 2.29268026e+00 9.47488070e-01 2.13977825e-02 -4.56278548e-02 -3.35382193e-01 1.01660812e+00 -1.32483318e-01 2.67186295e-02 -4.03461903e-01 -4.96445656e-01 8.70075524e-01 9.71362472e-01 1.89605325e-01 -1.29563272e+00 1.33329439e+00 6.28388166e+00 9.16626632e-01 -8.99824023e-01 1.62788689e-01 4.75019924e-02 3.69808853e-01 -6.66440278e-02 1.84878055e-02 -1.18345070e+00 4.17933673e-01 1.52334750e+00 -1.68542758e-01 9.29307267e-02 9.39657569e-01 -6.68014809e-02 1.56953201e-01 -8.01608860e-01 6.21292055e-01 -2.73588866e-01 -1.13327420e+00 -3.13061118e-01 9.75688845e-02 8.41146827e-01 4.28849459e-01 -5.23943245e-01 6.81604922e-01 7.78568447e-01 -8.98915410e-01 6.12002850e-01 5.28325081e-01 6.34882450e-01 -6.88842952e-01 1.00526571e+00 3.11149955e-01 -9.39698040e-01 1.64386421e-01 -4.54886526e-01 2.20231876e-01 1.68572590e-01 7.06615388e-01 -1.08257365e+00 8.02251816e-01 6.46338224e-01 2.29076281e-01 -4.87865239e-01 1.30134439e+00 -4.45530385e-01 8.36992025e-01 -4.30477291e-01 -2.02393740e-01 3.29765081e-01 1.46205530e-01 5.12407541e-01 1.72009182e+00 4.62072164e-01 1.31003290e-01 -1.68497145e-01 4.23616320e-01 -3.29334766e-01 5.48307836e-01 -5.25002301e-01 -2.72179514e-01 4.96318936e-01 1.42978513e+00 -7.55063653e-01 -3.49012226e-01 -4.40920472e-01 1.12365687e+00 3.76236588e-01 2.43494958e-01 -6.25583231e-01 -1.17172623e+00 5.13452888e-01 9.17451903e-02 5.58351874e-01 -1.89247444e-01 -2.75220782e-01 -1.34417081e+00 -2.49449626e-01 -5.02687275e-01 7.44391501e-01 -5.59882164e-01 -1.56991756e+00 7.47289062e-01 -3.38266343e-01 -9.56645906e-01 -3.30756754e-01 -8.50031614e-01 -5.45965135e-01 1.07147956e+00 -1.13381183e+00 -8.43219101e-01 3.57662350e-01 5.16370535e-01 -3.50502543e-02 -3.55118632e-01 1.14699137e+00 8.06058586e-01 -3.79986316e-01 6.27230227e-01 2.28903547e-01 9.29961503e-01 1.17517209e+00 -1.39477301e+00 7.24919140e-01 6.93900883e-01 4.75050867e-01 8.79753649e-01 6.03435397e-01 -7.67806709e-01 -4.61021662e-01 -9.78974283e-01 2.19460988e+00 -6.91173732e-01 9.84657466e-01 -2.19800681e-01 -8.15561831e-01 6.63858235e-01 3.03360522e-01 -2.00148359e-01 9.24084842e-01 5.55603981e-01 -6.24147594e-01 3.09897542e-01 -1.30391419e+00 3.96086007e-01 9.38949883e-01 -5.05407631e-01 -1.02126122e+00 3.44628572e-01 3.43378454e-01 -2.57045209e-01 -1.12401104e+00 -3.21140401e-02 2.89376765e-01 -6.24893367e-01 7.02909172e-01 -1.34533906e+00 8.63344222e-02 -1.47774577e-01 -5.78789413e-02 -1.30687332e+00 -3.65126044e-01 -4.62089390e-01 5.41427433e-01 1.91408169e+00 1.03872180e+00 -6.87384665e-01 5.82698524e-01 8.46529365e-01 -2.32712030e-01 2.16923550e-01 -9.96428967e-01 -1.06105983e+00 3.16305131e-01 -2.42423385e-01 7.96682954e-01 1.48662591e+00 3.03689271e-01 3.79622936e-01 -1.63395524e-01 3.67919534e-01 -1.23120460e-03 -2.23899350e-01 4.62839514e-01 -1.54806077e+00 1.44511938e-01 -2.92123944e-01 -5.68359554e-01 -4.72630054e-01 3.51279736e-01 -8.93679023e-01 1.13289058e-02 -1.58423793e+00 2.00164923e-03 -6.28755808e-01 -4.79577750e-01 9.32986677e-01 5.23008555e-02 1.44213252e-03 6.04003631e-02 9.49951038e-02 -5.51903009e-01 -1.61401093e-01 5.33583403e-01 1.65488884e-01 -2.26123095e-01 9.29159671e-02 -5.18877745e-01 7.23056138e-01 7.13472128e-01 -7.89101064e-01 4.80205268e-01 -2.16742560e-01 4.41347659e-01 -1.38025165e-01 -1.08532337e-02 -1.02658927e+00 4.44204658e-01 -2.08587527e-01 5.01182318e-01 -4.54286367e-01 -2.17433929e-01 -9.30785477e-01 3.60374451e-01 2.01012254e-01 -2.38930553e-01 1.24128915e-01 2.99441338e-01 2.17057675e-01 -3.86199951e-01 -6.44382298e-01 7.17086017e-01 -2.61982054e-01 -1.15699267e+00 1.95688251e-02 -4.51128155e-01 3.89227808e-01 1.02280641e+00 1.37777645e-02 -2.64595151e-01 5.59152626e-02 -1.03630209e+00 3.34448479e-02 3.43545437e-01 2.72689104e-01 -2.47521494e-02 -1.11323416e+00 -6.97542191e-01 2.73449551e-02 3.82469416e-01 -6.30810082e-01 2.98574418e-01 3.53651524e-01 -3.43963772e-01 5.92280865e-01 -2.69436777e-01 2.31173530e-01 -1.03371596e+00 5.23091733e-01 1.68506637e-01 -6.32613063e-01 -2.63671875e-01 5.65102041e-01 -4.28089261e-01 -1.11151361e+00 4.77989623e-03 -3.44316736e-02 -4.04798299e-01 3.52184802e-01 3.71457964e-01 5.30592203e-01 4.88969862e-01 -8.23268473e-01 -6.31260037e-01 1.67561993e-01 -7.28735328e-02 -1.19220920e-01 1.37987113e+00 7.73281828e-02 8.47103074e-04 3.51338267e-01 9.13513660e-01 6.30297065e-01 -3.06916237e-01 -1.25601575e-01 7.02113986e-01 1.23339586e-01 -2.12560281e-01 -1.22560132e+00 -5.30654252e-01 5.19118905e-01 3.54996681e-01 2.95443535e-01 6.65367842e-01 9.90197659e-02 7.32196510e-01 5.89726806e-01 4.97636586e-01 -1.32717299e+00 -1.00375259e+00 1.14064538e+00 1.36130497e-01 -9.91731286e-01 -3.44735146e-01 -1.08370021e-01 -7.19326138e-01 1.08596146e+00 2.14208782e-01 -4.50602872e-03 6.27380669e-01 2.25620717e-01 3.26284260e-01 -8.31961110e-02 -5.25460184e-01 -5.72992086e-01 3.04531097e-01 6.71143532e-01 6.70903027e-01 1.31121010e-01 -7.51265049e-01 9.40787494e-01 -3.49126875e-01 -2.16073856e-01 6.27420247e-01 7.24983633e-01 -2.33455420e-01 -1.17938232e+00 -2.74338365e-01 2.81760216e-01 -1.19077492e+00 -4.42021310e-01 -5.37916839e-01 1.12189031e+00 3.88168484e-01 8.22316289e-01 3.18242945e-02 -2.36378133e-01 8.00749302e-01 5.38623512e-01 -5.78671619e-02 -8.63571942e-01 -1.20508552e+00 -2.67753869e-01 5.80382884e-01 -1.04711130e-01 -1.88322484e-01 -8.18020225e-01 -1.50509048e+00 -2.12693840e-01 -4.29419458e-01 7.99506724e-01 6.20754540e-01 9.62177873e-01 4.66896415e-01 -2.74479687e-02 3.67916077e-01 -1.53985918e-01 -2.33727872e-01 -1.02118123e+00 -8.93931687e-01 4.66859043e-01 -1.46248281e-01 -2.99949318e-01 -2.16664225e-01 1.42515600e-01]

[9.751922607421875, 9.646135330200195]

52ea4c69-1a9c-4c33-95ed-8d760979fef8

explaining-chest-x-ray-pathologies-in-natural

2207.04343

https://arxiv.org/abs/2207.04343v1

https://arxiv.org/pdf/2207.04343v1.pdf

Explaining Chest X-ray Pathologies in Natural Language

Most deep learning algorithms lack explanations for their predictions, which limits their deployment in clinical practice. Approaches to improve explainability, especially in medical imaging, have often been shown to convey limited information, be overly reassuring, or lack robustness. In this work, we introduce the task of generating natural language explanations (NLEs) to justify predictions made on medical images. NLEs are human-friendly and comprehensive, and enable the training of intrinsically explainable models. To this goal, we introduce MIMIC-NLE, the first, large-scale, medical imaging dataset with NLEs. It contains over 38,000 NLEs, which explain the presence of various thoracic pathologies and chest X-ray findings. We propose a general approach to solve the task and evaluate several architectures on this dataset, including via clinician assessment.

['Thomas Lukasiewicz', 'Bartlomiej Papiez', 'Guy Parsons', 'Oana-Maria Camburu', 'Cornelius Emde', 'Maxime Kayser']

2022-07-09

['explainable-models']

['computer-vision']

[ 3.54221761e-01 9.75469410e-01 -3.40006977e-01 -8.21286380e-01 -7.36494958e-01 -2.39710584e-01 2.40688041e-01 2.80956209e-01 3.30987751e-01 9.84778762e-01 5.15133202e-01 -8.32187176e-01 -2.82235533e-01 -3.03624392e-01 -7.22997189e-01 -1.83852389e-01 1.03553228e-01 7.93929338e-01 -3.12237203e-01 2.22463757e-01 -1.30695954e-01 3.61599803e-01 -8.51008534e-01 9.32094157e-01 8.00275743e-01 8.96687925e-01 -5.15213609e-02 8.79794061e-01 -5.86069785e-02 1.46122789e+00 -5.28791487e-01 -7.04261899e-01 -4.17689145e-01 -7.74796188e-01 -9.77774262e-01 1.48150489e-01 3.76809865e-01 -5.81947982e-01 -3.23264569e-01 4.79740202e-01 4.44232732e-01 -3.47181529e-01 8.81699443e-01 -1.48626602e+00 -1.04969263e+00 1.00463200e+00 -3.96958180e-02 2.34965518e-01 4.06380773e-01 2.02350408e-01 9.73118067e-01 -8.41606200e-01 5.36286592e-01 1.02100599e+00 9.54677165e-01 1.34960878e+00 -1.13882637e+00 -4.10949349e-01 -4.35584364e-03 1.74249336e-01 -6.85520530e-01 -1.42059162e-01 4.83491749e-01 -4.19492185e-01 8.02430570e-01 6.07012451e-01 5.79949737e-01 1.58474052e+00 5.33362031e-01 9.05801296e-01 8.66921365e-01 -1.04498476e-01 1.03656769e-01 1.14002042e-01 1.80156380e-01 8.36881042e-01 4.06283200e-01 1.30518705e-01 -4.62111861e-01 -2.98356533e-01 7.27327824e-01 1.97431117e-01 -5.16777754e-01 -6.25880137e-02 -1.46227491e+00 9.32473540e-01 8.34424019e-01 3.96340601e-02 -6.34502351e-01 3.29885691e-01 2.75561839e-01 -1.30743459e-01 2.53884882e-01 8.84949088e-01 -6.88085854e-01 8.70822817e-02 -7.50081539e-01 4.24271643e-01 7.85778642e-01 7.91833222e-01 -2.37634662e-03 4.59431261e-02 -3.49316776e-01 3.77457500e-01 1.15708113e-01 2.00520664e-01 5.70854664e-01 -8.52248788e-01 2.70712823e-01 3.49424869e-01 5.01567824e-03 -9.70209301e-01 -9.51163948e-01 -6.08884156e-01 -1.28437996e+00 -1.85136154e-01 1.46092415e-01 -7.15337098e-02 -9.24603760e-01 1.69477153e+00 5.25188111e-02 2.01342478e-01 2.26633340e-01 1.08023310e+00 1.60423708e+00 1.98793441e-01 4.38191175e-01 -3.46608460e-02 1.19469476e+00 -1.02801919e+00 -9.06270325e-01 -3.29228729e-01 6.12207294e-01 -5.64455271e-01 1.21271896e+00 4.95158613e-01 -1.12378693e+00 -4.47239369e-01 -7.64328957e-01 -1.08059578e-01 3.72929871e-02 2.20148727e-01 7.94496298e-01 1.82707593e-01 -8.87581170e-01 7.58689940e-01 -8.44461501e-01 1.37186736e-01 7.25859642e-01 2.42091417e-01 -2.84772575e-01 -5.85951880e-02 -1.14250457e+00 9.60564435e-01 3.34624916e-01 -8.50939900e-02 -9.22492027e-01 -1.10281885e+00 -9.14948821e-01 3.78661782e-01 4.00308281e-01 -1.36000943e+00 1.64885509e+00 -7.21255720e-01 -8.61779571e-01 8.79960299e-01 -1.23317443e-01 -8.33714068e-01 6.46843255e-01 -2.40746766e-01 -4.23819929e-01 1.66007712e-01 1.04647666e-01 1.07992756e+00 5.73081851e-01 -1.35075998e+00 -1.09904200e-01 4.50325720e-02 -2.79914979e-02 -2.08789021e-01 1.37262661e-02 -4.75815594e-01 8.50319024e-03 -1.04037583e+00 2.24471033e-01 -1.07758570e+00 -7.47648656e-01 1.25666156e-01 -1.12187147e+00 -1.10516255e-03 3.65594357e-01 -6.55608892e-01 1.04313016e+00 -1.88863933e+00 -1.31989330e-01 -2.06345975e-01 9.98663306e-01 3.97068113e-02 1.91928983e-01 -1.21190548e-02 -5.16458750e-01 5.83210826e-01 -3.89224887e-01 -3.67250711e-01 -7.74886981e-02 7.16830909e-01 -5.62388539e-01 -8.10310170e-02 5.58030725e-01 1.33065963e+00 -8.86661708e-01 -7.87755430e-01 2.25885406e-01 5.67306697e-01 -6.88133121e-01 3.42911333e-01 -5.20556450e-01 9.81815398e-01 -6.78799212e-01 5.09390235e-01 2.38990486e-01 -1.06372511e+00 -7.90757462e-02 -8.21118131e-02 7.06781030e-01 4.60162044e-01 -4.18700337e-01 1.40740597e+00 -4.53951657e-01 7.82151401e-01 -5.09240866e-01 -5.12606561e-01 6.00540519e-01 6.85014367e-01 5.86144388e-01 -1.12721868e-01 3.62649262e-02 3.96297097e-01 9.74675715e-02 -8.76764655e-01 9.91804525e-02 -5.98152995e-01 1.08934447e-01 6.32001817e-01 -1.73834711e-01 -3.68894875e-01 -2.88615197e-01 4.12724018e-01 1.21076286e+00 -2.80255675e-01 7.02889442e-01 4.45456803e-02 3.79475921e-01 2.71923125e-01 3.52979243e-01 1.02360046e+00 -1.18815795e-01 1.16130531e+00 5.59563637e-01 -1.10615635e+00 -9.37657833e-01 -1.04606283e+00 -1.74216926e-01 4.39131021e-01 -1.33159518e-01 -3.72169912e-01 -7.61540651e-01 -1.12832487e+00 -4.86172251e-02 1.10886657e+00 -8.35761070e-01 -1.92042351e-01 -4.66814250e-01 -3.84387583e-01 4.88181561e-01 9.50419843e-01 7.12866187e-02 -1.40075302e+00 -8.91112089e-01 2.24815950e-01 -4.86004353e-01 -1.28916204e+00 -3.26998323e-01 2.48171344e-01 -1.18267274e+00 -1.16254187e+00 -7.21166253e-01 -2.45086074e-01 7.92623699e-01 -2.26797298e-01 1.91848624e+00 7.32624054e-01 -6.11610174e-01 2.90289909e-01 -2.98658967e-01 -6.70446396e-01 -7.85495281e-01 8.62590000e-02 -8.52341056e-02 -6.48141325e-01 1.18656553e-01 -2.27340102e-01 -7.74232447e-01 2.34956175e-01 -9.86548603e-01 6.01434231e-01 7.20002115e-01 1.18807828e+00 8.78104448e-01 -4.17115539e-01 5.85164249e-01 -1.39436412e+00 7.08959460e-01 -6.48580074e-01 2.94098675e-01 2.50573397e-01 -6.56649113e-01 2.26278365e-01 8.10965180e-01 -3.39048117e-01 -8.67540777e-01 8.03133100e-02 -4.40099448e-01 -4.71190482e-01 -5.09213507e-01 6.72066808e-01 2.87587345e-01 1.97226003e-01 9.30096805e-01 2.07262170e-02 -3.13024342e-01 -2.67965108e-01 3.73190224e-01 3.50170463e-01 9.85491753e-01 -3.59563231e-01 4.16337252e-01 3.43666017e-01 1.95731372e-01 -3.08045931e-02 -1.58342576e+00 1.82932079e-01 -3.47788423e-01 1.02036214e-03 9.27294910e-01 -5.15866160e-01 -5.87837040e-01 -5.56637585e-01 -1.41551256e+00 -2.24694133e-01 -4.86525744e-01 5.67455709e-01 -7.70743430e-01 1.26980126e-01 -6.26625359e-01 -4.08991039e-01 -4.73377764e-01 -1.23104835e+00 1.13177145e+00 3.80220227e-02 -9.89316463e-01 -1.07374871e+00 -4.09372687e-01 3.96638811e-01 4.16198969e-01 7.15066969e-01 1.15410638e+00 -9.88727927e-01 -4.36280221e-01 -1.19917549e-01 -3.06594074e-01 -6.25112141e-03 8.86231065e-02 -1.43460482e-01 -8.20917785e-01 2.14372203e-01 1.35320738e-01 -6.07352674e-01 7.17972159e-01 6.89108491e-01 2.11662865e+00 -5.20167053e-01 -4.10312921e-01 6.77393973e-01 8.85566115e-01 -6.74180090e-02 3.20115954e-01 1.70115754e-02 5.86121738e-01 6.56846404e-01 5.45827806e-01 5.04817724e-01 4.53099668e-01 2.48428315e-01 7.87008941e-01 -6.72987401e-01 -2.80090392e-01 -3.74014229e-01 -4.02706444e-01 6.30601525e-01 2.16167837e-01 -4.89292294e-01 -1.20457947e+00 3.78366500e-01 -1.85462523e+00 -4.77613330e-01 -5.90373993e-01 1.27037692e+00 7.42358267e-01 9.98519287e-02 -4.47660029e-01 -3.22442502e-02 2.82136589e-01 -1.70607880e-01 -8.91455293e-01 -4.82030928e-01 -8.86019841e-02 5.96451387e-02 6.09976277e-02 2.80998290e-01 -8.28929186e-01 4.31135237e-01 7.03233528e+00 3.27168763e-01 -9.57167029e-01 1.92081064e-01 1.26814079e+00 9.26263817e-03 -6.99402988e-01 -4.97597605e-01 -1.47426903e-01 2.55490959e-01 8.38028252e-01 6.95561171e-02 -4.42767106e-02 1.09496331e+00 2.36843482e-01 3.66179764e-01 -1.69930053e+00 9.97917712e-01 4.93680220e-03 -1.78735292e+00 2.77151108e-01 -1.93140373e-01 7.79966712e-01 -2.88077444e-01 3.52689922e-01 4.22841460e-02 2.77226269e-01 -1.62691498e+00 5.80707610e-01 5.42113423e-01 9.96965051e-01 -2.88923591e-01 9.24974501e-01 3.68422866e-01 -5.93102455e-01 -8.36747289e-02 -2.11610511e-01 1.18124090e-01 2.99081713e-01 5.24069011e-01 -1.30276728e+00 3.62967163e-01 5.70825934e-01 6.34978712e-01 -4.74599481e-01 7.39690125e-01 -7.20590830e-01 7.38977790e-01 1.02374651e-01 1.87518075e-01 4.16716933e-01 7.63230562e-01 2.42908224e-01 1.12563920e+00 3.82770598e-01 6.13969266e-01 5.60092591e-02 1.13718343e+00 -2.54407972e-01 -1.46358833e-01 -4.62858588e-01 -2.16865446e-02 2.55960468e-02 1.12094557e+00 -5.98812580e-01 -7.90025115e-01 -1.49092257e-01 8.76571536e-01 5.54784089e-02 6.18797056e-02 -1.24943638e+00 4.40258235e-01 2.64312923e-01 2.70332426e-01 -3.18974078e-01 3.93206984e-01 -9.27977145e-01 -1.21867561e+00 -6.18596748e-02 -1.20183194e+00 7.06948400e-01 -1.32749665e+00 -1.43899179e+00 1.18457842e+00 -1.54202148e-01 -1.13443661e+00 -7.80255318e-01 -7.98873246e-01 -4.35628712e-01 7.34472513e-01 -1.51201046e+00 -1.04160357e+00 -7.21305311e-01 4.99424368e-01 8.34063709e-01 -1.18148074e-01 1.10944557e+00 1.60004199e-01 -1.33818299e-01 4.49325472e-01 -5.93949854e-01 -1.62422150e-01 5.73832750e-01 -1.47740996e+00 4.96385694e-01 2.25084364e-01 3.16808224e-01 6.96883142e-01 1.08261168e+00 -6.88750863e-01 -7.40744889e-01 -1.10701489e+00 1.18905401e+00 -7.22503781e-01 3.28417540e-01 1.67448223e-01 -1.14262807e+00 9.19953346e-01 -2.30679871e-03 2.49517426e-01 1.05997109e+00 -7.67402500e-02 -2.02073112e-01 5.08323550e-01 -1.27783573e+00 6.13188982e-01 9.65245664e-01 -3.05235535e-01 -8.27176094e-01 7.73043156e-01 9.95396495e-01 -8.20672393e-01 -8.53902161e-01 6.23437107e-01 5.40151417e-01 -1.02546167e+00 1.12431526e+00 -1.31395662e+00 1.48831904e+00 1.25697896e-01 1.47689983e-01 -1.33273411e+00 -2.89316893e-01 -4.00024235e-01 -3.45673531e-01 5.77431262e-01 8.16524267e-01 -3.35180819e-01 1.04370499e+00 1.06829596e+00 -3.75907868e-01 -1.29305267e+00 -6.91007137e-01 -1.70648575e-01 -1.23728923e-02 -6.51835978e-01 1.00411093e+00 1.12122095e+00 -1.78942576e-01 8.97972882e-02 -5.95628023e-01 2.41586700e-01 4.30987090e-01 1.52752772e-01 5.46329796e-01 -1.08176148e+00 -6.15143895e-01 -3.84814739e-01 -1.52323082e-01 -7.27546096e-01 2.36661375e-01 -7.87881732e-01 1.45195156e-01 -1.91881812e+00 4.80659068e-01 -3.14600408e-01 -2.35659599e-01 7.61048019e-01 -6.25604272e-01 1.42963544e-01 8.41848478e-02 4.05863672e-01 -3.75316113e-01 2.67331481e-01 1.42251444e+00 -2.09429383e-01 3.48924428e-01 1.10020317e-01 -9.49682236e-01 1.02906537e+00 8.23487818e-01 -7.60961294e-01 -3.63141090e-01 -6.65956438e-01 2.15355337e-01 5.50633729e-01 6.44622505e-01 -6.14388704e-01 -9.15157348e-02 -9.46817994e-02 7.63589621e-01 -5.98509848e-01 6.58157915e-02 -8.23305786e-01 3.37945431e-01 9.15097952e-01 -1.08611929e+00 2.39929736e-01 1.56861559e-01 4.41645801e-01 -3.93178940e-01 -1.39823452e-01 5.57032108e-01 -4.16308969e-01 -2.46654749e-01 4.59330887e-01 -1.79090559e-01 6.75286679e-03 7.79080451e-01 1.65865272e-01 -3.04716080e-01 -6.87612176e-01 -1.18666184e+00 2.65681565e-01 -3.07751540e-02 2.57974148e-01 1.04594600e+00 -1.31510806e+00 -9.90871012e-01 -1.91764086e-01 2.31926218e-01 8.27517509e-02 5.24293303e-01 6.57623649e-01 -6.66147828e-01 6.79780722e-01 -7.59382397e-02 -6.43962681e-01 -1.21423709e+00 7.48452842e-01 4.98100877e-01 -6.59579754e-01 -9.54948664e-01 9.41172719e-01 5.87734282e-01 -5.89586318e-01 1.21367261e-01 -7.29424834e-01 -1.68397650e-01 -5.94421506e-01 5.32143414e-01 -7.47638941e-02 -1.82681859e-01 -2.96808749e-01 -3.52840990e-01 1.26207873e-01 3.36742476e-02 2.19895750e-01 1.38015163e+00 9.13751796e-02 2.14171767e-01 1.81394145e-01 6.81327820e-01 -4.99993056e-01 -1.00001335e+00 -8.74891654e-02 5.35954051e-02 -3.37066114e-01 -1.94089115e-01 -1.26123130e+00 -1.08482385e+00 1.31975639e+00 2.38701925e-01 2.48623073e-01 1.10629106e+00 4.13887173e-01 1.03096032e+00 3.18809420e-01 -1.59599230e-01 -3.01986396e-01 3.00675213e-01 -7.39388242e-02 1.27684796e+00 -1.59492505e+00 1.54261924e-02 -6.23461068e-01 -1.00983608e+00 1.34928238e+00 7.14872837e-01 3.42329293e-01 2.64981627e-01 2.78325647e-01 2.96103597e-01 -5.67073405e-01 -1.11045718e+00 2.41096497e-01 5.84289789e-01 5.31479895e-01 7.92447746e-01 4.31162983e-01 -7.88748860e-02 1.14673734e+00 -4.05815542e-01 2.36832872e-01 6.11479044e-01 4.59565639e-01 5.00732660e-02 -7.11096346e-01 -3.98253560e-01 7.52779245e-01 -8.76866102e-01 -3.19097608e-01 -7.24200547e-01 7.62716651e-01 1.26992047e-01 8.16732526e-01 -3.55329931e-01 -2.04287574e-01 2.60108948e-01 -1.32289112e-01 6.79850802e-02 -6.98086500e-01 -6.56139791e-01 -2.68975168e-01 1.72606662e-01 -6.51467621e-01 -1.44555062e-01 -2.73538202e-01 -1.60000134e+00 -1.74243331e-01 -3.41952480e-02 6.70131966e-02 4.32496727e-01 8.86577249e-01 3.06563318e-01 1.20244658e+00 1.68058529e-01 -3.83394003e-01 -7.05777168e-01 -6.93273365e-01 -1.57321379e-01 6.85985327e-01 6.15172029e-01 -4.84372318e-01 -2.11960077e-01 2.88223028e-01]

[8.626194953918457, 5.856684684753418]

73d07c97-5ccd-4099-b5f7-b66f10c0c366

boosting-on-the-shoulders-of-giants-in

2005.06194

https://arxiv.org/abs/2005.06194v1

https://arxiv.org/pdf/2005.06194v1.pdf

Boosting on the shoulders of giants in quantum device calibration

Traditional machine learning applications, such as optical character recognition, arose from the inability to explicitly program a computer to perform a routine task. In this context, learning algorithms usually derive a model exclusively from the evidence present in a massive dataset. Yet in some scientific disciplines, obtaining an abundance of data is an impractical luxury, however; there is an explicit model of the domain based upon previous scientific discoveries. Here we introduce a new approach to machine learning that is able to leverage prior scientific discoveries in order to improve generalizability over a scientific model. We show its efficacy in predicting the entire energy spectrum of a Hamiltonian on a superconducting quantum device, a key task in present quantum computer calibration. Our accuracy surpasses the current state-of-the-art by over $20\%.$ Our approach thus demonstrates how artificial intelligence can be further enhanced by "standing on the shoulders of giants."

['Mile Gu', 'Jayne Thompson', 'Felix Binder', 'Alex Wozniakowski']

2020-05-13

['multi-target-regression']

['miscellaneous']

[ 3.92880440e-01 -1.45462677e-02 -3.09345216e-01 -3.69720787e-01 -7.84772575e-01 -4.21898216e-01 5.57542086e-01 1.18727274e-01 -3.91306579e-01 1.04792345e+00 -5.28377593e-01 -6.56776905e-01 -1.51375413e-01 -7.65450418e-01 -9.79746878e-01 -8.50663483e-01 3.62055421e-01 5.57777584e-01 7.94373602e-02 -2.19345465e-01 7.80438602e-01 3.86802793e-01 -1.42815590e+00 -1.24849275e-01 8.87518585e-01 9.02532876e-01 1.99947774e-01 4.67753410e-01 -1.00538693e-01 7.85127103e-01 -2.69363225e-01 -5.45104861e-01 1.10812388e-01 -6.02787316e-01 -8.25675786e-01 -3.62061292e-01 3.89461100e-01 -1.06728273e-02 -3.25003326e-01 1.33427513e+00 5.84217943e-02 -2.85285354e-01 6.92481816e-01 -8.77750099e-01 -5.75951099e-01 4.90722895e-01 -5.00448763e-01 9.64294896e-02 1.60678178e-01 3.73646498e-01 1.20376134e+00 -2.41534635e-01 6.43917799e-01 5.10158837e-01 6.25032127e-01 5.62692404e-01 -1.55142736e+00 -7.77394712e-01 -4.75702465e-01 3.40525538e-01 -1.14736748e+00 -3.92238945e-01 9.51470256e-01 -5.61454296e-01 1.04078305e+00 4.36127260e-02 6.66450977e-01 8.39115024e-01 2.89872319e-01 3.00614208e-01 1.47906327e+00 -8.40176940e-01 3.11400771e-01 2.44729891e-01 3.73873740e-01 9.83621299e-01 5.37768424e-01 3.02426159e-01 -7.90484846e-01 -2.20606074e-01 5.85578859e-01 -3.95535350e-01 -2.59159058e-02 -4.87626135e-01 -1.23732412e+00 6.62184715e-01 8.84777084e-02 2.13229418e-01 -1.07010044e-01 2.64663249e-01 1.13245495e-01 4.21536207e-01 1.33374467e-01 1.04947615e+00 -6.41989589e-01 -3.78394336e-01 -1.02022278e+00 2.06126124e-01 9.02448535e-01 4.56820220e-01 1.00253308e+00 -1.29958272e-01 7.11788297e-01 3.02452832e-01 3.47308740e-02 3.41859341e-01 3.13405395e-01 -1.07322252e+00 -1.43339664e-01 5.51101327e-01 1.39164194e-01 -4.66037452e-01 -3.01372916e-01 -4.66115266e-01 -6.00222230e-01 4.02791291e-01 7.02914834e-01 -1.39095947e-01 -6.87739968e-01 1.72132456e+00 1.69449255e-01 2.22979948e-01 -2.73817573e-02 6.30005777e-01 4.01512891e-01 3.35204333e-01 -5.60397841e-02 -2.04305634e-01 9.80474889e-01 -4.13067818e-01 -2.77037024e-01 -1.46744698e-01 7.96317875e-01 -5.43515325e-01 8.96650374e-01 8.78453612e-01 -8.15950394e-01 -3.01976413e-01 -1.37518024e+00 -1.08020388e-01 -1.69707701e-01 -6.14627898e-02 1.42921901e+00 9.65166271e-01 -7.94251680e-01 1.17450428e+00 -7.39304781e-01 -1.96434543e-01 3.34861606e-01 6.55253351e-01 -5.19551218e-01 1.91667333e-01 -7.61120319e-01 1.15805507e+00 5.19233167e-01 -2.75045365e-01 -5.50801694e-01 -8.40468705e-01 -2.96223640e-01 -2.44419560e-01 4.58360553e-01 -7.01533496e-01 1.07019913e+00 -7.74731338e-01 -1.63478374e+00 1.28606403e+00 -1.04624338e-01 -4.60566670e-01 1.77779853e-01 1.53752491e-01 -3.22028279e-01 4.54263575e-02 -2.45707452e-01 2.45688424e-01 6.81502879e-01 -1.05737138e+00 -3.15248311e-01 -4.08172160e-01 6.34886138e-03 -5.83827913e-01 -2.89511085e-01 -1.17424995e-01 2.63180980e-03 -8.39345828e-02 3.30325961e-01 -1.15283382e+00 -1.69692233e-01 -1.98102191e-01 -2.43249446e-01 -2.69554257e-01 3.54759425e-01 -1.94261745e-01 9.59092021e-01 -1.78329337e+00 3.70757639e-01 9.57781747e-02 3.95846993e-01 4.95837592e-02 2.09184811e-01 2.55528867e-01 -1.66668758e-01 1.04322724e-01 -2.52305448e-01 8.08375105e-02 6.87064184e-03 3.44581567e-02 -2.74211258e-01 6.01847887e-01 1.69665277e-01 8.55914950e-01 -9.30937469e-01 -3.15479964e-01 3.18357497e-02 -9.15521476e-03 -6.29060864e-01 -5.02060540e-02 -4.61005330e-01 5.74673176e-01 -4.72916067e-01 4.98436779e-01 4.10129100e-01 -7.83059776e-01 4.52107906e-01 -3.99633236e-02 -2.66842216e-01 5.55746198e-01 -9.68471467e-01 2.07215619e+00 2.23890617e-02 5.48631251e-01 -7.47573525e-02 -1.55521333e+00 8.05280387e-01 -2.14119609e-02 6.52443409e-01 -5.87265968e-01 3.26565266e-01 4.92508948e-01 5.03041267e-01 -5.62127411e-01 4.04304206e-01 -6.07620060e-01 -9.90598500e-02 6.85362339e-01 2.47507080e-01 -6.02567911e-01 7.08071068e-02 5.72935864e-02 1.13306189e+00 4.28329498e-01 2.10242957e-01 -6.31866693e-01 2.97778904e-01 2.56787479e-01 4.81698930e-01 9.06323254e-01 -1.33279502e-01 2.13250741e-01 5.72903037e-01 -6.85203791e-01 -1.48496854e+00 -8.88544381e-01 -5.62942505e-01 7.21061051e-01 7.47363046e-02 -3.28635067e-01 -7.82081842e-01 -4.24603045e-01 3.82058173e-02 6.56554937e-01 -4.80713576e-01 -2.85052031e-01 -3.14609349e-01 -1.30550182e+00 3.88008028e-01 -2.49889586e-03 2.40538195e-01 -6.58669591e-01 -5.48276782e-01 6.87767789e-02 2.09347069e-01 -1.03002429e+00 3.59195262e-01 4.17512238e-01 -1.03032935e+00 -9.41461325e-01 -1.75327092e-01 -3.28436434e-01 3.87979954e-01 -1.13193765e-01 1.27706552e+00 1.67091772e-01 -5.35486937e-01 -6.56451955e-02 -4.98244092e-02 -4.90739137e-01 -6.71890378e-01 3.00982326e-01 4.72632557e-01 -3.29855323e-01 8.29800963e-01 -9.47117090e-01 -3.26633304e-01 -2.53971428e-01 -5.49248934e-01 1.77329391e-01 7.40322351e-01 9.12115335e-01 2.51061946e-01 8.53369385e-02 4.95846242e-01 -8.80737662e-01 1.44917458e-01 -3.89325023e-01 -7.95663178e-01 3.10521752e-01 -8.73760104e-01 6.90278709e-01 5.39286137e-01 -2.49427333e-01 -7.77757764e-01 7.71852657e-02 6.17813261e-04 5.25262430e-02 -2.42259398e-01 5.31272113e-01 -4.16254327e-02 -5.12987018e-01 8.07955086e-01 1.83803901e-01 -7.60397241e-02 -5.08346498e-01 1.57820791e-01 5.76931894e-01 6.07094705e-01 -1.12039077e+00 6.70546055e-01 3.28787178e-01 6.20225608e-01 -9.31637585e-01 -9.28777754e-01 -1.39298528e-01 -9.16452289e-01 1.96454916e-02 5.23573101e-01 -6.52450979e-01 -1.09111202e+00 3.50007355e-01 -9.22070503e-01 -7.64141157e-02 -1.39692649e-01 6.48397148e-01 -4.83740836e-01 5.35743117e-01 -3.55058610e-01 -9.11999941e-01 -7.94229582e-02 -9.44500089e-01 6.53721631e-01 3.97409678e-01 -1.91990867e-01 -8.20243776e-01 2.21786618e-01 6.78015769e-01 2.46420100e-01 1.72707111e-01 1.20075834e+00 -4.40169126e-01 -8.63420367e-01 -1.96974635e-01 -1.68535277e-01 2.62646735e-01 -1.54895544e-01 1.52622104e-01 -1.18446779e+00 -1.05295524e-01 2.34874949e-01 -6.88952446e-01 9.17583227e-01 1.05536124e-02 1.33461630e+00 2.26160824e-01 -3.08964401e-01 5.07647336e-01 1.58591199e+00 8.77034515e-02 6.29090130e-01 3.02443117e-01 5.76914907e-01 5.58386981e-01 -4.75640930e-02 8.95995796e-02 4.95163016e-02 5.00031531e-01 1.75865933e-01 3.93832445e-01 1.75468400e-01 -1.76732957e-01 -6.15603011e-03 9.10661817e-01 -4.95925725e-01 2.52444476e-01 -1.18290758e+00 1.71656758e-01 -1.49805689e+00 -9.74740982e-01 -1.90795347e-01 2.41824841e+00 1.11166108e+00 6.30136788e-01 -9.69759002e-02 1.52948588e-01 2.95575589e-01 -2.38533944e-01 -8.41145456e-01 -2.84424365e-01 -1.17004529e-01 6.67726278e-01 5.25411308e-01 2.22587541e-01 -7.82566369e-01 8.45778048e-01 7.46676064e+00 9.12012279e-01 -1.28870380e+00 2.05211882e-02 4.22049314e-01 -5.26094576e-03 -4.32527661e-01 3.31020266e-01 -6.93091989e-01 4.43489760e-01 1.38703227e+00 -1.52018396e-02 6.16721749e-01 6.32803559e-01 -2.92280018e-01 -3.95008922e-01 -1.49759614e+00 1.10518837e+00 -9.27376300e-02 -1.65689087e+00 -9.61613059e-02 3.85804892e-01 6.89256787e-01 2.59787798e-01 1.38751253e-01 2.11651519e-01 3.57431084e-01 -1.21446431e+00 4.02469844e-01 6.73883617e-01 9.60496485e-01 -4.47409689e-01 3.84060115e-01 6.87261820e-01 -2.43885338e-01 1.39661983e-01 -3.07739586e-01 -5.81075251e-01 -2.76422173e-01 6.84156001e-01 -4.77154940e-01 3.48723620e-01 3.32676262e-01 4.21230942e-01 -5.35352111e-01 7.91114569e-01 5.01215868e-02 7.19180286e-01 -4.01573688e-01 -3.70164901e-01 3.95396948e-02 -5.41788042e-01 2.27199078e-01 9.27501202e-01 1.88546464e-01 3.37680787e-01 -1.76402733e-01 1.02582562e+00 -3.88788462e-01 -1.72997504e-01 -5.15198529e-01 -5.27806938e-01 1.16866477e-01 1.18528831e+00 -6.28824532e-01 -1.68218657e-01 -4.17208791e-01 5.92377007e-01 5.28804898e-01 -1.09303735e-01 -6.29637897e-01 -2.73507386e-01 2.66638786e-01 -9.17622671e-02 1.33561566e-01 -4.96250272e-01 -6.73577070e-01 -1.52549708e+00 2.31733210e-02 -8.48843575e-01 -1.94541007e-01 -6.50219858e-01 -1.20525098e+00 5.21781147e-02 -3.34715873e-01 -6.68963253e-01 -2.36166447e-01 -1.07722795e+00 -3.14632237e-01 7.65897632e-01 -1.31516671e+00 -8.38616490e-01 1.38967440e-01 2.70481035e-02 -9.68633369e-02 -2.29315266e-01 1.04056478e+00 9.65522081e-02 -5.24671912e-01 3.44839990e-01 5.30287683e-01 -9.43739638e-02 5.48016906e-01 -1.29949629e+00 1.45957604e-01 5.87936163e-01 5.17983854e-01 7.70496666e-01 1.06889498e+00 -4.45837766e-01 -2.08887386e+00 -1.21211499e-01 7.33096123e-01 -8.82410228e-01 1.19103110e+00 -3.80272418e-01 -9.68459904e-01 3.83883715e-01 -7.05195125e-03 2.68953219e-02 9.03221369e-01 6.82863653e-01 -6.89431787e-01 -1.30061030e-01 -9.67033565e-01 2.83067226e-01 1.00621688e+00 -9.72801030e-01 -7.24189162e-01 3.26547146e-01 1.80173531e-01 -1.38051569e-01 -9.33395863e-01 3.79142851e-01 8.78950775e-01 -9.55953419e-01 8.15917611e-01 -9.60565090e-01 7.21815825e-01 -1.24058731e-01 -2.30491221e-01 -8.91675234e-01 -7.82766417e-02 -8.55152667e-01 -1.91585913e-01 7.76645422e-01 5.53661942e-01 -3.26317459e-01 1.15955949e+00 1.09038305e+00 1.63966730e-01 -5.24770200e-01 -1.03077853e+00 -6.63152397e-01 5.60408115e-01 -6.08204961e-01 3.41177940e-01 1.07000232e+00 2.70941019e-01 4.67485815e-01 -2.32538372e-01 1.78701337e-02 1.10010278e+00 3.87531698e-01 6.61755741e-01 -1.58503473e+00 -6.26657844e-01 -5.93238473e-01 -5.56293011e-01 -8.44050288e-01 3.76399875e-01 -1.03450966e+00 -2.11303800e-01 -6.81053102e-01 6.86390042e-01 -6.93619609e-01 -5.88846564e-01 8.98923352e-02 5.07733598e-02 4.34641838e-01 2.78960429e-02 3.00189674e-01 -4.80885178e-01 1.37578800e-01 8.94271612e-01 -9.90884677e-02 1.88901082e-01 -3.80600482e-01 -9.20799673e-01 9.87878978e-01 7.32249081e-01 -5.78806043e-01 6.86037168e-02 -2.49039382e-01 7.88563013e-01 -5.74621782e-02 5.05467117e-01 -1.22354805e+00 4.09226060e-01 -2.56210774e-01 3.07582200e-01 4.81527112e-03 2.67816365e-01 -4.92080688e-01 2.11096674e-01 4.24685299e-01 -3.31012696e-01 -4.62810278e-01 4.09414694e-02 4.77020502e-01 1.82741508e-01 -6.22068107e-01 7.96707332e-01 -3.90941411e-01 -5.80687463e-01 3.75452265e-02 4.60811779e-02 2.72074137e-02 6.98723674e-01 1.30215108e-01 -3.94917160e-01 1.77936494e-01 -4.48783755e-01 -3.12821686e-01 8.60221922e-01 -1.16441600e-01 9.26158279e-02 -8.31977785e-01 -4.74233508e-01 1.26494110e-01 3.31124254e-02 -4.70455557e-01 1.59319974e-02 7.48927116e-01 -5.13126969e-01 5.80017865e-01 -2.84358412e-01 -5.75070381e-01 -9.26304102e-01 7.44373381e-01 2.05280602e-01 -7.09443167e-02 -4.48258340e-01 8.35956812e-01 -1.97758377e-01 -1.52618468e-01 -1.73058003e-01 -1.45242110e-01 2.95466870e-01 -4.47778136e-01 3.23672533e-01 1.01820499e-01 1.65061802e-01 -2.60896713e-01 -1.50554344e-01 6.79824591e-01 -2.08321705e-01 -3.40208784e-02 1.50613308e+00 2.13755429e-01 -3.81533355e-01 6.74890876e-01 9.37178075e-01 8.93050358e-02 -1.14132369e+00 -3.36343944e-01 1.85884878e-01 -2.68306196e-01 3.25957119e-01 -8.46206069e-01 -5.59722960e-01 1.02724898e+00 3.89548153e-01 1.99755907e-01 7.57266521e-01 2.07920745e-01 7.69083381e-01 1.04761350e+00 8.45164776e-01 -1.26456976e+00 -1.70057774e-01 2.46110931e-01 2.20867664e-01 -1.67578089e+00 3.75608414e-01 -5.73486388e-02 -2.48843521e-01 1.32553184e+00 2.33483821e-01 -2.03236088e-01 6.10656261e-01 2.43343815e-01 -4.42616642e-01 -3.97893757e-01 -6.79155231e-01 6.71544522e-02 2.13167980e-01 3.34036320e-01 5.95639646e-01 9.30459350e-02 -2.06811562e-01 3.93390864e-01 -4.75727469e-01 3.49742770e-01 7.80579984e-01 9.50428724e-01 -6.81372106e-01 -1.48947644e+00 -2.33084038e-01 5.64921379e-01 -7.08744287e-01 -2.18691170e-01 -3.37478340e-01 5.24674118e-01 4.47941363e-01 6.96531713e-01 -2.84721136e-01 -4.46579486e-01 -1.52255863e-01 2.96240717e-01 1.16451812e+00 -5.92193544e-01 -1.13834664e-01 -3.26923460e-01 -2.12546889e-04 -1.99674144e-01 -6.77438915e-01 -8.23941171e-01 -1.12197804e+00 -7.00145841e-01 -3.61645639e-01 2.91500717e-01 9.04979169e-01 1.22401786e+00 1.58372611e-01 1.59457818e-01 2.64352381e-01 -7.04692185e-01 -6.82674468e-01 -5.95631838e-01 -7.44185030e-01 3.23677093e-01 4.32091773e-01 -7.15396643e-01 -3.53009492e-01 6.32132888e-02]

[5.722500324249268, 4.917947292327881]

c87f6d44-06a7-4920-8513-24d0f3e2ac73

sharp-bounds-for-generalized-causal

2305.16988

https://arxiv.org/abs/2305.16988v1

https://arxiv.org/pdf/2305.16988v1.pdf

Sharp Bounds for Generalized Causal Sensitivity Analysis

Causal inference from observational data is crucial for many disciplines such as medicine and economics. However, sharp bounds for causal effects under relaxations of the unconfoundedness assumption (causal sensitivity analysis) are subject to ongoing research. So far, works with sharp bounds are restricted to fairly simple settings (e.g., a single binary treatment). In this paper, we propose a unified framework for causal sensitivity analysis under unobserved confounding in various settings. For this, we propose a flexible generalization of the marginal sensitivity model (MSM) and then derive sharp bounds for a large class of causal effects. This includes (conditional) average treatment effects, effects for mediation analysis and path analysis, and distributional effects. Furthermore, our sensitivity model is applicable to discrete, continuous, and time-varying treatments. It allows us to interpret the partial identification problem under unobserved confounding as a distribution shift in the latent confounders while evaluating the causal effect of interest. In the special case of a single binary treatment, our bounds for (conditional) average treatment effects coincide with recent optimality results for causal sensitivity analysis. Finally, we propose a scalable algorithm to estimate our sharp bounds from observational data.

['Stefan Feuerriegel', 'Valentyn Melnychuk', 'Dennis Frauen']

2023-05-26

['causal-inference', 'causal-inference']

['knowledge-base', 'miscellaneous']

[ 3.97298008e-01 1.28511444e-01 -1.02540064e+00 -5.01080632e-01 -6.39079154e-01 -6.07977629e-01 3.02538693e-01 1.95913255e-01 -1.93877116e-01 1.14190912e+00 7.27039516e-01 -6.56819880e-01 -8.02471519e-01 -8.17929924e-01 -8.89309824e-01 -6.66088700e-01 -3.31634104e-01 2.58720130e-01 -1.76975295e-01 1.48224473e-01 7.96399787e-02 3.23977143e-01 -9.97731626e-01 -2.89303750e-01 1.26077378e+00 1.39890984e-01 -1.94260731e-01 3.26943815e-01 5.00586569e-01 1.71643093e-01 -1.12271741e-01 -2.44351372e-01 -5.40808812e-02 -4.57013220e-01 -8.18202138e-01 -2.20203131e-01 3.79434645e-01 -5.99596322e-01 -1.14617951e-01 1.05359578e+00 5.64823925e-01 -3.03297825e-02 7.97593951e-01 -1.58610201e+00 -9.15362060e-01 1.02315772e+00 -8.17780733e-01 2.11679310e-01 2.33274639e-01 -1.14627101e-01 9.47673738e-01 -5.36369205e-01 4.44705606e-01 1.79564142e+00 3.92605335e-01 3.19118887e-01 -1.57236814e+00 -8.14023197e-01 4.21632200e-01 -6.28926083e-02 -1.01490343e+00 -4.15525496e-01 2.17613816e-01 -5.62680304e-01 1.32720843e-01 3.07116657e-01 8.70747641e-02 1.33031940e+00 2.75930643e-01 3.48602861e-01 1.24084342e+00 -4.07847732e-01 3.79488111e-01 -2.69204557e-01 5.43112040e-01 2.70548463e-01 8.23706090e-01 5.68549573e-01 -8.74399319e-02 -6.36690855e-01 9.57120836e-01 1.08874954e-01 -3.84536952e-01 -2.70070225e-01 -1.28485978e+00 1.15964544e+00 2.51789182e-01 -1.09194763e-01 -6.15696251e-01 2.76039153e-01 2.05980986e-01 1.87961057e-01 4.34253126e-01 2.01620966e-01 -6.23117566e-01 2.04056397e-01 -5.04253030e-01 4.82837081e-01 6.92471504e-01 8.09399545e-01 2.46349812e-01 -3.03098083e-01 -7.15076089e-01 5.51674247e-01 7.62224421e-02 8.87486100e-01 -2.73928791e-01 -1.20064270e+00 2.54797429e-01 1.68094680e-01 5.75770259e-01 -6.83620214e-01 -5.83121836e-01 -1.88441962e-01 -9.56669748e-01 -2.58781165e-01 7.38547385e-01 -5.73844552e-01 -6.25713050e-01 2.62706304e+00 6.06295884e-01 3.61161470e-01 -2.53119946e-01 9.70652163e-01 3.21791679e-01 2.72805959e-01 4.58438993e-01 -9.61753607e-01 1.40964830e+00 -3.35988045e-01 -1.00283027e+00 2.85403460e-01 5.74312270e-01 -5.40103018e-01 1.06026518e+00 2.72476580e-02 -1.28846729e+00 2.14485005e-02 -2.58550018e-01 2.37145685e-02 6.95549995e-02 -4.62671548e-01 8.13481808e-01 7.19962180e-01 -8.80925000e-01 3.94682050e-01 -5.89881837e-01 -5.16120672e-01 1.72433540e-01 5.13807058e-01 -3.00027579e-02 -2.98277706e-01 -1.52478397e+00 4.87653136e-01 -3.67511995e-02 -1.22535720e-01 -8.95656049e-01 -1.39899206e+00 -6.16925001e-01 3.05056214e-01 8.30310285e-01 -1.24557924e+00 1.24525011e+00 -6.22501493e-01 -1.08703303e+00 3.59899551e-01 -4.25820321e-01 -6.33010045e-02 3.84387702e-01 -4.96213585e-02 -2.49427348e-01 -1.80303290e-01 6.06513679e-01 9.43094194e-02 3.83865595e-01 -9.97517347e-01 -5.44658720e-01 -7.11294472e-01 4.08559173e-01 7.70234168e-02 -9.85646918e-02 4.28127050e-01 3.74365374e-02 -6.13155663e-01 -9.94035304e-02 -8.72747719e-01 -6.03223145e-01 -3.82119447e-01 -5.15276432e-01 -1.97513238e-01 1.08600400e-01 -3.46597821e-01 1.34316468e+00 -2.04950118e+00 1.09226555e-01 1.33340761e-01 1.77209228e-01 -5.49913049e-01 -5.97517565e-02 3.82409096e-01 -4.61524487e-01 3.84807557e-01 -3.42288345e-01 2.39713728e-01 2.76695818e-01 -3.26344781e-02 -3.09531420e-01 7.20995009e-01 -4.85181473e-02 9.47415590e-01 -9.47119951e-01 -4.69262362e-01 -9.25960317e-02 1.04064949e-01 -8.38279903e-01 -2.39457842e-03 1.15028761e-01 6.27392888e-01 -6.87671661e-01 4.67982471e-01 9.25138474e-01 -4.23827082e-01 4.77288902e-01 1.48561671e-01 -4.07316864e-01 9.64373797e-02 -1.17459607e+00 1.13278627e+00 -3.37494940e-01 -8.38634595e-02 2.02567503e-01 -1.04873776e+00 1.44982502e-01 5.55208683e-01 5.25891185e-01 -3.83407474e-01 1.69231489e-01 8.84148777e-02 1.36448547e-01 -6.57171011e-01 3.58328931e-02 -7.50086963e-01 -4.66104150e-01 5.41966438e-01 -2.98176348e-01 5.94427824e-01 -3.07698920e-02 2.10746273e-01 9.35424387e-01 -2.71838635e-01 6.71244025e-01 -7.08770514e-01 -4.80356589e-02 -1.50889561e-01 7.08607674e-01 1.00193989e+00 -2.01734111e-01 2.72191256e-01 8.84793401e-01 1.28177911e-01 -8.37608814e-01 -1.21693242e+00 -6.82611465e-01 1.11299860e+00 1.30816191e-01 2.63636708e-01 -6.38478816e-01 -3.23652983e-01 3.64289373e-01 7.63219655e-01 -9.08264041e-01 -2.62960285e-01 -4.15402263e-01 -1.31994474e+00 3.80270958e-01 6.87943101e-01 6.01271316e-02 -4.39812005e-01 -1.62895143e-01 1.57074958e-01 -3.52126002e-01 -7.27542281e-01 -6.44737542e-01 -1.35425523e-01 -1.07500815e+00 -1.11543226e+00 -7.59739697e-01 -3.59036177e-01 3.41970742e-01 3.75292569e-01 8.40164423e-01 -4.22419071e-01 3.25765669e-01 3.96559447e-01 6.61465824e-02 -3.56342673e-01 -2.00458825e-01 -3.04244131e-01 3.50200474e-01 -7.01635107e-02 2.03156799e-01 -6.41796827e-01 -8.84269655e-01 4.02436882e-01 -8.03586006e-01 -2.79773682e-01 3.10570896e-01 8.49293768e-01 3.95043969e-01 -2.50335902e-01 1.01672196e+00 -1.09318054e+00 7.03841031e-01 -1.11567712e+00 -7.74137855e-01 3.26317310e-01 -7.14467168e-01 9.41569265e-03 4.18692648e-01 -6.83142781e-01 -1.51047909e+00 -5.54834247e-01 4.88308221e-01 -7.35117346e-02 -2.36317143e-01 7.68782616e-01 -5.16340375e-01 5.17398179e-01 4.88595247e-01 -7.24468172e-01 -2.87758499e-01 -5.70724189e-01 5.58890820e-01 5.13579845e-01 4.31281626e-01 -8.65144908e-01 2.77442306e-01 6.56748891e-01 3.99143726e-01 -2.75389314e-01 -8.75479162e-01 -1.73060238e-01 -3.32639545e-01 3.29142392e-01 8.71970952e-01 -9.17441845e-01 -1.30727637e+00 -9.19717774e-02 -9.34687614e-01 -3.59351009e-01 -7.65299052e-02 1.09332931e+00 -6.17917955e-01 1.22398019e-01 -5.76678097e-01 -9.87842262e-01 2.11122781e-01 -9.79639769e-01 1.03625059e+00 1.02778882e-01 -3.06041628e-01 -1.39247704e+00 3.01679909e-01 7.05088004e-02 -2.30119657e-03 4.48143095e-01 1.32224047e+00 -1.72046810e-01 -2.94322908e-01 3.77095006e-02 -4.01467651e-01 -4.59677696e-01 3.09714735e-01 4.75522224e-03 -6.64830804e-01 -1.80905491e-01 3.63495015e-02 5.45161106e-02 7.59708941e-01 1.50950968e+00 1.12623942e+00 -5.66498995e-01 -7.62473166e-01 4.29244608e-01 1.15234792e+00 2.40487665e-01 3.20720822e-01 -1.38331860e-01 5.94299495e-01 8.31022739e-01 7.18114376e-01 6.65023983e-01 5.72569370e-01 7.06192851e-01 4.08259511e-01 -3.18600446e-01 3.32602233e-01 -2.45595112e-01 7.02938512e-02 8.03569108e-02 -1.52554691e-01 -3.94330084e-01 -5.65865159e-01 7.71864891e-01 -2.00306416e+00 -9.37935114e-01 -8.67462456e-01 2.68899250e+00 9.01300848e-01 -3.40479910e-01 5.09684801e-01 -2.05858007e-01 1.28211510e+00 -4.16374892e-01 -5.54415584e-01 -4.50337976e-01 -1.64140463e-01 8.78686234e-02 8.97009611e-01 7.42549241e-01 -9.06903207e-01 4.77289081e-01 7.38620853e+00 4.82820213e-01 -6.53556883e-01 3.76915693e-01 6.46313608e-01 -2.17308238e-01 -6.05466187e-01 4.47216451e-01 -5.82333028e-01 4.26871270e-01 1.13253403e+00 -5.32960355e-01 1.86304957e-01 2.44626805e-01 1.05667698e+00 -1.99612066e-01 -1.32540417e+00 3.25785965e-01 -6.98772430e-01 -7.73401558e-01 -3.65043402e-01 4.91168052e-01 1.06179655e+00 -5.34609377e-01 1.94091469e-01 -7.47057945e-02 9.25046742e-01 -9.87408936e-01 3.95458728e-01 2.69902736e-01 1.12930548e+00 -6.79093063e-01 5.15514374e-01 1.54578805e-01 -5.17355204e-01 -3.57080489e-01 -4.11754102e-01 -4.52261597e-01 3.77792895e-01 1.03721726e+00 -3.18180263e-01 5.99026203e-01 4.50548828e-01 3.87000144e-01 1.86391026e-01 9.45008636e-01 -3.00568312e-01 9.34736311e-01 -2.05979317e-01 4.09521788e-01 -8.67619005e-04 -1.39403149e-01 3.63228440e-01 9.30793881e-01 3.82799655e-01 5.87463498e-01 -2.24801138e-01 1.00473511e+00 -1.84114218e-01 1.02068588e-01 -4.93502349e-01 2.56441712e-01 6.06135249e-01 6.27252519e-01 -4.75021213e-01 -2.59769231e-01 -4.96541232e-01 3.71513784e-01 1.25078514e-01 6.78337157e-01 -1.06723785e+00 2.67883956e-01 6.53653502e-01 5.87957986e-02 -1.68200716e-01 2.68903941e-01 -6.54635727e-01 -1.13615072e+00 -1.49441108e-01 -6.93417072e-01 9.53484774e-01 -3.33677769e-01 -1.30916798e+00 -5.61456501e-01 8.42101932e-01 -6.88326538e-01 5.96487662e-03 -1.11427985e-01 -5.14850736e-01 1.06025910e+00 -1.24618566e+00 -7.18096912e-01 1.84103608e-01 6.87319398e-01 8.77591148e-02 7.91606009e-01 6.55669451e-01 3.62640738e-01 -8.42144310e-01 3.88009012e-01 3.07956278e-01 -4.62288469e-01 1.08404005e+00 -1.28422368e+00 -5.69606312e-02 7.44485199e-01 -7.04393029e-01 9.65991616e-01 8.82090569e-01 -9.21568513e-01 -1.30246818e+00 -9.81592417e-01 1.02742386e+00 -2.95809060e-01 9.12968278e-01 -1.59768403e-01 -7.72133648e-01 9.46693540e-01 -4.01284685e-03 -3.13057989e-01 6.60659730e-01 8.85176420e-01 -3.37224513e-01 9.14006978e-02 -1.11953485e+00 8.84076595e-01 1.32866228e+00 -1.62369292e-02 -3.61086160e-01 4.36252683e-01 1.05951583e+00 -3.14720534e-02 -1.09201872e+00 5.59054554e-01 7.77042806e-01 -6.26084745e-01 9.27736223e-01 -1.26004064e+00 6.23527169e-01 1.55105710e-01 -1.55081913e-01 -1.33591342e+00 -7.43132472e-01 -5.87063551e-01 2.27031931e-01 1.16887808e+00 3.32423210e-01 -7.45392263e-01 1.48600638e-01 9.51050520e-01 1.14196628e-01 -2.38345236e-01 -7.12012947e-01 -8.46619129e-01 5.96347690e-01 -2.33364403e-01 8.45919192e-01 1.42736602e+00 3.73393774e-01 3.93226147e-01 -5.36299586e-01 5.47860026e-01 9.11910355e-01 4.70290214e-01 4.51680332e-01 -1.24796700e+00 -3.46195310e-01 -2.94813067e-01 2.48781651e-01 -7.38692641e-01 2.61954963e-01 -5.46545923e-01 -2.62793750e-01 -1.30854690e+00 1.00727391e+00 -5.14488399e-01 -3.32772434e-01 2.19473973e-01 -7.08277702e-01 -2.98035353e-01 -2.32315987e-01 -2.27611326e-02 -1.44103840e-01 4.53733444e-01 1.33619034e+00 1.21446505e-01 -3.52421492e-01 2.31449783e-01 -1.19750333e+00 6.08797669e-01 7.61602581e-01 -8.26179206e-01 -5.11802852e-01 -2.32481018e-01 2.68520117e-01 8.09755206e-01 6.69782281e-01 1.95357874e-01 -8.62630829e-02 -1.07538188e+00 -1.61702916e-01 -1.26416296e-01 -2.37660483e-01 -5.48685670e-01 3.93428952e-01 4.61202830e-01 -6.14662766e-01 -1.33401528e-01 4.82888557e-02 6.70064628e-01 1.65703177e-01 -8.97574145e-03 6.10405922e-01 2.03122124e-01 -5.22033125e-02 2.88424522e-01 -1.81410402e-01 2.96825886e-01 7.85689890e-01 1.19307086e-01 -6.71907008e-01 -5.00777543e-01 -1.00050128e+00 5.07913291e-01 2.30303019e-01 2.03458339e-01 1.43027112e-01 -1.51387095e+00 -9.05445695e-01 -4.00249243e-01 3.16310190e-02 -4.66454029e-01 6.90535247e-01 1.46085906e+00 4.22243714e-01 6.69630170e-01 8.84547606e-02 -3.01804692e-01 -1.07708848e+00 1.19385874e+00 7.29451329e-02 -1.17390282e-01 -8.80147889e-02 3.36618721e-01 1.30070639e+00 -1.13389306e-01 -3.21781822e-02 -4.39380050e-01 8.34014267e-02 -9.24009904e-02 4.91916955e-01 9.51015174e-01 -3.88508916e-01 -2.61829644e-01 -4.35044587e-01 3.33554924e-01 3.40336084e-01 -2.46151417e-01 1.14366078e+00 -5.90851486e-01 -3.70885879e-01 5.59125543e-01 1.10639751e+00 1.63637310e-01 -1.10252464e+00 -1.38443753e-01 -1.38250694e-01 -3.57966006e-01 -7.82716945e-02 -7.22600579e-01 -6.62106276e-01 6.39900565e-01 3.79173964e-01 2.51001209e-01 1.25366247e+00 1.32987648e-01 -9.53424945e-02 -2.32293203e-01 1.54865474e-01 -5.62229156e-01 -5.77825963e-01 -2.64656488e-02 6.67603791e-01 -1.20063853e+00 -8.03236887e-02 -7.85150468e-01 -1.39710069e-01 4.66865480e-01 2.22218528e-01 1.21965192e-01 7.43155301e-01 1.88196227e-01 -5.29826462e-01 -2.04526588e-01 -7.91425347e-01 -1.37842461e-01 1.76658496e-01 4.77873504e-01 6.70151532e-01 7.17864275e-01 -1.06476700e+00 7.63227105e-01 9.57485065e-02 2.12328374e-01 8.14595878e-01 4.18727487e-01 -6.88234568e-02 -9.64548290e-01 -8.61594021e-01 3.35595399e-01 -8.77566338e-01 -2.78701663e-01 -1.73377141e-01 9.69616830e-01 -1.21966630e-01 1.32618761e+00 2.88279299e-02 4.01856929e-01 5.63053012e-01 -1.87647089e-01 4.49034423e-01 -4.04521108e-01 -1.12019248e-01 4.51387495e-01 -1.63200498e-02 -5.49820364e-01 -8.49266171e-01 -9.65089619e-01 -8.96504939e-01 -7.71852672e-01 -5.30852854e-01 7.88199455e-02 8.56528878e-02 9.34500635e-01 1.78289026e-01 4.85605419e-01 7.62502074e-01 -9.60914269e-02 -6.70784891e-01 -7.52553761e-01 -8.58638644e-01 2.45709136e-01 4.73649770e-01 -9.72971499e-01 -3.59134555e-01 -1.09231053e-02]

[7.949225425720215, 5.251655101776123]

91f3b2f9-868e-425d-884e-664ed5110800

a-simple-and-general-strategy-for-referential

https://openreview.net/forum?id=IazZhsJK7wJ

https://openreview.net/pdf?id=IazZhsJK7wJ

A Simple and General Strategy for Referential Problem in Low-Resource Neural Machine Translation

This paper aims to solve a series of referential problems in sequence decoding caused by data sparsity and corpus scarce in low-resource Neural Machine Translation (NMT), including pronoun missing, reference error, bias and so on. It is difficult to find the essential reason of these problems because they are only shown in the prediction results and involve all aspects of the model. Different from the usual solutions based on complex mathematical rule setting and adding artificial features, we expect to turn the problems in the predictions into noise as much as possible, and use adversarial training to make the model find the balance between the noise and the golden samples, instead of exploring the reason of the problem during the complex training. In this paper, only a simple noise-based preprocessing operation and a slight modification of the adversarial training can make the model generalize to a series of referential problems in low-resource NMT task. On Korean-Chinese, Mongolian-Chinese and Arabic-Chinese tasks, the evaluation of BLEU score and the accuracy of pronouns in sequence have been significantly improved.

['Hongxu Hou', 'Nier Wu', 'Yatu Ji']

2021-01-01

['low-resource-neural-machine-translation']

['natural-language-processing']

[ 6.55942798e-01 2.16331601e-01 -3.05813104e-02 -4.02892441e-01 -7.99122751e-01 -3.88642639e-01 5.31031966e-01 -3.68938029e-01 -6.30346000e-01 1.19043076e+00 1.64912537e-01 -4.85909641e-01 1.14800997e-01 -4.85671610e-01 -8.85812283e-01 -7.64971316e-01 1.76647127e-01 6.17470264e-01 -1.82053939e-01 -6.82346225e-01 3.44766259e-01 3.63598042e-03 -1.11289930e+00 2.70880282e-01 1.37980437e+00 5.96555591e-01 4.11790341e-01 2.75169253e-01 -3.73097807e-01 5.97662270e-01 -8.89341414e-01 -7.32890368e-01 3.21428657e-01 -7.58429289e-01 -8.28495741e-01 -3.06404561e-01 3.00904959e-01 -1.96053475e-01 -5.16536757e-02 1.34035468e+00 8.52355838e-01 -1.06328078e-01 7.30260074e-01 -9.54851091e-01 -8.71616006e-01 8.65294278e-01 -4.38160777e-01 2.03923613e-01 2.94309050e-01 -6.92674294e-02 7.77599454e-01 -1.05292380e+00 6.65184736e-01 1.35525370e+00 7.62657464e-01 9.94105577e-01 -9.25807238e-01 -6.82476044e-01 9.75542888e-02 3.76727074e-01 -1.30260360e+00 -5.97565711e-01 5.08979142e-01 -2.19948351e-01 8.91429842e-01 6.80631399e-01 2.89624691e-01 1.60562193e+00 2.26554051e-01 8.16570580e-01 1.02643287e+00 -7.06333280e-01 -7.70067796e-02 2.08264105e-02 -1.18286416e-01 4.46889669e-01 1.47878751e-01 4.34480347e-02 -3.98481011e-01 1.82199627e-02 5.03517270e-01 -3.11338931e-01 -6.52873993e-01 3.07160109e-01 -1.47385240e+00 6.75393403e-01 1.64715145e-02 3.77174318e-01 -1.17638461e-01 -2.01743156e-01 3.75404388e-01 6.23112023e-01 4.71790582e-01 6.79606259e-01 -7.81592786e-01 -2.92845249e-01 -1.00395370e+00 1.10891007e-01 7.93701470e-01 1.34315062e+00 4.73656207e-01 2.26958141e-01 -3.04444343e-01 1.00701118e+00 -6.45302907e-02 6.62799358e-01 8.56069922e-01 -5.47192633e-01 9.49973762e-01 2.13478267e-01 2.34201401e-02 -7.72952378e-01 -1.03566438e-01 -5.50139368e-01 -8.67034256e-01 -1.59236461e-01 7.95789659e-01 -5.80884457e-01 -7.13570178e-01 1.87957251e+00 -3.91644202e-02 -1.08711012e-01 1.29788861e-01 1.23229837e+00 7.67117977e-01 6.40005052e-01 -2.25459918e-01 -6.05598748e-01 1.17412245e+00 -9.98073637e-01 -1.27238822e+00 -2.67227232e-01 7.53686368e-01 -9.96310174e-01 1.25260055e+00 2.61266589e-01 -1.09794784e+00 -3.43815476e-01 -1.12965119e+00 -2.28210852e-01 -2.22598329e-01 1.02578864e-01 5.77206612e-01 4.20624882e-01 -5.73761463e-01 8.11474502e-01 -4.97526586e-01 -4.38176125e-01 9.52948630e-02 2.46426493e-01 -2.04737350e-01 -2.57585719e-02 -1.57889223e+00 1.30560744e+00 4.17976171e-01 4.65029657e-01 -3.89699429e-01 -3.91015440e-01 -5.22077680e-01 -2.14657530e-01 5.26981413e-01 -5.00842154e-01 1.15559137e+00 -1.58033156e+00 -1.60823143e+00 5.58916986e-01 -4.00947064e-01 -2.78690428e-01 9.21520770e-01 -4.10947174e-01 -4.22999829e-01 -5.88032782e-01 -2.05549840e-02 3.99553865e-01 5.79015434e-01 -9.74422753e-01 -4.86703038e-01 -2.46519908e-01 -2.70187110e-01 2.90080875e-01 -1.17703862e-01 1.02068871e-01 -4.83147949e-01 -9.43728030e-01 2.40535945e-01 -8.07196856e-01 -5.04850410e-02 -5.54572403e-01 -4.16425556e-01 -2.99246728e-01 4.93679792e-01 -1.05418527e+00 1.28608370e+00 -1.87813127e+00 3.83617133e-01 -1.69998452e-01 -2.77593195e-01 3.82238328e-01 -3.15538526e-01 3.63764524e-01 -7.98249710e-03 3.95681649e-01 -2.45672643e-01 -4.77249660e-02 -1.34591609e-01 4.61390048e-01 -4.00954574e-01 3.07592541e-01 5.03160536e-01 1.09077179e+00 -9.15654063e-01 -4.88942534e-01 -4.06412542e-01 1.19731203e-01 -3.91643524e-01 1.71159565e-01 -3.50667745e-01 6.03026986e-01 -4.91265714e-01 5.85245073e-01 6.68285608e-01 2.23594770e-01 1.76902995e-01 -5.52801043e-02 -8.91396552e-02 6.23884678e-01 -9.19699728e-01 1.67792511e+00 -1.69496149e-01 4.97162700e-01 -1.35346860e-01 -9.99090135e-01 9.25584733e-01 3.13435525e-01 3.71927172e-02 -8.92888248e-01 2.31812119e-01 6.16542697e-01 4.71113652e-01 -8.90746295e-01 5.78529835e-01 -2.89359152e-01 1.24947459e-01 2.40217313e-01 -2.75755581e-03 2.45270923e-01 1.62633508e-02 -3.35094243e-01 8.85750353e-01 3.57792974e-01 -8.54368228e-03 -3.91278625e-01 4.52433616e-01 2.34085187e-01 1.08143556e+00 6.64402306e-01 -2.16739457e-02 8.30435872e-01 5.83206236e-01 -3.26828986e-01 -1.27196288e+00 -5.18687367e-01 -1.23467557e-01 7.98124433e-01 -5.32748140e-02 -2.19184190e-01 -9.02817249e-01 -7.39441633e-01 -3.64751101e-01 8.95012736e-01 -4.34086055e-01 -9.34829786e-02 -1.11596441e+00 -1.12453079e+00 7.65947580e-01 2.05280095e-01 4.58303690e-01 -1.01535749e+00 -4.11911160e-02 2.94656485e-01 -6.47415876e-01 -1.01262951e+00 -6.99026465e-01 3.25093865e-01 -9.79221284e-01 -7.19889045e-01 -7.42248535e-01 -9.51577783e-01 6.45506382e-01 -1.38470531e-01 1.03613842e+00 1.61432236e-01 3.18650514e-01 -4.40319628e-01 -4.48942751e-01 -6.63783371e-01 -6.76493466e-01 1.68545753e-01 1.30374908e-01 -2.21446171e-01 4.82367277e-01 -4.08357412e-01 -1.45530686e-01 2.72039443e-01 -5.02691984e-01 2.06555113e-01 6.94290698e-01 1.34616995e+00 4.01289195e-01 -2.23296091e-01 8.92187119e-01 -1.08625519e+00 6.91683471e-01 -3.85169059e-01 -3.69384438e-01 4.73936230e-01 -6.69664323e-01 2.48385713e-01 7.04131782e-01 -8.99633765e-01 -8.30430746e-01 -2.71965504e-01 -3.68082464e-01 -2.66857445e-01 1.93748578e-01 5.78662395e-01 -5.34696817e-01 1.94789901e-01 6.00616515e-01 5.30079126e-01 9.77507532e-02 -6.81914985e-01 6.72322586e-02 7.99274981e-01 3.05495709e-01 -5.81323206e-01 8.01277876e-01 -2.45200455e-01 -2.57124066e-01 -5.87643147e-01 -8.09209168e-01 1.19333610e-01 -6.29710436e-01 1.11924298e-01 7.29148090e-01 -7.70959735e-01 -4.47662324e-01 3.47228289e-01 -1.63998437e+00 -1.31287977e-01 -1.48109168e-01 4.39265877e-01 -4.06053901e-01 4.32522565e-01 -5.70310950e-01 -6.68843031e-01 -5.22772670e-01 -1.23695552e+00 7.31574297e-01 -9.29817483e-02 -2.16773778e-01 -7.95939684e-01 -2.84211487e-01 3.53501379e-01 4.97939527e-01 -8.51198509e-02 1.19468069e+00 -9.26512837e-01 -5.55033207e-01 -4.27733511e-02 -9.45072845e-02 6.04397893e-01 1.93994083e-02 -3.22241247e-01 -8.15124512e-01 -3.00285220e-01 4.71723706e-01 -6.45507351e-02 6.91274583e-01 -9.57602113e-02 1.03459942e+00 -7.50651896e-01 -4.69605438e-02 7.42561281e-01 1.35618043e+00 3.31074536e-01 8.19201648e-01 4.52370137e-01 7.15239346e-01 5.76767266e-01 7.92112052e-01 -8.71437490e-02 1.99284196e-01 7.80212939e-01 2.94626981e-01 1.03818318e-02 -6.77667335e-02 -3.76336843e-01 5.22434652e-01 1.39743900e+00 -3.00203055e-01 -3.38217288e-01 -8.04102063e-01 2.31197327e-01 -1.84075308e+00 -8.39580655e-01 -3.12314719e-01 2.31605077e+00 1.27780437e+00 -4.68477048e-02 -3.64017814e-01 6.00402663e-03 8.30198944e-01 -1.29026979e-01 -3.47847730e-01 -5.39779782e-01 -5.29994726e-01 -5.89555651e-02 6.22501910e-01 4.69245553e-01 -7.76543140e-01 1.11979818e+00 6.43002319e+00 1.26762521e+00 -1.24902654e+00 3.23648900e-01 5.39620280e-01 6.43504411e-02 -4.10972089e-01 -1.88837852e-02 -8.68239284e-01 6.21925771e-01 8.80299509e-01 3.96359675e-02 7.66118228e-01 3.79119456e-01 2.24703476e-01 1.56212807e-01 -1.39401531e+00 8.95516157e-01 2.52594590e-01 -9.52108026e-01 2.42302999e-01 -1.25714377e-01 5.43076992e-01 -2.58055162e-02 -1.47077769e-01 5.25927961e-01 -2.39115179e-01 -1.27328598e+00 1.00850451e+00 6.01758301e-01 7.34036922e-01 -4.79475528e-01 1.03313673e+00 8.40094864e-01 -2.12932572e-01 -1.35416407e-02 -6.58570886e-01 -2.59856910e-01 -1.20164253e-01 3.88742715e-01 -7.06649899e-01 6.63194180e-01 2.82646924e-01 3.99510413e-01 -3.19672465e-01 7.49812782e-01 -2.33201489e-01 7.45302439e-01 -1.46299347e-01 -3.37256461e-01 5.08875139e-02 -3.00913393e-01 7.16861963e-01 1.23958516e+00 5.94156325e-01 1.51838148e-02 -2.67454505e-01 9.37547684e-01 -1.21242315e-01 4.85938311e-01 -5.28026640e-01 -5.02277631e-03 5.06727695e-01 8.62812757e-01 -1.11655168e-01 -1.32244125e-01 -4.17637080e-01 1.17104805e+00 4.48418081e-01 4.69988704e-01 -8.15733850e-01 -2.86093295e-01 5.04126072e-01 1.07890882e-01 1.58948526e-01 -2.31906518e-01 -6.48075640e-01 -1.53861690e+00 3.81323278e-01 -1.50710821e+00 -1.30999774e-01 -5.27614594e-01 -1.32473540e+00 6.28710449e-01 -4.51706827e-01 -1.29833901e+00 -3.27844262e-01 -4.47477400e-01 -5.40950835e-01 1.31640089e+00 -1.46417570e+00 -9.75243151e-01 3.26094598e-01 3.44615012e-01 7.11530626e-01 -4.92510110e-01 1.01381242e+00 5.74893475e-01 -7.21794069e-01 1.02805948e+00 3.65447462e-01 3.27199161e-01 8.88288438e-01 -1.05628228e+00 4.03971553e-01 9.49610114e-01 -6.54612929e-02 7.22262204e-01 8.87945950e-01 -6.65155172e-01 -1.53373492e+00 -8.22577655e-01 1.60167754e+00 -6.26875997e-01 6.29373968e-01 -6.32208169e-01 -1.00963128e+00 5.07975459e-01 3.83512944e-01 -3.43963027e-01 2.82120764e-01 1.63612962e-01 -2.23758984e-02 1.41894087e-01 -8.99921536e-01 8.31699729e-01 1.29424465e+00 -1.67424709e-01 -8.37701321e-01 4.01091844e-01 9.02245343e-01 -6.88919306e-01 -6.90146863e-01 4.44133580e-01 3.17473650e-01 -3.48624229e-01 6.89403534e-01 -1.07028449e+00 6.23247623e-01 -1.19046859e-01 -2.68585145e-01 -1.30943131e+00 -9.55427140e-02 -7.55798697e-01 2.00863659e-01 1.49604428e+00 8.08938324e-01 -5.67898452e-01 4.86267149e-01 2.85587579e-01 -2.96344131e-01 -6.97410285e-01 -1.31699538e+00 -8.25812578e-01 6.24710321e-01 -2.42870077e-01 7.08894372e-01 1.21243429e+00 -8.98808613e-02 6.78200066e-01 -7.38396108e-01 -1.29515827e-02 3.44491035e-01 -2.07575828e-01 5.71301758e-01 -9.95040357e-01 -2.54294127e-01 -2.63775140e-01 6.08159713e-02 -1.26183379e+00 1.38506413e-01 -1.03417516e+00 1.24369636e-01 -1.16733837e+00 1.00638516e-01 -5.94751358e-01 1.27421618e-02 5.16132295e-01 -4.32344437e-01 -4.44768369e-02 2.33343691e-01 4.13290083e-01 -2.81110227e-01 5.80535173e-01 1.68644714e+00 -1.61773324e-01 1.42144069e-01 8.28350708e-02 -7.84052551e-01 5.79922378e-01 6.52755320e-01 -6.48799479e-01 -2.45987792e-02 -9.40908432e-01 3.92319441e-01 2.94494897e-01 9.70373824e-02 -4.61511552e-01 1.59705475e-01 -1.78040430e-01 3.62619817e-01 -3.80820006e-01 9.15838704e-02 -9.09977078e-01 -1.10978745e-01 4.09389496e-01 -3.41057450e-01 2.13648140e-01 -3.07379980e-02 2.43298993e-01 -1.32672414e-01 -6.40262187e-01 5.20901442e-01 -2.14214131e-01 -3.77510041e-01 -4.06162031e-02 -2.43968859e-01 2.60623336e-01 5.34871399e-01 -2.09500805e-01 -4.51347619e-01 -1.80615425e-01 -6.37312055e-01 2.06596196e-01 2.86852211e-01 6.91805780e-01 1.80887014e-01 -1.43776560e+00 -1.08055568e+00 3.40439647e-01 -2.09592447e-01 2.54105907e-02 -5.53536266e-02 9.87425268e-01 -2.99536914e-01 3.90561074e-01 -2.14516193e-01 -3.26089799e-01 -1.05877852e+00 4.59937215e-01 4.36787218e-01 -2.15778738e-01 -4.07330185e-01 7.14781106e-01 -1.75229728e-01 -6.29432440e-01 4.60282594e-01 -1.59294099e-01 -1.59802258e-01 1.58301797e-02 2.78280288e-01 2.15639994e-01 2.35790059e-01 -5.89644909e-01 -1.75602272e-01 3.39486241e-01 -5.45972176e-02 1.34572014e-01 1.26082814e+00 -2.93861747e-01 -3.34710091e-01 4.80374902e-01 9.54802275e-01 1.86819747e-01 -8.18959713e-01 -3.40026021e-01 2.89778531e-01 -2.24757805e-01 -3.06362510e-01 -1.15487599e+00 -8.07259023e-01 1.01010001e+00 5.40089905e-01 -1.03784755e-01 1.03991377e+00 -4.30314600e-01 8.85586500e-01 4.65530246e-01 5.25813401e-01 -1.21688390e+00 -4.16201979e-01 1.03860486e+00 1.12195563e+00 -1.39601755e+00 -3.00746709e-01 -2.77248323e-01 -5.71190298e-01 1.14214814e+00 4.87169325e-01 1.29933417e-01 1.33908749e-01 4.23336089e-01 3.20691139e-01 3.88350815e-01 -7.85963535e-01 1.37340173e-01 2.59861588e-01 5.44817686e-01 7.02525258e-01 -1.33943588e-01 -1.08213902e+00 9.67656791e-01 -5.09295762e-01 -5.78301214e-02 4.16723311e-01 5.51835120e-01 -3.09676707e-01 -1.46621275e+00 -3.53377193e-01 3.19475770e-01 -9.25563097e-01 -5.78793287e-01 -2.71079302e-01 8.28464210e-01 2.01708436e-01 7.53763974e-01 -8.73214379e-02 -5.37388742e-01 4.52170640e-01 3.81061137e-01 3.92387986e-01 -4.31257665e-01 -6.00639701e-01 1.32274941e-01 3.85438383e-01 -2.74601340e-01 -2.38987535e-01 -7.79997647e-01 -1.10617554e+00 -3.71917725e-01 -5.26912749e-01 2.08208427e-01 7.16664851e-01 1.19064415e+00 3.97346951e-02 5.60815156e-01 2.96432465e-01 -5.18581331e-01 -1.15101767e+00 -1.57128739e+00 -2.62435734e-01 3.79904747e-01 2.82969922e-01 -3.00176471e-01 -3.47980767e-01 -1.83648705e-01]

[11.598018646240234, 10.10793685913086]

74ff16bc-cbfd-44dc-906a-8a8de73e4cc7

performance-modeling-of-data-storage-systems

2307.02073

https://arxiv.org/abs/2307.02073v1

https://arxiv.org/pdf/2307.02073v1.pdf

Performance Modeling of Data Storage Systems using Generative Models

High-precision modeling of systems is one of the main areas of industrial data analysis. Models of systems, their digital twins, are used to predict their behavior under various conditions. We have developed several models of a storage system using machine learning-based generative models. The system consists of several components: hard disk drive (HDD) and solid-state drive (SSD) storage pools with different RAID schemes and cache. Each storage component is represented by a probabilistic model that describes the probability distribution of the component performance in terms of IOPS and latency, depending on their configuration and external data load parameters. The results of the experiments demonstrate the errors of 4-10 % for IOPS and 3-16 % for latency predictions depending on the components and models of the system. The predictions show up to 0.99 Pearson correlation with Little's law, which can be used for unsupervised reliability checks of the models. In addition, we present novel data sets that can be used for benchmarking regression algorithms, conditional generative models, and uncertainty estimation methods in machine learning.

['Mikhail Hushchyn', 'Artem Ryzhikov', 'Aziz Temirkhanov', 'Abdalaziz Rashid Al-Maeeni']

2023-07-05

['benchmarking', 'benchmarking']

['miscellaneous', 'robots']

[-6.51812136e-01 -1.10928118e-01 -9.40091684e-02 -4.30726945e-01 -3.67359728e-01 -8.60688686e-02 6.18941844e-01 2.05530003e-01 2.05298424e-01 9.23878014e-01 -4.01594132e-01 -3.99828762e-01 -3.15810323e-01 -7.64232993e-01 -6.46801949e-01 -9.25573289e-01 -1.76581025e-01 1.10403728e+00 7.05030501e-01 9.30233970e-02 3.00222307e-01 6.95818841e-01 -2.01737976e+00 8.49809349e-02 4.31055069e-01 1.47660196e+00 3.12031537e-01 5.82149923e-01 -1.74759850e-02 6.90633595e-01 -8.89629066e-01 3.34529757e-01 -1.54436216e-01 1.14726521e-01 -3.96395415e-01 -3.12101126e-01 -6.28810585e-01 -2.06588566e-01 -5.61059058e-01 7.12156117e-01 1.40051574e-01 -3.82367760e-01 9.89192903e-01 -1.44212925e+00 -2.24117473e-01 7.02706635e-01 -1.90000072e-01 -4.83167954e-02 -7.17147300e-03 1.69008866e-01 2.11768284e-01 -2.53439456e-01 2.56273717e-01 8.45572293e-01 8.93224254e-02 2.31931955e-02 -1.33779669e+00 -3.12828064e-01 -6.53938234e-01 1.77076638e-01 -1.69600475e+00 -1.83351651e-01 2.89669603e-01 -6.55186176e-01 1.27006972e+00 3.15713584e-01 2.04529986e-01 4.07569140e-01 1.15175343e+00 5.66107370e-02 1.27219498e+00 -5.28917789e-01 5.60115159e-01 7.32152522e-01 7.36456454e-01 2.41116226e-01 5.09399891e-01 4.02301490e-01 -3.54388952e-01 -3.29119891e-01 5.58785737e-01 -3.08662981e-01 1.85558662e-01 -5.09106696e-01 -5.63858986e-01 4.31346595e-01 -1.48294583e-01 2.59162039e-01 -3.70386332e-01 3.06477368e-01 3.31591547e-01 1.32636055e-01 1.62685484e-01 1.84679702e-01 -4.90295202e-01 -4.05011177e-01 -9.65867996e-01 2.64102727e-01 1.10947788e+00 1.16421425e+00 6.92804635e-01 1.65610626e-01 -1.52996331e-01 4.98840183e-01 6.35980964e-01 9.28800225e-01 4.55864519e-01 -7.31092274e-01 -1.21705875e-01 3.25177521e-01 3.99154603e-01 -2.97746480e-01 -3.79148781e-01 -4.78240997e-02 -5.04192472e-01 2.86629140e-01 -3.90629210e-02 2.14145720e-01 -9.32529628e-01 9.81447875e-01 -1.21883161e-01 -1.56267926e-01 -2.36256987e-01 3.94247949e-01 8.63402784e-02 9.10379589e-01 -9.99046564e-02 -5.85422456e-01 9.09963548e-01 -2.15675324e-01 -9.62129056e-01 4.65673268e-01 5.14151692e-01 -9.71730173e-01 6.55546367e-01 7.96047807e-01 -1.17061639e+00 -8.04296017e-01 -1.27139926e+00 5.90155780e-01 -2.17117608e-01 1.09897867e-01 5.81960641e-02 8.07213664e-01 -8.48528087e-01 9.01946425e-01 -1.20318913e+00 -3.55268568e-02 -2.74795592e-01 4.19797659e-01 3.53247553e-01 9.69211385e-02 -8.81951272e-01 1.08486414e+00 3.32216799e-01 -1.42407596e-01 -9.90660846e-01 -5.01418710e-01 -1.30039811e-01 1.37804061e-01 -4.51568374e-03 -3.08163196e-01 1.14729178e+00 2.82490999e-01 -1.46919131e+00 4.62914377e-01 1.08054087e-01 -4.20820355e-01 -6.34100288e-02 -2.57933289e-01 -9.66757596e-01 -3.10857356e-01 -4.25053775e-01 -3.57971042e-01 7.42561519e-01 -1.42090070e+00 1.88132562e-02 -5.08850932e-01 -7.66756892e-01 -7.57242382e-01 1.83354080e-01 1.71621978e-01 -2.70110190e-01 1.69427812e-01 -4.38868776e-02 -1.10070288e+00 7.41660148e-02 -7.02387571e-01 -8.29373062e-01 -1.01556249e-01 7.76694596e-01 -6.52656674e-01 1.55517066e+00 -1.94453919e+00 -2.86807790e-02 6.30249023e-01 -1.63822412e-01 1.73087522e-01 5.89324832e-01 7.57170081e-01 3.61287966e-02 -2.72601061e-02 2.03882977e-01 4.48968355e-03 2.32993305e-01 2.11194143e-01 -3.58697355e-01 1.25698686e-01 -4.02613096e-02 3.44061792e-01 1.82328988e-02 -1.56847417e-01 5.41774929e-01 2.17624843e-01 3.54539394e-01 7.09810615e-01 -4.18862402e-01 -1.37180248e-02 -3.97353917e-01 3.70150298e-01 6.79708362e-01 -3.40500712e-01 3.16500038e-01 -3.71463895e-01 -3.45277250e-01 3.95034939e-01 -1.25114202e+00 7.42076993e-01 -5.37424564e-01 1.76809505e-01 -2.20207915e-01 -6.68588579e-01 1.49946415e+00 -5.64685240e-02 4.20890927e-01 -7.82439590e-01 2.06565499e-01 2.13027045e-01 4.67229486e-02 -3.95786047e-01 6.26318991e-01 7.70727918e-02 7.39564141e-03 6.85624301e-01 -2.98998561e-02 -4.47827190e-01 7.74034485e-02 1.70449704e-01 1.10162342e+00 1.45899683e-01 -1.80569962e-01 -6.40410185e-01 1.71870500e-01 -2.52879530e-01 9.75634083e-02 2.24070162e-01 3.41724843e-01 4.09681439e-01 8.56277227e-01 -9.35841445e-03 -1.40703416e+00 -1.43572879e+00 -6.61334276e-01 2.95954525e-01 1.79925814e-01 -6.66219950e-01 -6.42243624e-01 1.96517378e-01 3.58238906e-01 1.24008846e+00 -1.93113938e-01 -4.22555238e-01 -1.93476111e-01 -1.01644707e+00 2.04220917e-02 6.70542836e-01 -2.12374762e-01 -5.41740000e-01 -3.14301550e-01 1.98888794e-01 5.59518397e-01 -9.95837569e-01 4.97275621e-01 7.26759911e-01 -9.35857415e-01 -9.28350687e-01 1.13713734e-01 1.45391494e-01 3.05943847e-01 -1.70433775e-01 1.48332727e+00 1.21510744e-01 -6.15866303e-01 3.06720406e-01 -4.07891162e-02 -3.51901442e-01 -1.01346421e+00 -2.22429693e-01 5.37751496e-01 -5.03442347e-01 4.03563738e-01 -4.63213205e-01 -2.66161919e-01 8.02780628e-01 -6.86966717e-01 -5.76696217e-01 3.77587020e-01 4.60581392e-01 1.02200842e+00 5.17561674e-01 2.23081216e-01 -6.24397933e-01 5.99783182e-01 -7.92485654e-01 -1.17563367e+00 5.39066970e-01 -1.54703772e+00 5.22696495e-01 3.89699370e-01 -1.70257777e-01 -1.01988375e+00 -2.42260456e-01 3.94139647e-01 -5.16886055e-01 -1.17617793e-01 1.93989858e-01 -3.75938177e-01 4.46556628e-01 5.68800509e-01 1.41235307e-01 3.11857492e-01 -6.30182087e-01 -1.08782157e-01 1.09597111e+00 4.25748646e-01 -9.36478734e-01 5.39464056e-01 -1.67821676e-01 9.13563371e-02 -8.06335747e-01 -6.76415935e-02 -3.22064720e-02 -5.83293021e-01 -2.35169023e-01 6.60387874e-01 -7.24446535e-01 -8.12155843e-01 6.45928741e-01 -6.48781478e-01 -3.48102123e-01 -2.30444744e-01 5.38704336e-01 -8.03575397e-01 9.17097479e-02 -7.71974444e-01 -1.25237978e+00 -2.19798014e-01 -1.30068231e+00 1.02490759e+00 2.72267133e-01 -1.45868391e-01 -6.25666082e-01 9.75140855e-02 1.47010341e-01 6.69968843e-01 -2.81582892e-01 1.33246160e+00 -4.42017078e-01 -9.63664889e-01 -5.01594245e-01 -2.90314015e-02 7.79741645e-01 -1.10920213e-01 8.71561229e-01 -6.53116286e-01 -3.72273624e-01 5.75105026e-02 -2.25083008e-01 2.28567168e-01 5.93203127e-01 1.41049361e+00 3.75465095e-01 -7.60406435e-01 2.35445201e-02 1.56677270e+00 4.09560114e-01 1.12443066e+00 -5.27044646e-02 1.22910239e-01 4.82068270e-01 9.04312253e-01 7.65453517e-01 -2.18975410e-01 9.88585472e-01 5.23521364e-01 6.75973177e-01 1.94760576e-01 2.80799326e-02 3.37994009e-01 8.17048550e-01 4.85915765e-02 -3.37122887e-01 -1.06764245e+00 1.31396893e-02 -1.52161908e+00 -5.88401258e-01 -7.40304589e-01 2.83745289e+00 4.41951096e-01 6.72708094e-01 2.84174055e-01 1.57451466e-01 6.07176006e-01 -4.42987502e-01 -5.62886178e-01 -5.80530822e-01 3.44934016e-01 3.12743932e-01 7.15167999e-01 1.28441229e-01 -3.38466167e-01 2.82002777e-01 7.09822607e+00 6.28000736e-01 -9.97820914e-01 -1.44102439e-01 5.73223293e-01 -1.07516594e-01 -4.72225845e-02 -4.00339440e-02 -1.33775294e+00 1.14909160e+00 2.15089798e+00 -2.96984434e-01 3.10540706e-01 8.19930971e-01 1.75649762e-01 -7.81928837e-01 -1.14151204e+00 6.13862455e-01 -2.58655071e-01 -1.04087448e+00 -3.66868258e-01 4.92937863e-01 5.17446876e-01 1.82346329e-01 -3.35706733e-02 2.31303185e-01 2.65462518e-01 -9.27245617e-01 5.29926360e-01 1.25800490e+00 8.06884110e-01 -7.42512584e-01 1.02280772e+00 2.45728374e-01 -5.88073075e-01 5.30755031e-04 -5.26037872e-01 2.40762413e-01 4.81003940e-01 1.20146811e+00 -6.86035454e-01 4.88704413e-01 8.45659614e-01 -2.63676584e-01 -4.37474459e-01 7.91835010e-01 2.49532178e-01 7.37648427e-01 -6.31287932e-01 -2.60946393e-01 -5.67349315e-01 -2.05006346e-01 -5.06872572e-02 5.07634878e-01 6.76006675e-01 -3.08425784e-01 -1.90923646e-01 1.20678544e+00 5.45282900e-01 -3.03937167e-01 -3.32111835e-01 -1.68617427e-01 8.40009511e-01 9.54730511e-01 -3.69940639e-01 -1.88401937e-01 -2.83173807e-02 7.03289583e-02 -2.71709025e-01 1.23822410e-02 -9.68181312e-01 -2.28241399e-01 5.25242686e-01 8.82727027e-01 2.20089048e-01 -6.27136230e-01 -3.58358473e-01 -5.47253668e-01 1.77243389e-02 -3.66420895e-01 -2.85318762e-01 -1.17756271e+00 -1.16831720e+00 7.54568100e-01 5.57134271e-01 -1.03128648e+00 -6.98428094e-01 -8.81932735e-01 -4.95024770e-01 1.22830319e+00 -8.34775269e-01 -7.09009826e-01 -1.78464472e-01 2.71765739e-01 -2.00247794e-01 -5.62595010e-01 8.04498076e-01 3.50892916e-02 -7.22993493e-01 1.80551231e-01 8.62441361e-01 -6.95999503e-01 6.40627265e-01 -1.12760174e+00 1.97519794e-01 1.87084466e-01 -3.04997325e-01 5.31072199e-01 1.20319009e+00 -8.26328754e-01 -1.42718220e+00 -7.80911684e-01 4.31777179e-01 -5.33919275e-01 8.78974140e-01 -1.59871474e-01 -1.22373402e+00 3.74750763e-01 2.55961623e-02 4.15879488e-02 6.74473882e-01 3.45673263e-01 -1.79552156e-02 -4.77868021e-01 -9.79878247e-01 -3.88672709e-01 1.29197553e-01 -4.71670240e-01 -1.02341667e-01 5.92682242e-01 3.97613347e-01 -2.03082055e-01 -1.49178553e+00 4.04607773e-01 4.69700515e-01 -1.14945686e+00 6.76907182e-01 -4.77337211e-01 2.96030939e-01 -2.53193498e-01 -4.96762842e-01 -1.12305951e+00 -1.68968648e-01 -1.63259402e-01 -6.53014123e-01 1.34844279e+00 4.73532945e-01 -4.39296156e-01 3.91634047e-01 1.17586625e+00 -2.19503194e-01 -7.94151843e-01 -8.31470609e-01 -1.29708803e+00 -9.22448467e-03 -4.38763171e-01 7.35857308e-01 2.77942643e-02 -8.97274911e-02 1.52736276e-01 -2.53735572e-01 1.43657029e-01 6.25699401e-01 1.67074189e-01 7.23006845e-01 -1.25808239e+00 -7.21831143e-01 -5.27998619e-02 -5.30424654e-01 -4.71140951e-01 4.37137708e-02 -2.17401490e-01 4.21569720e-02 -9.79070127e-01 2.70103544e-01 -1.01213777e+00 -3.69695038e-01 -1.65462017e-01 5.21983802e-01 -4.19882059e-01 -1.94404632e-01 4.70629483e-01 -2.00644404e-01 5.01232624e-01 3.31654549e-01 1.96158335e-01 -9.48346555e-02 3.84387732e-01 1.89016387e-01 1.39951989e-01 9.39862311e-01 -3.39012355e-01 -4.79042590e-01 -3.08117270e-02 1.13159440e-01 4.40948278e-01 2.01457635e-01 -1.34031594e+00 -1.92215964e-02 7.55156204e-03 2.72601157e-01 -1.09503162e+00 4.20749784e-01 -8.87496352e-01 1.05153847e+00 4.20149803e-01 -3.10744122e-02 3.31086338e-01 1.41856760e-01 5.32347441e-01 -2.19692603e-01 -5.72184026e-01 7.19745815e-01 4.29403454e-01 -1.25672221e-01 -1.42903075e-01 -3.78250003e-01 -6.94757879e-01 1.30735815e+00 4.97291982e-02 -5.57248235e-01 -2.12583780e-01 -8.10469210e-01 -2.55285539e-02 6.24022841e-01 2.88536280e-01 2.83688545e-01 -1.13390768e+00 -2.93624818e-01 4.41455901e-01 6.23301044e-02 -5.23377657e-01 3.54574740e-01 6.69426084e-01 -5.88871062e-01 7.51964688e-01 -2.91828871e-01 -7.42443383e-01 -1.08813858e+00 9.81668830e-01 1.97441250e-01 -2.89867878e-01 1.17505156e-01 1.76338047e-01 -4.76996630e-01 1.84415683e-01 -5.85061014e-02 -2.62711614e-01 3.23560774e-01 -4.61945444e-01 3.28689367e-01 7.19970942e-01 5.83233476e-01 -2.59849072e-01 -3.02681655e-01 2.79107481e-01 -1.24538034e-01 1.62128210e-01 1.13785613e+00 -9.63424370e-02 -4.29200500e-01 1.06731141e+00 8.72527421e-01 -3.83858562e-01 -6.32815957e-01 -1.49620503e-01 5.94913810e-02 -2.64143020e-01 2.87838817e-01 -1.00382769e+00 -9.46875036e-01 9.24981654e-01 1.02402091e+00 7.72946715e-01 8.24684381e-01 2.45698601e-01 3.69005978e-01 9.46891680e-02 8.97540510e-01 -1.41313183e+00 -1.93401366e-01 1.69294551e-01 7.22105145e-01 -7.68384457e-01 3.65982801e-01 -3.56330037e-01 -3.13300550e-01 1.17017746e+00 6.38144493e-01 5.52391708e-02 1.40006483e+00 1.06749737e+00 -4.62472200e-01 -9.52418894e-02 -1.19210291e+00 3.63015473e-01 1.94142740e-02 5.04416645e-01 5.40986419e-01 5.59882581e-01 -3.77611339e-01 7.09974527e-01 -6.76622614e-03 1.03075495e-02 5.00024378e-01 8.32448065e-01 -6.16356373e-01 -1.54616225e+00 -4.21467960e-01 8.24379802e-01 -1.16569186e-02 3.65347207e-01 1.89539954e-01 7.71446407e-01 -2.79352099e-01 8.04762602e-01 3.69886816e-01 -8.30881536e-01 2.42756814e-01 2.89747506e-01 5.40501475e-01 -4.40972149e-01 -8.13380405e-02 -2.63145287e-02 1.33358687e-01 -6.03584766e-01 1.80812284e-01 -8.90971303e-01 -1.33265793e+00 -5.35895467e-01 -6.58414900e-01 2.34070003e-01 1.33112979e+00 6.62140369e-01 5.41525304e-01 7.57739305e-01 8.68429184e-01 -4.85743076e-01 -1.16088879e+00 -1.17337942e+00 -1.37858629e+00 -1.24910027e-01 -5.27145803e-01 -1.05998981e+00 -3.18952292e-01 -6.58166111e-02]

[6.641984939575195, 2.845038890838623]

06136303-8f67-4e26-b3b0-3d92568c8e28

friends-in-need-how-chaperonins-recognize-and

2211.08623

https://arxiv.org/abs/2211.08623v1

https://arxiv.org/pdf/2211.08623v1.pdf

Friends in need: how chaperonins recognize and remodel proteins that require folding assistance

Chaperonins are biological nanomachines that help newly translated proteins to fold by rescuing them from kinetically trapped misfolded states. Protein folding assistance by the chaperonin machinery is obligatory in vivo for a subset of proteins in the bacterial proteome. Chaperonins are large oligomeric complexes, with unusual seven fold symmetry (group I) or eight/nine fold symmetry (group II), that form double-ring constructs, enclosing a central folding chamber. Dramatic large-scale conformational changes, that take place during ATP-driven cycles, allow chaperonins to bind misfolded proteins, encapsulate them into the expanded cavity and release them back into the cellular environment, regardless of whether they are folded or not. The theory associated with the iterative annealing mechanism, which incorporated the conformational free energy landscape description of protein folding, \textit{quantitatively} explains most, if not all, the available data. Misfolded conformations are associated with low energy minima in a rugged energy landscape. Random disruptions of these low energy conformations result in higher free energy, less folded, conformations that can stochastically partition into the native state. Group I chaperonins (GroEL homologues in eubacteria and endosymbiotic organelles), recognize a large number of misfolded proteins non-specifically and operate through highly coordinated cooperative motions. By contrast, the less well understood group II chaperonins (CCT in Eukarya and thermosome/TF55 in Archaea), assist a selected set of substrate proteins. Chaperonins are implicated in bacterial infection, autoimmune disease, as well as protein aggregation and degradation diseases. Understanding the chaperonin mechanism and their substrates is important not only for the fundamental aspect of cellular protein folding, but also for effective therapeutic strategies.

['D. Thirumalai', 'George H. Lorimer', 'George Stan']

2022-11-16

['protein-folding']

['natural-language-processing']

[ 1.65860891e-01 -2.93488920e-01 -1.12616926e-01 -2.77548563e-02 2.82494634e-01 -9.92530763e-01 -4.89831679e-02 1.40640855e-01 -2.67096698e-01 1.15105855e+00 5.13297468e-02 -6.99748814e-01 4.30592388e-01 -3.78039092e-01 -6.96559131e-01 -1.04185069e+00 -2.54326314e-01 5.82031429e-01 8.42975900e-02 -2.36007437e-01 8.97506475e-02 6.36325181e-01 -1.02459431e+00 1.79001495e-01 1.11524057e+00 1.81188256e-01 5.66745758e-01 7.01876998e-01 -2.74659097e-01 1.74761593e-01 -2.11107895e-01 7.66466111e-02 5.42184375e-02 -9.37192798e-01 -6.76256001e-01 1.07892960e-01 -3.54126275e-01 2.99680680e-01 3.00741136e-01 4.83334094e-01 3.95255476e-01 -1.54885918e-01 6.44422352e-01 -3.17180276e-01 -7.49829471e-01 -4.40152645e-01 3.44772860e-02 1.27974376e-01 4.41308320e-01 4.88003403e-01 6.16448879e-01 -1.14908147e+00 1.08223271e+00 1.00506663e+00 5.56549311e-01 6.83325648e-01 -1.60056758e+00 -7.19688460e-02 -5.19717753e-01 -2.51573831e-01 -7.71860600e-01 -4.98137325e-01 1.09934211e-01 -8.42023194e-01 1.63598979e+00 2.78272629e-01 6.94120347e-01 6.14013672e-01 1.34115982e+00 -8.88896808e-02 1.34664547e+00 -2.05064952e-01 6.57808423e-01 -5.38867950e-01 3.03323567e-01 5.81037700e-01 6.91577971e-01 -1.99935526e-01 -4.15030152e-01 -7.42713928e-01 2.07675174e-01 5.04931748e-01 -5.54908156e-01 -5.72134137e-01 -9.69483793e-01 5.48703134e-01 -2.80363113e-03 3.50941658e-01 -7.92359769e-01 -1.69416830e-01 4.55115795e-01 2.07676902e-01 -2.39960682e-02 4.44482625e-01 -8.19863498e-01 -2.54932106e-01 -3.40366930e-01 2.02052407e-02 9.25189257e-01 1.61622375e-01 9.19301331e-01 -6.43186152e-01 5.85058808e-01 4.66330618e-01 3.98217767e-01 4.68432248e-01 5.83537340e-01 -3.79506648e-01 3.06573026e-02 1.04157007e+00 5.20281792e-01 -6.29054308e-01 -6.71179235e-01 2.98950583e-01 -7.35238314e-01 -1.27815709e-01 7.72376955e-01 -1.39453217e-01 -8.18418622e-01 1.64948952e+00 3.98325294e-01 -3.84355068e-01 2.16654152e-01 1.09698164e+00 1.24751054e-01 6.08156323e-01 5.98359108e-01 -1.03304052e+00 1.84926951e+00 -2.60953546e-01 -4.93692219e-01 6.89798407e-03 4.70076293e-01 -1.03809977e+00 7.98566997e-01 -1.00399487e-01 -7.39249825e-01 -8.09859037e-02 -1.09257829e+00 1.01440236e-01 -2.67347664e-01 -2.26755619e-01 4.87618953e-01 5.85630715e-01 -6.54049754e-01 1.09703958e+00 -1.38205886e+00 -1.07208943e+00 -1.95469260e-01 3.53565246e-01 -3.83699924e-01 7.57262716e-03 -9.91101444e-01 1.08971739e+00 2.00078428e-01 -1.12658486e-01 -4.58801985e-01 -6.65129423e-02 -4.81463164e-01 -9.65271071e-02 -4.07220483e-01 -8.57529998e-01 5.68651378e-01 -4.47017163e-01 -1.49477375e+00 9.40279424e-01 -6.55327380e-01 -2.03548759e-01 -2.40959361e-01 -7.03271180e-02 5.26529551e-02 1.96837097e-01 4.73447107e-02 -5.37204407e-02 1.52618125e-01 -7.68839061e-01 2.34940171e-01 -8.39602768e-01 -8.03219259e-01 1.98464900e-01 7.37496793e-01 4.83280867e-01 7.85515130e-01 -2.29746297e-01 9.92775321e-01 -1.10865426e+00 -7.04433978e-01 -3.89825433e-01 -1.15163386e-01 -2.42515713e-01 9.99201596e-01 -3.71469826e-01 5.67386627e-01 -1.53159797e+00 4.99955595e-01 -2.00943708e-01 1.81755081e-01 3.67871851e-01 5.05806327e-01 1.19774592e+00 -1.66520029e-01 1.33469775e-01 -3.28303188e-01 4.56225634e-01 -3.56163085e-01 3.60891968e-01 -4.36544865e-01 9.79690254e-01 1.59450173e-01 7.74398267e-01 -9.89481986e-01 2.14104056e-01 1.17217228e-01 3.87907416e-01 -5.68592958e-02 1.80623040e-01 -1.43281311e-01 8.22067559e-01 -4.14584011e-01 9.50673580e-01 8.47184718e-01 -5.97766697e-01 1.25153792e+00 4.83373776e-02 -3.26040447e-01 3.64901245e-01 -4.37840492e-01 1.02955711e+00 2.75655866e-01 1.69081181e-01 7.96176642e-02 -5.23218334e-01 1.07912707e+00 3.54102969e-01 4.01077241e-01 -2.54176795e-01 1.49976835e-01 6.35923922e-01 1.63025618e-01 -3.99537504e-01 1.20961949e-01 -7.80552745e-01 1.97470769e-01 5.45479596e-01 -9.55440030e-02 3.70833188e-01 1.97389543e-01 7.34767988e-02 1.14769840e+00 5.53283930e-01 8.48019600e-01 -6.26813948e-01 3.94372046e-01 1.61421239e-01 1.02061379e+00 1.04261503e-01 -6.04896545e-01 4.36371773e-01 5.08804977e-01 -1.00489748e+00 -1.58684433e+00 -7.42713451e-01 -3.54571640e-01 9.51889098e-01 2.48695865e-01 -5.63390136e-01 -1.00978625e+00 3.63327265e-02 -1.21361502e-01 -1.85851473e-02 -2.90703052e-03 -9.43227708e-02 -8.11821222e-01 -1.47184324e+00 3.85087550e-01 -5.46985939e-02 1.39549989e-02 -1.13172269e+00 -1.28681087e+00 8.27343285e-01 -3.56413990e-01 -7.04295039e-01 -2.87204862e-01 8.82862508e-01 -1.22897315e+00 -1.71621573e+00 -4.26361173e-01 -2.40551129e-01 8.41549397e-01 3.16997081e-01 8.25239360e-01 2.88635880e-01 -5.22951961e-01 -2.61070192e-01 -3.18570137e-01 2.56087989e-01 -6.67516530e-01 -5.32440662e-01 8.31544518e-01 -4.58912671e-01 7.74509192e-01 -1.01437366e+00 -6.74688101e-01 7.05682456e-01 -4.91133273e-01 -1.00990228e-01 3.25998724e-01 9.54643369e-01 6.56434238e-01 -8.65190029e-01 2.86206633e-01 -4.40541267e-01 4.20766413e-01 -5.85823655e-01 -2.20900059e-01 1.30984873e-01 -6.72118962e-01 1.32516116e-01 1.03418612e+00 -1.42781064e-01 -9.30003881e-01 3.91557783e-01 3.20613086e-01 4.55068678e-01 -2.43497074e-01 -1.06394459e-02 -3.80338788e-01 1.07003562e-01 8.57001960e-01 6.31430030e-01 4.94890898e-01 -4.30888951e-01 -6.92695826e-02 3.90090376e-01 1.67851988e-02 -7.42316186e-01 3.06874752e-01 3.82908076e-01 3.44858438e-01 -9.82886016e-01 -1.05219603e-01 -4.82645482e-01 -5.59926331e-01 1.44750968e-01 1.06678748e+00 -4.53247249e-01 -1.04687357e+00 3.07592779e-01 -1.35790110e+00 -1.64149627e-01 1.24294639e-01 5.82483709e-01 -6.84667408e-01 9.03223157e-01 -8.41948688e-01 -6.19673252e-01 -4.97448504e-01 -1.12320042e+00 4.73668933e-01 4.71882075e-01 -2.91704595e-01 -5.36405087e-01 7.43915677e-01 2.47714639e-01 4.06958014e-01 4.74025726e-01 1.20420170e+00 -6.04140282e-01 -4.89634305e-01 5.02253354e-01 1.28462985e-01 -9.46827047e-03 2.08073437e-01 1.48916751e-01 -2.54737794e-01 -2.58543074e-01 4.36213464e-02 -2.53803909e-01 9.00609612e-01 3.64209533e-01 -3.14232439e-01 -3.37368101e-01 -7.46666372e-01 5.99221587e-01 1.34652054e+00 6.27888978e-01 6.62205696e-01 2.45529085e-01 1.35571271e-01 4.99404192e-01 6.88442290e-01 1.84556156e-01 -2.53308088e-01 3.49694580e-01 3.14668179e-01 2.65214175e-01 1.78768575e-01 2.25576580e-01 5.30095875e-01 7.30347812e-01 -6.89181447e-01 -5.85697778e-02 -1.01519263e+00 7.16961548e-02 -1.67663026e+00 -9.25981343e-01 -5.25842786e-01 2.27969265e+00 1.16631305e+00 -2.58283317e-01 -1.22709163e-01 -4.39776778e-01 7.95360923e-01 -3.20862591e-01 -9.01639044e-01 -7.27335870e-01 -3.12118590e-01 4.40872759e-01 6.87307775e-01 4.76722151e-01 -6.58261776e-01 8.89436245e-01 6.91855621e+00 -2.77360380e-01 -1.23338974e+00 -3.76164094e-02 3.04567099e-01 5.61598003e-01 1.64694488e-02 9.41424608e-01 -6.24907076e-01 5.76080203e-01 1.08424807e+00 4.25333763e-03 3.77176017e-01 7.60132670e-01 6.32225752e-01 -4.14091855e-01 -6.45694554e-01 4.99174654e-01 -6.27741933e-01 -1.47508299e+00 -5.83761990e-01 4.36765015e-01 5.40285766e-01 1.41341209e-01 -7.30413854e-01 -1.64179534e-01 1.33146554e-01 -8.87103498e-01 -9.33617912e-03 5.94322264e-01 6.92926109e-01 -4.36161697e-01 5.60967624e-01 5.10090709e-01 -1.06227100e+00 -1.12388181e-02 -9.74349439e-01 -5.01249075e-01 4.34945017e-01 9.44576323e-01 -8.34361136e-01 2.76627652e-02 4.98406887e-01 8.64131674e-02 -2.26716071e-01 4.10222858e-01 -7.97353312e-02 5.31627357e-01 -2.18496725e-01 -1.84176877e-01 -1.86560348e-01 -9.27647948e-01 7.46251702e-01 8.97383511e-01 -1.78588152e-01 9.43830669e-01 1.35891477e-03 8.13336492e-01 2.90159509e-02 3.41510624e-01 -3.77341360e-01 -1.82270229e-01 2.69438148e-01 1.24214375e+00 -1.06466091e+00 -4.24416870e-01 1.22051969e-01 1.11545777e+00 2.14310095e-01 3.70503217e-01 -2.75767535e-01 -3.22667271e-01 1.21155179e+00 3.54211062e-01 -8.43296945e-02 -6.58282876e-01 1.04650982e-01 -1.20293462e+00 -1.88988939e-01 -6.30183160e-01 -1.57264203e-01 -2.32383415e-01 -9.22591507e-01 3.55907410e-01 -6.63437247e-01 -5.10802567e-01 3.04859271e-03 -5.53617716e-01 -6.06666148e-01 8.29831839e-01 -7.02131629e-01 -8.98388565e-01 5.79510024e-03 -4.46698219e-02 3.11708272e-01 2.53326982e-01 1.09644330e+00 -3.25401634e-01 -3.99009079e-01 -3.08168232e-01 8.13165069e-01 -5.27562320e-01 8.58129501e-01 -1.35999191e+00 3.94312233e-01 7.57162631e-01 -7.49491572e-01 1.44539618e+00 1.18433201e+00 -1.21115959e+00 -1.55632544e+00 -9.70335484e-01 1.38219059e+00 -5.73048174e-01 -3.59510295e-02 -4.67035323e-01 -9.23425019e-01 4.65445936e-01 -2.19445869e-01 -4.86137159e-02 1.08924055e+00 -2.24485025e-01 -5.18523827e-02 5.68638086e-01 -1.31298459e+00 3.33772153e-01 6.23214960e-01 -8.25679377e-02 -7.42354810e-01 6.96150482e-01 5.77396512e-01 -2.56948978e-01 -7.57110059e-01 1.46772444e-01 8.24193954e-01 -1.38794863e+00 5.31169057e-01 -1.17703974e+00 1.04441345e-01 -6.57427013e-01 -2.89905548e-01 -8.11790466e-01 -4.37201381e-01 -1.10650480e+00 -1.45237252e-01 4.05984104e-01 1.32486457e-02 -7.47029483e-01 5.81075668e-01 6.51284754e-01 -3.34876448e-01 -8.08600068e-01 -1.10590303e+00 -6.14300907e-01 4.96228449e-02 7.11469769e-01 3.90067935e-01 1.00008011e+00 7.47706831e-01 4.58325684e-01 -1.31371930e-01 -4.41045702e-01 5.39884686e-01 1.46527410e-01 6.10606313e-01 -1.07528245e+00 6.39325380e-02 3.19128305e-01 -1.95373073e-01 -8.25586677e-01 -3.06243300e-01 -5.70558071e-01 -6.67355433e-02 -1.24873483e+00 5.23532271e-01 -2.59704709e-01 1.21804386e-01 4.79262680e-01 1.20657831e-01 1.35849357e-01 -8.50727595e-03 7.57562995e-01 -2.95822263e-01 2.42285177e-01 8.23070467e-01 6.45013690e-01 -6.64137721e-01 -2.19538406e-01 -3.23748946e-01 6.62089407e-01 8.59646201e-01 -6.64240241e-01 2.22228944e-01 6.77298605e-01 4.31755900e-01 4.87250313e-02 -8.90026540e-02 -4.59154814e-01 -2.71297663e-01 -6.65625393e-01 3.55608463e-01 -3.95619988e-01 -1.17616117e-01 -5.79838336e-01 7.58676350e-01 1.30981565e+00 4.04361576e-01 1.23111971e-01 -3.61195177e-01 4.11409497e-01 4.08212930e-01 1.15452841e-01 1.37682974e+00 -4.87616926e-01 1.01179689e-01 -1.58302441e-01 -1.46816599e+00 -2.48193681e-01 1.29977226e+00 -6.83210611e-01 -5.13932765e-01 3.91435891e-01 -1.12962604e+00 -3.27405512e-01 1.47651970e+00 1.23332376e-02 1.88999012e-01 -9.68112588e-01 -1.69064388e-01 -4.14459556e-02 -1.85099244e-01 -4.00707364e-01 3.34806949e-01 1.10889876e+00 -1.35640180e+00 8.50616038e-01 -4.77409184e-01 -7.47025251e-01 -9.86247003e-01 7.52585649e-01 6.67520940e-01 -3.02639306e-01 -3.36402476e-01 2.26589039e-01 -4.99966219e-02 -5.45685589e-01 -7.20411539e-01 -3.57217342e-01 3.12883347e-01 -4.24995691e-01 5.23947835e-01 4.75725293e-01 2.03119032e-02 -8.25020730e-01 -8.30911040e-01 4.57668930e-01 -3.51726599e-02 9.56431091e-01 1.15535355e+00 -2.69019306e-01 -1.07825875e+00 -3.01586706e-02 6.81807220e-01 -7.16349855e-02 -8.73347819e-01 5.94842434e-02 1.67201415e-01 -1.93291426e-01 -1.06891012e+00 -6.55317366e-01 1.49972260e-01 4.21431869e-01 4.80392367e-01 -1.79417223e-01 7.09368825e-01 2.50607818e-01 1.04067516e+00 5.78354359e-01 6.22203410e-01 -9.04013634e-01 -7.81913474e-02 4.89867061e-01 3.58260840e-01 -7.62831748e-01 -3.45534682e-02 -1.40245914e-01 -2.63339311e-01 1.32046020e+00 4.78561372e-01 -1.37580290e-01 4.31183614e-02 1.36615438e-02 3.63428891e-02 -2.39328280e-01 -1.18739557e+00 4.04870242e-01 -6.25324070e-01 7.83692956e-01 9.85727608e-01 2.67992944e-01 -1.55886209e+00 6.93151712e-01 3.72060925e-01 -2.73248434e-01 7.23164916e-01 1.18053842e+00 -1.35542130e+00 -1.48305464e+00 -6.96764410e-01 -9.21829119e-02 -5.54006994e-01 1.47874549e-01 -6.73737109e-01 2.63376951e-01 3.27086687e-01 8.52674544e-01 -2.39673495e-01 1.51573315e-01 -3.91656049e-02 7.46363223e-01 3.22426140e-01 -4.44174290e-01 -5.31105161e-01 1.77195162e-01 -1.17151827e-01 -6.60419166e-01 -1.47885531e-01 -5.95908880e-01 -1.87921607e+00 -3.30829710e-01 -3.57701898e-01 5.85132539e-01 7.18196273e-01 9.46436644e-01 8.53618979e-01 -1.61918461e-01 6.47910476e-01 -7.30164230e-01 -8.40937734e-01 -9.65420425e-01 -7.37822950e-01 3.20833862e-01 7.76416734e-02 -5.81655502e-01 -3.02501172e-01 2.86288112e-01]

[4.724524021148682, 5.276699542999268]

7b0d6954-827e-4316-b29c-f921cad64496

learning-target-domain-specific-classifier

2008.10785

https://arxiv.org/abs/2008.10785v1

https://arxiv.org/pdf/2008.10785v1.pdf

Learning Target Domain Specific Classifier for Partial Domain Adaptation

Unsupervised domain adaptation~(UDA) aims at reducing the distribution discrepancy when transferring knowledge from a labeled source domain to an unlabeled target domain. Previous UDA methods assume that the source and target domains share an identical label space, which is unrealistic in practice since the label information of the target domain is agnostic. This paper focuses on a more realistic UDA scenario, i.e. partial domain adaptation (PDA), where the target label space is subsumed to the source label space. In the PDA scenario, the source outliers that are absent in the target domain may be wrongly matched to the target domain (technically named negative transfer), leading to performance degradation of UDA methods. This paper proposes a novel Target Domain Specific Classifier Learning-based Domain Adaptation (TSCDA) method. TSCDA presents a soft-weighed maximum mean discrepancy criterion to partially align feature distributions and alleviate negative transfer. Also, it learns a target-specific classifier for the target domain with pseudo-labels and multiple auxiliary classifiers, to further address classifier shift. A module named Peers Assisted Learning is used to minimize the prediction difference between multiple target-specific classifiers, which makes the classifiers more discriminant for the target domain. Extensive experiments conducted on three PDA benchmark datasets show that TSCDA outperforms other state-of-the-art methods with a large margin, e.g. $4\%$ and $5.6\%$ averagely on Office-31 and Office-Home, respectively.

['PengFei Ge', 'Chuan-Xian Ren', 'Peiyi Yang', 'Shuicheng Yan']

2020-08-25

['partial-domain-adaptation']

['methodology']

[ 2.43037075e-01 8.19034427e-02 -3.04318786e-01 -6.44885957e-01 -8.84418488e-01 -5.23433387e-01 2.46378466e-01 9.83625948e-02 -1.28118217e-01 9.48942661e-01 -2.63590753e-01 1.95123702e-02 -1.10238671e-01 -7.08580077e-01 -6.23994231e-01 -9.57270682e-01 3.71259272e-01 7.51114190e-01 3.55518311e-01 -1.33610994e-01 -2.82052308e-02 2.31781229e-01 -1.47924960e+00 2.80370057e-01 1.28391898e+00 1.01360607e+00 2.17557892e-01 -1.97354585e-01 -4.00610566e-01 3.31867188e-01 -7.01614857e-01 -2.81964630e-01 4.39714521e-01 -5.24503589e-01 -6.37898982e-01 2.16559038e-01 3.49587590e-01 -1.46997711e-02 1.92923889e-01 1.21301889e+00 5.12860775e-01 1.26705408e-01 9.94170547e-01 -1.63395679e+00 -5.53296924e-01 3.63891244e-01 -8.18730831e-01 3.81049812e-02 1.38054371e-01 -1.20088629e-01 7.20547259e-01 -9.04820383e-01 4.76235807e-01 1.18747437e+00 6.74116075e-01 7.52958059e-01 -1.28591549e+00 -1.16112828e+00 5.29123783e-01 3.90729569e-02 -1.47374094e+00 -3.02074343e-01 9.71058488e-01 -4.90717292e-01 5.36961615e-01 -1.55381322e-01 4.04543988e-03 1.01910007e+00 -6.53657392e-02 7.18613565e-01 1.22800589e+00 -5.13025582e-01 5.40362835e-01 6.48956537e-01 2.87990689e-01 1.97141588e-01 2.42456675e-01 -6.04285905e-03 -4.45523471e-01 -3.27650875e-01 2.64497697e-01 3.68165337e-02 -9.08497199e-02 -9.70637381e-01 -9.71359134e-01 7.48284221e-01 2.00656205e-01 4.23184149e-02 -3.40124011e-01 -8.01327407e-01 4.58563626e-01 5.05728006e-01 5.53531826e-01 1.17160097e-01 -7.95716166e-01 2.13839263e-01 -4.53752518e-01 2.67997593e-01 6.40849471e-01 1.42525673e+00 9.86546278e-01 -9.08303335e-02 -6.38919249e-02 1.23019958e+00 4.38338459e-01 7.45536089e-01 6.79357052e-01 -5.71663380e-01 6.84485197e-01 9.29023087e-01 2.73769170e-01 -6.14591360e-01 -1.70859963e-01 -4.09162790e-01 -6.34265900e-01 3.12384844e-01 6.64004743e-01 -1.79707706e-01 -8.15911591e-01 1.89846969e+00 6.78576767e-01 2.71868855e-01 4.35066849e-01 7.96605229e-01 5.59868157e-01 5.18162131e-01 1.41448319e-01 -4.67917174e-01 9.33017015e-01 -7.72688508e-01 -4.79816347e-01 -4.01428133e-01 7.79229105e-01 -7.92868078e-01 1.19734740e+00 3.97450745e-01 -6.18261456e-01 -7.84650445e-01 -1.14515674e+00 5.10189056e-01 -3.89812559e-01 -1.89070571e-02 3.15985270e-02 6.74465716e-01 -3.45887452e-01 4.15779501e-01 -3.31982970e-01 -5.91924071e-01 5.39923370e-01 4.21712428e-01 -3.16046298e-01 -3.13010633e-01 -1.12920618e+00 6.47614777e-01 6.51366055e-01 -6.28572941e-01 -6.45661294e-01 -7.46803582e-01 -6.56810045e-01 -8.43558982e-02 2.83282280e-01 -3.03840250e-01 1.29155803e+00 -1.21801651e+00 -1.46245301e+00 9.55181777e-01 -2.16060564e-01 -2.36291409e-01 4.28314656e-01 -7.53943250e-02 -8.54138136e-01 -4.17653173e-01 2.96015501e-01 5.29954970e-01 1.06417167e+00 -1.41800463e+00 -1.18649495e+00 -5.51830828e-01 -4.72354442e-01 4.45935756e-01 -5.14700234e-01 -3.31906438e-01 -5.35529070e-02 -6.83333635e-01 1.31773204e-01 -9.09083009e-01 1.93134412e-01 -3.95160824e-01 -2.03234360e-01 -4.34941441e-01 1.22560525e+00 -3.15541178e-01 1.27339864e+00 -2.39293981e+00 -1.43771127e-01 4.27394927e-01 5.45280315e-02 4.87762302e-01 -1.55084491e-01 1.06061548e-01 -5.24130881e-01 -3.91449362e-01 -4.92506087e-01 -3.81980352e-02 -2.76694335e-02 2.32714638e-01 -3.94194812e-01 3.39063197e-01 3.96952964e-02 1.46770090e-01 -9.83211040e-01 -4.76216197e-01 5.52874692e-02 -6.24240115e-02 -4.02079493e-01 1.85557649e-01 -1.33916229e-01 5.43414354e-01 -4.93042290e-01 7.30415821e-01 1.14503992e+00 -8.24381113e-02 2.23812535e-01 1.71289314e-02 2.79386699e-01 -3.56829539e-02 -1.45862889e+00 1.44276333e+00 -2.84934551e-01 1.64495662e-01 -1.33042410e-01 -1.15305185e+00 1.57830238e+00 1.09754726e-01 6.64074659e-01 -6.78790748e-01 6.89180717e-02 4.70103413e-01 1.04002051e-01 -1.18926004e-01 6.96713254e-02 -2.63297528e-01 -2.04306543e-01 1.68331176e-01 -7.68753607e-03 1.18399367e-01 -2.25363895e-01 -1.41437501e-01 8.71454000e-01 2.75220275e-01 4.56777215e-01 -4.89407688e-01 7.54176259e-01 2.48431265e-01 1.10994220e+00 3.72550398e-01 -6.82118714e-01 5.80695152e-01 1.44273132e-01 -9.06009004e-02 -9.75480855e-01 -1.31641006e+00 -1.73755333e-01 1.25585663e+00 4.69929427e-01 2.07574695e-01 -6.49355412e-01 -1.28465939e+00 3.26287329e-01 1.05365610e+00 -3.04697126e-01 -5.39077997e-01 -4.71551478e-01 -6.66201413e-01 1.61014780e-01 4.36271727e-01 6.96153581e-01 -7.30125964e-01 2.47351658e-02 2.22878277e-01 -1.82355568e-01 -7.35469639e-01 -4.80365843e-01 2.98373431e-01 -9.04410660e-01 -1.07504749e+00 -8.51451874e-01 -1.14574385e+00 8.05132031e-01 2.87393391e-01 8.97980988e-01 -7.25802660e-01 3.22670072e-01 1.73443705e-01 -4.54923540e-01 -7.26477027e-01 -5.56498885e-01 6.65849149e-02 5.35011232e-01 2.33185202e-01 1.12009525e+00 -5.56812286e-01 -3.86720479e-01 8.15353215e-01 -5.25569737e-01 -4.30082023e-01 5.67478955e-01 1.00296843e+00 7.63293147e-01 2.84698069e-01 1.13434720e+00 -1.22015631e+00 7.08542645e-01 -8.69453907e-01 -4.71023411e-01 3.58547002e-01 -1.03302550e+00 -1.56175673e-01 9.84425783e-01 -8.63383889e-01 -1.35392213e+00 2.55573899e-01 3.79068553e-01 -6.11875176e-01 -5.14941633e-01 1.54482871e-01 -7.88501859e-01 2.08089963e-01 1.07145858e+00 2.51106769e-01 6.59764409e-02 -6.02298319e-01 -1.39904786e-02 1.08201981e+00 4.72700328e-01 -6.61475658e-01 9.62907612e-01 1.95682198e-01 -3.16760927e-01 -3.96765321e-01 -6.84395492e-01 -7.94937730e-01 -8.06165159e-01 5.65207005e-02 2.82219708e-01 -1.04313898e+00 -7.57406354e-02 6.06503189e-01 -7.66368687e-01 -2.41139352e-01 -5.64363718e-01 6.43242836e-01 -5.10337353e-01 2.72754908e-01 1.14982128e-01 -3.99603426e-01 -1.44953579e-01 -1.04195619e+00 6.89342082e-01 4.24913973e-01 -4.01933134e-01 -9.10671771e-01 5.67487106e-02 1.60888240e-01 8.55182558e-02 1.86480403e-01 1.04882073e+00 -1.37047505e+00 7.04841763e-02 -1.20725773e-01 -2.62111574e-01 6.66726112e-01 5.59279680e-01 -4.30631608e-01 -1.09766829e+00 -5.83458483e-01 -3.60589176e-02 -1.98437572e-01 3.49084675e-01 2.45290160e-01 8.54791820e-01 9.20294225e-02 -5.95060706e-01 2.85938680e-01 1.10980189e+00 7.27494836e-01 3.58763933e-01 4.31778878e-01 5.27429521e-01 5.81488550e-01 1.43947566e+00 5.59853673e-01 2.81457961e-01 6.63396239e-01 2.34222010e-01 1.47662044e-01 -1.42026141e-01 -2.37115175e-01 4.44168180e-01 7.40599871e-01 5.46754539e-01 -2.02907592e-01 -1.03560185e+00 7.78870940e-01 -1.81825209e+00 -5.24918258e-01 -5.28853387e-02 2.56083822e+00 9.50542271e-01 2.13234350e-01 3.44608873e-01 1.50274858e-01 1.19727838e+00 -2.87221372e-01 -1.14843369e+00 -2.46837333e-01 -1.89289153e-02 -3.90282180e-03 5.11584520e-01 2.47298524e-01 -1.25639331e+00 7.06034601e-01 5.15606260e+00 1.03519273e+00 -9.97874916e-01 6.75425828e-02 3.59977007e-01 2.41799325e-01 1.06976945e-02 -1.30570099e-01 -9.49011266e-01 8.14393938e-01 6.91328943e-01 -5.05326569e-01 1.18798278e-01 1.39763677e+00 -1.04142979e-01 1.46540716e-01 -1.34478533e+00 1.11346889e+00 -9.30780396e-02 -6.33961320e-01 -2.74309609e-02 3.77363190e-02 8.76193285e-01 -1.78780243e-01 1.03895888e-01 7.11372554e-01 4.36138242e-01 -4.22542900e-01 3.33257049e-01 2.19207585e-01 1.01996946e+00 -1.06103516e+00 8.52678120e-01 5.12413502e-01 -1.03060758e+00 -2.17245579e-01 -4.91243094e-01 1.66294828e-01 -3.72474611e-01 5.03680825e-01 -1.01139212e+00 6.16187692e-01 8.80684078e-01 7.94999421e-01 -3.94220769e-01 9.37893271e-01 1.62776664e-01 5.85777760e-01 -2.06195116e-01 3.32579970e-01 -1.20358296e-01 -1.67281792e-01 6.36359513e-01 8.33814025e-01 4.75670576e-01 8.79864916e-02 3.56164068e-01 5.12752891e-01 -1.48471832e-01 2.79348135e-01 -6.44479454e-01 3.50566834e-01 9.78124619e-01 8.04005444e-01 -3.69254619e-01 -4.23982978e-01 -4.39605296e-01 1.03263450e+00 1.37606621e-01 3.51166487e-01 -8.24623883e-01 -6.25127435e-01 7.67898738e-01 1.87430829e-01 2.40838781e-01 4.50108677e-01 -3.32145244e-01 -1.07049906e+00 1.24758676e-01 -9.23369229e-01 7.63268411e-01 -3.54154497e-01 -1.98246896e+00 3.25160444e-01 1.73035622e-01 -1.96021760e+00 -2.33093128e-01 -3.68471295e-01 -4.67686236e-01 1.00237966e+00 -1.47506893e+00 -8.46640050e-01 -4.33353305e-01 1.02619183e+00 5.73560774e-01 -6.90628350e-01 8.68963480e-01 4.73958820e-01 -5.50899148e-01 1.11423266e+00 6.57081008e-01 -4.06263880e-02 1.49974370e+00 -1.14514709e+00 -1.10476315e-02 5.80172718e-01 -5.36041975e-01 4.20059472e-01 5.68445742e-01 -6.19120419e-01 -8.04520965e-01 -1.44025993e+00 7.07233489e-01 -2.87662029e-01 3.07418793e-01 -1.56698093e-01 -1.32099617e+00 6.94506943e-01 -8.36249217e-02 4.37335260e-02 8.55937481e-01 6.55094013e-02 -4.90525216e-01 -6.36707306e-01 -1.78662276e+00 2.62786359e-01 8.59288871e-01 -2.12187573e-01 -7.68098950e-01 2.63353348e-01 5.36177337e-01 -3.16140354e-01 -1.02392006e+00 5.37861586e-01 1.84816256e-01 -8.85167658e-01 8.21938455e-01 -3.70686114e-01 -4.77147214e-02 -5.27072608e-01 -2.15432540e-01 -1.58110797e+00 -4.65604693e-01 -2.31036283e-02 -1.05232827e-01 1.65512776e+00 1.69455945e-01 -1.06386924e+00 9.13815081e-01 6.40696228e-01 -1.63702279e-01 -3.85557234e-01 -9.49615300e-01 -1.27173388e+00 3.72471899e-01 -5.84597625e-02 8.76130998e-01 1.48953021e+00 -5.22217192e-02 3.97653222e-01 -3.68093676e-03 3.39649260e-01 6.65478468e-01 1.19855434e-01 8.08054507e-01 -1.63309622e+00 5.82773723e-02 -3.00880581e-01 -1.84085846e-01 -8.59521747e-01 2.12704912e-01 -8.47297192e-01 3.94922346e-02 -9.65470195e-01 -3.55177782e-02 -8.64213169e-01 -6.69972718e-01 5.52040577e-01 -9.97904837e-02 -1.51059523e-01 2.43569259e-02 5.03087878e-01 -4.11768198e-01 5.47553360e-01 9.30099607e-01 -2.19031036e-01 -5.95995367e-01 2.20649078e-01 -7.63546646e-01 8.13915431e-01 1.02002823e+00 -7.45805800e-01 -8.24592054e-01 -1.23081654e-01 -6.43281460e-01 -2.41959080e-01 -9.54935178e-02 -1.21698844e+00 2.88599767e-02 -3.69997382e-01 3.98684561e-01 -6.18983507e-01 -1.33461077e-02 -1.25671506e+00 -1.08504429e-01 3.03409129e-01 -2.19754294e-01 -3.77702087e-01 2.67247021e-01 7.00686634e-01 -3.84972334e-01 -1.35288775e-01 1.27017558e+00 1.54470488e-01 -1.15652549e+00 1.03359394e-01 1.66275427e-02 1.69840887e-01 1.58542550e+00 -5.10259390e-01 -2.81043947e-01 -5.81071451e-02 -7.26552069e-01 3.58947843e-01 6.19533896e-01 5.32493234e-01 5.71587384e-01 -1.61928248e+00 -7.69229352e-01 2.64704525e-01 6.54218733e-01 3.26479673e-01 2.58267254e-01 4.24499571e-01 2.71879677e-02 8.36049393e-02 -3.65495950e-01 -7.67988563e-01 -1.28142691e+00 7.36716330e-01 3.00599784e-01 -1.87794015e-01 -3.24977636e-01 8.12084734e-01 4.96946156e-01 -9.44554150e-01 3.57629240e-01 4.69004139e-02 -1.54673934e-01 1.38936296e-01 3.62170517e-01 5.67204475e-01 8.22161958e-02 -5.79462349e-01 -6.48988485e-01 5.02740622e-01 -3.62653077e-01 3.48532915e-01 9.47118998e-01 -3.79346460e-01 2.92622507e-01 3.94352913e-01 1.08571982e+00 -8.61304104e-02 -1.29290795e+00 -8.41122210e-01 1.20113909e-01 -4.81422156e-01 -4.23123032e-01 -1.07992887e+00 -8.06125939e-01 7.19150603e-01 1.07659662e+00 -1.89322710e-01 1.44938684e+00 -1.62557378e-01 6.95380986e-01 2.31304944e-01 4.20967549e-01 -1.50290823e+00 1.26496926e-01 4.64971155e-01 5.80264628e-01 -1.52828574e+00 -8.13939422e-02 -4.26066905e-01 -9.51512873e-01 8.40688646e-01 1.19474745e+00 -7.69144073e-02 5.96415758e-01 -1.57669008e-01 1.64198086e-01 3.28582704e-01 -4.33704346e-01 -4.01763292e-03 1.07796855e-01 1.14215434e+00 3.17652188e-02 1.58349156e-01 -6.13499209e-02 9.18491662e-01 1.55529201e-01 -2.71445531e-02 1.79110020e-01 8.88290524e-01 -4.37109828e-01 -1.34297144e+00 -7.44305372e-01 3.90093714e-01 -4.90312837e-02 2.89534539e-01 -2.92457730e-01 8.58734429e-01 6.17262661e-01 1.00511253e+00 -1.59327853e-02 -5.72809041e-01 6.55906856e-01 4.83714104e-01 2.49584410e-02 -6.94962680e-01 -2.01967821e-01 7.27539957e-02 -1.99560151e-01 -1.72809035e-01 -3.10453534e-01 -8.40933263e-01 -1.47506118e+00 -2.04452008e-01 -2.48585492e-01 2.19327167e-01 2.31876701e-01 6.22971177e-01 5.68408251e-01 3.79817784e-01 9.17007029e-01 -2.20879287e-01 -7.03222394e-01 -8.44017267e-01 -8.43688548e-01 6.73093677e-01 1.33039877e-01 -9.00761843e-01 -2.27130845e-01 1.56139120e-01]

[10.354382514953613, 3.1214656829833984]

05a471bf-1775-47b4-820f-319d019ab58c

a-optimization-framework-for-herbal

2304.12828

https://arxiv.org/abs/2304.12828v1

https://arxiv.org/pdf/2304.12828v1.pdf

A optimization framework for herbal prescription planning based on deep reinforcement learning

Treatment planning for chronic diseases is a critical task in medical artificial intelligence, particularly in traditional Chinese medicine (TCM). However, generating optimized sequential treatment strategies for patients with chronic diseases in different clinical encounters remains a challenging issue that requires further exploration. In this study, we proposed a TCM herbal prescription planning framework based on deep reinforcement learning for chronic disease treatment (PrescDRL). PrescDRL is a sequential herbal prescription optimization model that focuses on long-term effectiveness rather than achieving maximum reward at every step, thereby ensuring better patient outcomes. We constructed a high-quality benchmark dataset for sequential diagnosis and treatment of diabetes and evaluated PrescDRL against this benchmark. Our results showed that PrescDRL achieved a higher curative effect, with the single-step reward improving by 117% and 153% compared to doctors. Furthermore, PrescDRL outperformed the benchmark in prescription prediction, with precision improving by 40.5% and recall improving by 63%. Overall, our study demonstrates the potential of using artificial intelligence to improve clinical intelligent diagnosis and treatment in TCM.

['Xuezhong Zhou', 'Tiancai Wen', 'Zhuang Liu', 'Feidie Yu', 'Qiguang Zheng', 'Ning Wang', 'Xiong He', 'Xin Su', 'Zecong Yu', 'Kuo Yang']

2023-04-25

['sequential-diagnosis']

['medical']

[ 1.78150818e-01 1.26638874e-01 -9.53725219e-01 -3.13053995e-01 -4.61681664e-01 -1.28064722e-01 -7.40127638e-02 3.07670951e-01 -9.82502624e-02 9.43331361e-01 3.32686037e-01 -2.55462945e-01 -5.92045307e-01 -1.02558637e+00 -3.79606932e-01 -7.53581703e-01 -5.39591797e-02 1.11015940e+00 -7.22455084e-01 5.79337478e-02 1.02968663e-01 1.84928000e-01 -8.70152652e-01 6.05956972e-01 1.43388236e+00 8.31555128e-01 3.13399047e-01 3.03306997e-01 3.53062361e-01 8.42249453e-01 -3.26198071e-01 -1.55902058e-01 2.60646325e-02 -4.77320701e-01 -9.14339900e-01 7.96093643e-02 -2.73637444e-01 -4.87852782e-01 -2.01285422e-01 7.40749002e-01 8.97270918e-01 -2.36898258e-01 6.17620587e-01 -7.67608523e-01 -1.03936374e+00 1.13162684e+00 -3.28190416e-01 -3.61132741e-01 3.46967459e-01 5.51674545e-01 8.84609699e-01 -1.61798045e-01 5.18169999e-01 9.36827600e-01 5.17407298e-01 8.51534724e-01 -1.30508232e+00 -5.12867451e-01 -1.67930573e-01 1.63421869e-01 -1.03763473e+00 1.24954715e-01 1.62396237e-01 -4.38910544e-01 8.17095041e-01 2.29494646e-01 1.01339424e+00 7.14674115e-01 4.10000533e-01 1.20978689e+00 1.06344891e+00 -8.83593932e-02 3.67444813e-01 -9.39134359e-02 7.07030371e-02 2.80175418e-01 1.37767419e-01 5.63420653e-01 7.47280195e-02 -7.55821466e-02 7.21068025e-01 3.22629452e-01 -2.28150681e-01 3.83867264e-01 -1.18757033e+00 1.09747648e+00 7.66934156e-01 3.08547497e-01 -1.02034473e+00 4.25722003e-02 1.93537295e-01 -2.13034712e-02 1.81778029e-01 1.07593143e+00 -9.34647977e-01 1.47556186e-01 -6.69487715e-01 5.89641333e-01 7.01159716e-01 6.79906189e-01 4.50238958e-03 3.74644808e-02 -1.03443170e+00 6.81681931e-01 9.07626450e-02 7.33794391e-01 4.24378484e-01 -1.01250756e+00 3.13480981e-02 7.81310558e-01 1.94762409e-01 -6.55691326e-01 -7.93788254e-01 -7.59292901e-01 -1.28298819e+00 -3.99745643e-01 3.88046391e-02 -4.56592768e-01 -9.25091445e-01 1.49076068e+00 2.61571348e-01 2.73617152e-02 1.98715091e-01 8.24921966e-01 9.31768715e-01 5.03092766e-01 6.45821452e-01 -5.52321136e-01 1.37648845e+00 -8.23215723e-01 -1.11273408e+00 1.07937250e-02 9.98425305e-01 -4.06282395e-01 7.01481700e-01 5.93914747e-01 -1.17666852e+00 -1.21629857e-01 -5.46175539e-01 3.33244622e-01 6.19221032e-02 4.84989524e-01 1.02105105e+00 3.64697307e-01 -5.26661396e-01 1.17747509e+00 -5.47456145e-01 2.02746391e-02 1.03783977e+00 7.58479595e-01 1.26160249e-01 -2.94278562e-01 -1.47127759e+00 9.63677585e-01 6.54203713e-01 -3.31559777e-02 -8.32503200e-01 -1.38541138e+00 -6.83276355e-01 3.16915125e-01 4.29721981e-01 -1.14921868e+00 1.59520435e+00 -5.61101139e-01 -1.77483904e+00 5.55275559e-01 2.18285799e-01 -7.47783720e-01 2.71714330e-01 -3.76283467e-01 -3.24887484e-01 1.19600900e-01 -6.57298416e-03 8.89147222e-01 1.09472282e-01 -8.18352342e-01 -5.44273973e-01 -4.48029101e-01 -5.19334823e-02 3.65511328e-01 -1.42322034e-02 -3.85254920e-01 -3.90479304e-02 -5.74236512e-01 -3.39199930e-01 -1.03803802e+00 -7.67976582e-01 -2.57131398e-01 -4.08434898e-01 -4.50341463e-01 2.72670597e-01 -7.94640422e-01 1.49315238e+00 -1.52907491e+00 4.17418003e-01 -8.18676129e-02 2.67024577e-01 4.83701855e-01 -7.67795071e-02 2.38817230e-01 -5.92378415e-02 4.21933472e-01 -1.79484993e-01 1.94453090e-01 -1.93611726e-01 3.28876704e-01 1.73879340e-01 2.14479625e-01 5.07327378e-01 1.51058340e+00 -1.24288297e+00 -3.68781507e-01 3.29599142e-01 4.10124123e-01 -9.51575875e-01 1.52680919e-01 -9.17261541e-01 7.52880156e-01 -7.74322510e-01 9.45680380e-01 4.75773364e-01 -8.71649861e-01 4.99212027e-01 -1.42451733e-01 3.06723267e-01 1.91933721e-01 -7.15574861e-01 1.44800484e+00 -2.68239826e-01 -6.21669441e-02 -6.58559740e-01 -9.27012920e-01 4.50196534e-01 6.50324702e-01 1.24508810e+00 -1.13239729e+00 3.81920874e-01 1.00673109e-01 3.48347515e-01 -8.48022699e-01 -1.69497970e-02 -2.77039409e-01 1.30174190e-01 3.13389719e-01 -3.78749430e-01 -1.91264391e-01 1.36363775e-01 -1.51215136e-01 8.40992928e-01 1.04268193e-01 4.95375842e-01 -1.74726203e-01 2.28188530e-01 4.78204370e-01 7.12885439e-01 5.26856065e-01 -9.68498364e-02 2.27431178e-01 3.68385166e-01 -4.63668674e-01 -6.59904122e-01 -5.62929571e-01 -3.94249827e-01 2.27460459e-01 -1.77829489e-01 -7.06381164e-03 -5.57182074e-01 -6.24530494e-01 2.72191972e-01 8.52953255e-01 -6.76726222e-01 -3.85850519e-01 -3.26808393e-01 -1.24412036e+00 3.02302450e-01 5.12903571e-01 5.79160869e-01 -1.56607604e+00 -5.39654016e-01 5.49256980e-01 -2.30188891e-01 -8.64904881e-01 -5.03576756e-01 1.19918749e-01 -1.26296830e+00 -1.13686442e+00 -1.03904366e+00 -7.50426888e-01 3.50043088e-01 -1.09755479e-01 1.27684271e+00 2.01098636e-01 -2.75400281e-01 -1.58291116e-01 -2.65717208e-01 -5.51871300e-01 -4.12886471e-01 1.10743409e-02 -1.09158948e-01 -5.18143535e-01 3.47014099e-01 -9.70916301e-02 -1.25984895e+00 9.73977670e-02 -5.53178310e-01 2.49895781e-01 7.25034595e-01 1.24140847e+00 9.10558999e-01 6.26264140e-02 9.87781644e-01 -1.14452755e+00 6.51604533e-01 -5.93185902e-01 -3.54678124e-01 2.68570513e-01 -1.14938569e+00 -1.11627437e-01 3.51369321e-01 -6.02082431e-01 -7.82812715e-01 2.41296843e-01 -2.27401376e-01 -2.54061818e-01 -5.63837364e-02 8.73721123e-01 2.02758148e-01 2.88759083e-01 6.13648891e-01 1.67056397e-01 -4.66920668e-03 -4.36980546e-01 1.70202583e-01 6.18716896e-01 -4.84716371e-02 -3.13131481e-01 -5.73717952e-02 -7.72577673e-02 1.41882628e-01 -3.07023019e-01 -1.08623016e+00 -9.50117037e-02 -1.07249640e-01 2.00818837e-01 1.10339963e+00 -9.44873869e-01 -1.58935356e+00 2.87861854e-01 -7.40391195e-01 -7.18329728e-01 -4.31287587e-01 7.85278916e-01 -4.52104747e-01 1.24426074e-01 -8.09534669e-01 -6.24339223e-01 -1.14611387e+00 -1.24736798e+00 9.75308776e-01 3.87646288e-01 -3.51949304e-01 -9.69995618e-01 1.08444490e-01 4.20191675e-01 3.07853281e-01 2.91744411e-01 1.25490439e+00 -4.73118246e-01 -3.69013071e-01 -9.55674499e-02 -1.56831548e-01 5.23150563e-02 4.00930166e-01 -2.27136329e-01 -6.40777707e-01 6.31798953e-02 -2.00077519e-01 -3.65509599e-01 4.07964319e-01 1.31133318e+00 1.57020545e+00 -3.84781122e-01 -3.80506337e-01 4.72576320e-01 1.48536551e+00 8.43846619e-01 8.94672155e-01 1.60227776e-01 5.05108893e-01 3.65040034e-01 9.60726857e-01 8.33186805e-01 4.89546329e-01 4.64539558e-01 5.70420682e-01 -3.01819175e-01 7.66107291e-02 2.17061356e-01 -1.86452150e-01 2.54538447e-01 -1.29870176e-01 -3.18759918e-01 -8.40537786e-01 4.08715576e-01 -1.83178687e+00 -9.15628254e-01 -2.22022742e-01 1.88757122e+00 1.35749125e+00 -3.01998854e-01 1.37010366e-01 6.66812342e-03 2.83780575e-01 -7.97691226e-01 -7.64402688e-01 -4.56851810e-01 4.12473865e-02 6.54076219e-01 3.50877464e-01 5.16135842e-02 -1.04535580e+00 8.72376442e-01 6.33159351e+00 8.25572848e-01 -1.09768116e+00 -2.64696211e-01 1.10744870e+00 1.01650417e-01 -6.87044635e-02 -3.14819187e-01 -6.37563765e-01 6.38547301e-01 7.80062556e-01 -5.29568233e-02 5.29003739e-01 6.51845038e-01 6.54513001e-01 2.97605116e-02 -1.10467327e+00 6.13582909e-01 -7.11989641e-01 -1.65252221e+00 6.05832748e-02 2.86727607e-01 1.18252349e+00 -6.38868436e-02 1.48464561e-01 5.02478302e-01 5.99510789e-01 -1.46216881e+00 -9.57650840e-02 5.97678959e-01 6.22231364e-01 -7.55537331e-01 8.95812690e-01 4.58111852e-01 -6.58069015e-01 -1.75753906e-01 -7.41676316e-02 6.77593872e-02 9.92297083e-02 6.65292144e-01 -1.07629728e+00 7.99778342e-01 4.25337821e-01 1.29811144e+00 -7.79204071e-02 1.27446175e+00 -1.50407568e-01 5.63352287e-01 -4.43062633e-02 -1.65390089e-01 1.27597496e-01 -3.46169710e-01 -2.51684695e-01 8.18585873e-01 3.07063758e-01 6.09866142e-01 4.43129063e-01 7.28893220e-01 -1.78509027e-01 1.65788919e-01 -3.01382035e-01 -6.88719571e-01 1.07522734e-01 1.04988623e+00 -2.54729956e-01 -4.81778890e-01 3.55743319e-02 6.02067232e-01 -2.02430665e-01 9.64134708e-02 -1.14792728e+00 2.51595289e-01 3.92216146e-01 -2.04001173e-01 5.93978986e-02 6.93241894e-01 -9.07103658e-01 -9.39968228e-01 -7.84897447e-01 -1.37341154e+00 6.58045948e-01 -6.75314844e-01 -1.24789643e+00 4.24955070e-01 -3.39468747e-01 -1.42161560e+00 4.81951013e-02 -3.99567515e-01 -2.32970506e-01 8.65131736e-01 -1.63480186e+00 -1.10854697e+00 -9.86368954e-02 3.41967732e-01 5.63194513e-01 1.60380527e-01 1.22457075e+00 5.64825952e-01 -7.94125438e-01 4.17366028e-01 1.36507601e-02 -2.23583594e-01 4.85521436e-01 -1.20397472e+00 -4.20258611e-01 -2.79063821e-01 -5.54499328e-01 4.18049932e-01 4.42807168e-01 -1.05697823e+00 -1.31002951e+00 -1.33743632e+00 9.10653591e-01 -1.27621710e-01 1.55683205e-01 7.21372783e-01 -7.15398312e-01 4.15243834e-01 3.64788681e-01 -5.97968102e-01 8.84535372e-01 -6.66755950e-03 4.97338146e-01 1.15326703e-01 -1.47548378e+00 8.23816001e-01 6.20739818e-01 1.87641948e-01 -6.87752366e-02 9.46474731e-01 9.68360603e-01 -6.67896509e-01 -1.74145555e+00 8.96142066e-01 6.15830719e-01 -4.07028012e-02 1.13975549e+00 -9.63855505e-01 1.17708528e+00 8.17918628e-02 2.04692051e-01 -1.32797074e+00 -6.80515885e-01 -2.28413895e-01 -4.67442453e-01 3.80104989e-01 4.75835860e-01 -2.41179273e-01 8.34847510e-01 6.06040120e-01 -1.28713753e-02 -1.51115990e+00 -1.50401697e-01 -2.83688784e-01 2.98039705e-01 3.12752314e-02 7.52564907e-01 1.20945334e+00 5.03633544e-02 3.27102244e-01 -7.71387994e-01 5.32058673e-03 4.50560600e-01 2.99267292e-01 1.57609880e-01 -1.04374456e+00 -5.23516476e-01 -4.99639213e-01 3.06422859e-01 -6.94755614e-01 -3.46638530e-01 -9.61952925e-01 -1.38694823e-01 -1.97623312e+00 5.20852208e-01 -4.87871051e-01 -4.23104763e-01 8.75671685e-01 -3.99637192e-01 -4.60655242e-02 -6.03273511e-02 -8.85227229e-03 -1.55863732e-01 6.96688354e-01 2.17597556e+00 -5.56906700e-01 -6.24137521e-01 2.69943774e-01 -9.12912130e-01 4.36334848e-01 1.20939493e+00 -6.60598993e-01 -3.35868835e-01 -2.65812755e-01 1.27991557e-01 5.95020354e-01 1.13190208e-02 -6.03900909e-01 -3.34704846e-01 -7.51531661e-01 6.35017633e-01 -8.27962041e-01 1.13667455e-02 -6.42867088e-01 3.80532384e-01 1.55357277e+00 -5.19740999e-01 -2.98773438e-01 3.00790846e-01 3.91203910e-01 2.53630608e-01 1.33841401e-02 7.19857395e-01 -2.05935940e-01 -3.96671236e-01 5.04518032e-01 -3.16212595e-01 -2.35003784e-01 1.17346859e+00 2.07053676e-01 -2.90545523e-01 -1.13017567e-01 -1.09243262e+00 8.03760588e-01 -2.07163379e-01 5.16095608e-02 7.42747784e-01 -1.30781329e+00 -8.82733464e-01 -3.33592832e-01 -4.56283502e-02 2.98848543e-02 6.19408488e-01 1.24533594e+00 -7.32109010e-01 7.05669343e-01 -7.11418018e-02 -5.64139366e-01 -1.10871637e+00 8.65475535e-01 5.08153498e-01 -9.32409883e-01 -5.70160925e-01 4.24662471e-01 1.89554729e-02 -4.81568098e-01 2.99407482e-01 -5.98665535e-01 -7.49684155e-01 -1.48706555e-01 2.33434051e-01 2.55623430e-01 -1.35970920e-01 4.63039577e-02 -1.67003483e-01 4.68403429e-01 -1.60854131e-01 5.44515908e-01 1.70294213e+00 3.34517717e-01 3.26299816e-02 -2.90368766e-01 6.61417246e-01 -7.67945707e-01 -7.40390718e-01 -1.90230906e-01 -2.74766982e-02 -1.64161175e-01 3.06750536e-01 -1.62151814e+00 -1.37785399e+00 5.86417735e-01 9.10121500e-01 -2.16405466e-01 1.43131518e+00 -4.42584753e-01 1.01138330e+00 4.81674701e-01 1.97724789e-01 -8.77587020e-01 1.40917525e-01 2.13812813e-01 9.33566391e-01 -1.58493996e+00 3.34340304e-01 -2.31293768e-01 -1.22633135e+00 8.68547142e-01 5.12846887e-01 -1.15409940e-01 6.90874875e-01 -1.36956111e-01 -1.76137239e-01 -4.44428921e-01 -7.76293457e-01 -2.01345414e-01 4.65845168e-01 2.60336310e-01 6.58545196e-01 5.26845396e-01 -7.34033465e-01 7.11019039e-01 4.29952651e-01 7.71369696e-01 5.38772009e-02 9.26294565e-01 -6.75405487e-02 -1.39520633e+00 5.75176999e-02 8.39113772e-01 -7.49665558e-01 -3.81530255e-01 -1.61880136e-01 5.88067651e-01 2.41675794e-01 9.03766394e-01 -4.32135195e-01 -3.63241583e-01 2.67684400e-01 -3.06332022e-01 6.32197678e-01 -8.69237244e-01 -1.17482126e+00 5.46747804e-01 4.80951853e-02 -3.80581647e-01 -5.59653878e-01 -3.53648931e-01 -1.66955793e+00 -2.76702672e-01 -4.55146462e-01 -1.65478624e-02 4.25404698e-01 9.67918515e-01 6.14954472e-01 1.22088361e+00 7.48404682e-01 -2.83982217e-01 -7.91679382e-01 -8.45426142e-01 -5.09316444e-01 2.81327128e-01 9.41605121e-02 -3.20015788e-01 4.41470385e-01 -1.21981725e-01]

[7.972714900970459, 5.624350070953369]

8cddc061-1b01-47b3-b54c-2f1470a1a422

nemf-neural-motion-fields-for-kinematic

2206.03287

https://arxiv.org/abs/2206.03287v3

https://arxiv.org/pdf/2206.03287v3.pdf

NeMF: Neural Motion Fields for Kinematic Animation

We present an implicit neural representation to learn the spatio-temporal space of kinematic motions. Unlike previous work that represents motion as discrete sequential samples, we propose to express the vast motion space as a continuous function over time, hence the name Neural Motion Fields (NeMF). Specifically, we use a neural network to learn this function for miscellaneous sets of motions, which is designed to be a generative model conditioned on a temporal coordinate $t$ and a random vector $z$ for controlling the style. The model is then trained as a Variational Autoencoder (VAE) with motion encoders to sample the latent space. We train our model with a diverse human motion dataset and quadruped dataset to prove its versatility, and finally deploy it as a generic motion prior to solve task-agnostic problems and show its superiority in different motion generation and editing applications, such as motion interpolation, in-betweening, and re-navigating. More details can be found on our project page: https://cs.yale.edu/homes/che/projects/nemf/.

['Yi Zhou', 'Holly Rushmeier', 'James Zachary', 'Jun Saito', 'Chengan He']

2022-06-04

['miscellaneous']

['miscellaneous']

[-1.71027452e-01 -1.90976679e-01 -2.58852899e-01 -2.14269698e-01 -4.95928854e-01 -4.95085239e-01 7.02255726e-01 -8.77721965e-01 -2.44306728e-01 6.39736176e-01 5.85013330e-01 -1.21759564e-01 1.36610135e-01 -9.50222015e-01 -1.00094569e+00 -7.00983167e-01 8.38338286e-02 2.63177127e-01 -3.91330793e-02 -1.32170409e-01 2.31076181e-02 3.20899308e-01 -1.26536393e+00 5.19734286e-02 6.88844323e-01 4.30518478e-01 3.71850133e-01 9.10189986e-01 2.08758682e-01 9.64529693e-01 -4.73842293e-01 -2.11427674e-01 2.80254066e-01 -6.54203951e-01 -8.84127080e-01 5.78579679e-03 1.38933197e-01 -5.47663391e-01 -9.57412779e-01 8.17560554e-01 4.07763392e-01 6.70257151e-01 8.66765440e-01 -1.38291728e+00 -1.10812783e+00 3.63401264e-01 -2.42860943e-01 -9.49697271e-02 3.36263865e-01 5.12567163e-01 9.52934206e-01 -8.31720412e-01 1.06733739e+00 1.23933387e+00 5.58096051e-01 8.93423319e-01 -1.32588124e+00 -3.77404571e-01 -5.30796684e-02 2.15920374e-01 -1.22357631e+00 -3.70274007e-01 9.04107213e-01 -7.04004169e-01 7.76466131e-01 1.53137341e-01 8.89108837e-01 1.65774500e+00 2.83129901e-01 1.12436450e+00 2.16854185e-01 -5.58670498e-02 3.40153188e-01 -5.86826086e-01 -4.83147085e-01 7.38187909e-01 -2.49805495e-01 1.56815886e-01 -3.73436302e-01 7.79104158e-02 1.46901143e+00 1.64720058e-01 -4.23406661e-01 -5.15105546e-01 -1.57435811e+00 9.87811029e-01 4.80639428e-01 1.11966744e-01 -5.43548644e-01 7.02473044e-01 1.53578952e-01 1.02781370e-01 1.65959403e-01 4.17956084e-01 -2.41479397e-01 -3.23293775e-01 -9.20397162e-01 9.45151567e-01 5.95399320e-01 8.77298117e-01 6.95342541e-01 3.93786520e-01 -3.65149438e-01 8.03266704e-01 2.47210726e-01 3.94728720e-01 7.19271660e-01 -1.80261493e+00 3.07674855e-01 2.49601044e-02 3.54137689e-01 -9.20538843e-01 -6.07851744e-02 -9.33959037e-02 -1.02452195e+00 8.52113664e-02 3.62891644e-01 -3.70664328e-01 -9.09034908e-01 2.04595995e+00 2.70584404e-01 4.73714888e-01 -1.78901121e-01 9.33605790e-01 5.65669239e-01 1.04004109e+00 1.10288098e-01 2.82093823e-01 9.21479642e-01 -1.18953371e+00 -7.49955773e-01 -1.20492175e-01 6.20459080e-01 -4.06152844e-01 1.16394508e+00 4.02276926e-02 -1.31261492e+00 -6.85551405e-01 -8.37745905e-01 -4.25214171e-01 -1.46128073e-01 2.43172854e-01 6.98585629e-01 -1.47561044e-01 -1.13034713e+00 9.86987412e-01 -1.32304430e+00 -2.14833528e-01 3.41770679e-01 8.77477974e-02 -3.66629481e-01 1.37587085e-01 -1.20870817e+00 4.88626033e-01 3.23462814e-01 1.90978706e-01 -1.12625468e+00 -6.53181016e-01 -1.09544492e+00 -1.71330437e-01 -2.32238625e-03 -1.42070246e+00 1.30941200e+00 -7.60345817e-01 -1.89918208e+00 5.26806176e-01 -2.75482655e-01 -3.42182875e-01 7.37110674e-01 -3.83875281e-01 -2.22243831e-01 3.59039828e-02 3.41290653e-01 1.14277112e+00 9.76835668e-01 -9.15441811e-01 -3.94265890e-01 9.85904559e-02 1.14739425e-01 2.25084007e-01 1.08703166e-01 -4.19529229e-01 -6.15509629e-01 -1.20621669e+00 -4.41090949e-02 -1.10947037e+00 -2.56331891e-01 1.07592598e-01 -5.03117979e-01 -1.00506209e-01 8.27069759e-01 -9.21179116e-01 1.28271735e+00 -2.00943685e+00 8.43102276e-01 -2.21970543e-01 4.56036329e-02 -6.81776628e-02 -3.01456481e-01 4.39096898e-01 -1.03331491e-01 3.07409130e-02 -5.50604105e-01 -4.80798572e-01 1.74179971e-01 4.06053990e-01 -3.29100251e-01 3.56928259e-01 3.57096374e-01 1.29376400e+00 -1.08088291e+00 -1.75516412e-01 3.04867834e-01 7.47133255e-01 -8.78136873e-01 4.38574046e-01 -4.82034802e-01 1.01540387e+00 -4.69413370e-01 3.99440348e-01 2.29967937e-01 -3.63499254e-01 -1.08890243e-01 -3.04163173e-02 4.46588658e-02 8.50888118e-02 -1.01415467e+00 2.31649828e+00 -2.96766400e-01 6.36127293e-01 -1.94743678e-01 -8.52999270e-01 6.33859336e-01 3.10061544e-01 6.50175571e-01 -3.57323855e-01 5.59958592e-02 -8.11611935e-02 -3.19270253e-01 -7.43649602e-01 7.09990382e-01 -9.86408535e-03 4.02894663e-03 3.30147952e-01 1.80872187e-01 -6.67627975e-02 3.93600278e-02 1.46408767e-01 1.05823708e+00 8.47411036e-01 -9.17316973e-02 -1.34276394e-02 3.37608635e-01 2.15413906e-02 6.13701642e-01 4.85429466e-01 -2.13021830e-01 8.68555665e-01 3.53345335e-01 -5.32586038e-01 -1.41183674e+00 -1.23175454e+00 2.13460952e-01 9.47106421e-01 1.51027748e-02 -2.97116995e-01 -8.30842376e-01 -3.57759744e-01 1.04505494e-02 7.38905132e-01 -7.48970747e-01 -1.61312863e-01 -9.80810463e-01 -4.66328382e-01 3.90325010e-01 8.77393842e-01 5.09821653e-01 -1.38142502e+00 -6.97476089e-01 2.79597968e-01 -5.31623185e-01 -8.00446808e-01 -8.68064046e-01 -2.76757658e-01 -7.60268986e-01 -7.24157035e-01 -1.24159956e+00 -7.74233401e-01 4.24729913e-01 -8.40592459e-02 1.04592001e+00 -1.75749391e-01 -2.11117849e-01 3.86852682e-01 -2.01873407e-01 2.05264241e-01 -3.25501382e-01 5.03170155e-02 -1.07192419e-01 9.05568711e-03 -7.37900361e-02 -8.33341300e-01 -1.05533481e+00 1.01319283e-01 -9.38117087e-01 3.34081322e-01 1.40024051e-01 1.02178597e+00 5.53438067e-01 -3.26580316e-01 2.68260747e-01 -6.20797634e-01 4.39224958e-01 -7.65879691e-01 -3.53540748e-01 -1.58696920e-01 1.04855917e-01 2.85357833e-01 3.79314363e-01 -6.42228603e-01 -9.89996195e-01 1.15156800e-01 -3.76732081e-01 -8.77116621e-01 -2.61571944e-01 2.67525345e-01 -9.44646820e-02 4.81058657e-01 5.36148548e-01 3.49711061e-01 -1.01333223e-01 -3.92575055e-01 7.83335745e-01 9.63694006e-02 9.33488190e-01 -6.89434707e-01 7.92843103e-01 5.71915150e-01 -4.07131016e-02 -6.60925508e-01 -3.57005358e-01 -8.90991986e-02 -7.14037478e-01 -1.69304386e-01 1.36646175e+00 -9.76115108e-01 -5.75424790e-01 5.27895987e-01 -1.24969375e+00 -1.04163980e+00 -3.82707775e-01 5.66273153e-01 -1.06170082e+00 1.92752019e-01 -9.32352543e-01 -4.71659303e-01 3.68691273e-02 -1.21907127e+00 1.25323665e+00 7.93355703e-02 -6.98187888e-01 -1.23920166e+00 4.19319183e-01 1.26837837e-02 2.89750367e-01 7.32195020e-01 6.85206413e-01 2.92478293e-01 -7.55693913e-01 -4.08381484e-02 3.47401559e-01 1.13550670e-01 1.17629860e-02 3.10465783e-01 -6.92760110e-01 -2.54321605e-01 -2.99729407e-01 -2.01622471e-01 8.89994264e-01 7.99366415e-01 1.42661417e+00 -3.91693383e-01 -2.65142143e-01 1.19716787e+00 1.17590082e+00 2.83187717e-01 8.12764049e-01 3.25252891e-01 1.10006785e+00 5.72894514e-01 1.27789125e-01 4.88032311e-01 5.29524267e-01 7.51767278e-01 3.04294676e-01 2.85082549e-01 -6.75629228e-02 -6.99461937e-01 5.06102324e-01 9.16860104e-01 -4.00091469e-01 -2.35323846e-01 -7.97040939e-01 7.13246703e-01 -2.04813409e+00 -1.31320071e+00 2.22627327e-01 1.74527562e+00 6.52516603e-01 -2.92293042e-01 3.13787967e-01 -1.98230058e-01 5.93764305e-01 5.46600163e-01 -6.43078506e-01 -5.01802051e-03 6.58278540e-02 8.15015063e-02 1.73463345e-01 7.10790277e-01 -1.23995936e+00 1.10098827e+00 6.11790943e+00 6.66034400e-01 -1.24101865e+00 -1.26377568e-02 4.78829056e-01 -1.32724732e-01 -6.41037941e-01 -7.59075657e-02 -2.97073036e-01 6.34880424e-01 7.55494893e-01 -1.11009590e-01 6.40359044e-01 7.48422682e-01 4.43340719e-01 4.05200422e-01 -1.04473877e+00 9.66726243e-01 -4.05546993e-01 -1.60205615e+00 2.83138245e-01 -4.55454290e-02 8.23894441e-01 -5.81064858e-02 2.63329387e-01 3.92690897e-01 6.13061786e-01 -1.21010959e+00 1.06603086e+00 8.56522202e-01 8.40709627e-01 -5.10284245e-01 1.11833446e-01 3.99573803e-01 -1.17164850e+00 1.90500617e-02 -3.26752901e-01 -7.05294535e-02 4.30426776e-01 2.35226713e-02 -2.48598352e-01 4.60136652e-01 5.96356630e-01 1.15177500e+00 -4.89778817e-02 7.12569892e-01 -2.70264655e-01 4.74754095e-01 4.65056673e-02 2.34220147e-01 2.10905954e-01 -4.42633808e-01 5.74521303e-01 9.79296744e-01 5.83899796e-01 1.31248280e-01 -3.19647975e-02 1.34672654e+00 -3.67074236e-02 -3.41665238e-01 -7.09089816e-01 -1.69351518e-01 5.01292348e-01 9.23692226e-01 -3.19274038e-01 -3.55732590e-01 -2.65534192e-01 1.47112787e+00 3.78936380e-01 9.87264037e-01 -1.04132485e+00 -3.71671975e-01 9.25756633e-01 -1.29386216e-01 4.77994472e-01 -6.81410253e-01 -3.83530068e-03 -1.50789332e+00 -1.25210419e-01 -4.41374779e-01 8.40726420e-02 -1.07140493e+00 -1.06005681e+00 4.57138538e-01 3.81639004e-02 -1.47718954e+00 -7.63777018e-01 -4.97509599e-01 -9.08130348e-01 9.89899337e-01 -8.52362931e-01 -1.04714024e+00 -2.89602429e-01 7.23020256e-01 6.99478567e-01 -6.45000488e-02 6.48710787e-01 2.91199058e-01 -5.98269641e-01 3.21685255e-01 7.33160526e-02 4.35602129e-01 4.80011940e-01 -1.28344011e+00 8.46164465e-01 7.90290475e-01 1.29354075e-01 6.91653967e-01 6.55042410e-01 -6.29827440e-01 -1.50667143e+00 -1.43000460e+00 6.04403794e-01 -8.24663520e-01 5.77961922e-01 -2.32668817e-01 -9.53707695e-01 1.19271779e+00 1.51762441e-01 9.62523222e-02 3.46507460e-01 -4.88264561e-01 2.93443091e-02 4.44528133e-01 -6.83957756e-01 9.91538346e-01 1.43290496e+00 -5.62488377e-01 -4.71581101e-01 8.69750306e-02 8.65113795e-01 -7.79143035e-01 -8.63725483e-01 2.31373325e-01 6.44940376e-01 -7.62988985e-01 1.21226656e+00 -7.90503323e-01 1.03734386e+00 -3.87902290e-01 -1.90076292e-01 -1.45542359e+00 -7.12504447e-01 -6.66730404e-01 -3.96611780e-01 8.72216702e-01 5.10698408e-02 -3.41033936e-01 9.97152746e-01 4.97016132e-01 -2.21291542e-01 -7.84590423e-01 -7.25466132e-01 -5.74006677e-01 4.13705230e-01 -4.59760487e-01 7.66899407e-01 1.15561545e+00 -5.10800004e-01 1.12041667e-01 -6.65051639e-01 5.32661900e-02 4.73278105e-01 1.69698428e-03 1.02788484e+00 -5.42081118e-01 -7.17873633e-01 -5.37876844e-01 -1.48280710e-01 -1.58279169e+00 4.95388567e-01 -8.18941653e-01 9.73355323e-02 -1.45756018e+00 -2.04965919e-01 1.82986371e-02 1.93455040e-01 2.63585538e-01 -1.63672045e-01 3.65341245e-03 1.90211803e-01 4.40044999e-01 -2.43515953e-01 1.11480916e+00 1.76350164e+00 -1.50142521e-01 -2.67134309e-01 -6.49988428e-02 -2.29427025e-01 6.37807012e-01 6.37382388e-01 -1.51731566e-01 -5.40845096e-01 -8.02457809e-01 -6.20163754e-02 6.32759750e-01 6.26737237e-01 -8.86367261e-01 1.53372034e-01 -5.31313181e-01 5.51767647e-01 -4.39293563e-01 4.99961346e-01 -3.74957472e-01 6.16436541e-01 3.70861650e-01 -4.94875848e-01 2.44405195e-01 -1.07846573e-01 6.34964824e-01 -2.65035748e-01 1.99055254e-01 4.49265748e-01 -2.36995995e-01 -9.50479269e-01 8.30040872e-01 -3.39197189e-01 6.37436137e-02 7.65470505e-01 -2.11700335e-01 -1.18607543e-02 -6.98874831e-01 -1.04806435e+00 2.45154917e-01 6.60688937e-01 4.81506020e-01 5.37241697e-01 -1.87064159e+00 -5.47682285e-01 2.40796685e-01 -1.25359863e-01 3.04942548e-01 5.67553461e-01 5.72928071e-01 -8.92161429e-01 1.72395796e-01 -4.68858361e-01 -6.66518867e-01 -4.41240460e-01 4.08304691e-01 5.01497030e-01 7.73263797e-02 -9.10385549e-01 9.41642582e-01 2.91028351e-01 -4.31001604e-01 2.56464966e-02 -4.08410549e-01 -1.93869509e-02 -4.83332872e-01 3.14936757e-01 4.85953987e-01 -7.07313955e-01 -7.51513541e-01 -1.28816382e-03 5.75141966e-01 5.53382576e-01 -4.86438662e-01 1.26031625e+00 -1.30416617e-01 5.34092635e-02 6.25135064e-01 1.37910402e+00 -2.88709641e-01 -1.99893987e+00 -2.78079584e-02 -1.85182482e-01 -4.21144038e-01 -3.09568554e-01 -1.81636617e-01 -1.12698162e+00 9.80819106e-01 1.37524724e-01 -8.02987963e-02 8.34054708e-01 -1.69205844e-01 8.86492670e-01 1.73211485e-01 2.88213968e-01 -8.98891747e-01 2.26723492e-01 6.84136331e-01 1.13859308e+00 -9.62444007e-01 -5.54328322e-01 -1.04770903e-02 -8.74587238e-01 1.03327930e+00 4.80298549e-01 -5.81465364e-01 6.51486218e-01 4.18574922e-02 -4.60349359e-02 3.25327478e-02 -7.36601055e-01 1.17958136e-01 3.19645643e-01 6.65523708e-01 5.89831710e-01 2.44754165e-01 8.88084769e-02 4.10551637e-01 -5.56231737e-01 2.93385446e-01 3.03921938e-01 7.83205986e-01 5.63079081e-02 -1.16334891e+00 -2.16879681e-01 1.01388894e-01 -1.20181873e-01 2.37560228e-01 4.44862992e-02 7.96514928e-01 1.75964728e-01 3.00207645e-01 2.06827655e-01 -5.03985941e-01 1.78895816e-01 1.70809448e-01 5.32680213e-01 -3.48728538e-01 -2.50879340e-02 3.16871665e-02 -2.27695420e-01 -9.72154915e-01 -4.00088459e-01 -7.39354730e-01 -1.21697640e+00 -5.53257287e-01 5.46137631e-01 -1.11042023e-01 2.08023116e-01 5.59124410e-01 3.51524800e-01 7.46188879e-01 2.78248698e-01 -1.34506130e+00 -4.32708800e-01 -7.58862674e-01 -5.71342826e-01 8.91658366e-01 5.61783552e-01 -6.77734792e-01 -9.12746862e-02 6.05946779e-01]

[7.335374355316162, -0.17010927200317383]

91653456-fb70-4a65-b9bd-f6979819508c

gmd-controllable-human-motion-synthesis-via

2305.12577

https://arxiv.org/abs/2305.12577v1

https://arxiv.org/pdf/2305.12577v1.pdf

GMD: Controllable Human Motion Synthesis via Guided Diffusion Models

Denoising diffusion models have shown great promise in human motion synthesis conditioned on natural language descriptions. However, it remains a challenge to integrate spatial constraints, such as pre-defined motion trajectories and obstacles, which is essential for bridging the gap between isolated human motion and its surrounding environment. To address this issue, we propose Guided Motion Diffusion (GMD), a method that incorporates spatial constraints into the motion generation process. Specifically, we propose an effective feature projection scheme that largely enhances the coherency between spatial information and local poses. Together with a new imputation formulation, the generated motion can reliably conform to spatial constraints such as global motion trajectories. Furthermore, given sparse spatial constraints (e.g. sparse keyframes), we introduce a new dense guidance approach that utilizes the denoiser of diffusion models to turn a sparse signal into denser signals, effectively guiding the generation motion to the given constraints. The extensive experiments justify the development of GMD, which achieves a significant improvement over state-of-the-art methods in text-based motion generation while being able to control the synthesized motions with spatial constraints.

['Siyu Tang', 'Supasorn Suwajanakorn', 'Konpat Preechakul', 'Korrawe Karunratanakul']

2023-05-21

['motion-synthesis', 'imputation', 'imputation', 'imputation']

['computer-vision', 'computer-vision', 'miscellaneous', 'time-series']

[-1.48242256e-02 -2.24709794e-01 -3.59709077e-02 -2.93857474e-02 -4.69711870e-01 -3.01155329e-01 6.91455662e-01 -4.19577926e-01 -1.60699219e-01 6.06348753e-01 8.54442179e-01 2.23882809e-01 -1.30160317e-01 -8.33874941e-01 -4.43401188e-01 -8.54500294e-01 2.49618858e-01 1.18658632e-01 2.75327384e-01 -3.22441161e-01 2.30337873e-01 4.11187708e-01 -1.18114066e+00 7.39941671e-02 9.43800509e-01 4.43868637e-01 6.77219450e-01 6.04576945e-01 -4.88844104e-02 9.53093112e-01 -3.21665943e-01 1.40376553e-01 1.97120562e-01 -7.37508118e-01 -4.71955985e-01 2.93181092e-01 1.15609340e-01 -4.88686204e-01 -7.00406849e-01 8.43158364e-01 5.16068280e-01 5.84034324e-01 5.69702208e-01 -1.04799771e+00 -6.54733419e-01 1.82217956e-01 -6.65193439e-01 -1.37676135e-01 5.62905550e-01 4.62615043e-01 7.17544258e-01 -9.85320508e-01 1.12810123e+00 1.30414677e+00 4.13517147e-01 7.76598096e-01 -1.14897203e+00 -3.24594408e-01 3.31222683e-01 9.65190157e-02 -1.37254739e+00 -4.69729990e-01 1.05756748e+00 -5.81431508e-01 6.63898110e-01 3.51889133e-01 7.58854628e-01 1.24571621e+00 3.60256046e-01 8.70192647e-01 4.60832298e-01 -3.01033914e-01 3.36389184e-01 -2.94317544e-01 -3.22246194e-01 5.98891795e-01 -7.93925896e-02 7.58375898e-02 -7.95212567e-01 -1.25738367e-01 1.04452729e+00 -8.56796578e-02 -6.17698491e-01 -5.73666573e-01 -1.66337419e+00 7.45858490e-01 3.99506927e-01 2.87253469e-01 -6.36199832e-01 2.11258382e-01 -1.59348547e-02 -4.39893067e-01 3.53978515e-01 3.65190327e-01 6.80963248e-02 -1.46359459e-01 -1.18829763e+00 6.44537508e-01 4.61305112e-01 8.69489193e-01 8.06444347e-01 3.23570639e-01 -4.24463540e-01 6.46208882e-01 5.48160255e-01 6.60693526e-01 2.92329371e-01 -1.34728897e+00 5.85243702e-01 3.98620814e-01 3.21115643e-01 -1.65583134e+00 -1.20679840e-01 -2.64766634e-01 -1.22482550e+00 3.98416035e-02 3.06161225e-01 -2.36034796e-01 -8.23916614e-01 1.87967396e+00 5.53042591e-01 5.48680246e-01 5.29426821e-02 1.22506809e+00 4.74174440e-01 9.97149408e-01 -2.08167270e-01 -2.70384341e-01 7.72532940e-01 -1.06862068e+00 -1.08995175e+00 -2.66462803e-01 5.27327597e-01 -7.63152659e-01 8.17313731e-01 7.04487637e-02 -1.18740416e+00 -6.45526469e-01 -8.87901664e-01 -4.87407260e-02 2.91247636e-01 -3.13453786e-02 4.33999836e-01 1.59508780e-01 -1.06686223e+00 4.29905623e-01 -1.07031834e+00 -3.18754911e-01 5.40339909e-02 5.77713586e-02 -2.91455537e-01 -4.04197246e-01 -1.10801840e+00 5.50083935e-01 1.13978878e-01 3.68730038e-01 -8.30931902e-01 -5.50304592e-01 -1.16704392e+00 -2.25624263e-01 2.97417372e-01 -1.11483371e+00 8.52999330e-01 -5.56241870e-01 -1.63637638e+00 7.82815143e-02 -4.76001084e-01 -7.06087649e-02 6.98281467e-01 -2.08574891e-01 -6.07874170e-02 2.87392765e-01 2.73340076e-01 1.05543494e+00 7.98334658e-01 -1.35827434e+00 -5.02442539e-01 7.95689896e-02 -2.10242242e-01 5.02599061e-01 -2.59016424e-01 -2.52950996e-01 -9.11001027e-01 -1.09044969e+00 3.43974382e-01 -1.09711301e+00 -7.68974662e-01 1.44560114e-01 -4.78765070e-01 1.58100948e-01 1.01508212e+00 -7.36939669e-01 1.44989216e+00 -2.18108296e+00 9.64334667e-01 4.19645816e-01 1.58049539e-01 1.64091624e-02 -2.59978354e-01 4.66926306e-01 4.39806432e-01 -1.43097445e-01 -4.80280489e-01 -6.56162441e-01 -3.10909245e-02 3.43395591e-01 -2.35479981e-01 4.12410021e-01 1.31874710e-01 8.51553559e-01 -1.11364186e+00 -2.92829096e-01 4.18021917e-01 7.94783592e-01 -8.75474811e-01 3.23057353e-01 -2.50780672e-01 1.17464244e+00 -7.82230437e-01 2.61948824e-01 6.26194000e-01 -1.64665431e-02 -3.97082753e-02 -2.00740606e-01 -2.62413830e-01 -1.41222954e-01 -1.47095907e+00 2.27824283e+00 -2.32942998e-01 3.97287160e-01 2.03023672e-01 -4.83576894e-01 9.42707121e-01 2.49257013e-01 6.42112374e-01 -4.54079598e-01 -4.20941152e-02 1.19103834e-01 -3.09425414e-01 -4.45429862e-01 6.94468617e-01 7.50335380e-02 5.77741526e-02 2.73537725e-01 -3.51374805e-01 -3.58599991e-01 3.07388846e-02 3.22851807e-01 1.00250673e+00 4.54718024e-01 -4.71249409e-02 -1.32852510e-01 6.67114794e-01 1.57952085e-01 9.33916986e-01 4.71672893e-01 3.31679918e-02 1.22143114e+00 1.50627851e-01 -3.35396022e-01 -1.05522251e+00 -7.02715218e-01 4.50830549e-01 3.73883367e-01 3.94655943e-01 -4.66034085e-01 -8.41047645e-01 -2.93297380e-01 -3.24468166e-01 5.46538830e-01 -5.29266298e-01 -1.12386949e-01 -9.61017132e-01 -6.68693721e-01 6.82331100e-02 4.84587222e-01 6.76041901e-01 -9.78458107e-01 -5.03532946e-01 5.88911414e-01 -5.47658324e-01 -1.07941544e+00 -1.01667976e+00 -4.31416690e-01 -6.61118269e-01 -6.40169740e-01 -1.31960618e+00 -6.59426630e-01 7.24949658e-01 5.50613105e-01 6.21813953e-01 1.38681129e-01 -1.09436922e-01 3.48866194e-01 -4.49802190e-01 4.19927359e-01 -3.59732151e-01 -7.68615901e-02 3.16819176e-02 2.82671601e-01 -2.29325712e-01 -5.79194069e-01 -8.82792830e-01 4.18629318e-01 -8.72807980e-01 4.95027542e-01 3.97587597e-01 8.98678184e-01 5.54789484e-01 -1.82996355e-02 2.09044546e-01 -3.70993495e-01 5.02158940e-01 -6.17897332e-01 -1.76101491e-01 -7.86969904e-03 -6.45884275e-02 9.66331661e-02 3.71702611e-01 -5.20464540e-01 -1.28193808e+00 4.13533688e-01 -1.71489730e-01 -7.00715899e-01 4.38662358e-02 4.77989405e-01 -3.22679698e-01 -1.50983958e-02 4.73674774e-01 4.04971570e-01 5.19101731e-02 -3.17827404e-01 6.84185922e-01 2.25989595e-01 5.43217838e-01 -6.29344225e-01 9.03115094e-01 7.32754529e-01 3.77657972e-02 -1.05196416e+00 -3.14440131e-01 -4.56923038e-01 -7.99382448e-01 -2.90319651e-01 1.28411674e+00 -9.11934316e-01 -2.05531120e-01 6.26828969e-01 -1.33425832e+00 -4.75258231e-01 -1.22791767e-01 6.34861827e-01 -5.56759834e-01 6.60546064e-01 -6.56660199e-01 -6.59042239e-01 -7.41862832e-03 -1.23883319e+00 1.10832107e+00 3.44256200e-02 -5.78309774e-01 -1.05658984e+00 2.78082609e-01 1.78355768e-01 3.66645098e-01 5.36470532e-01 5.09588540e-01 3.68559808e-01 -1.00923800e+00 1.03555053e-01 1.51359528e-01 1.05836384e-01 3.16390514e-01 1.27507851e-01 -3.96489501e-01 -2.45705158e-01 2.95935255e-02 1.61451638e-01 7.32263446e-01 6.38122261e-01 5.56490421e-01 -2.77701676e-01 -4.59858030e-01 8.36603224e-01 1.05835485e+00 1.75553530e-01 6.71810448e-01 4.86681849e-01 1.16536939e+00 7.87716329e-01 7.15983570e-01 6.18514001e-01 4.75581318e-01 8.89956832e-01 9.01009142e-02 -1.48555130e-01 -3.50658774e-01 -6.16313994e-01 2.80254334e-01 1.05894649e+00 -1.04195893e-01 -3.55060339e-01 -8.11967194e-01 5.56906343e-01 -2.31131697e+00 -9.73384678e-01 -2.35614985e-01 1.95131183e+00 5.37909329e-01 -1.07972547e-01 -1.95552066e-01 5.04334969e-03 6.03744805e-01 6.16969883e-01 -3.70636046e-01 1.68228135e-01 -1.81824729e-01 -3.12866896e-01 1.45607024e-01 9.17751491e-01 -8.59126687e-01 1.01524031e+00 6.05117083e+00 7.55534112e-01 -1.02263701e+00 -1.07485197e-01 2.89147824e-01 6.83793351e-02 -7.69826293e-01 -3.85772176e-02 -6.20507240e-01 4.93101537e-01 3.84903431e-01 -1.57760382e-01 3.03839147e-01 5.94218373e-01 9.79645491e-01 -1.03125244e-01 -7.63858199e-01 8.72475624e-01 3.33376899e-02 -1.57219136e+00 3.15605432e-01 1.07538104e-01 1.12657094e+00 -4.31890249e-01 1.83958840e-02 -1.55790403e-01 2.41466314e-01 -7.68406212e-01 1.00857437e+00 7.39630461e-01 4.39122856e-01 -7.14684904e-01 4.08208221e-01 7.10575581e-01 -1.38372719e+00 1.64094180e-01 -3.03014249e-01 2.98936460e-02 7.86435604e-01 5.66072285e-01 -3.11723858e-01 7.18136549e-01 3.22989553e-01 1.07648051e+00 -2.09733695e-02 8.35007250e-01 -3.81574363e-01 3.02795589e-01 -1.32971615e-01 2.85164952e-01 4.30749863e-01 -4.67715949e-01 8.77330422e-01 1.02772176e+00 6.78516388e-01 3.38122845e-01 4.28685069e-01 1.04989302e+00 5.17391622e-01 2.24220920e-02 -7.56994724e-01 2.49604583e-01 4.24481690e-01 9.64493930e-01 -5.33114374e-01 -2.08676487e-01 -3.57730508e-01 1.27988958e+00 9.55747887e-02 7.44485617e-01 -7.75863051e-01 1.33889213e-01 7.48083055e-01 1.87766060e-01 1.37856349e-01 -8.98858905e-01 -1.73842922e-01 -1.23146701e+00 2.86473949e-02 -7.45705783e-01 -1.49438217e-01 -7.69260824e-01 -9.27384198e-01 5.51650703e-01 -1.09061062e-01 -1.51839554e+00 -5.12097955e-01 -1.96585041e-02 -6.92215800e-01 1.07926762e+00 -1.25038815e+00 -1.18273401e+00 -4.20229346e-01 7.10883975e-01 7.43775129e-01 2.41210330e-02 4.75069433e-01 3.36956233e-01 -5.60678124e-01 6.39867708e-02 -8.67618248e-02 -1.25073403e-01 5.65811217e-01 -8.46888065e-01 4.79727358e-01 1.17646933e+00 1.71627216e-02 6.16501391e-01 6.94605887e-01 -1.05188251e+00 -1.50622070e+00 -1.20852411e+00 7.17674434e-01 -3.17346781e-01 3.85391444e-01 -1.28078267e-01 -1.02200925e+00 4.42426234e-01 4.11116779e-02 -7.84069896e-02 2.54360408e-01 -5.61142385e-01 2.80405104e-01 3.11234564e-01 -8.28796685e-01 9.64709044e-01 1.31822956e+00 -2.60784984e-01 -4.22615230e-01 7.07381219e-02 7.24115074e-01 -7.07060456e-01 -3.76215160e-01 3.14675242e-01 3.64349306e-01 -8.07048976e-01 1.00977504e+00 -1.43166199e-01 4.82761979e-01 -7.31575012e-01 -1.63882717e-01 -1.39992809e+00 -6.65775239e-01 -1.05276453e+00 -1.74117267e-01 1.25275016e+00 1.94140658e-01 -1.36333123e-01 9.30345178e-01 7.76138306e-01 -1.48831725e-01 -5.53325832e-01 -8.05818021e-01 -5.91145992e-01 -1.04998589e-01 -3.89675230e-01 3.92150939e-01 1.06538975e+00 -1.74134552e-01 -3.55519243e-02 -9.56873298e-01 3.25234234e-01 5.62761486e-01 -2.43362233e-01 1.03262472e+00 -4.94611979e-01 -3.23970765e-01 -1.12793133e-01 -3.50557476e-01 -1.74797702e+00 1.66978523e-01 -5.52769721e-01 4.82636124e-01 -1.85273659e+00 5.25389239e-02 -4.22388762e-01 1.18536569e-01 8.33434537e-02 -4.40702140e-01 6.26841467e-03 3.49431336e-01 4.95866299e-01 -2.25143299e-01 1.12748158e+00 1.72392881e+00 -1.14349358e-01 -6.04109883e-01 -2.23253384e-01 -3.84049147e-01 6.79106653e-01 4.90451753e-01 -3.24447215e-01 -7.33719826e-01 -7.54978776e-01 -4.87223677e-02 4.73724425e-01 2.11459398e-01 -1.18048346e+00 5.64975142e-01 -4.90902603e-01 1.82173684e-01 -6.02876127e-01 5.10944068e-01 -6.72747135e-01 5.77175617e-01 4.56865698e-01 -1.67062268e-01 1.41209990e-01 -4.90633063e-02 8.49353492e-01 -4.03532833e-01 1.57881811e-01 5.74021757e-01 -1.53839245e-01 -8.92779708e-01 4.38091964e-01 -5.29533923e-01 -1.01001568e-01 1.04386675e+00 -3.44028920e-01 1.44583181e-01 -6.22144580e-01 -8.88010204e-01 4.44329083e-01 6.78746521e-01 4.50906098e-01 8.61372352e-01 -1.59659398e+00 -7.88378954e-01 3.21068078e-01 -2.62835681e-01 3.44839692e-01 5.75763583e-01 7.22885668e-01 -6.69883728e-01 1.86334997e-01 -8.17687213e-02 -8.05729508e-01 -1.05034316e+00 4.73399282e-01 1.78351596e-01 -1.85493082e-02 -1.03514099e+00 8.00917327e-01 4.46371883e-01 -1.43207476e-01 4.27537300e-02 -4.84152704e-01 -1.29536062e-01 -2.20388740e-01 5.89748800e-01 4.02365088e-01 -4.51584339e-01 -9.17146862e-01 -2.43550807e-01 9.84210253e-01 3.52141231e-01 -6.13741517e-01 1.18827796e+00 -4.33826506e-01 1.32008582e-01 1.64586887e-01 9.02289927e-01 2.90081590e-01 -1.85544419e+00 -1.12234205e-01 -1.50914297e-01 -8.66624057e-01 6.07213750e-03 -3.14435661e-01 -1.11759722e+00 7.62924969e-01 8.92536342e-02 -2.60492563e-01 9.18863356e-01 -3.15648645e-01 1.15608871e+00 -5.73495124e-03 5.94687700e-01 -9.00494516e-01 4.15607303e-01 4.91376072e-01 1.11388135e+00 -9.61850405e-01 -6.15197644e-02 -7.53803790e-01 -7.68347740e-01 9.22234595e-01 4.94323164e-01 -1.62133366e-01 4.64084983e-01 2.02432498e-01 1.33220375e-01 6.88284263e-02 -5.12097001e-01 -4.80160676e-02 3.97218227e-01 6.47918820e-01 1.62202865e-01 -8.88088420e-02 -2.86060899e-01 2.58008599e-01 7.49695674e-02 4.71618921e-02 4.17277306e-01 9.28332925e-01 -3.71707290e-01 -1.18677032e+00 -5.56243122e-01 -2.47729078e-01 9.94986370e-02 2.46279221e-02 -1.58459201e-01 6.44590199e-01 1.93257570e-01 9.79878664e-01 -1.70445949e-01 -3.72696400e-01 3.82827759e-01 -3.44371557e-01 1.53326690e-01 -6.69008017e-01 -7.95031786e-02 5.33112466e-01 6.91953748e-02 -8.52722704e-01 -5.43848395e-01 -7.55309641e-01 -1.39228630e+00 -3.76748115e-01 -2.44656615e-02 -7.68657262e-03 2.40586922e-01 8.18138838e-01 4.56797153e-01 5.75194597e-01 5.39217889e-01 -1.31755316e+00 -1.41125217e-01 -6.55498326e-01 -5.49718320e-01 4.88673478e-01 4.22871917e-01 -6.15365267e-01 -4.18217421e-01 3.15478295e-01]

[10.776041984558105, -0.7806311249732971]

aca9bb80-d1f3-419c-a5bc-9b1a415e46f6

convolutional-neural-networks-can-be-deceived

http://openaccess.thecvf.com/content_CVPR_2019/html/Gomez-Villa_Convolutional_Neural_Networks_Can_Be_Deceived_by_Visual_Illusions_CVPR_2019_paper.html

http://openaccess.thecvf.com/content_CVPR_2019/papers/Gomez-Villa_Convolutional_Neural_Networks_Can_Be_Deceived_by_Visual_Illusions_CVPR_2019_paper.pdf

Convolutional Neural Networks Can Be Deceived by Visual Illusions

Visual illusions teach us that what we see is not always what is represented in the physical world. Their special nature make them a fascinating tool to test and validate any new vision model proposed. In general, current vision models are based on the concatenation of linear and non-linear operations. The similarity of this structure with the operations present in Convolutional Neural Networks (CNNs) has motivated us to study if CNNs trained for low-level visual tasks are deceived by visual illusions. In particular, we show that CNNs trained for image denoising, image deblurring, and computational color constancy are able to replicate the human response to visual illusions, and that the extent of this replication varies with respect to variation in architecture and spatial pattern size. These results suggest that in order to obtain CNNs that better replicate human behaviour, we may need to start aiming for them to better replicate visual illusions.

[' Marcelo Bertalmio', ' Javier Vazquez-Corral', ' Adrian Martin', 'Alexander Gomez-Villa']

2019-06-01

cvpr-2019-6

['color-constancy']

['computer-vision']

[ 9.18366387e-02 -5.29212430e-02 5.39921761e-01 -2.63313740e-01 6.64914131e-01 -6.63261652e-01 8.65078330e-01 -1.36913419e-01 -3.79322588e-01 3.74414742e-01 1.60897553e-01 -2.94482529e-01 2.45877877e-01 -5.57523608e-01 -7.77271748e-01 -7.12658048e-01 2.26720661e-01 -1.93607926e-01 2.71710545e-01 -5.86078048e-01 5.60980439e-01 8.22228372e-01 -1.65653551e+00 3.83118212e-01 6.21190250e-01 5.67587733e-01 1.02978840e-01 7.62830675e-01 2.74053931e-01 9.99189973e-01 -5.89224875e-01 -3.78542870e-01 2.66057283e-01 -8.07705522e-01 -5.79478681e-01 2.10635364e-01 8.82174730e-01 -2.82830954e-01 -5.19850433e-01 1.25410414e+00 3.35887104e-01 2.84873396e-02 9.40916181e-01 -1.22101688e+00 -1.53003109e+00 -4.50020060e-02 -3.91952425e-01 6.08290434e-01 2.34869421e-01 6.59758627e-01 5.28319836e-01 -6.00989997e-01 5.92653334e-01 1.41025007e+00 7.51100183e-01 6.66713238e-01 -1.65878177e+00 -3.09361190e-01 3.68196629e-02 2.33010337e-01 -9.65069652e-01 -5.34069836e-01 5.05229890e-01 -3.41249883e-01 1.02567589e+00 2.91496545e-01 9.15619254e-01 1.22716951e+00 6.62179530e-01 4.13613021e-01 1.67931747e+00 -3.98394257e-01 2.50846654e-01 2.76611239e-01 2.20904406e-02 6.20027781e-01 7.41567791e-01 5.39788663e-01 -2.53400296e-01 2.78460234e-01 1.09679270e+00 3.71542834e-02 -6.19671881e-01 -2.51317233e-01 -1.18994045e+00 5.89904368e-01 8.34651649e-01 6.99928343e-01 -3.36674869e-01 4.02921766e-01 1.34559751e-01 5.93278527e-01 1.53692409e-01 9.96598661e-01 -2.21499920e-01 2.86391795e-01 -8.70244443e-01 1.74413472e-01 6.53705299e-01 3.08429301e-01 6.30831718e-01 5.96467435e-01 -4.68497425e-02 5.49666762e-01 1.00118421e-01 4.41837430e-01 7.41296768e-01 -1.17053914e+00 -4.06470090e-01 5.36034942e-01 -7.07196742e-02 -1.48284769e+00 -5.48681915e-01 -4.21453059e-01 -9.98660564e-01 1.02373469e+00 7.09653437e-01 1.99223474e-01 -1.07240224e+00 1.69979072e+00 -3.65927339e-01 -3.75044942e-02 1.28464982e-01 1.22772014e+00 6.20492458e-01 2.95783877e-01 -1.46530434e-01 8.50496143e-02 1.43387222e+00 -5.58578849e-01 -5.06873250e-01 -4.84828293e-01 3.31884295e-01 -8.34048390e-01 1.38973475e+00 5.33723772e-01 -1.11144507e+00 -8.64542007e-01 -1.27056468e+00 -2.04221562e-01 -5.07301092e-01 -2.41797730e-01 6.25644624e-01 9.03590560e-01 -1.31728566e+00 8.03600371e-01 -4.48830545e-01 -6.56884253e-01 4.52780366e-01 -1.18355177e-01 -4.60522234e-01 -1.73651293e-01 -7.26649940e-01 1.33858943e+00 2.18018621e-01 3.39917064e-01 -9.11418617e-01 -4.48771030e-01 -6.88906491e-01 1.26293346e-01 -2.23944649e-01 -7.25924909e-01 8.20615530e-01 -1.60247731e+00 -1.02537727e+00 1.23268723e+00 -1.91318467e-01 -4.23935115e-01 4.02801961e-01 1.92953482e-01 -3.77732605e-01 1.09068245e-01 -3.42555821e-01 7.22152770e-01 1.16138053e+00 -1.80972016e+00 -4.02460061e-02 -4.81395096e-01 8.53962824e-03 -1.42770559e-01 -4.35381792e-02 -8.95794183e-02 2.27042183e-01 -7.06270039e-01 3.42382848e-01 -1.01977706e+00 -1.92330275e-02 2.90257305e-01 -3.42833608e-01 2.55239069e-01 6.61633790e-01 -6.18057370e-01 4.28448051e-01 -2.25161767e+00 -2.69795824e-02 1.89988241e-02 5.57197928e-01 5.07227540e-01 -4.53060180e-01 4.23594326e-01 -4.99446124e-01 2.96227306e-01 -1.32162988e-01 -2.92605937e-01 1.06609918e-01 3.45019668e-01 -3.50380808e-01 8.54598522e-01 3.41837496e-01 1.26732099e+00 -6.74811959e-01 -1.36830937e-02 1.37330264e-01 6.31750345e-01 -3.97907048e-01 3.87909189e-02 -1.22084610e-01 6.54858351e-01 2.93442518e-01 4.01991725e-01 1.01307666e+00 -1.16568923e-01 -2.39925802e-01 -1.18143991e-01 -2.24192411e-01 -2.15220094e-01 -7.41463542e-01 1.16101718e+00 -3.33341025e-02 1.44415188e+00 -1.38967112e-01 -1.06126440e+00 6.95039988e-01 1.00901634e-01 -2.56018758e-01 -1.39776313e+00 3.44690233e-01 1.09805547e-01 5.63872159e-01 -7.02081442e-01 3.71304572e-01 -4.95010704e-01 5.19562900e-01 4.79311496e-01 -1.00349285e-01 -4.50989753e-01 -2.90573183e-02 1.84529647e-02 8.95763278e-01 -3.42155248e-01 -2.99875792e-02 -3.34443450e-01 3.08686703e-01 -1.46782190e-01 1.17423467e-01 9.78922784e-01 -2.07498178e-01 9.10015464e-01 5.63731492e-01 -7.25492716e-01 -1.20725501e+00 -1.16570437e+00 -6.79291561e-02 7.54067183e-01 6.50240555e-02 4.13510650e-01 -7.96372473e-01 -5.93049005e-02 -1.27049610e-01 9.92778480e-01 -1.09769642e+00 -5.70433497e-01 -2.51617849e-01 -6.65535688e-01 5.79305887e-01 1.13812290e-01 8.05229127e-01 -1.41360438e+00 -1.11231029e+00 -2.63219237e-01 1.42927527e-01 -9.72232878e-01 -7.49556199e-02 2.11496577e-01 -7.52889156e-01 -1.11998868e+00 -1.00156093e+00 -7.13787138e-01 8.53110552e-01 5.90501249e-01 1.17900443e+00 6.16330147e-01 -4.07343209e-01 4.14915115e-01 -3.69687885e-01 -4.82385874e-01 -6.15845799e-01 -3.59423369e-01 -2.97697484e-02 -1.34393666e-02 7.39286989e-02 -7.72522986e-01 -8.21895897e-01 1.82642370e-01 -1.45605206e+00 -1.96167260e-01 6.33239031e-01 7.87036359e-01 -1.55479729e-01 2.70246528e-02 8.57301131e-02 -6.15210593e-01 1.00584090e+00 -4.19302508e-02 -2.09665313e-01 5.51284626e-02 -5.64359128e-01 2.05707297e-01 6.20207489e-01 -6.35249078e-01 -9.30288017e-01 -2.71717399e-01 7.45124221e-02 -2.21792728e-01 -6.60447419e-01 2.96064198e-01 1.98744208e-01 -5.48933148e-01 1.05651736e+00 6.04754031e-01 1.91471428e-01 -1.57509342e-01 3.91066730e-01 1.56439587e-01 4.51401085e-01 8.91880542e-02 9.74868417e-01 8.15963745e-01 5.49094602e-02 -1.11623502e+00 -4.29617345e-01 8.74857455e-02 -7.17366695e-01 -2.83050030e-01 9.80519950e-01 -4.63003516e-01 -9.61486995e-01 7.99955368e-01 -1.38423228e+00 -5.53758383e-01 -3.39629352e-01 2.55630732e-01 -5.54788709e-01 4.22081828e-01 -4.52424914e-01 -7.71870434e-01 2.16640145e-01 -9.72145498e-01 5.44624567e-01 2.88000882e-01 -1.60485148e-01 -1.03388321e+00 8.31541866e-02 1.29406571e-01 7.81645417e-01 1.64483085e-01 1.22070539e+00 -2.86808103e-01 -5.55073619e-01 -3.33964750e-02 -7.33475208e-01 6.41996622e-01 -1.51225263e-02 -3.32443528e-02 -1.21695542e+00 -4.10983145e-01 2.96487063e-01 -3.34014654e-01 1.29026306e+00 5.01768708e-01 7.14118481e-01 -4.06418920e-01 1.33226708e-01 6.94124877e-01 1.44638336e+00 -1.67118292e-02 1.36697352e+00 3.86702865e-01 3.92332107e-01 8.17943931e-01 -2.12011188e-01 -1.06299348e-01 -9.91913974e-02 4.42610472e-01 6.87620580e-01 -6.26901090e-01 -4.15112376e-01 5.96382692e-02 2.47180656e-01 3.12579960e-01 -3.68433177e-01 -2.26609766e-01 -6.74018800e-01 5.13355970e-01 -1.54734635e+00 -1.27472579e+00 -5.46302080e-01 2.11604762e+00 6.14554644e-01 1.35138273e-01 -5.59367836e-02 1.24709629e-01 2.01844260e-01 1.16253778e-01 -5.05821466e-01 -8.22365820e-01 -5.83482862e-01 7.22114295e-02 4.87029135e-01 2.56887466e-01 -5.60352564e-01 9.19171095e-01 7.24571896e+00 1.51758000e-01 -1.42898357e+00 -2.00738803e-01 7.38252401e-01 3.56371164e-01 -2.43413419e-01 -4.96309735e-02 8.30763206e-02 1.87104017e-01 6.52762532e-01 2.60039479e-01 8.23076665e-01 1.88186273e-01 3.27120423e-01 -4.01584417e-01 -1.04409897e+00 8.89828265e-01 3.68449390e-01 -1.19778228e+00 1.61605358e-01 5.29088676e-02 6.76544726e-01 -1.22314103e-01 7.99265325e-01 9.07926820e-03 3.37137938e-01 -1.66099334e+00 6.56593144e-01 8.31634879e-01 2.20659524e-01 -3.22520643e-01 5.94974458e-01 2.87278056e-01 -3.54161739e-01 -7.95498267e-02 -7.48179615e-01 -4.51103449e-01 -1.66178942e-01 2.39704579e-01 -6.62185252e-01 -1.73850909e-01 7.64797747e-01 3.95350546e-01 -1.05506957e+00 1.16548491e+00 -1.86733738e-01 3.95809829e-01 2.70202637e-01 2.08161529e-02 1.56127200e-01 -1.02974758e-01 5.66676497e-01 1.11791646e+00 1.39121905e-01 -2.43195891e-01 -5.99869668e-01 1.40720510e+00 3.31342995e-01 -2.85955787e-01 -7.90172637e-01 2.49200803e-03 -1.07242085e-01 9.53640521e-01 -9.65319216e-01 -1.01838380e-01 -4.33538377e-01 1.65220368e+00 1.74752980e-01 7.76601315e-01 -5.73170841e-01 -1.61564320e-01 7.56140590e-01 1.88390389e-01 3.41039717e-01 -3.96361023e-01 -4.85334009e-01 -1.01509881e+00 -1.82317182e-01 -1.07824314e+00 -3.62637639e-01 -1.41968417e+00 -1.42917657e+00 6.81465626e-01 -3.27770442e-01 -7.47809112e-01 1.24092232e-02 -8.71590555e-01 -7.60136127e-01 9.37585831e-01 -1.57385004e+00 -1.09517479e+00 -6.68468297e-01 7.63213038e-01 8.91801864e-02 -7.22100809e-02 7.41775990e-01 -3.60166579e-01 -2.18245476e-01 4.28631037e-01 -6.15965128e-02 1.92469403e-01 6.36525512e-01 -1.23061669e+00 6.71846271e-01 8.23921204e-01 4.15136576e-01 7.76416719e-01 1.37230432e+00 -2.65261382e-01 -1.09347820e+00 -6.10270619e-01 5.67117333e-01 -4.57374573e-01 4.29898798e-01 -2.36780882e-01 -1.20346737e+00 5.90040624e-01 6.54814363e-01 8.11511949e-02 2.04274461e-01 -2.29891166e-01 -8.24285686e-01 9.65535492e-02 -1.16571844e+00 9.44554865e-01 8.74715805e-01 -5.31103551e-01 -7.57144809e-01 8.74544680e-02 4.01755393e-01 1.67252228e-01 -1.46435678e-01 3.92128229e-02 5.05139589e-01 -1.63366568e+00 1.16790175e+00 -8.55389416e-01 5.79636335e-01 -1.55779526e-01 3.14709805e-02 -1.67545116e+00 -5.43366253e-01 -2.47772306e-01 5.08725047e-01 6.21875703e-01 3.03179979e-01 -9.73121822e-01 4.87895757e-01 5.26642740e-01 2.22074091e-02 -7.10980780e-03 -8.18397760e-01 -8.40538621e-01 3.78008127e-01 -9.36633050e-02 -7.51208048e-03 1.11607242e+00 -2.77267694e-01 7.04569370e-02 -3.17865610e-01 6.40320629e-02 6.33411348e-01 -1.46007791e-01 6.89793766e-01 -1.35974860e+00 -1.90900981e-01 -7.63177693e-01 -6.27828658e-01 -7.29300797e-01 9.83584449e-02 -6.06390297e-01 -7.07520992e-02 -1.33702970e+00 1.84417471e-01 -3.64369527e-02 -2.28561282e-01 3.25340509e-01 1.02182642e-01 8.03047836e-01 3.86620581e-01 2.78920561e-01 -1.66063339e-01 2.58884639e-01 1.60983765e+00 -8.55159108e-03 -5.13047874e-02 -2.39109755e-01 -1.03577065e+00 7.67327249e-01 9.35334921e-01 -3.31280768e-01 -2.53058314e-01 -4.18349147e-01 4.53150600e-01 -4.14811462e-01 1.05515563e+00 -1.22399914e+00 2.61257023e-01 -2.01107442e-01 8.36297452e-01 4.75409627e-02 4.16406929e-01 -9.12286758e-01 -5.28900698e-02 8.07870865e-01 -2.46493191e-01 1.92005903e-01 4.19343680e-01 4.88148600e-01 -1.06088363e-01 -3.47576916e-01 1.03806746e+00 -5.87488711e-01 -7.41179705e-01 -2.33584434e-01 -8.85154963e-01 -2.13265400e-02 7.88071096e-01 -6.63335979e-01 -6.41996562e-01 -6.25570953e-01 -6.47265136e-01 -5.14894962e-01 8.84619355e-01 3.18940103e-01 7.84526467e-01 -9.28561211e-01 -7.34170079e-01 2.78368503e-01 9.72419232e-02 -8.03805113e-01 2.87740469e-01 9.20778811e-01 -7.52313673e-01 3.66589904e-01 -9.64714527e-01 -2.87287176e-01 -1.22424150e+00 8.31888974e-01 7.66536772e-01 4.29091752e-01 -4.71363425e-01 8.85544062e-01 4.27758455e-01 3.48895639e-02 -8.05084854e-02 -5.20343006e-01 -6.54460117e-02 -1.74477488e-01 5.75102925e-01 1.77899431e-02 -1.64969310e-01 -5.54986954e-01 -4.07865942e-02 3.56282353e-01 4.75103669e-02 4.50091399e-02 1.42991889e+00 -1.43989280e-01 -4.66271520e-01 4.88346875e-01 9.58256245e-01 -5.34370989e-02 -1.26830482e+00 5.97667247e-02 -1.77748933e-01 -6.27405822e-01 1.31718069e-01 -9.65381145e-01 -1.20785177e+00 1.12658250e+00 8.20377707e-01 6.73642337e-01 1.20326483e+00 -1.61522124e-02 2.37245008e-01 3.79313469e-01 5.19838855e-02 -7.63275504e-01 6.14455044e-01 5.83635211e-01 1.28518188e+00 -1.14718544e+00 -1.28263593e-01 -7.89382756e-02 -6.29786789e-01 1.10762954e+00 5.18398762e-01 -6.30776703e-01 4.15047348e-01 1.07615776e-02 2.25337341e-01 -4.84811097e-01 -5.66527009e-01 -5.85333407e-01 2.97358513e-01 1.14188325e+00 4.06822115e-01 -2.19180614e-01 8.07812586e-02 2.98548513e-03 -1.92021742e-01 -2.06354156e-01 8.82231534e-01 3.94779742e-01 -6.28833294e-01 -6.57280982e-01 -5.58038712e-01 5.17080491e-03 -1.79735854e-01 -2.22893760e-01 -9.42616284e-01 8.97969902e-01 3.08128774e-01 8.83996725e-01 2.49625668e-01 -3.21082056e-01 4.12839502e-01 1.67407526e-03 8.34202766e-01 -3.32254648e-01 -6.15838885e-01 -4.75332350e-01 -3.20745945e-01 -4.72060919e-01 -3.49226505e-01 -5.03925800e-01 -7.84792662e-01 -6.95602238e-01 5.87349981e-02 -6.31918788e-01 5.90345085e-01 7.89345026e-01 1.11752842e-02 5.30482233e-01 1.85314223e-01 -1.02098513e+00 -1.93920672e-01 -1.13561463e+00 -7.88761199e-01 9.38532352e-01 6.70319319e-01 -5.36863267e-01 -6.80417776e-01 2.56770402e-01]

[10.105268478393555, 2.3059496879577637]

59b212a8-be00-4f47-bd33-aaaa70a9ff86

d-score-a-white-box-diagnosis-score-for-cnns

2304.00697

https://arxiv.org/abs/2304.00697v1

https://arxiv.org/pdf/2304.00697v1.pdf

D-Score: A White-Box Diagnosis Score for CNNs Based on Mutation Operators

Convolutional neural networks (CNNs) have been widely applied in many safety-critical domains, such as autonomous driving and medical diagnosis. However, concerns have been raised with respect to the trustworthiness of these models: The standard testing method evaluates the performance of a model on a test set, while low-quality and insufficient test sets can lead to unreliable evaluation results, which can have unforeseeable consequences. Therefore, how to comprehensively evaluate CNNs and, based on the evaluation results, how to enhance their trustworthiness are the key problems to be urgently addressed. Prior work has used mutation tests to evaluate the test sets of CNNs. However, the evaluation scores are black boxes and not explicit enough for what is being tested. In this paper, we propose a white-box diagnostic approach that uses mutation operators and image transformation to calculate the feature and attention distribution of the model and further present a diagnosis score, namely D-Score, to reflect the model's robustness and fitness to a dataset. We also propose a D-Score based data augmentation method to enhance the CNN's performance to translations and rescalings. Comprehensive experiments on two widely used datasets and three commonly adopted CNNs demonstrate the effectiveness of our approach.

['Fei Zuo', 'XiaoFeng Wang', 'Yuqi Song', 'Xin Zhang']

2023-04-03

['medical-diagnosis']

['medical']

[ 5.31283498e-01 9.56071168e-02 2.05653712e-01 -5.98523140e-01 -2.81885564e-01 -3.12386751e-01 3.41105014e-01 -7.40735680e-02 -5.70285022e-01 7.03045905e-01 -3.83101195e-01 -3.96939009e-01 -2.75858074e-01 -7.63172448e-01 -5.86183727e-01 -6.39570177e-01 2.16729820e-01 2.53132153e-02 3.66581947e-01 -1.44847050e-01 1.99103907e-01 4.40483242e-01 -1.90713584e+00 6.18909374e-02 1.09365916e+00 1.16175473e+00 -1.89116314e-01 4.71219391e-01 3.37625444e-01 6.37085617e-01 -1.00380063e+00 -6.41583622e-01 1.73290953e-01 -5.94797313e-01 -6.28583550e-01 -2.27430731e-01 5.52052706e-02 -5.43009602e-02 -9.83080454e-03 1.26802528e+00 5.48322201e-01 -4.77726385e-02 3.99797827e-01 -1.62604249e+00 -5.55405140e-01 2.89209783e-01 1.90899428e-02 1.29351601e-01 -1.87103301e-01 3.38240474e-01 8.19500864e-01 -4.09099817e-01 2.47679710e-01 1.00855136e+00 5.34893095e-01 7.27987468e-01 -6.30666018e-01 -8.15440178e-01 -1.00868285e-01 2.42493570e-01 -1.30167484e+00 -3.14530909e-01 5.45746744e-01 -2.35773027e-01 7.11374342e-01 3.17758530e-01 5.99498034e-01 8.86802256e-01 2.83784598e-01 5.02770841e-01 7.77398169e-01 -3.41255635e-01 4.13804978e-01 1.67053565e-01 -1.56026363e-01 7.95813560e-01 5.71130931e-01 3.16794872e-01 -2.34706372e-01 1.69084117e-01 5.51333725e-01 -2.87554651e-01 -1.95667505e-01 -2.79616356e-01 -9.64609981e-01 7.74558365e-01 3.70083630e-01 3.32901120e-01 -2.34608799e-01 8.64422764e-04 5.47177196e-01 2.46271297e-01 1.66362166e-01 7.79633045e-01 -5.67610621e-01 -2.94743270e-01 -5.49571991e-01 3.50233018e-01 3.63076419e-01 6.50221765e-01 3.78712714e-01 1.84473231e-01 -3.18221956e-01 7.19679713e-01 8.27942044e-02 2.40972757e-01 6.79270148e-01 -6.89579129e-01 2.19703361e-01 1.07485473e+00 -3.16727191e-01 -9.97503817e-01 -3.14241141e-01 -5.91727316e-01 -9.26626503e-01 3.23398620e-01 5.67009933e-02 -1.44502923e-01 -9.33836102e-01 1.94224536e+00 2.37711787e-01 1.87972516e-01 1.33425817e-01 8.44665110e-01 9.50620472e-01 1.41624272e-01 -8.48899409e-02 2.81569123e-01 9.63586390e-01 -8.00041795e-01 -5.55325329e-01 -2.72825658e-01 8.34358096e-01 -4.08489406e-01 1.17508221e+00 2.93075264e-01 -9.46598113e-01 -7.67855108e-01 -1.41477811e+00 2.94230163e-01 -3.98675561e-01 2.46579066e-01 2.76354522e-01 9.59102809e-01 -9.94163454e-01 6.07530415e-01 -7.87690282e-01 -1.10856712e-01 6.07744694e-01 5.27438462e-01 -3.20411831e-01 -2.61203479e-03 -1.46743810e+00 1.02696455e+00 5.97110868e-01 3.28514427e-01 -8.34932745e-01 -3.86462897e-01 -7.91279912e-01 5.64757660e-02 2.65212119e-01 -5.09021997e-01 1.16280234e+00 -9.43891525e-01 -1.17781842e+00 5.93882442e-01 3.25584143e-01 -4.32788312e-01 4.85490739e-01 2.22630426e-01 -8.33532810e-01 -2.00865254e-01 -4.90255095e-02 8.87810469e-01 6.09668612e-01 -9.24828172e-01 -8.43947947e-01 -1.89809576e-01 8.33706558e-02 -3.13419886e-02 -5.65476775e-01 -5.06479740e-02 -6.74151897e-01 -5.91603279e-01 7.23563433e-02 -9.68106687e-01 -8.37542862e-02 3.39653809e-04 -5.29995203e-01 9.29569304e-02 7.71574557e-01 -4.36162502e-01 1.28959918e+00 -2.29741073e+00 -1.28638208e-01 3.54426801e-01 2.05126032e-01 6.89048052e-01 -3.25901479e-01 -3.08846593e-01 -4.53079417e-02 3.28794777e-01 -4.84158903e-01 1.53353542e-01 -6.25149757e-02 3.61066371e-01 2.47506112e-01 2.61337638e-01 9.14162159e-01 9.69399750e-01 -6.27571583e-01 -3.65188658e-01 -1.46043133e-02 6.20234311e-01 -5.50741613e-01 1.09353669e-01 1.69109125e-02 4.23578888e-01 -4.14307892e-01 6.44905508e-01 4.18078333e-01 -9.05293003e-02 -2.05607072e-01 -1.70870826e-01 3.04298431e-01 7.44645670e-02 -1.00030482e+00 1.04035568e+00 -8.43240842e-02 6.26954973e-01 -4.81284380e-01 -9.18035448e-01 1.19912493e+00 7.83211589e-02 3.19039077e-02 -1.01408112e+00 5.07020354e-01 2.93538392e-01 6.30412757e-01 -6.02205992e-01 5.37119448e-01 8.96175280e-02 -2.64788717e-02 3.27723265e-01 -1.42933011e-01 -3.22101638e-02 2.51725405e-01 -2.28525639e-01 1.23736691e+00 1.03180781e-01 -3.43388021e-02 8.69502053e-02 5.57982922e-01 -9.95036513e-02 8.08184803e-01 5.16666055e-01 -4.14420307e-01 7.62742698e-01 7.27033019e-01 -5.11072516e-01 -1.03563583e+00 -4.77708608e-01 -1.28856599e-01 5.77066541e-01 -4.50169183e-02 7.71350693e-03 -1.08159733e+00 -9.30934668e-01 -1.67736813e-01 7.58172631e-01 -8.80381227e-01 -9.40010130e-01 -2.53239751e-01 -9.17706192e-01 9.79819894e-01 7.51622260e-01 7.61024237e-01 -1.27846432e+00 -1.06909668e+00 -3.48838344e-02 -4.91454639e-03 -1.00005341e+00 -1.01945579e-01 1.48275226e-01 -7.01877296e-01 -1.26190460e+00 -4.50373709e-01 -6.77220404e-01 9.79940593e-01 -1.05961204e-01 9.14718449e-01 5.89332819e-01 -2.32006267e-01 -1.72205687e-01 -4.53717589e-01 -5.65552890e-01 -7.38683283e-01 1.55629605e-01 -5.30800521e-02 -8.82147159e-03 3.39424968e-01 -9.51021537e-02 -4.18665409e-01 6.70061886e-01 -1.33367872e+00 -7.88703188e-02 8.22055578e-01 8.57410371e-01 4.19832617e-01 3.64109993e-01 6.96612179e-01 -8.71953428e-01 1.05301058e+00 -1.11070663e-01 -5.02150595e-01 4.36902583e-01 -8.53243172e-01 1.79915264e-01 4.07364637e-01 -4.51146513e-01 -6.65390730e-01 -2.70999610e-01 -2.02436194e-01 -3.63124698e-01 -6.05270527e-02 6.74208999e-01 -3.50452274e-01 -2.22270578e-01 5.88865817e-01 -1.55726537e-01 2.20245644e-01 2.91704033e-02 -1.11973226e-01 4.93680328e-01 6.03961766e-01 -3.06900591e-01 6.19874775e-01 -2.75988523e-02 8.93712789e-02 -3.39568615e-01 -4.45739686e-01 1.24501385e-01 -4.66448933e-01 -3.26827377e-01 5.78343928e-01 -3.73225063e-01 -6.40855134e-01 5.82085252e-01 -9.44129765e-01 -1.78665504e-01 -1.74803644e-01 3.46674263e-01 -1.36972994e-01 1.05982900e-01 -7.85330459e-02 -5.73836565e-01 -3.58270228e-01 -1.61017668e+00 6.68878078e-01 1.21315099e-01 -2.74692386e-01 -9.11775172e-01 -1.72372296e-01 2.59188056e-01 6.13640904e-01 3.94679934e-01 1.19766164e+00 -7.58540630e-01 -8.48294273e-02 -5.98899126e-01 -2.46365353e-01 9.56960022e-01 3.13095711e-02 2.26690158e-01 -1.08689380e+00 -1.42629147e-01 -1.36802614e-01 -2.90340602e-01 6.70106590e-01 1.70255095e-01 1.30012858e+00 -1.55518278e-01 -8.96858647e-02 4.51868892e-01 1.01256227e+00 4.72105533e-01 9.84814823e-01 5.04433393e-01 4.15648311e-01 6.64226472e-01 6.75503135e-01 2.28913590e-01 4.32498604e-02 3.59622985e-01 6.42174959e-01 -2.46142745e-01 2.09056027e-02 -8.37432034e-03 1.56805739e-01 7.10201859e-01 3.68064977e-02 -4.16980952e-01 -9.53944802e-01 3.43746603e-01 -1.63128126e+00 -5.47297299e-01 8.43780860e-02 2.14934635e+00 6.20757043e-01 5.60317695e-01 -1.67868882e-01 7.12173462e-01 7.79618859e-01 -3.32115680e-01 -7.43798018e-01 -4.60360557e-01 -2.15411350e-01 2.23002717e-01 2.88013667e-01 -6.21133111e-02 -9.03380632e-01 5.44836760e-01 6.10371733e+00 5.54789245e-01 -1.35061216e+00 -6.19126856e-02 9.35455263e-01 -1.35562792e-02 -2.26687089e-01 -2.84851938e-01 -5.21215916e-01 4.65355396e-01 8.38336885e-01 7.45625943e-02 3.14326547e-02 7.85391212e-01 6.54128492e-02 -7.74717927e-02 -1.01877916e+00 7.43723929e-01 1.32947996e-01 -1.05252552e+00 -1.59227923e-02 2.39839870e-02 7.03896582e-01 -2.07600147e-01 3.99319738e-01 4.04896975e-01 -2.20352827e-04 -1.13428175e+00 8.46398413e-01 3.19610626e-01 8.02056432e-01 -9.38141108e-01 1.25881660e+00 1.27505586e-01 -6.79962218e-01 -1.05309322e-01 -2.82984704e-01 1.47685647e-01 -2.59758264e-01 3.86992365e-01 -1.13786554e+00 3.21608275e-01 8.47310424e-01 2.25653097e-01 -1.15123808e+00 1.14944935e+00 -4.52890337e-01 4.96145844e-01 -1.09938963e-03 -4.34261620e-01 2.28353351e-01 1.86567560e-01 7.30178058e-02 7.72551775e-01 4.90697831e-01 -3.97335291e-01 -5.00445604e-01 8.84590685e-01 -1.65785417e-01 -1.36952177e-01 -4.04251456e-01 -1.02686353e-01 4.12978828e-01 1.11982703e+00 -7.15020657e-01 -9.61989909e-02 -2.14073062e-01 6.58745110e-01 8.36219490e-02 1.40808687e-01 -1.00742435e+00 -5.47241330e-01 5.62079132e-01 5.03107049e-02 1.72610983e-01 2.45775729e-01 -3.94896448e-01 -5.98002672e-01 2.46292517e-01 -1.08726871e+00 1.25769272e-01 -6.99633062e-01 -8.54454517e-01 1.01079845e+00 -1.74013957e-01 -1.32818460e+00 -1.19237244e-01 -6.01548612e-01 -5.03064156e-01 6.28504932e-01 -1.43093967e+00 -9.10713732e-01 -4.19133723e-01 4.11588490e-01 2.09385511e-02 -4.84847784e-01 7.74620533e-01 5.29541254e-01 -9.60133493e-01 9.42143977e-01 -2.47295439e-01 3.18365425e-01 3.83651555e-01 -7.62222588e-01 5.54305494e-01 9.35487628e-01 -1.94612250e-01 4.17239368e-01 5.35460472e-01 -5.14990509e-01 -1.01827645e+00 -1.41167045e+00 7.72602022e-01 -1.89755723e-01 2.29159370e-01 -2.85775717e-02 -1.06476831e+00 2.30623156e-01 -1.46215469e-01 1.73575670e-01 6.21780574e-01 -1.84770033e-01 -1.53160319e-01 -1.95703268e-01 -1.33918226e+00 5.86782992e-01 8.19365203e-01 -1.83330610e-01 -2.20653132e-01 -1.14869714e-01 5.29325068e-01 -4.14479196e-01 -5.33332348e-01 8.23737621e-01 6.17558718e-01 -1.01243234e+00 5.21028936e-01 -6.83153987e-01 4.49217707e-01 -3.23797464e-01 5.80430776e-02 -1.22629559e+00 -1.49179131e-01 6.84352070e-02 3.53219181e-01 1.08862257e+00 7.66437709e-01 -7.78821528e-01 6.58215582e-01 7.38557041e-01 -2.35121205e-01 -9.10116732e-01 -9.91974235e-01 -6.58879042e-01 -2.03284249e-01 -7.18951881e-01 1.02450323e+00 8.60427618e-01 -3.21008831e-01 1.03015885e-01 -8.25855583e-02 1.65567130e-01 1.29787773e-01 -5.27614057e-01 6.26970708e-01 -1.27748811e+00 4.60356139e-02 -6.53711736e-01 -8.91881526e-01 -9.30690244e-02 9.99617130e-02 -6.30281627e-01 1.86604545e-01 -1.25992012e+00 5.12743033e-02 -5.21375716e-01 -5.01195133e-01 7.15035141e-01 -1.35285154e-01 5.10522306e-01 -2.87096240e-02 -1.22949839e-01 -5.79512239e-01 4.46664333e-01 1.24100089e+00 -2.65892208e-01 -5.07758707e-02 2.01176740e-02 -8.22318316e-01 4.56022114e-01 1.10205972e+00 -4.95889634e-01 -5.88866711e-01 -5.09833276e-01 5.38621187e-01 -4.21832383e-01 4.64362562e-01 -1.39852691e+00 2.36951634e-01 1.24717541e-01 2.84162849e-01 -1.96437910e-01 5.81629202e-02 -6.34359896e-01 2.47420311e-01 6.44031227e-01 -5.49932778e-01 4.10215944e-01 4.00216818e-01 2.54391044e-01 -4.19279844e-01 -5.46326697e-01 7.09130764e-01 1.91662684e-01 -7.00517654e-01 2.74853796e-01 -1.03399619e-01 -7.88484961e-02 1.14501512e+00 -3.37235540e-01 -1.44758552e-01 -1.66179091e-01 -2.16093808e-01 7.29380101e-02 4.12827164e-01 6.55355573e-01 1.01393712e+00 -1.37697911e+00 -7.57392287e-01 6.30546689e-01 4.24173653e-01 -1.87337715e-02 7.47951120e-02 6.95707381e-01 -5.66366315e-01 3.01346034e-01 -3.50152701e-01 -5.75504541e-01 -1.26862550e+00 3.04469347e-01 4.65389162e-01 4.66867648e-02 -2.02491611e-01 7.37323940e-01 -1.37334049e-01 -5.27554989e-01 4.16193843e-01 -5.47962189e-01 -3.34235698e-01 -2.57995367e-01 4.06898290e-01 2.03406855e-01 6.95411742e-01 -5.39031863e-01 -3.33584934e-01 1.88222170e-01 1.96414366e-02 7.89699182e-02 1.12808836e+00 3.28205496e-01 -4.69312705e-02 2.83548236e-02 1.05069649e+00 -7.18158782e-01 -1.07597172e+00 -2.15056781e-02 1.45938236e-03 -3.11955065e-01 2.25468993e-01 -9.26216543e-01 -1.53194809e+00 9.42817092e-01 8.62874627e-01 2.29420558e-01 1.31424260e+00 -3.61892641e-01 5.02320051e-01 3.91582072e-01 5.19937277e-02 -1.17726123e+00 9.45219249e-02 3.48052800e-01 7.39280581e-01 -1.14199054e+00 -3.50826800e-01 -8.42000842e-02 -7.88748205e-01 9.92655396e-01 7.85014927e-01 4.43160415e-01 4.72635508e-01 3.20530444e-01 1.11481339e-01 -6.13263324e-02 -7.33779311e-01 4.94399890e-02 3.49398226e-01 6.22371733e-01 2.53392637e-01 -9.91385654e-02 -1.20241046e-01 6.37436569e-01 -1.88456282e-01 2.36308455e-01 2.52144188e-01 1.15998495e+00 -2.59038985e-01 -9.46170926e-01 -3.58969390e-01 6.01146400e-01 -3.00771624e-01 1.33041829e-01 -4.03061271e-01 7.32031822e-01 5.12229264e-01 8.75373006e-01 1.13832675e-01 -9.88489985e-01 5.57689726e-01 -6.01210706e-02 2.22707063e-01 -2.97553986e-01 -5.60153008e-01 -4.06886965e-01 4.14022766e-02 -3.15425277e-01 -4.19784904e-01 -5.02188265e-01 -1.17065310e+00 -3.24727178e-01 -5.52894652e-01 -7.69477338e-03 8.07654798e-01 9.49711502e-01 3.92067164e-01 9.42690015e-01 5.15278161e-01 -1.52704149e-01 -6.07076526e-01 -9.38135147e-01 -2.12545604e-01 3.66100401e-01 5.41748255e-02 -6.88003123e-01 -1.67212710e-01 -2.08879739e-01]

[6.555032730102539, 7.592701435089111]

a1576a31-a666-40b3-9b11-e55c216576cb

sequential-embedding-induced-text-clustering

1811.12500

http://arxiv.org/abs/1811.12500v1

http://arxiv.org/pdf/1811.12500v1.pdf

Sequential Embedding Induced Text Clustering, a Non-parametric Bayesian Approach

Current state-of-the-art nonparametric Bayesian text clustering methods model documents through multinomial distribution on bags of words. Although these methods can effectively utilize the word burstiness representation of documents and achieve decent performance, they do not explore the sequential information of text and relationships among synonyms. In this paper, the documents are modeled as the joint of bags of words, sequential features and word embeddings. We proposed Sequential Embedding induced Dirichlet Process Mixture Model (SiDPMM) to effectively exploit this joint document representation in text clustering. The sequential features are extracted by the encoder-decoder component. Word embeddings produced by the continuous-bag-of-words (CBOW) model are introduced to handle synonyms. Experimental results demonstrate the benefits of our model in two major aspects: 1) improved performance across multiple diverse text datasets in terms of the normalized mutual information (NMI); 2) more accurate inference of ground truth cluster numbers with regularization effect on tiny outlier clusters.

['Sargur N. Srihari', 'Tiehang Duan', 'Xiaohui Xie', 'Qi Lou']

2018-11-29

['text-clustering']

['natural-language-processing']

[-4.44404602e-01 -3.36472601e-01 -2.83070207e-01 -5.01564622e-01 -6.47031367e-01 -2.42085919e-01 9.86249268e-01 3.74560863e-01 -5.44085205e-01 3.67815346e-01 5.73300421e-01 5.21345288e-02 -2.02839002e-01 -5.97058475e-01 -2.35534802e-01 -9.79387164e-01 -1.54319257e-01 8.41939688e-01 1.02767251e-01 4.00751591e-01 5.06784856e-01 2.32163090e-02 -1.16651297e+00 9.94765833e-02 6.56120956e-01 4.50954407e-01 3.10370266e-01 7.25868881e-01 -4.76930380e-01 3.43091875e-01 -6.30494475e-01 -2.85139024e-01 3.31284441e-02 2.52551045e-02 -2.70110548e-01 1.40943751e-01 -2.52731860e-01 -3.89345169e-01 -6.31154597e-01 1.05903172e+00 5.21279573e-01 4.89046454e-01 1.33335364e+00 -1.31503546e+00 -9.70193267e-01 8.97675097e-01 -1.20188332e+00 6.45984784e-02 4.00139578e-02 -1.92856848e-01 1.31271827e+00 -1.17422426e+00 -4.12204452e-02 1.67452896e+00 5.76409638e-01 1.43573508e-01 -1.22681868e+00 -8.01610231e-01 2.84794301e-01 2.55785942e-01 -1.93920100e+00 -1.07818671e-01 5.40928781e-01 -5.34779608e-01 1.09188199e+00 8.29694420e-03 3.31129581e-01 1.06292427e+00 2.23899961e-01 8.77639294e-01 5.18475592e-01 -5.72780728e-01 3.61624539e-01 2.81117618e-01 6.73255801e-01 4.06855226e-01 6.01601839e-01 -3.74103278e-01 -8.05237830e-01 -7.38550246e-01 2.85980880e-01 4.16150540e-01 -1.64201945e-01 -2.37106755e-01 -9.12639976e-01 1.33604205e+00 -2.94957310e-01 1.18490167e-01 -4.46472429e-02 4.23633099e-01 2.83419371e-01 -4.31307822e-01 6.21217787e-01 -2.55807728e-01 -2.69390255e-01 -3.07690382e-01 -1.30144358e+00 5.65302186e-02 7.98568666e-01 1.41790390e+00 6.41596973e-01 -2.76609719e-01 -1.26021177e-01 9.55037594e-01 1.24444735e+00 3.63103986e-01 1.06238306e+00 -3.74223918e-01 4.48453337e-01 3.29462230e-01 7.07627833e-02 -1.11808038e+00 -1.15923405e-01 1.05393171e-01 -8.91033709e-01 -5.18453360e-01 -1.02886066e-01 8.76164660e-02 -1.00119197e+00 1.39362669e+00 2.27799445e-01 4.34572279e-01 -5.30839078e-02 4.59666729e-01 6.41350389e-01 1.05231452e+00 1.09637469e-01 -1.79979026e-01 1.58541977e+00 -8.17267656e-01 -1.09362698e+00 -3.85119803e-02 5.47172487e-01 -7.53762245e-01 7.59438634e-01 3.73226523e-01 -7.39046991e-01 -3.30067873e-01 -9.71021771e-01 -4.91231978e-02 -4.91610378e-01 -5.59967905e-02 5.80706477e-01 8.20162117e-01 -8.04917455e-01 2.46137887e-01 -1.08564222e+00 -3.17880392e-01 4.35182571e-01 4.36432660e-01 -4.27390188e-02 -1.52235210e-01 -1.17996621e+00 3.63869876e-01 7.44489491e-01 -5.42464815e-02 -7.88981616e-01 -5.45091093e-01 -7.99223185e-01 3.43662530e-01 5.90860797e-03 -4.28947955e-01 8.16419244e-01 -1.43390834e-01 -1.27026355e+00 6.10570371e-01 -4.41546470e-01 -2.61365801e-01 8.52724239e-02 -2.45192483e-01 -3.61223102e-01 1.88754395e-01 1.67375147e-01 7.09927678e-01 9.75356340e-01 -1.31495988e+00 -5.62446058e-01 -3.35464090e-01 -9.59828138e-01 1.87712133e-01 -8.88058364e-01 -1.99249506e-01 -8.29668224e-01 -9.65406656e-01 1.98432490e-01 -7.82554567e-01 3.00144758e-02 -2.64385819e-01 -6.51297569e-01 -6.12344801e-01 9.18934822e-01 -5.47191739e-01 1.62006748e+00 -2.46041608e+00 -2.46992141e-01 2.84542888e-01 3.24657977e-01 -1.21410526e-01 6.29399940e-02 7.79236317e-01 4.82670888e-02 2.62626886e-01 -4.72328924e-02 -8.27587485e-01 3.98350805e-01 3.83378237e-01 -6.65692925e-01 8.46690476e-01 -1.73018172e-01 5.46143413e-01 -7.76080549e-01 -8.44645023e-01 7.72278458e-02 4.69601780e-01 -5.59246838e-01 2.19530657e-01 -1.30743325e-01 -2.41583660e-01 -4.47268963e-01 2.63587445e-01 9.58810806e-01 -2.85207629e-01 2.84986824e-01 -1.95005164e-02 3.93611580e-01 -1.46431282e-01 -1.53714657e+00 1.65407670e+00 -1.35196120e-01 4.39172685e-01 -2.30289087e-01 -8.23974311e-01 8.46179366e-01 4.45487112e-01 4.08157498e-01 2.72049993e-01 8.73502865e-02 -2.71283656e-01 -2.34435767e-01 -3.94965291e-01 8.04462314e-01 -1.44278139e-01 -1.88296661e-01 8.56732011e-01 3.46103370e-01 -4.40991148e-02 1.66766167e-01 6.68302715e-01 8.53865921e-01 -3.27602684e-01 2.54408896e-01 -3.99149030e-01 9.55211148e-02 -3.53235871e-01 4.01237130e-01 9.79707837e-01 -9.45329145e-02 5.80086768e-01 3.73707622e-01 4.00575921e-02 -9.57718432e-01 -1.38833380e+00 -4.07920510e-01 1.25035393e+00 2.89770514e-01 -8.85806978e-01 -2.90822238e-01 -3.52042258e-01 1.86687857e-01 1.08734322e+00 -5.34398437e-01 -2.15565920e-01 1.36980200e-02 -1.36489034e+00 6.37184322e-01 5.52409232e-01 6.99006766e-02 -4.40390021e-01 -6.93877041e-02 1.21731311e-01 -1.26968995e-01 -1.01417542e+00 -7.77682185e-01 3.39760333e-01 -8.63836110e-01 -8.10811698e-01 -4.80767071e-01 -7.31126904e-01 6.82606399e-01 4.05086815e-01 7.57794321e-01 -3.73868018e-01 -5.52299798e-01 5.63333809e-01 -5.99564910e-01 -3.57560217e-01 -1.04802199e-01 -3.44229162e-01 4.91856843e-01 -3.12355720e-02 1.19917190e+00 -5.57952225e-01 -4.85800982e-01 1.36311427e-01 -1.22634816e+00 -3.28178287e-01 3.76908123e-01 9.17208374e-01 2.22286865e-01 6.17445052e-01 3.41246337e-01 -6.81198657e-01 9.06797111e-01 -7.45896220e-01 -3.39568555e-01 8.65934342e-02 -8.97170603e-01 2.29622066e-01 3.12604964e-01 -5.82973123e-01 -1.07114589e+00 -2.54392356e-01 3.82157326e-01 -3.02630901e-01 -1.57339618e-01 3.31718475e-01 -9.99594107e-02 8.50503206e-01 2.21598700e-01 3.50089520e-01 -2.51104385e-01 -5.38657308e-01 8.34634662e-01 1.41550171e+00 3.49032283e-01 -7.34577358e-01 5.92280865e-01 7.51684308e-01 -3.51631254e-01 -8.44424188e-01 -6.21313751e-01 -1.35693657e+00 -7.51144469e-01 2.84723669e-01 1.01020122e+00 -1.36742020e+00 -6.66454613e-01 2.19671592e-01 -1.21496391e+00 2.58819938e-01 1.55803517e-01 7.73735940e-01 -3.16602588e-01 1.03953063e+00 -6.50911093e-01 -1.10240304e+00 -2.30006844e-01 -9.39575016e-01 1.24431813e+00 -1.60697892e-01 -4.01037961e-01 -1.32773960e+00 2.16437086e-01 3.04174542e-01 -1.51339725e-01 -3.38142514e-01 1.27915061e+00 -1.08326411e+00 -8.90403539e-02 -4.91845161e-01 -3.02210420e-01 3.10195893e-01 1.88577250e-01 1.62868634e-01 -9.50106561e-01 -4.36656117e-01 -3.72659862e-02 -1.20125309e-01 1.16838670e+00 4.63551164e-01 1.14018118e+00 -2.29745328e-01 -7.08241642e-01 2.89927959e-01 1.51736867e+00 -6.30536079e-02 4.51881170e-01 -7.01952055e-02 7.59832621e-01 3.43210697e-01 3.21824998e-01 1.04457772e+00 3.86062890e-01 3.87719721e-01 2.52513915e-01 5.90865910e-01 3.39139104e-01 -3.04529905e-01 4.87987816e-01 1.41857016e+00 7.00108230e-01 -6.38290823e-01 -1.08729672e+00 9.58073676e-01 -2.10371900e+00 -8.58754337e-01 -4.54256356e-01 1.96454489e+00 1.09122741e+00 1.34106815e-01 -9.27760229e-02 8.78335834e-02 1.20743239e+00 7.59483427e-02 -1.84629545e-01 -2.15363860e-01 1.30890280e-01 -2.29836345e-01 5.44554710e-01 4.95509505e-01 -1.01794970e+00 9.32891309e-01 5.91314125e+00 1.26733255e+00 -1.31269693e-01 3.43676716e-01 3.58439952e-01 -1.97864920e-01 -1.88147292e-01 3.44844200e-02 -1.26097906e+00 1.00788391e+00 1.00578785e+00 -8.47328007e-02 1.12307131e-01 6.70456588e-01 1.98020309e-01 -3.12476546e-01 -1.09880495e+00 1.26604509e+00 4.12835300e-01 -8.55827868e-01 3.37270737e-01 1.95037842e-01 8.22612107e-01 -2.21464455e-01 1.85504273e-01 2.15321377e-01 7.91135967e-01 -1.05706644e+00 4.12140220e-01 5.54474711e-01 5.92142224e-01 -9.81218338e-01 7.02944636e-01 4.88991439e-01 -8.83540511e-01 -3.98110561e-02 -8.37395012e-01 2.87147790e-01 2.51067579e-01 1.21070385e+00 -1.05111468e+00 3.76649439e-01 7.74243474e-01 6.44495189e-01 -3.31338257e-01 7.20416903e-01 -2.37868997e-04 9.29457426e-01 -4.81626451e-01 -3.59540939e-01 3.68838042e-01 -4.02984589e-01 4.32540804e-01 1.64795208e+00 2.31347248e-01 -1.24114543e-01 9.70802084e-02 8.88973117e-01 -7.26008974e-03 9.46244970e-02 -3.45549196e-01 -4.08205420e-01 9.17626977e-01 1.21437383e+00 -8.92639697e-01 -3.43637317e-01 -3.71020347e-01 9.66643512e-01 6.39099479e-02 3.96832287e-01 -9.11709845e-01 -4.69096690e-01 4.40696836e-01 -1.89186141e-01 7.80882955e-01 -5.49140453e-01 -2.97180921e-01 -1.08485031e+00 -3.49928081e-01 -3.13071549e-01 3.54287714e-01 -5.24000287e-01 -1.86961889e+00 -1.72114391e-02 3.15631032e-01 -9.56065357e-01 -3.45312767e-02 -6.20704591e-01 -6.02938116e-01 7.28488266e-01 -1.05120361e+00 -9.96589839e-01 -5.04494011e-02 5.24110675e-01 5.76534390e-01 -4.10285324e-01 9.57478344e-01 1.06056489e-01 -5.77190399e-01 6.23822987e-01 1.00207829e+00 1.85394913e-01 1.09389269e+00 -1.47122824e+00 1.92778602e-01 4.66487855e-01 3.28422129e-01 9.29941416e-01 6.45215034e-01 -7.85280347e-01 -1.15376794e+00 -1.13304210e+00 6.38811171e-01 -5.67722917e-01 8.65068197e-01 -7.46543169e-01 -6.51328862e-01 6.13712311e-01 1.17972709e-01 -3.45221490e-01 1.41978276e+00 1.14526875e-01 -4.68458742e-01 1.93677142e-01 -9.72146988e-01 6.39539063e-01 3.13457191e-01 -2.86197633e-01 -9.13977444e-01 5.85510433e-01 8.70579779e-01 3.69210273e-01 -7.85130560e-01 -2.32415691e-01 3.69024813e-01 -5.54297566e-01 9.21796978e-01 -5.94132543e-01 3.05183530e-01 -2.98691303e-01 -5.81146300e-01 -1.16518247e+00 -4.56053644e-01 -5.13492405e-01 -6.88455224e-01 1.59336662e+00 7.84641057e-02 -3.28244328e-01 6.82760239e-01 4.31109935e-01 2.71812856e-01 -5.20357311e-01 -1.05225992e+00 -7.28583336e-01 1.04767315e-01 -5.32168686e-01 4.62985009e-01 9.49458838e-01 4.97146070e-01 4.61827874e-01 -4.49293166e-01 2.77723432e-01 8.11171293e-01 -1.60807997e-01 6.60054505e-01 -1.25261748e+00 -3.91670018e-01 -2.14891091e-01 -5.83567679e-01 -1.39093876e+00 4.98014003e-01 -9.22453105e-01 3.02987814e-01 -1.37467980e+00 8.02450955e-01 -1.27914354e-01 -1.71668455e-01 6.18214682e-02 -4.40545022e-01 -8.72317553e-02 -2.15909362e-01 5.15115142e-01 -8.37749958e-01 9.51958656e-01 3.79150778e-01 -2.01087400e-01 9.86135602e-02 -3.96276176e-01 -5.66609561e-01 7.38836229e-01 6.42661154e-01 -8.36418986e-01 -4.42351282e-01 -3.91847432e-01 1.46577567e-01 -4.08692986e-01 2.25921441e-02 -5.56390584e-01 6.86054111e-01 1.30383551e-01 3.62986445e-01 -9.91649508e-01 5.41932464e-01 -8.17857027e-01 -3.35535914e-01 5.51780313e-02 -2.25008339e-01 5.40370643e-02 9.38597247e-02 1.49843562e+00 -1.17477357e-01 -6.15921736e-01 5.73008776e-01 1.24608770e-01 -3.41611564e-01 1.20914519e-01 -7.18763769e-01 -1.76745001e-02 1.06402528e+00 -2.08035201e-01 -2.03229636e-02 -4.77329165e-01 -5.19748628e-01 3.79385918e-01 1.84471071e-01 2.23223880e-01 7.19747841e-01 -1.37835145e+00 -6.87793851e-01 3.21384631e-02 1.86694771e-01 1.96744993e-01 8.09305385e-02 5.37464380e-01 -4.09001619e-01 5.11497200e-01 6.10117316e-01 -7.27919996e-01 -1.01746774e+00 7.46064782e-01 -2.44717717e-01 -3.03381979e-01 -6.07622027e-01 1.00624561e+00 2.53718227e-01 -4.15726811e-01 5.77418745e-01 4.56174351e-02 -2.72519253e-02 2.48153955e-01 5.07048786e-01 4.62294817e-01 -1.94903836e-01 -3.81943643e-01 -3.58259499e-01 3.26568753e-01 -5.56382060e-01 -4.31497127e-01 1.35258293e+00 -4.68864411e-01 -4.47193176e-01 8.20487022e-01 1.39853442e+00 -7.43633583e-02 -1.11506677e+00 -4.01514053e-01 1.54620066e-01 -5.14046788e-01 2.10206583e-01 -2.97544628e-01 -3.96379977e-01 9.51181173e-01 5.57228267e-01 1.47080928e-01 4.99854416e-01 -2.10673753e-02 9.01825726e-01 2.89940894e-01 -1.61456332e-01 -1.44526756e+00 4.13481712e-01 4.77734447e-01 1.74342901e-01 -1.21366596e+00 3.52466673e-01 -2.77530998e-01 -6.34675324e-01 1.30166626e+00 2.52171665e-01 -3.23324911e-02 1.31569934e+00 3.17630589e-01 -2.17920884e-01 -3.04575115e-01 -7.99847007e-01 2.13969827e-01 1.59286126e-01 5.94453752e-01 4.40862149e-01 7.58058624e-04 -1.31218135e-01 9.93377686e-01 -7.37945437e-02 -5.51186681e-01 5.12500107e-01 7.30794191e-01 -7.10999787e-01 -8.64731371e-01 -6.47322297e-01 6.41097188e-01 -5.00934720e-01 -5.56269705e-01 -3.58691402e-02 3.79537374e-01 -1.03326946e-01 1.36925530e+00 4.03727651e-01 -1.02953732e-01 -5.18247306e-01 2.84986436e-01 8.78693312e-02 -9.85921383e-01 1.14148945e-01 3.50824863e-01 -4.16376054e-01 1.27715051e-01 -8.37421715e-02 -5.64752460e-01 -1.33293545e+00 -3.44439685e-01 -6.81550324e-01 4.24498171e-01 8.11515927e-01 8.94749463e-01 3.78142446e-01 3.61184120e-01 4.68337476e-01 -8.41300428e-01 -9.91571486e-01 -1.33505845e+00 -1.26041877e+00 4.00682747e-01 -4.59030457e-02 -5.10206223e-01 -7.11435020e-01 3.16335738e-01]

[10.397241592407227, 6.928426742553711]

ced6984f-7d25-4d1f-9a00-99f5b8887381

automatic-video-object-segmentation-via

1912.01373

https://arxiv.org/abs/1912.01373v1

https://arxiv.org/pdf/1912.01373v1.pdf

Automatic Video Object Segmentation via Motion-Appearance-Stream Fusion and Instance-aware Segmentation

This paper presents a method for automatic video object segmentation based on the fusion of motion stream, appearance stream, and instance-aware segmentation. The proposed scheme consists of a two-stream fusion network and an instance segmentation network. The two-stream fusion network again consists of motion and appearance stream networks, which extract long-term temporal and spatial information, respectively. Unlike the existing two-stream fusion methods, the proposed fusion network blends the two streams at the original resolution for obtaining accurate segmentation boundary. We develop a recurrent bidirectional multiscale structure with skip connection for the stream fusion network to extract long-term temporal information. Also, the multiscale structure enables to obtain the original resolution features at the end of the network. As a result of two-stream fusion, we have a pixel-level probabilistic segmentation map, which has higher values at the pixels belonging to the foreground object. By combining the probability of foreground map and objectness score of instance segmentation mask, we finally obtain foreground segmentation results for video sequences without any user intervention, i.e., we achieve successful automatic video segmentation. The proposed structure shows a state-of-the-art performance for automatic video object segmentation task, and also achieves near semi-supervised performance.

['Wonkyo Seo', 'Sungkwon Choo', 'Nam Ik Cho']

2019-12-03

['foreground-segmentation']

['computer-vision']

[ 5.19755781e-01 -2.91789085e-01 -1.86081558e-01 -2.25241706e-01 -8.74263883e-01 -3.17942977e-01 3.21783751e-01 -7.97387734e-02 -4.87331122e-01 4.33555514e-01 -2.25048587e-01 1.13024630e-01 4.30444404e-02 -8.44438970e-01 -7.03822017e-01 -9.67168570e-01 1.42510682e-01 1.34349540e-01 9.92649853e-01 2.99285620e-01 1.71472445e-01 3.51804346e-01 -1.53751457e+00 5.39670646e-01 8.91566217e-01 1.25152588e+00 3.70644480e-01 8.38027298e-01 -6.35255218e-01 9.01732564e-01 -4.47108269e-01 1.82716809e-02 7.08474442e-02 -5.83204210e-01 -8.10612142e-01 5.81314385e-01 4.06732708e-01 -4.69271839e-01 -1.56257048e-01 1.26888967e+00 1.10524178e-01 1.44592971e-01 3.98880988e-01 -1.15348065e+00 1.96739007e-02 6.74805880e-01 -9.29656744e-01 4.46741313e-01 6.46132976e-02 2.12647915e-01 7.72346556e-01 -8.12462747e-01 6.39875472e-01 1.21212339e+00 4.02438909e-01 1.03115477e-01 -9.77585316e-01 -4.41322505e-01 5.01776159e-01 2.98495024e-01 -1.24410808e+00 -2.46986777e-01 9.54332948e-01 -5.15300155e-01 4.66927737e-01 1.17888771e-01 8.96315634e-01 6.33448780e-01 4.23742011e-02 1.34516418e+00 9.36474860e-01 -6.49648309e-02 1.90010205e-01 -7.60692582e-02 3.58250827e-01 6.41278982e-01 -4.52493243e-02 -2.47187003e-01 -3.85017276e-01 1.94947958e-01 1.12062681e+00 2.36208513e-01 -2.27022335e-01 2.27716994e-02 -1.24836731e+00 1.95900619e-01 2.50106245e-01 7.40698874e-01 -5.93883812e-01 1.06952712e-01 1.75736964e-01 -2.49139920e-01 5.81843436e-01 -4.47150320e-01 -2.48286307e-01 -1.26418620e-02 -1.72444022e+00 4.93729347e-03 5.10936797e-01 1.02395570e+00 8.95192862e-01 1.51706651e-01 -4.48455930e-01 6.67789161e-01 5.28121054e-01 5.29144883e-01 3.23253453e-01 -1.29810095e+00 2.74376035e-01 6.24315321e-01 3.44474353e-02 -9.12261784e-01 -1.39428407e-01 -4.21552271e-01 -9.33689833e-01 -6.29788041e-02 4.97184724e-01 -6.27614791e-03 -1.06312215e+00 1.42020071e+00 5.51588178e-01 7.17473328e-01 2.32368335e-02 8.71846199e-01 9.06403601e-01 1.06512594e+00 1.50716767e-01 -6.81125343e-01 1.42042768e+00 -1.15267253e+00 -1.01701236e+00 1.38874650e-01 1.02165774e-01 -7.30149388e-01 4.99496162e-01 3.90268713e-01 -1.39346111e+00 -7.80630827e-01 -9.44726944e-01 1.88752398e-01 -6.41480312e-02 1.70332164e-01 2.15991288e-01 1.57668099e-01 -9.69396949e-01 6.51601374e-01 -9.52420115e-01 -1.44863456e-01 4.16792542e-01 2.51679480e-01 1.94541942e-02 -4.65512499e-02 -1.03042781e+00 2.18269661e-01 6.65273249e-01 4.11266387e-01 -8.59668791e-01 -5.06467700e-01 -7.29311287e-01 -3.21767703e-02 4.43227619e-01 -6.39624357e-01 1.06070054e+00 -1.26274168e+00 -1.29285789e+00 4.26186919e-01 -5.50118208e-01 -2.65351027e-01 6.51435494e-01 -9.31322798e-02 -2.22732574e-01 7.26399839e-01 3.40563506e-01 9.07073617e-01 1.11879325e+00 -1.44787455e+00 -1.41654909e+00 -1.99905425e-01 -1.86304331e-01 3.51441145e-01 -1.80447876e-01 9.72873271e-02 -9.80524719e-01 -6.68968499e-01 3.95969361e-01 -4.04856473e-01 -2.13863254e-01 -1.76516593e-01 -1.80163011e-01 -2.86611348e-01 1.21015227e+00 -9.38152194e-01 1.48351705e+00 -2.37286901e+00 1.66845709e-01 1.49484277e-01 2.38155216e-01 1.78650960e-01 8.41114745e-02 -3.16575438e-01 6.22771643e-02 2.21406326e-01 -7.31984198e-01 -2.63010174e-01 -4.87596840e-01 2.25077234e-02 2.16395166e-02 2.48938218e-01 2.95378834e-01 7.08360136e-01 -1.00554991e+00 -1.17384195e+00 4.34789449e-01 3.50314409e-01 -2.44516104e-01 3.43665957e-01 -2.74831146e-01 4.49982941e-01 -5.04248202e-01 6.82144761e-01 7.20937729e-01 -2.02198401e-01 -4.29570414e-02 -4.08043027e-01 -3.74332279e-01 -2.51670361e-01 -1.47237265e+00 1.59489572e+00 3.79609205e-02 4.26104963e-01 4.17947531e-01 -7.85016537e-01 6.70951664e-01 4.97159064e-01 8.95364702e-01 -3.24699908e-01 2.11584389e-01 1.99925408e-01 -4.15462375e-01 -5.61546087e-01 5.39746046e-01 -1.84247456e-03 2.10625678e-01 6.63738549e-02 7.56115094e-02 8.96088686e-03 6.20680571e-01 2.40185082e-01 7.18452632e-01 6.31880462e-01 -1.56388268e-01 -9.10818353e-02 9.59938109e-01 6.40240032e-03 9.85163391e-01 5.62440932e-01 -3.88496757e-01 7.80217707e-01 3.66795421e-01 -9.62676853e-02 -9.21751797e-01 -9.46052194e-01 1.23980880e-01 7.63063490e-01 7.27761984e-01 -3.60111557e-02 -1.13212454e+00 -5.25584459e-01 -4.25616622e-01 3.41619343e-01 -2.64219880e-01 3.06614399e-01 -9.36926365e-01 -7.44709253e-01 2.98679203e-01 6.35020554e-01 1.05249751e+00 -1.06338024e+00 -2.76011765e-01 4.48849887e-01 -6.80276871e-01 -1.44890070e+00 -5.93456507e-01 -2.75266826e-01 -1.11683977e+00 -9.72467601e-01 -1.01625431e+00 -1.08243835e+00 5.43543875e-01 4.24236566e-01 6.77420616e-01 1.60745308e-01 -2.76871473e-02 2.44176313e-01 -1.96481749e-01 3.26324433e-01 -3.14041257e-01 -1.01451859e-01 -3.52721423e-01 5.75133979e-01 -9.15185586e-02 -5.16119719e-01 -6.88700378e-01 3.32273602e-01 -1.13627517e+00 3.30703288e-01 5.90452492e-01 3.87422502e-01 9.76593614e-01 3.74201894e-01 3.42636287e-01 -6.08756721e-01 1.12782583e-01 -4.52980906e-01 -6.52663231e-01 2.19397128e-01 -1.88872784e-01 -3.25954080e-01 3.72006804e-01 -4.29485738e-01 -1.41194761e+00 1.97906718e-01 -1.58523127e-01 -6.86663985e-01 -4.94554609e-01 1.98515877e-01 -3.41333896e-01 2.94855952e-01 -1.56490635e-02 4.60958958e-01 -1.68886364e-01 -4.63079900e-01 5.21031439e-01 7.21604645e-01 7.80698299e-01 -4.28963274e-01 6.00561976e-01 6.02364898e-01 -2.70195246e-01 -9.03884768e-01 -6.77278817e-01 -7.13429928e-01 -9.28942800e-01 -7.37435341e-01 1.51760066e+00 -8.77394915e-01 -5.19695222e-01 9.24587965e-01 -1.32127750e+00 -1.72843724e-01 -3.02028507e-01 4.71697062e-01 -4.69144076e-01 6.43261790e-01 -1.07413888e+00 -9.32167590e-01 -2.56485820e-01 -1.25932574e+00 1.15065145e+00 6.36748970e-01 2.11135477e-01 -9.19299126e-01 -4.75200742e-01 4.97709394e-01 2.15395284e-03 2.37982959e-01 5.55070758e-01 -3.42881620e-01 -1.10046208e+00 1.08244844e-01 -4.07199562e-01 4.72134709e-01 1.01500146e-01 3.94820809e-01 -8.50363493e-01 8.57886150e-02 2.47965097e-01 3.95832062e-01 1.05548739e+00 8.37953210e-01 8.67702127e-01 -8.47363546e-02 -3.11014831e-01 5.67569256e-01 1.33756208e+00 5.20826578e-01 5.01285315e-01 9.41483825e-02 9.82251585e-01 6.91345751e-01 8.11508656e-01 2.04943731e-01 2.81810731e-01 2.94161886e-01 1.91743836e-01 -3.88867199e-01 -2.10437179e-01 -5.34283929e-02 5.17820537e-01 1.05876029e+00 -1.27389967e-01 -1.76854387e-01 -7.12509274e-01 7.35079348e-01 -2.06846309e+00 -1.09935844e+00 -4.70240235e-01 1.87518418e+00 7.38498449e-01 3.60691965e-01 4.20104295e-01 2.23753482e-01 1.28863776e+00 3.22989523e-01 -4.72425222e-01 7.22782314e-02 -2.54934579e-01 -3.56305093e-01 4.18887079e-01 5.12864351e-01 -1.25470066e+00 9.98629928e-01 5.80059958e+00 1.23474300e+00 -8.43246400e-01 1.90450817e-01 8.72637391e-01 8.61717910e-02 -1.84109911e-01 -6.12374060e-02 -8.68987083e-01 6.31331503e-01 5.75351894e-01 2.30808988e-01 1.91280186e-01 4.35252309e-01 4.72880661e-01 -4.60985273e-01 -7.68296599e-01 8.25030088e-01 -4.28849235e-02 -1.26009393e+00 1.88215613e-01 -1.48691460e-01 7.36475527e-01 -3.17403138e-01 -2.36004218e-01 -1.38511553e-01 -6.74956217e-02 -5.48111916e-01 1.00701773e+00 6.48359716e-01 4.31323439e-01 -7.30352402e-01 8.20867419e-01 5.66056013e-01 -1.87554860e+00 -4.27583866e-02 1.73401892e-01 3.71900767e-01 6.46958113e-01 8.21563244e-01 -1.71616971e-01 8.07218790e-01 7.14011967e-01 9.54926252e-01 -2.51072913e-01 1.09762847e+00 -1.47319391e-01 7.41042614e-01 -4.12160933e-01 2.78358221e-01 3.69941294e-01 -5.51710248e-01 7.46646643e-01 1.38726413e+00 1.09900333e-01 2.74100065e-01 4.94556278e-01 9.81389046e-01 7.85872936e-02 -3.08041163e-02 -5.67013957e-02 8.48096609e-02 2.65977681e-01 1.48492789e+00 -1.41485417e+00 -9.32834506e-01 -2.65371889e-01 9.56391335e-01 -2.85248995e-01 6.34426832e-01 -8.94861221e-01 -1.85818598e-01 2.27337211e-01 -1.02245897e-01 4.93659765e-01 -2.91971534e-01 -3.82199734e-01 -1.06923473e+00 9.44098830e-02 -5.29874623e-01 4.09190357e-01 -6.53152943e-01 -9.82023716e-01 4.55078185e-01 2.24866018e-01 -1.13522780e+00 -1.01955095e-02 -1.38271675e-01 -5.84559858e-01 6.94915056e-01 -1.44958472e+00 -1.11593354e+00 -3.08267534e-01 6.06342018e-01 1.03909397e+00 2.58777082e-01 -3.38943563e-02 4.50681061e-01 -8.51087272e-01 -3.21255997e-02 -1.33343548e-01 2.43782893e-01 3.46973538e-01 -1.12971365e+00 2.08356511e-02 1.25347495e+00 -1.23803742e-01 1.82029173e-01 3.25840026e-01 -1.01366782e+00 -8.48053336e-01 -1.30960321e+00 4.97003704e-01 -3.20644900e-02 4.06109691e-01 -1.85883138e-04 -1.06648612e+00 2.45076701e-01 2.17178598e-01 -9.77372527e-02 1.36379525e-01 -5.60602784e-01 2.37393513e-01 -2.20011637e-01 -1.06914365e+00 4.19608444e-01 8.29613268e-01 -2.85536915e-01 -5.50681353e-01 1.00143366e-01 9.40452158e-01 -2.86794990e-01 -9.83747542e-01 6.42621696e-01 3.70553911e-01 -8.77329648e-01 8.58261704e-01 7.79538378e-02 5.03137469e-01 -8.12063217e-01 2.13819072e-02 -5.08018911e-01 -4.18008447e-01 -5.98914802e-01 -1.21457592e-01 1.67545795e+00 2.51604736e-01 -1.48868293e-01 6.59767985e-01 3.42474431e-01 -1.27495406e-03 -7.25835681e-01 -7.75897682e-01 -4.75806296e-01 -4.31797177e-01 -4.89308178e-01 2.97709137e-01 6.30499661e-01 -5.93416452e-01 2.33356237e-01 -9.70916897e-02 2.91688710e-01 9.88063276e-01 2.64905840e-01 2.62005419e-01 -8.35581005e-01 -4.72459011e-02 -7.15286136e-01 -1.35912880e-01 -1.40123856e+00 5.47143854e-02 -6.19201183e-01 3.48574877e-01 -1.77691281e+00 3.62627655e-01 -1.21059276e-01 -3.99049222e-01 2.35701889e-01 -4.53373373e-01 3.47838849e-01 2.76286304e-01 4.80337441e-01 -8.84013712e-01 3.18450868e-01 1.47070146e+00 -1.82025746e-01 -4.78375375e-01 1.03288591e-01 -1.87288046e-01 1.00376391e+00 5.47494352e-01 -3.35772008e-01 -2.30601802e-01 -1.56092197e-01 -4.01015371e-01 3.49619061e-01 2.71243811e-01 -9.57009375e-01 4.25920188e-01 -1.90690652e-01 4.82456207e-01 -1.03729808e+00 2.72303522e-01 -7.94777751e-01 1.55260965e-01 2.90508807e-01 -8.01106021e-02 -2.46665776e-01 3.00016534e-02 6.60706520e-01 -4.31997091e-01 -2.13494435e-01 8.41705084e-01 -2.80551195e-01 -8.85563016e-01 4.23939079e-01 -6.16322100e-01 -1.12499364e-01 1.09785640e+00 -6.31164491e-01 6.19586967e-02 -1.20946683e-01 -1.02217591e+00 4.65450376e-01 2.51170248e-01 2.94067353e-01 7.12121725e-01 -1.19309592e+00 -5.73955119e-01 1.27125680e-01 -5.94375253e-01 5.29735386e-01 3.57055843e-01 1.10217559e+00 -5.62859416e-01 -1.18533336e-01 -1.24692600e-02 -1.16743171e+00 -1.35322773e+00 3.70023280e-01 2.60783732e-01 -2.11186081e-01 -6.45452559e-01 7.12138176e-01 3.83497447e-01 3.00138533e-01 2.53858507e-01 -5.47984004e-01 -4.49327886e-01 4.41748768e-01 4.89281356e-01 4.80262130e-01 -4.25310642e-01 -8.19352806e-01 -2.78587341e-01 8.10985029e-01 2.02700898e-01 -4.30106878e-01 1.07030332e+00 -4.39840376e-01 -4.39203352e-01 7.77850151e-01 9.99696851e-01 -3.53507876e-01 -1.46291614e+00 -2.29622051e-01 -1.08805694e-01 -3.01911026e-01 8.63974681e-04 -3.88916939e-01 -1.26570141e+00 8.33530307e-01 5.61056376e-01 3.38946402e-01 1.45872712e+00 -1.64823413e-01 1.16669023e+00 -2.33774468e-01 1.66887552e-01 -1.18815553e+00 -5.75155877e-02 3.81932646e-01 2.81022221e-01 -9.03177083e-01 -7.09658787e-02 -7.69017160e-01 -4.03748184e-01 1.02293026e+00 6.04126513e-01 -7.12087229e-02 5.39880872e-01 3.26900035e-01 -1.81349128e-01 1.95919007e-01 -4.77988780e-01 -3.84212673e-01 3.61758769e-01 2.07191885e-01 1.44454733e-01 -1.72498628e-01 -3.07360053e-01 6.20537102e-01 4.16770756e-01 -1.24194948e-02 1.99106306e-01 8.54402661e-01 -7.28712499e-01 -7.83433378e-01 -6.51533782e-01 2.81607479e-01 -5.18625557e-01 1.41068073e-02 -7.76159689e-02 5.09206295e-01 4.66740429e-01 1.00007248e+00 2.16980129e-01 -3.02848756e-01 6.83201179e-02 1.37027368e-01 3.64455432e-01 -3.30828398e-01 -6.35454774e-01 8.40888321e-01 -8.91830698e-02 -6.47722006e-01 -9.16012824e-01 -6.47153556e-01 -1.59106648e+00 9.59457457e-03 -4.75329250e-01 1.32796183e-01 5.06787837e-01 1.08026671e+00 -3.14058512e-02 9.22494829e-01 5.50841689e-01 -1.09759891e+00 2.69528657e-01 -8.02846491e-01 -4.98757124e-01 3.93151760e-01 2.72146583e-01 -2.55931318e-01 -2.46236727e-01 6.55833483e-01]

[9.165498733520508, -0.3150150775909424]

e023b9e4-7c34-4404-b73b-dfca73b556d5

semi-supervised-video-paragraph-grounding

http://openaccess.thecvf.com//content/CVPR2022/html/Jiang_Semi-Supervised_Video_Paragraph_Grounding_With_Contrastive_Encoder_CVPR_2022_paper.html

http://openaccess.thecvf.com//content/CVPR2022/papers/Jiang_Semi-Supervised_Video_Paragraph_Grounding_With_Contrastive_Encoder_CVPR_2022_paper.pdf

Semi-Supervised Video Paragraph Grounding With Contrastive Encoder

Video events grounding aims at retrieving the most relevant moments from an untrimmed video in terms of a given natural language query. Most previous works focus on Video Sentence Grounding (VSG), which localizes the moment with a sentence query. Recently, researchers extended this task to Video Paragraph Grounding (VPG) by retrieving multiple events with a paragraph. However, we find the existing VPG methods may not perform well on context modeling and highly rely on video-paragraph annotations. To tackle this problem, we propose a novel VPG method termed Semi-supervised Video-Paragraph TRansformer (SVPTR), which can more effectively exploit contextual information in paragraphs and significantly reduce the dependency on annotated data. Our SVPTR method consists of two key components: (1) a base model VPTR that learns the video-paragraph alignment with contrastive encoders and tackles the lack of sentence-level contextual interactions and (2) a semi-supervised learning framework with multimodal feature perturbations that reduces the requirements of annotated training data. We evaluate our model on three widely-used video grounding datasets, i.e., ActivityNet-Caption, Charades-CD-OOD, and TACoS. The experimental results show that our SVPTR method establishes the new state-of-the-art performance on all datasets. Even under the conditions of fewer annotations, it can also achieve competitive results compared with recent VPG methods.

['Heng Tao Shen', 'Zuo Cao', 'Fumin Shen', 'Jingran Zhang', 'Xing Xu', 'Xun Jiang']

2022-01-01

cvpr-2022-1

['video-grounding']

['computer-vision']

[ 2.51551539e-01 -9.19912606e-02 -4.80964780e-01 -2.87553757e-01 -1.15672696e+00 -3.67088526e-01 7.29252934e-01 2.15618461e-01 -3.38482618e-01 5.30631363e-01 5.86438656e-01 8.14069211e-02 2.44065851e-01 -4.77212578e-01 -1.05450249e+00 -4.77118134e-01 1.02032468e-01 1.20759912e-01 5.87069273e-01 -1.75247923e-01 2.31113568e-01 -1.88078314e-01 -1.53427422e+00 7.50550270e-01 6.26412511e-01 9.71322358e-01 5.71460485e-01 4.33526903e-01 -3.47415805e-01 1.31487083e+00 -4.64179099e-01 -2.93890685e-01 -1.57365680e-01 -6.72691345e-01 -8.90651107e-01 1.05605453e-01 4.32435453e-01 -3.21822435e-01 -6.31066203e-01 7.80514717e-01 4.50657457e-01 1.94085479e-01 2.67419040e-01 -1.44611251e+00 -4.56324011e-01 6.78876042e-01 -4.22375709e-01 2.52255619e-01 8.83869112e-01 -1.43291578e-01 1.18686557e+00 -1.00249159e+00 9.14362848e-01 1.03969038e+00 5.00595272e-01 4.61841017e-01 -8.32484841e-01 -3.88006449e-01 4.47285593e-01 5.39085150e-01 -1.48545623e+00 -4.10618603e-01 8.26795816e-01 -3.41369063e-01 1.03124642e+00 3.26368451e-01 6.55754328e-01 1.51941550e+00 -8.02878290e-02 1.06178498e+00 7.03675866e-01 -3.01782310e-01 1.07317060e-01 -1.34669870e-01 -8.07206407e-02 6.43654108e-01 -1.68854743e-01 -4.57255781e-01 -1.18288016e+00 -5.44636026e-02 5.40841103e-01 9.17363763e-02 -5.16855538e-01 -1.67732999e-01 -1.74324167e+00 5.82666874e-01 2.78545599e-02 5.03679335e-01 -3.39145035e-01 2.02477481e-02 6.28138363e-01 9.58605334e-02 5.06765902e-01 2.09768906e-01 -1.01811878e-01 -3.42562586e-01 -1.08809853e+00 2.99760103e-01 6.90203726e-01 1.34947538e+00 7.67195165e-01 -3.00656140e-01 -6.26497269e-01 6.51332974e-01 3.04654628e-01 5.24263561e-01 4.79893774e-01 -8.16077054e-01 1.02963448e+00 6.08750045e-01 1.39593020e-01 -1.20447481e+00 -3.07380687e-02 6.88090846e-02 -4.90514725e-01 -7.66880989e-01 1.00172805e-02 1.08018778e-01 -6.80495560e-01 1.74804854e+00 2.31975973e-01 6.57595694e-01 1.02943212e-01 9.44083869e-01 1.32218373e+00 1.05143619e+00 1.17823794e-01 -4.86596823e-01 1.30777442e+00 -1.21305871e+00 -9.35360014e-01 -2.50973701e-01 5.83693326e-01 -6.34368300e-01 1.19475818e+00 1.19899139e-01 -8.70945632e-01 -5.42808294e-01 -1.03709102e+00 -2.10280627e-01 -2.56075680e-01 1.18710324e-02 2.34953210e-01 -8.05929229e-02 -8.56539607e-01 2.55931795e-01 -8.31546009e-01 -6.09124839e-01 9.50362235e-02 2.63781548e-02 -4.99463379e-01 -2.06267968e-01 -1.31771958e+00 4.67371166e-01 5.78943431e-01 1.15719989e-01 -1.03469169e+00 -4.54269260e-01 -1.13636112e+00 -5.50875440e-03 7.55199909e-01 -5.78363717e-01 1.13061237e+00 -9.68660772e-01 -1.31632483e+00 8.69149148e-01 -5.93314707e-01 -4.40447032e-01 3.85529518e-01 -4.69453782e-01 -4.57447171e-01 7.80541956e-01 3.83339375e-01 7.29923010e-01 8.40814352e-01 -1.03011620e+00 -5.42805731e-01 -9.80341211e-02 3.19597781e-01 4.56343651e-01 -3.92166138e-01 1.07243069e-01 -1.00450635e+00 -7.11530268e-01 -2.22985744e-02 -8.38599205e-01 7.99801722e-02 -3.59710336e-01 -4.45375472e-01 -3.09178472e-01 9.84363616e-01 -7.87651718e-01 1.55577600e+00 -2.17864633e+00 4.72897559e-01 -2.15235040e-01 6.72023296e-02 2.01494500e-01 -1.68387696e-01 9.71547067e-01 3.36296670e-02 -6.56596348e-02 -4.04194072e-02 -5.11258364e-01 -6.46158680e-02 3.63248318e-01 -5.50088644e-01 2.16571286e-01 1.83736801e-01 9.97397006e-01 -1.17710495e+00 -1.04053843e+00 2.21818537e-01 4.57012326e-01 -5.14924228e-01 4.88063693e-01 -4.56589490e-01 5.23165762e-01 -4.92628813e-01 6.59697354e-01 1.76708162e-01 -2.91810870e-01 1.86465830e-01 -4.47578698e-01 -1.14429213e-01 3.43665928e-02 -9.94407117e-01 2.26187730e+00 -2.16951177e-01 7.07508445e-01 -3.30532402e-01 -1.02908623e+00 5.27015984e-01 6.41372919e-01 7.08595455e-01 -6.89328313e-01 -1.66012362e-01 4.37428355e-02 -7.69373894e-01 -1.03143227e+00 6.54596686e-01 2.52406299e-01 -2.25769356e-01 2.53362488e-02 3.80692720e-01 3.49314362e-01 3.29637200e-01 6.17099583e-01 1.08486831e+00 5.51414073e-01 2.76157320e-01 4.71594557e-02 6.49928749e-01 -3.77618447e-02 6.04570568e-01 6.45710647e-01 -1.25456780e-01 6.35186553e-01 5.99330425e-01 -1.94118991e-01 -7.05862164e-01 -7.74205863e-01 3.16907853e-01 1.12587059e+00 5.64331710e-01 -1.05527318e+00 -8.93571436e-01 -7.10700512e-01 -4.21073794e-01 5.71500897e-01 -4.98905122e-01 8.52515325e-02 -6.00612104e-01 -2.83751756e-01 4.35685366e-01 5.75139403e-01 6.98555112e-01 -9.51316178e-01 -3.31201077e-01 7.94550776e-03 -1.11180151e+00 -1.82017899e+00 -7.16945231e-01 -2.84931123e-01 -5.86371481e-01 -1.13022733e+00 -6.99764848e-01 -9.41037118e-01 5.99750280e-01 6.32672846e-01 1.14066446e+00 8.27970579e-02 2.47447625e-01 7.26294100e-01 -9.51039553e-01 -1.43530339e-01 -1.53754875e-01 8.88955072e-02 -1.29628718e-01 4.28765208e-01 5.15683413e-01 -3.43247503e-01 -6.45558000e-01 5.44031978e-01 -9.73842859e-01 4.44084167e-01 5.39404392e-01 7.07948446e-01 9.46666360e-01 -2.76256859e-01 4.81472164e-01 -6.40119493e-01 1.69130221e-01 -6.19249344e-01 -2.55203396e-01 4.12356377e-01 -1.99542657e-01 -1.12185672e-01 5.04621625e-01 -4.00174588e-01 -9.83120859e-01 6.44441182e-03 4.02981490e-02 -8.00697565e-01 -3.85886170e-02 6.73056245e-01 -3.25997829e-01 1.80742919e-01 2.70218372e-01 4.75721896e-01 -3.31722409e-01 -3.24938208e-01 2.65827179e-01 7.04686701e-01 6.81705534e-01 -5.93152881e-01 7.21356452e-01 4.95578647e-01 -7.67497122e-02 -9.35483038e-01 -1.06146371e+00 -8.07545900e-01 -5.16237915e-01 -5.39737225e-01 1.25898051e+00 -1.21065962e+00 -4.79917854e-01 1.74506694e-01 -1.14856935e+00 -1.55124694e-01 -6.18323963e-03 5.57612896e-01 -7.62781143e-01 8.06003690e-01 -5.32254994e-01 -4.65484977e-01 -2.73342460e-01 -1.04584062e+00 1.42909861e+00 -4.49424013e-02 -1.53703643e-02 -6.72005296e-01 2.21776348e-02 5.92502654e-01 -8.29389915e-02 4.56455350e-01 4.68698174e-01 -6.65076256e-01 -8.04330409e-01 -1.35954589e-01 -1.04690135e-01 1.67910859e-01 -3.55982371e-02 -1.31147712e-01 -7.93297350e-01 -2.52732068e-01 -2.13707000e-01 -3.50327849e-01 8.11758816e-01 1.18640132e-01 1.05720973e+00 -4.46241558e-01 -3.38150978e-01 5.55116832e-01 1.37983263e+00 4.25079055e-02 7.31001258e-01 3.86266768e-01 9.69701588e-01 4.41930681e-01 1.07610297e+00 3.65087330e-01 5.98868668e-01 9.41036701e-01 5.80746591e-01 2.16982305e-01 2.05036253e-03 -8.05948317e-01 5.96261322e-01 1.12150431e+00 -3.60528618e-01 -4.19661820e-01 -7.32593954e-01 6.13445878e-01 -2.49935627e+00 -1.35779691e+00 7.21506774e-02 1.98092353e+00 5.78255415e-01 2.81357970e-02 1.05105795e-01 -4.58013527e-02 7.05406308e-01 5.22711933e-01 -2.18527392e-01 1.24054439e-01 -9.90440845e-02 -3.87906551e-01 1.23760417e-01 2.18874097e-01 -1.26183391e+00 9.44072127e-01 5.20974398e+00 7.92557776e-01 -7.81967878e-01 3.13188881e-01 1.62029848e-01 -1.92400903e-01 -1.36885375e-01 -7.10337535e-02 -8.52609217e-01 7.80354559e-01 8.24588537e-01 8.47692974e-03 3.63283008e-01 6.76705241e-01 4.32600915e-01 -4.80917431e-02 -1.34582388e+00 1.32652450e+00 6.18986905e-01 -1.32577455e+00 3.09413254e-01 -2.96750575e-01 5.84537745e-01 -4.36985269e-02 -3.70132238e-01 4.42706168e-01 -5.19189179e-01 -6.57117963e-01 1.02939141e+00 4.98723686e-01 7.31003165e-01 -5.52655995e-01 8.23056757e-01 3.12461644e-01 -1.56399536e+00 9.58725251e-03 -1.93211287e-01 1.18118495e-01 3.52761090e-01 1.89773515e-01 -5.67152321e-01 1.02662528e+00 7.25079477e-01 1.12714052e+00 -5.40305793e-01 8.46328974e-01 -3.78908902e-01 7.09932029e-01 -9.79624167e-02 -2.60822251e-02 4.18473721e-01 2.80500632e-02 6.77191556e-01 1.31088042e+00 3.40024412e-01 8.40605348e-02 4.23457831e-01 4.12843585e-01 -6.46901578e-02 2.97686070e-01 -7.28872299e-01 -1.58318251e-01 5.30603051e-01 1.01014292e+00 -7.79762506e-01 -4.25343752e-01 -8.36568415e-01 1.10148847e+00 1.69743255e-01 4.82076466e-01 -1.16997397e+00 -3.36480558e-01 2.45831788e-01 1.44436330e-01 4.49824393e-01 -2.09879160e-01 4.89490837e-01 -1.63468552e+00 3.85911673e-01 -8.20096612e-01 5.14684081e-01 -1.12484813e+00 -8.37593496e-01 6.01175010e-01 3.52714181e-01 -1.64252019e+00 -2.99728811e-01 -1.52756676e-01 -3.98488164e-01 1.17289945e-01 -1.55445027e+00 -1.54041791e+00 -6.36217952e-01 8.43815207e-01 9.38249648e-01 3.75253111e-02 6.02603674e-01 5.27373850e-01 -6.28061652e-01 3.22752267e-01 -2.70612955e-01 2.85537660e-01 8.88165176e-01 -1.06300545e+00 -1.17866974e-02 1.04034770e+00 4.54279244e-01 4.21082705e-01 6.68028116e-01 -6.01857424e-01 -1.62929749e+00 -1.29986680e+00 1.11064363e+00 -5.70148706e-01 5.50708711e-01 -4.24599856e-01 -7.83897281e-01 1.02314365e+00 4.65031564e-01 -7.80465007e-02 6.94210827e-01 -2.29275301e-01 -2.82322526e-01 -1.11355036e-01 -5.83640754e-01 6.11221075e-01 1.33844459e+00 -8.48428845e-01 -8.67130101e-01 5.76618075e-01 1.14663565e+00 -6.61111057e-01 -7.21184433e-01 2.52065897e-01 3.11254889e-01 -8.29144359e-01 9.78009999e-01 -5.60947835e-01 6.98323190e-01 -4.37677413e-01 -4.96186554e-01 -8.90070438e-01 5.87898158e-02 -9.00458753e-01 -3.82487744e-01 1.56208551e+00 3.94560844e-02 -8.95920992e-02 4.87979531e-01 2.99992085e-01 -3.93084735e-01 -7.32716501e-01 -9.16576564e-01 -7.21780777e-01 -7.09250271e-01 -5.83517432e-01 5.16884565e-01 1.07612586e+00 2.50905275e-01 4.39627290e-01 -6.21686101e-01 2.58987725e-01 4.87764895e-01 1.88479051e-01 7.34575272e-01 -7.44967043e-01 -1.98680043e-01 5.79521246e-02 -5.79826653e-01 -1.20827186e+00 3.45598489e-01 -7.68359184e-01 2.10443199e-01 -1.66080356e+00 5.50920904e-01 4.61557567e-01 -3.37133169e-01 5.54930210e-01 -2.74775296e-01 1.77583769e-01 3.07144791e-01 3.31884295e-01 -1.40276361e+00 4.82472032e-01 9.90859270e-01 -2.19173804e-01 -1.32018834e-01 -3.29335690e-01 -3.59368116e-01 7.70089626e-01 3.33835483e-01 -2.98196048e-01 -6.76454961e-01 -6.07044458e-01 3.49064499e-01 3.70755434e-01 5.44861495e-01 -9.77019846e-01 3.61352503e-01 -1.76380053e-01 3.48536856e-03 -9.49137866e-01 3.15789521e-01 -6.93723619e-01 9.84253138e-02 -5.13670333e-02 -3.78475726e-01 1.70501918e-01 2.13144831e-02 9.13865924e-01 -5.84714055e-01 -1.08081140e-01 2.03999922e-01 -1.67766511e-01 -1.22294867e+00 3.01763654e-01 -3.19774449e-01 2.37729296e-01 1.11017978e+00 -1.31417423e-01 -3.69878680e-01 -4.84614342e-01 -5.84105968e-01 4.33357507e-01 3.73975843e-01 6.18609071e-01 7.56895900e-01 -1.59439754e+00 -5.33479333e-01 -2.29006410e-01 4.89922822e-01 4.64835577e-02 3.28830659e-01 9.27225828e-01 -3.92927617e-01 5.91829360e-01 1.00334674e-01 -8.56349587e-01 -1.32824624e+00 7.45937645e-01 -3.98461521e-02 -1.57945782e-01 -7.39961326e-01 6.25738263e-01 2.39486694e-01 1.73233151e-01 3.68311673e-01 -3.99823964e-01 -3.26162636e-01 2.92738557e-01 6.72312260e-01 8.67673606e-02 -5.66453151e-02 -1.02807057e+00 -4.19657737e-01 6.71949029e-01 7.79471397e-02 -3.98788005e-02 1.14710736e+00 -3.62630665e-01 2.91359462e-02 5.33436537e-01 1.24353659e+00 -1.59895197e-01 -1.10806715e+00 -3.49516362e-01 -7.77154602e-03 -4.65242893e-01 -1.01073213e-01 -3.61792386e-01 -7.91560352e-01 6.93455458e-01 1.55071288e-01 1.02889352e-01 1.19226182e+00 2.10921094e-01 1.11799836e+00 5.87063313e-01 5.71104705e-01 -1.00051832e+00 4.40378636e-01 2.72348404e-01 8.40890229e-01 -1.18967426e+00 -1.66236460e-02 -4.11764681e-01 -7.69428551e-01 9.24852371e-01 6.71579361e-01 2.19590276e-01 2.86083102e-01 -1.34196430e-01 -1.67573616e-01 -1.48701832e-01 -7.84696102e-01 -3.32544506e-01 3.10677797e-01 3.96146953e-01 1.68429032e-01 -3.08972478e-01 -3.02183896e-01 8.17591965e-01 2.62021363e-01 2.42777050e-01 3.03129286e-01 1.04051375e+00 -3.32828254e-01 -8.57989192e-01 -1.17938705e-01 4.40986753e-02 -4.56664979e-01 -1.23298075e-02 -3.51148099e-01 7.86143482e-01 4.09359671e-02 9.20713007e-01 -1.02542788e-02 -5.32406211e-01 3.85238260e-01 1.02911234e-01 3.10422450e-01 -6.84871852e-01 -3.34502488e-01 1.74879149e-01 2.08914116e-01 -9.91555214e-01 -1.09953940e+00 -6.43191814e-01 -1.08783591e+00 6.83469977e-03 -1.59966975e-01 3.77093226e-01 2.88076103e-01 1.15320873e+00 5.00240266e-01 4.31161165e-01 5.21915615e-01 -9.75324631e-01 -9.00477543e-02 -8.70515049e-01 -2.77775645e-01 6.39582336e-01 2.45098799e-01 -6.21840239e-01 -2.24920407e-01 5.16508520e-01]

[10.188530921936035, 0.7441869378089905]

f4477abf-4ee8-4aa9-babc-c5abe058b395

uir-pku-twitter-opinminer-system-for

https://aclanthology.org/S15-2111

https://aclanthology.org/S15-2111.pdf

UIR-PKU: Twitter-OpinMiner System for Sentiment Analysis in Twitter at SemEval 2015

['Kam-Fai Wong', 'Gaoyan Ou', 'Jing Ma', 'Yuxiao Zhang', 'Xu Han', 'Tengjiao Wang', 'Binyang Li']

2015-06-01

semeval-2015-6

['twitter-sentiment-analysis']

['natural-language-processing']

[-8.63703638e-02 1.71006292e-01 -6.22772932e-01 -4.08054382e-01 -8.41685571e-03 -9.08429027e-01 6.55310392e-01 -6.53472245e-01 -2.85945535e-01 1.06888819e+00 -4.63127941e-02 -1.01159286e+00 -3.91567826e-01 -9.63214397e-01 -4.95059669e-01 -6.31337762e-01 -9.79754329e-01 7.25764990e-01 3.30370307e-01 -6.93831444e-01 7.03166842e-01 7.88774848e-01 -1.68942046e+00 7.18545914e-01 7.04417467e-01 8.52217197e-01 2.49141872e-01 1.14950800e+00 -1.95044339e-01 1.55633950e+00 -7.48382092e-01 -5.46825826e-01 3.13719302e-01 -1.23176083e-01 -7.22945035e-01 -1.01074085e-01 9.28529128e-02 -8.59008506e-02 -2.09758401e-01 9.22211111e-01 5.37373662e-01 4.49454933e-02 1.08379531e+00 -1.42548037e+00 -5.91619551e-01 6.10313773e-01 -4.01565880e-02 1.21627934e-01 1.03678203e+00 -5.39447069e-01 1.19919395e+00 -1.13026452e+00 7.20913768e-01 1.26888943e+00 8.66221786e-01 5.44149756e-01 -1.22286928e+00 -1.94712028e-01 -3.26822817e-01 -9.51717794e-02 -1.46558487e+00 -3.25250506e-01 4.25783843e-02 -2.08119690e-01 1.66093647e+00 1.26596653e+00 1.20609856e+00 1.01401424e+00 1.26658809e+00 8.34431887e-01 1.04267764e+00 -5.13792276e-01 3.35295945e-01 3.66983831e-01 1.54683650e-01 6.33519173e-01 8.40953708e-01 5.26628852e-01 -7.06372619e-01 -9.13127720e-01 9.33553874e-01 -2.94925272e-01 1.71355158e-01 -5.05680561e-01 -9.05919552e-01 6.91228509e-01 1.78732842e-01 3.83959889e-01 -1.39880210e-01 9.89067405e-02 1.26390755e-01 5.30987144e-01 -2.58292928e-02 6.47037446e-01 -9.11868811e-01 -1.33165747e-01 -8.71728659e-01 5.10332465e-01 1.25398111e+00 1.52653182e+00 1.24482810e-01 2.94908643e-01 -9.34252143e-02 3.17179203e-01 8.92314315e-01 1.01808000e+00 4.28362608e-01 -1.36146402e+00 -6.87414408e-02 1.72361732e-01 5.01781464e-01 -8.52631688e-01 -6.33224547e-01 -9.64177120e-03 -8.93263519e-01 4.49267089e-01 3.49161088e-01 4.57367361e-01 -8.02827001e-01 5.07305264e-01 4.33481112e-02 -2.34125629e-01 4.53833073e-01 5.55570945e-02 4.99930978e-01 3.76208365e-01 -1.34477139e-01 -5.73289394e-01 1.06082785e+00 -1.36716676e+00 -1.35299087e+00 2.33215362e-01 9.05734658e-01 -1.07320261e+00 4.35900748e-01 5.33875942e-01 -1.55548143e+00 -1.37560293e-01 -1.08699942e+00 1.78573877e-01 -7.27255583e-01 -3.14239264e-01 8.57801437e-01 1.43120694e+00 -1.60129595e+00 9.73287821e-01 -4.91727620e-01 6.59165755e-02 1.20568443e-02 8.24621081e-01 -2.64718989e-03 4.62812334e-01 -1.33193445e+00 1.08501506e+00 2.22979754e-01 -1.21242590e-01 -1.65216476e-01 -2.13068098e-01 -8.23704481e-01 -5.63443303e-01 -4.78693932e-01 -5.29636025e-01 1.44139910e+00 -2.59346128e-01 -1.65295815e+00 9.71794367e-01 -1.42069459e-01 -1.97814897e-01 6.14786744e-01 -1.28011424e-02 -8.31891418e-01 2.42498964e-01 -1.89849049e-01 5.76383233e-01 9.28263724e-01 -1.35132408e+00 -7.59897232e-01 -1.67359829e-01 -1.23336017e-01 2.66287565e-01 -1.25510961e-01 1.89734384e-01 2.11616129e-01 -1.12999000e-01 3.27147305e-01 -7.20919967e-01 -2.53068686e-01 -5.32041907e-01 -1.46512717e-01 -7.10518599e-01 7.70373225e-01 -4.51523662e-01 1.83705616e+00 -1.67618537e+00 -2.06720144e-01 3.98590982e-01 3.57815564e-01 -1.24705513e-03 2.40583986e-01 1.08380008e+00 -2.76906848e-01 7.32199550e-01 4.11965609e-01 -1.10722095e-01 2.24991128e-01 4.85861301e-01 -4.16602850e-01 3.05609167e-01 -1.29282743e-01 1.15307164e+00 -1.16605783e+00 -5.23096442e-01 4.11106765e-01 9.42391157e-02 -4.58732933e-01 4.79237735e-01 2.99364805e-01 1.47170946e-01 -3.56553018e-01 1.39399457e+00 1.15709066e+00 -1.31984919e-01 1.45911396e-01 5.30878425e-01 -3.79135728e-01 3.55090618e-01 -6.96863770e-01 1.05554795e+00 7.08333924e-02 5.00173986e-01 1.02364108e-01 -7.94621468e-01 3.33247900e-01 8.61540735e-01 4.37155962e-01 -9.67555881e-01 -2.26398129e-02 6.92409754e-01 1.12803578e-01 -6.07703328e-01 7.58228302e-01 3.94563079e-02 -3.64872098e-01 6.40070081e-01 -2.37588286e-01 -6.59476995e-01 9.64643434e-02 3.08100313e-01 6.22585893e-01 -5.18246442e-02 5.62923312e-01 -1.05613089e+00 7.86340594e-01 -1.82965681e-01 -1.81299388e-01 1.03415680e+00 -3.09923887e-01 3.19085121e-01 2.99841821e-01 -6.65102363e-01 -6.45341039e-01 -1.12307119e+00 -4.89381433e-01 1.30636716e+00 3.24267983e-01 -4.39044595e-01 -9.54439282e-01 -2.49762803e-01 1.77620783e-01 6.89606130e-01 -5.90509653e-01 3.84124845e-01 -5.03739953e-01 -8.32535863e-01 7.39044368e-01 3.45434904e-01 -5.07752821e-02 -1.33414865e+00 -6.58416986e-01 1.25490099e-01 -2.20292807e-01 -6.63697243e-01 -6.23428151e-02 4.48765576e-01 -1.35989368e+00 -5.18594682e-01 -6.66252747e-02 -8.20914626e-01 5.87345481e-01 2.46782884e-01 1.27047324e+00 5.39230824e-01 -2.31483161e-01 4.26904231e-01 -1.21292919e-01 -4.95818377e-01 -4.59671497e-01 -8.00336525e-02 5.28869390e-01 -5.87835789e-01 5.19427478e-01 -2.50617653e-01 -7.29350567e-01 5.37953973e-01 -6.88540697e-01 1.62748516e-01 1.79803044e-01 1.04410267e+00 1.35816500e-01 -9.34035778e-02 1.22507080e-01 -6.38007045e-01 8.72274399e-01 -1.69219792e-01 -3.78732830e-01 5.77745810e-02 -6.77108407e-01 -3.74140263e-01 3.21430594e-01 -3.25342178e-01 -1.01981449e+00 -4.87835288e-01 -9.82677937e-02 2.45538145e-01 1.11353043e-02 -1.46784872e-01 6.47139177e-02 -5.24923325e-01 8.02199244e-01 9.25758183e-02 1.99174434e-02 -6.80815242e-03 3.01039815e-01 7.09525108e-01 -6.82967342e-03 -6.68678164e-01 8.44880998e-01 4.91470337e-01 7.98524171e-02 -9.57177758e-01 -1.52186140e-01 -2.60129690e-01 -9.51962709e-01 -6.54426932e-01 6.56643391e-01 -6.78531289e-01 -9.10833478e-01 3.91110867e-01 -9.38691139e-01 -3.38627815e-01 -3.91645581e-01 4.25431967e-01 -1.01278400e+00 2.75717527e-02 -3.90154392e-01 -1.27895141e+00 -5.10977268e-01 -1.02017939e+00 9.43384409e-01 5.30070923e-02 -5.10597289e-01 -1.26927447e+00 5.87685481e-02 2.71537274e-01 1.81734428e-01 -1.73075795e-01 6.90226793e-01 -2.38256708e-01 -4.24233019e-01 -1.53791070e-01 2.34436691e-02 -1.39755070e-01 1.70832314e-02 4.95917559e-01 -9.81751978e-01 -5.31145096e-01 6.65065646e-02 -1.92070693e-01 -1.08835101e-01 6.52520418e-01 5.91872573e-01 -2.29931593e-01 -8.56000841e-01 5.40386558e-01 1.38545322e+00 3.85070026e-01 5.32770038e-01 7.28214979e-01 1.41836226e-01 5.53460240e-01 9.17806149e-01 4.63203549e-01 1.30579369e-02 3.28798652e-01 2.40537539e-01 1.49327129e-01 1.11720070e-01 -1.54819340e-01 3.77893507e-01 1.16112018e+00 -8.18235934e-01 -2.69281328e-01 -5.07867396e-01 4.42987174e-01 -1.72482407e+00 -1.40330648e+00 -4.32368398e-01 6.90478683e-01 6.25676990e-01 1.56016424e-01 -1.48347050e-01 3.35214496e-01 4.99015123e-01 -2.03574806e-01 -1.19133167e-01 -1.06291151e+00 -1.43546045e-01 3.15233678e-01 7.37729073e-01 1.00061214e+00 -7.20721722e-01 1.03317809e+00 1.29781246e+01 1.02230716e+00 2.21112028e-01 1.03134915e-01 5.16071796e-01 3.48020852e-01 -4.36954498e-01 -4.56139445e-02 -1.04416132e+00 2.72933897e-02 1.38140702e+00 -4.30666685e-01 6.85999811e-01 5.44219851e-01 3.44648361e-01 -4.23268199e-01 -1.26188684e+00 5.26221812e-01 9.73738134e-02 -1.40886843e+00 -2.83300440e-04 6.85225725e-01 7.73699820e-01 -5.08050561e-01 6.22419357e-01 3.24184299e-01 6.09259963e-01 -1.14389277e+00 8.60300779e-01 2.53660440e-01 1.03040910e+00 -6.05088234e-01 5.67372203e-01 1.68872893e-01 -1.14389896e+00 -2.20873043e-01 -8.77727985e-01 -1.00755692e+00 3.93533185e-02 -1.81779593e-01 -4.29956943e-01 3.48861217e-01 9.58353162e-01 2.99398601e-01 -3.93658698e-01 9.95779395e-01 -4.78694476e-02 1.04875881e-02 -2.71853864e-01 -4.48467314e-01 4.83122796e-01 -3.54241252e-01 4.66730654e-01 1.00164843e+00 2.48499006e-01 3.51035744e-01 -9.84472036e-02 4.01770771e-01 5.45058846e-01 3.29446048e-02 -1.19659424e+00 -1.78908288e-01 2.83276141e-01 9.16795909e-01 -4.83487815e-01 -4.22520459e-01 -2.00212970e-01 8.62069130e-01 -3.55488248e-02 5.01107454e-01 -6.11489356e-01 -4.35615242e-01 9.72222984e-01 -1.27327025e-01 -1.14700586e-01 -3.48497719e-01 -6.23769283e-01 -7.30352640e-01 -5.89872956e-01 -4.54965204e-01 5.93606755e-02 -5.53365827e-01 -1.39813089e+00 5.79277515e-01 -2.27688253e-02 -1.40553558e+00 -6.99901402e-01 -1.27676582e+00 -4.76714373e-01 4.92853165e-01 -1.11898029e+00 -1.10984349e+00 2.50124663e-01 4.52870727e-01 1.64141744e-01 -5.34416080e-01 1.39563632e+00 3.57715860e-02 1.00637585e-01 9.24474537e-01 6.69434488e-01 -7.37814724e-01 5.56605101e-01 -1.27867436e+00 5.68737745e-01 -1.37897313e-01 -4.31265175e-01 9.05828118e-01 6.28349900e-01 -5.39804697e-01 -1.41196322e+00 -3.66917729e-01 1.08350635e+00 -9.83769417e-01 6.55218959e-01 -3.86345625e-01 4.23767231e-02 7.88592756e-01 7.15902448e-01 -6.18741751e-01 8.21781039e-01 -1.83753878e-01 1.80774391e-01 5.75296998e-01 -1.39248300e+00 6.12354755e-01 1.66275799e+00 -4.63594139e-01 -6.25784039e-01 7.60327101e-01 8.13696027e-01 -6.94087505e-01 -1.30082703e+00 3.34633321e-01 8.65424156e-01 -8.75409484e-01 1.61978090e+00 -1.32660246e+00 -4.43697497e-02 2.81152606e-01 -2.61993498e-01 -9.32519078e-01 -5.96193194e-01 -1.23518765e+00 -5.33532679e-01 -5.83747849e-02 5.96577883e-01 -1.13057327e+00 3.42365682e-01 8.74560475e-01 -2.82833427e-01 -6.42737269e-01 -1.06996536e+00 -1.32016802e+00 -3.35779637e-02 -1.45572275e-01 4.90409225e-01 7.63798356e-01 6.83744550e-01 1.09839931e-01 -6.36873543e-02 -1.00294888e-01 5.33176839e-01 9.55312885e-03 4.41501856e-01 -1.34294486e+00 3.86843324e-01 -5.75816095e-01 -3.07655483e-01 -9.45992947e-01 -8.85957032e-02 -8.12076271e-01 -6.53862000e-01 -1.28511906e+00 -8.32044985e-03 -1.91056758e-01 -1.11109078e-01 -1.65725678e-01 3.67937148e-01 2.13746816e-01 1.20859891e-02 1.03788137e-01 -3.71160030e-01 6.18435517e-02 1.29639816e+00 6.91750320e-05 -1.62315920e-01 4.85058486e-01 -4.81304944e-01 7.84440815e-01 8.58408585e-02 -2.99253196e-01 -6.78878546e-01 6.11881316e-02 6.69384480e-01 4.61409837e-02 3.24159935e-02 -7.42885649e-01 5.37211418e-01 -3.75702560e-01 4.78586555e-01 -1.32223868e+00 1.30741090e-01 -9.61415648e-01 6.95283338e-02 9.40189242e-01 2.74610907e-01 1.20424610e-02 8.66204947e-02 5.39500564e-02 -1.44871444e-01 -5.70943117e-01 9.21121240e-01 -4.07591403e-01 -4.92852688e-01 -5.20386267e-03 -1.03226590e+00 8.97834301e-02 9.92593169e-01 -7.84614205e-01 -3.59281451e-01 -4.20183957e-01 -8.29068601e-01 -1.95836127e-02 6.50830388e-01 3.03609259e-02 7.30431557e-01 -1.51530886e+00 -2.30721906e-01 7.18729138e-01 -3.22939813e-01 -3.74400020e-01 -1.70157343e-01 6.58265352e-01 -1.32361674e+00 1.02442718e+00 -5.41665435e-01 -4.55340147e-01 -1.14228773e+00 4.78126436e-01 4.28307921e-01 -2.41845414e-01 -2.15481281e-01 1.11954463e+00 2.71224789e-02 -8.23025763e-01 1.85185194e-01 -9.21545625e-02 -7.54407048e-01 3.55081353e-03 6.88606799e-01 1.05194807e+00 -2.91290224e-01 -6.04341030e-01 -4.56784427e-01 6.65885091e-01 2.32151806e-01 -2.87484169e-01 9.17833567e-01 -2.19243199e-01 -9.89108324e-01 4.28274393e-01 8.38715494e-01 -1.36269778e-01 -3.69319022e-02 4.16855574e-01 1.25943512e-01 -8.02164078e-01 -4.32406247e-01 -3.55811834e-01 -1.85641110e-01 5.54822803e-01 5.34874737e-01 8.99602413e-01 8.57008278e-01 -3.02566767e-01 8.18335712e-01 9.66778398e-01 5.72402716e-01 -1.68019545e+00 -2.13140488e-01 6.89524531e-01 9.12339568e-01 -9.28350806e-01 5.44190466e-01 -7.27165341e-01 -4.14997995e-01 1.32979155e+00 4.68304873e-01 -1.55325383e-01 1.27306652e+00 5.74917436e-01 1.14069022e-02 -3.70670199e-01 -9.44949508e-01 1.12705544e-01 3.75366658e-01 1.11147714e+00 5.06513238e-01 5.07374525e-01 -9.66500878e-01 3.21953118e-01 -7.28706717e-01 -2.34555230e-01 4.90474731e-01 1.41972518e+00 -6.43810987e-01 -1.20391917e+00 -7.23931909e-01 4.61561680e-01 -5.49773455e-01 -1.16372630e-01 -5.28106689e-01 8.46754074e-01 -5.76629937e-02 1.49448860e+00 -1.97535474e-03 -5.03491640e-01 4.11356747e-01 1.41089618e-01 7.21762300e-01 -1.23501487e-01 -9.04846430e-01 4.15413082e-01 3.84890139e-01 -1.23056793e+00 -8.58632207e-01 -1.05834293e+00 -1.40667629e+00 -1.19437599e+00 -5.12782812e-01 1.89310342e-01 3.83317530e-01 3.90289724e-01 -2.06836104e-01 2.85260603e-02 9.80917513e-01 -1.07949340e+00 -3.90341938e-01 -9.39418614e-01 -1.00026262e+00 -7.84516707e-02 2.89751232e-01 -8.00943017e-01 -7.83523321e-01 2.83909619e-01]

[-7.202101230621338, 3.728757858276367]

58f39764-adaa-4e9e-bfe7-dde1e56646a5

reusing-keywords-for-fine-grained

2210.11806

https://arxiv.org/abs/2210.11806v2

https://arxiv.org/pdf/2210.11806v2.pdf

Multi-view Semantic Matching of Question retrieval using Fine-grained Semantic Representations

As a key task of question answering, question retrieval has attracted much attention from the communities of academia and industry. Previous solutions mainly focus on the translation model, topic model, and deep learning techniques. Distinct from the previous solutions, we propose to construct fine-grained semantic representations of a question by a learned importance score assigned to each keyword, so that we can achieve a fine-grained question matching solution with these semantic representations of different lengths. Accordingly, we propose a multi-view semantic matching model by reusing the important keywords in multiple semantic representations. As a key of constructing fine-grained semantic representations, we are the first to use a cross-task weakly supervised extraction model that applies question-question labelled signals to supervise the keyword extraction process (i.e. to learn the keyword importance). The extraction model integrates the deep semantic representation and lexical matching information with statistical features to estimate the importance of keywords. We conduct extensive experiments on three public datasets and the experimental results show that our proposed model significantly outperforms the state-of-the-art solutions.

['Yueguo Chen', 'Denghao Ma', 'Li Chong']

2022-10-21

['keyword-extraction']

['natural-language-processing']

[ 1.31309882e-01 2.52543781e-02 -2.30781198e-01 -4.86929208e-01 -1.38209236e+00 -4.48509932e-01 6.91570342e-01 3.51823270e-01 -4.68223214e-01 2.94405192e-01 6.68052614e-01 7.86627680e-02 -3.26071292e-01 -9.18054342e-01 -6.50302589e-01 -2.44841829e-01 6.40141070e-01 5.23395181e-01 4.23465014e-01 -2.73677200e-01 5.45264900e-01 -1.73462451e-01 -1.54690611e+00 4.25593913e-01 9.36909974e-01 1.33251727e+00 2.92608678e-01 2.09748939e-01 -8.67544115e-01 6.02066815e-01 -6.19618595e-01 -4.91991013e-01 -1.41724244e-01 -3.37817878e-01 -1.17616880e+00 8.08650181e-02 3.03365082e-01 -1.14510804e-01 -2.68566042e-01 9.75804269e-01 2.88053632e-01 1.34526759e-01 6.37844741e-01 -9.91644442e-01 -8.57734442e-01 5.07128298e-01 -4.20455366e-01 1.40827909e-01 4.01944757e-01 -4.08549488e-01 1.74473524e+00 -1.15060103e+00 2.62726665e-01 1.40153337e+00 3.44913602e-01 2.67459303e-01 -7.82708466e-01 -5.78770876e-01 4.98102695e-01 6.01823330e-01 -1.23777413e+00 -1.64282233e-01 8.57841492e-01 -2.01085448e-01 6.88559234e-01 1.29894510e-01 1.10661283e-01 9.69585717e-01 -1.80633575e-01 9.28578496e-01 8.85465741e-01 -4.09248173e-01 2.04665691e-01 2.28217050e-01 5.62901258e-01 5.92996955e-01 6.60428181e-02 -5.61941862e-01 -4.29090500e-01 -4.66871858e-01 3.72250378e-01 3.23394775e-01 -2.54483938e-01 -1.72055513e-01 -1.15691817e+00 1.18763113e+00 3.89069140e-01 5.03019035e-01 -4.94286031e-01 2.60694921e-01 2.79399395e-01 2.14326724e-01 4.98749495e-01 5.56414485e-01 -7.75647342e-01 2.55597144e-01 -8.38818133e-01 4.62115854e-01 8.04812968e-01 8.98832083e-01 1.02060640e+00 -6.79158926e-01 -5.92660367e-01 1.08454299e+00 5.02281666e-01 2.18133181e-01 7.09428549e-01 -9.42717493e-01 5.59250891e-01 9.98401284e-01 1.41789183e-01 -1.02935755e+00 -1.76403224e-01 -4.95450437e-01 -4.80695963e-01 -7.42055893e-01 8.61187950e-02 8.92734900e-02 -8.89135718e-01 1.60302532e+00 3.88970762e-01 1.67482838e-01 8.95020142e-02 8.72300088e-01 1.03042161e+00 7.86536157e-01 2.19449148e-01 2.39999950e-01 1.98614323e+00 -1.27854359e+00 -8.71661127e-01 -3.45183969e-01 5.34130037e-01 -8.24813485e-01 1.08898675e+00 3.67100388e-02 -7.86788285e-01 -4.91927862e-01 -8.56496453e-01 -3.42047930e-01 -6.23360395e-01 1.47806630e-01 4.57535416e-01 2.36678138e-01 -6.12334669e-01 3.21759045e-01 -3.35691184e-01 -4.31516498e-01 2.47423276e-01 -9.61991921e-02 -2.25345388e-01 -3.91895294e-01 -1.70851684e+00 7.55137205e-01 4.39404726e-01 -2.19911158e-01 -7.25378096e-01 -6.41133428e-01 -8.65730822e-01 4.98592734e-01 5.95070958e-01 -1.03523791e+00 1.37926388e+00 -7.64167905e-01 -1.12192583e+00 8.13617706e-01 -4.39217895e-01 -1.41100854e-01 -2.92329013e-01 -3.99202615e-01 -2.73118079e-01 5.28818846e-01 6.10587955e-01 3.05278420e-01 9.88038361e-01 -1.22937214e+00 -8.04044664e-01 -5.93278527e-01 2.38429070e-01 1.79239720e-01 -5.93998790e-01 2.54546344e-01 -8.58232915e-01 -7.70663679e-01 3.14951003e-01 -4.80629891e-01 -1.36562556e-01 -1.67470336e-01 -1.77053317e-01 -8.74627411e-01 6.00971639e-01 -8.63217592e-01 1.33011782e+00 -1.89149082e+00 2.23478705e-01 3.27612436e-03 3.10406148e-01 -2.03205459e-02 -2.50054628e-01 5.98725438e-01 9.67004001e-02 -1.07956193e-02 -1.90783963e-01 -2.39703462e-01 3.30565423e-01 9.26321670e-02 -4.76171643e-01 4.92958575e-02 2.25462064e-01 1.15643668e+00 -9.48450089e-01 -5.42344689e-01 -3.54624569e-01 1.51557177e-01 -5.34684360e-01 7.18533278e-01 -4.87533033e-01 -2.24217206e-01 -1.22316253e+00 5.07408798e-01 5.44650316e-01 -8.16310823e-01 -2.58617271e-02 -3.47911328e-01 4.27620679e-01 5.81344664e-01 -1.01083934e+00 2.00839210e+00 -7.15996563e-01 -6.96506798e-02 2.66329758e-02 -1.28509724e+00 1.00125062e+00 4.11442220e-01 2.84921050e-01 -7.88827240e-01 5.74294031e-02 3.77735853e-01 -6.25642538e-01 -5.65347552e-01 5.82382023e-01 -2.03833088e-01 -5.61777532e-01 6.64534509e-01 4.16749328e-01 -1.00114837e-01 -1.73172817e-01 4.53483641e-01 9.15813744e-01 2.56232545e-02 2.47621819e-01 -1.83830097e-01 7.67987311e-01 -1.68310761e-01 2.86032617e-01 5.40540636e-01 1.26713634e-01 6.06252909e-01 4.85403895e-01 -2.13503063e-01 -6.22277975e-01 -8.96710992e-01 1.05916604e-01 1.41846704e+00 3.49679917e-01 -5.41660070e-01 -8.41728926e-01 -9.68650043e-01 1.23907097e-01 5.78798592e-01 -7.46190667e-01 -3.07979017e-01 -4.93640900e-01 -4.13812637e-01 2.17553556e-01 5.14122784e-01 4.63223130e-01 -9.46064353e-01 -1.52622193e-01 3.31018955e-01 -6.28778756e-01 -1.37990284e+00 -6.17412865e-01 4.53358851e-02 -5.88794410e-01 -1.04620945e+00 -9.37895894e-01 -9.04648960e-01 3.62165868e-01 5.55163801e-01 1.22606492e+00 1.23205028e-01 -2.64643952e-02 7.78158665e-01 -7.49514878e-01 -2.34525338e-01 1.73034631e-02 4.00106490e-01 -4.00711745e-01 3.01890492e-01 7.48425722e-01 -4.62459207e-01 -8.94415140e-01 2.69867748e-01 -1.14998138e+00 -2.48280928e-01 6.84072137e-01 7.43278563e-01 5.70303321e-01 -3.19024801e-01 1.05920577e+00 -8.15270185e-01 9.57413852e-01 -8.90648782e-01 -4.40181166e-01 4.77941930e-01 -6.50473535e-01 5.05047202e-01 5.14515877e-01 -1.00985087e-01 -1.01255393e+00 -5.10824680e-01 -3.45338613e-01 -4.08241719e-01 -2.33376920e-01 5.63844204e-01 -3.56327266e-01 3.71898085e-01 2.20900893e-01 4.81075495e-01 -2.88359284e-01 -8.93741131e-01 6.77820563e-01 8.86548579e-01 2.13318363e-01 -7.10519373e-01 7.95821071e-01 3.78679812e-01 -5.17859638e-01 -4.40931439e-01 -1.46874261e+00 -1.00929594e+00 -1.74329117e-01 1.70272470e-01 1.01786947e+00 -9.91689801e-01 -5.46822309e-01 1.95302576e-01 -1.38578463e+00 5.78864813e-01 -1.53385654e-01 2.86850303e-01 -4.22370791e-01 5.80391645e-01 -4.12296027e-01 -4.16509479e-01 -6.05776966e-01 -9.72730100e-01 1.72155774e+00 3.52624923e-01 -8.14682841e-02 -9.52734888e-01 9.83302593e-02 8.73692632e-01 4.99591112e-01 -2.56163448e-01 1.18992615e+00 -1.17039156e+00 -6.91595316e-01 -2.81861633e-01 -5.65215230e-01 2.99426764e-01 2.40777507e-01 -7.52132654e-01 -1.03572559e+00 3.44242826e-02 2.57448882e-01 -5.29105484e-01 1.27281725e+00 5.99606819e-02 1.35399330e+00 -3.90876919e-01 -3.34968299e-01 2.26453319e-01 1.21788192e+00 -3.03873658e-01 3.83057624e-01 4.84821588e-01 6.05704129e-01 9.55962241e-01 7.92483449e-01 3.68105143e-01 9.70984399e-01 6.58960819e-01 3.28530371e-01 2.41356090e-01 9.82924923e-02 -4.73040551e-01 -2.80862432e-02 1.00174582e+00 5.61451435e-01 -8.92462209e-02 -5.73364198e-01 7.70590663e-01 -1.90601408e+00 -8.10120761e-01 2.40364105e-01 1.83868027e+00 9.00523663e-01 -1.53428957e-01 -2.32514277e-01 -1.07532449e-01 6.67645216e-01 5.21144927e-01 -3.11127633e-01 -1.34391382e-01 2.87152231e-01 2.52893895e-01 -2.04341095e-02 2.76416421e-01 -1.01710725e+00 1.10741675e+00 5.14895439e+00 1.07506514e+00 -5.87127328e-01 2.70319492e-01 3.53706717e-01 3.52227539e-01 -9.81968641e-01 1.59649342e-01 -9.77280915e-01 3.36996228e-01 7.70465136e-01 -2.11877212e-01 6.44772425e-02 7.46500671e-01 -1.73134312e-01 1.30676523e-01 -8.81128609e-01 8.04031432e-01 4.14172590e-01 -1.38847625e+00 5.37796915e-01 -2.27350205e-01 5.95366061e-01 -1.78584069e-01 -2.65444577e-01 5.66979945e-01 1.05482303e-01 -7.85692275e-01 3.94586146e-01 7.08882391e-01 3.73740643e-01 -6.33528233e-01 6.85979843e-01 4.99855578e-01 -1.29334736e+00 -9.36977193e-02 -5.61528802e-01 3.13121557e-01 1.04619853e-01 6.12224460e-01 -3.38299334e-01 8.58683884e-01 7.48535156e-01 5.76330245e-01 -5.41175783e-01 6.97467923e-01 -3.22613269e-01 5.48623919e-01 4.51701664e-04 -2.22486034e-01 4.42233354e-01 -5.08446991e-02 7.75081664e-02 9.83175039e-01 2.16806605e-01 8.31566453e-02 7.61813298e-02 9.01198506e-01 -5.49136639e-01 4.48713303e-01 -1.27616093e-01 -1.59554705e-01 5.10583043e-01 1.43781722e+00 -3.76739144e-01 -5.25986016e-01 -7.57992208e-01 1.06822157e+00 4.41444725e-01 3.86954129e-01 -5.30641437e-01 -7.67928064e-01 7.46377587e-01 -1.38570249e-01 6.43330514e-01 9.18277875e-02 7.08744600e-02 -1.34934378e+00 2.55790532e-01 -9.94875014e-01 6.92364395e-01 -6.34010077e-01 -1.52380967e+00 5.21795928e-01 -4.39440049e-02 -8.95449817e-01 -3.72234583e-01 -3.42221498e-01 -5.63787222e-01 1.20699489e+00 -2.18115425e+00 -1.32410991e+00 -3.70015323e-01 6.61207438e-01 8.22284997e-01 -5.26819527e-02 8.51989686e-01 2.51664132e-01 -1.23857126e-01 4.93365407e-01 -6.55570403e-02 1.44396752e-01 7.13086605e-01 -1.10572124e+00 5.93014598e-01 4.43430841e-01 1.81266144e-01 7.13458002e-01 3.94530267e-01 -3.77490193e-01 -1.40795457e+00 -1.12301075e+00 1.26788688e+00 -4.67694104e-01 7.95658231e-01 -3.33097130e-01 -1.27222919e+00 4.42397892e-01 3.52119625e-01 1.54305678e-02 8.90270531e-01 1.22941084e-01 -7.14746535e-01 -7.27055147e-02 -7.73278415e-01 1.84574157e-01 6.78106785e-01 -7.74339795e-01 -1.35180378e+00 2.82014519e-01 1.42841887e+00 1.58718035e-01 -8.93579841e-01 3.63761783e-01 4.14440155e-01 -6.16554260e-01 1.02374482e+00 -8.14812541e-01 3.52705151e-01 -8.19718018e-02 -3.91886026e-01 -1.03536892e+00 -3.14218879e-01 -1.91728361e-02 -2.51742154e-01 1.42291558e+00 3.90357554e-01 -6.67443752e-01 5.43931961e-01 3.17438096e-01 2.91613229e-02 -8.89006138e-01 -8.14780891e-01 -4.15011853e-01 1.30131260e-01 -3.03233206e-01 8.72249365e-01 8.19152296e-01 -2.66052842e-01 7.95069814e-01 -4.29091789e-02 2.57945031e-01 5.67941546e-01 8.69380713e-01 5.48401773e-01 -1.41944408e+00 -3.11020613e-01 -3.71157885e-01 -3.70266587e-02 -1.44909918e+00 5.04254758e-01 -9.41709816e-01 1.68050453e-02 -1.95218110e+00 5.05808473e-01 -2.04352126e-01 -5.93055725e-01 1.80263549e-01 -7.64770508e-01 -2.09764168e-01 -6.45648614e-02 1.13998801e-01 -8.83253038e-01 1.02147007e+00 1.18488455e+00 -2.00888559e-01 4.18960363e-01 -2.55464818e-02 -1.30318761e+00 6.17837846e-01 5.98669112e-01 -6.91174030e-01 -4.11795437e-01 -5.99973738e-01 3.88152480e-01 5.75583614e-02 3.04164529e-01 -5.21437168e-01 2.80415177e-01 -5.90392388e-02 8.68379250e-02 -5.11641264e-01 2.83569217e-01 -9.01842475e-01 -7.71823764e-01 9.48455408e-02 -6.70682073e-01 -2.00900257e-01 -1.36216551e-01 8.21919501e-01 -6.48203313e-01 -5.11524200e-01 4.34918433e-01 -3.52750093e-01 -6.94253922e-01 4.70047235e-01 -3.51188742e-02 6.41501665e-01 5.81042171e-01 3.41108680e-01 -1.75820172e-01 -3.66071820e-01 -3.05199772e-01 6.10799313e-01 -5.22486027e-03 9.34054136e-01 7.62717366e-01 -1.36768126e+00 -7.07949042e-01 -4.89270687e-02 6.85495019e-01 -1.04557358e-01 2.41151556e-01 2.96392858e-01 1.90407932e-01 7.37268746e-01 3.23213398e-01 -2.66968101e-01 -7.43845224e-01 6.55231357e-01 1.04295716e-01 -6.27826989e-01 -2.44397089e-01 8.69762719e-01 5.27056694e-01 -6.74307168e-01 1.90840080e-01 -1.84036896e-01 -6.96855426e-01 3.51437181e-01 5.83933771e-01 4.57128324e-02 1.41764686e-01 -4.15033132e-01 -3.48886281e-01 8.74512136e-01 -3.27570230e-01 1.82406157e-02 1.29043496e+00 -3.47431928e-01 -1.91842720e-01 2.09510252e-01 1.64086378e+00 -2.10774735e-01 -5.60213566e-01 -9.42837596e-01 4.88452554e-01 -3.03559333e-01 1.71984687e-01 -4.68513638e-01 -8.31749082e-01 1.01878560e+00 2.45687574e-01 1.99742407e-01 1.21805239e+00 6.30616486e-01 1.34318185e+00 4.90207464e-01 1.54532850e-01 -1.04010701e+00 4.31401372e-01 4.64393228e-01 9.44323480e-01 -1.21920455e+00 -3.31144005e-01 -3.44837248e-01 -3.93568069e-01 9.13097262e-01 6.29854500e-01 -8.80294815e-02 5.26705801e-01 -3.03670257e-01 5.25191100e-03 -4.96132314e-01 -8.31195712e-01 -5.46410620e-01 6.60923719e-01 1.53640151e-01 3.22402239e-01 -2.52884060e-01 -4.11754668e-01 1.27386367e+00 2.18657017e-01 -1.72222674e-01 -5.20541519e-02 8.30275953e-01 -9.68710184e-01 -1.22941196e+00 -2.62046605e-01 5.87960005e-01 -6.89825535e-01 -3.57101768e-01 -2.45219618e-01 2.93676823e-01 -3.23530883e-01 9.64718580e-01 -1.62177086e-01 -2.44456396e-01 5.58117628e-01 3.73758674e-01 7.18425736e-02 -8.78504634e-01 -6.21272027e-01 -8.77058133e-02 -6.33344129e-02 -6.39432430e-01 -3.35300326e-01 -1.90381363e-01 -1.07318330e+00 3.20745081e-01 -3.66976798e-01 7.22359717e-01 5.98392010e-01 1.27798474e+00 5.13495028e-01 5.52140713e-01 8.99509013e-01 -2.39833429e-01 -8.90737593e-01 -9.75405812e-01 -4.51912999e-01 4.79045302e-01 2.06194773e-01 -5.37313044e-01 -3.90755206e-01 -2.60274261e-01]

[11.042142868041992, 8.11634635925293]

9496406d-b32e-403e-a685-284bcc88d585

ecg-segmentation-by-neural-networks-errors

1812.10386

http://arxiv.org/abs/1812.10386v1

http://arxiv.org/pdf/1812.10386v1.pdf

ECG Segmentation by Neural Networks: Errors and Correction

In this study we examined the question of how error correction occurs in an ensemble of deep convolutional networks, trained for an important applied problem: segmentation of Electrocardiograms(ECG). We also explore the possibility of using the information about ensemble errors to evaluate a quality of data representation, built by the network. This possibility arises from the effect of distillation of outliers, which was demonstarted for the ensemble, described in this paper.

['Aleksandra Koneva', 'Roman Kataev', 'Grigory Osipov', 'Sergey Alekseev', 'Iana Sereda']

2018-12-26

['electrocardiography-ecg']

['methodology']

[ 2.93503910e-01 4.01745111e-01 6.51749492e-01 -5.83732486e-01 -2.88375527e-01 -2.29525372e-01 1.36813402e-01 5.87764323e-01 -7.41301894e-01 1.04318392e+00 -8.54101554e-02 -4.11213189e-01 -4.14379895e-01 -7.43685722e-01 -7.72156119e-01 -7.24406540e-01 -4.77201253e-01 2.62023091e-01 -3.16861838e-01 -4.12694328e-02 3.27101946e-01 7.49714494e-01 -1.28613150e+00 4.76297498e-01 1.11059666e+00 9.92767870e-01 -6.56638563e-01 9.09700036e-01 2.43358359e-01 5.83129823e-01 -9.96942461e-01 -2.67971724e-01 2.03307390e-01 -5.23189127e-01 -7.01503515e-01 -2.61812478e-01 4.38505769e-01 -7.16060400e-02 -4.15937304e-02 1.18437481e+00 7.21545815e-01 -9.16771069e-02 7.42961824e-01 -9.81466174e-01 -1.16483793e-01 9.57134008e-01 7.09414855e-02 6.58551276e-01 -1.04755022e-01 2.87385225e-01 4.57223594e-01 -7.66415060e-01 6.01471722e-01 6.80268466e-01 1.12828541e+00 2.21319884e-01 -1.32366908e+00 -3.78272295e-01 -4.64668959e-01 1.08110562e-01 -1.53997481e+00 -3.43220294e-01 4.53706205e-01 -5.40265977e-01 6.87974989e-01 3.35422486e-01 5.68906248e-01 9.57742572e-01 6.76283419e-01 2.31944472e-01 1.08706498e+00 -2.83019334e-01 3.55697542e-01 4.26741913e-02 3.83756667e-01 3.68313611e-01 7.84600139e-01 3.76882166e-01 -1.37039602e-01 -3.58000427e-01 6.63332999e-01 -2.76315898e-01 -4.28268552e-01 2.42560834e-01 -9.68091249e-01 5.83345354e-01 3.87970507e-01 7.24904180e-01 -6.26201868e-01 1.82081223e-01 5.67008913e-01 4.88187820e-01 5.15634239e-01 1.10274506e+00 -7.17148185e-01 -2.27681354e-01 -1.18526733e+00 1.46836221e-01 8.73995662e-01 2.70444632e-01 6.50935769e-01 3.04244906e-01 -2.56557614e-01 3.15284938e-01 1.09082591e-02 8.95471498e-02 5.24085462e-01 -9.28911567e-01 2.25303143e-01 7.33697057e-01 -2.25192174e-01 -8.64532530e-01 -8.84336829e-01 -6.83446229e-01 -1.14814961e+00 4.85999793e-01 6.07101679e-01 -6.68693781e-01 -1.13669479e+00 1.34039640e+00 -1.97336338e-02 2.46381253e-01 3.17187011e-02 6.74069524e-01 8.45229149e-01 1.64583400e-02 2.86335312e-02 -7.27341101e-02 8.20745707e-01 -2.56442010e-01 -8.69567513e-01 2.14755043e-01 9.90689993e-01 -3.83680850e-01 2.29193687e-01 9.08107400e-01 -8.85745227e-01 -8.39276612e-01 -1.04004157e+00 3.36756259e-01 -4.61350322e-01 1.10008091e-01 3.87852013e-01 6.27347231e-01 -1.08034825e+00 1.60181391e+00 -7.18256474e-01 -1.45716682e-01 5.78840196e-01 4.86743003e-01 -3.57574463e-01 4.83707845e-01 -1.38414574e+00 1.19655395e+00 5.40560603e-01 3.99954319e-01 -4.96105880e-01 -4.19066995e-01 -6.23534322e-01 3.76461148e-02 -2.25831226e-01 -6.06207430e-01 8.77431750e-01 -1.55971611e+00 -9.11327362e-01 8.30516219e-01 3.79077643e-01 -8.86217177e-01 7.84487426e-01 -4.02455568e-01 -6.49108171e-01 -1.30663663e-01 -3.72523665e-01 9.34088528e-02 7.18942881e-01 -9.59325492e-01 -8.23041946e-02 -7.45442688e-01 -3.24706823e-01 -3.09946299e-01 3.98342639e-01 -2.71527112e-01 2.35239744e-01 -9.73437071e-01 1.64501771e-01 -9.03114736e-01 -5.76814711e-01 -3.86436790e-01 -5.35353005e-01 -9.84745566e-03 2.88903266e-01 -7.71820128e-01 1.40660989e+00 -2.27865458e+00 5.69626223e-03 7.64350057e-01 3.71917933e-01 5.95137656e-01 5.90588264e-02 4.08712506e-01 -5.26842356e-01 5.99931479e-01 -6.17709577e-01 -7.23950053e-03 -5.29048681e-01 3.71852040e-01 -5.48936576e-02 6.15957201e-01 4.27703530e-01 7.62631118e-01 -4.97479171e-01 -2.36973107e-01 2.43161514e-01 5.37468612e-01 -3.11317652e-01 2.97589935e-02 4.07237083e-01 7.25883842e-01 -4.28649396e-01 1.78664252e-01 7.32446969e-01 3.75923738e-02 1.82588086e-01 -3.14496577e-01 2.11125948e-02 2.45061964e-01 -1.24045789e+00 1.57373977e+00 6.24770075e-02 7.86774278e-01 -6.12062633e-01 -1.01165771e+00 1.12170756e+00 4.30622816e-01 6.07378781e-01 -4.50475365e-01 6.60395801e-01 3.69112343e-01 6.48844182e-01 -5.05466700e-01 4.56492901e-01 -1.30839646e-01 2.19254330e-01 3.37206662e-01 4.21172827e-01 1.88080877e-01 4.82363719e-03 -1.31423205e-01 1.15917277e+00 2.65450049e-02 4.34603274e-01 -5.65961003e-01 2.28884131e-01 -1.61445558e-01 5.44171393e-01 1.16453671e+00 -2.36938819e-01 1.01440895e+00 5.96584976e-01 -7.66094387e-01 -9.45679486e-01 -7.38300264e-01 -6.35087550e-01 2.38341346e-01 -3.76927614e-01 -3.09309542e-01 -1.07423151e+00 -7.38446534e-01 -1.20354295e-02 7.03685641e-01 -8.54514897e-01 -4.98885274e-01 -6.35981560e-01 -1.11778831e+00 9.35058653e-01 7.78809786e-01 2.17198059e-01 -1.24774802e+00 -1.11090779e+00 5.26053727e-01 2.36652903e-02 -6.57008708e-01 3.16404194e-01 6.63903117e-01 -1.47247922e+00 -1.20956624e+00 -3.35106790e-01 -4.70073558e-02 5.92834890e-01 -8.66071463e-01 1.38181472e+00 7.76801407e-01 -3.25318456e-01 1.39962539e-01 -3.91924024e-01 -8.16947997e-01 -6.24102354e-01 1.63806930e-01 1.18590118e-02 -2.73523387e-02 4.31948811e-01 -5.84484279e-01 -6.89095378e-01 -6.46813586e-02 -1.02764773e+00 -5.20470262e-01 4.35676485e-01 6.68852508e-01 2.92120457e-01 1.47766650e-01 5.09414434e-01 -1.19713461e+00 7.74081647e-01 -3.68471861e-01 -8.43144655e-02 8.32321942e-02 -7.70083427e-01 4.46078658e-01 4.42480057e-01 -1.68663159e-01 -4.40878719e-01 -1.70302376e-01 -6.85546815e-01 -3.81585360e-01 -5.42036176e-01 5.43308198e-01 1.73191860e-01 -3.36727351e-01 1.03992486e+00 -2.52635926e-01 6.86719716e-02 -3.16658676e-01 -2.23873928e-02 4.09762800e-01 4.83514786e-01 -2.90905625e-01 2.89245695e-01 2.33282149e-01 1.50151491e-01 -5.62102616e-01 -3.29519391e-01 -1.76879857e-02 -8.60450506e-01 -3.17796916e-01 7.74033129e-01 -4.22890604e-01 -3.69936585e-01 4.86126631e-01 -1.46308494e+00 -4.11579870e-02 -8.22845399e-01 5.63331425e-01 -3.85137528e-01 7.28687719e-02 -2.33302101e-01 -7.91837752e-01 -3.29370111e-01 -9.59083915e-01 4.88842517e-01 9.63991508e-02 -4.63786781e-01 -9.66005564e-01 2.96292931e-01 -3.83895040e-01 4.58651811e-01 8.62933457e-01 6.07567191e-01 -1.38237250e+00 -3.62894870e-02 -2.58147269e-01 2.87058875e-02 8.78891170e-01 -4.38683154e-03 3.33838880e-01 -1.40325034e+00 -7.58081675e-02 2.29250759e-01 1.18662834e-01 1.09269822e+00 5.44274032e-01 1.67415667e+00 -3.46464999e-02 1.76702756e-02 5.09969413e-01 1.38142049e+00 2.54365206e-01 1.16153800e+00 2.48463318e-01 2.63754576e-01 5.15367210e-01 1.83344305e-01 4.47897792e-01 -2.40091532e-01 1.73190132e-01 4.07551706e-01 -5.57354569e-01 2.11276188e-01 2.71273375e-01 -3.69288594e-01 7.41905034e-01 -7.95194149e-01 -2.36931011e-01 -1.17978501e+00 1.31396204e-01 -1.79774511e+00 -6.49813592e-01 -5.86566746e-01 2.06548214e+00 3.36722255e-01 4.62201297e-01 -1.69662282e-01 5.76051772e-01 6.15168929e-01 -1.96228996e-01 -4.86360431e-01 -7.25501239e-01 -1.88423902e-01 5.58411598e-01 6.91765845e-01 5.50171211e-02 -8.31291974e-01 5.33748455e-02 7.53862715e+00 3.56206864e-01 -1.03226268e+00 -2.02472769e-02 1.00283730e+00 5.84426463e-01 6.80650249e-02 -1.16043240e-01 -5.69790043e-02 4.71881896e-01 1.29422545e+00 2.04262435e-01 -6.90629855e-02 5.83246887e-01 2.37870082e-01 -2.30082929e-01 -1.10512793e+00 6.04917884e-01 4.06842791e-02 -1.28185081e+00 -1.59683228e-01 1.29428014e-01 3.85594934e-01 5.29408343e-02 -1.46756262e-01 1.24752149e-01 -7.90293515e-02 -1.19174755e+00 4.83118564e-01 1.15483236e+00 5.06099164e-01 -8.08473647e-01 1.34847283e+00 1.55989036e-01 -4.12489951e-01 -8.95085037e-02 -1.81502327e-01 -2.13864014e-01 -2.41147950e-02 9.54602242e-01 -8.56217444e-01 7.10833728e-01 7.70570517e-01 5.33924460e-01 -9.40260172e-01 1.16377437e+00 -4.94544487e-03 8.85650098e-01 -1.69110790e-01 4.22077954e-01 1.19895786e-02 -1.11787453e-01 6.92120671e-01 1.14873660e+00 2.90501714e-01 7.52569139e-02 -4.33779061e-01 9.02905285e-01 1.10567763e-01 4.81958799e-02 -8.01619828e-01 5.03818356e-02 9.04175639e-02 1.00027609e+00 -7.44915068e-01 -5.35058618e-01 1.99794203e-01 5.81288576e-01 -2.48837601e-02 2.52772063e-01 -7.40805745e-01 -5.78818500e-01 2.40335807e-01 1.00556165e-01 -4.14081402e-02 1.65690407e-01 -6.83025002e-01 -9.77119982e-01 -3.71642644e-03 -9.39641118e-01 4.13272023e-01 -8.16198051e-01 -1.27950573e+00 1.02165556e+00 -1.85718834e-01 -1.14809430e+00 -1.56936705e-01 -6.64675653e-01 -9.92805839e-01 9.28011417e-01 -9.40144241e-01 -1.37890220e-01 -2.32792854e-01 2.49576151e-01 6.45911030e-04 -5.12125343e-02 8.32420528e-01 3.71617675e-01 -3.38884175e-01 4.82668638e-01 3.87061648e-02 2.60503620e-01 6.16000175e-01 -1.46418071e+00 4.00171697e-01 7.90486634e-01 2.38974914e-02 6.30439162e-01 9.73960519e-01 -5.09385943e-01 -3.43179494e-01 -1.22736299e+00 8.91246378e-01 -5.79727411e-01 2.84184426e-01 3.48974228e-01 -1.45848536e+00 5.32149911e-01 4.19178516e-01 2.15890482e-01 7.92277932e-01 1.64023861e-01 -3.94909363e-03 -7.78415799e-02 -1.42113662e+00 -5.68501800e-02 7.50443161e-01 -2.28434071e-01 -8.46675873e-01 -2.26110592e-01 3.25282007e-01 -4.34808940e-01 -1.26573753e+00 8.63633871e-01 4.96805668e-01 -1.28416884e+00 6.70966685e-01 -8.91724586e-01 3.89878839e-01 -1.57813858e-02 1.96915075e-01 -1.70349932e+00 -1.08997941e-01 -2.70245224e-01 1.09214172e-01 9.56977308e-01 5.17995715e-01 -8.60531509e-01 4.65471596e-01 5.52417338e-01 -3.33595365e-01 -7.54016101e-01 -1.07464874e+00 -4.28439707e-01 3.54280710e-01 -5.83545506e-01 8.32341909e-01 1.03945112e+00 -4.49834913e-01 -7.57020861e-02 -2.23839104e-01 8.67061317e-02 3.22337985e-01 -5.00888467e-01 4.34063971e-01 -1.63347936e+00 5.00994101e-02 -2.01226696e-01 -8.84406626e-01 8.08728188e-02 -1.72226161e-01 -7.44405389e-01 -7.86602423e-02 -1.22150290e+00 -4.61455882e-01 -3.87693197e-01 -8.37819099e-01 1.25865087e-01 -2.60443002e-01 3.24106127e-01 2.80969292e-01 -4.57860753e-02 -3.74742419e-01 4.42222431e-02 9.35406327e-01 2.37198591e-01 -1.47615671e-01 6.29996043e-03 -5.25533140e-01 9.76308882e-01 1.13156557e+00 -9.08320844e-01 3.72311100e-02 -3.40899676e-01 5.55934906e-01 4.00871821e-02 6.59034371e-01 -1.51289570e+00 -4.88243364e-02 3.43847811e-01 7.17227340e-01 -5.69850326e-01 -2.37259597e-01 -9.87234235e-01 6.39636338e-01 5.65539360e-01 -4.62672859e-01 5.10796130e-01 2.48032302e-01 2.17640862e-01 -2.25803718e-01 -5.51993906e-01 7.41720259e-01 -2.50350744e-01 -1.84231460e-01 2.25704145e-02 -6.49413228e-01 4.41556573e-02 6.79964542e-01 -4.59179342e-01 3.83282043e-02 7.18925074e-02 -1.20442510e+00 -8.10940340e-02 3.03571969e-01 3.01969312e-02 5.59540808e-01 -1.15174937e+00 -7.58305371e-01 3.91269982e-01 1.77187026e-02 -2.06374526e-01 2.52195239e-01 1.07361317e+00 -6.55356228e-01 -1.93502262e-01 -3.54310960e-01 -7.03053892e-01 -1.01627803e+00 1.78309932e-01 9.12208736e-01 -2.05512747e-01 -6.09530747e-01 4.11892354e-01 -3.88526618e-01 -4.10326898e-01 5.27881260e-04 -6.98924422e-01 -4.00008172e-01 2.15708725e-02 3.14908206e-01 5.71986318e-01 6.37805462e-01 -3.66646737e-01 -3.69499832e-01 2.56394148e-01 3.11325908e-01 2.29858965e-01 1.37284887e+00 2.28848219e-01 -2.69591928e-01 8.09598863e-01 8.43681335e-01 -4.89290625e-01 -5.96826971e-01 2.60541588e-01 2.46057719e-01 -2.53133893e-01 -6.68855244e-03 -1.02878690e+00 -1.13029027e+00 1.07133126e+00 1.17995048e+00 3.95819694e-01 1.15376890e+00 -5.57124913e-01 1.95130661e-01 3.60498458e-01 2.63613075e-01 -1.21746492e+00 -2.99018443e-01 3.82868350e-01 7.78697431e-01 -1.19076371e+00 4.72212806e-02 1.73372820e-01 -6.26801193e-01 1.56341469e+00 3.28822285e-01 -5.31171739e-01 9.70718384e-01 2.35507309e-01 1.45015150e-01 -4.54806954e-01 -5.50843477e-01 4.40690443e-02 4.07632798e-01 5.65602005e-01 5.36605835e-01 1.48956776e-01 -8.00084770e-01 5.17777443e-01 6.24509826e-02 4.19064671e-01 8.15905988e-01 9.32517827e-01 -3.33314627e-01 -7.15087056e-01 -3.93472850e-01 9.11737919e-01 -7.37158179e-01 -1.00463614e-01 -4.65848446e-01 8.18130672e-01 8.61688673e-01 6.25049233e-01 1.47473618e-01 -5.58362722e-01 3.78393292e-01 2.65971869e-01 2.87465513e-01 -5.12957454e-01 -1.17794144e+00 -3.15127730e-01 1.53391883e-01 -4.14289176e-01 -6.49119198e-01 -5.14873445e-01 -1.14705205e+00 1.55308589e-01 -5.79650640e-01 3.78143787e-01 7.85469413e-01 1.08317435e+00 5.16879976e-01 1.02259159e+00 3.91685396e-01 -5.11569619e-01 -4.54741597e-01 -1.32334411e+00 -7.42678523e-01 5.40479302e-01 5.56389451e-01 -5.80318689e-01 -6.90433025e-01 -5.99993058e-02]

[14.328688621520996, 3.3062567710876465]

52749405-0de8-4900-9922-4cc62541718e

content-aware-unsupervised-deep-homography

1909.05983

https://arxiv.org/abs/1909.05983v2

https://arxiv.org/pdf/1909.05983v2.pdf

Content-Aware Unsupervised Deep Homography Estimation

Homography estimation is a basic image alignment method in many applications. It is usually conducted by extracting and matching sparse feature points, which are error-prone in low-light and low-texture images. On the other hand, previous deep homography approaches use either synthetic images for supervised learning or aerial images for unsupervised learning, both ignoring the importance of handling depth disparities and moving objects in real world applications. To overcome these problems, in this work we propose an unsupervised deep homography method with a new architecture design. In the spirit of the RANSAC procedure in traditional methods, we specifically learn an outlier mask to only select reliable regions for homography estimation. We calculate loss with respect to our learned deep features instead of directly comparing image content as did previously. To achieve the unsupervised training, we also formulate a novel triplet loss customized for our network. We verify our method by conducting comprehensive comparisons on a new dataset that covers a wide range of scenes with varying degrees of difficulties for the task. Experimental results reveal that our method outperforms the state-of-the-art including deep solutions and feature-based solutions.

['Jue Wang', 'Nianjin Ye', 'Lanpeng Jia', 'Shuaicheng Liu', 'Chuan Wang', 'Jirong Zhang', 'Jian Sun', 'Ji Zhou']

2019-09-12

https://www.ecva.net/papers/eccv_2020/papers_ECCV/html/2503_ECCV_2020_paper.php

https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123460630.pdf

eccv-2020-8

['homography-estimation']

['computer-vision']

[ 3.01708400e-01 -2.73674816e-01 -3.70348891e-04 -4.38250422e-01 -5.85978031e-01 -2.97977597e-01 3.82078439e-01 -3.17393392e-01 -2.79040039e-01 5.83016455e-01 9.55323577e-02 2.57547021e-01 -2.45612606e-01 -8.04942667e-01 -9.00566280e-01 -7.82187104e-01 3.41208726e-01 3.82122755e-01 1.48980394e-01 -2.22208023e-01 5.03397167e-01 5.04610419e-01 -1.45603120e+00 -2.61220559e-02 1.12116563e+00 1.00667179e+00 1.47909343e-01 -1.20980507e-02 2.15435922e-01 5.50529242e-01 -1.81183562e-01 -5.44987917e-01 9.46212888e-01 -4.84170318e-01 -3.86317521e-01 5.50503016e-01 1.04548097e+00 -5.76419950e-01 -6.17341459e-01 1.34761369e+00 5.14966249e-01 1.92387953e-01 4.35637116e-01 -1.21171141e+00 -5.82681358e-01 1.01740457e-01 -9.22633588e-01 -2.10482344e-01 2.44749218e-01 1.44516006e-01 1.01951027e+00 -9.58811700e-01 6.69509530e-01 9.29246187e-01 7.51019657e-01 1.15881927e-01 -1.18753326e+00 -6.23905778e-01 -7.98314586e-02 2.80486912e-01 -1.42735279e+00 -4.81790364e-01 1.19713700e+00 -4.34931129e-01 6.31246626e-01 -2.66057868e-02 4.78118747e-01 9.24216509e-01 2.37487003e-01 6.07743263e-01 1.06695426e+00 -3.46036315e-01 -8.49611498e-03 -2.67971426e-01 -2.17152670e-01 7.40499437e-01 3.77710253e-01 2.71504045e-01 -5.65211535e-01 1.35587081e-01 1.00225079e+00 4.27533239e-01 -4.98637885e-01 -9.85130668e-01 -1.51710165e+00 7.25344300e-01 5.74396193e-01 4.36371714e-02 -3.04790884e-01 9.17729363e-02 1.56732023e-01 2.07517803e-01 2.35167176e-01 3.47366929e-01 -1.76145807e-01 1.45551220e-01 -1.13909829e+00 7.86027908e-02 6.50166333e-01 1.19083345e+00 1.13730395e+00 3.92669290e-02 1.57310799e-01 8.04507136e-01 7.38967881e-02 2.92079210e-01 4.85080987e-01 -9.17225182e-01 6.16442502e-01 5.68461955e-01 9.41097438e-02 -1.63275468e+00 -3.14917266e-01 -6.30423307e-01 -1.25795877e+00 1.51340112e-01 4.49551165e-01 1.31346524e-01 -8.10013890e-01 1.36176229e+00 3.06228250e-01 4.07056332e-01 -2.36104399e-01 1.15548098e+00 4.62034494e-01 4.11629140e-01 -6.92341268e-01 -1.73620552e-01 9.66069698e-01 -1.26068401e+00 -7.01839149e-01 -2.42058620e-01 4.92216766e-01 -9.86286461e-01 1.00681424e+00 5.61066747e-01 -7.61221826e-01 -5.63819289e-01 -1.21197426e+00 -3.73343587e-01 -2.27973759e-01 3.09398264e-01 6.86358690e-01 4.09541100e-01 -7.97564924e-01 7.14354575e-01 -6.76992595e-01 -3.40846092e-01 1.10212170e-01 2.93168962e-01 -5.22057772e-01 -8.42738226e-02 -8.72134805e-01 6.16625786e-01 3.50862414e-01 3.00202310e-01 -4.93339062e-01 -4.88266051e-01 -9.89730299e-01 6.05386086e-02 4.84559983e-01 -7.51418948e-01 8.05941284e-01 -9.36790407e-01 -1.62017214e+00 8.55338633e-01 7.50645027e-02 -3.22497338e-01 7.96108425e-01 -4.90293294e-01 -8.74975622e-02 1.64190620e-01 8.44882056e-02 4.50932682e-01 1.02226150e+00 -1.27190065e+00 -5.78061461e-01 -3.05935830e-01 -3.81873697e-02 1.61042958e-01 -3.12925518e-01 -4.13275987e-01 -4.94696677e-01 -7.12280869e-01 6.88088894e-01 -9.62134480e-01 -3.10788721e-01 2.26165429e-01 -4.66755718e-01 3.73268306e-01 8.24041903e-01 -6.30959988e-01 9.05482650e-01 -2.03080988e+00 1.92960516e-01 3.51675779e-01 2.66401380e-01 4.18746881e-02 -4.19088975e-02 3.88027817e-01 -1.53756374e-02 -4.82335955e-01 -4.85755235e-01 -4.07183945e-01 7.73842260e-02 1.11742504e-01 -4.78649110e-01 8.59244823e-01 1.26376189e-03 6.40107989e-01 -9.03233707e-01 -4.88043696e-01 5.23791432e-01 4.40047979e-01 -7.46334374e-01 2.50557989e-01 1.68768466e-01 5.95681310e-01 -2.79822946e-01 6.25051677e-01 1.01109385e+00 -2.86774546e-01 2.06499696e-02 -6.44499362e-01 -1.89563498e-01 2.80680321e-02 -1.35418785e+00 2.05492568e+00 -5.15909791e-01 5.57164550e-01 -1.82742774e-01 -1.23099256e+00 1.10366654e+00 -1.27188981e-01 7.63780832e-01 -6.13126934e-01 2.86317706e-01 4.20063734e-01 1.54587151e-02 -3.80063176e-01 5.08711338e-01 2.27023438e-01 1.84638187e-01 4.20181938e-02 1.15467310e-01 -2.43608773e-01 8.27133581e-02 -1.26912087e-01 7.95347452e-01 2.12228119e-01 3.72933537e-01 -1.81333631e-01 5.68524182e-01 -1.16401687e-01 8.31025898e-01 5.54842889e-01 -7.90829957e-02 1.09401190e+00 2.42895707e-01 -6.93887651e-01 -1.02850103e+00 -7.96819985e-01 -2.40654781e-01 2.53589332e-01 6.70899272e-01 -3.34015369e-01 -5.38140357e-01 -5.18539846e-01 -4.30936553e-02 1.81478649e-01 -3.56553972e-01 -1.16409563e-01 -6.35895729e-01 -7.41560519e-01 3.51675004e-02 3.68942857e-01 1.10651481e+00 -6.14032507e-01 -6.46023095e-01 8.84230137e-02 -1.85255006e-01 -1.44864345e+00 -5.59036970e-01 -9.75492522e-02 -8.12759936e-01 -1.11156583e+00 -7.50809193e-01 -8.56049359e-01 9.57786381e-01 5.54045856e-01 9.92853582e-01 7.37489089e-02 -1.96040526e-01 1.91955611e-01 -3.58793169e-01 2.56489187e-01 2.07645088e-01 1.63109407e-01 3.51166613e-02 3.48726749e-01 3.49912584e-01 -7.70771623e-01 -9.24371004e-01 6.69120371e-01 -9.03342903e-01 1.38924465e-01 8.28368664e-01 1.16885042e+00 7.61834919e-01 2.53493339e-02 -1.60132663e-03 -6.91216528e-01 1.52477160e-01 -1.48586005e-01 -1.13260102e+00 2.25729585e-01 -5.16226292e-01 3.21890265e-02 7.92585135e-01 -3.41177046e-01 -7.80668378e-01 5.21830976e-01 5.78818098e-02 -7.63227880e-01 -3.30806524e-02 3.14012676e-01 -3.56695086e-01 -6.77708805e-01 3.18142921e-01 4.06335205e-01 -1.91996284e-02 -4.88978803e-01 1.91153511e-01 3.41609210e-01 7.92264938e-01 -4.73272413e-01 1.45860410e+00 7.84317970e-01 3.13277483e-01 -6.65844977e-01 -8.40426624e-01 -5.79111755e-01 -1.03613663e+00 -7.65137374e-02 7.23106563e-01 -8.90774786e-01 -6.61158264e-01 4.83410269e-01 -1.24921036e+00 -2.03484502e-02 2.58089956e-02 7.63400197e-01 -6.70347631e-01 8.80485952e-01 -3.22926432e-01 -3.24818939e-01 -1.73436716e-01 -1.40397048e+00 1.21780193e+00 6.32101968e-02 2.31921911e-01 -8.11735809e-01 1.53399393e-01 2.79964656e-01 1.95925191e-01 2.97281623e-01 4.11996454e-01 -3.62189412e-01 -1.11896646e+00 -1.40468419e-01 -4.20467407e-01 4.43556875e-01 2.10754365e-01 -1.05329938e-01 -8.04231524e-01 -4.09037709e-01 2.69105971e-01 -3.32020968e-01 8.88365328e-01 1.60165355e-01 1.42702866e+00 -1.25305414e-01 -2.88201887e-02 1.47597909e+00 1.70124304e+00 3.02270129e-02 9.19031799e-01 6.49748862e-01 1.12126505e+00 7.13891506e-01 7.86306024e-01 4.52733517e-01 2.51976877e-01 8.56036901e-01 6.14857852e-01 -2.85718113e-01 2.24610001e-01 -3.09965521e-01 1.01335697e-01 9.40578163e-01 -3.79683264e-02 -3.57348248e-02 -8.04593146e-01 3.69262308e-01 -2.03438020e+00 -8.71866167e-01 -2.31161807e-02 2.48333812e+00 6.17233157e-01 -1.09570771e-01 -2.39099726e-01 1.66662365e-01 5.37654102e-01 3.86940360e-01 -4.66818810e-01 1.85372934e-01 -2.10717261e-01 -4.29955386e-02 6.87018752e-01 3.90125424e-01 -1.28990102e+00 9.37787056e-01 5.14831972e+00 6.70605659e-01 -1.28501570e+00 -1.83317006e-01 4.04339433e-01 1.13510638e-01 -9.45252627e-02 3.40655237e-01 -7.23016977e-01 4.76838201e-01 -3.44182812e-02 1.15178160e-01 5.48799753e-01 8.06321144e-01 1.97601225e-02 -4.76815104e-02 -1.15068221e+00 1.54668915e+00 4.31740344e-01 -1.27625608e+00 8.95515829e-02 9.54457670e-02 1.08747852e+00 -7.57923499e-02 5.61026335e-02 -1.60885677e-01 -1.02172695e-01 -6.64004862e-01 4.99093831e-01 4.98121440e-01 5.75560629e-01 -5.44448256e-01 6.85794771e-01 1.93446800e-01 -1.06679463e+00 5.17694727e-02 -6.83159471e-01 6.73873052e-02 2.13275507e-01 8.38641822e-01 -2.23906189e-01 8.25625658e-01 6.64456189e-01 1.08465528e+00 -5.26229322e-01 1.40552258e+00 -1.77121863e-01 7.55153671e-02 -4.36807513e-01 4.92967963e-01 2.81023890e-01 -7.65748799e-01 4.16784137e-01 7.51388907e-01 6.34321272e-01 -2.97885835e-02 3.59032989e-01 8.72483969e-01 -1.68709919e-01 2.33123437e-01 -8.60422850e-01 2.77572125e-01 3.19671154e-01 1.27904558e+00 -6.98271453e-01 -1.12651758e-01 -6.04283929e-01 1.24848342e+00 1.16129689e-01 3.99579376e-01 -7.48464882e-01 -6.68982208e-01 4.57460612e-01 1.27345413e-01 2.71114349e-01 -6.06758296e-01 -1.86652124e-01 -1.71500409e+00 2.23476991e-01 -8.43105733e-01 2.24223603e-02 -5.90482056e-01 -1.30292594e+00 4.05687660e-01 -7.67725855e-02 -1.92733645e+00 -2.62812406e-01 -7.34595418e-01 -6.67981148e-01 5.69144845e-01 -1.80619609e+00 -1.14926445e+00 -8.57733250e-01 6.28238320e-01 5.21169424e-01 -2.66971886e-01 2.23490462e-01 6.05530322e-01 -5.66077232e-01 5.19172072e-01 4.09573197e-01 2.57731855e-01 9.84311283e-01 -9.64625478e-01 3.04796040e-01 1.13161075e+00 1.97777659e-01 7.79560447e-01 5.12608469e-01 -4.58121896e-01 -1.57637501e+00 -8.86675537e-01 6.46001637e-01 -1.11340962e-01 5.94878435e-01 -4.49016869e-01 -8.57563734e-01 6.56724811e-01 1.77473158e-01 1.71482161e-01 3.52623433e-01 -3.10434133e-01 -2.84953475e-01 -3.97554964e-01 -9.67982829e-01 5.43039620e-01 1.36455452e+00 -4.62247580e-01 -3.09848428e-01 4.02184844e-01 4.66467500e-01 -6.64649487e-01 -7.42615461e-01 5.68560243e-01 6.02690339e-01 -1.43910038e+00 9.69341576e-01 -1.26280054e-01 5.40963888e-01 -6.44054174e-01 -2.63676107e-01 -1.17012036e+00 -1.93947375e-01 -8.01340461e-01 1.41436830e-01 1.11034477e+00 -1.77595735e-01 -7.67677486e-01 7.89886355e-01 1.71776369e-01 -2.20103294e-01 -6.78276360e-01 -7.44716585e-01 -8.00980568e-01 -3.38497221e-01 8.22619870e-02 7.11704016e-01 1.24381316e+00 -4.62845951e-01 1.58222895e-02 -7.51634181e-01 4.54394370e-01 1.04028022e+00 5.71006417e-01 1.21289980e+00 -1.04328680e+00 -2.71913320e-01 -2.83272147e-01 -8.00552368e-01 -1.36444998e+00 2.43131667e-01 -6.70532763e-01 8.94969702e-02 -1.12996292e+00 7.69534484e-02 -2.07559258e-01 6.58331905e-03 1.78618222e-01 2.10896537e-01 4.21876043e-01 5.49978912e-02 3.75786245e-01 -3.46015513e-01 9.71879840e-01 1.27711308e+00 -1.79971576e-01 -9.56259947e-03 -1.77103713e-01 -2.13123053e-01 8.76014709e-01 5.85056007e-01 -2.58436978e-01 -5.72247505e-01 -7.64668047e-01 1.80948511e-01 -2.55855173e-01 5.25500596e-01 -1.25426173e+00 4.37703162e-01 -2.58005619e-01 4.37338144e-01 -8.12870741e-01 2.26403072e-01 -1.20811355e+00 6.48830235e-02 1.90059558e-01 -5.23638465e-02 2.44972989e-01 -2.29015112e-01 4.19825017e-01 -5.29836953e-01 -1.01934105e-01 7.41607785e-01 -1.66782867e-02 -8.31975460e-01 6.19358599e-01 3.74090612e-01 -1.01472303e-01 9.23655868e-01 -4.42830116e-01 -3.26657057e-01 -4.65748578e-01 -2.22946718e-01 2.06892282e-01 9.61748898e-01 2.60665178e-01 9.40412760e-01 -1.33842731e+00 -3.06761265e-01 5.95823288e-01 2.25704372e-01 2.51434803e-01 1.03906699e-01 1.07287669e+00 -9.28816915e-01 1.35730699e-01 -6.62356496e-01 -7.45593846e-01 -7.69716144e-01 5.06122470e-01 4.26172435e-01 -1.81515832e-02 -7.56719589e-01 2.55250663e-01 5.04932106e-01 -4.49520618e-01 3.59153062e-01 -3.36850762e-01 1.21408574e-01 -1.82463631e-01 1.60267293e-01 2.84749240e-01 9.71306786e-02 -7.96160579e-01 -8.70570987e-02 1.12706912e+00 -7.62833655e-02 1.12554759e-01 1.25943875e+00 -1.56841800e-01 -1.86280668e-01 1.37619868e-01 1.47498238e+00 8.41955934e-03 -1.38681364e+00 -2.49559626e-01 -6.47517219e-02 -1.08666611e+00 6.68975338e-02 -1.01329975e-01 -1.48365498e+00 1.01945150e+00 6.75045669e-01 -2.94618487e-01 1.21630883e+00 -5.50584674e-01 1.09555686e+00 6.78606033e-01 5.02336502e-01 -1.09006381e+00 2.02487737e-01 3.89458001e-01 8.56414676e-01 -1.52852845e+00 2.53778517e-01 -5.42438209e-01 -4.04345751e-01 1.19014776e+00 8.55760634e-01 -4.05580103e-01 6.04273617e-01 -1.25050649e-01 9.51354429e-02 -1.66589394e-01 -2.11344898e-01 -2.24843398e-01 4.13328916e-01 5.02183914e-01 1.80973232e-01 -4.91838604e-01 -4.07471538e-01 4.72124033e-02 -1.56457454e-01 -1.75841868e-01 4.58564460e-01 8.11688662e-01 -1.68787852e-01 -1.13116789e+00 -2.76547343e-01 2.44236931e-01 -6.62371814e-02 -2.33181310e-03 -3.13875496e-01 9.15502846e-01 -2.54523419e-02 4.63792890e-01 8.00251737e-02 -5.48538923e-01 3.13085765e-01 -4.91696388e-01 4.58665550e-01 -3.16039383e-01 -2.04373032e-01 1.88227743e-01 -3.25640827e-01 -8.56799960e-01 -5.98295748e-01 -3.62154484e-01 -8.02954614e-01 -2.56936073e-01 -2.99028516e-01 -2.59637862e-01 4.66887593e-01 8.20771873e-01 2.02348068e-01 4.23387578e-03 9.50760722e-01 -1.18176460e+00 -7.77227461e-01 -5.69408715e-01 -6.60626531e-01 7.04892039e-01 4.36094910e-01 -9.29232359e-01 -4.15750712e-01 5.00043817e-02]

[8.726795196533203, -2.3044636249542236]

6235f4ad-ff86-4913-a730-3b9a3082229e

self-calibrating-neural-probabilistic-model

2106.11196

https://arxiv.org/abs/2106.11196v1

https://arxiv.org/pdf/2106.11196v1.pdf

Self-Calibrating Neural-Probabilistic Model for Authorship Verification Under Covariate Shift

We are addressing two fundamental problems in authorship verification (AV): Topic variability and miscalibration. Variations in the topic of two disputed texts are a major cause of error for most AV systems. In addition, it is observed that the underlying probability estimates produced by deep learning AV mechanisms oftentimes do not match the actual case counts in the respective training data. As such, probability estimates are poorly calibrated. We are expanding our framework from PAN 2020 to include Bayes factor scoring (BFS) and an uncertainty adaptation layer (UAL) to address both problems. Experiments with the 2020/21 PAN AV shared task data show that the proposed method significantly reduces sensitivities to topical variations and significantly improves the system's calibration.

['Robert M. Nickel', 'Dorothea Kolossa', 'Benedikt Boenninghoff']

2021-06-21

['authorship-verification']

['natural-language-processing']

[-2.22543761e-01 3.57076749e-02 -3.88504326e-01 -4.97104645e-01 -1.16736603e+00 -7.36158669e-01 1.09066498e+00 3.20618033e-01 -3.44807714e-01 1.01944494e+00 1.44461304e-01 -5.12477338e-01 -1.77248418e-02 -5.03336549e-01 -6.90565050e-01 -2.69568145e-01 5.41417897e-01 8.87292624e-01 3.00104674e-02 1.28882095e-01 5.50927758e-01 1.80761874e-01 -9.30284858e-01 2.12987006e-01 1.04291844e+00 8.53447556e-01 -4.35405433e-01 6.82347178e-01 -3.13626170e-01 5.91979146e-01 -1.37505174e+00 -9.99666393e-01 6.61204830e-02 -1.38288010e-02 -8.92405927e-01 -3.48771900e-01 8.85092854e-01 -3.68905246e-01 -2.65499830e-01 1.06417596e+00 2.56689221e-01 -2.66333342e-01 1.25504386e+00 -1.65992570e+00 -5.36996365e-01 1.28360093e+00 -9.39387321e-01 4.45851505e-01 7.57582933e-02 -1.25830218e-01 1.30260921e+00 -8.80578220e-01 6.41391635e-01 1.21087384e+00 1.01900172e+00 2.34563828e-01 -1.17920971e+00 -1.04802787e+00 -7.17909113e-02 3.38334054e-01 -1.50862288e+00 -3.54560584e-01 6.55309200e-01 -5.47632694e-01 4.49534178e-01 -3.67043470e-03 1.20284416e-01 1.77595603e+00 5.03110290e-01 8.19204807e-01 1.00107038e+00 -3.59102368e-01 2.26497129e-01 4.28606331e-01 6.09716892e-01 7.86654204e-02 8.71890843e-01 -1.18213318e-01 -8.69524956e-01 -7.46669173e-01 3.17523450e-01 -5.07027924e-01 -3.05536270e-01 -1.74897015e-02 -9.55894172e-01 9.47933316e-01 -6.25362573e-03 1.91515446e-01 -1.35751441e-01 2.93482125e-01 5.66290259e-01 1.85542211e-01 3.65229160e-01 5.91949880e-01 -6.07588053e-01 -4.32571620e-01 -1.34419978e+00 6.50961161e-01 1.19301927e+00 1.00495684e+00 2.92535245e-01 -1.11854501e-01 -3.59191597e-01 7.36601472e-01 5.33764064e-01 6.38673246e-01 2.67221153e-01 -1.00889671e+00 6.59880877e-01 1.71950057e-01 3.60082805e-01 -1.11826754e+00 -2.30773300e-01 -5.10520101e-01 -6.47520125e-01 -1.76119730e-01 1.02834070e+00 -3.93079549e-01 -8.59320879e-01 1.71547449e+00 7.21284226e-02 -1.44838607e-02 -2.14802146e-01 4.79235768e-01 7.79362619e-01 4.33220953e-01 3.00268829e-01 -1.14749648e-01 1.35964036e+00 -2.74589479e-01 -1.12741053e+00 -2.43178502e-01 4.77250278e-01 -8.79241824e-01 7.35856950e-01 4.04267907e-01 -7.16342747e-01 -1.15526922e-01 -1.07400858e+00 -3.34752910e-02 -2.97694951e-01 2.73407400e-02 4.95644242e-01 8.13445449e-01 -5.73472381e-01 5.92112780e-01 -4.94022936e-01 -9.14747193e-02 6.44094408e-01 -3.66066722e-03 -1.28736824e-01 3.79270375e-01 -1.67142963e+00 8.62103224e-01 2.54323870e-01 2.42947992e-02 -5.45867503e-01 -1.11941576e+00 -6.54264629e-01 2.02479273e-01 3.18939596e-01 -2.86740780e-01 1.60408986e+00 -3.79794538e-01 -1.27222562e+00 4.52321708e-01 -1.55745387e-01 -5.91743410e-01 8.81304741e-01 -3.95269454e-01 -3.22330177e-01 -1.66497514e-01 1.25209793e-01 1.00765146e-01 8.92130256e-01 -1.39873302e+00 -5.92508137e-01 -2.11767614e-01 -6.74311340e-01 -2.38639414e-01 -8.70733634e-02 -2.41609886e-02 -1.94665983e-01 -8.48714232e-01 -1.02542996e-01 -7.88239300e-01 4.51768935e-01 -1.04320072e-01 -6.85578227e-01 -6.22114301e-01 8.21596265e-01 -9.59570169e-01 1.34520233e+00 -1.76132679e+00 -1.24200083e-01 5.60491383e-01 4.14953738e-01 2.14339852e-01 1.56664431e-01 2.82304615e-01 3.38100642e-01 3.75425279e-01 -2.10975856e-01 -5.81614494e-01 1.15916021e-01 1.22224661e-02 -7.80608833e-01 4.82608080e-01 -7.39233894e-03 7.50576437e-01 -6.77712619e-01 -7.40489244e-01 -1.93335950e-01 3.04055303e-01 -1.42178893e-01 1.60548445e-02 -3.29866141e-01 4.85870652e-02 -5.39641455e-03 6.25554979e-01 8.16964388e-01 -4.82312322e-01 3.03085595e-01 -9.98323224e-03 3.35083008e-01 7.20575690e-01 -1.20980144e+00 1.14089847e+00 -1.43214941e-01 1.16842258e+00 -1.98624253e-01 -2.52313942e-01 8.82965267e-01 2.39690959e-01 6.92722052e-02 -9.49779749e-02 4.00708050e-01 2.69220173e-01 1.17500663e-01 1.34695515e-01 1.06667638e+00 2.02998370e-01 -2.56400555e-01 8.23421478e-01 3.13010752e-01 -1.56253010e-01 -8.36373642e-02 5.42003751e-01 6.77276492e-01 -3.33302766e-01 1.34073496e-01 -2.72979468e-01 1.93746552e-01 -2.12419093e-01 9.04519796e-01 1.36152112e+00 -4.31742072e-01 7.12244630e-01 1.12277651e+00 -2.01675091e-02 -1.23632812e+00 -1.20740390e+00 -8.22621107e-01 6.94147348e-01 -7.93689340e-02 -6.01150811e-01 -7.10334480e-01 -1.01839674e+00 3.92395049e-01 1.15590954e+00 -7.87648976e-01 -3.73754650e-02 -3.36414248e-01 -7.02990472e-01 9.43036139e-01 6.21270537e-01 2.55552590e-01 -6.83170855e-01 -2.06802636e-01 1.30617946e-01 -3.24877769e-01 -1.12493074e+00 -4.78828371e-01 5.08381240e-02 -5.12064695e-01 -1.15536046e+00 -6.68707609e-01 -1.85526088e-02 1.58683524e-01 -2.23686367e-01 1.16268766e+00 -1.87763005e-01 1.06658638e-01 1.29563197e-01 -1.62412018e-01 -8.49434257e-01 -8.19095492e-01 4.74779546e-01 1.29305556e-01 -1.96157709e-01 7.42414534e-01 -8.56281221e-02 -1.23523585e-01 1.60080716e-01 -5.33149898e-01 -4.17807341e-01 2.98361242e-01 1.08573151e+00 -6.86493609e-03 -2.84210056e-01 5.71526706e-01 -1.40260708e+00 8.79330397e-01 -4.83979702e-01 -7.68345714e-01 4.21669304e-01 -1.19779181e+00 1.90893486e-02 7.28133693e-02 -5.30596554e-01 -1.30041325e+00 -7.11848378e-01 3.08629096e-01 -2.92813748e-01 2.58318707e-02 3.55825752e-01 1.12144485e-01 2.39716038e-01 7.67311931e-01 -1.40681878e-01 -9.46579315e-03 -4.15848255e-01 1.16454870e-01 1.24437904e+00 7.36342192e-01 -7.16165900e-01 6.95344388e-01 8.50186311e-03 -3.26712310e-01 -4.58219886e-01 -1.00225258e+00 -2.11062893e-01 -3.75519395e-01 1.33151589e-02 3.46046567e-01 -9.39370632e-01 -6.73771501e-01 7.46962786e-01 -1.50854695e+00 -2.00185210e-01 -1.58085376e-02 3.91924739e-01 -8.01803172e-02 6.35647357e-01 -7.93353617e-01 -9.62091267e-01 -4.65056449e-01 -9.92394090e-01 1.07646286e+00 4.17996705e-01 -8.15392137e-01 -9.55467522e-01 2.27028280e-01 3.76109809e-01 3.64910066e-01 -8.71291310e-02 1.05238855e+00 -1.07774413e+00 -3.27993631e-02 -3.02705854e-01 -4.33296055e-01 1.71687454e-01 -1.85276866e-01 5.56376278e-01 -1.14445889e+00 -9.71682519e-02 -3.17618728e-01 -1.79675072e-01 7.94121802e-01 4.46088433e-01 1.02063251e+00 -2.60782808e-01 -3.98131698e-01 2.50043780e-01 8.88772607e-01 -1.90611646e-01 3.97108078e-01 2.91874081e-01 4.92315561e-01 4.77082908e-01 3.68359327e-01 5.94409883e-01 4.78202105e-01 7.19450116e-01 -5.03110141e-03 5.08178890e-01 -1.17277555e-01 -4.44640905e-01 9.25131589e-02 4.55876648e-01 2.65625149e-01 -5.84113896e-01 -1.24444127e+00 5.38204551e-01 -1.77325106e+00 -8.96972537e-01 -5.10786474e-01 2.13909245e+00 1.12414885e+00 2.89151222e-01 -6.23444468e-02 1.60087839e-01 9.21926856e-01 1.68106407e-01 -4.61520642e-01 -5.04203200e-01 -2.84212410e-01 -7.96797350e-02 6.89058661e-01 6.01634860e-01 -9.84630883e-01 8.50940526e-01 7.59302235e+00 7.93158948e-01 -6.22664571e-01 1.75265387e-01 5.44624150e-01 1.51440054e-01 -4.42943156e-01 -1.34980516e-03 -1.07339585e+00 7.42346287e-01 1.12401545e+00 -4.35347944e-01 -6.40342236e-02 7.43288934e-01 -3.83963615e-01 -1.81580111e-01 -1.13198459e+00 1.00074804e+00 1.13138124e-01 -1.37866557e+00 1.71853378e-01 2.89867610e-01 8.79621327e-01 -6.80662170e-02 1.69412747e-01 3.77671838e-01 7.05141902e-01 -8.90608490e-01 8.23166609e-01 4.30800229e-01 8.26048970e-01 -6.86070740e-01 1.25005448e+00 1.15367137e-02 -2.66638935e-01 8.33204761e-02 -2.88458467e-01 4.11237150e-01 1.32662266e-01 8.49754870e-01 -1.25968707e+00 2.00207606e-01 6.09259605e-01 3.82909656e-01 -5.97808838e-01 9.21113610e-01 -5.22941768e-01 9.55019236e-01 -3.39726627e-01 -1.31869078e-01 -2.11452469e-01 3.78682226e-01 7.73797214e-01 1.07678831e+00 -6.62942836e-03 -3.54489237e-01 -6.30444884e-01 1.12535405e+00 -7.83583343e-01 -3.10099781e-01 -3.25276554e-01 -2.82143295e-01 9.24618483e-01 9.72438276e-01 -2.31389537e-01 -3.61357361e-01 -2.01295987e-01 6.49481654e-01 3.31786633e-01 3.70334804e-01 -7.74385929e-01 -3.22715044e-01 7.01793492e-01 -1.48525104e-01 2.04797253e-01 -1.23264156e-01 -8.41515183e-01 -1.25023758e+00 -2.56117642e-01 -9.53266740e-01 6.30681932e-01 -6.42618656e-01 -1.78187561e+00 3.46985698e-01 5.98368123e-02 -6.90756083e-01 -4.33409363e-01 -5.70644379e-01 -4.43101168e-01 1.01746428e+00 -1.23594999e+00 -8.57225597e-01 -9.20012742e-02 1.15319662e-01 4.41916078e-01 -4.45577353e-01 6.22274280e-01 3.26858973e-03 -5.34319997e-01 1.27821493e+00 3.05195928e-01 3.35498661e-01 1.30741060e+00 -1.56824422e+00 5.11263132e-01 8.27015042e-01 3.34860504e-01 7.76131272e-01 1.14557958e+00 -9.41697001e-01 -9.43071425e-01 -6.35904014e-01 1.21284294e+00 -1.18046153e+00 8.82753968e-01 -4.26222265e-01 -1.21599770e+00 7.49249518e-01 8.77211839e-02 -4.12326515e-01 7.90854156e-01 7.06934869e-01 -9.36318278e-01 1.95286766e-01 -1.17966199e+00 4.52582389e-01 3.81842077e-01 -7.35238254e-01 -9.91788805e-01 2.90803790e-01 4.97454047e-01 -6.28842711e-01 -9.89284813e-01 1.72854319e-01 7.97714114e-01 -4.51574266e-01 4.54231173e-01 -8.08849216e-01 4.82987314e-01 -4.96007539e-02 4.64598089e-02 -1.21811640e+00 -1.37195632e-01 -5.95951676e-01 -6.00653291e-01 1.60312712e+00 6.85325027e-01 -6.67893589e-01 7.46445596e-01 1.13683391e+00 4.83393192e-01 -6.89805821e-02 -1.16875124e+00 -5.88512003e-01 5.31040370e-01 -6.53814137e-01 7.41834283e-01 1.14402127e+00 -2.88556404e-02 3.12531143e-01 -3.41631144e-01 3.01816285e-01 9.16479766e-01 -7.23458380e-02 8.62548411e-01 -1.62148285e+00 -1.45708025e-01 -5.45812190e-01 -1.36369511e-01 -5.26810050e-01 3.69276822e-01 -4.75150704e-01 7.30434135e-02 -1.07418561e+00 5.20381510e-01 -6.24758422e-01 -1.92788109e-01 3.35548848e-01 -3.71584237e-01 -1.08640529e-01 7.20272437e-02 5.57463825e-01 -2.43292183e-01 4.34672952e-01 5.26955605e-01 -1.62156165e-01 1.14537984e-01 1.15786996e-02 -7.57861674e-01 5.42147040e-01 5.66709578e-01 -8.19782972e-01 3.69191319e-02 -3.28323007e-01 4.46999937e-01 -8.80784765e-02 2.21925214e-01 -5.29883742e-01 3.86050522e-01 -7.25463480e-02 4.61497426e-01 -7.78544962e-01 -9.86166745e-02 -7.01712608e-01 -1.03611253e-01 1.24054536e-01 -7.00395107e-01 -2.30032392e-02 2.59363651e-01 8.61547112e-01 4.27987091e-02 -3.77459943e-01 5.52253425e-01 3.70981127e-01 -8.40950981e-02 -4.77892766e-03 -3.90580624e-01 3.57947409e-01 5.46221375e-01 1.29899725e-01 -8.77802253e-01 -4.13573384e-01 -1.20718077e-01 3.66340041e-01 4.20073926e-01 5.29945672e-01 9.27245021e-02 -1.08152211e+00 -8.50238979e-01 -5.73520735e-02 1.04858167e-01 -1.12508245e-01 4.31373157e-02 6.60770178e-01 -3.44227344e-01 5.42754710e-01 8.41056779e-02 -6.26347184e-01 -1.10900974e+00 1.44858072e-02 2.38280758e-01 -3.08190465e-01 -4.00502741e-01 8.55872154e-01 -2.39776894e-01 -4.25549984e-01 4.56387222e-01 -5.20551428e-02 -3.30932178e-02 1.49597168e-01 5.83207548e-01 5.91670692e-01 1.31332651e-01 -5.60031295e-01 -3.43599975e-01 -1.71934869e-02 -6.13069177e-01 -3.69116873e-01 1.08761728e+00 1.45334620e-02 1.68945059e-01 7.19648778e-01 7.13095605e-01 2.09893897e-01 -1.03838873e+00 -5.76580226e-01 2.89526463e-01 -4.61430937e-01 2.04677060e-01 -1.11103797e+00 -7.77666867e-01 8.98457825e-01 2.91316003e-01 1.42337784e-01 2.24323973e-01 -1.16018929e-01 7.45002031e-01 4.07479733e-01 1.48875743e-01 -1.34918892e+00 -3.36115986e-01 5.20636797e-01 8.52831900e-01 -1.37475753e+00 2.67541468e-01 -1.70353502e-01 -7.17701018e-01 1.18834949e+00 6.03458703e-01 3.66361469e-01 5.53065419e-01 4.21372235e-01 3.06931019e-01 -1.80321455e-01 -7.58535624e-01 5.86352944e-01 4.98265862e-01 5.01670241e-01 3.82920504e-01 -5.34340506e-03 -4.67496604e-01 8.03560317e-01 -3.81855160e-01 -3.26769710e-01 1.03921127e+00 4.63274360e-01 -8.39395262e-03 -9.97548699e-01 -4.98062044e-01 7.35843480e-01 -7.16533661e-01 -5.18302023e-02 -6.72566652e-01 9.67510939e-01 -2.79272497e-01 7.42422342e-01 2.94161767e-01 -2.84706295e-01 -8.71783644e-02 3.84846002e-01 1.84879228e-01 -4.29676890e-01 -6.65663362e-01 -2.72316307e-01 2.71225095e-01 -1.49355963e-01 1.52903229e-01 -1.13743877e+00 -9.23422813e-01 -6.33296549e-01 -6.80846155e-01 3.07289720e-01 8.77439737e-01 9.29425418e-01 4.79913652e-01 2.48281226e-01 1.70333996e-01 -2.30541542e-01 -1.17976785e+00 -1.34943509e+00 -6.78718865e-01 2.76583791e-01 1.79067284e-01 -8.47275734e-01 -7.02597857e-01 -1.84632957e-01]

[9.62933349609375, 10.555913925170898]

d7c2c725-ec6d-44bb-bb85-631f1cbc04bb

similarity-kernel-and-clustering-via-random

1908.10506

https://arxiv.org/abs/1908.10506v1

https://arxiv.org/pdf/1908.10506v1.pdf

Similarity Kernel and Clustering via Random Projection Forests

Similarity plays a fundamental role in many areas, including data mining, machine learning, statistics and various applied domains. Inspired by the success of ensemble methods and the flexibility of trees, we propose to learn a similarity kernel called rpf-kernel through random projection forests (rpForests). Our theoretical analysis reveals a highly desirable property of rpf-kernel: far-away (dissimilar) points have a low similarity value while nearby (similar) points would have a high similarity}, and the similarities have a native interpretation as the probability of points remaining in the same leaf nodes during the growth of rpForests. The learned rpf-kernel leads to an effective clustering algorithm--rpfCluster. On a wide variety of real and benchmark datasets, rpfCluster compares favorably to K-means clustering, spectral clustering and a state-of-the-art clustering ensemble algorithm--Cluster Forests. Our approach is simple to implement and readily adapt to the geometry of the underlying data. Given its desirable theoretical property and competitive empirical performance when applied to clustering, we expect rpf-kernel to be applicable to many problems of an unsupervised nature or as a regularizer in some supervised or weakly supervised settings.

['Donghui Yan', 'Songxiang Gu', 'Ying Xu', 'Zhiwei Qin']

2019-08-28

['clustering-ensemble']

['graphs']

[ 2.17271775e-01 -1.02618419e-01 -2.72988319e-01 -4.31703061e-01 -1.45502836e-01 -5.19958913e-01 7.57299423e-01 1.80560336e-01 -2.21554488e-01 5.99537730e-01 2.69159853e-01 -3.17973286e-01 -6.45495594e-01 -1.00854468e+00 -4.34371799e-01 -1.19393897e+00 -3.72972816e-01 5.86087584e-01 4.68237251e-01 1.46487445e-01 4.24469858e-01 5.59617639e-01 -1.67987168e+00 2.18067467e-01 9.28931415e-01 8.18807781e-01 6.03952147e-02 1.96545601e-01 -5.55513613e-02 3.06506276e-01 2.85894908e-02 -2.07931668e-01 3.51584435e-01 -2.92189270e-01 -9.07043576e-01 2.47686505e-02 5.94550632e-02 6.33683264e-01 -8.35749432e-02 8.84691000e-01 1.45515710e-01 2.86968946e-01 1.10774624e+00 -1.15843630e+00 -3.46059322e-01 4.68690515e-01 -8.96687746e-01 -9.68375504e-02 1.51667073e-01 -3.45304817e-01 9.98325646e-01 -8.21290195e-01 6.25923097e-01 9.22553003e-01 8.97783279e-01 1.15027897e-01 -1.62863708e+00 -4.63124514e-01 -3.40192318e-02 2.53237218e-01 -1.59907627e+00 -1.85617208e-01 4.63487327e-01 -5.04051864e-01 5.50337553e-01 4.62005138e-01 5.97807407e-01 6.94659293e-01 1.35152683e-01 5.26469827e-01 1.55543387e+00 -6.28426850e-01 6.27132237e-01 2.09881678e-01 2.17638284e-01 4.98511165e-01 2.11855307e-01 3.61256488e-02 -4.43369359e-01 -5.10494411e-01 6.07042313e-01 4.18911874e-01 -1.60824671e-01 -9.89404321e-01 -1.25147128e+00 8.73270214e-01 5.35098851e-01 4.47434902e-01 -3.11234504e-01 -1.18069045e-01 1.21584862e-01 1.17483757e-01 4.10679728e-01 1.43776551e-01 -1.72239721e-01 1.20278321e-01 -1.17310894e+00 8.72875899e-02 8.70865107e-01 4.97428179e-01 9.90506053e-01 -5.36136866e-01 2.07577363e-01 8.86921883e-01 1.51590750e-01 3.71421337e-01 4.20596778e-01 -8.81432295e-01 5.74894547e-02 7.09703445e-01 -1.06773905e-01 -1.04631448e+00 -1.87086090e-01 -4.62116420e-01 -1.31018353e+00 3.16133022e-01 4.62127030e-01 3.73791546e-01 -6.34450436e-01 1.57408452e+00 5.26493967e-01 1.79789394e-01 -3.56300324e-02 5.45270920e-01 1.68561757e-01 5.19105792e-01 -1.19744360e-01 -4.38578188e-01 1.04398942e+00 -4.47795272e-01 -4.91562076e-02 1.58139810e-01 4.75082606e-01 -5.83554327e-01 9.93304193e-01 4.70351696e-01 -4.84766394e-01 -2.78676540e-01 -6.01706147e-01 6.98260307e-01 -2.92849541e-01 -8.73136595e-02 7.28103340e-01 5.48863888e-01 -1.03483951e+00 8.21738660e-01 -6.99639678e-01 -7.80525684e-01 3.99806142e-01 3.19730043e-01 -6.35651529e-01 -1.11106232e-01 -5.20599484e-01 7.30609953e-01 4.96855736e-01 -2.63926327e-01 -1.82493567e-01 -4.38572705e-01 -2.49912322e-01 -6.89391941e-02 3.21370453e-01 -6.49767458e-01 6.17952466e-01 -7.94693530e-01 -1.15977824e+00 9.51869905e-01 -4.17361766e-01 -3.43165487e-01 2.35979110e-01 -9.08182338e-02 -3.77179921e-01 -5.24602085e-02 2.62246132e-01 2.06922993e-01 8.30891490e-01 -1.29929078e+00 -5.81549168e-01 -8.02178919e-01 -6.20940506e-01 1.03767678e-01 -2.89164990e-01 -2.36767873e-01 1.61432162e-01 -4.81327355e-01 6.59024417e-01 -1.02277124e+00 -6.28218532e-01 -2.24100918e-01 -4.79173332e-01 -4.38871622e-01 1.01643765e+00 -7.90385231e-02 1.20226598e+00 -2.09701061e+00 1.37023643e-01 8.96832883e-01 3.13157022e-01 -5.32607883e-02 3.07716638e-01 8.24766815e-01 -2.67248899e-01 -1.47258520e-01 -5.93918025e-01 5.09010494e-01 -2.68109620e-01 2.99282461e-01 -3.48478615e-01 7.15509772e-01 -1.98764116e-01 5.32609344e-01 -1.01165450e+00 -5.31993926e-01 3.94092172e-01 1.90250129e-01 -1.60983190e-01 -8.66174474e-02 1.95955887e-01 2.16871113e-01 -4.08107489e-01 3.08121294e-01 6.12336457e-01 -4.70001400e-01 2.02508569e-01 1.13026172e-01 -2.48379439e-01 -3.11987307e-02 -1.37580442e+00 1.39241362e+00 -1.66178554e-01 3.15090448e-01 -7.33389258e-02 -1.27855742e+00 1.17034388e+00 5.93231060e-02 6.57858610e-01 -4.40111049e-02 -3.13701063e-01 2.01139256e-01 -4.98211943e-02 -3.91529053e-02 1.85188413e-01 -2.68327355e-01 1.98795184e-01 7.85980761e-01 -2.84875594e-02 1.14730962e-01 1.47063300e-01 3.74009252e-01 1.32795680e+00 -1.78223908e-01 7.66674459e-01 -8.22695613e-01 4.94145602e-01 -1.42339198e-03 3.17224950e-01 6.97000504e-01 1.26294479e-01 5.08112311e-01 9.22604427e-02 -4.34427828e-01 -8.36860478e-01 -1.49435413e+00 -4.38768685e-01 1.16344070e+00 1.13565037e-02 -4.25003171e-01 -6.18446410e-01 -6.16225541e-01 1.18701011e-01 5.12802243e-01 -6.86718524e-01 -2.93788556e-02 -1.66972205e-01 -9.52978909e-01 3.09615105e-01 4.26642060e-01 4.91851687e-01 -9.09172237e-01 -4.02239114e-01 1.54745623e-01 2.88391300e-02 -5.52002192e-01 -1.87266544e-01 6.50961518e-01 -1.17072999e+00 -1.14400327e+00 -4.82737899e-01 -5.99597871e-01 6.41279578e-01 6.48025453e-01 1.11459219e+00 -3.32336903e-01 -1.85989723e-01 3.23635876e-01 -3.96433026e-01 -2.06956506e-01 -2.36711875e-01 8.58596042e-02 3.18435609e-01 1.91846460e-01 6.26309216e-01 -1.05212581e+00 -5.76247513e-01 6.32165074e-01 -7.01046705e-01 -2.22190022e-01 5.82069099e-01 8.21013033e-01 6.79648757e-01 5.74463367e-01 2.95402437e-01 -9.60474849e-01 4.90903556e-01 -6.10780776e-01 -2.81897843e-01 3.73246849e-01 -8.23949814e-01 2.10631549e-01 6.89699292e-01 -2.79138923e-01 -7.73295820e-01 4.28398192e-01 3.16256970e-01 -2.75607169e-01 -5.32863379e-01 2.49538869e-01 -2.09928706e-01 5.58713228e-02 8.90222192e-01 5.24629891e-01 1.00702852e-01 -4.05452669e-01 4.96190041e-01 7.31190979e-01 7.33701229e-01 -6.94341421e-01 8.67491066e-01 9.73321259e-01 3.14695299e-01 -1.07192802e+00 -7.00856209e-01 -1.01185739e+00 -1.00475633e+00 -1.57049388e-01 5.61098039e-01 -6.16455436e-01 -5.63515663e-01 2.13779867e-01 -6.17190003e-01 9.67250764e-02 -4.73007262e-01 3.44313025e-01 -6.53099358e-01 4.23362404e-01 -9.47436318e-03 -9.40927148e-01 -2.08323881e-01 -4.98234421e-01 6.79742336e-01 1.19931579e-01 -3.63423228e-01 -1.14052391e+00 4.34561849e-01 1.19579971e-01 1.06195547e-01 2.58274466e-01 9.72623110e-01 -7.57898211e-01 -2.44216353e-01 -2.12419316e-01 -5.74750006e-02 2.04395056e-01 3.16986084e-01 2.02253655e-01 -7.91123867e-01 -1.62726372e-01 -1.90068126e-01 3.57696228e-02 1.10845995e+00 4.96694148e-01 1.23436284e+00 -1.57803953e-01 -7.39659727e-01 4.31915522e-01 1.36227596e+00 -1.18480891e-01 5.20557284e-01 1.55027211e-01 4.47195679e-01 7.64230430e-01 3.68043542e-01 4.31405067e-01 1.55682430e-01 5.96466839e-01 2.28454694e-01 1.06995240e-01 4.52334225e-01 -3.93022627e-01 1.98232159e-01 8.41126978e-01 -5.02056599e-01 1.81897566e-01 -9.55008626e-01 4.96978074e-01 -2.22469378e+00 -1.40925717e+00 -5.25305450e-01 2.67395163e+00 6.13806248e-01 -1.29866362e-01 4.97414470e-01 5.37523150e-01 9.64223564e-01 6.81002811e-02 -2.52763897e-01 -2.78697073e-01 -3.05359334e-01 3.24152172e-01 3.95377785e-01 9.95698106e-03 -1.21703327e+00 6.54579461e-01 6.65164042e+00 1.16874754e+00 -8.64839911e-01 -8.51539746e-02 4.96057093e-01 2.34261215e-01 -1.83771312e-01 2.45104730e-01 -5.24979830e-01 5.74087441e-01 7.12699056e-01 -1.92833334e-01 4.31216568e-01 9.96150792e-01 5.63618131e-02 -3.27258557e-01 -1.00911844e+00 8.35351825e-01 -3.49811107e-01 -1.33156919e+00 -1.96698278e-01 4.21072334e-01 7.81679988e-01 7.69240260e-02 3.18545140e-02 -2.68859297e-01 7.65837491e-01 -1.14222229e+00 2.46119261e-01 4.23460513e-01 7.81775951e-01 -7.57860899e-01 4.03365910e-01 6.50471330e-01 -1.12974584e+00 2.37378967e-03 -6.37637258e-01 -2.24664211e-01 -1.43407866e-01 1.28097975e+00 -9.89900291e-01 5.90207458e-01 1.07894123e+00 8.41246009e-01 -4.28394496e-01 9.78198588e-01 -6.01224229e-03 8.02514434e-01 -6.49142385e-01 3.99641097e-02 2.14912415e-01 -7.15909123e-01 2.95226276e-01 1.22114766e+00 3.43385965e-01 -3.05512901e-02 1.80976782e-02 4.87002075e-01 3.55681300e-01 2.67370701e-01 -9.22362745e-01 1.17698371e-01 5.96771479e-01 1.35033751e+00 -1.16167390e+00 -1.65082768e-01 -2.22939759e-01 8.79569650e-01 2.70458996e-01 1.80874139e-01 -2.20006555e-01 -2.44439989e-01 5.66850543e-01 5.09340286e-01 4.11992460e-01 -1.43895090e-01 -4.98466223e-01 -9.53220129e-01 -3.42851132e-02 -5.45412779e-01 5.42159677e-01 -4.45984364e-01 -1.71430707e+00 5.04858315e-01 6.81080446e-02 -1.27778041e+00 -2.20025212e-01 -5.73268116e-01 -8.62142980e-01 5.48148453e-01 -6.92565441e-01 -9.64133978e-01 -1.84695005e-01 8.88966799e-01 -3.86905372e-02 -2.56403923e-01 1.09640396e+00 -4.83256012e-01 -1.29777431e-01 4.86018248e-02 9.22819734e-01 -1.12013958e-01 5.54962575e-01 -1.46575725e+00 3.09912506e-02 6.48269475e-01 6.60431266e-01 8.12150002e-01 6.56058550e-01 -4.54477191e-01 -9.35681045e-01 -1.07992721e+00 7.16096938e-01 -5.37054718e-01 6.57006085e-01 -1.97838083e-01 -8.85051370e-01 2.27460891e-01 3.04240976e-02 2.14306429e-01 1.09540129e+00 6.06596887e-01 -6.62492394e-01 -2.43706182e-01 -1.13609302e+00 3.85143101e-01 1.12609160e+00 -4.08497214e-01 -6.27846718e-01 3.32321346e-01 -9.24429893e-02 4.44170713e-01 -9.26811397e-01 4.08393443e-01 7.46330559e-01 -1.52455902e+00 9.96021986e-01 -4.46830660e-01 3.27584386e-01 -4.49722856e-01 -3.92059803e-01 -1.43352950e+00 -8.00817847e-01 -3.64164561e-01 -2.97654644e-02 1.07422626e+00 1.05926499e-01 -7.40447223e-01 8.88850868e-01 6.26896173e-02 3.58956218e-01 -9.10756648e-01 -1.06573856e+00 -9.87689555e-01 1.25647992e-01 -2.83210158e-01 4.93450671e-01 1.09403646e+00 1.50527492e-01 3.05462211e-01 -1.45374626e-01 5.52523471e-02 1.03564847e+00 5.53353250e-01 8.25542152e-01 -1.94088972e+00 -2.94154823e-01 -3.40203106e-01 -6.58147693e-01 -7.21812487e-01 9.91308391e-02 -1.00909853e+00 -1.58694223e-01 -1.26159441e+00 5.25141239e-01 -8.84671330e-01 -3.46987486e-01 3.05870444e-01 -1.32557869e-01 2.13926330e-01 -6.12647831e-02 6.44628167e-01 -7.51013696e-01 2.71856219e-01 7.44985044e-01 2.68403053e-01 -3.17728400e-01 5.36982775e-01 -6.44623816e-01 8.53225946e-01 7.20853448e-01 -4.91458297e-01 -2.72298545e-01 3.44609112e-01 8.96885172e-02 -2.58378446e-01 3.57274145e-01 -1.04073036e+00 1.70311972e-01 -3.22224021e-01 4.31841820e-01 -6.09850168e-01 5.61418347e-02 -9.84774530e-01 5.73860765e-01 4.15495336e-01 -1.19427182e-01 -1.29042819e-01 -5.78081846e-01 8.15737545e-01 -1.07380465e-01 -2.28120387e-01 7.87051737e-01 -2.06399426e-01 -5.80194652e-01 1.65221274e-01 -4.11147743e-01 -9.09903347e-02 1.20725870e+00 -8.11768353e-01 -1.27665505e-01 -2.89174169e-01 -6.56227648e-01 -1.30245492e-01 9.03885424e-01 1.10546872e-01 4.36076462e-01 -1.21399581e+00 -6.08154535e-01 2.68720001e-01 3.26353848e-01 2.16419511e-02 -1.10911742e-01 9.00400996e-01 -1.70090377e-01 2.51201093e-01 -8.30455199e-02 -1.03952146e+00 -1.34504294e+00 5.15454948e-01 -1.58956513e-01 -3.18492830e-01 -7.15928078e-01 5.73807955e-01 2.65498370e-01 -7.08106697e-01 -1.36286914e-01 2.56481200e-01 1.48074985e-01 -1.33314729e-01 3.48909497e-01 6.54272735e-01 8.85567665e-02 -6.39578223e-01 -5.63017368e-01 5.42271554e-01 -6.17775507e-02 3.76067944e-02 1.56627572e+00 1.07374370e-01 -4.34207052e-01 7.54730821e-01 8.50313962e-01 3.79108787e-02 -9.23590064e-01 -3.80201072e-01 4.76889104e-01 -4.76935178e-01 -3.05795550e-01 -5.35173118e-01 -6.28205121e-01 7.27723300e-01 3.71591657e-01 4.80961174e-01 1.15380728e+00 4.50896114e-01 6.47520944e-02 4.55831289e-01 6.96936667e-01 -1.03353369e+00 -8.88099000e-02 3.66661400e-01 4.87157881e-01 -1.04554772e+00 2.03482434e-01 -5.97223341e-01 -7.61455536e-01 1.03413415e+00 8.58392641e-02 -2.62438148e-01 8.97756636e-01 3.21345508e-01 -6.55670762e-02 -2.48168912e-02 -7.62885332e-01 -2.63504773e-01 1.79144442e-01 9.02521431e-01 4.64622498e-01 3.49427760e-01 -2.50853032e-01 1.96106434e-01 -3.53267312e-01 -7.52918273e-02 1.31332144e-01 6.83630228e-01 -7.43792593e-01 -1.12928092e+00 -5.65585434e-01 9.30617154e-01 -9.64528620e-02 1.13278896e-01 -5.83766639e-01 5.98498166e-01 1.72506809e-01 7.12838709e-01 1.10724762e-01 -4.91132408e-01 -1.88725799e-01 3.09556246e-01 3.77848089e-01 -4.90982056e-01 -4.15248930e-01 -6.21806365e-03 -3.87899160e-01 -5.09445429e-01 -7.89867818e-01 -8.57957363e-01 -1.07943439e+00 -4.26650941e-01 -4.58070993e-01 5.20980537e-01 4.50081468e-01 9.17536378e-01 3.40164721e-01 -2.59488136e-01 8.57280552e-01 -6.25710607e-01 -6.17359042e-01 -8.04577291e-01 -1.16018713e+00 3.73215109e-01 -2.56348819e-01 -6.48755491e-01 -3.58704180e-01 3.83074172e-02]

[7.589538097381592, 4.533164978027344]

9d60fb4d-b3ca-41c8-a349-93cbe11d23a3

bias-and-fairness-on-multimodal-emotion

2205.08383

https://arxiv.org/abs/2205.08383v1

https://arxiv.org/pdf/2205.08383v1.pdf

Bias and Fairness on Multimodal Emotion Detection Algorithms

Numerous studies have shown that machine learning algorithms can latch onto protected attributes such as race and gender and generate predictions that systematically discriminate against one or more groups. To date the majority of bias and fairness research has been on unimodal models. In this work, we explore the biases that exist in emotion recognition systems in relationship to the modalities utilized, and study how multimodal approaches affect system bias and fairness. We consider audio, text, and video modalities, as well as all possible multimodal combinations of those, and find that text alone has the least bias, and accounts for the majority of the models' performances, raising doubts about the worthiness of multimodal emotion recognition systems when bias and fairness are desired alongside model performance.

['Jimi Cao', 'Rehan Ahmed', 'Matheus Schmitz']

2022-05-11

['multimodal-emotion-recognition', 'multimodal-emotion-recognition']

['computer-vision', 'speech']

[ 1.58548132e-01 1.40559733e-01 -5.44185221e-01 -7.43758678e-01 -2.42362767e-01 -5.85042894e-01 7.30937541e-01 3.72057974e-01 -6.31328881e-01 6.88291490e-01 5.05538821e-01 -5.21662176e-01 -6.37812614e-02 -3.49478126e-01 -6.79351389e-02 -3.76764417e-01 3.15087050e-01 2.30360925e-02 -5.35437882e-01 -3.30917239e-01 5.37882805e-01 3.98573130e-01 -1.68242025e+00 4.19019163e-01 7.96543479e-01 1.02087641e+00 -1.02269471e+00 6.30188525e-01 -1.43212229e-01 1.07034469e+00 -7.08333671e-01 -1.07580054e+00 -1.16098151e-01 -5.95852196e-01 -6.66467965e-01 -4.24471676e-01 7.96591461e-01 -3.89899611e-01 -1.84514433e-01 8.35253596e-01 9.73778009e-01 -9.21932608e-02 8.64920616e-01 -1.63056684e+00 -5.18362284e-01 6.55986667e-01 -4.43564862e-01 3.60129029e-02 6.07786834e-01 1.46490067e-01 8.77077341e-01 -6.12884104e-01 5.12240946e-01 1.55821097e+00 6.63551867e-01 7.99996376e-01 -1.35328627e+00 -9.77199018e-01 -1.79580487e-02 -3.36459465e-02 -1.12494683e+00 -1.08370912e+00 6.45549655e-01 -5.03845334e-01 5.41762054e-01 5.68540454e-01 5.84201276e-01 1.22856581e+00 2.73879737e-01 4.12279636e-01 1.42372060e+00 -4.45277214e-01 1.38167784e-01 5.92511952e-01 1.78465948e-01 4.44645733e-01 2.64873743e-01 2.33235732e-01 -8.11279416e-01 -5.99525809e-01 1.51182830e-01 -4.81624275e-01 -6.14403514e-03 -1.00006297e-01 -9.62969661e-01 9.29667354e-01 7.35933660e-03 1.64144940e-03 -1.98204681e-01 -2.61502303e-02 6.35858953e-01 2.96848357e-01 4.72894877e-01 6.95933759e-01 -1.19303018e-01 -4.10410881e-01 -1.04343092e+00 1.84963122e-01 7.04942942e-01 1.64856806e-01 4.43315864e-01 1.32270038e-01 -1.37742579e-01 1.05085671e+00 1.69011608e-01 6.23654187e-01 2.54168630e-01 -1.29040504e+00 -7.26190209e-03 5.08400500e-01 9.87201259e-02 -1.52122927e+00 -5.10518491e-01 1.58482239e-01 -5.02215803e-01 4.40922916e-01 5.86640656e-01 -4.95225698e-01 -4.42771286e-01 1.86682808e+00 8.76618326e-02 -6.37037694e-01 -8.25532302e-02 1.02308738e+00 9.77494538e-01 2.19166249e-01 6.98935509e-01 -1.01102382e-01 1.06034100e+00 -2.12933999e-02 -6.39091730e-01 -3.85795981e-01 5.49588442e-01 -1.00514174e+00 8.70832324e-01 3.81671697e-01 -1.22522283e+00 -2.48996943e-01 -7.38919556e-01 5.19679636e-02 -4.03646827e-01 -1.31488554e-02 7.94483423e-01 1.53721321e+00 -7.62433410e-01 4.57587093e-01 -7.68189952e-02 -4.65079576e-01 3.53284478e-01 2.28805706e-01 -3.70916039e-01 7.32134208e-02 -1.34139633e+00 1.26811063e+00 -9.63727310e-02 1.68247491e-01 -2.77108759e-01 -4.52089518e-01 -7.14928389e-01 -1.88113227e-01 -6.03919029e-02 -5.51096678e-01 8.84993792e-01 -1.82359695e+00 -1.24447882e+00 9.47307467e-01 -3.03597599e-01 1.40279792e-02 4.49257731e-01 1.34130061e-01 -6.73999548e-01 3.81910279e-02 -2.79157102e-01 1.11057520e+00 7.36130476e-01 -1.32575095e+00 -4.12672460e-01 -3.41592997e-01 7.81082213e-02 3.83856714e-01 -4.30811763e-01 5.90998650e-01 4.98830914e-01 -2.19748363e-01 -2.60907710e-01 -8.20530176e-01 1.17798805e-01 -3.58721279e-02 -1.27712667e-01 5.97590860e-03 4.30834949e-01 -7.57675827e-01 1.36267960e+00 -2.17116451e+00 -8.38390440e-02 4.33835149e-01 -1.23121262e-01 -1.27103269e-01 -4.90978919e-02 3.35424900e-01 -2.06631690e-01 5.96494794e-01 2.43427753e-01 -9.35956538e-02 4.64345932e-01 9.63056460e-02 -1.61293730e-01 5.80838442e-01 2.56820798e-01 6.69059217e-01 -3.50394100e-01 -8.59258413e-01 2.49415874e-01 5.61771750e-01 -5.90677023e-01 -2.45757714e-01 3.73977900e-01 1.07999533e-01 1.60161361e-01 8.14299285e-01 5.23522079e-01 4.23177242e-01 2.19170570e-01 -1.55799553e-01 -1.11389965e-01 1.56580925e-01 -8.10303986e-01 7.98386753e-01 -7.24496394e-02 9.45710719e-01 3.38510096e-01 -5.74334562e-01 9.48227167e-01 2.80284792e-01 3.68955076e-01 -7.47923553e-01 5.97865343e-01 1.54239774e-01 4.40811634e-01 -5.75313807e-01 8.07575107e-01 -6.76214755e-01 -2.45001107e-01 5.00126958e-01 -1.48552895e-01 -2.19362468e-01 -2.12235630e-01 2.15007469e-01 4.43522811e-01 -2.90730953e-01 1.29699290e-01 -1.32291585e-01 1.19254619e-01 -6.54669628e-02 4.10011292e-01 7.70612180e-01 -7.74172783e-01 5.12738705e-01 8.10433328e-01 -1.86084330e-01 -7.57713318e-01 -6.93232417e-01 -3.51329297e-01 1.58582520e+00 2.62628943e-02 -9.66389291e-03 -5.72373271e-01 -5.03613591e-01 2.36940324e-01 1.02098882e+00 -8.16534281e-01 -5.38308799e-01 3.00057173e-01 -8.13843012e-01 9.13182557e-01 2.74341226e-01 -1.45971462e-01 -7.30498850e-01 -8.70012105e-01 -4.98146325e-01 -2.56975651e-01 -7.96395421e-01 -7.69718513e-02 -1.81972370e-01 -7.09381521e-01 -9.38838422e-01 -4.37220305e-01 -6.12810776e-02 3.38241637e-01 -1.78833246e-01 1.19315720e+00 9.10811685e-03 -1.42658770e-01 8.33431959e-01 -2.27123395e-01 -7.70806551e-01 -3.27378005e-01 -3.14153820e-01 1.39922332e-02 7.98744261e-02 5.64601243e-01 -5.03737060e-03 -3.89553487e-01 2.83096910e-01 -6.64170980e-01 -1.05339035e-01 3.19381982e-01 6.53725386e-01 -2.95612544e-01 -2.46298015e-01 7.31904268e-01 -7.01040030e-01 1.10005045e+00 -4.98428196e-01 1.92279249e-01 1.70910776e-01 -7.73743927e-01 -5.67388654e-01 -1.28639162e-01 -4.44912791e-01 -1.12516260e+00 -2.03331098e-01 2.07832158e-01 -4.10030521e-02 -4.32115227e-01 5.34003496e-01 -6.00186437e-02 -3.46861035e-01 8.49067867e-01 -4.22234029e-01 2.92542607e-01 1.11658826e-01 2.43827134e-01 8.49254727e-01 1.75170988e-01 -8.04639816e-01 5.48710525e-02 2.96202779e-01 -2.82731652e-01 -9.02175486e-01 -2.90079266e-01 1.34395778e-01 -2.60478973e-01 -9.37120557e-01 7.06809640e-01 -6.94101751e-01 -1.08912301e+00 2.23252669e-01 -7.56571949e-01 -3.99684310e-02 4.42199446e-02 6.77555561e-01 -2.53328681e-01 1.39491588e-01 -3.78632814e-01 -1.51550305e+00 -1.21904887e-01 -8.63267839e-01 6.94352090e-01 3.54191482e-01 -1.08880281e+00 -7.66463101e-01 -1.39470726e-01 5.28873444e-01 6.34549260e-01 2.58018017e-01 1.03139055e+00 -6.59064412e-01 5.54294825e-01 -4.92797107e-01 -3.82805288e-01 9.67324153e-02 -1.98290914e-01 4.91585523e-01 -1.18652284e+00 8.88040662e-03 -4.94849831e-01 -7.42548823e-01 6.18737459e-01 4.93139684e-01 8.12932432e-01 -1.16953999e-01 9.02749673e-02 -8.63392726e-02 9.50196564e-01 1.52235717e-01 5.98994732e-01 1.54929116e-01 2.82544732e-01 1.26028192e+00 5.06653428e-01 5.64807236e-01 4.89805430e-01 2.33249828e-01 4.40949321e-01 -2.20053658e-01 4.37030733e-01 -7.38539174e-02 4.55805242e-01 1.21462792e-01 -1.08033329e-01 4.92087603e-02 -9.95006323e-01 3.17506671e-01 -1.48543274e+00 -1.23244524e+00 4.07531410e-02 2.13375807e+00 6.35910749e-01 -5.56728803e-03 4.93855596e-01 2.54060209e-01 8.75640631e-01 3.12704593e-01 -3.71551514e-01 -1.21365786e+00 -4.55610275e-01 -5.75779937e-02 2.77391911e-01 5.33536911e-01 -8.92414749e-01 4.72434521e-01 7.89714479e+00 3.94475341e-01 -1.21192348e+00 -3.97612780e-01 1.19063544e+00 -5.82083404e-01 -5.62816203e-01 -6.49811178e-02 -1.05871506e-01 2.42112324e-01 1.03963649e+00 -2.01679111e-01 4.17641312e-01 5.00294805e-01 4.31594729e-01 -5.54262817e-01 -1.10085166e+00 8.44528019e-01 4.01962936e-01 -4.59689140e-01 -2.26693638e-02 -3.14243697e-02 4.28604394e-01 -5.01265705e-01 4.17716563e-01 3.18087280e-01 4.40598466e-02 -1.57469070e+00 1.06381571e+00 5.91520607e-01 6.35565996e-01 -8.11105490e-01 7.89498866e-01 -1.62644610e-02 -9.93991196e-02 -2.05144599e-01 -1.17919803e-01 -4.57519114e-01 -2.14597322e-02 3.12588334e-01 -3.73463333e-01 1.35299936e-01 6.99577570e-01 2.38626748e-01 -6.25233948e-01 6.45146549e-01 1.95456475e-01 7.02855885e-01 -2.35655680e-01 -2.77669370e-01 -5.72720617e-02 8.58340114e-02 2.02002242e-01 1.43927622e+00 2.19617665e-01 1.13675207e-01 -3.16806197e-01 4.38196361e-01 3.50557640e-02 4.83959317e-01 -6.75948977e-01 -4.37161803e-01 4.81096536e-01 1.10120142e+00 -3.97627711e-01 -2.79730827e-01 -3.83657813e-01 5.03401339e-01 -9.76111665e-02 4.25180912e-01 -6.32560134e-01 -2.62824416e-01 7.35658228e-01 -1.97879419e-01 -5.07079482e-01 1.56522766e-01 -9.25738633e-01 -8.93567801e-01 -5.26440859e-01 -1.44340622e+00 6.15771830e-01 -9.80697393e-01 -1.36677945e+00 8.54550302e-02 -1.93955109e-01 -7.31309474e-01 -2.53451541e-02 -4.89676625e-01 -4.32869315e-01 9.76115763e-01 -1.06025457e+00 -8.85830879e-01 -1.13894634e-01 4.15936798e-01 -2.33586177e-01 -1.34958148e-01 9.33290184e-01 2.64650285e-01 -5.79221129e-01 7.50328183e-01 -3.39049220e-01 7.97204673e-02 1.41917717e+00 -8.27643216e-01 -6.75792456e-01 1.95568457e-01 -2.72423714e-01 5.71606934e-01 9.89067733e-01 -3.50231588e-01 -1.02655411e+00 -2.06786484e-01 1.06685317e+00 -5.61448932e-01 3.08584839e-01 1.46217331e-01 -4.50247645e-01 3.15971494e-01 4.89266038e-01 -6.28305674e-01 1.28666687e+00 5.85792482e-01 -7.41739273e-01 -1.49400039e-02 -1.37142241e+00 7.58307159e-01 3.84893209e-01 -7.94806004e-01 -3.14594150e-01 -4.85762328e-01 -1.37317255e-01 -2.29597032e-01 -9.32687223e-01 4.57121313e-01 1.38426352e+00 -1.38640738e+00 7.92521358e-01 -1.06759667e+00 9.53249216e-01 2.39632607e-01 -3.48326355e-01 -1.26958966e+00 -7.71916956e-02 -4.87921424e-02 4.81181145e-01 1.45138502e+00 6.00153506e-01 -7.28576839e-01 6.93934798e-01 1.54737902e+00 4.31037784e-01 -4.70967293e-01 -8.92324567e-01 -4.21455503e-02 3.43112320e-01 -6.40191555e-01 6.55225575e-01 1.49328804e+00 6.16810083e-01 2.76632965e-01 -6.14739716e-01 -1.81231886e-01 4.22104478e-01 -2.73183659e-02 7.46761739e-01 -1.10942328e+00 2.22940177e-01 -1.10493791e+00 -5.26505053e-01 -2.71462109e-02 3.64839613e-01 -4.72814143e-01 -2.93586135e-01 -9.98470306e-01 3.42095405e-01 -2.53537744e-01 -2.21828163e-01 4.61437434e-01 -2.21329525e-01 4.79898989e-01 3.97878766e-01 -1.30235672e-01 -3.14871192e-01 3.54318529e-01 7.49922514e-01 -7.82611966e-02 7.88241625e-04 -3.73354435e-01 -1.37277865e+00 7.32413530e-01 9.50605154e-01 -1.40653342e-01 -7.58567229e-02 -3.27713162e-01 5.46522915e-01 7.21203387e-02 5.23444295e-01 -5.17649055e-01 -5.70957661e-02 -5.42810500e-01 8.26309621e-01 1.11284733e-01 6.34513378e-01 -8.40204775e-01 4.00310233e-02 1.66903555e-01 -8.30764174e-01 3.21094692e-03 4.97129917e-01 1.41232088e-01 -2.53737867e-01 -1.67062461e-01 7.03445792e-01 2.12503180e-01 -2.49011680e-01 -3.65317911e-01 -6.30078435e-01 1.11887589e-01 7.03717470e-01 -2.90362984e-01 -5.75923264e-01 -1.04125702e+00 -7.23686635e-01 2.62716144e-01 5.49675822e-01 5.72456241e-01 2.74467677e-01 -1.34942186e+00 -6.83297575e-01 -1.32196739e-01 1.16364323e-01 -1.17141259e+00 3.15127641e-01 1.10344601e+00 -6.84900135e-02 1.20312534e-01 -5.64357042e-01 -2.48028442e-01 -1.66515732e+00 2.47470513e-01 5.14944375e-01 6.34169042e-01 5.58192730e-01 6.49962187e-01 -2.37239182e-01 -4.05687869e-01 2.47493505e-01 4.71634269e-01 -3.85019749e-01 6.29216373e-01 2.69175529e-01 6.18882775e-01 -3.74556959e-01 -1.13897347e+00 -4.60586667e-01 3.69519085e-01 3.95566106e-01 -5.36118209e-01 8.14920962e-01 -2.95524448e-01 -2.75610983e-01 7.14412808e-01 7.64289141e-01 8.28843117e-02 -4.93143409e-01 3.10805172e-01 -2.45748818e-01 -8.70186746e-01 9.91733372e-02 -1.07956266e+00 -7.48774529e-01 1.23143244e+00 7.19284713e-01 3.54858726e-01 1.11333907e+00 -3.04092079e-01 7.11770132e-02 -5.23766726e-02 4.44845185e-02 -1.38050008e+00 -3.51582825e-01 1.72385037e-01 5.87333798e-01 -1.37342978e+00 1.91802979e-01 -4.99965921e-02 -1.22190166e+00 1.01277244e+00 6.33357644e-01 3.36410731e-01 4.15307313e-01 -6.89426735e-02 4.38014567e-01 1.94817316e-02 -8.32924604e-01 -2.87910439e-02 2.46188983e-01 6.62305593e-01 1.10331547e+00 3.69767189e-01 -6.74973071e-01 7.96204031e-01 -2.69542545e-01 8.08924958e-02 4.14691478e-01 6.21447027e-01 -3.44200611e-01 -8.30418944e-01 -8.31375897e-01 6.85384512e-01 -7.78504789e-01 5.42727634e-02 -1.24963522e+00 5.74237585e-01 2.28006423e-01 1.31750274e+00 3.15798335e-02 -5.91967344e-01 3.91095519e-01 5.83525658e-01 3.83386225e-01 5.21115586e-02 -9.81401026e-01 -1.97104439e-01 7.47888029e-01 -4.85572815e-01 -5.43986678e-01 -9.33227241e-01 -6.74901545e-01 -9.42140162e-01 -1.92869380e-01 3.96266915e-02 6.74228132e-01 6.88253939e-01 2.12579429e-01 -4.53381762e-02 4.27716583e-01 -9.37863886e-01 -3.70657533e-01 -8.27737689e-01 -3.88592243e-01 5.38590014e-01 1.73053712e-01 -4.84381080e-01 -5.41411281e-01 -2.14220837e-01]

[13.003338813781738, 1.4409419298171997]

df07ea95-145d-47f5-9b3e-bf6468d14b86

eeg-decoding-for-datasets-with-heterogenous

2306.13109

https://arxiv.org/abs/2306.13109v1

https://arxiv.org/pdf/2306.13109v1.pdf

EEG Decoding for Datasets with Heterogenous Electrode Configurations using Transfer Learning Graph Neural Networks

Brain-Machine Interfacing (BMI) has greatly benefited from adopting machine learning methods for feature learning that require extensive data for training, which are often unavailable from a single dataset. Yet, it is difficult to combine data across labs or even data within the same lab collected over the years due to the variation in recording equipment and electrode layouts resulting in shifts in data distribution, changes in data dimensionality, and altered identity of data dimensions. Our objective is to overcome this limitation and learn from many different and diverse datasets across labs with different experimental protocols. To tackle the domain adaptation problem, we developed a novel machine learning framework combining graph neural networks (GNNs) and transfer learning methodologies for non-invasive Motor Imagery (MI) EEG decoding, as an example of BMI. Empirically, we focus on the challenges of learning from EEG data with different electrode layouts and varying numbers of electrodes. We utilise three MI EEG databases collected using very different numbers of EEG sensors (from 22 channels to 64) and layouts (from custom layouts to 10-20). Our model achieved the highest accuracy with lower standard deviations on the testing datasets. This indicates that the GNN-based transfer learning framework can effectively aggregate knowledge from multiple datasets with different electrode layouts, leading to improved generalization in subject-independent MI EEG classification. The findings of this study have important implications for Brain-Computer-Interface (BCI) research, as they highlight a promising method for overcoming the limitations posed by non-unified experimental setups. By enabling the integration of diverse datasets with varying electrode layouts, our proposed approach can help advance the development and application of BMI technologies.

['A. Aldo Faisal', 'Xiaoxi Wei', 'Jinpei Han']

2023-06-20

['eeg-decoding', 'transfer-learning', 'eeg-decoding']

['medical', 'miscellaneous', 'time-series']

[ 4.58698183e-01 -4.52290565e-01 1.80134535e-01 -3.05525780e-01 -4.78576571e-01 -5.74095786e-01 5.42580858e-02 9.22064483e-02 -5.71622014e-01 1.05717874e+00 -2.04260185e-01 -2.87410647e-01 -7.87255585e-01 -4.88988638e-01 -9.25418437e-01 -6.29032075e-01 -4.15647119e-01 2.04996616e-01 -1.06118537e-01 -7.28173628e-02 4.24245358e-01 6.27484441e-01 -1.53632092e+00 3.60908568e-01 9.66687143e-01 1.07677555e+00 4.23797369e-01 3.10519964e-01 1.47661105e-01 -6.26501217e-02 -8.86215508e-01 3.90751660e-02 3.02182764e-01 -2.77329236e-01 -3.64426702e-01 -3.45364690e-01 3.33959699e-01 5.42677231e-02 -1.20596781e-01 9.05570149e-01 8.86301994e-01 -8.98475125e-02 8.04177463e-01 -1.45840514e+00 -5.63123822e-01 5.70536375e-01 -4.89850819e-01 2.93238550e-01 1.98151648e-01 1.49156287e-01 3.20999324e-01 -3.15373302e-01 3.36364031e-01 6.46592796e-01 6.94643795e-01 5.11489809e-01 -1.64324486e+00 -1.28823304e+00 -3.98947224e-02 5.57913899e-01 -1.61496365e+00 -5.52918240e-02 7.48709440e-01 -6.24581337e-01 9.81983542e-01 1.74753442e-01 1.07332993e+00 1.53964818e+00 5.51870108e-01 6.89346716e-02 1.47546244e+00 -2.48078778e-01 4.79310662e-01 3.06487113e-01 8.91860053e-02 -6.76607564e-02 6.96616292e-01 -1.32347912e-01 -8.27384651e-01 -3.91351618e-02 6.76016748e-01 1.10026307e-01 -5.01484871e-01 -5.35238206e-01 -1.25920713e+00 5.28523803e-01 5.04838943e-01 5.95719516e-01 -3.43600392e-01 -1.64719552e-01 3.50849152e-01 5.90337753e-01 -9.68010873e-02 8.10755253e-01 -8.89325261e-01 -3.57185215e-01 -8.53039682e-01 -1.07690170e-01 9.42064703e-01 1.12251043e+00 7.63311863e-01 7.63508230e-02 2.05566153e-01 8.33607078e-01 -5.99984974e-02 3.50249171e-01 7.38980770e-01 -3.19779545e-01 6.24637783e-01 6.49438262e-01 -2.75537223e-01 -1.01073980e+00 -8.08657169e-01 -5.18649995e-01 -1.06097341e+00 2.03161135e-01 2.60579675e-01 -4.63202894e-01 -6.71963632e-01 1.78414321e+00 -2.87047327e-01 1.68621734e-01 -2.73123980e-01 6.83223963e-01 4.92033899e-01 -1.99519768e-02 1.63336679e-01 -4.92835678e-02 1.11236930e+00 -2.84427226e-01 -4.61099625e-01 -7.48585612e-02 6.61186934e-01 -1.28932938e-01 1.13209021e+00 8.96989524e-01 -5.79856098e-01 -4.58919615e-01 -1.42344749e+00 4.24320966e-01 -8.95272911e-01 -3.92109016e-03 7.24932969e-01 9.68306184e-01 -1.10929382e+00 6.07129216e-01 -6.54373050e-01 -5.70578754e-01 7.61643171e-01 1.13228142e+00 -6.65571213e-01 3.09150070e-02 -9.96698380e-01 8.69468272e-01 6.05919242e-01 9.56368968e-02 -3.97595823e-01 -9.93811131e-01 -3.95655513e-01 -6.76789507e-02 -1.09024107e-01 -6.23899519e-01 2.02827320e-01 -9.35131073e-01 -1.56766200e+00 4.13264513e-01 3.78402293e-01 -1.60948157e-01 6.78955987e-02 -4.75654453e-02 -5.78767121e-01 -2.51388282e-01 -2.28707150e-01 5.40521562e-01 6.82214081e-01 -1.01489198e+00 -1.38184234e-01 -8.49180222e-01 -4.21725094e-01 -2.12250903e-01 -8.13582063e-01 -2.32372418e-01 2.22681373e-01 -3.48009557e-01 -2.03547155e-04 -8.53106737e-01 2.48128235e-01 -4.90445971e-01 -2.26886030e-02 3.59542727e-01 5.40003955e-01 -6.60839856e-01 1.02372253e+00 -2.07533169e+00 5.14358222e-01 6.00393891e-01 2.08537012e-01 1.76321834e-01 -1.97870150e-01 3.31994563e-01 -3.78893495e-01 7.70741478e-02 1.06239263e-02 3.09927076e-01 -2.21636236e-01 3.82670537e-02 3.11548978e-01 3.53526741e-01 6.07094094e-02 8.29352736e-01 -6.43127620e-01 -6.48448616e-02 2.81707823e-01 4.79053974e-01 -4.16794986e-01 9.69150439e-02 4.12945986e-01 1.10560036e+00 -9.04167667e-02 4.36987251e-01 7.33760595e-01 -5.13441376e-02 1.81774557e-01 -6.03722453e-01 1.20094925e-01 -4.95352373e-02 -1.18072248e+00 2.04469395e+00 -5.12213349e-01 5.64729273e-01 -1.37497023e-01 -1.34652364e+00 1.05063784e+00 1.71624094e-01 7.57785320e-01 -9.55880642e-01 4.28482205e-01 2.67622143e-01 6.12225950e-01 -6.20212495e-01 -2.54807085e-01 3.78722586e-02 1.91835314e-01 1.89393789e-01 5.82887173e-01 -4.12469804e-02 -1.85255017e-02 -3.58198702e-01 1.21027768e+00 -1.28536746e-01 7.15862960e-02 -4.60981667e-01 2.19350308e-01 -3.32092375e-01 2.90343314e-01 7.34476328e-01 -1.08425289e-01 3.26937139e-01 3.61119121e-01 -2.00507656e-01 -8.09939921e-01 -1.02667105e+00 -5.85452616e-01 9.03684974e-01 -5.52156456e-02 -2.21569553e-01 -7.95291841e-01 -1.63665414e-01 9.43484232e-02 4.60309684e-01 -6.31571531e-01 -4.76068169e-01 -1.42128825e-01 -9.81061935e-01 6.16030812e-01 5.04799068e-01 3.17829281e-01 -8.53277683e-01 -6.34480655e-01 2.39231840e-01 2.06619725e-01 -9.76396084e-01 1.96152195e-01 7.99554110e-01 -1.18412578e+00 -1.02649224e+00 -5.56261480e-01 -6.77519858e-01 4.85641927e-01 -1.48074344e-01 7.83505619e-01 -2.99553066e-01 -4.21782166e-01 3.67363125e-01 -2.56780177e-01 -7.93253481e-01 1.74129661e-02 4.32940483e-01 3.72202128e-01 -1.10530742e-01 6.44613147e-01 -1.25374985e+00 -7.52026677e-01 2.81184584e-01 -9.12045836e-01 -1.68522224e-02 8.50623012e-01 8.81131351e-01 3.47923905e-01 6.75085559e-02 1.04853785e+00 -5.84482253e-01 9.75053847e-01 -7.68902361e-01 -3.51941288e-01 3.90739501e-01 -7.32975900e-01 -1.44395670e-02 6.46048844e-01 -8.54470789e-01 -4.89007264e-01 -2.91774958e-01 2.46312261e-01 -6.19165748e-02 -5.66293836e-01 5.29429734e-01 -5.05317688e-01 -5.88458359e-01 9.12745059e-01 -2.28434708e-03 2.98749451e-02 -3.00025940e-01 5.40708601e-02 1.01266742e+00 3.04095000e-01 -5.92518866e-01 2.94829309e-01 -3.08972877e-02 1.31362587e-01 -7.32218981e-01 6.86972663e-02 -1.17467299e-01 -1.03639984e+00 -2.26415217e-01 7.09655285e-01 -8.34773779e-01 -6.42801821e-01 6.63768828e-01 -6.26149178e-01 -4.35023665e-01 2.11637333e-01 8.90940130e-01 -3.87452364e-01 -2.01823674e-02 -3.32673371e-01 -3.94018203e-01 -4.54395771e-01 -1.12281716e+00 5.65926552e-01 2.79963940e-01 -4.45429176e-01 -8.86772513e-01 -1.07938230e-01 1.94800705e-01 7.13965535e-01 3.62011075e-01 1.23170197e+00 -6.99032485e-01 -1.86318070e-01 -2.56515235e-01 -9.14468020e-02 3.18786919e-01 6.44189358e-01 -4.69224721e-01 -8.69842649e-01 -6.23144209e-01 4.29303311e-02 -3.16258818e-01 2.43850291e-01 2.16303617e-01 1.37122881e+00 1.06835634e-01 -3.80695492e-01 7.91437685e-01 1.44780266e+00 4.43963677e-01 8.05273533e-01 6.86384499e-01 7.70449221e-01 3.53040665e-01 -1.15083918e-01 2.40493596e-01 1.07679330e-01 6.52348161e-01 2.60365605e-01 7.58120716e-02 1.95496455e-01 1.61571950e-01 2.03491032e-01 8.49905133e-01 -2.42629677e-01 3.68802957e-02 -9.13508952e-01 3.12613785e-01 -1.43876481e+00 -4.80393738e-01 -2.04158481e-03 2.52021337e+00 7.21525729e-01 -1.05672576e-01 1.52403414e-01 2.84276545e-01 4.93637025e-01 -8.18268836e-01 -8.79590333e-01 -3.60019267e-01 -1.17961064e-01 7.73489296e-01 4.80333090e-01 -3.14739227e-01 -6.08551264e-01 3.98139298e-01 6.00671053e+00 4.54872608e-01 -1.62985408e+00 1.38135627e-01 2.55621284e-01 -3.69653881e-01 2.44509488e-01 -4.35145259e-01 -4.68745470e-01 7.10443318e-01 1.38950384e+00 -2.31508240e-02 1.07339227e+00 3.67869914e-01 8.47719163e-02 -1.30943075e-01 -1.43802536e+00 1.44788659e+00 1.34719178e-01 -8.85697186e-01 -5.03877476e-02 2.25212231e-01 6.51189268e-01 3.48844320e-01 -6.94527701e-02 2.13753834e-01 -4.17040169e-01 -1.20304406e+00 3.61144185e-01 6.26782179e-01 9.11256552e-01 -6.23693049e-01 7.27158308e-01 2.92607099e-01 -7.36672938e-01 -4.07132894e-01 -3.38590473e-01 -2.89061546e-01 -4.04776216e-01 3.33112359e-01 -6.44350111e-01 6.35623872e-01 1.08553684e+00 6.05135918e-01 -8.05744112e-01 1.20557880e+00 3.24626952e-01 3.82786393e-01 -3.94533008e-01 -2.08438933e-01 -4.89664912e-01 -4.67554688e-01 3.36569697e-02 1.10497093e+00 7.81547606e-01 -1.89349070e-01 -4.71429408e-01 1.05865490e+00 1.05018601e-01 1.30162761e-01 -8.01444232e-01 7.27915913e-02 4.16748703e-01 1.32135844e+00 -5.84485590e-01 1.86904758e-01 -6.19009912e-01 7.77363479e-01 4.94562894e-01 6.32558942e-01 -7.12696612e-01 -5.41048884e-01 5.07239223e-01 1.33670792e-01 1.04772449e-01 -3.06417286e-01 -7.40126014e-01 -1.07081044e+00 1.38178587e-01 -8.74368370e-01 1.51822627e-01 -7.38408029e-01 -1.53469038e+00 5.17257094e-01 2.53605574e-01 -1.12229049e+00 -2.91869286e-02 -1.05671728e+00 -3.98726076e-01 1.03750932e+00 -1.14082432e+00 -8.59020591e-01 -5.02649784e-01 7.80850232e-01 -1.62627086e-01 -3.14220667e-01 1.07153285e+00 6.03371322e-01 -6.09279871e-01 7.58468330e-01 3.05693656e-01 -1.02587812e-01 7.93638289e-01 -8.99073839e-01 -2.92022794e-01 3.03698957e-01 1.66736066e-01 9.21844482e-01 1.01736158e-01 -3.37013334e-01 -1.72809446e+00 -8.52266729e-01 -3.19207534e-02 -3.08440268e-01 7.17023432e-01 -7.77786493e-01 -1.02106667e+00 4.20562387e-01 2.15653971e-01 -3.52047384e-01 1.17957211e+00 2.26280048e-01 2.79856212e-02 -5.16621232e-01 -1.14653397e+00 5.54980338e-01 1.16698313e+00 -3.64385962e-01 -4.73697484e-01 4.00238670e-02 -4.98615727e-02 -1.33093402e-01 -1.33878481e+00 4.35720116e-01 7.12614834e-01 -7.58475423e-01 7.47979641e-01 -5.55709600e-01 -6.92350343e-02 -1.14202194e-01 -7.44701698e-02 -1.77651393e+00 -4.20204282e-01 -2.54945278e-01 6.69767857e-02 1.07950354e+00 5.17128408e-01 -1.07569170e+00 4.36028153e-01 9.61839557e-01 1.71586908e-02 -8.05648029e-01 -1.03731096e+00 -8.19144726e-01 3.44539434e-01 -4.99022186e-01 7.38355696e-01 6.48806095e-01 5.22599995e-01 3.08877200e-01 -3.23638655e-02 4.71545346e-02 3.82051736e-01 -3.64194393e-01 7.50111163e-01 -1.61688256e+00 -2.42924675e-01 -4.51595157e-01 -1.05981469e+00 -4.09825623e-01 2.42019147e-01 -1.14240980e+00 -2.55989850e-01 -1.20884514e+00 3.06558132e-01 -4.77005363e-01 -8.02074254e-01 5.15513659e-01 9.89865661e-02 2.39102647e-01 -1.55699283e-01 -3.89240608e-02 -1.59500256e-01 2.63933867e-01 9.67256129e-01 -2.48814672e-01 -4.67115283e-01 -3.43233585e-01 -8.76671731e-01 3.33149970e-01 1.07545149e+00 -4.78892267e-01 -5.93346655e-01 -5.56980431e-01 2.28261337e-01 -3.50477874e-01 3.79527837e-01 -1.62714899e+00 2.69318759e-01 1.74618348e-01 9.02615905e-01 5.14587685e-02 4.66544442e-02 -1.23197591e+00 6.21390045e-01 1.79700315e-01 3.31194401e-02 2.47116312e-02 6.80778742e-01 4.97754574e-01 1.30891070e-01 6.98765069e-02 3.25128317e-01 1.52032286e-01 -5.05487382e-01 1.95146546e-01 -4.22260940e-01 -1.07349865e-01 1.15342307e+00 -7.16636419e-01 -1.61100760e-01 2.34445296e-02 -8.32690656e-01 -1.98864639e-01 3.44139874e-01 6.13064408e-01 4.50023353e-01 -1.17954755e+00 -3.90791237e-01 7.62048364e-01 3.24458152e-01 -3.76831502e-01 2.77117997e-01 1.12600398e+00 -9.53142941e-02 3.14710438e-01 -1.04466271e+00 -8.29635739e-01 -9.92741585e-01 3.31024587e-01 3.27177703e-01 2.23179802e-01 -3.06868017e-01 4.13922071e-01 -1.98641643e-01 -3.83572668e-01 2.37351403e-01 -4.94605452e-01 -2.50412166e-01 5.83095150e-03 2.33472839e-01 3.94387096e-01 6.45228922e-01 -1.68791220e-01 -3.98321778e-01 5.60527384e-01 3.12905908e-02 1.19785480e-01 1.70649230e+00 -3.78311351e-02 -1.71560168e-01 6.87183917e-01 1.12984204e+00 -4.13955510e-01 -1.05760467e+00 3.65989715e-01 -7.04128742e-02 -2.60565668e-01 -1.19811319e-01 -1.03936768e+00 -1.15280735e+00 8.92840803e-01 1.28497243e+00 -8.29660892e-02 1.29898822e+00 -2.86722064e-01 2.88221061e-01 4.48517114e-01 1.16484642e+00 -1.13051987e+00 -8.48131105e-02 1.25906974e-01 9.34015572e-01 -9.75503147e-01 -1.17931344e-01 9.33434665e-02 -2.96259582e-01 1.23361838e+00 6.95501149e-01 1.74680009e-01 7.82251060e-01 3.61465096e-01 -1.16015710e-01 -2.99491525e-01 -1.59140617e-01 2.92321622e-01 2.92689025e-01 1.26459777e+00 4.89634246e-01 2.31114522e-01 -2.05666587e-01 9.86411631e-01 -3.69248092e-01 4.46895868e-01 2.74252683e-01 8.25223923e-01 5.97323328e-02 -1.34134710e+00 -2.54626215e-01 1.11288083e+00 -1.63059965e-01 -1.72129795e-01 -2.89505750e-01 8.50617111e-01 3.96650791e-01 1.07840729e+00 -2.11521938e-01 -7.65508711e-01 4.93185669e-01 3.43245536e-01 9.19157207e-01 -4.37210023e-01 -5.39383054e-01 -4.03031558e-01 -5.38929939e-01 -4.23276305e-01 -3.74811351e-01 -5.16628027e-01 -9.86555219e-01 -1.42556638e-01 -4.23994631e-01 -1.44709408e-01 9.10962999e-01 8.13917279e-01 9.04652119e-01 8.48017097e-01 2.00747579e-01 -1.08836174e+00 -2.93568045e-01 -1.28925538e+00 -9.74875808e-01 5.19010246e-01 -1.33995235e-01 -9.77297664e-01 -3.15573424e-01 -6.51456863e-02]

[13.1327543258667, 3.4434077739715576]

79e8da24-be4a-4cb3-8306-62da2b9f0e80

multi-modal-deep-learning-system-for

2212.14490

https://arxiv.org/abs/2212.14490v1

https://arxiv.org/pdf/2212.14490v1.pdf

Multi-modal deep learning system for depression and anxiety detection

Traditional screening practices for anxiety and depression pose an impediment to monitoring and treating these conditions effectively. However, recent advances in NLP and speech modelling allow textual, acoustic, and hand-crafted language-based features to jointly form the basis of future mental health screening and condition detection. Speech is a rich and readily available source of insight into an individual's cognitive state and by leveraging different aspects of speech, we can develop new digital biomarkers for depression and anxiety. To this end, we propose a multi-modal system for the screening of depression and anxiety from self-administered speech tasks. The proposed model integrates deep-learned features from audio and text, as well as hand-crafted features that are informed by clinically-validated domain knowledge. We find that augmenting hand-crafted features with deep-learned features improves our overall classification F1 score comparing to a baseline of hand-crafted features alone from 0.58 to 0.63 for depression and from 0.54 to 0.57 for anxiety. The findings of our work suggest that speech-based biomarkers for depression and anxiety hold significant promise in the future of digital health.

['Jekaterina Novikova', 'Marija Stanojevic', 'Brian Diep']

2022-12-30

['anxiety-detection']

['medical']

[ 1.59392595e-01 2.76295841e-01 -2.88422614e-01 -7.32551455e-01 -1.40088308e+00 -3.18209589e-01 3.69795471e-01 6.88322544e-01 -1.61483273e-01 2.21581578e-01 6.96861982e-01 -7.74992332e-02 -2.82220840e-01 -6.70950413e-01 -1.14578359e-01 -1.58358067e-01 -2.10149705e-01 2.89004803e-01 -2.46986732e-01 -2.52341777e-01 -3.45051587e-01 2.39246026e-01 -1.45045662e+00 7.13562548e-01 8.25986207e-01 1.37152433e+00 -5.10302838e-03 5.41291058e-01 -1.32053122e-01 5.59840322e-01 -4.43529278e-01 -4.96652693e-01 -3.59864652e-01 -3.54647458e-01 -7.11838543e-01 6.55393824e-02 1.52601376e-01 -6.32644653e-01 -3.08300704e-01 8.52026105e-01 7.96842158e-01 -3.05052578e-01 2.74436504e-01 -7.26371884e-01 -6.32601857e-01 4.78925347e-01 -1.58703029e-01 7.40563348e-02 9.17807639e-01 -3.38291167e-03 8.06762397e-01 -4.13477033e-01 3.20793539e-01 1.38421118e+00 9.04455662e-01 7.90631711e-01 -1.06501889e+00 -5.81463575e-01 7.17204437e-02 4.22131233e-02 -1.07034457e+00 -8.65301371e-01 4.89892364e-01 -6.05686605e-01 1.01484728e+00 2.42523909e-01 9.63407099e-01 1.37256575e+00 -8.54473785e-02 5.52177131e-01 9.10619974e-01 -1.80133581e-01 1.93524808e-02 3.77811641e-01 1.21288262e-01 8.72191608e-01 2.14875881e-02 -2.57377058e-01 -5.88476241e-01 -6.75287545e-01 2.44350970e-01 1.09495811e-01 2.25104347e-01 4.82163072e-01 -7.67256320e-01 7.96737134e-01 3.80278714e-02 3.01659405e-01 -5.60111701e-01 -1.94019392e-01 5.16347051e-01 1.60306171e-01 8.36371601e-01 2.78562158e-01 -5.83775222e-01 -6.11171246e-01 -1.12720490e+00 -8.71335641e-02 5.60131252e-01 6.39768317e-02 4.31645185e-01 -4.25575934e-02 -5.20180285e-01 1.34055805e+00 5.10350525e-01 7.55792499e-01 7.79198349e-01 -7.85121083e-01 3.04175228e-01 7.40524828e-01 -1.78283662e-01 -1.23035729e+00 -9.26271200e-01 -4.05275106e-01 -4.81682718e-01 -6.98357642e-01 -8.73581469e-02 -4.39045012e-01 -7.31464386e-01 1.86040533e+00 3.69953096e-01 8.46069977e-02 -1.09834783e-01 4.62864906e-01 9.65684831e-01 3.47140223e-01 5.26638068e-02 -4.91022691e-02 1.55626333e+00 -3.08837652e-01 -7.88426161e-01 -6.92559719e-01 7.27472246e-01 -3.38479578e-01 9.03824329e-01 6.30719185e-01 -9.89302933e-01 -2.54173398e-01 -7.07300127e-01 1.64237201e-01 -1.73334137e-01 3.81660193e-01 8.17211092e-01 1.30089223e+00 -1.12853920e+00 3.28257084e-01 -1.09805298e+00 -4.18478340e-01 6.79802477e-01 5.68817079e-01 -4.00268137e-01 -2.33049974e-01 -1.16737449e+00 5.37418127e-01 -2.47489512e-01 -2.55887804e-04 -8.68852317e-01 -5.88683665e-01 -7.94033349e-01 3.19440775e-02 4.85703349e-02 -6.08375311e-01 1.31387734e+00 -7.70363510e-01 -1.54490483e+00 8.71827483e-01 -2.06010878e-01 -4.80762422e-01 4.33829613e-02 -3.00924689e-01 -8.42577279e-01 6.36665225e-01 1.53518692e-01 5.55424035e-01 6.33659601e-01 -2.73221612e-01 -3.92963201e-01 -8.70706260e-01 -2.61975616e-01 -1.71931952e-01 -1.09128869e+00 3.78888011e-01 -3.86233389e-01 -2.88360596e-01 2.18779910e-02 -7.16692150e-01 -7.02641755e-02 2.52072871e-01 -5.15266001e-01 9.17762071e-02 4.67970192e-01 -1.13469243e+00 1.18278968e+00 -2.23682284e+00 -3.73691648e-01 1.49203509e-01 4.86195296e-01 5.21278143e-01 -2.76359886e-01 3.04539323e-01 1.67026550e-01 2.57845849e-01 2.43752941e-01 -4.57026988e-01 -8.13637525e-02 -3.64458449e-02 3.70708317e-01 2.58959025e-01 7.00965345e-01 7.74145424e-01 -8.69305491e-01 -2.14769259e-01 1.57751471e-01 9.92019415e-01 -9.67769563e-01 1.95993379e-01 1.27176968e-02 3.07721198e-01 -7.08086848e-01 8.18878055e-01 3.30443829e-01 -2.26654932e-01 2.45315686e-01 2.06209496e-01 3.95597965e-01 9.48946416e-01 -7.09523857e-01 1.44070804e+00 -5.26134789e-01 4.91660416e-01 3.27366322e-01 -1.03176737e+00 1.04795086e+00 5.61888874e-01 6.10437334e-01 -7.23385572e-01 2.40064636e-01 -2.79245786e-02 1.04748890e-01 -9.22042966e-01 -2.94634789e-01 -2.04671174e-01 1.91312265e-02 1.66561976e-01 2.53339142e-01 9.34254080e-02 -3.07859659e-01 1.88546777e-01 1.74614191e+00 -7.48294115e-01 9.19889063e-02 2.55509257e-01 3.08648199e-01 -5.09153187e-01 3.20955694e-01 6.31225705e-01 -3.50589424e-01 4.00335401e-01 5.79520285e-01 8.32696557e-02 -3.44930738e-01 -7.53525496e-01 -2.81949967e-01 1.22666264e+00 -1.00735128e+00 -7.85922468e-01 -7.45760798e-01 -4.75821018e-01 1.75640106e-01 3.31421226e-01 -7.25530744e-01 -6.46025658e-01 3.68827701e-01 -8.39075625e-01 9.90064681e-01 4.84758347e-01 3.12199920e-01 -6.80070043e-01 -5.12620360e-02 4.29557800e-01 -3.64918977e-01 -1.34773254e+00 -4.96974541e-03 2.06738226e-02 -8.34168673e-01 -5.80496252e-01 -7.39411891e-01 -3.53861153e-01 1.12425491e-01 -8.65080431e-02 6.52527153e-01 -1.29590750e-01 -4.64414597e-01 9.35472250e-01 -2.76421100e-01 -4.30828393e-01 -5.09811282e-01 -1.44900337e-01 1.97216600e-01 3.49310338e-01 3.44520450e-01 -6.15301609e-01 -5.50543904e-01 -1.87924266e-01 -5.77004671e-01 -4.94427145e-01 6.74318492e-01 4.37383413e-01 2.58856654e-01 -2.93610543e-01 8.78247023e-01 -6.09530091e-01 9.62204993e-01 -7.02448964e-01 9.76732820e-02 -4.95150462e-02 -4.12295967e-01 -2.38305852e-01 9.86554399e-02 -6.59854233e-01 -8.04277658e-01 6.40002862e-02 -7.58374035e-01 2.84886602e-02 -5.00748277e-01 7.73118258e-01 -1.66596532e-01 1.10830516e-01 5.47384918e-01 2.91638188e-02 9.58126932e-02 -6.42427981e-01 2.07179755e-01 1.25195658e+00 4.35256481e-01 -2.86064386e-01 -6.53159767e-02 3.64637047e-01 -3.13856155e-01 -1.24136257e+00 -1.08808076e+00 -4.60566849e-01 -2.16977000e-01 1.48488162e-02 8.70671988e-01 -1.12701964e+00 -5.99194944e-01 3.43756974e-01 -9.07343745e-01 1.58254690e-02 4.14814688e-02 5.27648151e-01 -1.69286519e-01 2.32907996e-01 -7.30085671e-01 -1.22817087e+00 -6.46562874e-01 -9.48378623e-01 1.53773224e+00 -1.24891169e-01 -6.59647405e-01 -9.18988764e-01 2.18422264e-01 7.11623192e-01 3.88438314e-01 1.40986636e-01 7.66643584e-01 -1.03051209e+00 4.28853631e-01 -9.21838284e-01 -1.68672249e-01 6.07507586e-01 3.89003366e-01 -2.20286757e-01 -1.19955361e+00 7.40934163e-02 9.04929563e-02 -5.90961456e-01 7.71854579e-01 5.09820521e-01 1.08236897e+00 -3.11182052e-01 -3.73590618e-01 1.45462498e-01 5.64890444e-01 1.42974079e-01 3.30835760e-01 -2.95412596e-02 3.52741122e-01 5.47386587e-01 9.55678299e-02 9.34565842e-01 6.45504117e-01 5.64862669e-01 1.08677074e-01 -1.86609495e-02 -5.02370819e-02 -2.07693726e-01 6.59526408e-01 7.59440660e-01 4.70551997e-01 -6.38498142e-02 -1.19617701e+00 5.62223256e-01 -1.12994850e+00 -7.59163797e-01 8.94585475e-02 1.94679677e+00 1.04274547e+00 1.93882018e-01 4.91698891e-01 4.22394961e-01 4.60050046e-01 -2.51130611e-01 -3.68987262e-01 -3.97800773e-01 1.44185603e-01 4.22211736e-01 -2.16978505e-01 1.88104898e-01 -1.03031325e+00 4.74994928e-01 6.56766510e+00 4.57797974e-01 -1.40883493e+00 4.33778167e-01 6.79864585e-01 -3.24634224e-01 -1.31459057e-01 -6.97264612e-01 -8.12202573e-01 3.45212013e-01 1.76935172e+00 2.94888377e-01 3.76257539e-01 4.97606814e-01 5.55261910e-01 8.93850699e-02 -1.03237832e+00 9.17345226e-01 1.76511541e-01 -8.35581779e-01 -2.85159618e-01 1.37103781e-01 2.19209030e-01 2.96750039e-01 5.16734064e-01 1.10516779e-01 -8.69301111e-02 -9.87087011e-01 2.36395150e-01 5.17022014e-01 9.05019701e-01 -5.77973068e-01 5.35880923e-01 2.14227259e-01 -6.58713639e-01 -2.73836434e-01 9.13946424e-03 -2.12154329e-01 -1.91606387e-01 1.02999365e+00 -1.32076919e+00 1.00649394e-01 4.72164989e-01 8.51176023e-01 -6.05780900e-01 8.88295889e-01 9.30749625e-02 1.09110272e+00 -4.55363721e-01 -3.95625643e-02 1.53446361e-01 4.27109361e-01 3.81430537e-02 1.36901724e+00 5.85223854e-01 2.05883160e-02 8.56473520e-02 7.52770126e-01 7.22341686e-02 2.20390871e-01 -5.66313207e-01 -8.83376181e-01 2.86112756e-01 1.23498535e+00 -4.15998042e-01 -3.54397416e-01 -5.64640045e-01 7.72419035e-01 8.41317475e-02 -9.14357528e-02 -4.62865919e-01 -2.81336904e-01 7.58683085e-01 2.19858766e-01 1.98836606e-02 3.63970697e-02 -1.24199353e-01 -1.04289734e+00 -3.10282499e-01 -1.12151074e+00 3.55900049e-01 -5.97596169e-01 -1.30942559e+00 6.13135517e-01 -5.74168444e-01 -6.59860432e-01 -4.32099074e-01 -5.56964874e-01 -4.05016690e-01 7.16150582e-01 -1.01389229e+00 -9.48637486e-01 -1.11116596e-01 4.76648033e-01 3.13090831e-01 -1.58094525e-01 1.23284340e+00 5.82336724e-01 -9.06732202e-01 7.38809466e-01 1.79283824e-02 2.08520338e-01 7.11538792e-01 -7.08678246e-01 1.43209621e-01 2.12379754e-01 2.13507682e-01 5.01518130e-01 4.44913030e-01 -7.92731225e-01 -1.41419923e+00 -1.18457735e+00 8.90814006e-01 -6.13790095e-01 8.10634792e-01 -5.17867029e-01 -7.94867218e-01 3.85814011e-01 -3.59394997e-01 -1.30111620e-01 1.18133509e+00 5.39885998e-01 -3.33783090e-01 -2.73324877e-01 -1.49855983e+00 7.22781047e-02 7.85100341e-01 -1.18169963e+00 -3.34679008e-01 6.32984340e-01 6.47591233e-01 -2.36410331e-02 -1.02078259e+00 1.82143912e-01 8.55620384e-01 -6.84151947e-01 1.22712052e+00 -6.72762334e-01 4.77038741e-01 7.38044202e-01 -3.48837376e-01 -1.25375450e+00 -5.60045652e-02 -6.06868923e-01 -5.54081984e-02 1.38414347e+00 5.82587004e-01 -6.13847017e-01 5.07299125e-01 7.88345337e-01 1.70751754e-03 -6.74796581e-01 -8.08079720e-01 -4.26494747e-01 -1.23787783e-01 -1.03207898e+00 2.48474270e-01 7.60092139e-01 5.04256666e-01 2.95474648e-01 -2.23739862e-01 3.54927748e-01 3.38775665e-02 -6.07474625e-01 2.04995692e-01 -1.57157457e+00 -4.24335331e-01 -1.89195305e-01 -6.82604671e-01 -4.38195109e-01 1.40535653e-01 -1.04839849e+00 -2.94326484e-01 -1.52077115e+00 4.18924421e-01 -4.50848155e-02 -3.47512335e-01 7.87058413e-01 -9.51961130e-02 1.11052252e-01 -1.81235433e-01 -5.49370825e-01 -4.49102938e-01 5.19551575e-01 7.15497017e-01 -2.34288171e-01 -3.69589537e-01 2.02761605e-01 -1.24546146e+00 7.43811071e-01 8.88746679e-01 -4.95116264e-01 -4.36486274e-01 -2.74609774e-01 1.22792758e-01 4.46353257e-01 3.09723616e-01 -9.75715399e-01 -1.32035717e-01 9.28162858e-02 4.55771893e-01 -1.84186772e-01 1.05304396e+00 -2.36771688e-01 -4.82622772e-01 5.23538649e-01 -6.06952846e-01 -3.43697190e-01 5.48586845e-01 5.81160188e-01 -2.82475706e-02 2.53400803e-02 3.07056397e-01 -5.67972921e-02 3.16553116e-02 8.61066673e-03 -7.41958857e-01 8.80674347e-02 2.83028245e-01 3.14460903e-01 -1.21253677e-01 -9.17634249e-01 -1.17761242e+00 -1.40133411e-01 -4.42043692e-01 3.94757807e-01 7.19724238e-01 -1.01111495e+00 -5.60453117e-01 4.29112136e-01 -2.27249768e-02 -8.69655788e-01 3.10970515e-01 1.16192114e+00 2.81033605e-01 6.61172330e-01 6.12139292e-02 -4.85136330e-01 -1.37523377e+00 2.92881340e-01 2.44850352e-01 4.19648699e-02 -2.27756992e-01 1.06083357e+00 -1.23037726e-01 -3.98392469e-01 4.50995475e-01 -6.69659197e-01 -1.04231186e-01 3.91153485e-01 9.32529688e-01 2.70367235e-01 5.38527787e-01 -3.69519114e-01 -6.33067548e-01 6.20462708e-02 6.56574033e-03 -3.49312872e-01 1.68440402e+00 -8.36706907e-02 3.28962468e-02 4.44080323e-01 1.12452710e+00 2.45860349e-02 -3.51161301e-01 -1.71749130e-01 7.36418441e-02 -2.16926821e-02 5.58911741e-01 -1.26544356e+00 -9.00021553e-01 1.23000252e+00 1.03694630e+00 3.08684468e-01 1.18255067e+00 3.41953367e-01 6.61734343e-01 5.91689050e-01 -1.89592820e-02 -7.62652814e-01 4.38645661e-01 1.80892706e-01 6.28992617e-01 -1.01796663e+00 -4.04082447e-01 -2.82708764e-01 -6.71234071e-01 8.28528345e-01 2.31532946e-01 2.05979943e-01 8.29552352e-01 3.76281828e-01 -1.72942057e-02 -2.24290013e-01 -1.03271353e+00 -4.00861979e-01 6.59692824e-01 1.03001201e+00 7.90919542e-01 2.90062815e-01 2.01000199e-02 1.58741581e+00 -2.12111712e-01 6.97585866e-02 1.56526729e-01 6.95351899e-01 -7.12339997e-01 -1.16228986e+00 -4.44809258e-01 1.06780493e+00 -8.74374211e-01 -1.66730523e-01 -9.40084577e-01 8.86559412e-02 1.41219392e-01 1.42908072e+00 -5.70494756e-02 -5.97760320e-01 2.22324699e-01 5.68042457e-01 2.90588707e-01 -1.09287822e+00 -5.50096929e-01 3.13669711e-01 6.03698969e-01 -6.72944605e-01 -2.04352930e-01 -8.25014949e-01 -9.42096412e-01 2.34158009e-01 -1.29768521e-01 -2.43169442e-01 6.45364821e-01 1.08598208e+00 8.12713027e-01 6.03915870e-01 3.08797747e-01 -3.24337602e-01 -4.33656961e-01 -1.00513172e+00 -4.99003679e-01 -2.14294091e-01 7.15280235e-01 -4.83702123e-01 1.06426559e-01 -3.86736780e-01]

[13.767669677734375, 5.0406928062438965]

bbd4a7f9-c67d-4229-9118-c29f4198ec9f

evaluating-multilingual-sentence

https://aclanthology.org/2022.lrec-1.314

https://aclanthology.org/2022.lrec-1.314.pdf

Evaluating Multilingual Sentence Representation Models in a Real Case Scenario

In this paper, we present an evaluation of sentence representation models on the paraphrase detection task. The evaluation is designed to simulate a real-world problem of plagiarism and is based on one of the most important cases of forgery in modern history: the so-called “Protocols of the Elders of Zion”. The sentence pairs for the evaluation are taken from the infamous forged text “Protocols of the Elders of Zion” (Protocols) by unknown authors; and by “Dialogue in Hell between Machiavelli and Montesquieu” by Maurice Joly. Scholars have demonstrated that the first text plagiarizes from the second, indicating all the forged parts on qualitative grounds. Following this evidence, we organized the rephrased texts and asked native speakers to quantify the level of similarity between each pair. We used this material to evaluate sentence representation models in two languages: English and French, and on three tasks: similarity correlation, paraphrase identification, and paraphrase retrieval. Our evaluation aims at encouraging the development of benchmarks based on real-world problems, as a means to prevent problems connected to AI hypes, and to use NLP technologies for social good. Through our evaluation, we are able to confirm that the infamous Protocols are actually a plagiarized text but, as we will show, we encounter several problems connected with the convoluted nature of the task, that is very different from the one reported in standard benchmarks of paraphrase detection and sentence similarity. Code and data available at https://github.com/roccotrip/protocols.

['Simon Levis Sullam', 'Rexhina Blloshmi', 'Rocco Tripodi']

lrec-2022-6

['paraphrase-identification']

['natural-language-processing']

[ 3.53700593e-02 8.85004997e-02 1.29710034e-01 -3.22884880e-02 -7.52633452e-01 -8.34383309e-01 1.03376365e+00 6.44436598e-01 -3.67003262e-01 6.28740132e-01 6.83978617e-01 -4.03632909e-01 -1.73534408e-01 -7.09260166e-01 -5.08751988e-01 -3.88260752e-01 1.97157577e-01 4.39440489e-01 -5.05066849e-02 -8.96870494e-01 1.01607466e+00 3.42361897e-01 -1.44804108e+00 5.89182615e-01 7.58035362e-01 3.84255111e-01 1.39128994e-02 6.13251805e-01 -7.29040429e-02 1.10685861e+00 -1.03929746e+00 -9.88352358e-01 -8.06114003e-02 -5.51249921e-01 -1.38275182e+00 -3.76529127e-01 4.97487068e-01 -9.77855846e-02 -4.29217488e-01 1.23553574e+00 2.51760662e-01 -1.42848998e-01 6.77533567e-01 -1.12502551e+00 -7.63896704e-01 7.69502461e-01 -6.04574122e-02 4.18832690e-01 1.16513216e+00 1.33649871e-01 1.02808714e+00 -7.26768672e-01 8.84115398e-01 1.36415482e+00 7.03285277e-01 4.28507358e-01 -1.01575327e+00 -4.59042102e-01 -7.97197044e-01 5.88252604e-01 -1.04246354e+00 -5.23451447e-01 7.89369345e-01 -5.00549734e-01 6.57016695e-01 6.22391999e-01 5.88258326e-01 1.52868485e+00 2.26992518e-01 5.50806463e-01 1.06305671e+00 -5.77849150e-01 9.56378207e-02 5.43812871e-01 4.37438220e-01 4.53395396e-01 1.64524823e-01 -9.65786502e-02 -5.25550663e-01 -5.68983555e-01 1.53405383e-01 -1.91147014e-01 -6.00713491e-01 4.30856310e-02 -9.70980048e-01 1.25035119e+00 2.81570673e-01 1.05783093e+00 -1.73093751e-01 -1.35668278e-01 7.86917090e-01 7.78162599e-01 3.69618326e-01 9.39574301e-01 6.68924376e-02 -4.72738087e-01 -1.13698983e+00 6.13743722e-01 1.34657681e+00 5.93267560e-01 4.31811959e-01 -6.16674960e-01 8.69404078e-02 8.13035965e-01 -3.66380028e-02 1.33345619e-01 8.66775036e-01 -1.05875909e+00 7.18500257e-01 6.87061131e-01 2.41825029e-01 -1.45873213e+00 -3.16940472e-02 -2.31811866e-01 -5.86836874e-01 -1.13893025e-01 6.78722143e-01 2.75201023e-01 -9.24781263e-02 1.45988441e+00 -2.27617919e-01 -3.47419202e-01 1.64855823e-01 9.14315045e-01 8.73315930e-01 6.08031154e-01 -3.27540994e-01 -2.18861118e-01 1.60722792e+00 -5.44516206e-01 -5.80038190e-01 -1.85087398e-01 8.94743323e-01 -1.10173929e+00 1.36007094e+00 3.89842629e-01 -1.33600593e+00 -1.72649950e-01 -1.12411857e+00 -1.87818825e-01 -3.77524376e-01 -9.94225368e-02 2.29130626e-01 7.88194895e-01 -8.08865011e-01 1.08464062e+00 -1.77869961e-01 -6.93449259e-01 1.62190869e-01 -2.99657464e-01 -4.12487030e-01 -3.70517210e-03 -1.51332533e+00 1.33742785e+00 2.55838662e-01 -4.83090393e-02 -5.29924333e-01 -5.68631709e-01 -6.27971888e-01 1.83723181e-01 2.12084979e-01 -4.19434637e-01 1.17880714e+00 -1.12415314e+00 -1.07357621e+00 1.66066968e+00 6.67155758e-02 -5.33212721e-01 7.11439550e-01 -2.27056399e-01 -4.28666204e-01 3.83589864e-01 3.35406691e-01 -1.24084309e-01 7.29231179e-01 -9.41380560e-01 1.39102310e-01 -4.62568909e-01 1.58063844e-01 -4.42221090e-02 -3.63663584e-01 4.74800825e-01 3.34048718e-01 -7.58521259e-01 -1.03470035e-01 -6.64783895e-01 2.63063222e-01 -3.13253373e-01 -4.22015727e-01 -2.59731978e-01 4.56552833e-01 -9.52962041e-01 1.43272007e+00 -2.10648918e+00 1.74691930e-01 -7.49130845e-02 3.32317948e-01 3.87714654e-01 9.42841247e-02 1.34962618e+00 -3.49849254e-01 4.34973627e-01 -4.10681248e-01 -2.58221567e-01 -3.85630317e-02 -4.23270501e-02 -6.75804853e-01 5.32534003e-01 -1.12231672e-01 7.57593870e-01 -1.05862415e+00 -4.53282982e-01 -1.49352700e-01 2.93865092e-02 -1.84059024e-01 1.53197274e-01 8.33706260e-02 3.75171304e-02 -2.20002979e-01 2.04818681e-01 4.85600024e-01 -1.99870139e-01 3.81308980e-03 -1.56511351e-01 -4.29773182e-02 7.99981833e-01 -6.72017515e-01 1.78694296e+00 -1.46733493e-01 7.55748808e-01 -4.10023555e-02 -8.48190069e-01 9.32208061e-01 3.39265347e-01 4.88372482e-02 -5.65366149e-01 3.34493697e-01 4.54430670e-01 -1.73116192e-01 -7.63841689e-01 9.78987515e-01 -2.69249082e-01 -3.36367905e-01 8.80102396e-01 -1.24148965e-01 -6.32452726e-01 3.39493901e-01 7.01446235e-01 1.30758369e+00 -3.21886688e-01 5.00312507e-01 -3.49910617e-01 9.01606917e-01 1.46215633e-01 -6.10411018e-02 7.90175080e-01 -3.08586836e-01 6.04959488e-01 9.16676283e-01 -3.06741208e-01 -1.23623645e+00 -1.08297825e+00 -3.21961306e-02 6.71209276e-01 4.65707965e-02 -8.64837825e-01 -8.21740210e-01 -4.95284796e-01 -2.35627964e-01 1.12990117e+00 -5.36491930e-01 -3.82151663e-01 -5.63778996e-01 -2.95561135e-01 1.06997037e+00 -2.86618114e-01 3.23921412e-01 -1.17846859e+00 -7.96356082e-01 5.31371422e-02 -5.90814710e-01 -8.38219643e-01 -2.62297720e-01 -3.87155175e-01 -5.30912936e-01 -1.29810488e+00 -5.81103504e-01 -4.81228769e-01 2.07207315e-02 1.29358992e-01 1.43657446e+00 4.65016127e-01 -2.73233920e-01 3.20382506e-01 -6.61169350e-01 -1.64261069e-02 -1.36780238e+00 -1.15050226e-01 -1.78979039e-01 -3.39635015e-01 3.61546218e-01 -9.03365195e-01 -4.16084856e-01 1.15251750e-01 -1.17573559e+00 -2.13787064e-01 2.73927659e-01 7.77213454e-01 -6.12823367e-01 -4.47245747e-01 4.95709866e-01 -8.76563072e-01 1.17028344e+00 -7.12235630e-01 -8.08576867e-02 2.76811540e-01 -4.08146799e-01 -2.08240077e-02 7.77673483e-01 -2.51126230e-01 -6.65198147e-01 -6.59800351e-01 -2.47521564e-01 -1.24750018e-01 -1.61765262e-01 3.89112264e-01 3.23635817e-01 1.34216145e-01 1.14123356e+00 3.45107883e-01 2.17737243e-01 -5.35588741e-01 4.30199713e-01 1.03752124e+00 6.34153605e-01 -5.07206440e-01 6.79926336e-01 2.96804756e-01 -4.00438040e-01 -9.44948316e-01 -5.91304243e-01 -4.60681915e-01 -2.01506644e-01 -1.67150244e-01 2.52020955e-01 -5.39241016e-01 -9.02495861e-01 3.64459455e-01 -1.67374742e+00 2.49111533e-01 -4.33032423e-01 1.03266969e-01 -6.37012005e-01 1.12887204e+00 -9.54920709e-01 -5.75162053e-01 -6.34694159e-01 -7.42499352e-01 6.62916005e-01 -1.34041801e-01 -8.10868561e-01 -9.33998525e-01 4.71672803e-01 6.55958116e-01 4.00272757e-01 9.32106823e-02 1.20586991e+00 -9.68259871e-01 1.19950145e-01 -2.19447732e-01 -1.17173687e-01 4.34058577e-01 -4.05987725e-03 -1.37299746e-01 -7.68462241e-01 -3.00982684e-01 7.87408292e-01 -6.14298761e-01 5.52978575e-01 -5.28188884e-01 5.35592079e-01 -6.60240591e-01 1.49515778e-01 -9.73945558e-02 1.28840232e+00 -3.54015440e-01 1.01099086e+00 4.31833625e-01 6.80998638e-02 9.12644506e-01 2.70825356e-01 4.48875278e-01 1.08745128e-01 5.21659672e-01 2.98953950e-01 5.73051035e-01 2.26465650e-02 -2.89971530e-01 5.70555568e-01 9.18580949e-01 2.72177070e-01 -2.02373102e-01 -7.91928887e-01 5.74377120e-01 -1.75113058e+00 -1.27213728e+00 -5.07367849e-01 2.23489666e+00 8.26457262e-01 -1.66902803e-02 2.32510448e-01 3.88983786e-01 7.21579432e-01 3.09079170e-01 1.89555600e-01 -8.40362847e-01 -1.39039055e-01 2.39979967e-01 -6.33035302e-02 4.58618581e-01 -5.79346657e-01 6.87432230e-01 5.84962416e+00 9.63146865e-01 -6.48239434e-01 1.21052250e-01 2.33403608e-01 1.45262510e-01 -3.64775568e-01 4.06446844e-01 -2.53072739e-01 8.67402673e-01 1.05549014e+00 -5.07183433e-01 4.74170297e-01 3.60618472e-01 2.17160180e-01 -3.24607193e-01 -1.26353002e+00 9.23126221e-01 5.58407485e-01 -1.36839938e+00 3.60042937e-02 -2.40859762e-01 1.08959049e-01 -1.50887504e-01 -3.57409954e-01 1.70114115e-01 7.02056885e-02 -1.01188385e+00 7.20863223e-01 3.97749811e-01 1.79884359e-01 -3.40797573e-01 6.86017990e-01 6.47082150e-01 -4.58481580e-01 -1.18378915e-01 -3.25212121e-01 -3.03306937e-01 1.38872787e-01 5.40924191e-01 -6.14937603e-01 7.25106239e-01 3.90471905e-01 6.85534298e-01 -8.78698468e-01 8.58121753e-01 -3.61617208e-01 5.34006119e-01 -1.12792760e-01 -3.23292255e-01 1.54018342e-01 -4.29302603e-01 1.01506841e+00 1.24459863e+00 2.24516809e-01 -1.08105808e-01 -5.61654925e-01 1.43431687e+00 -3.36158499e-02 2.89991051e-01 -8.79222155e-01 -2.65408665e-01 5.19129097e-01 1.13143587e+00 -3.52266729e-01 -2.95981437e-01 -6.41106069e-02 1.16650426e+00 4.36195254e-01 -1.02810994e-01 -7.12438166e-01 -5.70088208e-01 1.41974792e-01 2.69774139e-01 -1.85606167e-01 1.81421079e-02 -1.36594385e-01 -1.33986628e+00 2.03849673e-01 -1.14903426e+00 3.60976160e-01 -9.95530963e-01 -1.71419477e+00 4.71644402e-01 1.04355566e-01 -1.01782024e+00 -4.67985809e-01 -2.76489198e-01 -9.15786505e-01 7.78826714e-01 -8.26342344e-01 -9.03121829e-01 -7.16352388e-02 5.59655368e-01 3.29254001e-01 8.32172483e-03 9.09913480e-01 3.49141300e-01 -1.95439890e-01 3.44939858e-01 4.35989499e-02 1.84622899e-01 7.38421321e-01 -9.10311759e-01 4.68331456e-01 6.97836816e-01 1.95735946e-01 7.08637834e-01 1.22831929e+00 -5.22249997e-01 -1.12215245e+00 -2.22694650e-01 1.46835351e+00 -5.97786963e-01 1.17053592e+00 -2.51490533e-01 -1.21308374e+00 3.88694882e-01 4.96699423e-01 -8.85038376e-01 3.32860321e-01 -2.15905949e-01 -6.60738587e-01 4.04553890e-01 -1.17794740e+00 5.20593584e-01 8.27896237e-01 -7.95036733e-01 -1.38005674e+00 7.94853330e-01 5.20945609e-01 1.34045690e-01 -7.30129123e-01 -1.92089528e-01 4.21409905e-01 -1.41150618e+00 8.43468547e-01 -7.09972143e-01 9.81242776e-01 2.63823837e-01 -5.50760366e-02 -1.14293420e+00 1.48816720e-01 -7.32263625e-01 3.67400706e-01 1.43155992e+00 8.76195431e-02 -7.05007136e-01 6.92427158e-01 4.27145481e-01 5.50866053e-02 -1.97191879e-01 -1.19361019e+00 -7.34598041e-01 5.27560830e-01 -2.38426641e-01 2.39864364e-01 1.19946945e+00 5.71706474e-01 5.96674263e-01 -1.77689359e-01 -4.23206091e-01 5.14159858e-01 1.51479959e-01 6.88008070e-01 -1.02811611e+00 -3.71346682e-01 -7.29475796e-01 -3.72737855e-01 -6.51381373e-01 2.61361808e-01 -1.09163296e+00 -2.51720399e-01 -1.16447759e+00 3.18999320e-01 2.02525482e-02 4.48773623e-01 4.36507389e-02 1.00655578e-01 8.76617134e-02 4.98062640e-01 6.54537082e-01 -2.31144533e-01 4.14740086e-01 7.34763920e-01 -1.16630390e-01 6.66470826e-02 -2.04684772e-02 -7.22747266e-01 6.99380934e-01 8.53015065e-01 -8.60270262e-01 -1.61859374e-02 6.15542196e-02 8.24164689e-01 2.66083091e-01 8.29011381e-01 -8.35154951e-01 2.02927738e-01 7.39132091e-02 -4.66968387e-01 -1.95012003e-01 3.88298780e-01 -6.06512129e-01 2.07015440e-01 8.03777277e-01 -4.82921064e-01 2.51190126e-01 -1.49994284e-01 2.15432420e-01 -1.58775628e-01 -1.16103244e+00 6.64323211e-01 -4.64032143e-01 -7.74712488e-02 -5.08434951e-01 -6.97442532e-01 3.29403490e-01 6.65769041e-01 -1.01116739e-01 -9.42361712e-01 -7.28907764e-01 -4.59942758e-01 -1.22685932e-01 7.91135788e-01 3.36532980e-01 6.28859878e-01 -9.91468549e-01 -1.04132879e+00 -1.41910300e-01 1.24826953e-01 -8.82829368e-01 7.37595782e-02 7.81640410e-01 -8.14630747e-01 1.19283833e-01 -3.91947150e-01 -6.58815205e-02 -1.43047571e+00 6.93820417e-01 2.08469540e-01 -4.98967439e-01 -4.84663039e-01 3.55070174e-01 -3.76351804e-01 -4.49690253e-01 -8.55798945e-02 -2.95571052e-02 -2.79937685e-01 1.94294557e-01 4.99485046e-01 6.28416955e-01 1.30008131e-01 -6.40235484e-01 -2.56379992e-01 2.55282342e-01 -9.01665241e-02 -1.80528939e-01 1.13704956e+00 -1.26109734e-01 -6.82516575e-01 6.53567135e-01 1.46834457e+00 1.75671890e-01 7.54828146e-03 -4.02814820e-02 2.36810043e-01 -5.40241838e-01 -5.41617990e-01 -7.58950591e-01 -2.75342286e-01 8.69955719e-01 1.92829996e-01 8.35687339e-01 6.41869783e-01 2.99659241e-02 9.12525356e-01 5.53705633e-01 3.01900506e-01 -7.61339426e-01 1.92934975e-01 4.77786511e-01 1.37841535e+00 -8.13508987e-01 2.32937694e-01 -3.36235791e-01 -5.80811739e-01 1.20576799e+00 -1.87436063e-02 -3.04579645e-01 1.45774558e-01 -1.68387160e-01 -2.10503101e-01 -5.71872950e-01 -5.30490756e-01 3.54849339e-01 -6.86788037e-02 1.05315410e-01 3.73314559e-01 -2.04425290e-01 -1.06603360e+00 4.83497500e-01 -7.08737731e-01 -7.34139606e-02 1.12260568e+00 8.08037102e-01 -4.16198283e-01 -1.11510217e+00 -6.16219461e-01 2.56517082e-01 -5.76872051e-01 -2.46863693e-01 -1.08429039e+00 7.46818900e-01 -2.95190573e-01 1.16915917e+00 -1.57798678e-01 -3.23865116e-01 2.57461041e-01 1.19732693e-01 5.80508828e-01 -4.35921341e-01 -1.44678819e+00 -7.56244957e-01 4.31954145e-01 -4.38969165e-01 -4.11442906e-01 -5.05525947e-01 -6.40710294e-01 -1.00864613e+00 -2.21036658e-01 6.29126728e-01 6.70003891e-01 9.13155138e-01 1.84440389e-01 -2.23456696e-01 6.24400198e-01 -6.32225215e-01 -1.03276157e+00 -1.19262838e+00 -5.52113771e-01 1.08787620e+00 1.14315294e-01 -1.45689249e-01 -9.24133182e-01 -2.36377686e-01]

[8.842330932617188, 10.022235870361328]

7b5ca296-36da-4425-850b-ef1c939c7c27

zero-shot-event-causality-identification-with

https://aclanthology.org/2022.clib-1.13

https://aclanthology.org/2022.clib-1.13.pdf

Zero-shot Event Causality Identification with Question Answering

Extraction of event causality and especially implicit causality from text data is a challenging task. Causality is often treated as a specific relation type and can be considered as a part of relation extraction or relation classification task. Many causality identification-related tasks are designed to select the most plausible alternative of a set of possible causes and consider multiple-choice classification settings. Since there are powerful Question Answering (QA) systems pretrained on large text corpora, we investigated a zero-shot QA-based approach for event causality extraction using a Wikipedia-based dataset containing event descriptions (articles) and annotated causes. We aimed to evaluate to what extent reading comprehension ability of the QA-pipeline can be used for event-related causality extraction from plain text without any additional training. Some evaluation challenges and limitations of the data were discussed. We compared the performance of a two-step pipeline consisting of passage retrieval and extractive QA with QA-only pipeline on event-associated articles and mixed ones. Our systems achieved average cosine semantic similarity scores of 44 – 45% in different settings.

['Sven Schlarb', 'Daria Liakhovets']

clib-2022-9

['relation-classification', 'passage-retrieval', 'event-causality-identification']

['natural-language-processing', 'natural-language-processing', 'natural-language-processing']

[ 3.21733981e-01 4.04905051e-01 -1.84161320e-01 -3.83891612e-01 -1.20647001e+00 -7.50765324e-01 1.02198756e+00 1.06104183e+00 -6.28656387e-01 1.21770990e+00 7.19958425e-01 -3.55500251e-01 -5.64272642e-01 -9.36326325e-01 -7.33214915e-01 -1.67601705e-01 -2.35820308e-01 8.38520586e-01 6.28833652e-01 -2.50229895e-01 4.64402407e-01 2.93448240e-01 -1.54392052e+00 6.48212552e-01 1.04173625e+00 4.34564620e-01 3.88151780e-02 9.05430555e-01 -1.90855324e-01 1.53770888e+00 -7.24449277e-01 -6.55052364e-01 -4.09532398e-01 -5.85194230e-01 -1.52489972e+00 -1.01410747e+00 3.38607669e-01 -5.62540665e-02 -9.04537067e-02 6.53934896e-01 5.58364987e-01 6.41678041e-03 9.68137085e-01 -1.36820877e+00 -2.85700679e-01 9.28559244e-01 2.59352386e-01 9.52617526e-01 1.16175711e+00 -1.30964622e-01 1.33982491e+00 -6.95556521e-01 9.41988647e-01 1.35024714e+00 2.48293713e-01 2.55112231e-01 -9.71882045e-01 -4.51241225e-01 -1.40315339e-01 1.09846115e+00 -9.52294767e-01 -2.79680103e-01 4.33765113e-01 -4.83458489e-01 1.62413919e+00 4.53706086e-01 2.68429071e-01 1.30554628e+00 1.96100622e-01 4.80173856e-01 9.02282417e-01 -5.71694791e-01 4.14076000e-01 -2.47375846e-01 5.05790353e-01 4.09321338e-01 3.06198746e-01 -2.46954799e-01 -8.87083530e-01 -5.51784337e-01 1.48132786e-01 -7.97631681e-01 -3.51086050e-01 3.83005053e-01 -1.67215633e+00 8.56864810e-01 2.61949718e-01 2.04636455e-01 -7.78856337e-01 1.47004977e-01 5.51815987e-01 2.69968510e-01 2.16707379e-01 7.89443731e-01 -8.38027179e-01 -1.90738589e-01 -5.94376564e-01 7.37997830e-01 1.25935209e+00 5.84898472e-01 1.45649225e-01 -6.97592199e-01 -6.74140751e-01 5.98929286e-01 1.00397140e-01 3.19056243e-01 9.67528000e-02 -5.84261656e-01 6.80631697e-01 6.98228002e-01 2.39089742e-01 -1.06707132e+00 -6.15098238e-01 7.95330256e-02 -2.78470993e-01 -2.75354534e-01 8.07647169e-01 -2.76382297e-01 -5.67000926e-01 1.91414738e+00 7.03658104e-01 1.61015481e-01 2.37909406e-01 9.23546553e-01 1.15171337e+00 5.98248005e-01 7.03341305e-01 -5.52541614e-01 1.88535130e+00 -4.38130260e-01 -1.19872200e+00 -7.53245577e-02 6.02017462e-01 -8.31081092e-01 8.43785763e-01 3.07533890e-01 -9.21760082e-01 7.72654861e-02 -9.44718361e-01 -3.95241410e-01 -6.59729123e-01 -2.88403571e-01 6.39275193e-01 -6.59709275e-02 -5.50792515e-01 6.16116405e-01 -2.66255677e-01 -6.73859417e-01 9.17229727e-02 1.77738681e-01 -3.65385652e-01 1.44050673e-01 -2.20282865e+00 1.39159250e+00 7.07269430e-01 -3.26836556e-01 -8.33083749e-01 -1.03429794e+00 -7.28188396e-01 1.12260930e-01 5.75633824e-01 -1.00442481e+00 1.28270292e+00 -3.62768501e-01 -9.24106002e-01 8.02361012e-01 -4.24143225e-01 -5.60272872e-01 2.25132614e-01 -5.53975165e-01 -6.20868504e-01 5.40837109e-01 5.92158139e-01 1.42053053e-01 4.66084450e-01 -5.17626405e-01 -6.11367285e-01 -1.25104442e-01 1.50245488e-01 2.86736131e-01 5.22741824e-02 6.68442428e-01 1.48947209e-01 -5.92568696e-01 -3.57343912e-01 -3.78393143e-01 3.36634554e-02 -3.84309530e-01 -4.10851955e-01 -8.74360740e-01 2.46683180e-01 -8.80773723e-01 1.51694155e+00 -1.36598718e+00 4.26138239e-03 -1.63755134e-01 -5.19113317e-02 -1.95234403e-01 5.93116283e-02 7.67330825e-01 -5.20162702e-01 1.33876994e-01 -1.46118224e-01 3.16809267e-01 3.05445623e-02 9.15285125e-02 -6.51400745e-01 3.74677218e-02 6.80430412e-01 6.63903952e-01 -1.51930189e+00 -9.20350432e-01 -1.96085662e-01 1.01308428e-01 -2.22519562e-01 4.03772175e-01 -7.12582827e-01 2.53115505e-01 -3.99640828e-01 2.81603396e-01 1.05873540e-01 -2.39522710e-01 1.64937451e-01 -3.40766191e-01 -3.05716880e-02 1.31550479e+00 -1.08634377e+00 1.64498627e+00 -2.03013211e-01 4.86124933e-01 -7.50443578e-01 -9.10011113e-01 4.61686283e-01 9.66607630e-01 2.41251454e-01 -6.91191912e-01 -6.81225061e-02 3.63712400e-01 2.23720849e-01 -9.52354193e-01 2.53885448e-01 -5.94327599e-02 -1.15372814e-01 3.59597832e-01 3.41354549e-01 1.74730003e-01 8.51355910e-01 5.39801538e-01 1.72471797e+00 4.29008789e-02 7.13285029e-01 -3.40751261e-01 5.84576845e-01 7.68852830e-01 4.53300983e-01 6.59172118e-01 2.38113310e-02 5.35022080e-01 1.00427878e+00 -3.06196958e-01 -6.10918343e-01 -1.14970708e+00 -2.55847216e-01 7.54477739e-01 -1.75880328e-01 -7.23313153e-01 -4.51231658e-01 -9.71112430e-01 -3.43814611e-01 1.36201966e+00 -7.31856406e-01 1.71226501e-01 -7.03001738e-01 -1.03806591e+00 7.24266827e-01 1.86369091e-01 2.54063427e-01 -1.10467553e+00 -6.61978841e-01 5.07385969e-01 -7.67865658e-01 -1.18501520e+00 9.65266973e-02 1.63341358e-01 -3.88279080e-01 -1.77553844e+00 -3.03432405e-01 -3.87530386e-01 7.70986453e-02 -4.67846811e-01 1.52362823e+00 -2.26899430e-01 -1.07147634e-01 2.02896774e-01 -3.82477254e-01 -7.71130562e-01 -5.42150021e-01 -1.06889658e-01 -3.85245457e-02 -4.16550875e-01 6.88886225e-01 -4.40180868e-01 -6.17211580e-01 -1.24701798e-01 -5.33592939e-01 -1.26022309e-01 2.88692206e-01 5.63199818e-01 2.81175107e-01 -9.29594189e-02 9.50115204e-01 -6.46940887e-01 7.86197484e-01 -9.00055885e-01 -1.94305316e-01 4.60965633e-01 -5.93986750e-01 2.55798995e-01 5.20393014e-01 -4.26648378e-01 -1.38337016e+00 -2.73088753e-01 -1.09818965e-01 3.02782714e-01 -4.02918994e-01 7.14129269e-01 -3.09595644e-01 8.46746266e-01 1.02813065e+00 -2.98293263e-01 -4.45915580e-01 -1.08487904e-01 4.97566849e-01 3.38344872e-01 5.46951056e-01 -6.86894774e-01 4.16173458e-01 2.41035521e-01 1.54070079e-01 -3.81287575e-01 -1.21178472e+00 -4.74763125e-01 -4.59406912e-01 -1.43159360e-01 1.31441724e+00 -8.21027458e-01 -6.81573749e-01 1.43700920e-03 -1.91954410e+00 -2.09332258e-01 -1.82961211e-01 6.75940394e-01 -2.11826146e-01 8.42965022e-02 -4.69820559e-01 -5.46836793e-01 -5.37502944e-01 -4.95757282e-01 8.61402571e-01 8.24231952e-02 -9.40606833e-01 -9.52036500e-01 5.67816913e-01 2.93539703e-01 -1.07168369e-01 2.77688265e-01 1.15888047e+00 -1.13698471e+00 -2.04787895e-01 -7.92307630e-02 -2.28065580e-01 -5.10746479e-01 7.65957637e-04 -2.94290744e-02 -9.76034045e-01 7.21734941e-01 -3.47598612e-01 -2.95733869e-01 6.00502551e-01 6.65597022e-02 4.75812167e-01 -6.21156633e-01 -4.74277616e-01 -3.38783056e-01 1.12265253e+00 1.23055801e-01 6.68836117e-01 4.50953931e-01 5.13545454e-01 9.17606652e-01 7.36626267e-01 1.33290991e-01 7.48025477e-01 4.50563788e-01 2.67515182e-01 2.32717127e-01 -2.57076442e-01 -2.78755218e-01 1.64575607e-01 3.54018778e-01 -1.33407697e-01 -4.22138423e-01 -1.11982727e+00 9.33438897e-01 -1.88402677e+00 -1.36871481e+00 -1.03207505e+00 1.87817383e+00 1.59481299e+00 1.44575551e-01 -3.46622220e-03 4.08356518e-01 4.64888245e-01 -2.06691340e-01 -6.45104423e-03 -4.99602914e-01 -1.31644115e-01 3.73935491e-01 5.39608821e-02 4.89678055e-01 -1.21365070e+00 6.23266637e-01 5.83613157e+00 8.47385287e-01 -5.74289978e-01 2.13767335e-01 1.89913332e-01 1.64468467e-01 -3.28039885e-01 3.71820033e-01 -5.35696208e-01 3.51666778e-01 1.30678475e+00 -2.58300126e-01 -1.65908992e-01 2.27853477e-01 3.49310666e-01 -5.09761810e-01 -1.48769808e+00 3.69972438e-01 -2.68043190e-01 -1.48784435e+00 2.29118437e-01 -5.24168193e-01 3.06785464e-01 -1.66979596e-01 -7.55500436e-01 7.30658546e-02 2.67593414e-01 -9.71804440e-01 7.08051920e-01 7.65640497e-01 4.48632866e-01 -2.84948051e-01 7.17333078e-01 1.10795155e-01 -1.02093041e+00 1.32362172e-01 3.79887968e-02 -5.09115636e-01 3.54581028e-01 9.28124726e-01 -1.10928333e+00 9.12763357e-01 8.18068802e-01 6.39726996e-01 -5.86212456e-01 9.35449243e-01 -9.74525034e-01 1.04418492e+00 -4.96294677e-01 -5.07727802e-01 -3.56034905e-01 4.66006935e-01 8.05890143e-01 1.50408065e+00 1.55934626e-02 5.20671070e-01 -3.91891629e-01 1.02723444e+00 1.20363794e-01 2.07563624e-01 -4.01140064e-01 -4.78194430e-02 5.83106935e-01 1.10855389e+00 -6.35585368e-01 -5.59532762e-01 -1.16054423e-01 7.76292741e-01 4.26384807e-01 1.34807006e-01 -9.84700024e-01 -5.10398388e-01 3.22395444e-01 -1.07411340e-01 -1.39322728e-01 1.00912124e-01 -3.51089031e-01 -1.12160397e+00 -1.17397256e-01 -6.03013158e-01 1.19931471e+00 -1.00551546e+00 -1.46804023e+00 3.14792901e-01 4.64770019e-01 -8.04983795e-01 -3.79541636e-01 -2.88089454e-01 -9.14824367e-01 9.31868732e-01 -1.50447166e+00 -7.63724864e-01 -7.12939426e-02 4.92893845e-01 4.01513100e-01 4.40029413e-01 1.07153535e+00 3.54732692e-01 -4.49723005e-01 2.81249341e-02 -6.87652946e-01 2.19412357e-01 1.09862614e+00 -1.68838167e+00 -5.99717908e-02 1.02952063e+00 5.76473027e-02 6.31929517e-01 1.14817107e+00 -9.91710782e-01 -1.08372521e+00 -9.54913855e-01 1.99648035e+00 -8.86268437e-01 1.02001655e+00 -6.06897213e-02 -1.05396080e+00 3.30007881e-01 7.83039093e-01 -4.94268596e-01 7.23287940e-01 1.87120020e-01 -4.95872974e-01 1.35602087e-01 -8.89785528e-01 6.45907342e-01 9.84167874e-01 -3.61292094e-01 -1.53577244e+00 5.80170333e-01 8.35434198e-01 -1.69853508e-01 -9.23951387e-01 4.76856500e-01 1.35273412e-01 -5.34890831e-01 1.03759623e+00 -8.12181711e-01 9.75132406e-01 -4.77861017e-01 2.34805718e-01 -1.06714046e+00 5.20983362e-04 -5.89388967e-01 -3.02555382e-01 1.42511749e+00 8.55347574e-01 -4.45368320e-01 -1.44772723e-01 7.80805886e-01 2.72841126e-01 -1.62670359e-01 -8.92315805e-01 -3.52135986e-01 6.48781136e-02 -6.11904681e-01 2.80032218e-01 1.19826770e+00 4.82907027e-01 1.13030863e+00 3.17479730e-01 6.67613208e-01 6.74376547e-01 8.86567011e-02 2.76639074e-01 -1.17460251e+00 -1.09014446e-02 -1.67277738e-01 7.74374977e-02 -8.36812407e-02 1.64140239e-01 -9.80024934e-01 -5.89200668e-02 -2.04481816e+00 2.46091962e-01 -6.32714331e-02 -1.98423743e-01 6.03390574e-01 -8.66910994e-01 -2.72793710e-01 -3.26054245e-01 4.74737994e-02 -5.12645066e-01 2.73154199e-01 8.86874676e-01 -9.42665804e-03 -1.45012997e-02 -4.60002035e-01 -2.63307959e-01 7.24569082e-01 6.07212365e-01 -7.63068259e-01 -5.50120592e-01 -1.87893480e-01 1.00224316e+00 3.42112482e-01 8.02018464e-01 -5.96841812e-01 4.67275679e-01 -3.54882896e-01 3.74948792e-02 -5.80092728e-01 -1.87997222e-01 -3.53618801e-01 1.39049992e-01 2.94566274e-01 -6.45011663e-01 -1.01909384e-01 1.88703716e-01 4.93725270e-01 -2.26129591e-01 -3.49746257e-01 6.43499941e-02 -1.13853484e-01 -6.85547233e-01 -3.63632828e-01 -5.64978778e-01 4.55253333e-01 7.67301440e-01 3.54825079e-01 -7.57522047e-01 -1.64000183e-01 -4.92953300e-01 2.51600832e-01 -4.20303881e-01 4.03683007e-01 4.85012770e-01 -1.03175855e+00 -1.10960722e+00 -7.68924057e-01 1.64525807e-01 -1.53578624e-01 -4.57947962e-02 1.19512093e+00 -3.76613081e-01 5.86696923e-01 2.13007390e-01 -1.20275542e-01 -1.22046351e+00 7.27040827e-01 9.84794199e-02 -7.03457057e-01 -3.73952985e-01 8.72805774e-01 -1.02728248e-01 -6.69632927e-02 -2.19264925e-02 -3.41768473e-01 -8.28228116e-01 4.83434588e-01 7.72425592e-01 3.71070325e-01 2.26586252e-01 -3.65872495e-02 -8.15036595e-01 1.17834173e-01 2.58908898e-01 -3.82745296e-01 1.08188725e+00 1.07915476e-01 -5.65758824e-01 4.73230392e-01 7.68912256e-01 -4.74750847e-02 -3.46635699e-01 8.96981061e-02 6.85635507e-01 1.86142370e-01 -1.86598137e-01 -1.36751080e+00 -1.90107599e-01 6.04225934e-01 3.42541300e-02 4.60634381e-01 9.74479079e-01 3.07080001e-01 5.70405900e-01 2.31055737e-01 1.30683914e-01 -7.98508048e-01 -1.63661048e-01 6.22724473e-01 1.42559838e+00 -1.16821158e+00 1.98128391e-02 -6.51888192e-01 -3.95397067e-01 1.25039041e+00 4.57333624e-01 -1.08290995e-02 4.16144490e-01 1.25632733e-01 -1.46412596e-01 -7.04493761e-01 -1.23306537e+00 -5.02046883e-01 7.72359967e-01 2.73294449e-01 7.63042986e-01 -1.86679274e-01 -1.15663433e+00 9.45555031e-01 -1.78567529e-01 2.09057573e-02 4.86887872e-01 6.56720340e-01 -1.26057770e-02 -9.21024024e-01 -4.79628712e-01 3.74655336e-01 -8.14659894e-01 -5.56317389e-01 -5.64335346e-01 6.92369580e-01 1.41883746e-01 1.50051212e+00 -1.37391701e-01 -8.46572500e-03 3.89284343e-01 3.29840690e-01 5.40852129e-01 -5.30178428e-01 -8.73511016e-01 -3.45974356e-01 9.71381783e-01 -6.69538736e-01 -7.36280799e-01 -9.03877020e-01 -1.74026799e+00 1.12915277e-01 -1.61734208e-01 3.66888523e-01 2.33264491e-01 1.44636965e+00 2.13988706e-01 8.78135145e-01 7.73135573e-03 7.23650446e-03 -1.65770605e-01 -9.32655573e-01 1.97576627e-01 5.65749705e-01 1.99355185e-01 -5.89639664e-01 -3.45490098e-01 4.24357861e-01]

[9.150571823120117, 9.141562461853027]

0502354d-ee12-4673-949b-48214f6cbb3f

a-deep-learning-approach-to-solar-irradiance

1901.04881

http://arxiv.org/abs/1901.04881v1

http://arxiv.org/pdf/1901.04881v1.pdf

A deep learning approach to solar-irradiance forecasting in sky-videos

Ahead-of-time forecasting of incident solar-irradiance on a panel is indicative of expected energy yield and is essential for efficient grid distribution and planning. Traditionally, these forecasts are based on meteorological physics models whose parameters are tuned by coarse-grained radiometric tiles sensed from geo-satellites. This research presents a novel application of deep neural network approach to observe and estimate short-term weather effects from videos. Specifically, we use time-lapsed videos (sky-videos) obtained from upward facing wide-lensed cameras (sky-cameras) to directly estimate and forecast solar irradiance. We introduce and present results on two large publicly available datasets obtained from weather stations in two regions of North America using relatively inexpensive optical hardware. These datasets contain over a million images that span for 1 and 12 years respectively, the largest such collection to our knowledge. Compared to satellite based approaches, the proposed deep learning approach significantly reduces the normalized mean-absolute-percentage error for both nowcasting, i.e. prediction of the solar irradiance at the instance the frame is captured, as well as forecasting, ahead-of-time irradiance prediction for a duration for upto 4 hours.

['Talha A. Siddiqui', 'Shivkumar Kalyanaraman', 'Samarth Bharadwaj']

2019-01-15

['solar-irradiance-forecasting']

['time-series']

[-4.25685272e-02 -6.16462111e-01 2.31382266e-01 -6.22762203e-01 -3.35283190e-01 -9.08265829e-01 6.67126715e-01 -3.31909955e-01 -2.38652363e-01 1.12310290e+00 2.34260529e-01 -1.53692141e-01 2.31900234e-02 -1.04986274e+00 -1.01106358e+00 -1.13592994e+00 -1.45818129e-01 -1.05476283e-01 -4.09996837e-01 -2.18059085e-02 2.72714067e-02 7.45789349e-01 -1.72950649e+00 -7.90032223e-02 6.30479336e-01 1.22959065e+00 3.59513760e-01 1.21426380e+00 4.15708572e-01 7.45888054e-01 -4.98536140e-01 2.40592793e-01 5.05417526e-01 -1.12326793e-01 -9.23805013e-02 2.10448369e-01 1.08991802e+00 -9.44066167e-01 -3.01472604e-01 7.86825120e-01 5.53981900e-01 1.94473341e-01 4.92339849e-01 -9.16961730e-01 -4.91015226e-01 -1.24079347e-01 -4.24725860e-01 6.40442669e-01 2.40614265e-01 3.79051507e-01 6.24346495e-01 -6.42957628e-01 3.94341141e-01 4.29065973e-01 6.96361244e-01 -6.10423610e-02 -7.74822652e-01 -3.57208222e-01 -2.13043198e-01 2.91340381e-01 -1.18004560e+00 -4.61925030e-01 4.71445322e-01 -7.12931812e-01 1.11738646e+00 4.46238697e-01 9.92753983e-01 9.06378329e-01 4.15235937e-01 -8.66371840e-02 1.42032981e+00 -3.25114697e-01 2.93178499e-01 -1.17400728e-01 -3.42610925e-01 3.25688839e-01 1.86292008e-01 6.39877379e-01 -4.22887981e-01 2.51589626e-01 6.80401742e-01 -4.11719456e-02 -8.95922720e-01 3.52164596e-01 -1.12960780e+00 6.86457455e-01 5.86473882e-01 -1.51768615e-02 -7.34862149e-01 4.43383753e-02 -5.49879670e-02 1.70848370e-01 8.65758359e-01 -1.49717614e-01 -9.19032931e-01 -1.76870860e-02 -1.27045965e+00 8.52897689e-02 7.36265063e-01 4.77034688e-01 8.14925730e-01 6.32071137e-01 1.39127314e-01 3.88587803e-01 2.24934906e-01 1.21649742e+00 4.29690480e-01 -8.61417055e-01 1.27177611e-02 8.11365619e-02 5.61583340e-01 -9.05595779e-01 -6.52083516e-01 -4.36802119e-01 -9.02051091e-01 5.42640090e-01 2.15873212e-01 -8.52626741e-01 -8.46953511e-01 1.28847969e+00 4.37868029e-01 5.94949424e-01 1.35386467e-01 1.38717663e+00 9.29426730e-01 1.29757798e+00 -3.56719583e-01 -5.16150594e-01 1.23168814e+00 -1.00671661e+00 -5.30991077e-01 -1.70695752e-01 3.10927153e-01 -7.61581838e-01 5.92282534e-01 3.61469537e-01 -6.43279612e-01 -6.56574726e-01 -1.09645987e+00 2.10256893e-02 -6.51468694e-01 4.96443629e-01 3.15813243e-01 5.16819775e-01 -1.40347040e+00 5.52382708e-01 -6.18280292e-01 -3.68060559e-01 4.25719395e-02 -1.03023015e-01 -2.40297779e-01 1.73216701e-01 -9.45314348e-01 6.78330839e-01 1.07600980e-01 5.38658082e-01 -1.27021229e+00 -8.59365880e-01 -5.03369391e-01 3.42780918e-01 3.28749865e-02 -6.93678141e-01 1.23980701e+00 -1.36667943e+00 -1.67034519e+00 3.85480642e-01 -1.39997408e-01 -6.37422323e-01 1.84801310e-01 -3.77415359e-01 -7.21520960e-01 5.73985614e-02 -4.22625810e-01 3.71053129e-01 9.69225049e-01 -9.80055273e-01 -7.63485193e-01 -1.64303973e-01 4.01222020e-01 5.04812777e-01 -2.84050584e-01 -1.33079350e-01 1.57808021e-01 -4.01836067e-01 -6.40474856e-01 -9.14420903e-01 5.64989541e-03 3.06338787e-01 -1.78901732e-01 2.28042603e-01 8.86714458e-01 -1.06266320e+00 5.48188210e-01 -1.73219693e+00 -2.19043776e-01 -4.13689226e-01 -2.73170590e-01 2.67975360e-01 1.19382836e-01 4.14840311e-01 -3.88410091e-01 -2.81447858e-01 -2.67999806e-02 -3.92690040e-02 -3.35462660e-01 2.46954098e-01 -6.13768637e-01 7.05846131e-01 -1.84520289e-01 4.73509014e-01 -5.08882344e-01 5.27842864e-02 7.52541244e-01 7.68348932e-01 1.35056615e-01 4.38116640e-01 -5.45693636e-01 6.20068610e-01 2.02628202e-03 4.40031528e-01 1.01150703e+00 -1.26513928e-01 -4.91824448e-02 -3.76873255e-01 -8.25142980e-01 8.20664614e-02 -9.73242581e-01 1.46854842e+00 -7.40467012e-01 1.34940803e+00 5.30757271e-02 -7.46420503e-01 6.67494714e-01 5.08753598e-01 1.91296190e-01 -7.10417151e-01 -1.21258516e-02 -1.61670163e-01 -6.13000214e-01 -7.25018024e-01 5.62445045e-01 -2.07077876e-01 6.58079386e-01 1.77313268e-01 -3.75165462e-01 -1.06917359e-01 1.22275226e-01 -3.03805113e-01 5.32375693e-01 1.93973675e-01 3.43159258e-01 -3.02010894e-01 4.79407072e-01 2.01769043e-02 5.42931020e-01 3.43611449e-01 1.04790658e-01 5.80679357e-01 -1.99115515e-01 -9.62532580e-01 -1.13628781e+00 -6.73256278e-01 -3.02638799e-01 8.60808313e-01 -3.03864568e-01 2.89487720e-01 -6.37360156e-01 -6.55127317e-02 -1.37340143e-01 9.89511609e-01 -4.77154672e-01 6.27544045e-01 -2.10113242e-01 -1.05921173e+00 -1.05411075e-01 5.35080433e-01 6.37233913e-01 -7.97605813e-01 -1.04515481e+00 -4.67148237e-02 -6.30786046e-02 -1.19334686e+00 -9.39140916e-02 -1.14207767e-01 -7.09693074e-01 -1.02210319e+00 -9.06979382e-01 -2.01401100e-01 3.44526827e-01 6.26415133e-01 1.46539199e+00 -2.05576465e-01 -4.17290241e-01 4.64251459e-01 -2.07978055e-01 -5.64146280e-01 1.24272786e-01 -4.97407347e-01 1.84667259e-01 8.64588916e-02 9.18331072e-02 -6.58937216e-01 -9.62741435e-01 5.69848046e-02 -7.47017860e-01 1.70827344e-01 3.40026200e-01 3.31674784e-01 4.50456172e-01 2.65579373e-01 7.17088394e-03 -1.95905119e-01 7.07476586e-02 -5.80071509e-01 -1.71640682e+00 1.79024488e-01 -4.61260766e-01 -5.96622527e-01 7.18369901e-01 7.80450627e-02 -1.51857173e+00 2.36982424e-02 1.91109911e-01 -4.57011133e-01 -6.24607682e-01 6.80161774e-01 4.98855472e-01 -3.17450136e-01 5.17025471e-01 3.62590760e-01 -7.80061960e-01 -3.43894541e-01 2.32966244e-01 6.18591905e-01 7.11180508e-01 1.66945066e-02 7.89463758e-01 8.29478025e-01 1.87029734e-01 -1.39754105e+00 -8.63773644e-01 -2.59556234e-01 -4.16774541e-01 -6.82885230e-01 1.06150484e+00 -1.68413782e+00 -8.01767886e-01 8.26343358e-01 -9.42103207e-01 -5.35254896e-01 -3.84485275e-02 7.38211095e-01 -1.03702106e-01 1.96469411e-01 -2.09383547e-01 -9.55664754e-01 -6.01122022e-01 -5.64133167e-01 1.11298358e+00 8.59901309e-01 5.61579466e-01 -1.27923763e+00 5.90269685e-01 3.21590155e-01 5.83368063e-01 5.45492232e-01 1.67133927e-01 2.97515541e-01 -8.27111304e-01 -6.18414804e-02 -3.61771673e-01 5.68578839e-01 4.73352820e-02 4.61727440e-01 -1.50839376e+00 -5.11935890e-01 2.47240946e-01 -2.23085895e-01 7.87305355e-01 1.07166219e+00 1.17601359e+00 -3.78634989e-01 5.68701997e-02 1.11002803e+00 2.15191007e+00 9.26310420e-02 5.50317109e-01 3.99983972e-01 6.05635643e-01 1.79547325e-01 3.69581789e-01 9.07343149e-01 3.71250927e-01 4.55631405e-01 8.76630664e-01 -3.14699024e-01 -1.26542151e-02 5.11471748e-01 4.67348695e-01 5.24460256e-01 -4.74079609e-01 -7.80401111e-01 -5.22716641e-01 7.95191884e-01 -1.26842129e+00 -1.07328022e+00 -3.63712102e-01 2.38906765e+00 3.61346602e-01 -4.33948457e-01 -2.34164670e-01 -2.59097219e-01 5.20557761e-01 7.44452417e-01 -6.69412792e-01 -3.70920956e-01 -3.14104319e-01 -6.60806801e-03 1.03930831e+00 5.88984609e-01 -1.17065370e+00 2.38582000e-01 6.34994316e+00 1.97754815e-01 -1.55271435e+00 1.16041377e-01 5.60496628e-01 -5.37977219e-01 -4.69361804e-02 -1.14742881e-02 -7.95197725e-01 6.28156066e-01 1.41037631e+00 3.37034836e-02 8.81434619e-01 7.18051970e-01 8.64891648e-01 -6.41921878e-01 -7.56739438e-01 7.40599573e-01 1.12556465e-01 -1.61653662e+00 -4.27406639e-01 7.96034262e-02 1.25165498e+00 7.80526102e-01 -2.31128812e-01 -3.74918044e-01 1.55544639e-01 -8.84470642e-01 4.72668380e-01 1.19459546e+00 8.42544258e-01 -4.85204488e-01 7.14955151e-01 4.15011704e-01 -1.22007370e+00 -1.55385584e-01 -7.05044866e-01 -4.95652556e-01 3.06451410e-01 1.37307727e+00 -6.23717666e-01 8.23429465e-01 1.21070063e+00 8.94558907e-01 -1.23737454e-01 1.00683260e+00 -2.67128468e-01 7.76466131e-01 -7.32824147e-01 1.90540820e-01 2.43481740e-01 -6.05491757e-01 3.42509538e-01 1.03936529e+00 9.67218816e-01 3.51682961e-01 -3.54845598e-02 4.13639873e-01 -8.24911818e-02 -4.28957820e-01 -6.88928366e-01 1.62053779e-01 2.26722017e-01 1.66043413e+00 -1.89780042e-01 -6.10486805e-01 -8.48888397e-01 6.81749642e-01 -1.18407175e-01 5.30564129e-01 -8.83737743e-01 2.60372385e-02 8.28482330e-01 -1.83741152e-01 7.78236926e-01 -4.28368270e-01 1.74743012e-02 -1.31136978e+00 8.35596621e-02 -3.52946371e-01 1.19239047e-01 -1.75468791e+00 -1.02962136e+00 3.84812564e-01 -9.31408331e-02 -1.26688671e+00 -9.67441350e-02 -7.85236418e-01 -1.07554030e+00 1.14269161e+00 -2.19504690e+00 -9.15772617e-01 -1.13372958e+00 4.53150690e-01 7.96285868e-01 6.56590157e-04 8.89569879e-01 1.59102291e-01 -7.56914258e-01 -3.72049689e-01 1.00450134e+00 -2.37205252e-01 5.73001146e-01 -1.24202859e+00 3.82562697e-01 1.05444515e+00 1.41193628e-01 -2.97216862e-01 9.53205884e-01 -2.73230940e-01 -1.43266106e+00 -1.19339764e+00 6.49939895e-01 -5.35838418e-02 5.29337823e-01 2.54877865e-01 -6.38028204e-01 8.17766964e-01 8.93250406e-01 3.08416516e-01 4.06655282e-01 -3.82108778e-01 9.92347077e-02 -6.52145088e-01 -9.65577364e-01 7.00027402e-03 2.72615075e-01 -5.01618981e-01 -1.73111126e-01 1.03764224e+00 3.48985791e-01 -6.58751428e-01 -1.03157938e+00 2.21094415e-01 6.49175882e-01 -1.45799041e+00 8.10924470e-01 -6.67886883e-02 4.92458969e-01 -5.06684184e-01 -3.29828292e-01 -1.65135729e+00 -2.92994529e-01 -3.78954619e-01 -2.00799659e-01 9.46682572e-01 6.65815324e-02 -5.15432298e-01 6.31588399e-01 2.77013570e-01 -2.17020005e-01 -3.32492769e-01 -8.15745890e-01 -5.91235697e-01 -2.51484424e-01 -2.07658231e-01 6.31029844e-01 8.29754055e-01 -9.29354072e-01 -1.64324239e-01 -6.43735707e-01 1.14227307e+00 5.63313305e-01 5.93379855e-01 7.66285956e-01 -1.14795101e+00 -1.02772683e-01 1.69721857e-01 -5.74924685e-02 -9.04874861e-01 -1.23908855e-01 -6.53043389e-02 1.19041197e-01 -1.42398322e+00 2.34094784e-02 3.35436791e-01 -1.11201733e-01 1.48142263e-01 2.67545640e-01 3.58172923e-01 -1.39840856e-01 1.03273675e-01 2.61861831e-01 5.87371051e-01 8.90200615e-01 5.85315861e-02 3.14276040e-01 -8.60439092e-02 1.79571837e-01 8.50893438e-01 9.01200473e-01 -2.68899947e-01 -2.96742976e-01 -1.13815403e+00 5.55897713e-01 1.89348280e-01 7.83210397e-01 -1.25218916e+00 7.60780647e-02 -4.50570285e-01 8.38981926e-01 -8.68566751e-01 5.36772847e-01 -8.86030197e-01 5.16970396e-01 1.42416209e-01 1.22317977e-01 9.08753648e-02 3.61618310e-01 6.93283260e-01 -5.98208159e-02 -5.57713211e-02 7.06817627e-01 -1.81553230e-01 -1.08941746e+00 2.55049646e-01 -4.85075951e-01 -5.03105879e-01 1.07176149e+00 4.53414433e-02 -7.77916789e-01 -6.01646364e-01 -3.22629184e-01 2.78632753e-02 5.15580833e-01 4.39351462e-02 2.63467640e-01 -9.40349817e-01 -8.99983048e-01 1.88743293e-01 -1.97634026e-01 2.90868171e-02 5.10649264e-01 7.42889404e-01 -9.93799746e-01 6.88333750e-01 -1.89866483e-01 -7.08389580e-01 -1.26332498e+00 4.35131818e-01 6.96375489e-01 1.01929314e-01 -6.63951218e-01 8.77955794e-01 2.56030023e-01 -1.72871530e-01 -2.49451622e-01 -6.60600662e-01 -2.79750675e-01 2.10049413e-02 5.41230977e-01 4.87480789e-01 3.12382966e-01 -6.26750112e-01 -1.48753986e-01 8.58820617e-01 5.24950445e-01 1.97273836e-01 1.46006727e+00 -5.26481092e-01 -7.81031772e-02 5.75568736e-01 1.00891590e+00 -2.36877248e-01 -1.89688456e+00 1.32103130e-01 -8.21328402e-01 -5.79581022e-01 9.22046006e-01 -1.17508411e+00 -1.49744844e+00 9.82842863e-01 1.16413832e+00 3.14211190e-01 1.61226988e+00 -6.41984522e-01 8.18900645e-01 5.75654328e-01 -6.44448772e-02 -9.95786607e-01 -6.71997607e-01 2.89657742e-01 8.23997438e-01 -1.36474144e+00 3.60323817e-01 6.26349077e-02 -2.93526918e-01 1.46598399e+00 2.30364248e-01 -4.50694822e-02 6.98707879e-01 1.44298151e-01 3.35568368e-01 -5.01500517e-02 -8.91102493e-01 8.36148337e-02 2.31994569e-01 3.13071579e-01 3.12615067e-01 2.81170428e-01 2.72674859e-01 -3.69143367e-01 -7.20543182e-03 4.10934746e-01 8.69225323e-01 6.39026165e-01 -6.43963873e-01 -1.24934837e-02 -7.85794795e-01 3.39165032e-01 -1.15713231e-01 -2.33364448e-01 1.64168835e-01 3.59458208e-01 7.77395368e-02 1.09677911e+00 4.90196615e-01 1.02067836e-01 -1.21057168e-01 -2.10715696e-01 2.25539073e-01 9.54226404e-02 -5.08456305e-02 -1.52770862e-01 3.39636087e-01 -4.58043456e-01 -9.64902699e-01 -7.18516707e-01 -4.10274029e-01 -6.18601024e-01 -4.35980409e-01 -3.55650663e-01 1.17889225e+00 1.20339274e+00 3.11666161e-01 4.64400083e-01 1.22284269e+00 -1.64205909e+00 -1.43720761e-01 -1.09257925e+00 -8.05636168e-01 -1.86812937e-01 9.11167741e-01 -4.17792410e-01 -7.50049949e-01 4.96903121e-01]

[6.41500997543335, 2.7090020179748535]

decf02e9-3277-47ad-9e0e-b31a4ee85827

learning-transferable-pedestrian

2304.05554

https://arxiv.org/abs/2304.05554v1

https://arxiv.org/pdf/2304.05554v1.pdf

Learning Transferable Pedestrian Representation from Multimodal Information Supervision

Recent researches on unsupervised person re-identification~(reID) have demonstrated that pre-training on unlabeled person images achieves superior performance on downstream reID tasks than pre-training on ImageNet. However, those pre-trained methods are specifically designed for reID and suffer flexible adaption to other pedestrian analysis tasks. In this paper, we propose VAL-PAT, a novel framework that learns transferable representations to enhance various pedestrian analysis tasks with multimodal information. To train our framework, we introduce three learning objectives, \emph{i.e.,} self-supervised contrastive learning, image-text contrastive learning and multi-attribute classification. The self-supervised contrastive learning facilitates the learning of the intrinsic pedestrian properties, while the image-text contrastive learning guides the model to focus on the appearance information of pedestrians.Meanwhile, multi-attribute classification encourages the model to recognize attributes to excavate fine-grained pedestrian information. We first perform pre-training on LUPerson-TA dataset, where each image contains text and attribute annotations, and then transfer the learned representations to various downstream tasks, including person reID, person attribute recognition and text-based person search. Extensive experiments demonstrate that our framework facilitates the learning of general pedestrian representations and thus leads to promising results on various pedestrian analysis tasks.

['Qi Tian', 'Houqiang Li', 'Wengang Zhou', 'Xiaoyu Qiu', 'Longhui Wei', 'Liping Bao']

2023-04-12

['person-re-identification', 'person-search', 'unsupervised-person-re-identification']

['computer-vision', 'computer-vision', 'computer-vision']

[ 3.67522538e-02 -3.80501360e-01 -7.47777820e-02 -8.19870055e-01 -2.63152242e-01 -1.88417047e-01 6.14114940e-01 -1.79937715e-03 -8.04554164e-01 7.12839186e-01 2.84746975e-01 -5.50710876e-03 2.42268905e-01 -8.38571370e-01 -7.16350496e-01 -8.19741905e-01 1.78713724e-01 7.23996818e-01 -1.81750372e-01 -1.82696328e-01 -2.31639490e-01 3.94114673e-01 -1.74686909e+00 3.95704567e-01 9.82693613e-01 8.85033011e-01 8.57066289e-02 5.76067626e-01 -2.37928227e-01 7.08325207e-01 -5.52952468e-01 -8.20713222e-01 1.51964575e-01 -1.22446105e-01 -9.64143515e-01 4.05623972e-01 7.47214079e-01 -3.13735902e-01 -5.21759152e-01 1.15598094e+00 6.25559986e-01 2.77050912e-01 1.09786630e+00 -1.44171095e+00 -1.12256324e+00 2.78574020e-01 -7.94611692e-01 4.88067687e-01 3.99001360e-01 4.43512052e-01 6.42075360e-01 -1.21265185e+00 2.60522157e-01 1.75566459e+00 8.36069643e-01 7.44069815e-01 -1.34523320e+00 -6.85935140e-01 4.94489700e-01 5.23517549e-01 -1.54112673e+00 -4.61829841e-01 7.07598209e-01 -4.25916344e-01 6.14999175e-01 2.62132019e-01 6.93466008e-01 1.33839178e+00 -2.74827391e-01 1.34762776e+00 1.11511445e+00 -2.60938644e-01 -3.60287309e-01 3.88677955e-01 2.93854356e-01 7.35841036e-01 1.42103136e-01 1.70295164e-01 -2.59293288e-01 2.11360484e-01 6.81729853e-01 1.46652609e-01 9.41219032e-02 -2.42904171e-01 -1.25165927e+00 4.11624670e-01 7.11707115e-01 -1.31893441e-01 -2.26115629e-01 -1.13984220e-01 7.16107547e-01 2.70248294e-01 4.79611486e-01 -1.44884557e-01 -2.42263034e-01 5.31555653e-01 -3.21929783e-01 2.11109474e-01 1.59991279e-01 1.28193355e+00 9.79755819e-01 1.07135579e-01 -4.67010736e-01 1.11887741e+00 3.59524101e-01 8.32865536e-01 2.45889604e-01 -5.18260121e-01 6.48141205e-01 7.21398234e-01 -6.72749579e-02 -5.79313874e-01 -3.69170070e-01 -3.41428697e-01 -1.33316827e+00 5.28074987e-02 5.78529835e-01 -9.67128724e-02 -1.17441857e+00 1.66485071e+00 3.47065479e-01 -8.96501988e-02 6.14196584e-02 9.87491906e-01 1.25860536e+00 5.81907213e-01 9.66834307e-01 2.05378145e-01 1.55164564e+00 -1.05565286e+00 -4.65278417e-01 -3.00613970e-01 2.45231181e-01 -6.08508766e-01 1.13208592e+00 -1.12803310e-01 -9.56833601e-01 -1.26703274e+00 -8.24160755e-01 -5.78387417e-02 -7.11404443e-01 4.94898975e-01 4.39260960e-01 8.05937529e-01 -1.09787607e+00 3.97674330e-02 -2.45647162e-01 -6.46540105e-01 7.47448146e-01 5.95373571e-01 -5.96830070e-01 -2.23466277e-01 -9.77836311e-01 7.16394603e-01 5.52553415e-01 1.72037825e-01 -8.04362476e-01 -3.76252621e-01 -1.01130927e+00 -2.26181038e-02 8.85441527e-02 -1.34686625e+00 8.34111392e-01 -1.13520575e+00 -1.13647532e+00 1.39811122e+00 -3.43263388e-01 -1.87476382e-01 5.87336481e-01 -1.29520848e-01 -5.31140149e-01 -8.96853686e-04 5.68719685e-01 1.11907518e+00 9.30939734e-01 -1.51990891e+00 -9.06386256e-01 -5.31020522e-01 -4.23554629e-02 3.46504331e-01 -5.17654538e-01 1.87729374e-01 -5.12374878e-01 -8.28186333e-01 -2.70080984e-01 -8.05815518e-01 -1.77502885e-01 -7.21239522e-02 -5.77825010e-01 -5.88150263e-01 8.43458056e-01 -4.99328196e-01 5.78398943e-01 -2.06365967e+00 1.28919929e-01 1.24120183e-01 2.84932464e-01 2.49584109e-01 -4.03436601e-01 -1.41992539e-01 -2.31333435e-01 -1.23502687e-01 7.50710368e-02 -6.27418816e-01 -9.76658706e-03 1.70866400e-01 2.23287627e-01 4.95707124e-01 4.16651309e-01 1.22406721e+00 -8.13518584e-01 -9.69931483e-01 5.58245122e-01 4.14985299e-01 -2.69923329e-01 3.81235391e-01 2.36279756e-01 7.50982106e-01 -5.30502677e-01 9.57706928e-01 7.98362374e-01 -1.02704182e-01 -2.86211103e-01 -4.40851718e-01 -6.25402257e-02 -4.22731906e-01 -1.05001915e+00 1.28264701e+00 -1.18416786e-01 4.96686310e-01 3.57135683e-02 -1.21687841e+00 9.16136920e-01 -1.16624422e-01 3.77471447e-01 -8.10861349e-01 2.58228481e-01 -2.68287182e-01 -3.23162407e-01 -4.31824625e-01 4.72338468e-01 3.05978805e-01 -1.61128581e-01 1.98311418e-01 1.56733438e-01 5.62817633e-01 2.17870593e-01 1.55183643e-01 1.41626298e-01 1.43638730e-01 6.88063055e-02 -4.69948232e-01 1.24108601e+00 6.32972736e-03 5.10629177e-01 7.74423659e-01 -5.14680624e-01 2.08789140e-01 -2.80291647e-01 -9.18528914e-01 -1.27607512e+00 -1.31955218e+00 -1.19285852e-01 1.78342533e+00 5.38825333e-01 -1.08085871e-01 -9.33655858e-01 -8.38093817e-01 9.25808325e-02 3.31838399e-01 -6.95622504e-01 -1.66208342e-01 -7.47426569e-01 -1.11321437e+00 6.05774999e-01 9.80568707e-01 1.24466574e+00 -1.01291382e+00 1.88758239e-01 6.76118769e-03 -4.59017843e-01 -1.06992996e+00 -7.60319054e-01 -2.33171806e-01 -3.70976597e-01 -9.98982847e-01 -8.98097575e-01 -1.43016815e+00 1.06552434e+00 4.59336370e-01 1.08788216e+00 1.00380488e-01 -4.70841289e-01 8.24407339e-01 -1.54418707e-01 -4.65759754e-01 -1.26029715e-01 -1.08464062e-01 4.71997917e-01 3.63877684e-01 6.58286929e-01 -1.70818880e-01 -8.19026828e-01 6.95429504e-01 -2.63073713e-01 -3.00251041e-02 7.99354374e-01 1.10809326e+00 7.50265837e-01 -9.89421234e-02 6.17174864e-01 -6.06071711e-01 4.02984679e-01 -1.73534214e-01 -4.43941317e-02 4.68122542e-01 -3.74089509e-01 -9.77015123e-02 6.06368661e-01 -5.54660618e-01 -1.57386339e+00 1.77641302e-01 -2.91270196e-01 -1.86726838e-01 -7.27218211e-01 -1.03866503e-01 -7.08537877e-01 -2.29666725e-01 5.08133411e-01 6.33363128e-01 -1.18591882e-01 -3.37566018e-01 5.39322793e-01 6.52986467e-01 1.05727232e+00 -1.03491700e+00 9.04277802e-01 5.18896580e-01 -2.31444582e-01 -8.63468051e-01 -7.59863734e-01 -6.71745181e-01 -1.16337967e+00 -4.70546901e-01 1.22390795e+00 -1.19527888e+00 -1.00219285e+00 7.50473619e-01 -8.18372726e-01 -1.20322384e-01 -2.50786841e-01 2.16900542e-01 -5.68103135e-01 6.39565468e-01 -6.90956473e-01 -7.36497402e-01 -5.87036729e-01 -1.06411242e+00 9.81661022e-01 5.36964774e-01 1.59173965e-01 -8.32947135e-01 -4.92522657e-01 6.17999136e-01 1.50337085e-01 -2.08204329e-01 8.39990616e-01 -6.59964919e-01 -4.53451037e-01 -7.78205097e-02 -7.60749042e-01 2.68524081e-01 1.09602295e-01 -4.26193446e-01 -1.22450328e+00 -5.27458310e-01 -6.61183178e-01 -4.61754858e-01 1.02769840e+00 4.23182219e-01 1.52128243e+00 -2.38109857e-01 -5.54903090e-01 8.32221568e-01 8.23278546e-01 -9.03415531e-02 5.59380412e-01 7.15837061e-01 1.02060688e+00 6.78170204e-01 6.56418085e-01 2.50229090e-01 9.87933457e-01 6.11207128e-01 1.58003017e-01 -6.12860739e-01 -4.41730678e-01 -3.26581448e-01 1.72773629e-01 3.46026629e-01 -6.83321476e-01 -8.09201896e-02 -7.10842550e-01 4.76138949e-01 -1.93912542e+00 -1.20993125e+00 -1.42501384e-01 1.94048607e+00 4.98150587e-01 -1.07835129e-01 7.51064479e-01 -1.42809495e-01 1.18441617e+00 -1.66555256e-01 -5.75734794e-01 6.72440380e-02 -4.53345299e-01 -2.43070170e-01 4.62226331e-01 8.83308763e-04 -1.84447658e+00 1.11935747e+00 6.14004183e+00 6.83398545e-01 -4.29610759e-01 -1.53707201e-02 9.61039841e-01 5.21880150e-01 1.28867671e-01 -6.02025807e-01 -1.08563447e+00 4.61328954e-01 3.95231456e-01 -1.60807297e-01 4.05775290e-03 9.59673345e-01 -8.40001404e-02 2.67226428e-01 -1.26849365e+00 1.39226794e+00 1.24864399e-01 -1.02089548e+00 6.32026851e-01 -1.15923055e-01 5.97503722e-01 -4.23992038e-01 3.82439762e-01 6.84105277e-01 3.72481048e-01 -9.67616141e-01 5.93442917e-01 5.88622332e-01 8.03729236e-01 -1.00317311e+00 8.26008499e-01 1.75686270e-01 -1.88839841e+00 -2.21505448e-01 -6.01313293e-01 2.43686676e-01 -3.89703512e-02 -2.24106550e-01 -6.89821124e-01 5.88761389e-01 1.06191945e+00 8.65293443e-01 -8.45237076e-01 1.08164823e+00 -5.34782410e-02 2.75853183e-02 1.24331884e-01 -1.02412077e-02 1.93840995e-01 -2.78402448e-01 3.90867889e-01 1.80079818e+00 -1.72583893e-01 2.08661929e-01 7.78856695e-01 7.61242390e-01 -4.04325053e-02 1.66605040e-01 -3.35985124e-01 5.15954137e-01 3.20908517e-01 1.08290732e+00 -4.07645226e-01 -6.42540455e-01 -6.32176459e-01 1.24446452e+00 3.97876590e-01 6.23468339e-01 -6.62782729e-01 -3.11504882e-02 8.72123957e-01 4.14459081e-03 1.20362595e-01 -1.01090640e-01 -2.42818430e-01 -1.12981963e+00 -1.65846780e-01 -6.68917537e-01 9.52008188e-01 -4.99024153e-01 -1.99722993e+00 5.65887570e-01 1.23790152e-01 -1.06533492e+00 -1.79439053e-01 -7.91912377e-01 -7.01439142e-01 1.05207062e+00 -1.62473357e+00 -1.98510599e+00 -6.02374971e-01 1.36589313e+00 8.86829019e-01 -6.98398650e-01 5.61003447e-01 5.57569146e-01 -9.01333094e-01 1.13479996e+00 -2.37548977e-01 7.30046034e-01 8.71724188e-01 -1.39774859e+00 6.16665065e-01 8.07893276e-01 -2.58178830e-01 6.62865877e-01 4.24468875e-01 -7.19883144e-01 -1.27812755e+00 -1.46217155e+00 4.64515656e-01 -5.56061625e-01 3.57148319e-01 -2.64502406e-01 -7.91233361e-01 8.02950442e-01 1.38349414e-01 1.05451569e-01 7.65428662e-01 2.52478689e-01 -3.84031504e-01 -2.51655281e-01 -1.14244723e+00 4.31437790e-01 1.37171400e+00 -5.95314741e-01 -5.62480748e-01 3.38513345e-01 2.77200252e-01 -9.99097973e-02 -8.80582273e-01 4.50128525e-01 3.96386176e-01 -5.19080102e-01 1.79450989e+00 -8.98661435e-01 -9.87177566e-02 -4.13564920e-01 -3.81214340e-04 -1.19276690e+00 -7.68969834e-01 1.09888911e-01 2.90065169e-01 1.50606740e+00 1.34152487e-01 -5.10190785e-01 8.31827402e-01 7.11343348e-01 -2.93073595e-01 -2.01225892e-01 -7.21470535e-01 -7.44598687e-01 6.82902411e-02 5.57441451e-02 6.81535542e-01 9.32475507e-01 -4.01236385e-01 3.80133182e-01 -5.24518609e-01 4.37934816e-01 1.33764803e+00 -5.48867956e-02 1.12679398e+00 -1.34243131e+00 1.68119460e-01 -5.34832776e-01 -5.94397664e-01 -1.30901861e+00 4.26758975e-01 -9.44608688e-01 -9.20869932e-02 -1.31536734e+00 8.25685918e-01 -6.71369553e-01 -3.17944795e-01 3.99735212e-01 -6.42467320e-01 2.77791470e-01 1.61723390e-01 3.67431670e-01 -9.05189514e-01 6.70926750e-01 1.30206239e+00 -9.52788830e-01 -4.23390642e-02 2.34882116e-01 -5.67442238e-01 7.67694831e-01 6.75486982e-01 1.58256412e-01 -2.37546057e-01 -4.99337584e-01 -5.47784865e-01 -5.56811035e-01 8.83164525e-01 -9.93104458e-01 4.39568847e-01 3.89860757e-02 1.42448270e+00 -9.00691152e-01 4.51229811e-01 -5.31201065e-01 -2.63398379e-01 2.58772403e-01 -1.66755617e-01 1.98040232e-01 2.28202865e-01 6.74320638e-01 -1.32235065e-01 1.16543964e-01 8.53168130e-01 -4.02648360e-01 -1.43641353e+00 7.77614653e-01 -3.33958954e-01 -1.10707708e-01 8.70165884e-01 -4.76541162e-01 -4.28922027e-01 -1.76719785e-01 -1.04483414e+00 5.45732558e-01 3.06543320e-01 3.79053682e-01 9.72788036e-01 -1.56146681e+00 -9.88347232e-01 3.25941801e-01 4.17598665e-01 -2.12780148e-01 4.39227879e-01 4.55028474e-01 -2.52658203e-02 1.76329270e-01 -7.45906532e-01 -1.00512135e+00 -1.52227342e+00 8.54265571e-01 3.88463020e-01 2.20332801e-01 -7.12296307e-01 9.73394573e-01 7.85188675e-01 -5.74797213e-01 3.08955491e-01 3.39359879e-01 -6.36252880e-01 -2.23688465e-02 6.89086676e-01 4.50519830e-01 -3.74766111e-01 -1.23336089e+00 -4.71511483e-01 7.15530097e-01 -4.80920941e-01 3.90913934e-01 9.50803578e-01 -5.97884297e-01 3.50161307e-02 -1.52123734e-01 1.02310455e+00 -5.42697847e-01 -1.26509237e+00 -5.45565605e-01 -9.99122187e-02 -5.33102155e-01 -3.27774525e-01 -6.05902791e-01 -8.92888844e-01 7.05889165e-01 9.86944139e-01 -3.81166279e-01 1.05814946e+00 1.21641584e-01 6.27086878e-01 7.02413559e-01 4.25283879e-01 -1.23748088e+00 3.79720837e-01 3.30794394e-01 4.59561288e-01 -1.57985628e+00 6.95243338e-03 -6.41063273e-01 -8.56811047e-01 9.41832185e-01 1.23662865e+00 2.01065689e-01 2.44470522e-01 -1.56165302e-01 -9.88536607e-03 3.73337306e-02 -6.14978722e-04 -7.33359456e-01 5.91102242e-01 1.16616166e+00 1.59358367e-01 2.00515389e-01 3.36190462e-01 5.33846319e-01 -7.27280974e-02 -4.18389291e-01 -7.97150880e-02 5.63306093e-01 -4.74075437e-01 -1.05976057e+00 -6.43678427e-01 3.82442892e-01 6.93210065e-02 -1.22983744e-02 -4.44810122e-01 7.87487447e-01 4.39930201e-01 9.43895519e-01 1.81415841e-01 -4.00791407e-01 4.99015212e-01 -1.13857441e-01 5.57339013e-01 -2.24641860e-01 -4.90539581e-01 -1.87340870e-01 2.87499934e-01 -1.35191798e-01 -6.59443498e-01 -8.20698440e-01 -8.37868452e-01 -6.66503429e-01 -1.36646274e-02 -3.51552218e-02 1.61295354e-01 1.02387714e+00 -2.37881597e-02 4.17873085e-01 5.74506879e-01 -9.80534196e-01 -2.05009505e-02 -7.46284723e-01 -1.96052298e-01 8.50838304e-01 2.25558564e-01 -8.48007500e-01 2.37962916e-01 3.71789962e-01]

[14.592623710632324, 0.9568402171134949]

9b184c38-389c-42a8-b754-cedd19204671

learning-multilingual-sentence

2306.06919

https://arxiv.org/abs/2306.06919v1

https://arxiv.org/pdf/2306.06919v1.pdf

Learning Multilingual Sentence Representations with Cross-lingual Consistency Regularization

Multilingual sentence representations are the foundation for similarity-based bitext mining, which is crucial for scaling multilingual neural machine translation (NMT) system to more languages. In this paper, we introduce MuSR: a one-for-all Multilingual Sentence Representation model that supports more than 220 languages. Leveraging billions of English-centric parallel corpora, we train a multilingual Transformer encoder, coupled with an auxiliary Transformer decoder, by adopting a multilingual NMT framework with CrossConST, a cross-lingual consistency regularization technique proposed in Gao et al. (2023). Experimental results on multilingual similarity search and bitext mining tasks show the effectiveness of our approach. Specifically, MuSR achieves superior performance over LASER3 (Heffernan et al., 2022) which consists of 148 independent multilingual sentence encoders.

['Haifeng Wang', 'Hua Wu', 'Zhongjun He', 'Liwen Zhang', 'Pengzhi Gao']

2023-06-12

['nmt']

['computer-code']

[ 1.35065421e-01 -2.15651006e-01 -6.97281420e-01 -4.97842252e-01 -1.36707234e+00 -6.40946209e-01 7.30661690e-01 -8.53344947e-02 -5.73931217e-01 8.69871378e-01 4.11707103e-01 -9.07434702e-01 2.46902227e-01 -5.29929042e-01 -1.02400875e+00 -5.63473292e-02 3.03894788e-01 7.17515826e-01 -4.16352391e-01 -6.33681893e-01 -2.13367790e-01 -1.50934279e-01 -9.10512388e-01 4.95177597e-01 1.09606481e+00 5.37434936e-01 4.99427408e-01 2.42486805e-01 -2.47988969e-01 8.21544230e-01 -1.63758844e-01 -1.02248788e+00 2.15286255e-01 -3.91452193e-01 -9.04905438e-01 -4.17227507e-01 7.70138264e-01 -2.48391312e-02 -3.35452884e-01 1.13690329e+00 4.46620911e-01 -3.96904111e-01 4.91567343e-01 -6.75724626e-01 -1.08780503e+00 1.41676760e+00 -7.58452594e-01 3.12548459e-01 1.67331532e-01 -9.70362201e-02 1.51236081e+00 -1.13924062e+00 7.80269980e-01 1.36045110e+00 5.13546348e-01 4.33354855e-01 -1.23697233e+00 -8.07786345e-01 -5.74247278e-02 3.77667278e-01 -1.33335054e+00 -7.17474163e-01 5.56834638e-01 -5.67577742e-02 1.73859751e+00 2.00466469e-01 3.41693223e-01 1.43555379e+00 6.77146852e-01 1.12671793e+00 1.16616952e+00 -8.10260475e-01 -4.91122216e-01 1.55313641e-01 2.33129058e-02 8.24468911e-01 1.92906022e-01 3.59870866e-02 -6.26718760e-01 -2.20350493e-02 4.28635359e-01 -4.93790805e-01 2.01450944e-01 -1.34985626e-01 -1.40848041e+00 1.00356984e+00 4.98651005e-02 4.54387426e-01 -3.30507308e-01 -4.28006873e-02 8.68753076e-01 8.95331740e-01 7.41023660e-01 1.86339766e-01 -5.59160948e-01 4.12720032e-02 -8.20559025e-01 -2.30938107e-01 3.49610776e-01 1.17598462e+00 6.02904856e-01 8.55372697e-02 -4.86704558e-02 1.36953437e+00 2.84414053e-01 9.79636669e-01 9.18643653e-01 -5.71196735e-01 1.12613404e+00 3.08245897e-01 -7.32312322e-01 -3.82654011e-01 1.23210154e-01 -4.34013367e-01 -7.97826946e-01 -6.52286828e-01 -2.49418780e-01 -3.67312081e-04 -3.57367605e-01 1.96677470e+00 4.06152233e-02 -3.79039645e-01 4.85324234e-01 6.60671771e-01 5.40820539e-01 6.23624504e-01 2.26830319e-02 -1.87748730e-01 1.44892716e+00 -1.18230748e+00 -6.55411005e-01 -4.26755607e-01 1.06720531e+00 -1.32659030e+00 1.16796994e+00 8.80789757e-02 -1.28538346e+00 -4.90092367e-01 -1.11778092e+00 -5.45047760e-01 -2.15845540e-01 3.16773772e-01 8.16554844e-01 3.46679807e-01 -9.40614760e-01 2.98395246e-01 -7.88339972e-01 -5.60614109e-01 6.64081275e-02 2.88084839e-02 -4.89317924e-01 -4.37489331e-01 -1.56473923e+00 1.41911113e+00 3.03791970e-01 -1.49079233e-01 -4.32515413e-01 -4.17259008e-01 -1.13834488e+00 -3.74754041e-01 -7.67112449e-02 -8.79143000e-01 1.31833684e+00 -7.52791405e-01 -1.47393787e+00 1.11702728e+00 -3.36983979e-01 -6.47479832e-01 1.82196245e-01 -3.78336340e-01 -6.00974679e-01 -2.37554997e-01 5.55927098e-01 6.67322695e-01 4.08474505e-01 -5.22746742e-01 -5.55097997e-01 -4.09977436e-01 -3.10654432e-01 3.28900635e-01 -4.43940282e-01 5.91389000e-01 -5.32739341e-01 -7.82623291e-01 -1.97640881e-01 -8.99999380e-01 -5.11328951e-02 -7.52454102e-01 -5.03678203e-01 -5.32224953e-01 3.01399857e-01 -9.63458896e-01 1.19620419e+00 -1.71330047e+00 3.39545161e-01 -1.28149182e-01 -3.28596085e-01 9.30709839e-02 -6.95442379e-01 6.09082222e-01 -1.81195512e-01 -1.11472264e-01 -1.14880912e-01 -6.23123407e-01 1.76845416e-01 3.81807595e-01 -5.41642189e-01 2.95632482e-01 3.59834164e-01 1.35791337e+00 -8.53385150e-01 -8.31025422e-01 1.07324854e-01 4.17005867e-01 -2.70259738e-01 -3.43767665e-02 9.22851451e-03 2.11464629e-01 -2.16428638e-01 7.08512306e-01 1.74417764e-01 -6.20900132e-02 5.84908366e-01 -2.10415527e-01 -2.73685485e-01 1.09501767e+00 -3.12339991e-01 2.22676206e+00 -1.09720564e+00 4.72647637e-01 -2.89276868e-01 -6.45452976e-01 9.40967619e-01 4.79837239e-01 3.68329793e-01 -1.16820323e+00 3.10266495e-01 7.38873720e-01 -1.30606636e-01 -1.38145059e-01 6.80386424e-01 -3.15228581e-01 -5.85784018e-01 8.39331388e-01 4.23390239e-01 -2.71594450e-02 5.25160849e-01 3.28016639e-01 8.93465757e-01 2.85812646e-01 4.51209635e-01 -3.79555374e-01 5.00315666e-01 5.78275211e-02 4.61342573e-01 3.29647005e-01 2.28078529e-01 1.78900376e-01 -2.01108813e-01 -4.02815014e-01 -1.38157976e+00 -1.08099723e+00 -3.18993956e-01 1.11476576e+00 -4.02370960e-01 -5.78109324e-01 -5.80136836e-01 -7.43561327e-01 -1.31562790e-02 8.72187078e-01 -2.66299427e-01 -2.10124731e-01 -1.06164992e+00 -7.32173264e-01 8.65869761e-01 2.21306637e-01 2.81633466e-01 -1.00562847e+00 4.47082799e-03 4.09068376e-01 -9.18812335e-01 -1.33181274e+00 -9.97596323e-01 3.15759778e-01 -8.91569614e-01 -7.03072250e-01 -4.39204037e-01 -1.07211661e+00 3.34686160e-01 3.02829295e-01 1.45721877e+00 -5.43596566e-01 -1.68719217e-01 -3.87029201e-02 -1.79204449e-01 -7.59700239e-02 -9.28768516e-01 4.16352302e-01 5.84696770e-01 -3.97381723e-01 7.49401450e-01 -6.13405108e-01 6.53254613e-02 -1.13746740e-01 -5.78447461e-01 9.77414325e-02 9.41545248e-01 9.22040641e-01 9.13559377e-01 -6.21215999e-01 6.18713737e-01 -8.09905827e-01 8.78681004e-01 -4.51582968e-01 -6.04242861e-01 7.34045327e-01 -8.04718792e-01 4.53195393e-01 6.91099405e-01 -3.89029622e-01 -8.48852634e-01 -3.25909346e-01 -2.17475936e-01 -4.09798771e-01 4.23342049e-01 7.32203364e-01 -2.27482662e-01 1.40700981e-01 4.11394686e-01 6.15034759e-01 -1.87712014e-01 -7.24194109e-01 8.12010944e-01 9.29217041e-01 5.76958418e-01 -8.39466155e-01 5.81855714e-01 -3.20415944e-01 -4.19302046e-01 -3.84990931e-01 -9.63416636e-01 -2.41501555e-01 -7.44571507e-01 2.40495577e-01 6.39930606e-01 -1.26894188e+00 -2.67290652e-01 5.71286902e-02 -1.38697290e+00 3.20246909e-03 -9.88572612e-02 7.45822370e-01 -5.44313371e-01 4.22402918e-01 -1.25254154e+00 -2.69599915e-01 -1.10649586e+00 -1.09462428e+00 1.16592622e+00 -5.30249178e-01 -3.75423580e-01 -9.97024775e-01 6.24818921e-01 7.18935251e-01 3.74324471e-01 -4.55293387e-01 1.12159038e+00 -5.36450565e-01 -2.21177295e-01 4.44663987e-02 -9.40304920e-02 5.60106456e-01 2.47711450e-01 -4.08032984e-01 -4.56148267e-01 -6.45940959e-01 -8.60838592e-02 -8.46181512e-01 7.54410207e-01 -5.04177157e-03 5.81252217e-01 -5.07666409e-01 -2.85624582e-02 6.49440885e-01 1.32075727e+00 -2.09770694e-01 2.67201334e-01 5.30852854e-01 7.58503318e-01 3.99556935e-01 5.10398388e-01 -1.40395582e-01 9.41156745e-01 7.93760002e-01 -1.17367499e-01 -6.63411990e-02 -4.09160495e-01 -5.83471775e-01 9.42021012e-01 2.22243547e+00 1.77496105e-01 7.49468282e-02 -6.44237339e-01 4.67063189e-01 -1.55505371e+00 -7.95958161e-01 3.08232177e-02 2.07512641e+00 1.54207122e+00 -5.67027070e-02 -3.38921934e-01 -4.32930112e-01 5.25183380e-01 5.68325669e-02 -3.90672982e-01 -8.35382819e-01 -6.39540792e-01 6.03907406e-01 5.94814479e-01 5.65113425e-01 -8.55452597e-01 1.44219005e+00 5.98640871e+00 1.02548456e+00 -1.07415879e+00 4.65705246e-01 1.98714420e-01 2.47304570e-02 -5.85226357e-01 -3.80296656e-03 -1.01047015e+00 3.15318584e-01 1.27274203e+00 -4.15950716e-01 6.22948885e-01 5.23705065e-01 -8.55560377e-02 4.98598844e-01 -1.17726350e+00 9.07497168e-01 3.56246918e-01 -1.18063045e+00 4.26545590e-01 -3.97575796e-02 7.00599492e-01 9.25832748e-01 8.16366747e-02 6.37797475e-01 8.70099127e-01 -8.09304118e-01 7.61667252e-01 9.61148068e-02 1.20204377e+00 -8.29817593e-01 5.10094881e-01 2.98461527e-01 -1.30354822e+00 3.87084067e-01 -5.40775657e-01 3.73335809e-01 3.71498585e-01 5.53822339e-01 -7.68711627e-01 8.87712240e-01 5.02452254e-01 1.04818404e+00 -4.63223100e-01 1.84049442e-01 -4.47576016e-01 7.06804991e-01 -2.71358281e-01 1.95618913e-01 2.73987442e-01 -3.43609661e-01 4.29482698e-01 1.44790423e+00 3.65923941e-01 -5.47648847e-01 1.13770336e-01 5.62606692e-01 -3.94587696e-01 6.57389104e-01 -8.70522916e-01 -2.38881201e-01 6.46115541e-01 1.09865880e+00 1.60548717e-01 -3.13459128e-01 -5.89147329e-01 1.27248263e+00 8.63340855e-01 2.89528556e-02 -6.08038783e-01 -1.19562671e-01 6.29865229e-01 -4.50545341e-01 1.48814037e-01 -2.22059011e-01 -1.55566111e-02 -1.66394567e+00 2.96283484e-01 -1.41069305e+00 3.31500232e-01 -4.81000394e-01 -1.67162311e+00 9.92664874e-01 -2.85992712e-01 -1.25634074e+00 -6.76673532e-01 -4.19452131e-01 -2.28879988e-01 1.16822469e+00 -1.83291149e+00 -1.89077246e+00 8.49524379e-01 7.24654734e-01 8.01242888e-01 -5.06831765e-01 1.07540345e+00 7.18711078e-01 -7.11231589e-01 1.08608222e+00 1.33678496e-01 3.34414452e-01 8.74125421e-01 -9.25237238e-01 9.82178628e-01 8.29607487e-01 7.78717041e-01 9.88271713e-01 1.82439238e-01 -6.54031992e-01 -1.72560024e+00 -1.36652744e+00 1.86154330e+00 -5.52314699e-01 1.29058266e+00 -4.98224348e-01 -8.09533596e-01 1.02770960e+00 7.51274168e-01 -3.68823618e-01 9.55196202e-01 3.95077378e-01 -8.75277996e-01 -5.09543903e-02 -5.08378386e-01 6.75974965e-01 1.04629540e+00 -1.03440094e+00 -8.35870624e-01 4.12331104e-01 9.23040926e-01 -1.55509293e-01 -1.18348777e+00 5.92473328e-01 5.31496584e-01 -2.63195515e-01 9.81746435e-01 -7.39313364e-01 5.77141643e-01 1.20062053e-01 -5.88876188e-01 -1.51184261e+00 -2.29164392e-01 -4.05783325e-01 3.34221907e-02 1.21267748e+00 7.79427767e-01 -6.26328945e-01 9.19210985e-02 -4.83034253e-01 -3.07799548e-01 -6.82127714e-01 -1.25583577e+00 -1.18485975e+00 7.06959009e-01 -4.38571036e-01 5.94024599e-01 1.27284956e+00 3.47789615e-01 1.25692332e+00 -5.52316129e-01 -1.48807347e-01 9.30042446e-01 2.62878656e-01 4.10347521e-01 -7.83673525e-01 -5.25644124e-01 -4.86998379e-01 9.86112878e-02 -1.19077015e+00 6.52665913e-01 -1.75373793e+00 -1.84419662e-01 -1.26122308e+00 5.66714466e-01 -4.03764457e-01 -4.28535551e-01 7.14687288e-01 -1.19684309e-01 2.99945772e-01 -6.08769059e-03 4.89068180e-01 -5.46394050e-01 8.45686078e-01 1.05265129e+00 -4.12671268e-01 3.11391592e-01 -4.36265945e-01 -5.95627725e-01 3.39783907e-01 8.37483644e-01 -5.18797219e-01 -2.20048651e-01 -1.23983502e+00 2.79159963e-01 7.86597207e-02 -2.15747938e-01 -4.16515768e-01 1.80844694e-01 -5.53119404e-04 -4.42624018e-02 -6.32853687e-01 2.23603353e-01 -4.71110880e-01 2.55736001e-02 3.69054288e-01 -5.87621808e-01 7.68589795e-01 1.74550653e-01 7.93246739e-03 -3.93084884e-01 7.12098256e-02 4.70491976e-01 -2.66394541e-02 -3.09046894e-01 2.42427126e-01 -1.41585782e-01 2.27473944e-01 4.36323404e-01 4.59420830e-01 -4.61539209e-01 1.46675259e-01 -1.06616929e-01 2.39669949e-01 1.89317629e-01 9.63342011e-01 4.07645553e-01 -1.75336325e+00 -1.34415710e+00 2.91880757e-01 4.70751375e-01 -6.37662649e-01 -2.86381125e-01 8.78165185e-01 2.18276158e-02 8.80804539e-01 -8.19493756e-02 -3.22134078e-01 -1.27425897e+00 4.06010896e-01 1.42201528e-01 -5.59998155e-01 -4.68583643e-01 6.80573165e-01 -1.50992483e-01 -1.04835606e+00 -2.18087792e-01 -5.44783324e-02 9.04152729e-03 -3.29735279e-01 2.96420336e-01 -3.45407948e-02 4.07490641e-01 -9.58776236e-01 -3.76490682e-01 4.48687822e-01 -5.95774829e-01 -2.55024344e-01 1.21519637e+00 -3.78917664e-01 -6.25713050e-01 5.91128647e-01 1.40300548e+00 1.81978092e-01 -3.31109911e-01 -8.30359638e-01 2.61283040e-01 -3.08809616e-02 -2.53567398e-02 -7.97730744e-01 -8.67605150e-01 9.42065835e-01 1.18833452e-01 -6.80324912e-01 8.63407791e-01 1.42655924e-01 1.44444036e+00 6.57523870e-01 8.00885856e-01 -1.00820494e+00 -3.24984819e-01 9.99631524e-01 7.54355013e-01 -1.37261665e+00 -3.53576034e-01 6.23316690e-03 -4.46538180e-01 8.83777440e-01 4.19461280e-01 1.88484102e-01 2.14371726e-01 3.62023383e-01 2.99327761e-01 1.22158583e-02 -1.26106167e+00 6.29548952e-02 4.61643010e-01 6.28716350e-02 8.85013103e-01 3.73993933e-01 -6.82221234e-01 4.94864285e-01 -6.16052806e-01 -2.07635432e-01 -6.99598417e-02 7.28987336e-01 -1.71431795e-01 -1.83062375e+00 -5.69959590e-03 2.29901448e-01 -4.08841968e-01 -1.00762308e+00 -3.51516545e-01 5.12664378e-01 -1.27876014e-01 8.18187833e-01 -9.99169424e-02 -5.49867511e-01 2.33854219e-01 1.56767130e-01 7.84391165e-01 -6.27624929e-01 -6.62878215e-01 7.88912326e-02 3.79678905e-01 -4.20020401e-01 -2.00442702e-01 -9.13322687e-01 -8.05983424e-01 -3.41275603e-01 -1.65488452e-01 3.06654334e-01 7.87251055e-01 9.56398666e-01 4.24103498e-01 2.44599864e-01 8.56362402e-01 -1.00470185e-01 -8.20463300e-01 -1.33635378e+00 -8.38990659e-02 3.57694365e-02 -6.76744729e-02 -1.04694940e-01 -2.89937388e-02 1.93132274e-02]

[11.357203483581543, 10.091445922851562]

c17fdae4-f0a8-4fb5-abf8-380aea8c285b

a-review-of-probabilistic-forecasting-and

2209.08307

https://arxiv.org/abs/2209.08307v1

https://arxiv.org/pdf/2209.08307v1.pdf

A review of probabilistic forecasting and prediction with machine learning

Predictions and forecasts of machine learning models should take the form of probability distributions, aiming to increase the quantity of information communicated to end users. Although applications of probabilistic prediction and forecasting with machine learning models in academia and industry are becoming more frequent, related concepts and methods have not been formalized and structured under a holistic view of the entire field. Here, we review the topic of predictive uncertainty estimation with machine learning algorithms, as well as the related metrics (consistent scoring functions and proper scoring rules) for assessing probabilistic predictions. The review covers a time period spanning from the introduction of early statistical (linear regression and time series models, based on Bayesian statistics or quantile regression) to recent machine learning algorithms (including generalized additive models for location, scale and shape, random forests, boosting and deep learning algorithms) that are more flexible by nature. The review of the progress in the field, expedites our understanding on how to develop new algorithms tailored to users' needs, since the latest advancements are based on some fundamental concepts applied to more complex algorithms. We conclude by classifying the material and discussing challenges that are becoming a hot topic of research.

['Georgia Papacharalampous', 'Hristos Tyralis']

2022-09-17

['additive-models']

['methodology']

[-2.30757892e-02 5.19358180e-02 -3.67650449e-01 -8.54093015e-01 -8.40072334e-01 -5.67552686e-01 7.19169319e-01 2.96728849e-01 -5.51359951e-02 1.02225041e+00 1.77905470e-01 -3.99970025e-01 -6.29725099e-01 -8.85266125e-01 -5.50235629e-01 -8.93264413e-01 -4.12724346e-01 7.19655573e-01 1.40956029e-01 -4.03370820e-02 4.79304820e-01 4.68698859e-01 -1.95375764e+00 3.26668173e-01 8.85000944e-01 1.53229856e+00 -5.69381043e-02 6.43915951e-01 -3.85287046e-01 6.43360257e-01 -4.83452410e-01 -6.67628706e-01 -2.82993197e-01 1.57360703e-01 -3.94109994e-01 -4.38810349e-01 1.83073521e-01 3.74040753e-03 2.25100100e-01 5.00143409e-01 3.67509723e-01 1.52364165e-01 1.28616488e+00 -1.62138677e+00 -6.74871981e-01 8.58813167e-01 -3.20966423e-01 7.16810003e-02 1.22691058e-01 -5.86542726e-01 9.63077724e-01 -7.78703392e-01 1.11973509e-02 1.08557451e+00 1.29892004e+00 1.46835417e-01 -1.03603137e+00 -5.68974793e-01 3.40256989e-01 5.48534095e-01 -1.24078310e+00 -5.98960519e-02 5.03831029e-01 -8.61662269e-01 8.47301483e-01 3.52608740e-01 2.41274655e-01 1.01146126e+00 5.66795051e-01 7.28046596e-01 1.02254581e+00 -5.57361126e-01 6.42078936e-01 5.40601730e-01 2.48829469e-01 5.61808422e-02 1.27281979e-01 3.38582247e-01 -7.13114142e-01 -4.73914027e-01 1.63818359e-01 1.48700520e-01 2.75813788e-01 -5.39173424e-01 -8.88228714e-01 1.17034256e+00 1.12164378e-01 4.04845238e-01 -5.83160400e-01 -3.46897752e-03 2.68689066e-01 -6.11464716e-02 1.02567935e+00 -2.58300230e-02 -9.65414345e-01 -2.47943416e-01 -1.37522042e+00 5.16493320e-01 9.74009454e-01 7.44872093e-01 4.87269402e-01 6.92470595e-02 -1.63497820e-01 8.39117050e-01 5.45991659e-01 5.83813429e-01 2.58594364e-01 -7.21270680e-01 2.71439031e-02 2.01452728e-02 4.83500421e-01 -1.07759798e+00 -7.72390902e-01 -4.30490494e-01 -8.79981339e-01 1.40822485e-01 4.13004667e-01 -4.06567782e-01 -5.62561393e-01 1.30080271e+00 2.76430488e-01 9.59889367e-02 -1.93860680e-01 4.13970262e-01 7.68571973e-01 8.69273961e-01 3.21436286e-01 -3.07465106e-01 1.12110960e+00 -5.40092409e-01 -6.73685849e-01 -7.42352232e-02 5.46728432e-01 -7.79815376e-01 4.94575322e-01 6.00993395e-01 -5.02306640e-01 -4.51326758e-01 -7.31862783e-01 2.52616048e-01 -8.31042171e-01 -6.41154721e-02 8.00399005e-01 1.13690972e+00 -9.04027343e-01 8.02975714e-01 -8.46572280e-01 -1.61687046e-01 3.40099871e-01 2.24793360e-01 3.40723060e-02 1.15461707e-01 -1.51903176e+00 1.29225087e+00 2.33452737e-01 2.68786494e-02 -2.59681910e-01 -7.78546393e-01 -7.93399572e-01 -5.47008701e-02 -2.13439330e-01 -4.91319805e-01 1.38824785e+00 -7.12465763e-01 -1.60743463e+00 4.15941983e-01 -1.01060726e-01 -5.18913627e-01 5.55204630e-01 -4.24554914e-01 -4.95706916e-01 -7.23299742e-01 -1.54862538e-01 3.99220437e-01 6.51468098e-01 -1.02675259e+00 -1.15797007e+00 -4.81146038e-01 -5.76757133e-01 -2.22635657e-01 -9.96678770e-02 2.06832916e-01 4.61340874e-01 -5.27770698e-01 7.73743019e-02 -7.40493298e-01 -4.97424364e-01 -3.94335330e-01 -2.46459857e-01 -6.11074030e-01 5.87245882e-01 -7.82085896e-01 1.30356824e+00 -1.78073537e+00 -1.74462423e-01 3.00773233e-01 -2.32668594e-01 -3.47290576e-01 2.54575998e-01 7.30713844e-01 -1.12980448e-01 -1.01773888e-01 -3.29416096e-01 -4.60812747e-01 2.27769017e-01 1.23439707e-01 -9.22175825e-01 4.62519914e-01 -1.32768407e-01 6.71197653e-01 -7.04411626e-01 -2.71429181e-01 5.30695140e-01 6.26049399e-01 -1.26336291e-01 4.74191718e-02 -3.76096606e-01 1.91067338e-01 -3.14027637e-01 7.20764160e-01 8.54011178e-01 -1.75130442e-01 -1.23734763e-02 1.44839719e-01 -4.20668513e-01 3.19739878e-01 -1.19099355e+00 9.18697238e-01 -6.05092883e-01 6.11335695e-01 -3.83264184e-01 -1.13965690e+00 1.08534229e+00 2.76107311e-01 5.67118704e-01 3.83148505e-03 -4.67802100e-02 3.49911094e-01 -4.58673567e-01 -4.77360226e-02 4.72055405e-01 -1.76445082e-01 -2.61639416e-01 2.49829680e-01 1.01696357e-01 -1.81618303e-01 -1.76078588e-01 -3.16668272e-01 5.31160235e-01 4.28238362e-01 6.03874087e-01 -1.75940201e-01 1.47247285e-01 3.26567627e-02 1.60869509e-01 9.56870258e-01 -1.49857253e-01 3.28469902e-01 2.86715627e-01 -7.53774405e-01 -9.10957992e-01 -1.07437980e+00 -6.53276503e-01 1.81386840e+00 -4.01440561e-01 -1.90404430e-01 -4.72107887e-01 -4.60349381e-01 4.14472669e-01 1.22420275e+00 -7.11783051e-01 3.03906649e-01 4.73909490e-02 -1.33958817e+00 -5.31715751e-02 7.71463752e-01 -1.91296309e-01 -9.24489856e-01 -3.13346952e-01 3.17277223e-01 -1.05831742e-01 -9.00777698e-01 3.93736750e-01 4.39060569e-01 -1.14165056e+00 -5.26137352e-01 -6.54750168e-01 -2.27049187e-01 -2.99940147e-02 -2.60560870e-01 1.22225535e+00 -6.16974413e-01 1.77828610e-01 4.39481497e-01 -3.39861184e-01 -1.03102863e+00 -7.71371797e-02 1.14083089e-01 3.20991278e-01 -1.95823327e-01 6.30770802e-01 -8.48104775e-01 -2.40205288e-01 3.30801159e-01 -4.92293715e-01 -1.45241648e-01 5.73869765e-01 7.59931862e-01 4.78486389e-01 7.13737831e-02 7.04155505e-01 -9.62343693e-01 4.79101837e-01 -9.51035559e-01 -7.04913378e-01 1.84470907e-01 -9.62948143e-01 -1.35732973e-02 2.99232751e-01 -2.30357766e-01 -1.03935921e+00 4.77587394e-02 -3.07466775e-01 6.76557645e-02 -3.55362654e-01 6.10848725e-01 2.15542912e-01 1.41969755e-01 9.23425972e-01 3.36321920e-01 -2.30883718e-01 -5.92722058e-01 6.97578192e-01 9.86455619e-01 2.33310610e-01 -4.87225473e-01 4.50892538e-01 3.64009142e-01 8.84441957e-02 -6.78052723e-01 -1.16870427e+00 -3.76887709e-01 -9.12548423e-01 -4.66681719e-01 4.75329429e-01 -6.59744859e-01 -4.38925862e-01 3.44683170e-01 -1.14721584e+00 -2.66444199e-02 -1.84894040e-01 6.80797279e-01 -8.52869630e-01 3.76700759e-02 -2.64064252e-01 -1.63483667e+00 -3.79253000e-01 -7.34834552e-01 1.11765242e+00 2.53925174e-01 -4.15330917e-01 -1.36953747e+00 4.57712531e-01 2.21184641e-01 7.34543800e-01 2.59748101e-01 8.33233953e-01 -1.07256424e+00 -2.08763957e-01 -5.61496198e-01 -7.96051025e-02 3.12770456e-01 -2.12227821e-01 3.93775880e-01 -1.32404721e+00 3.85432810e-01 -3.30105990e-01 -2.36857925e-02 8.44476223e-01 1.08026779e+00 1.29333520e+00 -1.95586473e-01 -7.78885901e-01 1.12032004e-01 1.02856588e+00 9.49701443e-02 3.67533386e-01 2.91672438e-01 1.64448649e-01 1.10914004e+00 6.52796388e-01 8.02170336e-01 6.10116482e-01 5.01010954e-01 4.31098163e-01 4.32453215e-01 4.92147565e-01 -1.55332774e-01 2.89396733e-01 6.02813840e-01 -4.85368222e-01 -1.06113583e-01 -9.91357565e-01 3.56095672e-01 -2.14269733e+00 -1.01076877e+00 -2.88838953e-01 2.40753794e+00 6.16447568e-01 1.00426458e-01 2.71066338e-01 1.33551359e-01 8.23218465e-01 2.77085938e-02 -3.47871989e-01 -5.18015027e-01 6.69736788e-02 -3.64343338e-02 6.58243299e-01 3.86443883e-01 -1.44098008e+00 6.44513071e-01 8.11857224e+00 1.20496321e+00 -1.16201973e+00 1.34407654e-01 1.02170885e+00 1.87585980e-01 -3.86472642e-01 -1.05256118e-01 -1.07531226e+00 6.23244226e-01 1.48661697e+00 3.91339026e-02 1.29905060e-01 1.49836183e+00 1.70128688e-01 -1.19814664e-01 -1.04941940e+00 7.35315621e-01 -2.52292544e-01 -1.41744244e+00 -4.61143970e-01 -6.13068305e-02 7.80338228e-01 3.71683478e-01 3.31599623e-01 6.12208247e-01 5.34697354e-01 -1.09098268e+00 7.39963531e-01 1.09721184e+00 3.95941049e-01 -8.59006941e-01 1.23091149e+00 5.41166127e-01 -7.01551259e-01 -1.09351024e-01 -4.43112850e-01 -4.34843779e-01 1.88534111e-01 1.44799173e+00 -6.75778389e-01 4.79747802e-01 1.14181411e+00 6.65906310e-01 -1.99841738e-01 1.22185290e+00 8.32850337e-02 8.42052221e-01 -6.22014761e-01 -6.09523535e-01 -5.56896999e-02 -3.84754688e-02 3.42479378e-01 1.33319795e+00 6.83246732e-01 -2.83608109e-01 -3.54547822e-03 6.55484736e-01 5.55611014e-01 3.07474464e-01 -4.54436183e-01 1.59228757e-01 5.10544419e-01 1.20383739e+00 -4.17234093e-01 -3.00692674e-03 -5.19836485e-01 2.76693702e-01 8.22107941e-02 9.38252732e-02 -7.84740508e-01 -1.35614470e-01 4.97407824e-01 2.98925102e-01 1.95484728e-01 -1.14126466e-01 -8.00788164e-01 -6.52240753e-01 -3.20861012e-01 -2.36590087e-01 5.57534575e-01 -8.61625433e-01 -1.89260602e+00 5.64839721e-01 4.09693092e-01 -1.14099491e+00 -7.02609181e-01 -9.35402036e-01 -5.48699915e-01 8.75842869e-01 -1.32026088e+00 -1.20097101e+00 1.29457787e-01 6.23064935e-02 1.87211826e-01 -2.91067243e-01 1.15234506e+00 -5.37926815e-02 -1.58030972e-01 2.03708902e-01 9.26053822e-01 -5.57359636e-01 8.03746045e-01 -1.31187773e+00 1.95432305e-01 2.03658208e-01 1.32866278e-01 3.15119833e-01 1.07146096e+00 -3.93226177e-01 -6.16701245e-01 -9.30322587e-01 1.20453930e+00 -9.31702435e-01 8.73035967e-01 -2.18297362e-01 -6.34392858e-01 6.16100192e-01 -9.03359614e-03 -1.93447575e-01 1.11864662e+00 8.81878674e-01 -3.40450644e-01 -3.03887129e-01 -1.27087271e+00 2.45514497e-01 2.73488462e-01 -2.18526050e-02 -3.29204589e-01 5.75300157e-01 4.36555266e-01 -2.10872903e-01 -9.70608115e-01 6.84134364e-01 1.12351084e+00 -1.06083131e+00 7.88538635e-01 -6.67537034e-01 4.07921612e-01 9.57330391e-02 -4.81563181e-01 -1.32393003e+00 -4.96898979e-01 -2.97183722e-01 -4.61900920e-01 1.06755280e+00 7.19295740e-01 -7.66054690e-01 8.12323809e-01 9.00569499e-01 1.21985963e-02 -1.08754957e+00 -1.20058322e+00 -4.62341845e-01 4.37825710e-01 -1.13045812e+00 4.53268707e-01 6.22543812e-01 1.50855735e-01 4.84278426e-03 -6.30747616e-01 3.44048664e-02 5.41510105e-01 2.31229797e-01 5.00583887e-01 -1.90352404e+00 1.07797921e-01 -5.92939019e-01 -6.38250053e-01 -7.53098488e-01 -8.13218672e-03 -3.06127101e-01 7.80276582e-02 -1.53402722e+00 1.11710034e-01 -5.26859045e-01 -4.50406492e-01 3.39502901e-01 -8.10821075e-03 -2.00333893e-02 -3.56256843e-01 1.78645521e-01 -4.99720991e-01 4.81277972e-01 4.70867306e-01 1.26960531e-01 -4.73155715e-02 1.01085484e+00 -6.53357744e-01 1.05065954e+00 7.66427398e-01 -4.57260728e-01 -2.19554290e-01 2.32696339e-01 5.90505958e-01 -2.09037885e-02 1.94611117e-01 -8.97690713e-01 1.10774897e-01 -4.57107216e-01 7.36740172e-01 -1.02892601e+00 2.30252013e-01 -8.36157560e-01 1.55420348e-01 4.22092862e-02 -2.55179852e-01 5.23872897e-02 2.63522536e-01 7.02540576e-01 -2.38013685e-01 -4.01142299e-01 5.82330465e-01 2.79102057e-01 -6.63255632e-01 1.55824274e-01 -5.18429339e-01 -4.29021001e-01 1.15047455e+00 -2.96860617e-02 7.20092952e-02 -9.06727254e-01 -1.01296949e+00 2.51366869e-02 -2.20172375e-01 3.63046676e-01 2.28483766e-01 -1.15018034e+00 -8.86006534e-01 -3.06393385e-01 1.41701519e-01 -3.99589866e-01 5.73454857e-01 7.36474037e-01 -2.54493326e-01 8.52920175e-01 -1.77798979e-02 -6.61741316e-01 -8.16462696e-01 5.13129830e-01 2.19909698e-01 -4.01369750e-01 -1.95342135e-02 9.62004542e-01 -1.41038820e-01 -7.54264057e-01 3.48653466e-01 -2.75447696e-01 -6.01029813e-01 1.72937065e-01 5.99305212e-01 7.23775744e-01 2.26780400e-01 -4.12655711e-01 -5.22584260e-01 3.46991420e-01 1.78463548e-01 -9.47202891e-02 1.53616786e+00 -1.84620425e-01 -1.43523648e-01 8.98840129e-01 5.89911282e-01 -3.05060130e-02 -1.00434077e+00 -1.26764014e-01 3.37435782e-01 -8.82446617e-02 1.84741214e-01 -1.06386435e+00 -5.15173733e-01 9.33844924e-01 7.11283922e-01 6.71613514e-01 8.92075181e-01 1.15729906e-01 3.40782136e-01 1.68382838e-01 5.67586064e-01 -1.05674016e+00 -6.05928659e-01 6.19973958e-01 9.85642314e-01 -1.44617486e+00 1.16535664e-01 -3.61347377e-01 -6.52106106e-01 1.29420757e+00 1.40801892e-01 6.85182586e-02 1.58949244e+00 4.07318175e-01 -1.82747096e-01 4.12626922e-01 -8.60321045e-01 1.78818889e-02 7.39533961e-01 1.05431128e+00 8.48716021e-01 4.75470722e-01 -3.19368452e-01 1.36549544e+00 -5.78198373e-01 5.24259657e-02 -1.27602834e-02 4.34217870e-01 -7.26959050e-01 -1.11198151e+00 -5.99227965e-01 8.89236927e-01 -4.73600119e-01 -1.85515001e-01 -3.09433192e-02 4.41374928e-01 2.58536994e-01 1.10483730e+00 9.08378661e-02 -4.69424188e-01 1.54385984e-01 4.07945156e-01 1.24567479e-01 -3.75446588e-01 1.30817052e-02 -2.87153512e-01 9.55034941e-02 -1.70031190e-01 -3.19203138e-01 -9.13434088e-01 -6.06040180e-01 -4.48586732e-01 -4.72650379e-01 2.07462192e-01 1.25008297e+00 1.10046220e+00 2.42208764e-01 3.30554277e-01 4.33978856e-01 -1.16613281e+00 -7.06027985e-01 -1.40066493e+00 -9.42181289e-01 -4.87852097e-01 1.68728065e-02 -9.78850245e-01 -4.33723003e-01 1.53114513e-01]

[7.677577018737793, 3.9699289798736572]

a37d72de-743e-4b70-b12b-dc6d20526a69

exploring-transformers-for-ranking-portuguese

https://aclanthology.org/2022.lrec-1.275

https://aclanthology.org/2022.lrec-1.275.pdf

Exploring Transformers for Ranking Portuguese Semantic Relations

We explored transformer-based language models for ranking instances of Portuguese lexico-semantic relations. Weights were based on the likelihood of natural language sequences that transmitted the relation instances, and expectations were that they would be useful for filtering out noisier instances. However, after analysing the weights, no strong conclusions were taken. They are not correlated with redundancy, but are lower for instances with longer and more specific arguments, which may nevertheless be a consequence of their sensitivity to the frequency of such arguments. They did also not reveal to be useful when computing word similarity with network embeddings. Despite the negative results, we see the reported experiments and insights as another contribution for better understanding transformer language models like BERT and GPT, and we make the weighted instances publicly available for further research.

['Hugo Gonçalo Oliveira']

lrec-2022-6

['word-similarity']

['natural-language-processing']

[-1.05292210e-02 6.45066500e-01 -2.35287428e-01 -2.26244569e-01 -1.23647936e-01 -7.33664572e-01 1.08137536e+00 9.49191988e-01 -6.63450599e-01 7.18636811e-01 8.16615939e-01 -6.61105573e-01 -8.38958263e-01 -1.15231335e+00 -4.40238416e-01 -6.16895854e-01 -4.36240703e-01 7.43236840e-01 5.43486595e-01 -5.63476503e-01 4.48723584e-01 4.03353631e-01 -1.43924797e+00 5.34363210e-01 5.81823409e-01 5.88959515e-01 6.87509328e-02 1.31236240e-01 -1.95760414e-01 1.09520280e+00 -6.13652408e-01 -1.00941920e+00 -1.11232288e-01 -4.42499876e-01 -1.12266231e+00 -3.37210268e-01 -2.50624985e-01 2.97529906e-01 -2.51810282e-01 9.69577909e-01 3.47027451e-01 4.55388017e-02 8.48598957e-01 -9.25111055e-01 -6.41704023e-01 1.11783290e+00 -6.90024421e-02 5.79239905e-01 5.82123995e-01 -1.54261380e-01 1.65610743e+00 -5.33295333e-01 9.04505789e-01 1.44327545e+00 5.99461913e-01 -6.48253644e-03 -9.58857179e-01 -2.93287873e-01 1.53981507e-01 3.86728793e-01 -8.81415606e-01 -5.04789710e-01 4.61827040e-01 -2.29368687e-01 1.30453181e+00 4.14452881e-01 8.84054661e-01 9.85267282e-01 1.44158497e-01 3.20973635e-01 1.26308060e+00 -4.90243196e-01 -3.95957008e-02 5.13611615e-01 2.14792844e-02 5.17325938e-01 6.75437927e-01 9.65192169e-02 -5.94981909e-01 -4.15557086e-01 2.83986956e-01 -6.14612341e-01 -1.66517332e-01 -2.25852370e-01 -1.06434977e+00 1.11880016e+00 4.85304654e-01 1.05814862e+00 -3.02127331e-01 -3.64173859e-01 6.68754637e-01 5.87453067e-01 5.85204303e-01 7.47574270e-01 -6.83718979e-01 -3.60223770e-01 -4.00377125e-01 1.05007365e-01 8.65327477e-01 7.36471653e-01 3.75244975e-01 -3.99220228e-01 1.77994058e-01 8.74849081e-01 1.22350186e-01 -7.37462938e-02 6.25931919e-01 -6.97949409e-01 4.20292109e-01 8.21445763e-01 -1.37593821e-01 -1.27608311e+00 -4.95140165e-01 -3.94961834e-01 -9.09288302e-02 -3.21092963e-01 5.98939121e-01 2.22795963e-01 -5.36383614e-02 1.59389722e+00 -1.37734130e-01 -5.06949067e-01 1.18204594e-01 9.27159667e-01 5.98492801e-01 3.00611973e-01 5.12319952e-02 -3.24125797e-01 1.73466516e+00 -2.38149896e-01 -6.54514492e-01 -2.96827435e-01 8.00820589e-01 -9.52194691e-01 1.17199910e+00 1.25355134e-02 -9.17209327e-01 -8.83865580e-02 -7.84642220e-01 2.17799563e-02 -6.26952112e-01 -5.51549554e-01 1.04568803e+00 7.19309270e-01 -7.88037956e-01 8.01700711e-01 -5.78112781e-01 -9.11919594e-01 1.63162142e-01 -7.15843663e-02 -9.47917700e-02 2.69218087e-01 -1.50510740e+00 1.42012262e+00 4.70361173e-01 2.79273670e-02 -2.08805904e-01 -2.30246931e-01 -7.51798391e-01 1.54278114e-01 3.72928828e-01 -2.07865268e-01 6.68824494e-01 -6.86995387e-01 -1.00488615e+00 1.13594401e+00 -1.48773134e-01 -5.84390938e-01 4.32535589e-01 1.61929443e-01 -6.38146996e-01 7.10366014e-03 1.11826807e-01 2.13266775e-01 3.13574255e-01 -1.02498996e+00 -5.68938434e-01 -1.95733741e-01 4.07467961e-01 8.68457854e-02 -4.02981818e-01 4.85270441e-01 2.49335989e-01 -4.58592027e-01 9.82489586e-02 -6.12390995e-01 5.76281957e-02 -3.96546632e-01 -1.38903931e-01 -5.69438517e-01 1.69619218e-01 -4.92911845e-01 1.34172857e+00 -1.94070375e+00 -1.41734965e-02 3.19388956e-01 -5.86329550e-02 2.19703108e-01 -1.92272231e-01 9.63169813e-01 -2.69060910e-01 5.18249750e-01 5.06363297e-03 6.02030218e-01 -4.65854034e-02 3.29894394e-01 -3.14373165e-01 4.32147920e-01 4.05389458e-01 8.33962321e-01 -1.18989050e+00 -5.01805365e-01 8.76398459e-02 2.97451884e-01 -3.78225595e-01 -2.36822709e-01 -7.64869079e-02 -2.57315673e-02 -3.76060903e-01 2.01801226e-01 1.83877107e-02 -7.26794228e-02 5.85994065e-01 -4.29533333e-01 -9.47907418e-02 1.31130433e+00 -7.78881729e-01 9.12931383e-01 -3.32258344e-01 8.07101727e-01 -3.40941042e-01 -1.29826760e+00 8.71582627e-01 4.01595950e-01 1.35460556e-01 -9.28610146e-01 1.92406937e-01 3.76846939e-01 7.72484004e-01 -7.08741188e-01 5.34357488e-01 -4.01085258e-01 1.46272153e-01 5.03343761e-01 -1.60011038e-01 -2.63070196e-01 6.40055895e-01 2.41593808e-01 1.00326192e+00 8.81927013e-02 4.35685396e-01 -5.49567044e-01 4.49727297e-01 5.28384782e-02 4.29815829e-01 4.04425144e-01 1.03375554e-01 1.86361641e-01 9.97825503e-01 -2.89323509e-01 -8.61824274e-01 -8.80508304e-01 -5.46213090e-01 8.49886179e-01 -3.86371501e-02 -9.10333753e-01 -2.71667391e-01 -7.55019307e-01 -1.93252340e-01 1.02876258e+00 -4.74763244e-01 -1.59710065e-01 -3.83987308e-01 -8.16848457e-01 4.78487283e-01 2.44180471e-01 -1.25996858e-01 -1.13880932e+00 -6.64460003e-01 2.14474484e-01 -2.14022234e-01 -9.62393284e-01 -2.51249806e-03 5.01429737e-01 -8.70818913e-01 -1.47073936e+00 -1.76046282e-01 -6.10687792e-01 4.59710687e-01 -5.74032366e-02 1.28279543e+00 6.13618314e-01 3.20718512e-02 1.23377621e-01 -5.90442777e-01 -2.36945465e-01 -6.59058988e-01 3.77841778e-02 -3.07126604e-02 -4.93954778e-01 6.14466846e-01 -7.47702897e-01 -1.68442294e-01 1.55915469e-01 -8.07432473e-01 -6.48375511e-01 3.43874544e-01 9.64723825e-01 -6.17205538e-02 3.72406065e-01 3.18860948e-01 -1.16066968e+00 1.19591343e+00 -5.24554253e-01 -2.08425209e-01 2.44150519e-01 -6.74464583e-01 2.77165949e-01 5.14123857e-01 -3.97188753e-01 -7.72202253e-01 -7.55016983e-01 -1.22442581e-01 3.91690731e-01 -4.79880832e-02 7.17055857e-01 2.30730519e-01 2.22809821e-01 6.49697065e-01 -2.35669851e-01 -1.50830716e-01 -3.08124870e-01 2.64964968e-01 5.51701725e-01 -2.54835904e-01 -8.65317464e-01 6.90671206e-01 2.62800485e-01 -1.34991005e-01 -9.91780937e-01 -8.92927825e-01 -2.18633443e-01 -3.44279885e-01 1.68978021e-01 6.90254867e-01 -6.31518185e-01 -7.34931231e-01 -3.12690973e-01 -1.08160233e+00 -3.73753347e-02 -3.41718763e-01 4.66867894e-01 -4.39136624e-01 3.82111698e-01 -7.81249881e-01 -8.69211495e-01 2.46881098e-01 -8.73722076e-01 3.94927055e-01 -1.60360932e-01 -1.03330278e+00 -1.44811213e+00 -2.61792928e-01 2.55911291e-01 4.18080777e-01 -8.05295482e-02 1.50116158e+00 -1.08449197e+00 -1.41333416e-01 -1.57973871e-01 -1.61159337e-01 7.13815615e-02 4.59814548e-01 -2.96335462e-02 -8.49568725e-01 -9.63905975e-02 1.23441488e-01 -2.07439840e-01 5.72307348e-01 -1.54455915e-01 5.45146286e-01 -6.04566872e-01 -2.69098282e-01 -7.94385597e-02 1.22768593e+00 8.51235464e-02 5.74834585e-01 5.86245060e-01 1.63181633e-01 1.31216109e+00 6.17503881e-01 1.48376644e-01 2.53167331e-01 7.83232629e-01 3.12596858e-01 1.55383110e-01 1.85470431e-04 -4.12529975e-01 3.70824754e-01 1.00862467e+00 -1.49356127e-01 -2.95204550e-01 -8.67943168e-01 9.20101702e-01 -1.47796607e+00 -1.13906229e+00 -6.25948012e-01 1.95814371e+00 6.95389867e-01 6.16484582e-01 3.07918966e-01 3.84381294e-01 4.62062120e-01 3.79591793e-01 1.79497823e-01 -7.82568276e-01 -3.90327483e-01 4.54380095e-01 3.00302237e-01 6.24924898e-01 -4.27528232e-01 8.91460478e-01 6.65319109e+00 6.70860767e-01 -6.90662146e-01 1.01265326e-01 3.78680617e-01 1.42836466e-01 -7.88307011e-01 4.64344084e-01 -3.11367899e-01 3.87259364e-01 1.03730345e+00 -2.24284217e-01 3.91369432e-01 1.75298855e-01 1.06785744e-01 -2.51113087e-01 -1.08635771e+00 2.31573164e-01 -2.62845993e-01 -9.47438717e-01 1.33786082e-01 7.68600106e-02 2.11099654e-01 2.05317931e-03 -2.77929842e-01 -6.21394441e-02 4.23850924e-01 -8.72263968e-01 7.10981250e-01 -6.26826137e-02 6.13524467e-02 -7.00227857e-01 1.04158843e+00 2.25293577e-01 -8.52640450e-01 -9.17537212e-02 -6.62699461e-01 -5.35606027e-01 1.64368063e-01 7.13564277e-01 -6.75659478e-01 6.44872427e-01 6.25016212e-01 6.65713429e-01 -6.05731010e-01 6.83725834e-01 -5.89810431e-01 7.96763539e-01 -2.88984537e-01 -5.22940338e-01 2.18577728e-01 -2.95855761e-01 6.90235972e-01 1.18589377e+00 1.94086313e-01 -1.28065363e-01 -4.52270150e-01 7.79839516e-01 3.33289415e-01 2.94099182e-01 -7.45956540e-01 -4.84793395e-01 6.83378160e-01 1.14693844e+00 -1.25430453e+00 -2.79831178e-02 -7.22865701e-01 7.31460512e-01 4.37700182e-01 -3.98593023e-02 -5.87529719e-01 -1.85981616e-01 7.66119599e-01 4.45002496e-01 1.19811274e-01 6.43806979e-02 -1.92809463e-01 -9.65600312e-01 5.66476509e-02 -8.05880368e-01 5.24411082e-01 -6.07039630e-01 -1.46940696e+00 6.18655801e-01 -1.60567218e-03 -8.11001182e-01 -9.83198509e-02 -6.85061455e-01 -3.64971101e-01 6.66957438e-01 -1.39118934e+00 -5.34117162e-01 5.53291738e-01 1.75042331e-01 1.22033373e-01 5.96005730e-02 1.02301180e+00 1.87030420e-01 5.88908419e-02 2.71107286e-01 -3.97489607e-01 -1.68766119e-02 5.11151493e-01 -1.07628965e+00 2.94355690e-01 5.57944119e-01 5.70427239e-01 1.05386651e+00 1.06227255e+00 -5.45475185e-01 -8.09656620e-01 -4.77274776e-01 1.70442402e+00 -7.00258732e-01 1.07641232e+00 -2.54042238e-01 -8.35417330e-01 5.70699573e-01 4.96546239e-01 -5.72266877e-01 5.06425023e-01 5.74097693e-01 -3.87034416e-01 2.11761624e-01 -1.03834403e+00 6.79074764e-01 1.57982624e+00 -5.30274212e-01 -1.04668415e+00 4.51125741e-01 5.55690885e-01 1.10905915e-01 -1.02172482e+00 1.84571415e-01 2.59647429e-01 -1.26789093e+00 9.29469824e-01 -7.27791846e-01 5.60955882e-01 -3.09104677e-02 -6.84627965e-02 -1.42493606e+00 -5.23324072e-01 -2.10422590e-01 5.45556605e-01 1.52962339e+00 8.95281494e-01 -9.33758557e-01 4.23094928e-01 2.19112098e-01 1.67257890e-01 -6.22875750e-01 -1.08630681e+00 -8.43297541e-01 2.38650560e-01 -5.38551569e-01 6.73490584e-01 1.31739485e+00 5.37956953e-01 5.74119031e-01 -5.60274981e-02 -1.78231418e-01 1.58451870e-01 -1.02308482e-01 -3.92310657e-02 -1.31177688e+00 -2.67117620e-01 -6.60937905e-01 -5.19826591e-01 -4.03814763e-01 2.79390395e-01 -1.20919371e+00 -4.38188374e-01 -1.56679010e+00 4.60134484e-02 -5.37387550e-01 -2.19541684e-01 3.50668669e-01 -1.25166818e-01 1.19974576e-01 1.17643207e-01 2.60005463e-02 -1.57845542e-01 3.08372080e-01 9.71990645e-01 -1.19049242e-02 6.26883134e-02 -4.69053946e-02 -9.91520166e-01 8.02857995e-01 8.03314328e-01 -7.44061828e-01 -5.08615494e-01 -4.13210362e-01 8.87208819e-01 -1.87625065e-01 -1.59061339e-04 -4.36117738e-01 -1.36367694e-01 -1.40615672e-01 5.05655408e-02 2.33852286e-02 2.06093714e-01 -1.07534361e+00 1.36152714e-01 7.51077592e-01 -6.34760976e-01 3.25569451e-01 1.06680937e-01 1.50993928e-01 -2.11191922e-01 -5.75808108e-01 2.75141567e-01 -2.95164526e-01 -2.57538825e-01 -3.40550661e-01 -7.06673801e-01 3.22518021e-01 5.71337402e-01 -2.02662498e-01 -1.80839017e-01 -5.30071855e-01 -4.86352831e-01 -1.23560481e-01 5.35804987e-01 7.74105668e-01 3.06063414e-01 -1.05768812e+00 -6.81820989e-01 -1.14949666e-01 2.34685957e-01 -7.35805690e-01 -4.77007866e-01 7.62245715e-01 -3.38311076e-01 4.99600619e-01 1.11001797e-01 -8.23299438e-02 -1.20589578e+00 6.85680807e-01 2.32330814e-01 -2.02508390e-01 -6.03159010e-01 7.11603761e-01 -2.12501630e-01 -4.83378738e-01 5.35139330e-02 -3.30795407e-01 -3.58057439e-01 5.36399424e-01 2.94398844e-01 2.36660197e-01 1.60101458e-01 -6.28612995e-01 -7.12881744e-01 4.27679151e-01 1.14385374e-01 -1.09113818e-02 1.69197011e+00 -4.26521935e-02 -5.75887442e-01 7.09689736e-01 8.21893096e-01 4.91027087e-01 -3.30239117e-01 -4.64303978e-02 7.40833580e-01 -3.98722827e-01 -5.60025692e-01 -6.54817998e-01 -8.57207000e-01 7.45082617e-01 1.14163823e-01 8.20231795e-01 5.75748444e-01 3.69201928e-01 2.83976883e-01 2.06018060e-01 4.56567973e-01 -9.60564375e-01 -2.68299133e-01 6.31006360e-01 7.50721455e-01 -6.97547495e-01 1.59964323e-01 -6.96987867e-01 -4.25877333e-01 1.33331549e+00 1.40100345e-01 9.21573639e-02 4.18694884e-01 3.12436402e-01 -5.98359713e-03 -4.90861654e-01 -9.63135779e-01 -5.09116709e-01 -1.03105880e-01 7.40275502e-01 1.28500271e+00 1.52656920e-02 -1.43716431e+00 1.95147842e-01 -7.35174656e-01 -4.02211845e-01 4.30311739e-01 5.73016822e-01 -1.57728657e-01 -1.60160470e+00 -1.48029432e-01 5.84070444e-01 -6.56319499e-01 -3.52623552e-01 -6.87762558e-01 8.73659372e-01 -7.05163693e-04 1.26115799e+00 1.11164719e-01 -2.27510363e-01 4.66570169e-01 2.47191638e-01 7.34438062e-01 -5.33302724e-01 -9.41145182e-01 -4.32318002e-01 9.48865533e-01 -3.98710638e-01 -5.62650204e-01 -7.19516575e-01 -9.74341631e-01 -5.02708018e-01 -5.48572004e-01 5.36684692e-01 2.20659658e-01 9.81614828e-01 -5.10226525e-02 4.94552314e-01 8.77415463e-02 -1.65767133e-01 -4.73533005e-01 -1.01527596e+00 -3.96896303e-01 6.98269188e-01 -1.92378774e-01 -5.77802896e-01 -5.24395525e-01 -4.31266934e-01]

[10.274543762207031, 9.050971031188965]

564d7d27-8c3e-43a4-8967-a7b5280de954

hltsuda-at-semeval-2019-task-1-ucca-graph-1

https://aclanthology.org/S19-2002

https://aclanthology.org/S19-2002.pdf

HLT@SUDA at SemEval-2019 Task 1: UCCA Graph Parsing as Constituent Tree Parsing

This paper describes a simple UCCA semantic graph parsing approach. The key idea is to convert a UCCA semantic graph into a constituent tree, in which extra labels are deliberately designed to mark remote edges and discontinuous nodes for future recovery. In this way, we can make use of existing syntactic parsing techniques. Based on the data statistics, we recover discontinuous nodes directly according to the output labels of the constituent parser and use a biaffine classification model to recover the more complex remote edges. The classification model and the constituent parser are simultaneously trained under the multi-task learning framework. We use the multilingual BERT as extra features in the open tracks. Our system ranks the first place in the six English/German closed/open tracks among seven participating systems. For the seventh cross-lingual track, where there is little training data for French, we propose a language embedding approach to utilize English and German training data, and our result ranks the second place.

['Wei Jiang', 'Zhenghua Li', 'Min Zhang', 'Yu Zhang']

2019-06-01

semeval-2019-6

['ucca-parsing']

['natural-language-processing']

[-7.09655136e-02 6.06640577e-01 -4.56022322e-01 -4.18398112e-01 -1.20067620e+00 -8.41700554e-01 4.99731958e-01 2.02905223e-01 -5.38647652e-01 4.63597417e-01 2.94753551e-01 -4.93465871e-01 1.44522220e-01 -8.97919595e-01 -7.49107897e-01 -5.50414741e-01 1.10742196e-01 5.81562221e-01 2.79619455e-01 -2.19849944e-01 -2.07636371e-01 7.82695860e-02 -9.97521698e-01 3.58663350e-01 7.69985557e-01 9.39810693e-01 2.14254975e-01 7.44435906e-01 -5.56027055e-01 6.78285778e-01 -3.51163745e-01 -7.40685284e-01 2.59145081e-01 -3.84548485e-01 -1.00167418e+00 -1.67759836e-01 4.60553318e-01 8.00081119e-02 -2.80454934e-01 1.15415096e+00 1.60340980e-01 -1.68552576e-03 3.59620184e-01 -9.68313992e-01 -7.19887316e-01 1.00888491e+00 -4.75018889e-01 -1.19972430e-01 2.96029449e-01 -5.78781605e-01 1.59819758e+00 -8.92059565e-01 1.01038480e+00 1.38364720e+00 6.88649774e-01 6.22144878e-01 -1.03964472e+00 -4.59073693e-01 5.79090714e-01 7.87952021e-02 -1.07259321e+00 -3.28465879e-01 9.05793369e-01 -3.20327431e-01 8.84234428e-01 -6.27975389e-02 2.51293629e-01 1.19801319e+00 1.43911526e-01 9.50531781e-01 1.08007777e+00 -6.27629995e-01 1.19706981e-01 2.04717383e-01 6.50110066e-01 1.24283433e+00 1.09797576e-02 5.68699017e-02 -1.37050435e-01 -7.80915841e-02 3.10943693e-01 -1.38227642e-01 -9.74520072e-02 -4.70351458e-01 -9.95584428e-01 1.02855861e+00 6.66413963e-01 3.70761573e-01 -3.63672227e-02 2.14616861e-02 5.78118503e-01 3.33734363e-01 6.23038411e-01 8.89305398e-02 -7.82291353e-01 1.77376956e-01 -7.87960529e-01 -4.70360853e-02 9.33225453e-01 1.08472836e+00 9.94446099e-01 -1.98386446e-01 1.63244516e-01 9.62761462e-01 3.79952788e-01 1.52165294e-01 4.51488107e-01 -8.36892784e-01 7.23378599e-01 6.02256417e-01 -4.71529543e-01 -4.86327291e-01 -4.35543686e-01 -3.01027596e-01 -5.49418092e-01 2.79186908e-02 5.26781976e-01 -2.09831342e-01 -1.05379784e+00 1.73464346e+00 2.28708163e-01 8.48036930e-02 3.54426593e-01 5.99471867e-01 6.88003004e-01 5.47056019e-01 3.04117262e-01 5.07875979e-02 1.61076152e+00 -1.42684793e+00 -7.12613106e-01 -3.67608666e-01 8.78132582e-01 -6.01011813e-01 1.02419376e+00 1.14166297e-01 -6.88521802e-01 -5.77664614e-01 -1.00931454e+00 -4.89555120e-01 -7.62775540e-01 2.86749989e-01 7.33495235e-01 3.08299124e-01 -1.18382037e+00 5.66324174e-01 -8.97016585e-01 -2.61062264e-01 1.50696203e-01 1.03777543e-01 -7.81619787e-01 -3.68167371e-01 -1.24723923e+00 8.22074831e-01 5.95115781e-01 1.76035389e-01 -4.71884042e-01 -3.96300107e-01 -1.35798919e+00 1.28628425e-02 3.82065058e-01 -4.03512806e-01 1.04219043e+00 -9.59963500e-01 -1.31234741e+00 1.07311594e+00 -1.25350982e-01 -3.07308018e-01 4.84662205e-01 -6.72728419e-02 -3.71751547e-01 9.88935381e-02 4.59848136e-01 8.22522998e-01 6.43721938e-01 -1.25739777e+00 -8.81174386e-01 -4.29236621e-01 2.31734395e-01 2.10303683e-02 -5.17217442e-02 -1.15535386e-01 -8.47627759e-01 -7.59846449e-01 3.49673539e-01 -1.04384005e+00 -2.34008491e-01 -3.28419983e-01 -4.45354521e-01 -5.75660348e-01 7.38378048e-01 -1.05214643e+00 1.08189642e+00 -2.26676440e+00 5.35171449e-01 1.72788128e-01 2.35535547e-01 -1.76701427e-01 -3.18439305e-01 1.72610000e-01 -1.35993943e-01 3.32447082e-01 -3.47446889e-01 -7.68641293e-01 1.45038264e-02 4.71414894e-01 -1.87951371e-01 2.07496524e-01 3.82074803e-01 8.53898525e-01 -1.02100396e+00 -5.04520297e-01 -1.04102390e-02 2.55694538e-01 -6.33282542e-01 1.08546115e-01 -1.97656676e-01 1.45458236e-01 -6.35809004e-01 7.43899524e-01 6.18048728e-01 1.16286054e-02 4.50307071e-01 -2.23396182e-01 -9.43880528e-02 4.74095732e-01 -9.56343114e-01 1.96270454e+00 -6.97871983e-01 3.21974903e-01 4.04098749e-01 -1.00805843e+00 9.83236074e-01 1.94480017e-01 2.69741029e-01 -6.46047890e-01 -1.00990258e-01 3.98765206e-01 -2.21965745e-01 -4.59273010e-02 4.99725908e-01 -2.25759335e-02 -6.20290816e-01 8.93079340e-02 5.05002856e-01 1.51805282e-01 1.65050849e-01 3.77393275e-01 1.19579268e+00 5.39445698e-01 1.32203549e-01 -3.55616778e-01 5.32608449e-01 1.26022950e-01 7.57021308e-01 5.00052750e-01 -3.97283107e-01 4.21570957e-01 8.27658594e-01 -4.88346040e-01 -8.74155879e-01 -1.16680634e+00 -5.42776249e-02 1.44794929e+00 -1.75608486e-01 -7.11852074e-01 -7.74707079e-01 -1.35683167e+00 -2.72278264e-02 5.23423135e-01 -6.24111474e-01 -5.32543212e-02 -8.11606348e-01 -5.32225251e-01 4.80157346e-01 6.23652279e-01 3.06367666e-01 -1.11028409e+00 2.12245598e-01 2.61312634e-01 -2.46062309e-01 -1.38144124e+00 -3.96372437e-01 7.49863565e-01 -7.69778609e-01 -1.16966820e+00 -3.29628110e-01 -1.42821503e+00 5.30395210e-01 -1.39885455e-01 1.07431412e+00 -8.64154622e-02 1.16670020e-01 1.44012377e-01 -5.69793880e-01 4.96979766e-02 -6.12472057e-01 4.55608338e-01 -2.28512153e-01 -7.80635625e-02 3.53073090e-01 -2.11238742e-01 -3.07050403e-02 -1.32497385e-01 -5.48186958e-01 -1.38932928e-01 5.26805162e-01 1.08041227e+00 9.48804438e-01 -2.61580050e-01 4.50933814e-01 -1.48841465e+00 3.52455497e-01 -4.78440374e-01 -6.02575064e-01 3.66821885e-01 -4.99344349e-01 4.63823259e-01 8.68376732e-01 -1.48751020e-01 -9.86960053e-01 4.57150757e-01 -4.23368722e-01 -4.07862186e-01 -1.57078698e-01 5.15213192e-01 -5.03691018e-01 2.41910204e-01 2.70991206e-01 -2.04761147e-01 -4.16870832e-01 -8.99857998e-01 7.65321434e-01 6.28889978e-01 6.95742488e-01 -6.72597647e-01 6.06325567e-01 1.61160409e-01 -3.57900620e-01 -6.10585928e-01 -1.12183797e+00 -4.92901266e-01 -1.02679968e+00 2.67658532e-01 1.29733014e+00 -1.10532737e+00 -1.31520018e-01 2.32359588e-01 -1.23242950e+00 -3.71482641e-01 -2.66898096e-01 4.00239319e-01 -3.18118125e-01 3.72779399e-01 -1.04585302e+00 -3.14614087e-01 -2.33957067e-01 -1.22491705e+00 1.28536880e+00 -4.12971824e-02 1.32465124e-01 -1.34135807e+00 8.51610899e-02 3.15957576e-01 -7.95870088e-03 2.75537875e-02 1.19288266e+00 -9.00501132e-01 -4.53867793e-01 -2.41887033e-01 -2.28889689e-01 4.30819035e-01 6.85592964e-02 -8.99424106e-02 -1.09993064e+00 -3.48884940e-01 -1.69610232e-01 -2.25397959e-01 1.27461898e+00 2.07912385e-01 9.23755765e-01 -4.64989850e-03 -4.22679573e-01 8.21093261e-01 1.33029866e+00 -1.10751450e-01 2.90182620e-01 3.01965952e-01 1.11011600e+00 7.00126290e-01 4.29613233e-01 -3.96586686e-01 8.30478072e-01 4.44546282e-01 2.89418727e-01 -2.24537551e-02 -3.46236140e-01 -6.35482490e-01 7.14483619e-01 1.44749010e+00 3.04582179e-01 -1.72243103e-01 -8.28417301e-01 5.80640972e-01 -1.67593181e+00 -5.20838499e-01 -1.61777541e-01 1.83521867e+00 4.79891479e-01 3.64903808e-01 -2.37842903e-01 -1.78649276e-01 7.66844094e-01 3.77279311e-01 -1.52927339e-01 -6.64631248e-01 -1.33829832e-01 5.73464096e-01 7.91300535e-01 8.49468946e-01 -1.51661289e+00 1.52965963e+00 6.12779284e+00 6.59367442e-01 -9.10344124e-01 4.14975196e-01 4.51283157e-01 5.36649108e-01 -3.12041759e-01 4.72920269e-01 -1.03252590e+00 2.78024137e-01 1.13550961e+00 2.58323163e-01 2.89058059e-01 9.22492623e-01 -3.48516971e-01 1.32686794e-01 -1.11362469e+00 6.06889188e-01 -3.13836557e-04 -1.16746628e+00 -1.10210747e-01 -1.96671393e-02 3.67261320e-01 3.22235018e-01 -3.38637114e-01 8.05072784e-01 6.60397828e-01 -6.66187942e-01 7.35148311e-01 1.56927615e-01 7.18074918e-01 -5.61130226e-01 7.58175492e-01 2.82044977e-01 -1.51694620e+00 1.37302736e-02 -4.09834385e-01 2.87343323e-01 2.12510571e-01 3.94657552e-01 -4.87761587e-01 8.91837299e-01 6.44004047e-01 8.59935045e-01 -6.65624917e-01 4.57577854e-01 -7.66834497e-01 6.26554430e-01 -1.01365492e-01 3.11279684e-01 3.94006342e-01 -5.86387575e-01 4.56495166e-01 1.44191754e+00 1.42945677e-01 -3.17292958e-01 6.25815094e-01 4.74397421e-01 -5.32160640e-01 2.99356222e-01 -6.57516181e-01 1.14857018e-01 2.62065709e-01 1.45494568e+00 -7.47989357e-01 -3.52210730e-01 -7.87584722e-01 1.34665334e+00 8.96414638e-01 2.11309031e-01 -5.59509337e-01 -3.27047616e-01 5.19877076e-01 -4.02996510e-01 3.51633191e-01 -3.11114699e-01 -3.31361257e-02 -1.46377993e+00 1.71838254e-01 -5.25777996e-01 8.38370919e-01 -4.16731596e-01 -1.33889449e+00 8.52209449e-01 -1.92503050e-01 -8.05430770e-01 -3.40307415e-01 -9.69635487e-01 -5.32804847e-01 7.97047257e-01 -1.69432378e+00 -1.49343729e+00 1.04211524e-01 4.74804789e-01 6.67970240e-01 -2.59386808e-01 1.17752016e+00 5.02970457e-01 -7.75318742e-01 6.50945842e-01 3.73897292e-02 6.34969592e-01 7.56940186e-01 -1.63379550e+00 6.78568900e-01 1.00992441e+00 5.57604909e-01 2.47129634e-01 1.68279141e-01 -6.66830599e-01 -1.27653658e+00 -1.20603132e+00 1.33846343e+00 -4.28792119e-01 1.11945343e+00 -8.16667020e-01 -1.02268183e+00 1.21887708e+00 8.05398673e-02 4.35114831e-01 3.90402824e-01 4.71702009e-01 -6.60166621e-01 4.06429321e-02 -8.12030554e-01 2.80659914e-01 9.81398046e-01 -9.40283179e-01 -7.65227616e-01 2.34635621e-01 1.05656028e+00 -3.98720294e-01 -8.63882840e-01 1.55495986e-01 1.34991631e-01 -2.78620005e-01 7.05232441e-01 -7.20904350e-01 2.85627067e-01 -4.15025800e-02 -3.64369810e-01 -1.59160542e+00 -3.76850069e-01 -2.57477403e-01 -7.21984869e-03 1.41934872e+00 8.58192325e-01 -6.52012289e-01 7.84379244e-01 9.05038118e-02 -5.19842446e-01 -3.09344500e-01 -1.01080739e+00 -6.26028597e-01 4.17940080e-01 -3.49116892e-01 1.26863763e-01 1.10222137e+00 -1.14465803e-01 6.92574561e-01 -1.53334990e-01 1.12945020e-01 5.88385165e-01 2.23095536e-01 2.96380758e-01 -1.40951133e+00 -3.42648804e-01 -1.93853036e-01 -2.78845429e-01 -1.03695512e+00 8.37530792e-01 -1.59803379e+00 1.16528094e-01 -1.46421552e+00 -3.67482007e-02 -7.11369991e-01 -5.68010449e-01 6.82331562e-01 -2.31275752e-01 1.01713873e-02 3.21076363e-01 9.02635679e-02 -4.90273088e-01 2.91615069e-01 8.82806182e-01 -1.12375766e-01 -2.07480699e-01 -1.63726091e-01 -8.46311092e-01 7.74241209e-01 6.47761703e-01 -6.60199463e-01 -3.10152490e-02 -7.07505465e-01 -1.31398970e-02 5.55862002e-02 1.51983336e-01 -5.78892291e-01 1.92089781e-01 3.29718918e-01 1.27202302e-01 -2.92208880e-01 1.05084851e-01 -8.06859016e-01 -2.62964070e-01 1.54020816e-01 -2.07200736e-01 2.75864959e-01 2.57104263e-02 5.59199274e-01 -4.90032345e-01 -2.07531378e-01 7.16313362e-01 -1.01123430e-01 -1.01991761e+00 3.22879016e-01 2.32265536e-02 1.67516753e-01 8.71666253e-01 2.73899198e-01 -2.81518430e-01 8.82235616e-02 -1.28789985e+00 4.35901552e-01 2.44966999e-01 7.90298700e-01 3.25524062e-01 -1.37448585e+00 -6.69681191e-01 3.78371477e-01 2.43123025e-01 6.47667190e-03 -1.59373078e-02 5.03376126e-01 -4.56949174e-01 1.81733951e-01 -2.63421219e-02 -4.52281654e-01 -9.82579231e-01 6.00373507e-01 2.95848817e-01 -5.96260965e-01 -8.54989409e-01 7.04981327e-01 1.52336076e-01 -9.62749839e-01 7.14607015e-02 -3.60186100e-01 -2.27635324e-01 3.61197293e-01 1.05489701e-01 -1.00495815e-02 2.77940482e-01 -7.98488855e-01 -3.70328575e-01 5.59616148e-01 -2.17241257e-01 -6.29841983e-02 1.41991818e+00 -3.68055739e-02 -1.59496263e-01 3.91278237e-01 1.64852762e+00 4.03187335e-01 -1.13032174e+00 -2.87991673e-01 5.71563780e-01 1.31805167e-01 2.20540985e-01 -5.85206687e-01 -1.23045921e+00 9.62182820e-01 3.63246977e-01 3.09360288e-02 8.24319780e-01 4.74926502e-01 7.43449986e-01 1.21995635e-01 3.53683323e-01 -1.23547363e+00 -4.23984826e-01 8.21290493e-01 4.65655416e-01 -1.25420105e+00 -5.11790395e-01 -7.77366936e-01 -6.20677769e-01 1.25651062e+00 4.76106882e-01 -3.00376683e-01 7.35440135e-01 3.71414810e-01 1.78629577e-01 -1.81261197e-01 -5.78994811e-01 -3.58872503e-01 1.33080021e-01 5.02528906e-01 4.88413185e-01 4.70074266e-01 -2.50811040e-01 9.85367358e-01 -3.13369930e-01 -5.80035150e-01 3.23873341e-01 6.46518886e-01 -3.43197852e-01 -1.63877177e+00 1.17506713e-01 1.99233085e-01 -5.85443974e-01 -2.54967749e-01 -4.28129017e-01 1.07821107e+00 9.58442464e-02 7.02878118e-01 2.27417693e-01 -3.36285055e-01 6.14306450e-01 6.31551385e-01 1.71204805e-01 -6.77768588e-01 -6.01911545e-01 1.23589449e-01 3.15752715e-01 -6.97285056e-01 -1.41720176e-01 -6.56900883e-01 -1.51993024e+00 -9.04241763e-03 -1.80351302e-01 2.52164006e-01 7.26592243e-01 7.98275292e-01 2.59911895e-01 6.66867971e-01 5.38533151e-01 -5.10884941e-01 -5.07687211e-01 -7.75194287e-01 -4.79757905e-01 5.10549128e-01 2.99781084e-01 -4.81633723e-01 -4.05280590e-01 -5.99197410e-02]

[10.52525806427002, 9.657309532165527]

1bba08ac-1839-4039-8cc2-62f79ab3ea94

multi-agent-path-finding-with-continuous-time

1903.09820

http://arxiv.org/abs/1903.09820v1

http://arxiv.org/pdf/1903.09820v1.pdf

Multi-agent Path Finding with Continuous Time Viewed Through Satisfiability Modulo Theories (SMT)

This paper addresses a variant of multi-agent path finding (MAPF) in continuous space and time. We present a new solving approach based on satisfiability modulo theories (SMT) to obtain makespan optimal solutions. The standard MAPF is a task of navigating agents in an undirected graph from given starting vertices to given goal vertices so that agents do not collide with each other in vertices of the graph. In the continuous version (MAPF$^\mathcal{R}$) agents move in an $n$-dimensional Euclidean space along straight lines that interconnect predefined positions. For simplicity, we work with circular omni-directional agents having constant velocities in the 2D plane. As agents can have different sizes and move smoothly along lines, a non-colliding movement along certain lines with small agents can result in a collision if the same movement is performed with larger agents. Our SMT-based approach for MAPF$^\mathcal{R}$ called SMT-CBS$^\mathcal{R}$ reformulates the Conflict-based Search (CBS) algorithm in terms of SMT concepts. We suggest lazy generation of decision variables and constraints. Each time a new conflict is discovered, the underlying encoding is extended with new variables and constraints to eliminate the conflict. We compared SMT-CBS$^\mathcal{R}$ and adaptations of CBS for the continuous variant of MAPF experimentally.

['Pavel Surynek']

2019-03-23

['multi-agent-path-finding']

['playing-games']

[ 3.18158984e-01 2.86677659e-01 2.55143363e-02 6.92655817e-02 1.13926586e-02 -8.09475660e-01 3.19557488e-01 5.81806064e-01 -6.79467618e-01 1.20900822e+00 -6.58328116e-01 -6.43942118e-01 -1.10665023e+00 -1.35089946e+00 -5.57413518e-01 -5.83438933e-01 -8.46127033e-01 1.15106726e+00 5.23392260e-01 -6.29750729e-01 3.85821730e-01 6.06179416e-01 -1.16125262e+00 -2.06455901e-01 6.88981950e-01 5.01135886e-01 3.73437941e-01 6.82914257e-01 -2.93004066e-01 2.46195361e-01 -7.18188226e-01 -2.05693170e-01 6.65256202e-01 -4.02823687e-01 -1.09353149e+00 4.59625386e-03 -2.96020150e-01 1.47459686e-01 8.52339491e-02 1.06523502e+00 -7.97322541e-02 3.22666109e-01 5.23619115e-01 -2.08430314e+00 6.22463040e-02 5.89265168e-01 -7.27202475e-01 -4.08472568e-02 6.39311194e-01 9.01905000e-02 7.77022719e-01 4.92209122e-02 9.51404035e-01 1.03267241e+00 3.23898852e-01 4.24271196e-01 -1.04666054e+00 -4.12079692e-01 5.95192194e-01 3.07686329e-01 -1.46783924e+00 2.00475827e-01 3.27713072e-01 -1.17651984e-01 1.57580256e+00 7.58567810e-01 7.86742151e-01 1.66741654e-01 4.85325724e-01 4.67737652e-02 8.34324539e-01 -6.03662789e-01 6.28665626e-01 -2.10143581e-01 -1.70265540e-01 6.02434874e-01 7.06655085e-01 3.65456492e-01 -1.05407611e-01 -1.28904611e-01 7.55963445e-01 -3.69230151e-01 2.73777470e-02 -5.21418095e-01 -1.28937793e+00 1.01767504e+00 2.94503689e-01 3.69493306e-01 -4.10630673e-01 3.51355851e-01 1.52243331e-01 5.29724419e-01 -4.19166476e-01 8.73368919e-01 -3.72661173e-01 1.38829900e-02 -4.80392605e-01 6.85130119e-01 8.39512825e-01 1.28001559e+00 7.80778825e-01 -1.46669880e-01 4.99587804e-01 1.51454151e-01 1.65197682e-02 5.83159268e-01 -3.72293830e-01 -1.24502611e+00 6.20161772e-01 6.72212720e-01 4.81234819e-01 -1.23656154e+00 -9.24113274e-01 -2.24529177e-01 -5.29937565e-01 9.01160121e-01 4.48563606e-01 -2.27848560e-01 -4.46601629e-01 1.74307370e+00 4.42683190e-01 -3.87361161e-02 2.22271085e-01 8.11089635e-01 1.11508392e-01 8.12713087e-01 -3.21564168e-01 -8.28590393e-01 9.54513073e-01 -1.10154998e+00 -3.05059552e-01 -2.36958534e-01 7.08042681e-01 -3.83745313e-01 1.97403133e-01 5.77323735e-01 -1.42432225e+00 1.35534912e-01 -1.12396359e+00 7.43316472e-01 -4.49532777e-01 -9.12652850e-01 5.48369408e-01 4.87520188e-01 -1.18683398e+00 3.79912972e-01 -6.35199666e-01 -2.04261109e-01 -1.21908911e-01 7.76924968e-01 -4.02292818e-01 -2.53093958e-01 -1.02477980e+00 8.68001997e-01 4.77869958e-01 -5.92896715e-02 -7.46966481e-01 -1.93188211e-03 -8.80970299e-01 -1.69375136e-01 1.25666928e+00 -6.43097639e-01 1.02096546e+00 -5.96888840e-01 -1.18296909e+00 4.34455246e-01 -1.97120517e-01 -4.47943926e-01 5.29365838e-01 6.43782318e-01 -4.17151481e-01 -3.13662109e-03 4.35923785e-01 4.25445557e-01 2.83397049e-01 -1.39304507e+00 -1.00818062e+00 -3.34274650e-01 6.79533660e-01 2.24213734e-01 5.01850009e-01 -1.69536024e-02 -2.79705256e-01 -2.07589660e-02 2.97795981e-01 -1.17966890e+00 -7.64967680e-01 -4.64255959e-01 -4.69883174e-01 -1.93651244e-01 2.86338627e-01 2.13209465e-01 1.29240263e+00 -1.52976811e+00 5.95858514e-01 9.05656338e-01 2.82805830e-01 -5.67758530e-02 -4.39311057e-01 9.71593440e-01 1.86780751e-01 2.86378086e-01 -1.49657384e-01 1.83399037e-01 1.31952301e-01 5.94303250e-01 3.20823491e-01 3.77699703e-01 -2.84801573e-01 4.61167902e-01 -1.14583004e+00 -3.20018619e-01 9.79789123e-02 -3.09346110e-01 -7.17797518e-01 -3.95424485e-01 -5.84754586e-01 6.84288293e-02 -6.03868663e-01 3.15483660e-01 8.67561102e-01 6.62896484e-02 4.79219615e-01 5.71517408e-01 -8.76316547e-01 1.96354575e-02 -1.79297996e+00 1.36528790e+00 -9.00728703e-02 1.42028362e-01 4.44574505e-01 -1.10665858e+00 8.32575440e-01 -5.04588820e-02 6.34048641e-01 -9.75664079e-01 2.33116299e-01 3.13060075e-01 2.13306159e-01 -2.57151783e-01 4.61511612e-01 -1.08100614e-03 -3.85821760e-01 6.20907843e-01 -7.63291001e-01 -4.22894955e-01 8.96450222e-01 2.06827745e-01 1.47811162e+00 -2.88206667e-01 2.01116532e-01 -3.06875706e-01 7.79890120e-01 6.68253124e-01 8.00104856e-01 1.03427804e+00 -1.42565861e-01 -2.68238544e-01 5.98316908e-01 -6.72286391e-01 -8.11593771e-01 -1.05433822e+00 3.51695031e-01 6.95715547e-01 8.80341947e-01 -6.69241011e-01 -5.06640434e-01 -4.17559296e-01 -9.47306752e-02 8.44465256e-01 -5.07217228e-01 2.04616487e-01 -8.52390945e-01 -2.98481286e-01 7.41186813e-02 -3.99056859e-02 4.99584198e-01 -1.09513962e+00 -1.36490834e+00 7.14858413e-01 -4.90621589e-02 -9.25268173e-01 -2.02278972e-01 3.64957273e-01 -2.98633873e-01 -1.38700783e+00 -6.79435506e-02 -8.28852177e-01 9.92951810e-01 5.59058189e-01 6.02252662e-01 3.45053047e-01 -2.83837169e-01 2.00444236e-01 -5.38542747e-01 -4.53521103e-01 -2.08382294e-01 -1.33849859e-01 6.24667816e-02 -6.64423227e-01 1.19763777e-01 -3.13932151e-01 -2.72639453e-01 6.46165013e-01 -6.30798042e-01 2.12487802e-01 4.86933142e-02 4.14935172e-01 7.66835213e-01 1.02367747e+00 3.41482580e-01 -3.54932249e-01 7.01292276e-01 -3.51648897e-01 -1.06459379e+00 2.44647577e-01 -4.50802475e-01 7.21024908e-03 7.51258850e-01 -1.56322673e-01 -3.47272724e-01 -2.60072440e-01 4.81063902e-01 -3.33850947e-03 -7.62275234e-02 7.56231844e-01 1.34347826e-01 -1.19376227e-01 4.10513222e-01 -5.19688763e-02 -7.49776438e-02 1.48771763e-01 6.72566295e-02 -6.11183383e-02 3.72835964e-01 -6.86137259e-01 7.31717944e-01 3.27806920e-01 7.50651658e-01 -2.77473986e-01 3.61934811e-01 6.69364035e-02 -2.51291096e-01 -3.06818515e-01 6.27651453e-01 -2.66823359e-02 -1.23614633e+00 1.60853729e-01 -1.04894662e+00 -6.33194208e-01 -1.77817881e-01 2.49338016e-01 -7.16914713e-01 1.73347399e-01 -7.56489858e-02 -1.17626405e+00 2.12034613e-01 -1.32119012e+00 4.31001663e-01 2.71879077e-01 -3.20391595e-01 -7.99697757e-01 2.09634885e-01 -3.52342948e-02 4.54211086e-01 5.63039184e-01 1.17135406e+00 -5.71334481e-01 -7.78630853e-01 1.24889398e-02 2.00816598e-02 -6.89195991e-01 1.37733286e-02 -7.97609091e-02 4.70672518e-01 -5.53049624e-01 -3.41450602e-01 3.74998361e-01 -8.07799026e-02 4.64747101e-01 3.95807683e-01 -5.48002005e-01 -8.55392277e-01 2.27299482e-02 1.64491677e+00 1.07291424e+00 3.89399052e-01 1.23697102e+00 -1.47434428e-01 8.41180563e-01 8.43848348e-01 5.03912091e-01 5.72463572e-01 8.05816054e-01 9.78746831e-01 2.65133917e-01 4.78038758e-01 4.20489639e-01 4.45240475e-02 -3.09649818e-02 -3.84879887e-01 -6.96382999e-01 -1.03932059e+00 5.80365300e-01 -1.87347269e+00 -1.00937438e+00 -3.45038205e-01 2.24664354e+00 3.28774363e-01 4.11537677e-01 3.18761408e-01 3.39489609e-01 9.56054866e-01 -2.73283064e-01 -4.35447723e-01 -1.06385767e+00 -3.90726216e-02 1.81655779e-01 7.10107565e-01 1.08187747e+00 -4.64061856e-01 7.59280860e-01 4.89731264e+00 2.57060111e-01 -8.48543584e-01 -1.87261686e-01 -7.22031966e-02 -3.11049789e-01 -3.59433949e-01 2.22083941e-01 -4.66841191e-01 2.49300212e-01 6.28634095e-01 -4.27565992e-01 9.41389143e-01 4.01755154e-01 4.02663618e-01 -5.50029039e-01 -9.54811931e-01 6.92251444e-01 -2.05036208e-01 -1.35154378e+00 -3.67247075e-01 3.04251701e-01 7.40903139e-01 -3.73969585e-01 -1.96035132e-01 1.98518597e-02 5.75021923e-01 -1.03683269e+00 8.05444658e-01 -2.24700227e-01 4.36356902e-01 -1.18234301e+00 5.89029253e-01 5.55869997e-01 -1.44610536e+00 -3.35100025e-01 -1.00723365e-02 -4.51831758e-01 7.10696220e-01 9.25588945e-04 -8.06924760e-01 1.07881045e+00 6.74980879e-01 -1.10481054e-01 3.52012396e-01 1.17057753e+00 -1.54988930e-01 -5.33350289e-01 -6.49317324e-01 -2.97599703e-01 6.69014335e-01 -5.96818745e-01 1.07191133e+00 7.04348922e-01 4.08416390e-01 4.24588859e-01 4.32845980e-01 7.43535876e-01 7.25919962e-01 -9.36851576e-02 -4.97279972e-01 2.81723946e-01 7.75374591e-01 6.59781098e-01 -1.20898569e+00 6.64819870e-03 -1.16414383e-01 5.79779148e-01 -4.64797840e-02 3.84349167e-01 -9.28334594e-01 -5.50587833e-01 7.25258410e-01 2.11533934e-01 1.82678103e-01 -6.14156365e-01 -1.72758162e-01 -4.26826328e-01 -4.17976603e-02 -7.70084202e-01 4.67519701e-01 -5.84211767e-01 -6.13141358e-01 9.56314564e-01 4.68210876e-01 -9.38413680e-01 -3.70875955e-01 -3.17395478e-01 -5.93963563e-01 6.59853041e-01 -1.28688622e+00 -5.38546324e-01 -2.15204760e-01 9.11446750e-01 4.35991526e-01 -1.16703995e-01 7.82950878e-01 -2.54011273e-01 -4.65927958e-01 2.06989139e-01 -1.78816691e-01 -4.43805784e-01 -1.49434194e-01 -9.14791584e-01 2.20204681e-01 9.38376546e-01 -3.26178461e-01 5.57588100e-01 9.60967004e-01 -8.76080513e-01 -1.62126660e+00 -7.79935241e-01 8.40920329e-01 2.83243358e-01 4.72450495e-01 8.26770291e-02 -2.29880095e-01 7.71498203e-01 2.41347089e-01 -5.45407891e-01 2.55555242e-01 -3.35114032e-01 1.14769526e-01 7.22069964e-02 -1.43764293e+00 9.44727898e-01 1.21277225e+00 4.15661186e-01 -2.74925977e-01 2.56953120e-01 4.42237109e-01 -3.44056606e-01 -4.94239271e-01 4.64804173e-01 1.30868807e-01 -8.35781991e-01 7.51056075e-01 -5.63225389e-01 -2.14046299e-01 -8.11244488e-01 -1.39806196e-01 -1.42550015e+00 -4.52454329e-01 -9.63631809e-01 6.97622657e-01 3.87757301e-01 6.35348499e-01 -1.06833994e+00 7.33445287e-01 6.29345953e-01 -3.28072757e-01 -6.56587899e-01 -1.50815535e+00 -1.21228182e+00 9.54946280e-02 -4.75670397e-01 8.99408877e-01 8.02275598e-01 7.29874134e-01 -4.66984250e-02 -6.74628392e-02 4.40049529e-01 7.38519430e-01 2.49107108e-01 6.89840257e-01 -1.17554271e+00 -3.47963482e-01 -7.77963221e-01 -2.44521156e-01 -6.53808415e-01 -8.40971917e-02 -7.15465486e-01 -7.75604248e-02 -2.01110840e+00 -2.58552879e-01 -1.06187570e+00 2.73882478e-01 6.88569248e-01 7.72523344e-01 -2.95284986e-01 4.81968135e-01 -1.21088721e-01 -6.86475337e-01 -2.01295376e-01 1.29344261e+00 -1.29445612e-01 -5.09717047e-01 -1.91479810e-02 -2.81166613e-01 4.54071343e-01 8.79822493e-01 -5.63645065e-01 -6.53389990e-01 -4.19841141e-01 8.88910592e-01 8.11627269e-01 -6.27686922e-03 -7.52920330e-01 5.48288107e-01 -1.07867444e+00 -5.45014441e-01 -4.34574425e-01 3.80358279e-01 -1.17938209e+00 9.11828876e-01 1.01261950e+00 -8.99418741e-02 6.53240860e-01 2.97294825e-01 2.31108740e-01 2.17731580e-01 -5.66412985e-01 2.88631946e-01 -2.40606964e-01 -6.46128654e-01 4.46869247e-02 -7.17466712e-01 -3.80524814e-01 1.89290369e+00 -6.62571907e-01 -7.02768445e-01 -4.56887633e-01 -8.65035832e-01 8.08526278e-01 5.84192157e-01 1.55820146e-01 6.51102364e-01 -8.81240368e-01 -5.73374808e-01 9.44412425e-02 -1.17523611e-01 2.99834430e-01 2.06499919e-01 8.84918749e-01 -8.00579727e-01 6.46303713e-01 -4.98479843e-01 -1.19967408e-01 -1.11942422e+00 1.00063038e+00 4.56697077e-01 -3.66804838e-01 -3.65574300e-01 7.91319370e-01 -1.60447940e-01 -1.15991563e-01 4.76959348e-02 -2.50673801e-01 -1.26966819e-01 -3.11874092e-01 5.33660471e-01 6.90978408e-01 -4.91152778e-02 -4.82708722e-01 -7.36372292e-01 6.27924740e-01 -2.07470218e-03 -5.38221538e-01 1.40677345e+00 -2.99163163e-01 -3.87725621e-01 -3.66594970e-01 4.88243848e-01 1.31603464e-01 -6.00189626e-01 1.88378245e-01 -1.82028472e-01 -6.13885581e-01 -4.30970162e-01 -8.69337142e-01 -1.01464677e+00 -2.21228320e-02 -5.22841103e-02 7.44820774e-01 1.00770688e+00 -1.98131148e-02 3.32866669e-01 2.60975659e-01 1.35515690e+00 -1.09550631e+00 -8.98155123e-02 6.95065737e-01 7.50960946e-01 -3.83630484e-01 -7.77406693e-02 -7.25685000e-01 -5.44462442e-01 1.16502237e+00 7.55478382e-01 -1.26688316e-01 8.51194635e-02 5.87825656e-01 -3.65025938e-01 -3.07696968e-01 -6.26874626e-01 -2.16613561e-01 -6.20443523e-01 8.93953919e-01 -6.07800603e-01 1.66033953e-01 -8.61649036e-01 2.81476468e-01 -5.02987385e-01 -1.97698548e-01 1.02967346e+00 1.43481112e+00 -7.09760427e-01 -1.46238983e+00 -6.56733394e-01 -7.07619190e-02 3.02235216e-01 3.12283754e-01 -3.23654979e-01 1.12041509e+00 4.24701571e-01 1.46480751e+00 3.03783357e-01 -3.08928251e-01 5.19926548e-01 -4.66133922e-01 7.44065166e-01 -3.92069221e-01 -4.90173608e-01 1.03484564e-01 5.17204940e-01 -4.99564618e-01 -3.57722044e-01 -8.96117330e-01 -2.03953242e+00 -6.72623813e-01 -1.39981747e-01 6.86639309e-01 6.11145139e-01 8.68014932e-01 3.65290381e-02 4.77037877e-01 6.39273405e-01 -5.97959340e-01 -9.54905078e-02 -1.29092634e-01 -4.61436808e-01 -2.14543968e-01 3.86357941e-02 -8.77599835e-01 -2.84188390e-01 -6.59794927e-01]

[4.982133865356445, 1.8188738822937012]

3e969341-0d79-442b-9667-ac0bec4329c1

r-monet-region-based-unsupervised-scene

https://openreview.net/forum?id=pAJ3svHLDV

https://openreview.net/pdf?id=pAJ3svHLDV

R-MONet: Region-Based Unsupervised Scene Decomposition and Representation via Consistency of Object Representations

Decomposing a complex scene into multiple objects is a natural instinct of an intelligent vision system. Recently, the interest in unsupervised scene representation learning emerged and many previous works tackle this by decomposing scene into object representations either in the form of segmentation masks or position and scale latent variables (i.e. bounding boxes). We observe that these two types of representation both contain object geometric information and should be consistent with each other. Inspired by this observation, we provide an unsupervised generative framework called R-MONet that can generate objects geometric representation in the form of bounding boxes and foreground segmentation masks simultaneously. While bounding boxes can represent the region of interest for generating foreground segmentation masks, the foreground segmentation masks can also be used to supervise bounding boxes learning with the Multi-Otsu Thresholding method. Through the experiments on CLEVR and Multi-dSprites datasets, we show that ensuring the consistency of two types of representation can help the model to decompose the scene and learn better object geometric representations.

['Shengxin Qian']

2021-01-01

['foreground-segmentation']

['computer-vision']

[ 6.77477300e-01 2.66463608e-01 3.58017795e-02 -5.91844022e-01 -3.52015167e-01 -7.78091788e-01 8.60527694e-01 -2.59355810e-02 1.25951588e-01 4.24850315e-01 -3.98414470e-02 -1.84567610e-03 -1.10856786e-01 -1.17887890e+00 -8.64896297e-01 -9.85234261e-01 2.35260099e-01 6.51670694e-01 5.53057730e-01 1.86396226e-01 3.43419760e-01 7.02773988e-01 -1.81049407e+00 4.62269813e-01 8.39652300e-01 7.66262650e-01 4.34915602e-01 6.37897074e-01 -4.27568376e-01 8.70991766e-01 -7.77179837e-01 7.99099505e-02 4.98421758e-01 -4.83344674e-01 -8.96067262e-01 8.64600539e-01 3.47367644e-01 -1.85919758e-02 1.09966286e-01 1.21679294e+00 -1.98352873e-01 1.88763723e-01 1.02230513e+00 -1.30390048e+00 -6.59589589e-01 5.76518774e-01 -6.68753207e-01 6.23013563e-02 -6.62845373e-02 -2.37912312e-01 9.05274510e-01 -6.22304499e-01 6.01831973e-01 1.50845706e+00 1.57652080e-01 2.20401585e-01 -1.45872796e+00 -1.52824044e-01 3.10756832e-01 2.55465116e-02 -1.29162073e+00 -2.23802067e-02 1.03587437e+00 -7.78916717e-01 3.77874911e-01 4.55348492e-01 4.31537330e-01 5.87918818e-01 1.60542965e-01 9.73298252e-01 1.18810058e+00 -4.67680931e-01 3.74438941e-01 3.50996405e-01 4.32252973e-01 6.44273162e-01 4.25946712e-01 -3.05175543e-01 -5.74550144e-02 -2.90429462e-02 1.17230475e+00 3.07120860e-01 -1.17831059e-01 -7.69239902e-01 -1.12263513e+00 1.10547507e+00 5.27315021e-01 3.80792588e-01 -3.27228397e-01 1.72276393e-01 -3.17864753e-02 -2.47404784e-01 3.27210099e-01 3.35429966e-01 -7.81315342e-02 5.11664450e-01 -1.15193176e+00 -4.01236713e-02 5.14988065e-01 1.00179446e+00 1.11335111e+00 1.52818128e-01 -1.16935499e-01 8.35029781e-01 5.23210466e-01 3.34527165e-01 3.28051776e-01 -6.45110011e-01 1.89708233e-01 9.77216661e-01 -1.75931886e-01 -1.05910695e+00 -1.02989808e-01 -6.41274825e-02 -8.91153395e-01 2.18424633e-01 3.27016830e-01 1.67118937e-01 -1.43592107e+00 1.37037909e+00 3.81363273e-01 2.02448100e-01 2.44157836e-02 7.31602669e-01 7.52120435e-01 9.48405504e-01 4.09246683e-02 -8.72169156e-03 1.36419308e+00 -9.54418421e-01 -6.07490003e-01 -1.47799894e-01 2.74808079e-01 -7.56666303e-01 5.91310918e-01 4.73083049e-01 -9.23740387e-01 -9.83070791e-01 -9.88321364e-01 -1.66354459e-02 -5.37180841e-01 4.92724687e-01 6.57736361e-01 7.16725469e-01 -8.54583919e-01 2.43318841e-01 -9.19510245e-01 -2.12294146e-01 3.86342913e-01 1.11055747e-01 -7.96631873e-02 1.30026445e-01 -5.52239478e-01 6.77431643e-01 9.01389897e-01 1.19651549e-01 -1.13356280e+00 -2.36567378e-01 -8.65220726e-01 -7.67930672e-02 3.58948588e-01 -4.43772435e-01 7.12348044e-01 -1.24730742e+00 -1.21829689e+00 1.02447736e+00 -2.08690763e-01 -4.40386623e-01 3.05478662e-01 -1.41970292e-01 -1.14889434e-02 3.67745161e-01 1.58117265e-01 7.29538560e-01 1.34261012e+00 -1.73337960e+00 -6.84228361e-01 -3.30541909e-01 1.02302395e-01 -5.35203610e-04 1.32748127e-01 1.92402881e-02 -4.06894177e-01 -7.46101439e-01 6.58683956e-01 -9.30458903e-01 -4.40138131e-01 -3.19355488e-01 -8.74971449e-01 -3.27711701e-01 1.31338298e+00 -5.14869869e-01 6.30319238e-01 -2.07302976e+00 5.60779274e-01 4.70972121e-01 2.00250328e-01 -6.54764920e-02 1.95738196e-01 -6.99864998e-02 -3.81680757e-01 2.39994794e-01 -5.55311322e-01 -2.44349450e-01 -2.39744484e-01 5.87872863e-01 -5.23212850e-01 5.54979742e-01 6.20529830e-01 7.21653461e-01 -5.80800593e-01 -7.22271085e-01 5.58926165e-01 3.98530036e-01 -3.92975509e-01 2.74319977e-01 -5.04550517e-01 6.03424847e-01 -5.29703796e-01 4.92039502e-01 8.67346704e-01 -2.51503110e-01 1.96936414e-01 -1.79431900e-01 -5.12725413e-02 -3.13594192e-02 -1.62892067e+00 1.14039111e+00 -8.59238859e-03 5.59190869e-01 4.37058844e-02 -1.39842713e+00 1.21946049e+00 -3.38380807e-03 4.65669453e-01 -1.28252000e-01 1.99109182e-01 -2.30983987e-01 -2.20451176e-01 -1.61857113e-01 5.85009038e-01 -1.42977878e-01 -4.46485057e-02 4.37765300e-01 1.83127865e-01 -5.06053627e-01 4.01796281e-01 2.37968698e-01 4.80146050e-01 1.78178281e-01 1.90204486e-01 -4.41665918e-01 7.32771516e-01 5.16572548e-03 4.77293015e-01 6.78414047e-01 3.90547097e-01 7.81768858e-01 4.19778109e-01 -3.53306592e-01 -9.71075952e-01 -1.30671787e+00 -2.96678364e-01 1.12940633e+00 2.70766884e-01 -1.74263820e-01 -8.51210713e-01 -5.53069472e-01 -1.99776858e-01 7.13388503e-01 -7.08437502e-01 1.33489728e-01 -6.17615044e-01 -7.70800889e-01 1.52895153e-01 6.68920696e-01 4.48588341e-01 -1.14049435e+00 -8.95324707e-01 9.19550378e-03 -1.30315751e-01 -1.12831223e+00 -6.42254576e-02 4.28604662e-01 -1.01705766e+00 -1.06240416e+00 -6.78261101e-01 -9.52354252e-01 1.20821679e+00 4.28788483e-01 1.08668220e+00 4.52899411e-02 -5.38875878e-01 5.06768286e-01 -5.52133679e-01 -5.59821069e-01 -5.16599476e-01 -2.96891600e-01 -1.87032923e-01 5.44871390e-01 4.65790294e-02 -4.59874064e-01 -4.60780635e-02 2.83690095e-01 -1.38119137e+00 4.30924535e-01 5.76772451e-01 3.29689533e-01 1.06750500e+00 3.40343386e-01 5.15208393e-03 -1.17423522e+00 1.36214584e-01 -3.89716417e-01 -9.89160538e-01 3.56355369e-01 3.14510576e-02 2.64819801e-01 4.80354965e-01 -3.85045409e-01 -1.13616264e+00 3.31719339e-01 4.38851058e-01 -5.34027576e-01 -6.80619299e-01 2.56879032e-01 -4.19658571e-01 1.92783296e-01 3.89325410e-01 5.52401543e-01 -3.55638057e-01 -4.95729923e-01 7.44752645e-01 4.14129525e-01 4.96339202e-01 -6.87445700e-01 1.15002751e+00 7.32102156e-01 1.78754911e-01 -1.17473018e+00 -7.31730700e-01 -6.80551529e-01 -1.21222067e+00 -1.96045280e-01 1.37538052e+00 -6.48704827e-01 3.79260406e-02 3.64510775e-01 -1.20212662e+00 -7.56744295e-02 -3.96977723e-01 1.01939626e-01 -7.10372448e-01 5.40951312e-01 -1.91016883e-01 -8.85879874e-01 1.30427644e-01 -1.23166919e+00 1.29308796e+00 4.07303780e-01 -3.02273110e-02 -9.42824423e-01 -1.49965942e-01 3.83050174e-01 -8.98803994e-02 5.80476999e-01 1.13744009e+00 -6.43475175e-01 -1.12608171e+00 5.79149053e-02 -3.06722254e-01 5.33927619e-01 3.91273916e-01 3.22227836e-01 -9.28233862e-01 -4.12241109e-02 2.52114773e-01 -1.14245387e-02 8.88122499e-01 3.79772961e-01 1.23403728e+00 -1.93721101e-01 -2.78236210e-01 6.51430309e-01 1.60256147e+00 3.36416513e-01 7.68475413e-01 8.33348781e-02 9.49200213e-01 6.61053360e-01 3.70705456e-01 3.57142091e-01 9.56722815e-03 5.62968910e-01 3.41969669e-01 -2.16345593e-01 -7.36154243e-02 -2.04239544e-02 2.29477867e-01 5.27193725e-01 -2.01268643e-01 -2.92984378e-02 -8.98749352e-01 5.22452235e-01 -1.73163235e+00 -1.02975690e+00 -4.58605140e-01 1.88988233e+00 6.30899668e-01 8.50825086e-02 1.73050985e-01 1.94912285e-01 9.77766931e-01 1.19656317e-01 -2.40216225e-01 -3.17322761e-01 -1.90505773e-01 2.02516571e-01 2.81243443e-01 2.66834348e-01 -1.28034949e+00 1.03664315e+00 6.19975853e+00 6.32040799e-01 -1.07634580e+00 -2.05252215e-01 7.25587010e-01 5.39883852e-01 -1.77395120e-01 2.12208584e-01 -8.97808194e-01 3.61719638e-01 4.63330686e-01 5.87662645e-02 1.54515088e-01 1.19777095e+00 -1.20696753e-01 -4.05910850e-01 -1.14523089e+00 8.51545095e-01 2.29713023e-01 -1.27731681e+00 5.94889343e-01 8.48233402e-02 9.84062314e-01 -4.05185759e-01 -4.67443913e-02 1.47843771e-02 5.22189379e-01 -1.01489115e+00 9.63268042e-01 5.89531302e-01 2.75298506e-01 -5.66823006e-01 4.21353430e-01 4.98660058e-01 -1.52148986e+00 1.38107851e-01 -6.62308872e-01 1.91026062e-01 1.99377969e-01 5.81867456e-01 -9.47879732e-01 7.25589752e-01 4.88129437e-01 4.69679564e-01 -9.33580875e-01 1.00764883e+00 -4.33245122e-01 6.84300840e-01 -4.12999034e-01 2.66289949e-01 2.83851862e-01 -6.91539049e-01 4.47294772e-01 1.06428099e+00 -6.71472698e-02 8.36902335e-02 4.66906607e-01 1.48361492e+00 2.27334127e-01 -6.25726804e-02 -4.44128215e-01 -1.01390995e-01 1.40635641e-02 1.38977110e+00 -1.46863651e+00 -6.61921084e-01 -1.71465844e-01 8.00528109e-01 6.72030300e-02 3.14486027e-01 -1.00803745e+00 -1.11976027e-01 2.59028345e-01 1.56974364e-02 6.81050003e-01 -5.42811632e-01 -5.35414934e-01 -1.08919966e+00 -2.06671253e-01 -5.02591908e-01 1.52036950e-01 -8.06162953e-01 -1.04837203e+00 5.42986989e-01 4.47382569e-01 -1.12353337e+00 -5.52743152e-02 -8.09262395e-01 -7.97145307e-01 6.31677687e-01 -1.27009749e+00 -1.25303710e+00 -4.35477555e-01 7.24213898e-01 8.34967911e-01 6.61288500e-02 4.83198762e-01 -1.95521384e-01 -5.52037597e-01 -1.20354161e-01 7.49823228e-02 3.57109100e-01 1.37621939e-01 -1.47388649e+00 5.24693839e-02 1.13900459e+00 7.20875323e-01 6.99937224e-01 6.86886191e-01 -6.83226824e-01 -1.07393837e+00 -1.35216749e+00 2.68332869e-01 -6.19842410e-01 2.22881600e-01 -5.20002365e-01 -8.60602438e-01 7.64816642e-01 3.89175974e-02 -2.21816555e-01 6.03006423e-01 -1.19638130e-01 -2.27094144e-01 1.29312217e-01 -9.48993027e-01 4.01314884e-01 5.97694516e-01 -3.10658485e-01 -1.02297688e+00 4.79320824e-01 5.01153231e-01 -1.33169696e-01 -4.49930310e-01 3.65037084e-01 -9.72352251e-02 -9.59642589e-01 1.05907083e+00 -5.08124352e-01 3.37307930e-01 -6.32604003e-01 -2.39249095e-01 -9.17627692e-01 -3.47877115e-01 -1.43772945e-01 8.53765756e-02 1.36659837e+00 1.05724782e-01 -2.52869993e-01 7.72612691e-01 4.19304013e-01 -4.20667715e-02 -3.72745216e-01 -5.86119175e-01 -6.60993695e-01 -1.60689652e-01 -3.91465694e-01 3.15567404e-01 7.96510994e-01 -7.45846331e-01 2.46861935e-01 -2.21491121e-02 3.91354352e-01 8.04888546e-01 7.07570910e-01 9.80219305e-01 -1.49701715e+00 -1.95738256e-01 -3.51716012e-01 -5.99701226e-01 -1.10314357e+00 6.63552433e-02 -8.93509924e-01 2.13987038e-01 -1.71919000e+00 2.37773001e-01 -3.62831026e-01 2.16527767e-02 5.05244672e-01 4.32899818e-02 2.90730566e-01 3.21729451e-01 1.98915005e-01 -5.82370102e-01 4.63122010e-01 1.15410578e+00 -2.23363250e-01 -2.71050334e-01 1.57737453e-02 -5.52401245e-01 1.17344356e+00 6.36063695e-01 -4.12654638e-01 -4.56246108e-01 -1.36536241e-01 -1.30338490e-01 -2.56620795e-01 5.17022967e-01 -1.03831351e+00 2.36906055e-02 -4.66393471e-01 4.98789430e-01 -1.02294242e+00 2.29859665e-01 -9.37549233e-01 2.27615431e-01 9.99807417e-02 -1.77079678e-01 -2.63171583e-01 6.76085055e-02 4.97846365e-01 -2.24758640e-01 -3.77978683e-01 1.00930619e+00 -4.01296794e-01 -8.92569244e-01 -1.93266105e-02 -5.84879458e-01 -1.80852994e-01 1.29669237e+00 -6.20568931e-01 -5.67497276e-02 -7.64144287e-02 -8.17928135e-01 -1.95847657e-02 2.91613251e-01 5.02747655e-01 7.76282847e-01 -1.20597160e+00 -5.09904027e-01 4.78350222e-01 -1.00464158e-01 3.73047501e-01 1.02819338e-01 5.06859124e-01 -7.02509522e-01 4.37512338e-01 -3.58184189e-01 -1.01874709e+00 -1.38285947e+00 6.82816029e-01 1.06573649e-01 -1.61975652e-01 -4.59946513e-01 9.72995758e-01 7.84245908e-01 -1.47992998e-01 1.05351299e-01 -6.42092049e-01 -3.56973916e-01 2.26885051e-01 3.19691181e-01 1.94121704e-01 -2.85692543e-01 -1.04020762e+00 -2.61658937e-01 7.76295483e-01 -9.14513320e-02 2.12478805e-02 1.34123135e+00 -4.09974530e-02 -5.27842641e-01 5.39078474e-01 9.59577322e-01 -6.22067079e-02 -1.14090323e+00 -1.70608789e-01 1.42748848e-01 -4.99517888e-01 -1.05969734e-01 -2.40082249e-01 -1.04462147e+00 8.54504943e-01 3.13354939e-01 7.15578020e-01 1.09690464e+00 4.07923162e-01 1.97212368e-01 2.84708768e-01 4.53503758e-01 -9.59212959e-01 4.62750375e-01 4.58666354e-01 9.16508794e-01 -8.60088050e-01 1.64069250e-01 -9.19999480e-01 -7.14076698e-01 1.25619280e+00 6.06829643e-01 -5.89582682e-01 5.88088572e-01 1.91082254e-01 -1.25956774e-01 -3.41981053e-01 -2.91775793e-01 -4.63817120e-01 7.29569197e-01 6.28762007e-01 2.54960239e-01 1.40074655e-01 1.55596301e-01 5.31812072e-01 -2.91038901e-01 -5.70974350e-01 5.47034442e-01 9.95620906e-01 -7.12003648e-01 -1.02808571e+00 -7.90100634e-01 2.74809062e-01 -1.49341464e-01 1.84294462e-01 -6.36326671e-01 6.52776539e-01 4.66334045e-01 7.96018302e-01 2.34215245e-01 -1.18020140e-01 1.20271510e-02 1.29938647e-01 5.53897798e-01 -1.10380471e+00 -2.52354175e-01 3.12971026e-01 -5.02995193e-01 -3.47170889e-01 -6.96144462e-01 -5.43476284e-01 -1.32983947e+00 3.02144736e-01 -5.97636998e-01 5.88629358e-02 4.04830575e-01 1.07434845e+00 -1.88523769e-01 6.79074645e-01 5.24237573e-01 -1.03605664e+00 -4.29260135e-02 -8.70768547e-01 -6.42336786e-01 5.34788489e-01 1.41042560e-01 -7.75038779e-01 -2.86457539e-01 8.06424141e-01]

[9.640202522277832, 0.7493057250976562]

b0486d5a-a4e1-4fa6-a92f-e7406286e631

combinational-q-learning-for-dou-di-zhu

1901.08925

http://arxiv.org/abs/1901.08925v2

http://arxiv.org/pdf/1901.08925v2.pdf

Combinational Q-Learning for Dou Di Zhu

Deep reinforcement learning (DRL) has gained a lot of attention in recent years, and has been proven to be able to play Atari games and Go at or above human levels. However, those games are assumed to have a small fixed number of actions and could be trained with a simple CNN network. In this paper, we study a special class of Asian popular card games called Dou Di Zhu, in which two adversarial groups of agents must consider numerous card combinations at each time step, leading to huge number of actions. We propose a novel method to handle combinatorial actions, which we call combinational Q-learning (CQL). We employ a two-stage network to reduce action space and also leverage order-invariant max-pooling operations to extract relationships between primitive actions. Results show that our method prevails over state-of-the art methods like naive Q-learning and A3C. We develop an easy-to-use card game environments and train all agents adversarially from sractch, with only knowledge of game rules and verify that our agents are comparative to humans. Our code to reproduce all reported results will be available online.

['Liangwei Li', 'Yang You', 'Baisong Guo', 'Cewu Lu', 'Weiming Wang']

2019-01-24

['card-games']

['playing-games']

[-1.14414811e-01 -7.33170286e-02 2.76200101e-03 1.30208924e-01 -6.15163624e-01 -9.34054732e-01 8.38382781e-01 -4.27020967e-01 -8.32856238e-01 8.85338843e-01 -1.64213404e-01 -3.31167936e-01 -5.88050112e-02 -1.07669628e+00 -8.91271830e-01 -6.05407417e-01 -2.35984668e-01 5.63074768e-01 4.06160831e-01 -8.88885677e-01 1.34303570e-01 2.23712951e-01 -1.22971404e+00 1.74671292e-01 8.93240333e-01 7.75045812e-01 -2.30768055e-01 7.62328506e-01 4.70748872e-01 1.32766581e+00 -9.10523951e-01 -9.27752316e-01 8.59037638e-01 -7.25359380e-01 -6.78029954e-01 -2.41914406e-01 2.28140578e-01 -8.37344527e-01 -7.32193232e-01 1.27600336e+00 7.92465508e-01 3.83969575e-01 3.77029777e-01 -1.48491311e+00 -5.32141447e-01 6.94130123e-01 -2.84090638e-01 7.41240829e-02 3.07636201e-01 7.45268106e-01 9.41760898e-01 -2.41672188e-01 3.97255719e-01 1.25601673e+00 4.44064707e-01 8.91445458e-01 -8.93231094e-01 -1.05442798e+00 9.22540948e-02 3.45904171e-01 -8.62795293e-01 -4.08506989e-02 6.63360357e-01 -2.87112892e-01 9.86050427e-01 -1.16398960e-01 8.17247212e-01 1.34807050e+00 4.88658488e-01 8.72154117e-01 1.29205036e+00 -1.13517165e-01 5.59160471e-01 -6.21411145e-01 -6.82965457e-01 6.13086939e-01 4.46668267e-03 7.70587027e-01 -2.72822857e-01 -7.54586160e-02 1.08583808e+00 1.52849510e-01 2.15536311e-01 -4.04187888e-01 -9.92655158e-01 1.25313616e+00 5.86590707e-01 -9.19322073e-02 -3.95387292e-01 5.74948072e-01 5.28634667e-01 7.22014427e-01 1.47740886e-01 8.70881379e-01 -3.70067149e-01 -3.02700162e-01 -3.92718256e-01 1.02313697e+00 7.79590726e-01 6.83275461e-01 3.52506101e-01 2.12359339e-01 -2.30429471e-01 4.99951601e-01 -1.55409649e-01 4.92342412e-01 4.94381726e-01 -1.31559503e+00 6.13357246e-01 1.97901964e-01 2.08921418e-01 -7.32955098e-01 -3.68827701e-01 -1.68863922e-01 -8.47809374e-01 9.19550300e-01 6.79799199e-01 -6.01081252e-01 -8.58183384e-01 1.92057824e+00 9.56496671e-02 2.91567743e-01 1.38321862e-01 9.61678803e-01 4.38610941e-01 5.01743793e-01 6.05073646e-02 7.05989823e-02 1.02471197e+00 -1.12228954e+00 -3.89550567e-01 -3.43675315e-01 4.67982054e-01 -3.04319978e-01 9.36688602e-01 7.02234805e-01 -1.30953944e+00 -4.98348534e-01 -1.03986800e+00 3.28479141e-01 -3.02859306e-01 -4.95328367e-01 1.06170535e+00 7.38858461e-01 -9.04520094e-01 9.80089784e-01 -8.93855691e-01 1.61964912e-02 6.72052920e-01 6.69285059e-01 -2.54789293e-01 -1.52623365e-02 -1.63525712e+00 9.88621771e-01 6.18432641e-01 -8.02515745e-02 -1.49257386e+00 -3.02996814e-01 -8.91542614e-01 -1.28979236e-01 1.09924960e+00 -5.49164116e-01 1.64589608e+00 -1.16880190e+00 -2.23354793e+00 4.26720858e-01 8.13793242e-01 -7.03344345e-01 9.22826529e-01 -4.52582240e-01 -5.88988792e-03 1.09807700e-01 -1.38521194e-02 7.15953171e-01 7.63577998e-01 -6.40704215e-01 -5.91246605e-01 -1.86403587e-01 1.08294427e+00 4.99541759e-01 9.15653184e-02 1.68329567e-01 -3.32725495e-01 -9.15485919e-01 -5.93053043e-01 -1.17288876e+00 -7.05710292e-01 -1.33079305e-01 -2.33235434e-01 -2.89872020e-01 1.93627372e-01 -4.18896317e-01 8.44396472e-01 -1.80064380e+00 3.54117155e-01 -8.15028474e-02 3.36959511e-01 3.67284477e-01 -3.51139128e-01 4.61019456e-01 1.49648085e-01 2.21460219e-02 -2.59610355e-01 -1.12887606e-01 3.14161718e-01 4.10749465e-01 -2.90108591e-01 4.68466789e-01 1.74608693e-01 1.17449141e+00 -1.24567020e+00 -1.91329017e-01 2.58619696e-01 -1.32919163e-01 -1.05466366e+00 4.56182957e-01 -5.29808819e-01 5.38743377e-01 -5.14614820e-01 4.30957973e-01 5.11810303e-01 3.34627509e-01 9.34854448e-02 5.12377739e-01 1.45032138e-01 1.09651782e-01 -1.05577803e+00 1.70627153e+00 -2.74151534e-01 8.05552453e-02 -1.35868937e-01 -9.87461686e-01 5.22544026e-01 2.46556416e-01 3.95870864e-01 -1.01388264e+00 4.38071072e-01 1.31266624e-01 4.21028495e-01 -1.77136198e-01 2.56540537e-01 -3.42244238e-01 -5.87198853e-01 4.34670150e-01 1.37389645e-01 -4.97988284e-01 4.31577057e-01 5.96477166e-02 1.27208412e+00 3.82653326e-01 5.07078111e-01 2.14819953e-01 2.73553073e-01 -1.35254813e-02 7.98531592e-01 1.27171016e+00 -4.91951227e-01 3.15948486e-01 8.42674673e-01 -5.10025382e-01 -1.05722916e+00 -1.15226614e+00 5.56726098e-01 1.31726849e+00 3.18172157e-01 -2.23915562e-01 -7.96306014e-01 -9.40378726e-01 -1.18543617e-01 4.25094575e-01 -7.64406741e-01 -4.22128260e-01 -9.24148262e-01 -5.68943977e-01 7.40178168e-01 7.62568593e-01 1.06680298e+00 -1.56130791e+00 -7.28564560e-01 2.22238332e-01 2.57725626e-01 -7.79512167e-01 -5.76257050e-01 2.20283270e-01 -4.84570891e-01 -1.25897598e+00 -7.60954797e-01 -6.32587731e-01 1.51228249e-01 -2.01721027e-01 1.15377343e+00 -5.40548116e-02 1.82549749e-02 -8.09467584e-02 -6.32034600e-01 -4.54184920e-01 -4.27737802e-01 2.65411218e-03 3.31617296e-01 -4.35390890e-01 1.77931324e-01 -5.40728927e-01 -6.24659419e-01 3.59122306e-01 -9.67949569e-01 -5.55657111e-02 6.94066763e-01 9.50236082e-01 2.25284383e-01 2.49047533e-01 4.43675190e-01 -9.30347860e-01 7.88089275e-01 -2.62516916e-01 -1.06834233e+00 -6.01973198e-02 -1.16762100e-03 2.74054017e-02 1.23774409e+00 -6.15318358e-01 -6.89796686e-01 -2.75435597e-02 -2.53899276e-01 -5.34341216e-01 2.68519279e-02 1.62848935e-01 -4.55760926e-01 -1.54417381e-01 7.04417109e-01 -1.17231775e-02 -7.00643361e-02 -4.21687253e-02 7.10159421e-01 2.36324713e-01 5.33675194e-01 -6.77439630e-01 1.18142891e+00 2.11249009e-01 -8.93293321e-02 -8.83581210e-03 -6.51831686e-01 5.19201942e-02 -3.81418586e-01 -1.85902789e-01 9.84431207e-01 -8.40545058e-01 -1.29499567e+00 1.08103991e+00 -9.06106651e-01 -8.29812944e-01 -3.53086352e-01 5.14881730e-01 -9.28375661e-01 1.18661173e-01 -9.73915935e-01 -5.76681435e-01 -8.30517188e-02 -1.23930395e+00 5.64118981e-01 3.82192820e-01 2.46557295e-01 -7.22119510e-01 4.49291050e-01 2.48831958e-01 3.37391943e-01 4.29186374e-01 4.78351623e-01 -8.60614896e-01 -5.42726278e-01 2.61901934e-02 2.32210591e-01 3.36929232e-01 -7.86178336e-02 -5.61200082e-01 -6.38457239e-01 -6.25895858e-01 -1.04356602e-01 -9.06153619e-01 8.10397923e-01 3.27619255e-01 1.18717790e+00 -4.43280667e-01 2.07817763e-01 6.86922967e-01 1.17888105e+00 6.43252909e-01 7.77789950e-01 5.62508166e-01 6.17908239e-01 2.73978502e-01 5.79296410e-01 5.23744822e-01 1.26929104e-01 7.90608823e-01 8.06213677e-01 1.10798948e-01 3.80397469e-01 -3.48296404e-01 6.12066984e-01 3.30104023e-01 -3.47241372e-01 -2.70442307e-01 -5.31222701e-01 1.79042578e-01 -1.95066476e+00 -1.32370412e+00 6.73558354e-01 1.97449172e+00 8.35636914e-01 5.89287817e-01 6.57377183e-01 -7.81084821e-02 4.72408473e-01 2.05272093e-01 -9.35730517e-01 -5.17894804e-01 8.73003826e-02 7.63249218e-01 7.57266164e-01 4.19726163e-01 -1.48271251e+00 1.44338381e+00 6.37151623e+00 1.10414231e+00 -7.66302764e-01 1.02701940e-01 5.28103709e-01 -1.99023768e-01 1.80919841e-01 -1.98777094e-01 -2.46409044e-01 4.39425975e-01 6.54848814e-01 -1.58398509e-01 9.17224526e-01 9.90777612e-01 7.37178251e-02 9.28500369e-02 -1.07169938e+00 8.93028915e-01 -9.87555683e-02 -1.16730487e+00 -8.81058201e-02 1.06177799e-01 8.45225751e-01 -1.23684645e-01 3.40263844e-01 1.06682789e+00 1.33469558e+00 -1.42406952e+00 7.50912905e-01 1.22369207e-01 6.59758508e-01 -1.14424312e+00 8.54833722e-01 3.83043200e-01 -1.09998655e+00 -3.98885936e-01 -4.65952903e-01 -5.69799542e-01 -1.39380559e-01 -3.34492058e-01 -3.10982257e-01 5.92875481e-01 6.10761404e-01 5.99719644e-01 -3.26641381e-01 9.78433967e-01 -6.94979250e-01 4.96407121e-01 -1.02271371e-01 -2.00963482e-01 5.43305218e-01 -2.54472375e-01 2.24437848e-01 5.79067290e-01 1.75564755e-02 4.15179193e-01 6.01535201e-01 7.69549727e-01 -2.32035488e-01 -1.28642365e-01 -6.22856736e-01 -1.14796318e-01 2.66822368e-01 8.46388042e-01 -6.03525817e-01 -1.22038782e-01 -4.84462708e-01 1.31548989e+00 3.95785600e-01 8.17728639e-02 -1.20221984e+00 -5.32424808e-01 8.57038140e-01 -3.24966669e-01 4.21444446e-01 -1.99535713e-01 1.58340037e-01 -1.25889456e+00 -1.41114801e-01 -1.35545158e+00 4.09958303e-01 -4.12377208e-01 -1.51783681e+00 4.55362976e-01 -5.14498539e-02 -1.48103750e+00 -5.25516748e-01 -8.33027899e-01 -9.57473874e-01 5.16886532e-01 -1.16249108e+00 -9.67449903e-01 1.33706033e-01 8.77623677e-01 5.85233271e-01 -5.36465108e-01 6.38364851e-01 6.15011379e-02 -5.07906556e-01 8.19437385e-01 1.32538140e-01 6.10642612e-01 4.92532283e-01 -1.52939224e+00 7.20232606e-01 8.68006647e-01 3.07385251e-02 2.21637875e-01 6.48549974e-01 -6.40225530e-01 -1.32730412e+00 -1.03293920e+00 1.68795232e-02 -4.52985257e-01 7.68956900e-01 -4.61992383e-01 -3.46403807e-01 6.77156270e-01 2.85209715e-01 -1.52824176e-02 3.46453667e-01 -2.20405415e-01 -2.68734813e-01 -1.04152910e-01 -1.01089680e+00 1.07723594e+00 1.22443807e+00 -3.84074152e-01 -8.38232040e-01 2.28605613e-01 8.47238362e-01 -6.57271087e-01 -4.25835490e-01 2.03980997e-01 2.74570674e-01 -1.11110759e+00 9.57398772e-01 -1.01049173e+00 5.93413293e-01 -2.25937307e-01 7.91887492e-02 -1.66180730e+00 -2.82285273e-01 -1.05141354e+00 6.79525957e-02 4.96572614e-01 1.07901461e-01 -6.57535255e-01 1.00922120e+00 4.08264548e-01 -2.20632385e-02 -6.52655005e-01 -1.02995980e+00 -1.09824729e+00 7.46298134e-01 -3.06691080e-01 5.15100300e-01 7.69744456e-01 1.48257673e-01 2.25449950e-02 -8.64928544e-01 2.17101760e-02 5.27947426e-01 -1.33546442e-01 9.30866182e-01 -7.82948196e-01 -9.29821968e-01 -6.26875997e-01 -4.95976865e-01 -1.05609000e+00 2.98652440e-01 -6.91242456e-01 9.15116295e-02 -1.16507638e+00 6.73404858e-02 -1.48361444e-01 -3.17296922e-01 5.13402104e-01 -3.13599139e-01 3.76088381e-01 5.14494717e-01 -8.65197182e-03 -9.45768178e-01 8.31269205e-01 1.54374433e+00 -2.25271061e-01 -1.99626997e-01 2.32638389e-01 -6.14399493e-01 7.11798608e-01 9.28293705e-01 -3.84345978e-01 -4.36460316e-01 -1.18838422e-01 3.03087860e-01 1.17139637e-01 4.16974962e-01 -1.20085919e+00 1.71898574e-01 -5.80061615e-01 2.87665904e-01 4.92495745e-02 3.37354958e-01 -5.29260516e-01 -1.77359849e-01 8.57603669e-01 -3.46431792e-01 2.70766079e-01 7.42725283e-02 4.13265884e-01 -1.09833144e-01 -1.33468479e-01 7.86670446e-01 -5.14094770e-01 -6.80724502e-01 6.23013139e-01 -5.87218761e-01 3.70150357e-01 1.29687417e+00 9.06567499e-02 -2.53654182e-01 -7.08609879e-01 -5.27602911e-01 4.04691666e-01 4.30498689e-01 2.95850843e-01 4.05065507e-01 -1.36918485e+00 -8.02124143e-01 -1.62693754e-01 -1.46689847e-01 -8.92637670e-02 3.12457621e-01 2.69774050e-01 -8.43699098e-01 9.23718065e-02 -7.01639235e-01 1.69972807e-01 -8.20009947e-01 9.12352622e-01 7.15059757e-01 -7.24435151e-01 -5.70418894e-01 8.78574014e-01 2.80510247e-01 -6.88518882e-01 1.32184476e-01 -1.04954161e-01 -2.55189687e-01 -3.17381322e-01 5.21084964e-01 1.85451850e-01 -3.20041358e-01 -2.74658322e-01 -2.48609632e-01 3.29912961e-01 -2.63783813e-01 -2.51859546e-01 1.28705692e+00 6.29601300e-01 3.31559956e-01 -9.92813259e-02 6.43522620e-01 -1.94293737e-01 -1.79202402e+00 -1.00181498e-01 -3.10493529e-01 -4.33583349e-01 -3.84381235e-01 -7.40887642e-01 -1.22575164e+00 6.92020655e-01 4.89156008e-01 6.86448589e-02 1.13454127e+00 -2.34183684e-01 8.69561136e-01 6.55694187e-01 7.26390183e-01 -1.19913423e+00 6.32561147e-01 9.01561320e-01 6.45994782e-01 -1.22254312e+00 -2.47038737e-01 5.28254472e-02 -8.85424733e-01 8.32491875e-01 9.74191070e-01 -8.15724611e-01 2.76035279e-01 2.08270162e-01 2.01308765e-02 9.22800750e-02 -7.18944311e-01 -4.96235788e-01 -1.64931729e-01 8.34877253e-01 -8.83171707e-02 1.34610072e-01 -1.37639835e-01 7.05829442e-01 -4.99494404e-01 -5.25975823e-02 5.42854428e-01 1.04585278e+00 -3.19006473e-01 -1.37738645e+00 -1.92947611e-01 1.94047973e-01 -6.37112081e-01 -1.54576600e-01 -3.96250159e-01 8.54035258e-01 3.23478550e-01 9.76813018e-01 -9.33696330e-02 -5.50578535e-01 4.34004635e-01 -3.78328860e-01 7.22423136e-01 -4.78841126e-01 -9.96325135e-01 -2.61371821e-01 -8.66513550e-02 -9.09403980e-01 -2.59499520e-01 -4.90416110e-01 -9.88214135e-01 -6.71841741e-01 4.96920235e-02 1.38855837e-02 -1.31918594e-01 1.09498262e+00 -1.93200037e-01 5.82516849e-01 6.68927431e-01 -1.03787374e+00 -9.37164605e-01 -7.84848928e-01 -6.66125119e-01 5.46200573e-01 6.24587573e-02 -8.11843038e-01 -1.76867545e-01 -3.71581972e-01]

[3.6487374305725098, 1.5051696300506592]

f24e3060-a55e-49ce-baae-dd658fc2ec16

one-step-bipartite-graph-cut-a-normalized

2305.07386

https://arxiv.org/abs/2305.07386v1

https://arxiv.org/pdf/2305.07386v1.pdf

One-step Bipartite Graph Cut: A Normalized Formulation and Its Application to Scalable Subspace Clustering

The bipartite graph structure has shown its promising ability in facilitating the subspace clustering and spectral clustering algorithms for large-scale datasets. To avoid the post-processing via k-means during the bipartite graph partitioning, the constrained Laplacian rank (CLR) is often utilized for constraining the number of connected components (i.e., clusters) in the bipartite graph, which, however, neglects the distribution (or normalization) of these connected components and may lead to imbalanced or even ill clusters. Despite the significant success of normalized cut (Ncut) in general graphs, it remains surprisingly an open problem how to enforce a one-step normalized cut for bipartite graphs, especially with linear-time complexity. In this paper, we first characterize a novel one-step bipartite graph cut (OBCut) criterion with normalized constraints, and theoretically prove its equivalence to a trace maximization problem. Then we extend this cut criterion to a scalable subspace clustering approach, where adaptive anchor learning, bipartite graph learning, and one-step normalized bipartite graph partitioning are simultaneously modeled in a unified objective function, and an alternating optimization algorithm is further designed to solve it in linear time. Experiments on a variety of general and large-scale datasets demonstrate the effectiveness and scalability of our approach.

['Jian-Huang Lai', 'Chang-Dong Wang', 'Dong Huang', 'Si-Guo Fang']

2023-05-12

['graph-partitioning']

['graphs']

[ 8.27371329e-02 1.24210175e-02 -2.30389744e-01 -9.54003334e-02 -6.18111908e-01 -7.18552887e-01 -3.03235371e-03 1.29826397e-01 -6.30358160e-02 5.68174362e-01 -1.82213098e-01 -3.76893610e-01 -6.11933708e-01 -5.75525522e-01 -4.80570048e-01 -9.87097502e-01 -2.33245388e-01 6.20730102e-01 1.61220461e-01 3.00833791e-01 -3.51965986e-02 5.61261117e-01 -1.08165145e+00 -2.99523979e-01 1.16441178e+00 4.48650271e-01 9.42703523e-03 2.97147483e-01 -5.15865237e-02 2.25240901e-01 -1.99873939e-01 -2.51272589e-01 4.27476883e-01 -4.90129888e-01 -8.07437479e-01 5.30379057e-01 1.89259231e-01 2.27523923e-01 -4.27676648e-01 1.40674293e+00 4.00156885e-01 2.56231040e-01 6.74816251e-01 -1.61770332e+00 -3.42084229e-01 6.24097884e-01 -1.09310901e+00 -1.45557851e-01 3.60112518e-01 -2.45767146e-01 1.16412675e+00 -7.11301506e-01 5.97744048e-01 1.02618086e+00 3.68040532e-01 6.46019429e-02 -1.79048014e+00 -6.98276401e-01 2.53119618e-01 1.93650469e-01 -1.92518020e+00 3.88932340e-02 1.15203190e+00 -6.44695401e-01 2.89381176e-01 5.14020801e-01 7.34063506e-01 5.89281321e-01 -2.53564656e-01 5.15535057e-01 1.05967152e+00 -3.44682842e-01 3.20033222e-01 -1.57141965e-02 3.99313271e-01 6.32165551e-01 5.68483531e-01 -4.01102483e-01 -1.84228539e-01 -3.41027617e-01 3.54838997e-01 6.48475811e-02 -5.22670329e-01 -1.06330216e+00 -1.10020339e+00 9.20010269e-01 5.07517219e-01 2.84782380e-01 -1.15157232e-01 -2.10922778e-01 2.70854622e-01 1.47536188e-01 2.86250919e-01 8.57772976e-02 -1.37456924e-01 4.34685379e-01 -1.14059877e+00 -9.69208926e-02 1.03123546e+00 1.17631233e+00 1.14716756e+00 -9.86096933e-02 1.00285225e-01 8.16075206e-01 1.71728045e-01 4.79896933e-01 1.72418803e-02 -7.57485390e-01 3.25937450e-01 1.00701261e+00 -1.09996855e-01 -1.60453880e+00 -6.41521156e-01 -4.22257572e-01 -1.41649795e+00 -2.71032333e-01 5.87050021e-01 2.29253881e-02 -6.92937851e-01 1.81541610e+00 6.35631084e-01 5.95141277e-02 -2.23772123e-01 1.24685693e+00 4.84255046e-01 3.45852435e-01 -3.52946401e-01 -6.70861065e-01 1.07838035e+00 -7.21903205e-01 -6.18934810e-01 9.26155504e-03 5.48628271e-01 -6.10625863e-01 8.81773055e-01 2.97159404e-01 -7.53795981e-01 -7.92747885e-02 -9.14991379e-01 2.60461956e-01 -1.76082835e-01 -7.63769075e-02 7.31364608e-01 6.99801505e-01 -1.00504589e+00 2.75509000e-01 -7.48183846e-01 -5.64720571e-01 1.36814848e-01 5.27581215e-01 -5.46529591e-01 -2.89912075e-01 -9.03389454e-01 1.57472655e-01 5.29915154e-01 2.52466232e-01 -5.15857577e-01 -3.35327327e-01 -7.14061558e-01 1.52964413e-01 8.24600875e-01 -5.46151936e-01 2.80120671e-01 -8.52852404e-01 -1.09738517e+00 8.53056192e-01 -1.95201680e-01 -7.34641925e-02 3.43380749e-01 3.18016142e-01 -2.87771285e-01 2.76009351e-01 9.52405185e-02 3.03879440e-01 6.47406399e-01 -1.45537555e+00 -1.24104597e-01 -6.50435269e-01 9.57561880e-02 2.68115401e-01 -5.37250936e-01 -1.06411539e-01 -9.55135345e-01 -4.85822707e-01 8.93378496e-01 -1.20816338e+00 -4.35403764e-01 -4.83035535e-01 -8.21066856e-01 -2.04568997e-01 6.75125003e-01 -5.72271764e-01 1.59262133e+00 -2.16678214e+00 5.00564873e-01 9.07460868e-01 6.03415310e-01 -2.86410689e-01 -6.31940290e-02 5.30599594e-01 -2.51109838e-01 5.07068560e-02 -5.76207817e-01 8.63099322e-02 -9.69811231e-02 1.26573771e-01 4.12474796e-02 9.15668368e-01 -3.12985569e-01 4.38169807e-01 -9.73931074e-01 -7.21076369e-01 5.00789508e-02 1.17557377e-01 -7.38818467e-01 -2.70698555e-02 1.20382361e-01 5.62690794e-01 -3.42864960e-01 6.45507812e-01 1.00315130e+00 -5.84773719e-01 7.93038130e-01 -4.84460622e-01 2.98534911e-02 -3.65484416e-01 -1.87726700e+00 1.69112337e+00 1.35168642e-01 3.86582434e-01 5.53508401e-01 -1.32877636e+00 7.65019536e-01 8.78527239e-02 9.46205437e-01 -2.25467384e-01 1.00404218e-01 1.74870983e-01 1.42278269e-01 -1.74319595e-01 1.84098959e-01 -5.29993465e-03 -1.91346481e-01 4.04917419e-01 -1.63354576e-01 3.24249208e-01 5.08581638e-01 6.73604071e-01 1.15344465e+00 -2.43469819e-01 1.82584539e-01 -6.53817236e-01 7.11831748e-01 -6.25750190e-03 8.64237368e-01 4.75519300e-01 -2.34373719e-01 6.43150985e-01 8.70210171e-01 -1.09785544e-02 -8.64917457e-01 -9.07525241e-01 -8.78539830e-02 7.48926938e-01 6.15394533e-01 -5.90160608e-01 -8.65850806e-01 -6.32838190e-01 -5.92990890e-02 1.30956739e-01 -4.78855968e-01 -1.42237902e-01 -3.84339631e-01 -9.07953501e-01 9.16897729e-02 1.33941546e-01 8.54410902e-02 -4.66762096e-01 3.97197679e-02 -8.63311142e-02 -3.17892522e-01 -1.05229616e+00 -7.51263082e-01 2.60922015e-01 -7.49228597e-01 -1.46202815e+00 -4.26547617e-01 -7.79341698e-01 1.14191902e+00 6.70509994e-01 7.69829750e-01 1.22880004e-01 -1.46873787e-01 5.00192583e-01 -4.05909508e-01 3.21718425e-01 1.25059634e-01 2.59155333e-01 2.29022533e-01 3.53570461e-01 2.60117114e-01 -8.40955257e-01 -5.67685425e-01 6.29135966e-01 -1.16163099e+00 6.64647296e-02 4.61464137e-01 8.91366303e-01 7.74290800e-01 3.52285326e-01 3.31764966e-01 -1.05356359e+00 4.61096168e-01 -5.08169293e-01 -7.78653681e-01 3.58238727e-01 -6.69888914e-01 -3.35419476e-02 6.48536086e-01 -2.84716904e-01 -5.77263176e-01 3.37530434e-01 3.71148467e-01 -7.59910941e-01 1.40599743e-01 7.14511573e-01 -7.52775490e-01 -2.40072846e-01 2.94555902e-01 2.81543255e-01 -1.39629379e-01 -3.03884983e-01 5.54772973e-01 4.91178274e-01 2.68603146e-01 -5.88460982e-01 1.24454224e+00 6.60743177e-01 3.56287956e-01 -8.48721445e-01 -5.67908049e-01 -9.65234637e-01 -9.84524488e-01 -2.87913948e-01 7.82211065e-01 -6.99261189e-01 -9.46187198e-01 1.25187919e-01 -7.39310563e-01 6.86346442e-02 1.47291049e-01 3.95929158e-01 -3.07791799e-01 9.08057809e-01 -3.84835780e-01 -7.21435785e-01 -5.86319081e-02 -1.02840126e+00 6.31496727e-01 2.65416317e-02 4.37108539e-02 -8.90566289e-01 1.95450008e-01 3.69809777e-01 -1.73980922e-01 2.94297069e-01 9.53164041e-01 -6.95964813e-01 -7.00506926e-01 -1.80468664e-01 -4.62544769e-01 -1.53043680e-02 5.09508289e-02 -1.81095019e-01 -5.54494083e-01 -7.95820475e-01 -2.79080242e-01 -6.34934157e-02 6.76789105e-01 2.13726640e-01 1.08999979e+00 -2.98886061e-01 -5.94468057e-01 7.49993086e-01 1.54308462e+00 8.23319703e-02 2.18560845e-01 -8.63615945e-02 1.07753766e+00 6.17030501e-01 4.19282854e-01 4.37501073e-01 2.58303553e-01 6.10824704e-01 4.09521937e-01 -2.19430417e-01 2.11070850e-01 -1.28351763e-01 1.77538954e-02 1.22809672e+00 2.97210254e-02 -1.53693035e-01 -9.23322618e-01 4.04178202e-01 -2.02099729e+00 -7.27388322e-01 -4.40064877e-01 2.51390576e+00 5.64890325e-01 -1.78286970e-01 2.66926020e-01 2.42143214e-01 1.16946113e+00 1.03267640e-01 -6.14917219e-01 8.52864608e-02 -3.38803619e-01 -2.41361097e-01 6.27705455e-01 5.22401094e-01 -1.15687621e+00 8.66701305e-01 5.62804747e+00 9.72354114e-01 -7.79488087e-01 1.13506829e-02 4.15879160e-01 7.71433339e-02 -2.70711213e-01 4.40532386e-01 -3.18478376e-01 4.03666168e-01 4.23145831e-01 -2.55795121e-01 8.49902093e-01 8.31749678e-01 1.61301628e-01 6.52577206e-02 -8.93118799e-01 1.11787641e+00 3.81512567e-02 -6.53172612e-01 2.22615548e-03 3.45852375e-01 9.52659547e-01 -2.96443999e-01 -3.58374864e-01 3.54592055e-02 2.91651905e-01 -5.47734857e-01 4.07042414e-01 1.95501462e-01 7.80297101e-01 -1.06159151e+00 3.94733995e-01 2.90890634e-01 -1.48361301e+00 -4.30193134e-02 -5.37854254e-01 2.29242429e-01 4.23557907e-02 8.68154943e-01 -4.04528230e-01 1.03380275e+00 6.48105979e-01 5.30786216e-01 -5.15315950e-01 1.28289878e+00 -9.00066420e-02 5.27678728e-01 -4.35615391e-01 2.49157697e-01 1.96304649e-01 -9.94012952e-01 9.35797632e-01 1.01880848e+00 1.43981606e-01 2.25532383e-01 7.09747970e-01 7.60193467e-01 -1.65030047e-01 4.65175301e-01 -4.87435728e-01 -8.21604952e-02 4.89250988e-01 1.63668859e+00 -1.36135149e+00 -3.77158932e-02 -5.04974306e-01 1.01319611e+00 4.61495757e-01 6.89062953e-01 -7.40457833e-01 -4.55214769e-01 4.41253811e-01 1.95040852e-01 3.13313454e-01 -4.29912716e-01 -2.51075625e-01 -1.30667925e+00 1.16168521e-01 -8.65711570e-01 7.48228669e-01 -2.74079293e-01 -1.30024970e+00 3.23506802e-01 -1.39640614e-01 -1.30170536e+00 3.19587111e-01 -2.90264606e-01 -4.80522007e-01 4.91885126e-01 -1.06072962e+00 -1.06161153e+00 -3.85284364e-01 8.76630425e-01 -1.19386822e-01 2.07239121e-01 5.07826269e-01 5.79252958e-01 -9.96558487e-01 4.88821208e-01 4.37955886e-01 1.61009103e-01 6.75525725e-01 -1.40230024e+00 -5.49040437e-01 1.09655762e+00 1.42944872e-01 7.30597913e-01 7.01316774e-01 -7.41956711e-01 -1.79514122e+00 -1.04298151e+00 2.81569958e-01 1.10400520e-01 8.60147834e-01 -5.95048249e-01 -9.11865532e-01 6.49121940e-01 5.20063154e-02 -6.87015578e-02 8.35767865e-01 3.43606889e-01 -4.44116712e-01 -2.94173539e-01 -8.51497591e-01 6.32994235e-01 1.21461439e+00 -2.53179282e-01 -1.14511468e-01 5.19114792e-01 6.56481862e-01 -1.09791212e-01 -8.55478764e-01 4.76764947e-01 2.80067593e-01 -8.66949677e-01 7.89391279e-01 -5.64149678e-01 -1.49713039e-01 -7.04504967e-01 -1.82315856e-01 -1.09018815e+00 -6.44204557e-01 -8.21973324e-01 2.10198108e-02 1.39853632e+00 2.87830830e-01 -5.91118217e-01 8.61257613e-01 4.68785316e-01 1.55036170e-02 -5.75650692e-01 -9.24282968e-01 -8.77026796e-01 -1.38752207e-01 -3.20392877e-01 3.18097532e-01 1.34952688e+00 2.42139041e-01 4.36444432e-01 -4.55210418e-01 5.13227046e-01 9.90260005e-01 6.79092467e-01 8.33534837e-01 -1.37493777e+00 -2.34780878e-01 -4.73885149e-01 -5.45405269e-01 -9.43879008e-01 2.89161384e-01 -1.17947841e+00 -1.15722977e-01 -1.50537574e+00 6.88451469e-01 -4.36077356e-01 -2.95120388e-01 1.88516572e-01 -2.63366818e-01 1.46899194e-01 1.02708183e-01 3.50419879e-01 -1.01938450e+00 5.09679794e-01 1.01323116e+00 -2.66065419e-01 -4.70995277e-01 3.78471091e-02 -5.66616774e-01 6.82223856e-01 4.06686723e-01 -4.92938519e-01 -4.73129570e-01 3.33808511e-02 3.64340752e-01 1.87185988e-01 1.14030451e-01 -8.35853100e-01 5.58391333e-01 -2.02148929e-01 4.66307327e-02 -7.41424322e-01 -3.69886272e-02 -1.26936924e+00 5.60144901e-01 3.08892727e-01 6.40249848e-02 -1.90938368e-01 -1.28864884e-01 8.78894806e-01 -2.63185471e-01 -5.04006222e-02 8.28031778e-01 1.39761269e-01 -2.34845504e-01 5.47062933e-01 -8.66209865e-02 1.35242596e-01 1.31764960e+00 -3.34160745e-01 1.16057992e-01 -2.68180281e-01 -9.74266708e-01 5.73757827e-01 6.75882876e-01 1.20461211e-01 4.12756622e-01 -1.48195446e+00 -5.72633564e-01 1.03094824e-01 1.96911395e-01 7.70718008e-02 4.55965132e-01 1.28385746e+00 -2.74017215e-01 4.02534038e-01 2.67169476e-01 -7.84878671e-01 -1.32067597e+00 9.57830787e-01 1.25946045e-01 -3.67136747e-01 -5.12879729e-01 3.98196459e-01 4.42791313e-01 -3.85460109e-01 2.41722181e-01 3.39832574e-01 -8.83638561e-02 2.22054467e-01 -4.75781597e-02 6.48887575e-01 -1.09535456e-01 -7.65205443e-01 -3.79176706e-01 6.53167725e-01 1.15979284e-01 1.99129209e-01 1.11412990e+00 -4.11516041e-01 -6.25018716e-01 3.60665560e-01 1.12781608e+00 1.44400656e-01 -9.27449644e-01 -2.56379932e-01 1.50300071e-01 -4.63352472e-01 -1.65136561e-01 -8.11983421e-02 -1.30623174e+00 5.52108705e-01 2.57613659e-01 5.48563600e-01 1.33611214e+00 1.65463418e-01 6.12569153e-01 3.17525953e-01 4.71563458e-01 -1.01909769e+00 -2.05387548e-01 1.95070639e-01 6.01414442e-01 -1.02133214e+00 4.45702784e-02 -8.57925594e-01 -3.28821361e-01 9.29034650e-01 6.15316570e-01 5.35203442e-02 7.34784603e-01 1.54262288e-02 -3.17870915e-01 -2.36168489e-01 -1.79246381e-01 -3.13966632e-01 4.29002911e-01 2.44160324e-01 1.59801885e-01 2.74094611e-01 -5.89612722e-01 5.40477276e-01 8.36114585e-02 -5.30972123e-01 3.07675779e-01 5.20424545e-01 -3.12916428e-01 -1.02472341e+00 -4.95197386e-01 4.70214486e-01 -2.75312901e-01 4.00216356e-02 -6.14407778e-01 6.95892334e-01 -9.46744438e-03 1.00703824e+00 -2.42929667e-01 -3.23241383e-01 1.84885174e-01 -1.40642345e-01 2.46108651e-01 -3.59147042e-01 -1.50546908e-01 5.34160972e-01 -3.38311136e-01 -7.07428098e-01 -4.49738145e-01 -6.37038708e-01 -1.19256353e+00 -2.37416580e-01 -6.86596930e-01 4.95505303e-01 2.83298880e-01 6.82134748e-01 3.38200420e-01 3.07313830e-01 9.12297785e-01 -5.77170491e-01 -2.23222941e-01 -7.69342542e-01 -1.08512211e+00 5.82355738e-01 -9.01091620e-02 -6.26711667e-01 -7.81378686e-01 -3.89261283e-02]

[7.3773345947265625, 4.892287254333496]

aa58b3ff-e136-4619-bef0-d1db024562a0

neural-texture-synthesis-with-guided

http://openaccess.thecvf.com//content/CVPR2023/html/Zhou_Neural_Texture_Synthesis_With_Guided_Correspondence_CVPR_2023_paper.html

http://openaccess.thecvf.com//content/CVPR2023/papers/Zhou_Neural_Texture_Synthesis_With_Guided_Correspondence_CVPR_2023_paper.pdf

Neural Texture Synthesis With Guided Correspondence

Markov random fields (MRFs) are the cornerstone of classical approaches to example-based texture synthesis. Yet, it is not fully valued in the deep learning era. This paper aims to re-promote the combination of MRFs and neural networks, i.e., the CNNMRF model, for texture synthesis, with two key observations made. We first propose to compute the Guided Correspondence Distance in the nearest neighbor search, based on which a Guided Correspondence loss is defined to measure the similarity of the output texture to the example. Experiments show that our approach surpasses existing neural approaches in uncontrolled and controlled texture synthesis. More importantly, the Guided Correspondence loss can function as a general textural loss in, e.g., training generative networks for real-time controlled synthesis and inversion-based single-image editing. In contrast, existing textural losses, such as the Sliced Wasserstein loss, cannot work on these challenging tasks.

['Hui Huang', 'Rongjun Xiao', 'Kaijian Chen', 'Yang Zhou']

2023-01-01

cvpr-2023-1

['texture-synthesis']

['computer-vision']

[ 7.36395895e-01 2.04444543e-01 2.89437152e-03 -3.25589240e-01 -5.98057210e-01 -1.33733690e-01 8.49242628e-01 -1.75640225e-01 -1.51069745e-01 7.96587884e-01 -7.47391433e-02 -5.05420088e-04 -3.47918421e-01 -1.14020479e+00 -1.08554482e+00 -8.72645855e-01 4.13692325e-01 4.53579575e-01 7.34894797e-02 -2.37772167e-01 4.26722139e-01 6.63371265e-01 -1.49250996e+00 2.56989807e-01 9.15114224e-01 1.28497076e+00 2.55695432e-01 4.36309844e-01 1.19780809e-01 9.09577549e-01 -5.99030741e-02 -3.87056798e-01 1.72611013e-01 -5.70463955e-01 -9.10758197e-01 9.76632759e-02 7.88438320e-01 -4.70408589e-01 -4.57725883e-01 1.07963562e+00 4.10410047e-01 3.98189515e-01 7.22307980e-01 -9.94168043e-01 -1.05886078e+00 3.41453642e-01 -2.73374915e-01 -2.72389174e-01 3.84611748e-02 -9.86811146e-03 9.64976788e-01 -1.11016071e+00 9.10416782e-01 1.42528474e+00 7.12147176e-01 4.80813891e-01 -1.50640380e+00 -2.15051174e-01 -7.24282786e-02 6.50544688e-02 -1.21406960e+00 -6.27545655e-01 8.03596675e-01 -3.39853168e-01 5.72384536e-01 2.07571968e-01 4.72132057e-01 1.13317573e+00 3.82001758e-01 7.97812462e-01 1.31334329e+00 -7.18023241e-01 2.78625399e-01 -1.42668799e-01 -6.93052292e-01 9.21843052e-01 -1.47963271e-01 5.56347668e-01 -7.74908245e-01 2.30796695e-01 1.35860538e+00 -5.92431836e-02 -3.32864404e-01 -7.52215922e-01 -1.35375631e+00 8.29041541e-01 4.97944742e-01 8.82648006e-02 -2.39271820e-01 5.08708775e-01 1.39676049e-01 4.41095501e-01 8.38432252e-01 5.48329115e-01 -1.08952999e-01 2.68900972e-02 -1.07509291e+00 5.39479554e-01 4.48580265e-01 8.95751178e-01 1.01139152e+00 2.21167758e-01 -3.76253635e-01 8.89022827e-01 2.43505433e-01 7.25627124e-01 9.60565824e-03 -1.21928358e+00 1.17509991e-01 1.63955167e-01 6.38930872e-02 -1.14768696e+00 1.26107886e-01 -1.09228157e-01 -1.01564038e+00 4.00592446e-01 4.32912916e-01 3.55972141e-01 -1.00655138e+00 1.61451304e+00 3.13098997e-01 2.24668071e-01 -2.32746005e-01 8.87514532e-01 4.22381401e-01 3.46752971e-01 -3.13259304e-01 2.10817456e-01 7.00980246e-01 -9.17781830e-01 -6.84239388e-01 1.27399698e-01 2.77601838e-01 -9.92551029e-01 1.25546086e+00 4.72854137e-01 -1.25459695e+00 -4.65074807e-01 -9.79057014e-01 -2.74125695e-01 -2.42520571e-01 -1.33208662e-01 6.30211830e-01 3.25622708e-01 -1.10179365e+00 1.22297037e+00 -9.53641236e-01 -1.58528790e-01 5.42264163e-01 9.14203748e-02 -2.72160739e-01 -3.21996778e-01 -1.12131441e+00 9.13910151e-01 7.90140871e-03 3.82139772e-01 -9.88479912e-01 -8.70267391e-01 -8.01212788e-01 -1.08075731e-01 4.47291136e-01 -8.96900475e-01 1.11958587e+00 -9.81939614e-01 -1.92080462e+00 1.00292361e+00 -2.16775034e-02 -3.43910128e-01 8.19976807e-01 -2.31818557e-01 -4.73369099e-02 1.07025824e-01 9.28857997e-02 8.89924347e-01 1.13594520e+00 -1.36966455e+00 -4.06573296e-01 -1.37761816e-01 7.60273561e-02 9.85864401e-02 3.34082879e-02 -4.35546875e-01 -1.10714972e-01 -1.03391707e+00 1.24172747e-01 -8.77732933e-01 -3.13532382e-01 5.66530764e-01 -6.64155900e-01 1.20986581e-01 7.10485280e-01 -5.74912667e-01 8.06825638e-01 -2.00205398e+00 5.15966952e-01 4.01900768e-01 2.68057555e-01 -9.35626552e-02 -1.67769626e-01 2.52374798e-01 1.04274511e-01 3.83880399e-02 -5.67664862e-01 -3.30363631e-01 2.95960933e-01 1.49667680e-01 -4.36511278e-01 4.33534235e-01 4.10168886e-01 1.09310007e+00 -8.82461846e-01 -1.78410694e-01 3.81409645e-01 7.22294509e-01 -6.91308320e-01 -1.80166606e-02 -4.42614436e-01 7.28402615e-01 -4.30217862e-01 4.75633234e-01 5.66107273e-01 -9.31395665e-02 -1.34973213e-01 -3.14305872e-01 -1.29158452e-01 1.68378115e-01 -9.40908134e-01 2.01159477e+00 -7.81596959e-01 7.68594682e-01 1.88130159e-02 -1.06218827e+00 9.40468788e-01 8.53037462e-02 2.43618652e-01 -9.68591213e-01 -4.43591438e-02 5.22902310e-01 -3.49784404e-01 -1.09063357e-01 5.26129246e-01 -2.51775712e-01 1.67297527e-01 3.68927568e-01 -1.04787223e-01 -4.39460188e-01 -8.11552629e-02 -1.91138133e-01 8.32436740e-01 6.19529903e-01 -1.01525582e-01 -4.77162480e-01 2.60645330e-01 -1.93640068e-01 3.42099190e-01 7.86231279e-01 2.55934566e-01 1.01364410e+00 3.43999475e-01 -5.97006738e-01 -1.47734225e+00 -1.15071821e+00 -1.81037948e-01 7.82468796e-01 1.04123391e-01 9.64491144e-02 -8.58674109e-01 -4.67810541e-01 -1.67288128e-02 4.84976798e-01 -8.07876647e-01 -3.21090251e-01 -6.32053912e-01 -2.97230870e-01 3.75984758e-01 2.91499704e-01 6.38590693e-01 -1.28986824e+00 -4.15172577e-01 3.28182191e-01 -2.44906262e-01 -8.52305233e-01 -5.32274663e-01 1.15077518e-01 -8.55855048e-01 -7.68803179e-01 -9.48377311e-01 -6.77538872e-01 5.78555584e-01 -7.01094344e-02 1.20731461e+00 -7.58404052e-03 -2.14637041e-01 2.25999132e-01 -2.43571833e-01 -1.05333529e-01 -4.72781897e-01 2.94667389e-02 -2.22909391e-01 3.19713593e-01 -9.29272696e-02 -8.48178208e-01 -6.77595079e-01 3.79665107e-01 -1.11492980e+00 2.82598108e-01 6.53279603e-01 1.14226604e+00 1.04701507e+00 -8.88399780e-02 3.79741907e-01 -9.95859623e-01 4.26384896e-01 -9.41753611e-02 -4.86653894e-01 4.25693274e-01 -6.65080249e-01 2.97143251e-01 4.96059805e-01 -3.74300629e-01 -1.03868628e+00 -1.04220033e-01 -1.50475517e-01 -6.25167310e-01 -2.61212494e-02 4.66986805e-01 -1.32501602e-01 -4.68916893e-01 5.23282766e-01 3.43035102e-01 4.18369621e-02 -4.87703711e-01 4.44795996e-01 1.91876009e-01 4.89954710e-01 -8.27793181e-01 8.78293514e-01 8.11350346e-01 2.96087056e-01 -6.77671432e-01 -7.48984516e-01 1.93609834e-01 -7.07487643e-01 -1.32856652e-01 9.14084494e-01 -5.51663220e-01 -4.39985186e-01 6.59689844e-01 -1.03084946e+00 -9.03237522e-01 -5.92067838e-01 2.06441835e-01 -1.22371900e+00 3.21226746e-01 -4.96841878e-01 -5.20412862e-01 -1.95819259e-01 -1.16736782e+00 1.40941060e+00 -6.54847100e-02 -1.40874892e-01 -1.13705218e+00 -1.03033416e-01 1.67091012e-01 7.81172037e-01 5.18209696e-01 1.05160332e+00 1.59354702e-01 -8.93952370e-01 -1.01947049e-02 -2.07890704e-01 4.35268670e-01 6.44510314e-02 4.69250791e-02 -1.01926088e+00 -1.49052724e-01 -1.19853616e-01 -5.22319496e-01 1.04536402e+00 5.71205914e-01 1.41269052e+00 -2.24395037e-01 3.69497687e-02 9.28734481e-01 1.31930280e+00 8.60170946e-02 1.02289188e+00 3.29572916e-01 7.84777701e-01 5.53790629e-01 5.19919097e-01 2.52230674e-01 1.89755678e-01 8.68578017e-01 3.27797085e-01 -2.59475350e-01 -5.32431662e-01 -5.24838030e-01 8.13678056e-02 6.73781574e-01 -3.27535868e-01 -1.72784567e-01 -7.58183956e-01 2.27225378e-01 -1.79554737e+00 -7.94506252e-01 1.06470384e-01 2.28544211e+00 7.90110409e-01 6.87866881e-02 -5.19702971e-01 1.44251451e-01 6.86263382e-01 1.93882748e-01 -4.97939259e-01 -2.51114935e-01 -3.46227676e-01 6.01640522e-01 3.40370625e-01 7.05885231e-01 -1.10020983e+00 1.17023456e+00 5.79680252e+00 1.19422424e+00 -1.30371583e+00 2.94250753e-02 8.89456928e-01 1.55270949e-01 -5.67102849e-01 -3.05422898e-02 -3.71719450e-01 2.86043406e-01 4.49163288e-01 1.73276588e-01 7.44858503e-01 6.39581025e-01 2.41548344e-01 -4.45874147e-02 -1.25879586e+00 8.05516005e-01 -4.17624414e-02 -1.60219586e+00 3.19399983e-01 8.42259303e-02 9.98969555e-01 -3.82712513e-01 5.33721209e-01 -7.85933286e-02 2.69141644e-01 -1.25031531e+00 1.02693546e+00 9.91204917e-01 1.15888917e+00 -6.06602609e-01 3.72647285e-01 -6.40150756e-02 -8.74884784e-01 4.00490671e-01 -3.83116961e-01 2.37899840e-01 9.17361304e-02 7.74393499e-01 -1.85391977e-01 5.07735431e-01 7.12985218e-01 8.23005736e-01 -1.63905874e-01 7.24457324e-01 -1.10671163e-01 3.08581948e-01 -2.06394970e-01 1.62864342e-01 2.31655136e-01 -3.62520307e-01 4.72287804e-01 6.71756029e-01 4.19652373e-01 -3.77471209e-01 8.14181045e-02 1.34156418e+00 -8.15232098e-02 -8.54345188e-02 -6.97271287e-01 -5.28242402e-02 2.36863568e-01 9.43826735e-01 -7.11009204e-01 -1.35120407e-01 -4.78960164e-02 1.24631059e+00 2.70336866e-01 5.41345596e-01 -6.96431577e-01 -4.91128832e-01 3.90274793e-01 3.17039013e-01 4.59789366e-01 -5.50209731e-03 -4.95936781e-01 -1.01877344e+00 2.29841605e-01 -8.86685193e-01 -4.28225249e-01 -9.20959651e-01 -1.39804137e+00 5.57639897e-01 -2.13913843e-01 -1.16495764e+00 -1.51632717e-02 -5.43206692e-01 -5.14444292e-01 9.29972470e-01 -1.70695174e+00 -1.51127434e+00 -1.43121213e-01 6.53003097e-01 3.05083752e-01 -1.78834721e-02 8.66058290e-01 4.31204110e-01 -1.96526214e-01 6.05689466e-01 2.83123940e-01 -5.15774377e-02 7.04836845e-01 -1.16776979e+00 4.96411175e-01 6.35708392e-01 -8.56057648e-03 5.63090086e-01 5.44124603e-01 -5.89094102e-01 -1.26214361e+00 -1.20815647e+00 8.67121994e-01 -2.70633966e-01 5.58478415e-01 -5.20702541e-01 -8.31219733e-01 5.23479402e-01 -7.72561654e-02 1.82452872e-01 5.04007898e-02 -1.93926200e-01 -2.79758513e-01 -7.75076747e-02 -1.08697093e+00 8.27015579e-01 1.22011566e+00 -7.31653512e-01 -6.69793040e-03 3.58674020e-01 5.01085937e-01 -4.78036612e-01 -1.03891873e+00 5.75236976e-01 7.27320850e-01 -1.07311308e+00 1.01670110e+00 -5.00627577e-01 1.03169096e+00 4.37133340e-03 -4.18226808e-01 -1.44223976e+00 -2.80439109e-01 -8.17402661e-01 7.09924996e-02 9.89951849e-01 4.25283045e-01 -4.79448795e-01 8.20182145e-01 2.21329287e-01 -3.77172977e-01 -1.01092613e+00 -9.27354813e-01 -7.21762002e-01 4.60911274e-01 -1.84378102e-01 5.37081301e-01 9.00620282e-01 -6.06744587e-01 -1.26425281e-01 -6.41095459e-01 -2.24057600e-01 7.99431324e-01 1.44761965e-01 5.36917746e-01 -1.11611569e+00 -3.06582093e-01 -6.76390350e-01 -2.37798795e-01 -1.12444615e+00 1.26024291e-01 -8.32627237e-01 2.74754375e-01 -1.40147173e+00 -5.98999346e-03 -8.45377743e-01 -1.48210391e-01 1.51791543e-01 1.16407968e-01 4.06239033e-01 -1.54796764e-02 1.39738202e-01 -3.73894542e-01 9.79715884e-01 1.94280088e+00 -2.89968938e-01 6.41516969e-02 -1.42002448e-01 -3.76077086e-01 6.48939133e-01 6.85592949e-01 -4.60040808e-01 -3.34204763e-01 -5.49333453e-01 4.93076235e-01 1.03508018e-01 6.64922893e-01 -8.97859514e-01 7.56839886e-02 -4.25736189e-01 2.73210138e-01 -2.08345175e-01 2.23369837e-01 -5.89892566e-01 1.08215861e-01 1.66491702e-01 -6.20092452e-01 -2.15084195e-01 -1.94342688e-01 7.01092899e-01 -3.80027324e-01 1.26204535e-01 9.72416580e-01 -1.78021982e-01 -4.99066591e-01 6.15619957e-01 -1.53445080e-01 1.01918221e-01 4.69561249e-01 -4.98192281e-01 -5.37937991e-02 -3.17977756e-01 -8.20926726e-01 -3.23412508e-01 5.82737982e-01 2.15831265e-01 8.74648869e-01 -1.69278431e+00 -6.87565804e-01 2.70564109e-01 3.67181450e-02 3.35625887e-01 3.95936280e-01 1.03241217e+00 -7.55979359e-01 2.92426318e-01 -3.66084963e-01 -5.97781658e-01 -7.62664616e-01 2.20310017e-01 5.77606916e-01 -4.48392123e-01 -8.20765197e-01 9.86153424e-01 4.37837452e-01 -6.73005164e-01 2.21955657e-01 -2.98498720e-01 3.48209530e-01 -4.56485510e-01 2.89503425e-01 2.72709131e-01 3.01481098e-01 -3.68333697e-01 1.03755496e-01 7.41930485e-01 1.10216513e-02 -2.46341988e-01 1.36877954e+00 -1.33093223e-01 -2.72973925e-01 4.46166962e-01 1.18454731e+00 -2.43499324e-01 -1.68686593e+00 -3.92859608e-01 -1.56240746e-01 -5.48448265e-01 3.06226164e-01 -7.06930459e-01 -1.26774418e+00 1.10770357e+00 5.74204028e-01 -1.03174649e-01 1.03520763e+00 -3.57152730e-01 9.85386670e-01 4.79957849e-01 4.03193027e-01 -1.21685421e+00 1.88941106e-01 5.86174250e-01 1.13794458e+00 -1.12276232e+00 -2.19704911e-01 -4.46378022e-01 -3.69776756e-01 1.16035521e+00 3.41875941e-01 -4.23240185e-01 7.44515896e-01 2.25493670e-01 -1.94878772e-01 -2.64533192e-01 -5.89891911e-01 4.31078635e-02 3.44235420e-01 6.54917121e-01 3.94473881e-01 5.18302023e-02 1.05189592e-01 -1.02697439e-01 -2.09300995e-01 1.32755280e-01 2.61110038e-01 8.61099541e-01 -2.13622287e-01 -9.36697960e-01 -7.77900666e-02 4.35942560e-01 -3.59686524e-01 -3.49366814e-01 -3.20470542e-01 5.98837554e-01 1.62113607e-01 4.87253636e-01 1.73759446e-01 -4.00504261e-01 3.55402976e-01 -1.40005842e-01 8.94411385e-01 -4.37155008e-01 -1.81160375e-01 -8.47307313e-03 -2.07656085e-01 -6.77489102e-01 -5.46241522e-01 -4.29388046e-01 -7.00491130e-01 -5.36369860e-01 -1.75889492e-01 -2.57873386e-01 4.82739449e-01 8.31358254e-01 1.59771577e-01 3.94581348e-01 6.07833862e-01 -1.23355973e+00 -5.29745877e-01 -7.92537212e-01 -6.86398983e-01 5.52168787e-01 4.14375663e-01 -9.46168005e-01 -2.69984007e-01 2.24404827e-01]

[11.488895416259766, -0.5958898663520813]

ed5e5612-fa72-4dec-a3e7-82bb33e07fb4

action-genome-actions-as-compositions-of

http://openaccess.thecvf.com/content_CVPR_2020/html/Ji_Action_Genome_Actions_As_Compositions_of_Spatio-Temporal_Scene_Graphs_CVPR_2020_paper.html

http://openaccess.thecvf.com/content_CVPR_2020/papers/Ji_Action_Genome_Actions_As_Compositions_of_Spatio-Temporal_Scene_Graphs_CVPR_2020_paper.pdf

Action Genome: Actions As Compositions of Spatio-Temporal Scene Graphs

Action recognition has typically treated actions and activities as monolithic events that occur in videos. However, there is evidence from Cognitive Science and Neuroscience that people actively encode activities into consistent hierarchical part structures. However, in Computer Vision, few explorations on representations that encode event partonomies have been made. Inspired by evidence that the prototypical unit of an event is an action-object interaction, we introduce Action Genome, a representation that decomposes actions into spatio-temporal scene graphs. Action Genome captures changes between objects and their pairwise relationships while an action occurs. It contains 10K videos with 0.4M objects and 1.7M visual relationships annotated. With Action Genome, we extend an existing action recognition model by incorporating scene graphs as spatio-temporal feature banks to achieve better performance on the Charades dataset. Next, by decomposing and learning the temporal changes in visual relationships that result in an action, we demonstrate the utility of a hierarchical event decomposition by enabling few-shot action recognition, achieving 42.7% mAP using as few as 10 examples. Finally, we benchmark existing scene graph models on the new task of spatio-temporal scene graph prediction.

[' Juan Carlos Niebles', ' Li Fei-Fei', ' Ranjay Krishna', 'Jingwei Ji']

2020-06-01

cvpr-2020-6

['few-shot-action-recognition']

['computer-vision']

[ 4.48740214e-01 -6.15686625e-02 -2.22366378e-01 -3.98051798e-01 -1.02764413e-01 -6.60490990e-01 9.69384611e-01 3.40663493e-01 -2.23541915e-01 3.83133560e-01 8.53706241e-01 1.08299397e-01 -2.32595906e-01 -6.16978049e-01 -9.24408257e-01 -5.86477578e-01 -6.26556098e-01 1.98359415e-01 6.21074021e-01 1.88497841e-01 1.83789134e-01 4.73606080e-01 -1.77392030e+00 9.97812986e-01 5.40326759e-02 8.20570290e-01 3.20511088e-02 8.48215640e-01 1.33865640e-01 1.50559115e+00 -4.77885514e-01 -1.70470346e-02 2.04260379e-01 -7.60253489e-01 -9.28266644e-01 5.79017878e-01 6.91564798e-01 -3.46456140e-01 -9.19227242e-01 5.82632601e-01 -5.02159446e-02 4.77944165e-01 6.29721582e-01 -1.64017749e+00 -5.62133014e-01 5.31483948e-01 -4.41270947e-01 7.62273312e-01 7.57432520e-01 2.28852198e-01 1.25708318e+00 -6.76605165e-01 9.94546652e-01 1.35571158e+00 3.45942169e-01 7.59795010e-02 -1.22845542e+00 -3.08411539e-01 3.01409990e-01 9.61226940e-01 -1.08852386e+00 -4.50262845e-01 5.15411973e-01 -6.51761651e-01 1.84879494e+00 2.72005379e-01 1.27933419e+00 1.04783189e+00 6.05595529e-01 9.63797092e-01 6.02569163e-01 -1.23728849e-01 2.63824105e-01 -8.34360957e-01 1.05925322e-01 5.89463055e-01 -4.51998487e-02 -2.20071539e-01 -9.68922257e-01 4.59251702e-02 7.96804249e-01 4.66762960e-01 -7.15719387e-02 -6.20730877e-01 -1.48610175e+00 4.90563840e-01 6.01579547e-01 2.35078067e-01 -4.71028298e-01 6.72012031e-01 6.25830412e-01 1.60771742e-01 1.45499915e-01 3.53161275e-01 -1.78125501e-01 -4.18690741e-01 -5.72258174e-01 1.92210525e-01 5.18148720e-01 8.27920258e-01 5.99804640e-01 -8.89379010e-02 -2.58921474e-01 5.23308754e-01 1.51961492e-02 -5.20590656e-02 3.35926235e-01 -1.19318032e+00 2.26277724e-01 1.10697556e+00 -1.70672372e-01 -1.18674767e+00 -4.23254639e-01 2.85195798e-01 -5.14605701e-01 2.50705779e-02 3.05905789e-01 4.93086189e-01 -8.88622999e-01 1.48791063e+00 2.74588794e-01 7.81734467e-01 -1.10869363e-01 5.79703152e-01 7.72160232e-01 6.72526479e-01 2.35392913e-01 -1.88323259e-01 1.30688596e+00 -8.87883127e-01 -5.79723358e-01 -2.33477771e-01 7.22160518e-01 -2.87286162e-01 7.43438184e-01 3.40611607e-01 -9.41935897e-01 -5.82145631e-01 -7.97189057e-01 -2.27430359e-01 -4.84941751e-01 -2.38193527e-01 1.03654039e+00 -1.20958567e-01 -8.99337709e-01 4.85727340e-01 -1.12665319e+00 -9.36797440e-01 4.93301302e-01 1.10618904e-01 -6.89904094e-01 -2.26904787e-02 -8.83787096e-01 7.15700448e-01 7.23999441e-01 -2.83057362e-01 -1.25913513e+00 -5.12714922e-01 -1.06218195e+00 1.07766345e-01 5.15615880e-01 -3.93353462e-01 1.02439165e+00 -9.07800317e-01 -8.55258226e-01 7.31304169e-01 -1.81474805e-01 -7.48271883e-01 -1.40529111e-01 -9.42388028e-02 -4.67450827e-01 5.48146963e-01 5.72394095e-02 8.59311581e-01 6.20631635e-01 -7.06001043e-01 -8.45540702e-01 -3.53415012e-01 4.18374032e-01 4.26388532e-01 -9.10448208e-02 2.86998600e-01 -6.54762626e-01 -5.52259445e-01 1.89820156e-01 -9.06296372e-01 -1.06381074e-01 7.99333900e-02 -1.94009200e-01 -3.12919378e-01 8.50648820e-01 -4.60060447e-01 1.04186654e+00 -2.37950516e+00 4.61960047e-01 -1.76398516e-01 2.99505860e-01 -2.59484380e-01 -1.62283674e-01 7.90087581e-01 -4.10944104e-01 -1.15592808e-01 -5.51573969e-02 5.72170736e-03 -1.15878128e-01 5.99116623e-01 -3.77527654e-01 5.57420969e-01 2.47569397e-01 1.13645959e+00 -1.10708892e+00 -5.47564447e-01 5.50784647e-01 4.62759048e-01 -6.68525577e-01 1.35178603e-02 -2.21082032e-01 2.71841586e-01 -3.07717323e-01 6.28485084e-01 -1.14235401e-01 -5.93191266e-01 4.30848509e-01 -2.58668959e-01 -7.06931874e-02 3.78311753e-01 -1.05726039e+00 1.79578006e+00 3.45411301e-01 9.37975049e-01 -4.97707844e-01 -1.16916394e+00 3.82681519e-01 2.86910743e-01 1.22896457e+00 -6.87931299e-01 -1.47284180e-01 -4.77428466e-01 2.41202079e-02 -4.59056646e-01 4.61498111e-01 2.95182347e-01 -1.86275691e-01 2.91217536e-01 4.07353163e-01 4.88617606e-02 6.81997359e-01 6.50452316e-01 1.85595179e+00 3.81301463e-01 8.09958994e-01 1.33072779e-01 1.95186988e-01 2.92777747e-01 3.99565786e-01 6.54723227e-01 -3.68208200e-01 4.32380617e-01 7.32514918e-01 -6.56865358e-01 -7.86794960e-01 -1.21140456e+00 1.12451091e-01 1.39853966e+00 1.44149020e-01 -1.01178277e+00 -3.07038218e-01 -5.21771491e-01 -1.78473908e-02 5.16621232e-01 -8.78479481e-01 -2.14704946e-01 -5.86255908e-01 -5.53130448e-01 4.21673775e-01 8.68746102e-01 5.22053182e-01 -1.31422746e+00 -9.15794134e-01 2.30255410e-01 -1.92642644e-01 -1.26793408e+00 -3.24753910e-01 3.06382596e-01 -6.90814197e-01 -1.44224131e+00 -8.11160803e-02 -6.79065108e-01 4.55569297e-01 5.21100938e-01 1.22788179e+00 -1.20926268e-01 -8.57175350e-01 1.06874561e+00 -6.16614223e-01 -1.08093657e-01 -6.37061968e-02 -7.09997654e-01 2.38495562e-02 1.21439673e-01 4.40843910e-01 -7.77847767e-01 -5.86332023e-01 2.97085881e-01 -7.58298159e-01 3.07735056e-01 3.75991374e-01 4.20615613e-01 8.09660017e-01 1.45928577e-01 -7.85695668e-03 -4.38908130e-01 -1.88135758e-01 -5.95112801e-01 -3.15334082e-01 3.24845225e-01 4.81533073e-02 -2.85431683e-01 3.35941762e-01 -4.19344664e-01 -9.42062557e-01 5.60594440e-01 5.74497461e-01 -6.05502307e-01 -3.95836681e-01 3.95034462e-01 7.93154463e-02 2.21771106e-01 7.97795355e-01 5.16674936e-01 -3.28361273e-01 -7.74667859e-02 6.11007690e-01 -5.06696813e-02 6.47563219e-01 -2.36812487e-01 2.82561600e-01 8.20815384e-01 1.60162687e-01 -1.05066860e+00 -7.63573050e-01 -8.36341381e-01 -1.01195323e+00 -5.89751542e-01 1.34889340e+00 -1.01906943e+00 -8.68342042e-01 3.27829450e-01 -9.24828351e-01 -4.32418376e-01 -5.34971356e-01 5.42610884e-01 -1.00073647e+00 5.42959213e-01 -6.77442431e-01 -4.42946792e-01 3.95182997e-01 -6.62415326e-01 1.27504909e+00 -1.33932427e-01 -5.04881740e-01 -7.17087388e-01 2.23889261e-01 2.69552529e-01 -3.18660021e-01 4.26266134e-01 6.97676361e-01 -4.34652954e-01 -8.08369100e-01 -1.02502247e-02 -2.01548934e-02 -1.58569887e-01 3.72175366e-01 1.78743050e-01 -4.37666476e-01 -1.16201788e-01 -2.33100623e-01 -3.86510313e-01 9.49611366e-01 4.48185265e-01 1.29046285e+00 -1.98662132e-01 -5.01504362e-01 5.40307820e-01 1.10548818e+00 4.40308154e-01 8.47725809e-01 1.37454808e-01 8.98297966e-01 6.73425496e-01 5.15802324e-01 6.17759287e-01 3.54250968e-01 6.79152668e-01 5.49609840e-01 3.44727427e-01 -3.58434945e-01 -3.16649526e-01 6.48556054e-01 4.98707414e-01 -3.76015186e-01 -2.86342770e-01 -9.50720251e-01 4.80966300e-01 -2.21899772e+00 -1.60174668e+00 -2.38468975e-01 1.78263128e+00 5.03516912e-01 1.49188519e-01 2.28726700e-01 -5.07632233e-02 4.57211256e-01 5.69349349e-01 -6.75429463e-01 6.84156269e-02 -1.10187039e-01 -1.14426970e-01 3.33506435e-01 7.92924687e-02 -1.32360029e+00 1.09264100e+00 7.09695148e+00 3.16519976e-01 -5.47384381e-01 -2.53659844e-01 2.91665584e-01 -5.08435726e-01 1.70657054e-01 5.88935316e-02 -4.98428285e-01 1.89646199e-01 8.06844413e-01 -2.67606288e-01 3.93121064e-01 8.08308363e-01 2.16042861e-01 -4.18958664e-01 -1.58455408e+00 1.13597381e+00 2.90161848e-01 -1.38488019e+00 3.12641919e-01 1.28065385e-02 4.75389689e-01 6.04396202e-02 -4.72016245e-01 3.92695606e-01 6.02139592e-01 -9.92230475e-01 5.95282495e-01 6.56807423e-01 3.05219591e-01 -2.94895142e-01 -3.22876014e-02 1.57954723e-01 -1.77733719e+00 -1.69366241e-01 -3.73685747e-01 -4.44336921e-01 2.75791883e-01 7.39662573e-02 -8.64067912e-01 3.69662225e-01 9.82704282e-01 1.72783875e+00 -8.71709228e-01 9.92343962e-01 -2.17491090e-01 6.55034304e-01 -9.85311270e-02 2.51045823e-01 2.27817267e-01 -1.97919697e-01 3.92325461e-01 1.18473792e+00 3.30894627e-02 5.48337400e-01 4.10863072e-01 4.99916315e-01 1.86606765e-01 -2.83742815e-01 -1.05592322e+00 -4.51445252e-01 1.82227753e-02 1.02937615e+00 -1.03118324e+00 -5.98159134e-01 -7.27332652e-01 1.24607801e+00 3.52371931e-01 3.49306613e-01 -1.08016253e+00 2.29029313e-01 8.97599757e-01 -8.62897038e-02 4.66095895e-01 -5.76893270e-01 1.74490497e-01 -1.15396869e+00 -1.76453680e-01 -7.63258517e-01 7.69319415e-01 -1.18972301e+00 -9.92868781e-01 2.77902167e-02 3.13082576e-01 -1.28987277e+00 -2.99393296e-01 -6.14220738e-01 -3.89311552e-01 1.75582189e-02 -6.25412881e-01 -1.09721255e+00 -4.39445585e-01 9.32910860e-01 9.14098084e-01 6.96834698e-02 8.06156933e-01 8.76136422e-02 -4.58099961e-01 -1.78607717e-01 -3.00996751e-01 2.12945849e-01 3.72639477e-01 -1.26468980e+00 5.02082348e-01 9.19343174e-01 9.06126738e-01 3.01392078e-01 4.75076765e-01 -9.57102776e-01 -1.70603454e+00 -1.17137444e+00 7.03073680e-01 -9.01899695e-01 1.04199767e+00 -5.23560762e-01 -8.66120040e-01 1.31488454e+00 2.07206905e-01 2.51161993e-01 6.88592076e-01 8.78718346e-02 -5.34695745e-01 1.48951322e-01 -5.05614400e-01 8.19818974e-01 1.84196627e+00 -7.68766403e-01 -9.26063299e-01 5.47955930e-01 4.99544829e-01 -2.07522705e-01 -8.72097552e-01 2.72694945e-01 4.71945494e-01 -9.88618374e-01 1.20919502e+00 -1.07534623e+00 4.90032941e-01 -3.64620954e-01 -3.88526112e-01 -1.00050902e+00 -8.07044804e-01 -2.49145314e-01 -5.28664052e-01 6.09182060e-01 -1.13177784e-01 -7.45156258e-02 7.75480390e-01 3.73983860e-01 -2.85973132e-01 -3.81102979e-01 -8.02481115e-01 -8.60501230e-01 -6.38379157e-01 -6.36832893e-01 1.48803443e-01 9.99608636e-01 2.71434724e-01 3.59580785e-01 -3.96045446e-01 5.54787852e-02 3.13614428e-01 8.52258578e-02 8.47105026e-01 -9.58518803e-01 -4.57545877e-01 -3.52052063e-01 -1.05929661e+00 -1.03307056e+00 1.28585801e-01 -9.48521614e-01 1.21052377e-01 -1.80128860e+00 6.10582590e-01 5.62526703e-01 -5.00042617e-01 9.28726912e-01 9.81725603e-02 1.90515533e-01 2.02020764e-01 2.57968485e-01 -1.20929873e+00 4.93751466e-01 1.02385688e+00 -2.17611000e-01 -1.97358914e-02 -7.02526033e-01 -1.38133779e-01 8.64730895e-01 4.58460063e-01 -1.91564858e-01 -7.14884758e-01 -2.59679943e-01 2.18033820e-01 7.71683156e-02 6.64177120e-01 -1.21405220e+00 3.51263434e-01 -4.89443928e-01 4.92314607e-01 -6.74527109e-01 7.63998270e-01 -7.87030160e-01 5.36777377e-01 2.96165615e-01 -3.92206281e-01 5.86315058e-02 1.30122870e-01 1.01681066e+00 -3.11793894e-01 4.52735364e-01 3.84015709e-01 -2.48511001e-01 -1.54037774e+00 3.81846458e-01 -7.20439494e-01 -4.49616313e-02 1.51558244e+00 -4.94773954e-01 -4.01533842e-01 -3.73673588e-01 -1.08552241e+00 1.15466222e-01 3.61709923e-01 6.01804435e-01 7.10670948e-01 -1.38734746e+00 -3.86600137e-01 -3.68361101e-02 4.09262985e-01 -3.96665961e-01 3.59578103e-01 8.11904252e-01 -3.87678057e-01 3.33011478e-01 -6.43317938e-01 -8.02720666e-01 -1.61162126e+00 7.02943325e-01 1.15444876e-01 2.37425864e-02 -1.04034591e+00 8.06159735e-01 5.70462704e-01 1.48588657e-01 2.58512437e-01 -5.25346279e-01 -3.39288682e-01 2.94366241e-01 6.01970613e-01 4.10957664e-01 -3.51305336e-01 -8.73479962e-01 -6.59882486e-01 3.45326275e-01 1.26434445e-01 8.53012875e-02 1.41158700e+00 1.90409780e-01 -2.84468591e-01 8.69742513e-01 9.81750309e-01 -5.57911694e-01 -1.66213441e+00 -2.99416389e-02 1.55823931e-01 -4.73418742e-01 -4.20390189e-01 -4.90472823e-01 -8.06872308e-01 6.04063869e-01 2.34614670e-01 4.15692627e-01 1.20658362e+00 3.81191641e-01 3.83677542e-01 7.21633315e-01 4.43047971e-01 -8.77164781e-01 7.79197454e-01 6.91849053e-01 1.08459568e+00 -1.07330537e+00 2.86261082e-01 -4.73525524e-01 -8.21847796e-01 8.62678647e-01 5.88663697e-01 -2.65132159e-01 5.95043421e-01 1.45908818e-02 -6.05058610e-01 -6.89124107e-01 -1.10542345e+00 -4.00301784e-01 4.96692955e-01 3.25549573e-01 2.74981439e-01 1.89184934e-01 -4.05670144e-02 2.96962317e-02 8.53929743e-02 -7.16080070e-02 4.60053444e-01 1.13419592e+00 -4.97913927e-01 -6.46718979e-01 9.33292508e-03 6.31778717e-01 -1.61746204e-01 -5.75426640e-03 -8.29925597e-01 8.61493766e-01 1.32122651e-01 8.29609871e-01 5.34409523e-01 -4.99231189e-01 3.08628857e-01 2.66155183e-01 7.56478608e-01 -1.02007282e+00 -4.16707963e-01 -7.75064155e-02 2.28760824e-01 -1.38449156e+00 -8.33972216e-01 -1.13082922e+00 -1.57365930e+00 -1.98970530e-02 4.65984970e-01 -3.87934357e-01 1.10419206e-01 8.79752219e-01 5.10478318e-01 7.56456017e-01 2.20964029e-01 -7.50486851e-01 1.98736280e-01 -7.86059260e-01 -6.87002480e-01 8.69034946e-01 2.11567525e-03 -8.04095924e-01 -3.34925532e-01 6.66057289e-01]

[8.54984188079834, 0.7400519847869873]

19bd2b95-3a40-4480-8dc1-0eb059421ba7

deep-reinforcement-learning-for-1

2209.05559

https://arxiv.org/abs/2209.05559v6

https://arxiv.org/pdf/2209.05559v6.pdf

Deep Reinforcement Learning for Cryptocurrency Trading: Practical Approach to Address Backtest Overfitting

Designing profitable and reliable trading strategies is challenging in the highly volatile cryptocurrency market. Existing works applied deep reinforcement learning methods and optimistically reported increased profits in backtesting, which may suffer from the false positive issue due to overfitting. In this paper, we propose a practical approach to address backtest overfitting for cryptocurrency trading using deep reinforcement learning. First, we formulate the detection of backtest overfitting as a hypothesis test. Then, we train the DRL agents, estimate the probability of overfitting, and reject the overfitted agents, increasing the chance of good trading performance. Finally, on 10 cryptocurrencies over a testing period from 05/01/2022 to 06/27/2022 (during which the crypto market crashed two times), we show that the less overfitted deep reinforcement learning agents have a higher return than that of more overfitted agents, an equal weight strategy, and the S&P DBM Index (market benchmark), offering confidence in possible deployment to a real market.

['Berend Jelmer Dirk Gort', 'Christina Dan Wang', 'Shuaiyu Chen', 'Jiechao Gao', 'Xinghang Sun', 'Xiao-Yang Liu']

2022-09-12

['algorithmic-trading']

['time-series']

[-7.87563145e-01 -1.17713720e-01 8.41073543e-02 5.01916185e-02 -6.74977183e-01 -7.77587891e-01 4.85175610e-01 -1.50973886e-01 -5.12301147e-01 1.09496319e+00 -4.76013511e-01 -7.43205190e-01 -5.47642484e-02 -1.04835963e+00 -8.04527044e-01 -9.24725711e-01 -5.71351051e-01 9.72853541e-01 1.11958366e-02 -3.10164988e-01 5.62393904e-01 1.03314228e-01 -1.02450383e+00 3.20791267e-02 6.28641725e-01 1.38686299e+00 -2.61763781e-01 5.33975244e-01 4.92465734e-01 8.95600140e-01 -1.19938838e+00 -6.20046794e-01 7.66711593e-01 -4.22103703e-01 -3.37392837e-01 -3.12097520e-01 -1.89914927e-01 -8.50276887e-01 -1.31337389e-01 1.32444310e+00 4.25971150e-01 -5.31078279e-01 2.25690782e-01 -1.29193783e+00 -1.13053992e-01 1.06301248e+00 -5.63045859e-01 3.83084416e-01 -3.14830035e-01 4.12745476e-01 1.42781079e+00 -5.57900608e-01 3.82773936e-01 8.12270284e-01 4.59745288e-01 1.03397571e-01 -9.28074062e-01 -1.02417445e+00 -1.69742286e-01 1.17615566e-01 -8.26367140e-01 8.54518116e-02 5.91655374e-01 -2.77414680e-01 9.23958719e-01 1.02870747e-01 1.27178490e+00 8.35170507e-01 7.74750054e-01 8.39495301e-01 1.46819007e+00 -6.71917126e-02 6.59199715e-01 8.04075599e-02 -3.31665516e-01 1.01780899e-01 6.43871069e-01 1.07612407e+00 -3.03038985e-01 -4.74145591e-01 6.33618832e-01 -1.71769187e-01 1.97528526e-01 -2.60833532e-01 -8.22842121e-01 1.48057699e+00 3.35650861e-01 1.48441613e-01 -7.42600858e-01 1.64611384e-01 3.12495649e-01 9.87386227e-01 2.98878789e-01 1.06676161e+00 -7.59128451e-01 -3.35818410e-01 -8.82246256e-01 6.37444735e-01 1.02280712e+00 3.83667320e-01 3.89003426e-01 5.73451102e-01 1.36866629e-01 1.74860716e-01 6.40014485e-02 5.85003376e-01 7.69893765e-01 -9.44839239e-01 2.12101817e-01 3.18982661e-01 2.39153117e-01 -4.29838121e-01 -2.03663856e-01 -9.52245593e-01 -4.46062237e-01 6.77792609e-01 2.42162764e-01 -6.62534237e-01 -4.41132724e-01 1.35788476e+00 -1.44356877e-01 -1.47412764e-02 3.92314941e-01 7.60056615e-01 -3.13556552e-01 6.08743250e-01 -5.88032007e-01 -5.71505129e-01 9.50748205e-01 -8.10735881e-01 -3.29872072e-01 -2.55927831e-01 5.20268559e-01 -6.84165061e-01 6.40244782e-01 9.79147971e-01 -1.06200898e+00 -6.85623363e-02 -1.34141374e+00 1.26027799e+00 -1.29036739e-01 -5.29179871e-01 8.43143106e-01 6.34312212e-01 -7.50248313e-01 6.97942793e-01 -6.71264529e-01 4.47299033e-01 2.31900036e-01 5.45756340e-01 3.07604820e-01 5.33033907e-01 -1.41082549e+00 8.10966551e-01 5.75094640e-01 3.32717448e-02 -1.31761801e+00 -3.20330739e-01 -1.82738423e-01 3.23449433e-01 6.84989512e-01 -3.04378550e-02 1.52421248e+00 -1.15094101e+00 -1.55472779e+00 3.58142197e-01 1.06488538e+00 -1.27571988e+00 9.57864821e-01 -1.97227180e-01 -4.18254107e-01 -2.94296712e-01 -1.52807355e-01 1.48140192e-01 6.51334703e-01 -9.00743663e-01 -9.78577435e-01 -2.77087390e-01 -6.01190329e-02 5.64880036e-02 8.18820857e-03 -2.60243922e-01 4.92011666e-01 -7.95724869e-01 -1.61088005e-01 -9.85468745e-01 -1.84373692e-01 -1.02744198e+00 -1.35008797e-01 -1.06136687e-01 3.74022037e-01 -4.36046481e-01 8.21411550e-01 -1.94460368e+00 -4.20718968e-01 5.31180441e-01 8.64243656e-02 1.38124794e-01 6.10470623e-02 3.45198125e-01 6.02517929e-03 2.99047381e-02 1.56737655e-01 5.37444711e-01 2.61579752e-01 1.39116555e-01 -6.10698342e-01 3.82280260e-01 8.93686041e-02 9.34528172e-01 -6.63976014e-01 1.28129423e-01 -2.22046837e-01 -3.94792140e-01 -6.25169694e-01 4.24050212e-01 -5.04690409e-01 -2.34862059e-01 -4.13702726e-01 7.87239194e-01 6.15680933e-01 -1.07383274e-01 4.02609438e-01 3.64463449e-01 -1.96009859e-01 3.22267562e-01 -8.77894998e-01 4.60189193e-01 -1.01569660e-01 5.07571518e-01 -2.24235922e-01 -9.92127776e-01 1.20473993e+00 3.51788588e-02 4.72502261e-01 -1.32704544e+00 1.93271756e-01 8.21219981e-01 5.51768839e-01 3.77525762e-03 5.64879835e-01 -5.75207412e-01 -2.18637779e-01 8.02996695e-01 -1.14923559e-01 -2.80816793e-01 1.26100734e-01 -2.14219570e-01 9.45362628e-01 -1.80607773e-02 -1.97692439e-02 -2.79171199e-01 -8.79571512e-02 8.69396105e-02 9.89586294e-01 9.41932201e-01 -3.42534751e-01 4.57555056e-02 1.14626920e+00 -8.25437546e-01 -9.96605992e-01 -1.07633018e+00 4.45463024e-02 6.22483015e-01 -1.42048955e-01 1.57831177e-01 -4.64574844e-01 -8.22964191e-01 4.53327596e-01 1.01821673e+00 -5.11972070e-01 -2.98337907e-01 -4.79878962e-01 -1.16847944e+00 4.90112364e-01 2.14723691e-01 5.59055567e-01 -1.29113686e+00 -1.16914058e+00 4.52673733e-01 3.09173226e-01 -2.46368378e-01 -2.32262284e-01 7.13003337e-01 -7.21526206e-01 -1.21692431e+00 -7.92154312e-01 -3.87869120e-01 2.34647214e-01 -2.51275599e-01 1.34901130e+00 1.05868109e-01 1.44758597e-01 -3.97990972e-01 -1.69185072e-01 -4.75739717e-01 -5.52460790e-01 -1.41920209e-01 2.86695927e-01 -4.16012377e-01 5.46747267e-01 -5.53475842e-02 -6.44085526e-01 3.99947912e-01 -5.53143442e-01 -5.77996135e-01 7.36612082e-01 1.27310991e+00 5.01472473e-01 5.24267554e-01 9.73257363e-01 -6.83868527e-01 7.06231356e-01 -4.05832857e-01 -1.67396677e+00 2.97903210e-01 -1.25330710e+00 2.72098720e-01 5.12968123e-01 -4.69932497e-01 -6.62699163e-01 -5.43898702e-01 1.98636681e-01 -4.45581794e-01 7.87921011e-01 5.70709169e-01 3.10081542e-01 4.23830785e-02 2.25141585e-01 3.84222180e-01 2.44740129e-01 -2.56925732e-01 -3.00802052e-01 4.30569857e-01 2.65400380e-01 -4.64962184e-01 7.72501647e-01 -1.45429417e-01 -3.36745381e-01 -9.60096493e-02 -2.85266310e-01 4.20565903e-01 3.06623280e-01 -2.34300986e-01 1.78084210e-01 -7.81402111e-01 -1.15207005e+00 7.71931648e-01 -4.16789144e-01 -5.62468767e-01 -4.72998589e-01 7.70703495e-01 -6.08229697e-01 6.39155805e-02 -6.78577244e-01 -9.50973749e-01 -2.95976907e-01 -1.30397606e+00 2.26081431e-01 3.03353548e-01 2.94539519e-02 -6.93712473e-01 5.32682180e-01 3.42688352e-01 4.55164135e-01 1.06161766e-01 8.71107638e-01 -1.41686392e+00 -6.27122104e-01 -3.28457326e-01 1.40707791e-01 6.01926565e-01 -1.97519705e-01 9.65874363e-03 -5.52554190e-01 -7.90104508e-01 5.15739918e-01 -6.58036411e-01 6.09204650e-01 4.83287930e-01 4.75101233e-01 -6.24384463e-01 2.58849174e-01 4.30013925e-01 1.15412414e+00 8.87531817e-01 6.03437006e-01 1.06952143e+00 -3.08357358e-01 4.83704537e-01 9.92127001e-01 6.99691892e-01 1.46346003e-01 3.50019872e-01 7.45955944e-01 1.07156120e-01 7.08331585e-01 -3.00962687e-01 8.45936418e-01 6.13653719e-01 2.23475575e-01 -8.70740488e-02 -6.23330474e-01 2.89397240e-01 -1.58760810e+00 -7.95600474e-01 4.50871378e-01 2.44680023e+00 8.43479097e-01 9.24231827e-01 3.99646014e-01 -1.22987278e-01 5.42816043e-01 -9.46409851e-02 -9.26373243e-01 -5.34665704e-01 -5.36499083e-01 1.59911871e-01 8.43549550e-01 3.05893928e-01 -6.78329587e-01 7.69077599e-01 6.29685974e+00 7.06290781e-01 -1.27279985e+00 -2.72672683e-01 1.22043216e+00 -2.69094646e-01 -5.90718567e-01 3.16951156e-01 -5.39956212e-01 6.89627290e-01 1.11774135e+00 -4.08052891e-01 5.79955816e-01 8.26171756e-01 -1.52475059e-01 -1.96220472e-01 -8.14764559e-01 4.27459717e-01 -2.43723184e-01 -1.46778011e+00 -3.70267242e-01 6.21264875e-01 7.35380888e-01 2.87881255e-01 2.52244383e-01 8.76180649e-01 6.49688423e-01 -7.87474215e-01 1.00226700e+00 3.52673024e-01 -4.99334857e-02 -1.20902860e+00 1.48994267e+00 1.58093929e-01 -3.52191061e-01 -3.20135206e-01 -3.56350422e-01 -1.41742930e-01 -3.96261543e-01 5.41777015e-01 -1.20251274e+00 3.00238103e-01 7.12517321e-01 7.11453930e-02 -3.51059109e-01 9.30869281e-01 -8.10911283e-02 6.92507863e-01 -4.33917850e-01 -4.50468570e-01 5.55585206e-01 -4.12294954e-01 3.48621339e-01 3.98571402e-01 5.53117335e-01 -5.29889822e-01 -8.60563070e-02 8.42647493e-01 -1.55234128e-01 -1.85124323e-01 -3.30884457e-01 -3.27501833e-01 4.41136539e-01 8.88969362e-01 -7.94417799e-01 -2.23193601e-01 -1.17343836e-01 3.86752337e-01 -1.07800476e-01 2.10614666e-01 -7.71143913e-01 -2.77806282e-01 3.10992360e-01 -1.61391020e-01 5.69085300e-01 2.42933258e-01 -2.37474144e-01 -1.11336625e+00 -8.90495535e-03 -1.49835432e+00 5.52207410e-01 -4.07585800e-01 -1.24995756e+00 5.96629739e-01 -4.89670008e-01 -1.32560325e+00 -7.95638382e-01 -5.54995656e-01 -7.29767203e-01 5.25890529e-01 -1.39156544e+00 -2.66320020e-01 6.38002872e-01 -6.31946092e-03 3.05100471e-01 -1.05623484e+00 3.58906984e-01 -1.12073302e-01 -3.91733080e-01 6.65507615e-01 4.83023345e-01 1.91002171e-02 3.62600058e-01 -1.37976444e+00 3.39180619e-01 5.78178167e-01 9.61122587e-02 2.27986917e-01 8.47817719e-01 -8.48955691e-01 -1.42716897e+00 -5.04291236e-01 4.58819747e-01 1.70373440e-01 1.07899213e+00 -1.95104435e-01 -7.42139339e-01 4.74462539e-01 3.90358418e-01 -3.12162697e-01 3.92440438e-01 -1.66424185e-01 -1.18548088e-01 -4.33385968e-01 -1.24827480e+00 4.96717334e-01 -1.88403562e-01 -2.49622501e-02 -7.10057676e-01 1.65280297e-01 3.76432925e-01 -1.41558275e-01 -8.92256558e-01 3.09906840e-01 5.89282393e-01 -1.25332332e+00 5.04718125e-01 -3.67768079e-01 -2.17451085e-03 9.08094347e-02 -1.01539969e-01 -1.39743114e+00 6.72185197e-02 -1.01972258e+00 -6.05279692e-02 7.73102641e-01 6.54104471e-01 -1.04627228e+00 1.07019114e+00 -4.24930379e-02 2.87898570e-01 -7.71598279e-01 -1.35552430e+00 -1.07207787e+00 6.24388933e-01 1.27698615e-01 7.22086191e-01 8.09350550e-01 5.76538872e-03 -1.71344474e-01 -5.13077080e-01 -1.85401782e-01 9.31700706e-01 5.49155772e-01 5.36675870e-01 -9.37992394e-01 -8.11313868e-01 -8.40394557e-01 -6.66469857e-02 -4.82877433e-01 -1.29603028e-01 -4.25976872e-01 7.58914798e-02 -4.62747902e-01 8.69530588e-02 -3.70738178e-01 -7.83457637e-01 2.80977398e-01 -5.22697531e-03 2.44612359e-02 1.63576603e-01 4.33513612e-01 -3.56210649e-01 5.46354890e-01 9.62446749e-01 -2.62157619e-01 2.89527066e-02 4.29008484e-01 -6.07125521e-01 3.57692450e-01 1.09685791e+00 -4.68866915e-01 -5.92134427e-03 5.07944729e-03 6.92902982e-01 4.97118622e-01 2.22843766e-01 -7.36907303e-01 -7.35493898e-02 -3.32369477e-01 3.79608721e-01 -7.13072956e-01 -1.83856994e-01 -5.94126284e-01 1.92141086e-01 1.38469672e+00 -1.34598419e-01 6.57166362e-01 5.91468513e-02 1.90878883e-01 -4.28188860e-01 -3.54573935e-01 7.05056429e-01 -1.48338661e-01 -2.66152680e-01 4.85478118e-02 -6.03367031e-01 1.92513153e-01 1.20401537e+00 2.04867534e-02 -2.89436698e-01 -4.67751175e-01 -1.90967694e-01 6.65446639e-01 3.41915876e-01 4.04277027e-01 6.62047625e-01 -1.08141112e+00 -9.35265481e-01 6.28137290e-01 -3.13612908e-01 -5.44369876e-01 -1.32620022e-01 5.92766941e-01 -7.80468702e-01 1.82660744e-01 -4.14634764e-01 -3.27581912e-01 -6.36532128e-01 2.81432033e-01 8.70340765e-01 -7.13941097e-01 -3.78144681e-01 6.14956200e-01 -3.08741070e-02 -1.18014932e-01 2.36459851e-01 -1.51434481e-01 2.06302553e-01 3.34863305e-01 1.48865536e-01 1.96838275e-01 2.38063186e-01 1.45452797e-01 -2.72569835e-01 5.59000820e-02 -4.11545992e-01 -3.40177000e-01 1.54827642e+00 4.17511195e-01 3.25702615e-02 2.95094341e-01 7.81965494e-01 -1.67072013e-01 -1.58694518e+00 -3.49506177e-02 5.08736789e-01 -3.54836136e-01 1.46953613e-01 -1.05253160e+00 -1.49596357e+00 5.90825915e-01 7.33586788e-01 8.54505181e-01 6.70374572e-01 -3.73214453e-01 7.09099472e-01 5.63169658e-01 5.67039549e-01 -1.60085750e+00 9.22989473e-02 5.85528612e-01 6.68621659e-01 -1.09127712e+00 6.03772840e-03 1.05044103e+00 -8.84616673e-01 1.21867514e+00 5.13395965e-01 -4.77174819e-01 4.60252941e-01 5.14864028e-01 2.71449357e-01 -1.97895929e-01 -1.21722043e+00 4.84555304e-01 -5.35002947e-01 1.37180790e-01 -1.32758424e-01 4.13178146e-01 -1.64132357e-01 6.55052066e-01 -4.86977458e-01 -9.12925750e-02 7.79549837e-01 8.76933217e-01 -6.33936048e-01 -1.00424683e+00 -2.73027182e-01 6.11864448e-01 -6.50640070e-01 -1.40261233e-01 -3.70220333e-01 1.01648211e+00 -2.27368146e-01 5.31153977e-01 5.41556656e-01 -4.42702264e-01 2.12295458e-01 -1.90030470e-01 2.32378542e-01 8.21970701e-02 -9.93812263e-01 6.52778387e-01 -7.79009750e-03 -2.56014645e-01 1.72954142e-01 -8.36344182e-01 -1.08689940e+00 -4.94770795e-01 -5.90915203e-01 6.99315131e-01 3.40905815e-01 6.56904578e-01 -1.39924929e-01 2.76684642e-01 1.42256343e+00 -2.82909691e-01 -1.62684679e+00 -9.59865272e-01 -1.05051303e+00 -3.88587900e-02 4.23569649e-01 -6.03120446e-01 -7.70723104e-01 -8.07269335e-01]

[4.457311630249023, 3.9754748344421387]

ca4f6807-f08c-4d13-957e-8a39ce0ec474

learning-to-see-the-wood-for-the-trees-deep

1902.10194

http://arxiv.org/abs/1902.10194v1

http://arxiv.org/pdf/1902.10194v1.pdf

Learning to See the Wood for the Trees: Deep Laser Localization in Urban and Natural Environments on a CPU

Localization in challenging, natural environments such as forests or woodlands is an important capability for many applications from guiding a robot navigating along a forest trail to monitoring vegetation growth with handheld sensors. In this work we explore laser-based localization in both urban and natural environments, which is suitable for online applications. We propose a deep learning approach capable of learning meaningful descriptors directly from 3D point clouds by comparing triplets (anchor, positive and negative examples). The approach learns a feature space representation for a set of segmented point clouds that are matched between a current and previous observations. Our learning method is tailored towards loop closure detection resulting in a small model which can be deployed using only a CPU. The proposed learning method would allow the full pipeline to run on robots with limited computational payload such as drones, quadrupeds or UGVs.

['Adrian Penate-Sanchez', 'Maurice Fallon', 'Georgi Tinchev']

2019-02-26

['loop-closure-detection']

['computer-vision']

[ 1.28573954e-01 7.57999066e-03 -5.69943860e-02 -5.44822693e-01 -4.69617844e-01 -7.34609604e-01 7.00910687e-01 6.13559902e-01 -7.16366768e-01 6.45208836e-01 -5.27651131e-01 -2.14859322e-01 -1.09581538e-01 -1.00495887e+00 -8.58576119e-01 -3.02795470e-01 -8.55973303e-01 8.69044185e-01 5.09325743e-01 -3.07250291e-01 1.95019364e-01 9.79054689e-01 -1.97258937e+00 -3.01565915e-01 5.06241918e-01 1.01492369e+00 6.55882537e-01 7.22772121e-01 1.26410639e-02 -8.20489526e-02 -4.78780493e-02 3.52112383e-01 6.11367524e-01 4.05920178e-01 -3.68451566e-01 1.17134256e-02 7.13542402e-01 -4.36734825e-01 1.46474585e-01 8.54352951e-01 3.94758940e-01 5.55978902e-02 5.13443172e-01 -1.42927384e+00 1.09182701e-01 8.07296485e-02 -4.00768518e-01 -1.27241701e-01 4.87846196e-01 1.98118418e-01 8.62245202e-01 -7.58416772e-01 6.26012027e-01 1.09683192e+00 9.94451225e-01 1.29538719e-02 -1.14869547e+00 -3.21189076e-01 1.40008166e-01 1.64745450e-01 -1.35540557e+00 -3.10453892e-01 3.42221588e-01 -5.08227587e-01 1.04525650e+00 -3.36986855e-02 8.84723663e-01 7.07026482e-01 1.23588599e-01 4.50113386e-01 9.15100813e-01 -2.98117876e-01 4.95849550e-01 -8.43191966e-02 -2.68664122e-01 9.86546457e-01 5.23947060e-01 2.82058299e-01 -4.68307674e-01 -2.17280701e-01 7.29762375e-01 2.96881258e-01 3.26466262e-02 -1.24703705e+00 -1.32569325e+00 8.49319160e-01 9.78433907e-01 -1.50983194e-02 -6.13370478e-01 3.10647815e-01 2.26871565e-01 4.25478607e-01 2.01331690e-01 2.27666304e-01 -7.09913373e-01 -1.13524543e-02 -8.53423059e-01 5.64885795e-01 9.04288173e-01 1.20612431e+00 1.38478649e+00 -2.86145419e-01 5.18449306e-01 3.60369682e-01 4.60794359e-01 9.09346998e-01 2.39439145e-01 -8.37526441e-01 2.23514631e-01 7.90873945e-01 5.86891890e-01 -9.64310229e-01 -8.41917157e-01 -3.06449354e-01 -5.40232956e-01 6.70556128e-01 1.59777805e-01 -2.04086751e-01 -8.98532748e-01 1.30272555e+00 6.94640636e-01 1.40575087e-02 4.36572172e-02 9.82227683e-01 6.18661702e-01 3.55316967e-01 -2.68872619e-01 2.15638846e-01 1.13264322e+00 -6.68111801e-01 -1.55857916e-03 -7.42819250e-01 7.23656476e-01 -4.02224600e-01 9.21833336e-01 3.34939808e-01 -4.18807268e-01 -3.39038223e-01 -1.15865529e+00 -1.38667434e-01 -6.83368564e-01 2.28239715e-01 7.11294353e-01 2.48565018e-01 -1.37299347e+00 8.25850725e-01 -1.27563858e+00 -1.13822186e+00 2.43291944e-01 6.68715000e-01 -6.55707061e-01 -8.01309645e-02 -6.41043067e-01 1.01477766e+00 4.14972037e-01 2.22262546e-01 -9.35910702e-01 -6.17157146e-02 -1.14658332e+00 -2.49128044e-01 1.15292773e-01 -5.76777995e-01 9.64115322e-01 -4.20542568e-01 -1.21334636e+00 1.04607248e+00 -1.43333122e-01 -9.22404528e-01 7.40983427e-01 -4.64008272e-01 4.49382588e-02 -1.36877626e-01 3.96186680e-01 9.72999096e-01 8.97353947e-01 -1.25907886e+00 -1.07875085e+00 -8.10614467e-01 5.66086322e-02 3.71650070e-01 1.62787601e-01 -5.33151925e-01 7.28791356e-02 7.95781761e-02 7.61478901e-01 -1.38351333e+00 -5.17851293e-01 6.26953423e-01 -7.83225819e-02 -6.75100982e-02 1.11030304e+00 -1.35049701e-01 1.64067626e-01 -1.67662561e+00 -5.91160804e-02 3.44659656e-01 -1.82837799e-01 -4.92018322e-03 -1.06908716e-01 7.31767654e-01 4.29803312e-01 -2.04611212e-01 -2.36040428e-01 -1.94849178e-01 1.47799462e-01 4.44788069e-01 -1.28915966e-01 8.80495191e-01 1.41555905e-01 5.83655834e-01 -1.08142757e+00 -2.77085930e-01 4.77799833e-01 3.60214770e-01 -3.18829745e-01 2.69308537e-01 -2.16081262e-01 4.86244440e-01 -3.84881347e-01 9.00725424e-01 8.39734912e-01 2.37233564e-01 7.14611867e-03 1.79639131e-01 -6.72564983e-01 2.42979452e-01 -1.21929526e+00 2.19559216e+00 -7.01926887e-01 6.39499784e-01 4.03245240e-01 -6.98823452e-01 1.37271285e+00 -2.56791383e-01 3.22819948e-01 -3.01087439e-01 -8.13313723e-02 4.26188380e-01 -4.21029508e-01 -3.23234320e-01 6.44564509e-01 1.15890764e-01 -8.19916725e-02 1.27039567e-01 6.37968034e-02 -5.35826564e-01 -5.41383326e-02 -2.83899456e-01 1.32013798e+00 4.25984323e-01 4.84669417e-01 -2.57117420e-01 4.10237402e-01 4.36026901e-01 3.18036169e-01 9.16942954e-01 -2.92964667e-01 2.39747450e-01 -6.71105757e-02 -9.18848693e-01 -9.75236416e-01 -1.11788607e+00 -2.04699591e-01 1.05042553e+00 3.79704863e-01 -2.52769530e-01 -2.26168204e-02 -4.20311242e-01 4.12111014e-01 2.57908702e-01 -2.65924275e-01 1.31002471e-01 -4.71850246e-01 -3.33156064e-02 1.57292590e-01 3.48380506e-01 5.12282670e-01 -8.96092117e-01 -1.47315121e+00 2.12739497e-01 1.91334948e-01 -8.99024129e-01 4.51307654e-01 6.88574612e-01 -1.03124261e+00 -9.79949951e-01 -3.05976242e-01 -9.38054919e-01 5.11896968e-01 5.47179461e-01 8.17996800e-01 -1.18226595e-01 -1.25637725e-01 4.93362069e-01 -4.35642332e-01 -6.74833119e-01 6.63492903e-02 3.12442243e-01 4.88639116e-01 -4.54633445e-01 5.00650704e-01 -9.03906047e-01 -4.96517181e-01 1.57471508e-01 -3.28571051e-01 -3.13282073e-01 7.66261220e-01 7.95924783e-01 7.44083583e-01 -2.36124992e-01 -4.33530025e-02 -3.56745690e-01 2.24258289e-01 -5.58396757e-01 -1.11382186e+00 -1.44316792e-01 -3.32586437e-01 3.34388204e-02 3.58878136e-01 -1.05051793e-01 -1.91450506e-01 9.15209830e-01 -2.95492690e-02 -2.63464689e-01 -6.81877851e-01 5.37425876e-01 8.14479217e-02 -5.31639218e-01 8.06805134e-01 8.29907507e-02 1.19663559e-01 -4.48297411e-01 4.67254132e-01 6.70367479e-01 6.21855259e-01 -4.00268883e-01 1.12984884e+00 7.03702271e-01 4.16636705e-01 -1.08922791e+00 -3.32255572e-01 -8.78022790e-01 -1.26913273e+00 -2.46417299e-02 2.36495629e-01 -1.15367270e+00 -8.14744294e-01 5.55572920e-02 -1.11414135e+00 -2.78036654e-01 -3.21968347e-01 5.91226876e-01 -8.93355787e-01 1.77879244e-01 6.77222088e-02 -8.24560702e-01 -3.51430953e-01 -7.94623315e-01 1.57107830e+00 2.64392167e-01 -5.72236702e-02 -7.64608860e-01 4.44533765e-01 -3.08767974e-01 2.16761768e-01 6.62318885e-01 1.60554826e-01 -6.72914863e-01 -8.59964490e-01 -5.50822437e-01 -9.66831818e-02 -1.82918727e-01 1.60456464e-01 -2.41293043e-01 -8.92329812e-01 -7.37595856e-01 -2.93393701e-01 -5.99511147e-01 7.69445240e-01 1.69422030e-01 5.20400465e-01 -1.55461177e-01 -7.49500573e-01 7.13665366e-01 1.52521300e+00 -2.07764238e-01 1.07617721e-01 7.63714194e-01 3.64286095e-01 6.44296110e-01 1.20417464e+00 5.82135081e-01 4.61307079e-01 5.81524193e-01 1.17859066e+00 -2.31831567e-03 5.86359918e-01 -3.71264040e-01 2.19780609e-01 -5.33223934e-02 1.03482930e-02 1.02465168e-01 -1.33163393e+00 9.02999699e-01 -2.03973913e+00 -5.66568375e-01 -2.00856656e-01 2.54293489e+00 -2.37610657e-03 1.72645539e-01 -1.40959606e-01 -3.34337875e-02 5.45167506e-01 5.25015146e-02 -7.40869343e-01 -1.97018445e-01 2.41217360e-01 1.41294822e-01 1.08423007e+00 4.55610633e-01 -1.50230539e+00 1.09741378e+00 5.50379848e+00 -4.43400741e-02 -1.27749455e+00 -1.61535859e-01 -3.79027903e-01 2.86492854e-01 2.30168790e-01 3.31827223e-01 -8.55443120e-01 -8.28172266e-03 8.28716815e-01 1.82364359e-01 1.56533971e-01 1.36335027e+00 2.64097154e-01 -3.87905121e-01 -1.24409068e+00 1.08348942e+00 -2.97114462e-01 -1.06109726e+00 -3.40933889e-01 1.71088487e-01 2.93605268e-01 8.40654373e-01 -1.38626695e-01 1.84342206e-01 4.47232783e-01 -8.36492240e-01 6.58496380e-01 2.68656701e-01 6.19007587e-01 -5.07743537e-01 7.20207930e-01 7.60873437e-01 -1.33702600e+00 -1.69727802e-01 -6.60033584e-01 -4.49894845e-01 1.12139188e-01 3.73673886e-01 -1.63735271e+00 3.39966178e-01 9.23459888e-01 8.55966568e-01 -5.63510776e-01 1.40438342e+00 -2.24073246e-01 1.01829968e-01 -9.92318273e-01 -1.50931478e-01 4.38497484e-01 -2.59995103e-01 7.81276703e-01 1.01968706e+00 7.74550915e-01 -3.19212884e-01 6.35012031e-01 6.41160071e-01 3.30716819e-01 -1.13413529e-02 -1.25178421e+00 3.40656728e-01 5.27067065e-01 1.52033126e+00 -7.89417088e-01 8.11165720e-02 -1.26606017e-01 9.38020349e-01 5.10275006e-01 -1.46727115e-01 -3.58881056e-01 -5.17567217e-01 8.84331107e-01 2.75988638e-01 3.70091200e-01 -9.74574864e-01 5.27533293e-02 -1.02352118e+00 5.11313491e-02 -1.86538905e-01 -6.77518994e-02 -6.54280961e-01 -8.22232723e-01 4.33473855e-01 -3.36602628e-01 -1.62092352e+00 -4.08787131e-01 -6.96818233e-01 -4.62103069e-01 6.54254615e-01 -1.88100290e+00 -1.42796683e+00 -8.66075814e-01 4.54062790e-01 2.62971342e-01 -6.23495877e-03 9.49790120e-01 -2.13819206e-01 2.23796412e-01 -1.98342830e-01 1.45433977e-01 -2.21162155e-01 6.91486180e-01 -1.39046693e+00 5.80195725e-01 7.69208729e-01 3.02800953e-01 4.75292355e-01 8.78953755e-01 -7.41701424e-01 -1.59233570e+00 -1.20567322e+00 7.87768602e-01 -3.15781116e-01 6.40320182e-01 -4.98459220e-01 -5.93472004e-01 7.46687651e-01 -2.92977631e-01 3.44024122e-01 3.03291738e-01 9.67852250e-02 -1.38034709e-02 -3.43658268e-01 -1.40732026e+00 2.14294478e-01 1.24723768e+00 -1.84870198e-01 -3.28741938e-01 6.96885705e-01 4.03973043e-01 -6.77724957e-01 -6.99847341e-01 5.09874821e-01 6.07038200e-01 -1.00300336e+00 7.89315164e-01 -4.22859341e-01 -1.69018105e-01 -6.16574526e-01 -4.17019337e-01 -1.46265137e+00 -1.67471185e-01 -4.99312729e-01 1.29587799e-01 8.12943578e-01 2.12600846e-02 -6.06195986e-01 1.02223134e+00 1.77383274e-01 -1.16639592e-01 -1.28975913e-01 -1.07059216e+00 -8.15673470e-01 -3.46303076e-01 -1.57758459e-01 6.34130001e-01 4.70547080e-01 -2.99812585e-01 1.26198977e-01 -9.57317352e-02 7.76869833e-01 6.85651183e-01 3.13577622e-01 1.30883050e+00 -1.84251118e+00 3.42363536e-01 2.59450525e-01 -1.05720448e+00 -1.07947922e+00 2.55690604e-01 -8.03570271e-01 4.31829810e-01 -1.72610068e+00 -6.23149216e-01 -8.95416439e-01 1.87721655e-01 6.37004316e-01 6.03488207e-01 1.95561219e-02 1.26650734e-02 2.29661301e-01 -4.88402307e-01 5.34804642e-01 4.47615027e-01 -1.57626465e-01 -3.22349310e-01 3.86429727e-01 5.75337075e-02 8.47162545e-01 8.18844736e-01 -4.43168104e-01 -2.09639952e-01 -5.95305204e-01 1.42787382e-01 -1.32562310e-01 7.56641507e-01 -1.53756464e+00 4.59458500e-01 -9.82044265e-03 3.64968956e-01 -8.57547581e-01 5.16373456e-01 -1.29906332e+00 -3.73758748e-02 7.55028248e-01 6.13399185e-02 1.79567114e-01 6.99484721e-02 8.34603071e-01 -1.78233720e-02 -2.36774489e-01 5.52124619e-01 -4.31931734e-01 -1.19720793e+00 3.01220179e-01 -2.15205371e-01 -5.57961345e-01 1.06434155e+00 -2.72940189e-01 -7.17469603e-02 -3.00154358e-01 -6.94230676e-01 5.08585632e-01 1.02206433e+00 3.74736637e-01 7.27640271e-01 -1.08555126e+00 -5.59847713e-01 4.36440706e-01 5.05713642e-01 4.76365030e-01 -3.28230411e-01 4.77154315e-01 -1.08065498e+00 5.46803355e-01 -5.46970367e-01 -1.15721607e+00 -1.12167501e+00 3.50293428e-01 1.85379371e-01 1.40226632e-01 -6.29299283e-01 6.72149301e-01 -4.64378953e-01 -1.05750370e+00 1.13539346e-01 -6.95280015e-01 -9.52980369e-02 2.87481099e-02 1.74218789e-01 2.79515415e-01 3.38206142e-01 -8.55164587e-01 -6.39792919e-01 7.38951385e-01 3.62824529e-01 -2.50251200e-02 1.50291693e+00 -2.60941446e-01 -6.47257715e-02 5.53205967e-01 9.82548296e-01 -2.51877934e-01 -1.42199934e+00 -1.89544126e-01 3.98547769e-01 -5.12544751e-01 8.93009827e-03 -3.23322058e-01 -4.08882320e-01 8.00016880e-01 1.33444214e+00 -6.04356006e-02 6.48801327e-01 -9.47957672e-03 4.82168883e-01 1.14631367e+00 1.24404073e+00 -8.26534808e-01 -3.48384202e-01 6.14936471e-01 7.62653708e-01 -1.68149161e+00 8.95083174e-02 -1.50394663e-02 -3.07785392e-01 1.24007595e+00 3.37045908e-01 -4.85929251e-01 5.72929800e-01 2.41859213e-01 8.02188143e-02 -1.10285625e-01 -5.55589497e-01 -8.28691185e-01 -1.58531636e-01 1.24125767e+00 -1.27338395e-02 1.08629294e-01 -8.31704810e-02 -8.43560994e-02 -3.97786289e-01 -3.39318328e-02 4.73937720e-01 1.33605790e+00 -1.02952325e+00 -9.31803823e-01 -3.48133951e-01 4.12524909e-01 3.35914940e-01 1.76230401e-01 -3.76509398e-01 7.04172254e-01 3.78239542e-01 7.85230935e-01 1.95999146e-01 -3.13297182e-01 4.72502470e-01 -1.56920522e-01 2.56159604e-01 -7.86560893e-01 -2.47291833e-01 -1.80708930e-01 -1.93035752e-02 -8.00120771e-01 -5.42858243e-01 -8.59276831e-01 -1.16405702e+00 1.20395407e-01 -2.14106381e-01 2.66075637e-02 1.25572383e+00 6.81684315e-01 5.14134228e-01 -3.17239583e-01 4.61374372e-01 -1.60193288e+00 -5.19082129e-01 -8.55140507e-01 -7.05984652e-01 -1.57963216e-01 6.38788521e-01 -9.37557578e-01 -2.53399946e-02 -2.06162289e-01]

[7.406903266906738, -2.065643548965454]

9fb104af-d757-43a7-ac35-45b29ca6f03e

deep-directional-statistics-pose-estimation

1805.03430

http://arxiv.org/abs/1805.03430v1

http://arxiv.org/pdf/1805.03430v1.pdf

Deep Directional Statistics: Pose Estimation with Uncertainty Quantification

Modern deep learning systems successfully solve many perception tasks such as object pose estimation when the input image is of high quality. However, in challenging imaging conditions such as on low-resolution images or when the image is corrupted by imaging artifacts, current systems degrade considerably in accuracy. While a loss in performance is unavoidable, we would like our models to quantify their uncertainty in order to achieve robustness against images of varying quality. Probabilistic deep learning models combine the expressive power of deep learning with uncertainty quantification. In this paper, we propose a novel probabilistic deep learning model for the task of angular regression. Our model uses von Mises distributions to predict a distribution over object pose angle. Whereas a single von Mises distribution is making strong assumptions about the shape of the distribution, we extend the basic model to predict a mixture of von Mises distributions. We show how to learn a mixture model using a finite and infinite number of mixture components. Our model allows for likelihood-based training and efficient inference at test time. We demonstrate on a number of challenging pose estimation datasets that our model produces calibrated probability predictions and competitive or superior point estimates compared to the current state-of-the-art.

['Peter Gehler', 'Sergey Prokudin', 'Sebastian Nowozin']

2018-05-09

deep-directional-statistics-pose-estimation-1

http://openaccess.thecvf.com/content_ECCV_2018/html/Sergey_Prokudin_Deep_Directional_Statistics_ECCV_2018_paper.html

http://openaccess.thecvf.com/content_ECCV_2018/papers/Sergey_Prokudin_Deep_Directional_Statistics_ECCV_2018_paper.pdf

eccv-2018-9

['probabilistic-deep-learning']

['computer-vision']

[ 2.01230153e-01 7.26073384e-02 6.88739121e-02 -5.74790716e-01 -1.20469964e+00 -5.10590613e-01 4.62817550e-01 -6.75677583e-02 -5.22520423e-01 7.08959341e-01 -1.35742813e-01 6.45696465e-03 -3.21379811e-01 -5.49196005e-01 -1.34025955e+00 -7.41714656e-01 8.41634721e-02 9.42067206e-01 3.05553466e-01 4.43191975e-01 1.00699596e-01 5.85712254e-01 -1.40494919e+00 -6.82887658e-02 5.31448841e-01 1.27791238e+00 2.54662991e-01 6.53844893e-01 2.53027081e-01 4.74770844e-01 -6.10867321e-01 -3.63667130e-01 3.27851385e-01 2.25506335e-01 -4.83446777e-01 9.90106314e-02 7.42429435e-01 -8.04498613e-01 -2.25576207e-01 1.12034011e+00 4.45220917e-01 -7.75481239e-02 1.20351899e+00 -1.22977185e+00 -2.92720407e-01 4.14606214e-01 -6.04815841e-01 -9.16770101e-02 1.01979613e-01 1.51060820e-01 8.22799861e-01 -7.03585744e-01 3.00437480e-01 1.02505982e+00 8.56892467e-01 1.78751573e-01 -1.39314485e+00 -5.21269500e-01 -4.57387697e-03 1.44923255e-01 -1.43216860e+00 -2.69563824e-01 5.42239070e-01 -7.29957521e-01 5.27144253e-01 -2.08990291e-01 2.25866541e-01 1.26903510e+00 4.44848299e-01 7.62969911e-01 1.01714337e+00 -1.00676380e-01 3.20021510e-01 3.63563336e-02 -2.83325821e-01 5.66686273e-01 3.65089118e-01 1.02858327e-01 -4.59396750e-01 -1.92562148e-01 1.11230803e+00 -2.02464014e-01 -2.16282845e-01 -8.31899822e-01 -1.07799995e+00 6.81376457e-01 5.32791734e-01 -2.08284870e-01 -4.26751137e-01 6.00746989e-01 -1.02504663e-01 -3.56464267e-01 2.19039202e-01 5.29791951e-01 -4.14498121e-01 -7.08882436e-02 -1.07127988e+00 4.83142793e-01 7.60208666e-01 8.32664430e-01 4.22075659e-01 -1.14690941e-02 -1.06555134e-01 8.65489006e-01 5.72398424e-01 6.51711643e-01 -4.67129909e-02 -1.41480136e+00 5.72436713e-02 -1.90844297e-01 5.67144752e-01 -8.31182420e-01 -5.09682894e-01 -7.82790482e-01 -7.33243406e-01 5.33540905e-01 8.30569625e-01 6.21474907e-02 -1.04006362e+00 1.91421318e+00 1.16154559e-01 7.28301182e-02 -3.39588910e-01 1.11612785e+00 3.92281771e-01 5.48115790e-01 -1.46888331e-01 -5.16649596e-02 1.12736380e+00 -3.48966181e-01 -4.92285907e-01 -2.51391888e-01 -1.00938775e-01 -9.18242455e-01 7.57350683e-01 9.43956733e-01 -1.08413672e+00 -4.31397647e-01 -1.16773891e+00 -3.84575166e-02 1.97389647e-01 1.58668518e-01 5.19331634e-01 6.11302078e-01 -7.75688469e-01 7.78950214e-01 -9.60905135e-01 8.53552446e-02 7.49437928e-01 3.50817949e-01 -3.14078152e-01 -1.29352286e-01 -7.02345192e-01 1.11725581e+00 2.30443910e-01 6.42531440e-02 -1.06432414e+00 -7.86113441e-01 -8.20134461e-01 1.76378209e-02 4.51340944e-01 -8.90502512e-01 1.51768470e+00 -5.66161871e-01 -1.45479226e+00 5.29882193e-01 8.30665231e-02 -5.29062688e-01 8.07535648e-01 -6.18900180e-01 2.20928013e-01 5.30498736e-02 -1.28250927e-01 9.30154562e-01 1.08029830e+00 -1.47924745e+00 -1.79400727e-01 -2.59007663e-01 1.12199381e-01 6.47406653e-02 1.78864986e-01 -4.01923329e-01 -3.75045151e-01 -3.97636265e-01 2.96555340e-01 -1.02523077e+00 -1.14169002e-01 4.97136950e-01 -3.41632158e-01 -3.34922336e-02 3.26718628e-01 -5.43413103e-01 3.32089424e-01 -1.89254022e+00 1.35576993e-01 1.02610789e-01 2.12435067e-01 -2.10072905e-01 4.82887365e-02 -8.55898857e-02 2.36132205e-01 -8.01345631e-02 -3.79309565e-01 -3.99199992e-01 3.10114622e-01 2.18856975e-01 -3.05019438e-01 8.02163363e-01 3.27101409e-01 5.69550574e-01 -7.14604437e-01 -4.06759024e-01 4.25884634e-01 7.17164755e-01 -7.11217642e-01 3.08025837e-01 -4.91099268e-01 4.65097725e-01 -1.25739202e-01 4.20034796e-01 7.72916496e-01 -4.07265335e-01 -3.04252710e-02 -6.32313490e-01 3.50908697e-01 1.48355901e-01 -1.43959689e+00 1.66418719e+00 -4.86213416e-01 6.39872551e-01 1.18654452e-01 -7.27248311e-01 7.31869400e-01 2.61772536e-02 5.12607992e-01 -1.29663527e-01 4.00234848e-01 9.18629616e-02 9.33432952e-02 -2.27745906e-01 4.35539603e-01 -3.99156958e-01 -7.39810392e-02 2.42454425e-01 3.75745475e-01 -6.43875360e-01 -2.06879765e-01 -1.68305710e-02 7.89470673e-01 5.24255872e-01 1.65962011e-01 -4.06103209e-02 -1.32130787e-01 -3.62284184e-01 4.43458080e-01 8.34864080e-01 -1.72705859e-01 1.05539513e+00 4.63985771e-01 -2.65212446e-01 -1.26362991e+00 -1.78099918e+00 -6.21049941e-01 6.99835360e-01 1.20861992e-01 -3.41609679e-02 -5.41599512e-01 -3.30511421e-01 1.78850126e-02 7.63112664e-01 -3.04022074e-01 -5.08613102e-02 -2.78993964e-01 -9.59600449e-01 2.98539579e-01 7.66999960e-01 3.61239314e-01 -5.18120885e-01 -5.38579524e-01 7.93653503e-02 -1.90715790e-01 -1.40707493e+00 -2.03160584e-01 1.14491440e-01 -5.98852694e-01 -8.60556662e-01 -6.53081238e-01 -2.62743324e-01 3.42745721e-01 -3.14151287e-01 1.25993812e+00 -7.69346297e-01 -4.10825074e-01 5.56332231e-01 6.97192475e-02 -6.60507381e-01 -2.82667369e-01 -3.22963297e-01 3.55099171e-01 -3.15510720e-01 6.67638257e-02 -6.64504826e-01 -6.76898658e-01 2.97457993e-01 -7.94495940e-01 -1.62854999e-01 7.66329229e-01 8.53640318e-01 7.70453513e-01 1.57327443e-01 1.89593434e-01 -3.86724293e-01 2.54683465e-01 -3.60952914e-01 -9.14790869e-01 -3.30795906e-02 -2.83730179e-01 5.12037933e-01 1.99299142e-01 -6.21613324e-01 -8.86731505e-01 1.80613518e-01 -2.25582048e-01 -7.08601236e-01 -2.08201423e-01 3.09840560e-01 -1.61525831e-02 -7.87163600e-02 7.22906113e-01 -3.05665936e-02 3.61679532e-02 -4.28807497e-01 2.63785303e-01 2.99724489e-01 8.22525203e-01 -8.73076737e-01 6.64135098e-01 4.76890415e-01 3.75872284e-01 -6.08930945e-01 -9.73426580e-01 -4.22015488e-02 -5.59416890e-01 -3.12684476e-01 8.33438456e-01 -9.40122306e-01 -1.13096094e+00 5.07600546e-01 -1.12814069e+00 -2.26165235e-01 -1.32728234e-01 8.38591576e-01 -9.05771971e-01 3.02132219e-01 -5.23388624e-01 -8.90846550e-01 1.19054362e-01 -1.36658394e+00 1.23231769e+00 1.75742730e-01 -1.39623061e-01 -6.60298586e-01 1.91369597e-02 3.30804944e-01 3.44008178e-01 2.69773573e-01 5.80344200e-01 -3.95086795e-01 -9.05548334e-01 -3.09374690e-01 -3.25360507e-01 5.28430581e-01 -2.60473281e-01 1.10446900e-01 -1.03335738e+00 -1.13505960e-01 6.96181729e-02 -6.65763319e-01 7.90611148e-01 9.65358198e-01 1.67289424e+00 -4.76390943e-02 -1.01511978e-01 6.99632108e-01 1.29742670e+00 -2.03273103e-01 4.81510848e-01 1.42293632e-01 6.24016285e-01 5.58784068e-01 4.16766942e-01 4.42606837e-01 2.67540872e-01 7.38307476e-01 8.46529841e-01 3.66973549e-01 1.02256790e-01 -1.20045073e-01 3.71484049e-02 2.14869708e-01 6.96262391e-03 -2.39679441e-01 -9.34559345e-01 3.04152071e-01 -1.54464090e+00 -7.93487966e-01 2.40859583e-01 2.38977265e+00 7.77591467e-01 5.39621115e-01 -1.98311564e-02 -9.54385102e-02 4.08978254e-01 -3.01237963e-02 -8.14782739e-01 -1.77895389e-02 6.03001714e-02 1.80902466e-01 6.94381177e-01 5.25151014e-01 -1.17439818e+00 4.57952380e-01 6.49087429e+00 8.85780811e-01 -1.03227055e+00 -7.54093900e-02 6.89728439e-01 -2.99018770e-01 -2.02474073e-01 -3.04868668e-01 -8.02840710e-01 3.67476970e-01 6.64153814e-01 2.51533657e-01 1.90938115e-01 9.67874229e-01 -2.86840379e-01 -3.45294416e-01 -1.37308931e+00 1.14348245e+00 6.53395280e-02 -1.11701608e+00 -2.79325068e-01 8.73331726e-02 5.28181970e-01 2.56440938e-01 4.33406323e-01 2.09418774e-01 5.10549605e-01 -1.33813369e+00 9.54100192e-01 8.18672597e-01 7.64966607e-01 -6.49116337e-01 7.44302809e-01 4.98075277e-01 -5.46013117e-01 2.52782315e-01 -4.87143397e-01 1.84057906e-01 3.05919141e-01 9.51841772e-01 -9.84517992e-01 2.80282825e-01 7.12637782e-01 2.58423805e-01 -2.56549597e-01 1.22079062e+00 -2.20163032e-01 4.70003039e-01 -8.20755899e-01 6.34748936e-02 -4.50796671e-02 -2.32582055e-02 6.39173508e-01 9.60539520e-01 3.78809124e-01 -6.55610263e-02 3.10757440e-02 1.18170059e+00 -1.40161484e-01 -4.41995770e-01 -3.26568186e-01 2.13206619e-01 4.99821365e-01 1.17426229e+00 -4.27843422e-01 -1.10974517e-02 -7.07019446e-03 7.36255765e-01 3.94006401e-01 2.62505829e-01 -9.46199596e-01 1.58341229e-01 5.36465049e-01 1.53177202e-01 3.78250360e-01 -5.50367057e-01 -3.70686501e-01 -1.01302266e+00 1.43175885e-01 -6.06557488e-01 -1.08979922e-02 -1.08648002e+00 -1.55688167e+00 4.64436620e-01 4.72795784e-01 -1.05167675e+00 -5.23961425e-01 -1.09315002e+00 -2.46987164e-01 8.33327055e-01 -1.18169069e+00 -1.03592813e+00 -1.35648757e-01 1.75118074e-01 2.65117228e-01 1.04794011e-01 5.44336081e-01 1.31254336e-02 -1.20553851e-01 4.20489430e-01 8.87990929e-03 7.71909812e-03 8.62086773e-01 -1.39080918e+00 2.63300147e-02 5.68423808e-01 2.39756033e-01 4.75230187e-01 1.23686934e+00 -3.50582778e-01 -1.13656044e+00 -7.32286453e-01 1.00841366e-01 -6.21421933e-01 4.59143072e-01 -3.54519069e-01 -8.65157306e-01 5.77070296e-01 -2.02163398e-01 3.69353503e-01 5.25685906e-01 2.76026636e-01 -6.11508369e-01 -1.43453106e-01 -1.13759053e+00 3.38661283e-01 6.23078406e-01 -4.12235439e-01 -5.44571280e-01 3.96048754e-01 4.19198990e-01 -7.44485021e-01 -9.43102896e-01 9.32724655e-01 1.01685417e+00 -1.11298931e+00 1.16980100e+00 -2.92154729e-01 5.46055675e-01 -2.99417824e-01 -4.86301899e-01 -1.26049471e+00 -2.96848178e-01 -1.75177827e-01 -3.18765700e-01 7.70446956e-01 2.60273665e-01 -2.85859108e-01 7.46411681e-01 6.68507397e-01 -4.44714576e-02 -8.07283461e-01 -1.12970650e+00 -8.26294601e-01 2.56095618e-01 -8.49140048e-01 1.13251805e-01 3.74339968e-01 -6.28668964e-01 1.42842233e-01 -4.17764872e-01 6.46222711e-01 1.18190658e+00 1.95255801e-02 7.35521793e-01 -1.19242907e+00 -7.23467350e-01 -5.40040195e-01 -5.51841140e-01 -1.14630890e+00 4.71933186e-02 -3.23304832e-01 4.39498663e-01 -1.40590262e+00 3.56536210e-01 -4.03335690e-01 -1.12833142e-01 -1.40279224e-02 2.22006906e-02 2.84975559e-01 7.26853609e-02 3.77584845e-02 -5.34756899e-01 6.24229252e-01 9.74126458e-01 -1.07252650e-01 1.80542365e-01 2.20727503e-01 -3.90640378e-01 1.04307818e+00 6.27944052e-01 -4.51563179e-01 -3.13908488e-01 -5.20812094e-01 5.18029034e-01 1.36912987e-02 6.78564429e-01 -1.20148075e+00 1.97410941e-01 -2.30055004e-02 9.64586496e-01 -6.92516685e-01 7.77572632e-01 -7.53949821e-01 1.42385691e-01 1.45284802e-01 -3.97474498e-01 -4.97534633e-01 3.27984482e-01 7.35283017e-01 1.32444322e-01 -1.91284120e-01 1.06305957e+00 5.29587604e-02 -2.69452929e-01 3.93137962e-01 -1.62203267e-01 -9.94121507e-02 7.26268232e-01 2.37589702e-02 -2.06063911e-01 -6.71270728e-01 -8.82131338e-01 4.22068387e-02 4.56360728e-01 3.20008576e-01 6.31141067e-01 -1.21392691e+00 -7.37110913e-01 -6.21526241e-02 6.70660511e-02 3.35629404e-01 2.55998135e-01 6.79981112e-01 -4.62162703e-01 1.01264074e-01 -2.47109622e-01 -1.14159727e+00 -8.45112979e-01 3.70771140e-01 4.07053381e-01 -8.08870196e-02 -1.75049022e-01 1.18135202e+00 3.69387567e-01 -3.59952748e-01 4.03102010e-01 -4.32790577e-01 1.76844954e-01 -3.36253285e-01 3.34050745e-01 1.64188027e-01 -1.30005985e-01 -4.45765465e-01 -2.82621801e-01 6.07675254e-01 -8.25974345e-03 -3.81914437e-01 1.23449361e+00 9.46713146e-03 2.65477031e-01 6.87659442e-01 1.04782343e+00 -1.14309222e-01 -1.90683424e+00 -2.22591117e-01 -4.11816150e-01 -4.11274165e-01 2.59107232e-01 -8.53839517e-01 -7.10839570e-01 1.01597130e+00 7.21478641e-01 -3.24580818e-01 8.72887254e-01 2.88682520e-01 4.79941696e-01 3.53742421e-01 4.10042971e-01 -8.84992599e-01 2.48901740e-01 3.36212218e-01 9.99172270e-01 -1.65804446e+00 2.52749652e-01 -2.71437585e-01 -5.16567588e-01 1.05006349e+00 5.74982345e-01 -3.48384202e-01 7.97004759e-01 6.46468163e-01 -2.72838175e-01 4.37544808e-02 -5.90453565e-01 -3.73436771e-02 7.51929641e-01 7.16401875e-01 1.85026303e-01 5.57741374e-02 2.47906759e-01 5.36578894e-01 -4.26766515e-01 -8.90832096e-02 4.09190476e-01 6.85491800e-01 -5.95067739e-01 -6.75619066e-01 -5.74069738e-01 5.42025149e-01 -5.41608930e-01 7.30472133e-02 1.07655838e-01 6.29127204e-01 1.56044915e-01 4.68955964e-01 3.34163904e-01 -2.00428873e-01 2.67405119e-02 -6.38236478e-02 1.16137898e+00 -4.49244857e-01 2.92116851e-01 2.85034209e-01 -1.47327766e-01 -6.36826396e-01 -3.71259481e-01 -6.91482484e-01 -8.83331358e-01 -2.41442733e-02 -3.03196877e-01 -3.44136655e-01 1.07153690e+00 1.01494229e+00 -8.52821097e-02 5.31988442e-01 1.24586254e-01 -1.12561870e+00 -9.85027611e-01 -1.08720434e+00 -6.32845759e-01 1.18832573e-01 4.72463608e-01 -1.09025180e+00 -3.37960482e-01 7.33490065e-02]

[7.456586837768555, -1.4058055877685547]

f846ad06-e30a-42f2-8711-0c54ffb43236

lilt-a-simple-yet-effective-language

2202.13669

https://arxiv.org/abs/2202.13669v1

https://arxiv.org/pdf/2202.13669v1.pdf

LiLT: A Simple yet Effective Language-Independent Layout Transformer for Structured Document Understanding

Structured document understanding has attracted considerable attention and made significant progress recently, owing to its crucial role in intelligent document processing. However, most existing related models can only deal with the document data of specific language(s) (typically English) included in the pre-training collection, which is extremely limited. To address this issue, we propose a simple yet effective Language-independent Layout Transformer (LiLT) for structured document understanding. LiLT can be pre-trained on the structured documents of a single language and then directly fine-tuned on other languages with the corresponding off-the-shelf monolingual/multilingual pre-trained textual models. Experimental results on eight languages have shown that LiLT can achieve competitive or even superior performance on diverse widely-used downstream benchmarks, which enables language-independent benefit from the pre-training of document layout structure. Code and model are publicly available at https://github.com/jpWang/LiLT.

['Kai Ding', 'Lianwen Jin', 'Jiapeng Wang']

2022-02-28

https://aclanthology.org/2022.acl-long.534

https://aclanthology.org/2022.acl-long.534.pdf

acl-2022-5

['document-image-classification', 'semantic-entity-labeling', 'key-information-extraction']

['computer-vision', 'natural-language-processing', 'natural-language-processing']

[ 1.75031349e-01 -2.70447463e-01 -3.45118791e-01 -4.89987612e-01 -9.28101301e-01 -9.90708292e-01 6.82356894e-01 1.71737716e-01 -3.77451986e-01 4.90124166e-01 4.62766111e-01 -7.32056797e-01 1.04685768e-01 -6.14069462e-01 -6.43303633e-01 -2.93055832e-01 3.84420037e-01 6.45514250e-01 1.01380877e-01 -2.34392434e-01 3.33281368e-01 1.99408885e-02 -8.91591132e-01 7.33277023e-01 1.30663860e+00 4.82304424e-01 7.28720725e-01 6.46417975e-01 -6.05445206e-01 6.38462067e-01 -4.29806888e-01 -3.99971724e-01 -1.22546129e-01 -3.63635749e-01 -1.02738965e+00 -2.74361130e-02 2.18036637e-01 -1.50321454e-01 -8.10706019e-02 8.65763426e-01 2.80845195e-01 -1.63320616e-01 4.18163121e-01 -7.91348815e-01 -9.92351949e-01 1.03307426e+00 -6.38611972e-01 1.87059924e-01 4.05839860e-01 -1.34402096e-01 1.16573024e+00 -9.82541740e-01 7.42562175e-01 1.30773067e+00 3.67856860e-01 3.41604412e-01 -9.99821007e-01 -5.43497562e-01 5.90589285e-01 1.69939578e-01 -1.31310248e+00 -4.05395955e-01 6.97535753e-01 -3.13787103e-01 1.16877770e+00 6.89850794e-03 1.64165869e-01 1.05502987e+00 -1.33593142e-01 1.20014083e+00 8.49625111e-01 -7.04628348e-01 -1.97071880e-01 8.51109624e-02 3.35417986e-01 6.82831466e-01 1.37560040e-01 -5.38192987e-01 -2.95613468e-01 3.35317135e-01 3.15688640e-01 -2.13945732e-01 -4.02596742e-01 -2.37476051e-01 -1.23689556e+00 6.65505350e-01 2.35466838e-01 5.26741624e-01 2.52973169e-01 -3.20723921e-01 8.92377913e-01 2.75853693e-01 4.98122782e-01 3.49973291e-01 -7.50565112e-01 -1.80813849e-01 -7.76490510e-01 -5.94149791e-02 7.87303627e-01 1.23430860e+00 5.85092366e-01 -3.39450806e-01 -1.71633437e-02 1.02567720e+00 1.76423922e-01 6.39718652e-01 5.73954821e-01 -3.47087562e-01 1.34497297e+00 7.65809774e-01 -1.65308133e-01 -6.76470816e-01 -3.09062213e-01 -4.63062584e-01 -8.94506454e-01 -7.16310203e-01 4.64995623e-01 9.66870505e-03 -7.50728130e-01 1.40567195e+00 5.29002845e-02 -3.46802264e-01 2.07174435e-01 6.06441259e-01 7.03756988e-01 1.12614942e+00 -1.86372310e-01 -3.71718146e-02 1.31220794e+00 -1.46317577e+00 -5.65386772e-01 -5.76698124e-01 1.16876626e+00 -9.57580328e-01 1.68093061e+00 5.12497723e-01 -7.08170533e-01 -6.57876909e-01 -9.10161793e-01 -5.84497809e-01 -5.71286321e-01 5.66183627e-01 5.33164322e-01 6.92140400e-01 -8.08150589e-01 -4.20512557e-02 -7.93463528e-01 -3.80792648e-01 4.12756920e-01 2.25452147e-02 -3.23765576e-01 -7.03208625e-01 -1.06074417e+00 5.18161297e-01 5.75280130e-01 3.06574851e-01 -5.96235931e-01 -7.77361691e-01 -8.69293034e-01 5.91182075e-02 6.90282464e-01 -4.49705839e-01 1.36051178e+00 -7.70250261e-01 -1.40852535e+00 8.12724710e-01 -4.70086157e-01 -1.20571166e-01 4.82403070e-01 -6.59631729e-01 -4.49271083e-01 -7.75120929e-02 3.05907220e-01 3.88671190e-01 3.79551858e-01 -1.19626772e+00 -5.96944928e-01 -4.25238937e-01 1.12769201e-01 2.31036335e-01 -5.71350694e-01 1.57967597e-01 -1.23351669e+00 -7.16110766e-01 -1.44260049e-01 -8.19219410e-01 -1.80599596e-02 -3.52713317e-01 -7.36188531e-01 -2.66918242e-01 9.50990796e-01 -9.73849297e-01 1.43648231e+00 -2.04839349e+00 4.46838625e-02 6.29160777e-02 -3.66717517e-01 4.88820881e-01 -5.26965797e-01 8.52871120e-01 2.52175242e-01 2.01845080e-01 -2.96768695e-01 -3.46008420e-01 3.54654938e-02 1.27831399e-01 -5.70293188e-01 2.32046545e-02 2.35823300e-02 1.05903292e+00 -8.69931698e-01 -4.99191880e-01 7.32729211e-02 3.66335452e-01 -5.83464622e-01 1.62398353e-01 -6.11487508e-01 6.60442650e-01 -6.41930640e-01 4.75838691e-01 6.63332582e-01 -3.73171270e-01 5.27176559e-01 -1.71527937e-01 -1.81343317e-01 6.56417489e-01 -8.25045228e-01 2.17919230e+00 -9.06991124e-01 5.75355947e-01 -1.08471908e-01 -9.54291403e-01 8.00696671e-01 9.46165994e-02 -1.12148464e-01 -9.78741050e-01 -6.90997839e-02 3.11227322e-01 2.78240778e-02 -4.54689384e-01 5.74761271e-01 2.58928150e-01 -2.76073962e-01 6.73747003e-01 -2.25556523e-01 -1.82234585e-01 7.34045029e-01 4.84686583e-01 8.59372497e-01 1.63797155e-01 1.31226793e-01 -1.70014441e-01 9.52135146e-01 2.11328957e-02 3.48727047e-01 5.33244491e-01 2.96089143e-01 5.08784175e-01 4.45411950e-01 2.96488944e-02 -1.06262565e+00 -7.73951709e-01 -2.43587103e-02 1.34969509e+00 1.53693467e-01 -7.69425929e-01 -8.61316562e-01 -8.55324209e-01 -1.12194933e-01 7.50884235e-01 -2.86140233e-01 2.26830870e-01 -7.83090472e-01 -4.30346638e-01 7.23553121e-01 8.48455608e-01 5.27030528e-01 -9.24009204e-01 5.01503125e-02 2.00235441e-01 -4.21777606e-01 -1.32107091e+00 -7.30946064e-01 -3.65858972e-02 -7.70747304e-01 -7.79183328e-01 -7.60678172e-01 -9.83268082e-01 8.52277040e-01 3.81434023e-01 1.02274609e+00 3.45399141e-01 4.49150950e-02 5.02936244e-02 -5.38180888e-01 -2.72243798e-01 -4.06580091e-01 7.70209610e-01 -2.63530016e-01 -2.08162352e-01 2.92669505e-01 -9.56627131e-02 -3.00804734e-01 2.61201441e-01 -1.06004333e+00 3.41929883e-01 5.61401546e-01 7.80589521e-01 3.69907200e-01 2.27056876e-01 5.55970192e-01 -1.47652268e+00 7.39212990e-01 -1.61064982e-01 -6.79122508e-01 7.86580801e-01 -4.79437172e-01 2.49243915e-01 1.16632414e+00 -1.02751993e-01 -1.61154950e+00 -3.06826800e-01 -3.11808765e-01 2.81634599e-01 -8.38374197e-02 1.09276199e+00 -7.04606354e-01 4.84622926e-01 1.80722430e-01 2.46596739e-01 -5.51202238e-01 -9.84333575e-01 6.05718017e-01 9.19395745e-01 5.85393310e-01 -8.65847647e-01 9.14356291e-01 4.01232392e-01 -6.09576881e-01 -7.36062944e-01 -1.20775652e+00 -6.02465630e-01 -1.12142396e+00 2.76227355e-01 4.60564494e-01 -8.23904872e-01 -1.31430626e-01 6.15099072e-01 -1.31418085e+00 -4.58823889e-01 3.04823399e-01 2.02513814e-01 4.77379486e-02 5.82006812e-01 -7.03391910e-01 -3.94374400e-01 -5.11678517e-01 -1.16199493e+00 1.10089052e+00 1.32806152e-01 -1.41103417e-01 -1.32838714e+00 1.05484672e-01 7.45140254e-01 2.12708384e-01 -4.62383181e-01 1.50471711e+00 -6.46705270e-01 -5.60016751e-01 -1.78228781e-01 -6.26520157e-01 3.95380437e-01 2.91041613e-01 7.25590512e-02 -8.03950012e-01 -4.56205547e-01 -4.85081792e-01 -5.40851951e-01 8.93298149e-01 1.13971382e-01 1.18778718e+00 -1.67091310e-01 -5.12380362e-01 6.37440979e-01 1.38626373e+00 1.67290047e-01 3.97216797e-01 5.23961902e-01 1.06406391e+00 5.93071997e-01 6.81828856e-01 1.71152338e-01 6.43179595e-01 4.39708084e-01 -1.32895216e-01 -1.99732587e-01 -1.27139583e-01 -6.93732381e-01 3.42183292e-01 1.40139246e+00 1.93407208e-01 -5.92114806e-01 -1.21598017e+00 6.12146676e-01 -1.71815073e+00 -4.93552178e-01 -1.75295383e-01 1.86906707e+00 1.03328943e+00 3.31788361e-01 -2.67614454e-01 2.49872152e-02 5.32619715e-01 2.37290695e-01 -5.33319354e-01 -4.15104300e-01 -2.03240246e-01 -4.67558801e-02 3.72629344e-01 5.14130533e-01 -1.06930673e+00 1.34674692e+00 5.46604824e+00 1.06903362e+00 -1.06001163e+00 2.31283959e-02 5.85155189e-01 2.02227637e-01 -5.72508931e-01 1.15035437e-01 -1.14976704e+00 4.87838984e-01 8.19612861e-01 -3.08763772e-01 2.34009683e-01 7.02589273e-01 3.72508466e-01 -5.15553132e-02 -1.12763536e+00 6.26032710e-01 -4.26770980e-03 -1.25759315e+00 3.74977380e-01 -1.02198564e-01 8.03713083e-01 4.21564914e-02 4.78440374e-02 4.00989264e-01 2.14328229e-01 -9.51583147e-01 7.01482117e-01 1.55072972e-01 9.51332748e-01 -8.55871081e-01 6.43315077e-01 6.13176048e-01 -1.38199258e+00 7.08972737e-02 -4.63444501e-01 8.96265954e-02 2.99689412e-01 6.02284908e-01 -7.37850189e-01 8.14560175e-01 5.92313528e-01 1.03613055e+00 -8.22851479e-01 7.32930899e-01 -6.12687647e-01 8.31592083e-01 -2.05400318e-01 1.51856178e-02 5.23109972e-01 -3.23025316e-01 1.12866454e-01 1.51690090e+00 2.40610987e-01 -2.12907016e-01 2.82028854e-01 6.11521423e-01 -2.42535174e-01 4.87472922e-01 -3.85794610e-01 -4.54285234e-01 3.59905422e-01 1.13670135e+00 -8.63417327e-01 -4.25015241e-01 -8.50755632e-01 9.79666829e-01 5.77667117e-01 3.15398157e-01 -5.93887150e-01 -5.72079301e-01 1.43347517e-01 -3.25543359e-02 4.58684504e-01 -4.20131236e-01 -3.71115804e-01 -1.42886245e+00 2.68089145e-01 -1.19446981e+00 5.08894026e-01 -6.91270590e-01 -1.13883376e+00 7.07142115e-01 -1.78112030e-01 -1.07265127e+00 -1.61072478e-01 -8.03926528e-01 -4.29130107e-01 8.83756816e-01 -1.57804942e+00 -1.48110914e+00 -6.44135335e-03 3.54408056e-01 9.31990445e-01 -1.62403971e-01 7.41374195e-01 3.29677045e-01 -9.52606022e-01 6.57046080e-01 6.28095627e-01 4.36036497e-01 8.47995579e-01 -1.18174303e+00 7.64968634e-01 1.06863952e+00 3.78768355e-01 1.03131628e+00 3.42058301e-01 -5.94250858e-01 -1.47046340e+00 -1.15010738e+00 1.16236627e+00 -5.05743444e-01 8.84635508e-01 -7.63154328e-01 -1.21526480e+00 8.11765313e-01 5.00523090e-01 -4.91602212e-01 7.22321808e-01 4.19889569e-01 -6.50830925e-01 -6.17661998e-02 -4.52392250e-01 6.11650407e-01 1.13274121e+00 -7.59442747e-01 -5.69109321e-01 4.28828806e-01 6.64789915e-01 -4.88984823e-01 -6.39391184e-01 2.35185549e-01 4.62843746e-01 -5.89379728e-01 7.98113644e-01 -5.24142385e-01 6.14982188e-01 -3.33543956e-01 9.94677693e-02 -1.31076157e+00 2.40354389e-02 -4.14984524e-01 1.93878457e-01 1.55020320e+00 7.32053280e-01 -5.12849033e-01 5.25728703e-01 3.47533822e-01 -1.41681731e-01 -6.88989997e-01 -3.42703432e-01 -9.15114164e-01 3.11681986e-01 -5.68380177e-01 5.86683035e-01 7.62998223e-01 1.32220730e-01 8.28323781e-01 -1.64340422e-01 1.06673762e-01 2.75882035e-01 5.12081265e-01 9.12272096e-01 -8.99813473e-01 -2.35315546e-01 -3.52706552e-01 3.05106103e-01 -1.81131375e+00 4.14921671e-01 -1.26808238e+00 1.38532355e-01 -1.85672009e+00 4.49730277e-01 -5.89184880e-01 -7.41973370e-02 5.92905760e-01 -2.33091444e-01 -3.94633040e-02 5.70322536e-02 2.47317135e-01 -7.84612179e-01 6.19568825e-01 1.29765427e+00 -3.86523932e-01 -1.70194611e-01 -1.24151088e-01 -8.77077401e-01 6.86273694e-01 7.30014682e-01 -2.94961512e-01 -7.30193138e-01 -1.06332219e+00 2.76826560e-01 -2.72350341e-01 -2.25052550e-01 -6.15524292e-01 2.02011064e-01 -1.45255961e-02 2.20292851e-01 -9.42840874e-01 -1.68105334e-01 -4.65689480e-01 -4.60377663e-01 1.43081740e-01 -5.00160933e-01 2.09839921e-02 4.24031049e-01 3.02984238e-01 -4.15938854e-01 -5.04065514e-01 2.81990111e-01 2.11351272e-02 -9.69286025e-01 1.84654012e-01 -2.35823050e-01 4.47398841e-01 4.91718739e-01 1.30573556e-01 -6.32999003e-01 -2.32617348e-01 -1.18417427e-01 3.35471839e-01 5.54567933e-01 7.85770416e-01 3.44346285e-01 -8.30798030e-01 -6.22954190e-01 1.91704810e-01 3.54791999e-01 1.51960149e-01 1.95382908e-01 6.07201159e-01 -5.92964947e-01 1.15912592e+00 1.81891724e-01 -4.84646261e-01 -1.24269950e+00 6.06375992e-01 -2.49441899e-02 -7.05231845e-01 -6.07694089e-01 8.57523143e-01 5.21409571e-01 -9.42329049e-01 1.33592457e-01 -5.64573050e-01 -4.63997051e-02 -2.73817360e-01 6.45960212e-01 -2.76088286e-02 2.04417408e-01 -6.10307276e-01 -4.35340524e-01 7.05080867e-01 -5.66185176e-01 -1.81795076e-01 1.16612232e+00 -3.58957320e-01 -1.98569879e-01 3.92318487e-01 1.34825146e+00 2.92343587e-01 -8.72097254e-01 -5.56573749e-01 4.92096454e-01 -3.32777590e-01 -4.50144289e-03 -8.55643213e-01 -9.45642471e-01 1.12181246e+00 -1.37673870e-01 -4.07628082e-02 1.08066797e+00 -1.51074911e-02 1.09780121e+00 8.25859427e-01 3.63813877e-01 -1.22480524e+00 1.67841520e-02 1.00961637e+00 8.96172106e-01 -1.20025456e+00 1.95734247e-04 -5.85332036e-01 -6.78105652e-01 9.80221272e-01 5.00111759e-01 4.04026896e-01 5.54921269e-01 2.76377678e-01 3.44938219e-01 6.24549389e-02 -6.71133697e-01 -7.68544376e-02 5.89952230e-01 4.71373796e-01 1.06109428e+00 -1.45835400e-01 -2.14061424e-01 8.28235269e-01 -3.47508490e-01 -8.96438211e-02 1.28463864e-01 9.95091498e-01 -1.14848278e-01 -1.59449947e+00 -1.67587236e-01 3.38484913e-01 -3.28313857e-01 -4.43527788e-01 -3.99759412e-01 8.19452643e-01 -1.49446219e-01 8.74343693e-01 -6.60594255e-02 5.13654277e-02 2.96164781e-01 2.02214476e-02 4.58717674e-01 -8.67419124e-01 -3.21281493e-01 2.18340173e-01 2.09861442e-01 -2.62438715e-01 -2.18602270e-01 -6.26951933e-01 -1.45490408e+00 -1.06047161e-01 -3.13940346e-01 2.33617395e-01 3.89640540e-01 1.13728297e+00 2.81628847e-01 3.93774807e-01 3.99887472e-01 -5.41107178e-01 -1.67598546e-01 -1.07396603e+00 -3.46687049e-01 2.90231526e-01 6.97692484e-02 -2.60168761e-01 -5.45266978e-02 3.87669474e-01]

[11.039701461791992, 8.693772315979004]

974d41b4-b3bd-4325-a2bc-8e854036bcca

a-counterfactual-collaborative-session-based

2301.13364

https://arxiv.org/abs/2301.13364v3

https://arxiv.org/pdf/2301.13364v3.pdf

A Counterfactual Collaborative Session-based Recommender System

Most session-based recommender systems (SBRSs) focus on extracting information from the observed items in the current session of a user to predict a next item, ignoring the causes outside the session (called outer-session causes, OSCs) that influence the user's selection of items. However, these causes widely exist in the real world, and few studies have investigated their role in SBRSs. In this work, we analyze the causalities and correlations of the OSCs in SBRSs from the perspective of causal inference. We find that the OSCs are essentially the confounders in SBRSs, which leads to spurious correlations in the data used to train SBRS models. To address this problem, we propose a novel SBRS framework named COCO-SBRS (COunterfactual COllaborative Session-Based Recommender Systems) to learn the causality between OSCs and user-item interactions in SBRSs. COCO-SBRS first adopts a self-supervised approach to pre-train a recommendation model by designing pseudo-labels of causes for each user's selection of the item in data to guide the training process. Next, COCO-SBRS adopts counterfactual inference to recommend items based on the outputs of the pre-trained recommendation model considering the causalities to alleviate the data sparsity problem. As a result, COCO-SBRS can learn the causalities in data, preventing the model from learning spurious correlations. The experimental results of our extensive experiments conducted on three real-world datasets demonstrate the superiority of our proposed framework over ten representative SBRSs.

['Minghao Yin', 'Xueyan Liu', 'Kunpeng Liu', 'Yan Wang', 'Shoujin Wang', 'Wenzhuo Song']

2023-01-31

['counterfactual-inference']

['miscellaneous']

[ 1.37589753e-01 8.30181409e-03 -6.55919135e-01 -4.96281743e-01 -3.55502740e-02 -3.06808144e-01 3.43310803e-01 -8.79614055e-02 4.91514988e-02 8.28373253e-01 7.04335213e-01 -4.26207691e-01 -5.94639540e-01 -9.14550900e-01 -1.01777184e+00 -4.70636874e-01 -2.64259189e-01 -1.71263572e-02 9.56601202e-02 -3.30790728e-01 3.17840010e-01 -1.16930500e-01 -1.52130914e+00 5.10368526e-01 1.14872348e+00 4.97859836e-01 3.12377065e-01 1.54695958e-01 -4.02844809e-02 7.38155901e-01 -4.17652935e-01 -4.67236608e-01 2.63468951e-01 -6.51058793e-01 -3.80804181e-01 -2.69471407e-01 -1.17942952e-01 -1.98162943e-01 -5.13236761e-01 6.00768209e-01 2.02932358e-02 4.07745600e-01 5.26577532e-01 -1.30226254e+00 -8.45068753e-01 1.00045109e+00 -4.42775786e-01 2.35225409e-01 6.10292852e-01 -1.07631728e-01 1.50540674e+00 -6.43583059e-01 2.39534825e-01 1.46070981e+00 5.25702238e-01 4.60117698e-01 -1.17373991e+00 -1.01210642e+00 7.93578744e-01 2.84840316e-01 -1.14173365e+00 -6.34361356e-02 7.04896748e-01 -2.81348556e-01 3.42221469e-01 6.94580197e-01 6.20374918e-01 1.24485338e+00 1.03417665e-01 9.53258514e-01 1.07941532e+00 -1.09716609e-01 3.70454043e-01 2.98292547e-01 3.54895145e-01 2.41802096e-01 3.81980032e-01 3.66686225e-01 -8.11750472e-01 -5.34678161e-01 7.72668183e-01 4.54537928e-01 -2.47990325e-01 -3.41057070e-02 -9.96105194e-01 8.82690310e-01 6.34115994e-01 1.15605459e-01 -4.29339528e-01 -2.78387249e-01 -4.26157564e-02 2.69035101e-01 3.29141438e-01 3.82854193e-01 -8.26937973e-01 3.68671864e-01 -4.40913111e-01 4.49523747e-01 7.66054928e-01 8.29469681e-01 4.53568190e-01 -4.10640895e-01 -2.70726353e-01 8.55963349e-01 6.37716055e-01 3.87418300e-01 5.15738249e-01 -3.28281730e-01 4.28645015e-01 7.41144061e-01 4.36658740e-01 -1.14516461e+00 -2.29453132e-01 -5.45479178e-01 -6.45886958e-01 -6.66704714e-01 3.79278302e-01 -2.90177941e-01 -5.55553913e-01 1.72281051e+00 4.80579436e-01 8.79516006e-01 -1.47405997e-01 1.18663883e+00 7.91370451e-01 6.38542295e-01 2.61146218e-01 -6.20781958e-01 9.54817295e-01 -5.91370463e-01 -7.71541357e-01 -7.51331821e-02 6.41790926e-01 -5.08178413e-01 1.25192416e+00 4.01996553e-01 -4.81754512e-01 -5.06638587e-01 -8.22089791e-01 5.11054873e-01 -1.90094002e-02 1.66001424e-01 1.05956614e+00 4.34169650e-01 -3.17055136e-01 7.46963680e-01 -3.92070651e-01 3.09287887e-02 2.00999081e-01 4.88563955e-01 2.57707953e-01 -1.05193049e-01 -1.79626691e+00 2.66987234e-01 1.95673868e-01 1.81683376e-01 -8.97235394e-01 -1.17517245e+00 -2.61831582e-01 1.61152154e-01 8.78037751e-01 -5.85976481e-01 1.12011778e+00 -9.72936749e-01 -1.30710065e+00 -6.49031028e-02 -4.57905889e-01 -1.96063906e-01 1.06657259e-01 -3.23256999e-01 -8.24657023e-01 -8.10193241e-01 4.68062684e-02 -2.33393282e-01 6.18356645e-01 -1.34186184e+00 -8.33543241e-01 -2.72064358e-01 3.15050364e-01 2.69796073e-01 -1.92591295e-01 -1.71433628e-01 -3.39504182e-01 -7.43030369e-01 1.01559646e-02 -1.06899107e+00 -4.42366987e-01 -7.86287725e-01 -4.89696056e-01 -5.49213886e-01 4.10658151e-01 -2.02319145e-01 1.68990719e+00 -2.05957079e+00 -2.00437933e-01 4.20358151e-01 3.40096676e-03 1.54621959e-01 -2.06448674e-01 4.36514527e-01 -2.18229041e-01 1.84692428e-01 2.72709191e-01 3.57530899e-02 -2.86962628e-01 3.63969803e-01 -6.74960613e-01 4.69917804e-02 -3.35897624e-01 6.27794027e-01 -1.12013948e+00 -1.09343357e-01 1.02402359e-01 2.39054579e-02 -8.47866714e-01 6.85671091e-01 -3.72537106e-01 5.72671235e-01 -7.71145225e-01 5.38865896e-03 6.33173764e-01 -2.79441833e-01 6.59244180e-01 -1.35851547e-01 -1.09675221e-01 9.01249349e-01 -1.34838748e+00 1.31558764e+00 -5.83753467e-01 -1.77618042e-01 -6.10950649e-01 -7.66043067e-01 7.65580237e-01 3.72103244e-01 5.59430778e-01 -6.02789044e-01 -1.90804705e-01 5.35049029e-02 2.83521742e-01 -6.36467934e-01 1.26098305e-01 -1.79263249e-01 1.41628757e-01 5.65801978e-01 -1.61416918e-01 7.65649617e-01 9.00039226e-02 5.83521843e-01 9.35810208e-01 9.57261473e-02 3.23554814e-01 -1.05812095e-01 6.54141963e-01 -1.76731661e-01 1.07465851e+00 1.01220870e+00 3.93817216e-01 2.21553922e-01 3.81009519e-01 -3.62233996e-01 -5.33383906e-01 -1.08418691e+00 -3.70972306e-02 1.36532438e+00 3.61911029e-01 -6.07342303e-01 -2.40161106e-01 -1.14700460e+00 1.83820099e-01 1.11019969e+00 -9.06726778e-01 -2.40244254e-01 -2.17268839e-01 -1.09138596e+00 -1.64851516e-01 1.91152364e-01 7.67450407e-02 -9.64389265e-01 3.39803457e-01 2.09736109e-01 -2.86970913e-01 -6.06857419e-01 -6.02763176e-01 -2.24649921e-01 -8.53845656e-01 -1.39402437e+00 -1.29289135e-01 -1.49394214e-01 6.72701597e-01 7.72624969e-01 9.29556668e-01 1.88565239e-01 3.32220703e-01 -8.28222856e-02 -4.67681587e-01 -3.41728330e-01 -1.40448377e-01 -1.97239712e-01 3.43609959e-01 4.13168460e-01 5.78421414e-01 -8.02719891e-01 -9.85235035e-01 8.58409286e-01 -5.35188615e-01 5.50457574e-02 6.00565195e-01 6.19260132e-01 3.18048865e-01 1.43045858e-01 1.16026819e+00 -1.63287389e+00 6.50909901e-01 -1.26817942e+00 -4.05479610e-01 3.28451335e-01 -1.04838192e+00 -1.37528613e-01 1.01376748e+00 -7.23920524e-01 -1.59429467e+00 -3.04628521e-01 6.24355562e-02 -1.66029006e-01 -1.99699551e-01 7.31186509e-01 -2.90827304e-01 6.27044201e-01 5.81570685e-01 4.73200753e-02 -5.26463926e-01 -7.30740190e-01 2.77756214e-01 7.01814830e-01 6.62704706e-02 -4.29304391e-01 6.81973815e-01 2.99104661e-01 -3.31120461e-01 -2.33919054e-01 -1.59593916e+00 -7.79025376e-01 -1.99779600e-01 -7.93422610e-02 2.34432966e-01 -1.07831275e+00 -9.67713952e-01 -1.03907049e-01 -7.53774405e-01 -7.19636604e-02 -8.49723667e-02 8.93476367e-01 -1.76897079e-01 -1.09680705e-01 -2.71385133e-01 -9.97453392e-01 -1.59569159e-02 -6.50792360e-01 4.54232752e-01 3.83384645e-01 -2.72910923e-01 -1.02446306e+00 2.67253369e-01 5.19152105e-01 -1.39381111e-01 -4.54853415e-01 7.85520732e-01 -8.65874588e-01 -5.64616859e-01 7.41833868e-03 -4.60270755e-02 -1.29486501e-01 5.97659945e-01 -2.01980069e-01 -6.67151988e-01 -5.66265136e-02 -1.42662153e-01 2.76443958e-01 4.99834269e-01 5.01952827e-01 1.28483403e+00 -7.31838286e-01 -4.77643192e-01 2.02918231e-01 1.26011884e+00 4.74573672e-01 4.59502786e-01 -1.26386315e-01 7.74961531e-01 5.09748340e-01 1.01259494e+00 5.61455369e-01 5.25780916e-01 6.54402912e-01 2.66681165e-01 -6.41187504e-02 1.61254033e-01 -9.20401335e-01 4.19165164e-01 8.45108688e-01 -2.48686254e-01 -2.17247918e-01 -1.73247889e-01 4.65772957e-01 -2.27897787e+00 -8.00313234e-01 -8.63549769e-01 2.49169874e+00 8.66834641e-01 -1.57513559e-01 2.45018706e-01 -1.73339784e-01 7.29783654e-01 -7.91115388e-02 -7.22895980e-01 -2.58960873e-01 2.30594173e-01 -1.83482885e-01 4.08550143e-01 1.62543908e-01 -7.00340092e-01 5.88203549e-01 5.62882948e+00 6.52838469e-01 -6.37284458e-01 1.36722863e-01 3.57209444e-01 -3.12966824e-01 -6.86892986e-01 2.62633234e-01 -9.01171029e-01 7.78410256e-01 8.68218958e-01 -2.48182997e-01 6.53532565e-01 6.19368076e-01 1.05408001e+00 -4.50759158e-02 -1.11787093e+00 5.04139185e-01 -2.50955492e-01 -1.16140306e+00 9.66326445e-02 9.57140177e-02 1.17433739e+00 -1.71607584e-01 -6.92158118e-02 3.98263037e-01 8.19589972e-01 -6.95144951e-01 3.32037181e-01 5.73312461e-01 1.43246382e-01 -7.01065421e-01 6.46812320e-01 6.48939908e-01 -7.79990196e-01 -4.50102270e-01 -5.07883728e-01 -3.66538048e-01 -1.90186679e-01 9.96252596e-01 -9.03952956e-01 8.32828939e-01 6.55750573e-01 9.26626325e-01 -2.26595625e-01 8.64336908e-01 -6.44915581e-01 1.37669277e+00 -6.36623651e-02 -1.28823668e-01 -1.74736887e-01 -3.14428836e-01 4.14574087e-01 8.01621377e-01 2.21785009e-01 4.86374825e-01 1.11078106e-01 8.67066264e-01 -1.66397437e-01 4.31407362e-01 -3.71508181e-01 3.47620636e-01 6.10697210e-01 1.02819753e+00 -1.76971242e-01 -2.23774359e-01 -5.90154171e-01 3.33492696e-01 1.27766848e-01 4.87003773e-01 -8.76705647e-01 2.53099442e-01 6.84247971e-01 3.50043237e-01 1.37865528e-01 3.17455202e-01 -3.51852834e-01 -1.50987923e+00 -3.27475101e-01 -9.72704887e-01 6.76563859e-01 -6.00542247e-01 -1.62180603e+00 -9.89936516e-02 -4.61144857e-02 -1.18395650e+00 1.45123810e-01 6.47889525e-02 -9.77730393e-01 8.60099196e-01 -1.22492397e+00 -7.71765113e-01 -2.40052305e-02 7.26415694e-01 4.95711595e-01 3.33420336e-02 6.60963774e-01 2.64213979e-01 -8.56941640e-01 5.09405851e-01 1.31491944e-01 -2.39124060e-01 8.20800781e-01 -1.00743973e+00 3.73111404e-02 7.15750933e-01 2.40259781e-01 1.39325988e+00 7.63684988e-01 -1.04497182e+00 -1.30396020e+00 -1.14281702e+00 7.36412883e-01 -2.81808108e-01 7.22302556e-01 -3.27343643e-01 -1.01445687e+00 7.15016901e-01 -2.71733403e-01 -2.59071529e-01 1.19954300e+00 1.26501572e+00 -4.08102274e-01 -2.50230670e-01 -7.47604668e-01 8.13561618e-01 1.39488661e+00 -1.08280517e-01 -8.80311787e-01 4.00508285e-01 7.81283975e-01 -2.95173936e-02 -7.96486139e-01 2.40612194e-01 7.17428982e-01 -8.37469161e-01 8.96315455e-01 -1.00007665e+00 7.00126112e-01 -4.21486378e-01 2.55185336e-01 -1.65792441e+00 -5.21974981e-01 -4.97970492e-01 -2.71892607e-01 1.34511220e+00 4.86642599e-01 -4.97245401e-01 7.40875244e-01 8.42012882e-01 8.59442055e-02 -6.50693953e-01 -3.35352182e-01 -5.19410610e-01 -3.98017973e-01 -4.50339079e-01 1.05903792e+00 1.13963556e+00 1.09893478e-01 5.45432031e-01 -7.27793157e-01 4.34172034e-01 7.26159930e-01 5.45114219e-01 9.25248802e-01 -1.17719638e+00 -6.99443042e-01 2.30331078e-01 2.93579310e-01 -1.16939998e+00 -1.10034943e-01 -7.99880683e-01 -1.25331566e-01 -1.25075746e+00 4.61359084e-01 -9.39043343e-01 -7.58414388e-01 3.98091346e-01 -7.73154616e-01 -2.93138742e-01 -7.70298345e-03 4.13871348e-01 -6.24147415e-01 4.85725641e-01 1.44812083e+00 3.64334524e-01 -7.42890179e-01 6.29420996e-01 -1.07660949e+00 8.27055156e-01 5.98284364e-01 -8.01751196e-01 -8.71631265e-01 9.13657174e-02 4.84442860e-01 7.37131909e-02 1.44225433e-01 -2.54407257e-01 2.42952257e-02 -8.38715732e-01 2.12232023e-01 -4.11751986e-01 -2.48443648e-01 -7.85359085e-01 4.02432531e-01 1.25823945e-01 -7.34995425e-01 -7.16262460e-01 -2.63318628e-01 9.94895995e-01 1.35155201e-01 -1.43058687e-01 -1.07678231e-02 9.02601983e-03 -3.73176485e-01 1.92591235e-01 -2.97759771e-01 -1.22237988e-01 7.43457496e-01 2.47501016e-01 -1.97806746e-01 -3.11782151e-01 -7.65148640e-01 3.88804823e-01 -2.36989945e-01 6.92949772e-01 4.69160646e-01 -1.22134566e+00 -4.56567347e-01 1.04798116e-01 6.81409389e-02 -1.05657950e-01 6.26683414e-01 7.39015698e-01 6.51299000e-01 6.09779716e-01 3.25523525e-01 -1.62831023e-01 -1.07323337e+00 7.30676591e-01 -5.01180068e-02 -4.64243561e-01 -2.21793354e-01 5.86718440e-01 8.00623238e-01 -4.93351966e-01 3.00606317e-03 -4.24942821e-02 -5.05401075e-01 -7.36753121e-02 4.78877276e-01 4.09416050e-01 -7.85148144e-02 -5.94023094e-02 -2.06072867e-01 -1.78619415e-01 -3.20428729e-01 3.20910186e-01 1.50885153e+00 -3.73736441e-01 3.04177236e-02 6.79101408e-01 9.00677979e-01 3.56923908e-01 -1.13081336e+00 -3.63204598e-01 -1.01083837e-01 -7.14887261e-01 5.45998402e-02 -1.21773911e+00 -1.02225733e+00 1.77050605e-01 1.80871755e-01 1.87523961e-01 1.00307763e+00 -1.22615427e-01 6.74951017e-01 7.20436051e-02 4.75235671e-01 -8.38660479e-01 -1.94750667e-01 1.33509994e-01 8.34381878e-01 -1.09230900e+00 7.41136745e-02 -8.16543221e-01 -7.25471377e-01 5.54707587e-01 6.95856750e-01 -2.90852994e-01 8.90818477e-01 -2.32420921e-01 -2.00744137e-01 4.06553820e-02 -1.13115633e+00 -8.32523704e-02 5.84834993e-01 1.40818402e-01 5.18023431e-01 3.95190388e-01 -8.03934038e-01 1.40554118e+00 -8.22333544e-02 3.96834940e-01 4.66478080e-01 3.26837301e-01 1.14426136e-01 -1.26998174e+00 -2.88000584e-01 9.20147896e-01 -4.51371968e-01 -1.81551203e-01 -2.84661889e-01 3.81587803e-01 3.71868700e-01 1.35539353e+00 -2.43031442e-01 -7.25745320e-01 5.99920034e-01 -2.27330983e-01 5.94363548e-02 -7.81764984e-01 -6.76602304e-01 1.37204722e-01 1.42739236e-01 -6.69903100e-01 -4.81777161e-01 -8.06399941e-01 -1.24927008e+00 -3.29706341e-01 -7.20974028e-01 6.76097393e-01 4.94167745e-01 1.19381988e+00 4.38952476e-01 6.64719284e-01 1.40199256e+00 -1.23175435e-01 -4.01004076e-01 -8.97172093e-01 -7.14818537e-01 8.00690055e-01 -2.44537368e-02 -8.77086163e-01 -6.21482611e-01 -3.06327119e-02]

[9.861831665039062, 5.5662617683410645]

d0139ecc-a2d6-4a2d-86ad-ec52ab31e596

audio-attacks-and-defenses-against-aed

2106.07428

https://arxiv.org/abs/2106.07428v4

https://arxiv.org/pdf/2106.07428v4.pdf

Audio Attacks and Defenses against AED Systems -- A Practical Study

In this paper, we evaluate deep learning-enabled AED systems against evasion attacks based on adversarial examples. We test the robustness of multiple security critical AED tasks, implemented as CNNs classifiers, as well as existing third-party Nest devices, manufactured by Google, which run their own black-box deep learning models. Our adversarial examples use audio perturbations made of white and background noises. Such disturbances are easy to create, to perform and to reproduce, and can be accessible to a large number of potential attackers, even non-technically savvy ones. We show that an adversary can focus on audio adversarial inputs to cause AED systems to misclassify, achieving high success rates, even when we use small levels of a given type of noisy disturbance. For instance, on the case of the gunshot sound class, we achieve nearly 100% success rate when employing as little as 0.05 white noise level. Similarly to what has been previously done by works focusing on adversarial examples from the image domain as well as on the speech recognition domain. We then, seek to improve classifiers' robustness through countermeasures. We employ adversarial training and audio denoising. We show that these countermeasures, when applied to audio input, can be successful, either in isolation or in combination, generating relevant increases of nearly fifty percent in the performance of the classifiers when these are under attack.

['Shirin Nilizadeh', 'Rodrigo dos Santos']

2021-06-14

['audio-denoising']

['audio']

[ 4.24646109e-01 1.87195778e-01 6.45101249e-01 3.03264648e-01 -8.06596398e-01 -1.06880641e+00 7.70652115e-01 -5.68444915e-02 -5.85447013e-01 7.50930369e-01 -2.48266295e-01 -5.71172118e-01 5.30235022e-02 -8.72338414e-01 -8.94399345e-01 -9.30042565e-01 -4.68135178e-01 -3.91689464e-02 1.51053369e-01 -3.86587888e-01 -1.65471971e-01 7.39450276e-01 -1.55985975e+00 8.55757296e-02 2.73501009e-01 1.03025758e+00 -6.21705055e-01 1.06295288e+00 6.26789868e-01 5.93641579e-01 -1.63487697e+00 -5.83017468e-01 4.77374494e-01 8.28155279e-02 -3.81216913e-01 -5.21676064e-01 3.63727450e-01 -4.81703609e-01 -4.17432398e-01 1.15639973e+00 1.02602959e+00 -3.69953692e-01 5.81134319e-01 -1.38836515e+00 -1.59877032e-01 4.13026839e-01 -1.85322985e-02 2.36691475e-01 5.76256573e-01 7.55175352e-01 3.25245202e-01 -2.38104001e-01 1.27227694e-01 1.05984688e+00 7.21022427e-01 9.88270998e-01 -1.29089475e+00 -1.14475870e+00 -2.13275909e-01 -3.92929345e-01 -1.12208343e+00 -4.93888527e-01 6.62845612e-01 -3.69242817e-01 7.33796895e-01 3.62133026e-01 3.46208304e-01 2.07368708e+00 3.67762655e-01 2.91282296e-01 1.03922987e+00 -2.33960509e-01 4.73878384e-01 -1.91952903e-02 -4.70790595e-01 1.24979444e-01 3.87345523e-01 6.94165051e-01 -2.81529516e-01 -5.51148772e-01 3.07312936e-01 -2.41846710e-01 -5.19964516e-01 2.89806485e-01 -8.36806595e-01 7.18727052e-01 2.87980318e-01 3.78526479e-01 -2.94261605e-01 4.58275169e-01 5.06147206e-01 7.29450524e-01 2.68466026e-01 9.21028554e-01 -4.25783187e-01 -2.00400785e-01 -4.21038866e-01 3.48031312e-01 9.77330327e-01 5.37595034e-01 2.75160670e-02 6.39595568e-01 -5.34646325e-02 3.01603943e-01 -1.14096858e-01 9.25267458e-01 3.49479020e-01 -4.45386767e-01 4.68201667e-01 -8.96047205e-02 1.66711256e-01 -9.86816108e-01 -3.75769824e-01 -3.70126396e-01 -7.77334273e-01 7.43284047e-01 4.84183401e-01 -7.16567814e-01 -1.01777434e+00 1.87778950e+00 1.47176743e-01 4.86745775e-01 3.65552038e-01 4.73219723e-01 5.86222947e-01 3.42014253e-01 1.42906442e-01 2.97734533e-02 1.15195310e+00 -6.03560321e-02 -5.87233961e-01 -1.25594169e-01 4.04808849e-01 -7.27657616e-01 1.08673811e+00 7.78609216e-01 -9.11965489e-01 -3.96977842e-01 -1.47103703e+00 8.52364480e-01 -8.08113396e-01 -5.39834082e-01 2.32102916e-01 1.45066333e+00 -8.38861763e-01 6.56269729e-01 -8.61615300e-01 2.40714729e-01 4.49584812e-01 6.62585676e-01 -3.97209525e-01 2.79146910e-01 -1.67675292e+00 9.44125175e-01 -6.40745088e-02 -1.51907310e-01 -1.77080226e+00 -7.26172030e-01 -5.53318977e-01 -2.48808451e-02 9.76936594e-02 -3.74959826e-01 1.22366273e+00 -9.57176507e-01 -1.26971459e+00 6.25117123e-01 5.77957213e-01 -9.48623419e-01 7.41940439e-01 -2.57361650e-01 -8.22140813e-01 3.52582544e-01 -3.86609286e-01 1.84828028e-01 1.37059069e+00 -1.14377213e+00 -2.25657642e-01 -1.38114214e-01 4.89087343e-01 -3.90604317e-01 -7.04647362e-01 1.82105735e-01 4.05432612e-01 -1.02373612e+00 -6.99973047e-01 -1.00463367e+00 -1.64600521e-01 -1.89701378e-01 -5.06238222e-01 4.52681482e-01 1.18983400e+00 -4.07285213e-01 9.27589834e-01 -2.26717949e+00 -1.06135882e-01 3.73955697e-01 2.05563471e-01 7.42554247e-01 -1.38540879e-01 2.38340899e-01 -2.51497418e-01 5.25254190e-01 -4.55614090e-01 -1.24322675e-01 1.45577148e-01 -2.77861431e-02 -7.41689026e-01 4.28519607e-01 3.59890372e-01 4.62502837e-01 -6.08342767e-01 3.24436843e-01 1.44714490e-01 6.19823158e-01 -4.31047082e-01 3.45582694e-01 5.47419637e-02 3.75006378e-01 -4.39805150e-01 5.47954559e-01 4.21376944e-01 5.96630752e-01 -3.80577445e-01 7.99289569e-02 4.10385668e-01 2.55808532e-01 -1.02247560e+00 7.30546653e-01 -7.03956068e-01 6.58028245e-01 2.95679003e-01 -8.86808753e-01 8.23038101e-01 6.64916217e-01 1.50099084e-01 -4.27978188e-01 3.83298784e-01 2.78085232e-01 2.95496434e-01 -2.27641404e-01 -1.06986456e-01 -2.49959648e-01 -4.19479758e-01 5.37713826e-01 1.40866116e-01 -5.77048659e-01 -4.77316767e-01 4.11103554e-02 1.77444351e+00 -6.74124062e-01 -4.14888114e-02 -3.81363668e-02 4.66607004e-01 -4.54847336e-01 7.09513202e-02 9.93142068e-01 -1.74850941e-01 4.16325212e-01 4.99702126e-01 -3.44722271e-01 -9.84576881e-01 -1.19882655e+00 -1.51550904e-01 5.06901979e-01 -2.81935260e-02 -1.78893477e-01 -1.07860744e+00 -8.27195525e-01 5.38507067e-02 6.83383524e-01 -6.49493635e-01 -8.38946402e-01 -4.27068949e-01 -7.36120999e-01 1.65740597e+00 4.39425051e-01 5.88633478e-01 -1.20401442e+00 -7.08346367e-01 1.06160574e-01 5.27491629e-01 -1.20512986e+00 1.06587045e-01 6.36316597e-01 -2.45025247e-01 -1.10699511e+00 -5.04034758e-01 -3.73031557e-01 3.62027645e-01 -3.39345485e-01 8.14164817e-01 2.37473488e-01 -4.89750892e-01 7.56379545e-01 -2.54809082e-01 -1.06017411e+00 -8.78726780e-01 -2.27760077e-01 6.58648014e-01 -1.10659622e-01 2.95005180e-02 -8.15638602e-01 -2.93641508e-01 4.32279915e-01 -1.17167246e+00 -9.83020365e-01 1.51691779e-01 7.90109932e-01 3.81636322e-02 4.32947278e-01 7.88023114e-01 -5.68895280e-01 9.37073410e-01 -4.69756216e-01 -5.41289628e-01 -5.86665161e-02 -2.17915196e-02 -1.37552425e-01 1.15711820e+00 -1.02105689e+00 -4.77072805e-01 -1.91209242e-01 -6.22071266e-01 -6.74956858e-01 -5.94687462e-01 -6.70638084e-02 -3.74411106e-01 -6.47067547e-01 1.06176853e+00 5.55051200e-04 -2.73247153e-01 -1.62153184e-01 8.02249182e-03 6.88785255e-01 5.69320381e-01 -5.66875160e-01 1.52781069e+00 3.53719592e-01 3.30725014e-02 -9.35205758e-01 -3.79825920e-01 3.95458102e-01 -3.83388028e-02 -3.19028705e-01 7.49647081e-01 -6.68976247e-01 -8.70572209e-01 9.43125069e-01 -1.00982428e+00 -4.89460379e-01 -8.06190595e-02 2.86835551e-01 -3.78914982e-01 -6.20714203e-02 -4.34074819e-01 -8.87141347e-01 -2.87408233e-01 -1.29742563e+00 8.68555903e-01 -3.90523225e-02 -1.75511986e-01 -8.61451089e-01 -1.82230905e-01 -4.18853834e-02 6.89471662e-01 7.93185532e-01 7.55158663e-01 -1.14540553e+00 -9.46701393e-02 -6.29569471e-01 5.93440890e-01 7.32069314e-01 7.35975653e-02 7.65140578e-02 -1.47244084e+00 -6.22161925e-01 4.35201794e-01 -5.65646529e-01 5.05962491e-01 -5.30977845e-02 1.10772264e+00 -5.25581658e-01 -9.10458192e-02 6.07107878e-01 9.17413414e-01 5.84456265e-01 6.93467140e-01 3.28002781e-01 4.35456812e-01 4.56951290e-01 3.14328551e-01 1.90378949e-01 -6.38691545e-01 6.85565054e-01 9.38579261e-01 -2.43784383e-01 3.47135872e-01 -1.31789401e-01 5.03514886e-01 4.64338660e-02 4.86321747e-02 -4.08700496e-01 -9.42302465e-01 2.28055879e-01 -1.04575145e+00 -1.01689363e+00 2.80498236e-01 2.40264511e+00 6.35808289e-01 7.22323060e-01 1.87062263e-01 8.64468038e-01 6.29247367e-01 7.59331882e-02 -4.79083031e-01 -5.72982490e-01 -2.50407785e-01 8.65419567e-01 6.19753599e-01 3.06780696e-01 -1.48504817e+00 7.47909009e-01 6.48133087e+00 7.12985337e-01 -1.32723606e+00 8.00767913e-02 4.79330540e-01 -3.81676972e-01 -1.18564852e-01 -5.35717964e-01 -3.51068914e-01 7.54404962e-01 1.30442691e+00 -1.54475588e-02 4.25101250e-01 7.19607592e-01 -1.01831704e-01 3.40777040e-01 -9.31506574e-01 8.82473171e-01 -1.87491551e-02 -1.06761837e+00 -1.02330960e-01 3.74621414e-02 3.30524474e-01 -2.57403910e-01 4.89687234e-01 3.12293679e-01 5.71757317e-01 -1.53966725e+00 6.01191401e-01 2.49614660e-02 6.71538830e-01 -1.16907108e+00 8.48525822e-01 2.83916593e-01 -7.33341336e-01 -2.59160936e-01 2.13171449e-02 -5.35396524e-02 -8.39405209e-02 2.67920762e-01 -6.03013396e-01 1.90810099e-01 9.39372361e-01 -1.07143007e-01 -4.30486172e-01 6.37574315e-01 -5.57224512e-01 1.09327662e+00 -5.25220037e-01 -1.23702720e-01 2.86114991e-01 5.37408113e-01 7.40098357e-01 9.91674602e-01 2.33329996e-01 1.01402543e-01 -2.65341073e-01 4.41482216e-01 -1.63054377e-01 -4.87342119e-01 -1.23566377e+00 7.20541459e-03 6.21522188e-01 9.37569082e-01 -2.35197857e-01 5.11709414e-02 -1.77499838e-02 7.28350401e-01 -2.58163184e-01 4.14140791e-01 -1.01311362e+00 -8.23825181e-01 9.83035624e-01 -2.27891486e-02 1.89814270e-01 -4.49407101e-02 5.70727699e-02 -7.99862981e-01 4.76812348e-02 -1.41715217e+00 1.37581587e-01 -4.66786951e-01 -1.23087263e+00 8.69844258e-01 -1.40946880e-01 -1.36746407e+00 -3.16658914e-01 -7.43468285e-01 -1.01968420e+00 6.66376114e-01 -9.48059201e-01 -8.78654063e-01 8.61247331e-02 9.41156983e-01 2.71229167e-02 -6.28017187e-01 1.06591880e+00 3.97156417e-01 -3.31430823e-01 1.03132474e+00 -2.01524511e-01 4.51356679e-01 4.36243087e-01 -9.49460268e-01 6.71868980e-01 9.68284190e-01 2.69907951e-01 5.24994850e-01 9.96118188e-01 -2.51903236e-01 -1.20890427e+00 -1.12760639e+00 5.40179200e-02 -5.54302990e-01 8.12924087e-01 -8.50081801e-01 -8.25897336e-01 3.78516376e-01 -3.52535695e-02 7.32780376e-04 5.06306112e-01 -2.76867062e-01 -4.63407576e-01 -4.10215482e-02 -1.58873582e+00 8.29479575e-01 7.63303697e-01 -8.47827137e-01 -7.15382397e-01 3.72465521e-01 9.72830832e-01 -2.97285795e-01 -7.87256420e-01 5.80298603e-01 3.87803644e-01 -8.42204750e-01 1.24546564e+00 -8.23566735e-01 1.76665932e-01 -2.85354346e-01 -1.16036035e-01 -1.52033234e+00 4.78481084e-01 -9.93130684e-01 -8.48495886e-02 1.35310233e+00 4.82522160e-01 -9.71766353e-01 4.66911197e-01 2.04316899e-01 -1.11132115e-02 -4.87660915e-01 -1.39021742e+00 -1.08459795e+00 2.28649408e-01 -8.45792949e-01 7.79237509e-01 8.28920960e-01 -2.72284657e-01 -1.12234928e-01 -4.42521393e-01 6.94707096e-01 5.00189662e-01 -7.54082263e-01 8.69610608e-01 -7.61442959e-01 -4.89269733e-01 -3.65847021e-01 -8.93472850e-01 -1.15612127e-01 1.80331826e-01 -3.16700041e-01 -5.19128852e-02 -4.46309566e-01 -7.61993945e-01 -3.68685603e-01 -2.34594777e-01 6.41424179e-01 -9.97079760e-02 6.66988254e-01 1.60878107e-01 -2.99690366e-01 6.83190748e-02 2.16206878e-01 6.30103827e-01 -5.64090848e-01 4.67382446e-02 3.94278526e-01 -5.07698357e-01 9.69920814e-01 9.27479327e-01 -8.70906830e-01 -3.12188327e-01 -4.33553830e-02 6.45829216e-02 -1.75992757e-01 7.36737192e-01 -1.50471842e+00 -4.02082466e-02 2.93025792e-01 3.11294347e-01 1.98365733e-01 5.40014565e-01 -1.06215751e+00 1.56881034e-01 8.62457097e-01 -2.41353914e-01 3.73346247e-02 7.53864646e-01 4.84630376e-01 -2.32424766e-01 -6.01550788e-02 9.12941933e-01 1.50952160e-01 4.61807102e-02 1.27254725e-01 -6.64970100e-01 8.29008296e-02 1.28042603e+00 -2.25573890e-02 -6.58669472e-01 -7.90810168e-01 -7.07729876e-01 -9.01106820e-02 3.01133960e-01 4.30729300e-01 6.25203431e-01 -1.05957425e+00 -4.50047851e-01 4.44661319e-01 -1.04200959e-01 -3.65509242e-01 5.01881838e-02 4.48490888e-01 -3.30975711e-01 -6.82074158e-03 -3.09454948e-01 -3.96048069e-01 -1.46761191e+00 7.71061480e-01 7.49714553e-01 7.04586655e-02 -2.37727389e-01 7.35929370e-01 1.38805270e-01 -2.47918591e-01 6.28567934e-01 -7.54600689e-02 7.38646835e-03 -8.29786658e-02 6.28699958e-01 1.27520403e-02 4.02442604e-01 -1.92626566e-01 -5.23340523e-01 4.52805459e-01 2.37498611e-01 -1.69515669e-01 1.08599591e+00 6.97513282e-01 4.32663590e-01 1.18540814e-02 1.05701578e+00 2.80392259e-01 -9.95432377e-01 4.04717028e-01 -3.08655173e-01 -1.59335539e-01 -1.71248391e-01 -8.78265083e-01 -1.08507669e+00 1.12124681e+00 8.82270634e-01 8.25231135e-01 1.26711845e+00 -3.12088221e-01 5.34474850e-01 5.15346348e-01 4.85811710e-01 -5.10556698e-01 3.24339837e-01 2.01321736e-01 8.34423482e-01 -8.32048893e-01 -3.35559845e-01 1.35514855e-01 -4.92281228e-01 7.86676466e-01 4.15882170e-01 -3.67632240e-01 6.93198204e-01 9.94831622e-01 2.61925399e-01 -4.42972034e-02 -6.78702474e-01 1.60813957e-01 1.43478364e-02 1.10643220e+00 -2.37818033e-01 -5.46158180e-02 1.97704330e-01 5.56122601e-01 -4.70751196e-01 -6.93234563e-01 6.10752225e-01 9.76644337e-01 -1.20956801e-01 -9.68087435e-01 -8.27982605e-01 3.51383358e-01 -9.46873248e-01 -7.34242722e-02 -7.63104260e-01 1.02068889e+00 1.56309694e-01 1.20156968e+00 -1.12346768e-01 -8.14331055e-01 5.13328791e-01 -7.86606744e-02 8.76964778e-02 -2.56698161e-01 -1.21453857e+00 -3.49564791e-01 2.61978805e-01 -4.54687655e-01 -1.93085611e-01 -3.36091936e-01 -7.42910206e-01 -3.29508930e-01 -1.66324243e-01 3.15066762e-02 6.43539250e-01 7.72566438e-01 4.06979807e-02 7.32490838e-01 9.51950848e-01 -9.83883440e-01 -7.77463436e-01 -8.18995237e-01 -4.50347006e-01 4.01398182e-01 6.28811955e-01 -7.67342031e-01 -1.20837975e+00 -3.21550697e-01]

[5.5536675453186035, 7.766383647918701]

e8ecae42-72a4-4a0c-a691-571197bffb94

dual-temporal-memory-network-for-efficient

2003.06125

https://arxiv.org/abs/2003.06125v1

https://arxiv.org/pdf/2003.06125v1.pdf

Dual Temporal Memory Network for Efficient Video Object Segmentation

Video Object Segmentation (VOS) is typically formulated in a semi-supervised setting. Given the ground-truth segmentation mask on the first frame, the task of VOS is to track and segment the single or multiple objects of interests in the rest frames of the video at the pixel level. One of the fundamental challenges in VOS is how to make the most use of the temporal information to boost the performance. We present an end-to-end network which stores short- and long-term video sequence information preceding the current frame as the temporal memories to address the temporal modeling in VOS. Our network consists of two temporal sub-networks including a short-term memory sub-network and a long-term memory sub-network. The short-term memory sub-network models the fine-grained spatial-temporal interactions between local regions across neighboring frames in video via a graph-based learning framework, which can well preserve the visual consistency of local regions over time. The long-term memory sub-network models the long-range evolution of object via a Simplified-Gated Recurrent Unit (S-GRU), making the segmentation be robust against occlusions and drift errors. In our experiments, we show that our proposed method achieves a favorable and competitive performance on three frequently-used VOS datasets, including DAVIS 2016, DAVIS 2017 and Youtube-VOS in terms of both speed and accuracy.

['Kaihua Zhang', 'Bo Liu', 'Zhu Li', 'Qingshan Liu', 'Long Wang', 'Dong Liu']

2020-03-13

['one-shot-visual-object-segmentation']

['computer-vision']

[ 3.21140047e-03 -1.74490273e-01 -5.01496494e-01 -3.58719617e-01 -4.04293120e-01 -2.16608554e-01 6.63518757e-02 -1.96308643e-01 -4.90711123e-01 3.01959187e-01 -8.65907595e-02 7.95081630e-02 1.93797022e-01 -6.77669406e-01 -1.02627134e+00 -6.23596668e-01 -2.60417819e-01 1.27739638e-01 1.07728910e+00 1.27237573e-01 4.36388887e-02 3.12008202e-01 -1.36641788e+00 1.93671823e-01 7.47484148e-01 1.35557210e+00 5.85427999e-01 7.46566176e-01 -1.23619027e-01 9.97436106e-01 -2.66780853e-01 7.89011791e-02 2.58842558e-01 -3.03442031e-01 -9.27402377e-01 4.20859516e-01 4.51827556e-01 -3.28854769e-01 -8.97378623e-01 9.74449575e-01 2.73244768e-01 4.91328478e-01 2.89946288e-01 -1.23383415e+00 -4.80891049e-01 3.90907019e-01 -8.73683572e-01 5.58819532e-01 -2.41704267e-02 3.69772702e-01 7.59208441e-01 -6.73025906e-01 1.10370553e+00 1.14902604e+00 5.84711373e-01 5.45229733e-01 -8.14029038e-01 -5.02532661e-01 7.87822604e-01 4.70137537e-01 -1.44943786e+00 -2.76450604e-01 6.54441714e-01 -5.34894288e-01 8.83077145e-01 -1.41786292e-01 9.63114381e-01 8.78745794e-01 3.94466490e-01 1.28093421e+00 6.31632566e-01 5.15552089e-02 1.52827442e-01 -5.39548099e-01 3.63445789e-01 9.02721941e-01 -1.80067196e-01 -1.10228546e-01 -7.55747259e-01 1.95918947e-01 1.06756032e+00 3.08730572e-01 -3.20054650e-01 -3.80040526e-01 -1.09791994e+00 5.55632889e-01 6.48907781e-01 4.00917947e-01 -6.00098193e-01 3.88161749e-01 3.83945465e-01 -1.64377540e-02 6.49702966e-01 -3.42257261e-01 -5.00812650e-01 6.05946742e-02 -1.40723705e+00 2.31297761e-02 3.93137842e-01 1.11584461e+00 6.93069994e-01 8.41516852e-02 -4.12205189e-01 6.89303458e-01 5.28577149e-01 2.00797230e-01 5.47649145e-01 -1.00553548e+00 4.48841423e-01 3.69114101e-01 7.91624486e-02 -9.67558742e-01 -1.36438325e-01 -4.01751548e-01 -8.31920922e-01 -5.20259887e-02 2.25226402e-01 -8.31240490e-02 -1.51293302e+00 1.88091671e+00 6.02813423e-01 9.23990726e-01 -1.76921755e-01 1.06438351e+00 1.06278515e+00 9.97729838e-01 2.35874087e-01 -5.12809873e-01 1.06774259e+00 -1.45462501e+00 -9.05829012e-01 -4.90773648e-01 3.00350428e-01 -4.60393310e-01 6.86687589e-01 -2.26428166e-01 -1.27483511e+00 -7.49018848e-01 -1.05231452e+00 -1.71453372e-01 -2.53990620e-01 -2.31486455e-01 2.21535861e-01 -1.73276901e-01 -1.14077210e+00 7.29766548e-01 -1.12669623e+00 -2.72435755e-01 6.27352238e-01 2.02205732e-01 -5.88913113e-02 -7.51134530e-02 -1.16035807e+00 3.07924569e-01 3.92484456e-01 5.19264102e-01 -1.11544597e+00 -6.53058946e-01 -1.04973614e+00 -1.36744514e-01 6.18698895e-01 -7.81676292e-01 1.05584431e+00 -9.79290545e-01 -1.11829293e+00 9.05039310e-01 -6.54677033e-01 -6.41089857e-01 7.06517398e-01 -7.02860281e-02 -2.30029538e-01 3.36113542e-01 3.76136690e-01 9.92881119e-01 1.08673608e+00 -1.12664402e+00 -8.19397211e-01 -3.36021215e-01 -2.82613754e-01 2.04576030e-01 -1.37748614e-01 -1.28858268e-01 -1.37677109e+00 -9.61521327e-01 2.78143615e-01 -9.71445262e-01 -2.82066703e-01 -2.70210840e-02 -4.30692524e-01 -2.32830659e-01 1.20929277e+00 -9.34386790e-01 1.33962512e+00 -2.12948608e+00 3.43347400e-01 -1.17291443e-01 1.94484934e-01 5.11591136e-01 -2.45141715e-01 -1.06891848e-01 1.57724619e-02 1.40799154e-02 -3.08426976e-01 -5.90609729e-01 -3.49935889e-01 3.34119260e-01 -2.86899865e-01 4.73963708e-01 -2.23086067e-02 1.40355992e+00 -9.33617473e-01 -8.18937242e-01 2.71996826e-01 5.26868165e-01 -1.04213268e-01 5.30421555e-01 -5.07546663e-01 5.02692223e-01 -4.90787953e-01 7.48966098e-01 4.72263068e-01 -4.65446055e-01 -2.08848596e-01 -1.59288466e-01 -2.94383597e-02 6.49665222e-02 -1.13890076e+00 2.07101297e+00 -3.57978512e-03 6.80375516e-01 6.54774159e-02 -6.76108420e-01 5.24410248e-01 1.53764054e-01 7.16912389e-01 -7.88571417e-01 5.17924838e-02 -5.82216755e-02 -4.87893522e-01 -7.19371974e-01 4.69593912e-01 2.55869627e-01 3.40633631e-01 3.10858727e-01 1.20154336e-01 3.15926105e-01 4.32235956e-01 3.61553967e-01 7.69456923e-01 3.85339260e-01 -1.92890406e-01 -4.86663356e-02 3.27562004e-01 -1.60246313e-01 1.09785628e+00 6.40436649e-01 -4.95888323e-01 9.94504988e-01 1.94182783e-01 -5.12436211e-01 -6.84790194e-01 -8.19533765e-01 2.52064586e-01 1.07270300e+00 6.04881346e-01 -2.86767215e-01 -7.70599604e-01 -7.08400309e-01 -2.18811765e-01 3.09448391e-01 -6.71638072e-01 -8.49775746e-02 -7.46004760e-01 -5.37679315e-01 2.47214809e-01 8.27583313e-01 8.01462650e-01 -1.31347072e+00 -6.39395952e-01 4.01149124e-01 -4.55111653e-01 -1.45174837e+00 -1.19278526e+00 -1.54427998e-02 -1.04499125e+00 -1.01283300e+00 -1.03755450e+00 -1.01462221e+00 4.10209388e-01 5.49325347e-01 9.83636200e-01 1.29149124e-01 -2.24978507e-01 2.34043211e-01 -8.31821337e-02 1.09504521e-01 6.90324754e-02 4.38848436e-02 -3.41563612e-01 3.50866497e-01 -2.14854125e-02 -5.39472282e-01 -7.84029901e-01 4.46402043e-01 -9.42297101e-01 1.74515530e-01 2.26675376e-01 5.98266721e-01 1.13898456e+00 -7.34937117e-02 2.82631963e-01 -7.82257259e-01 -7.26080984e-02 -4.45137382e-01 -5.08880973e-01 4.60985273e-01 -3.89228255e-01 -3.03061545e-01 4.66315672e-02 -5.38331509e-01 -7.84000397e-01 1.63913473e-01 3.96258719e-02 -1.01694381e+00 1.04012854e-01 4.33814824e-01 -1.82641745e-01 -1.02411427e-01 -4.18757275e-02 4.88682181e-01 -2.96428084e-01 -3.31723869e-01 2.68302828e-01 2.67055392e-01 9.16469395e-01 -3.33341300e-01 4.03208107e-01 5.60777962e-01 -1.83745563e-01 -7.64610767e-01 -1.10480964e+00 -7.36465156e-01 -9.07473266e-01 -5.91296494e-01 1.36648357e+00 -1.09925544e+00 -4.86423999e-01 8.33622515e-01 -1.16353321e+00 -8.18917215e-01 -3.51036668e-01 1.89056426e-01 -5.42202771e-01 4.41487551e-01 -9.50339139e-01 -4.79340345e-01 -2.89409697e-01 -1.22716784e+00 1.14958096e+00 6.05320454e-01 1.02285013e-01 -1.20029366e+00 -1.05658464e-01 4.40053284e-01 3.12948264e-02 5.28466463e-01 4.96019781e-01 -1.07314579e-01 -9.09022987e-01 1.26754433e-01 -2.05424622e-01 3.52236748e-01 -3.40331644e-02 4.24813852e-02 -7.27904975e-01 -3.63671422e-01 -3.64244133e-02 -7.75592774e-02 1.37433791e+00 9.79696751e-01 1.17015755e+00 4.36589420e-02 -3.83339494e-01 9.41206217e-01 1.24959648e+00 2.98179924e-01 6.57510579e-01 2.03725114e-01 1.39800203e+00 4.24027681e-01 8.55368376e-01 3.52972709e-02 6.07180655e-01 5.13868928e-01 4.68181610e-01 -2.17805535e-01 -4.26242948e-01 -3.25929940e-01 3.52549136e-01 6.98335230e-01 5.57676703e-02 -4.13355380e-01 -8.28489661e-01 8.68684769e-01 -2.55460262e+00 -9.83663380e-01 -2.19804734e-01 2.09851718e+00 5.48339307e-01 1.86954170e-01 1.89536780e-01 -3.70673776e-01 1.14720941e+00 7.88536131e-01 -1.04398048e+00 8.83660987e-02 -3.00119936e-01 -4.05482948e-01 5.90327203e-01 4.69758958e-01 -1.30552924e+00 1.22722018e+00 5.98167992e+00 7.74301648e-01 -1.16317821e+00 2.34163269e-01 1.03738308e+00 -2.49648020e-01 1.70512944e-01 -1.79758355e-01 -6.92350864e-01 5.33104658e-01 7.09188938e-01 2.16095701e-01 2.07963884e-01 6.15115881e-01 4.29256380e-01 -2.27977738e-01 -1.01240432e+00 1.08809423e+00 5.21482155e-02 -1.64768326e+00 -3.17444727e-02 -1.42939731e-01 1.05289781e+00 3.92183334e-01 7.01371431e-02 -8.91404506e-03 1.22856488e-02 -9.05491471e-01 1.13361883e+00 6.13838434e-01 6.36859596e-01 -6.87757552e-01 4.73986417e-01 2.30473012e-01 -1.70753503e+00 1.82886720e-01 -5.62897399e-02 3.38295460e-01 4.63865340e-01 5.08863032e-01 -2.26160139e-01 4.64286953e-01 9.67943907e-01 1.34750998e+00 -5.38746238e-01 1.02406859e+00 -1.91380262e-01 6.20525599e-01 -1.91491127e-01 5.36375105e-01 6.44932091e-01 -3.47181678e-01 5.19272149e-01 1.28430152e+00 -5.18154763e-02 2.15773866e-01 5.37482500e-01 7.18248248e-01 -2.24383384e-01 -2.64993519e-01 -2.02623755e-01 -8.97405520e-02 2.41894454e-01 1.07965934e+00 -1.19336176e+00 -6.05062246e-01 -3.25953126e-01 1.17198503e+00 2.54118949e-01 7.34705627e-01 -8.62955928e-01 2.09611915e-02 5.54087818e-01 2.09862918e-01 7.68621981e-01 -3.00531596e-01 -2.32394755e-01 -1.22132647e+00 2.45034471e-01 -4.21000540e-01 7.49047101e-01 -8.99151564e-01 -1.07315457e+00 6.19839609e-01 -1.53284371e-01 -9.44597125e-01 -1.79772556e-01 2.94424240e-02 -6.69189334e-01 7.62412965e-01 -1.67956841e+00 -1.20546699e+00 -3.52699429e-01 8.28514278e-01 9.26727593e-01 2.06802070e-01 -4.10678238e-02 3.30853969e-01 -8.42901051e-01 2.31840268e-01 -1.91726878e-01 5.03653467e-01 5.15974879e-01 -9.62699771e-01 5.50608873e-01 1.27950883e+00 1.95004657e-01 2.53792197e-01 5.09173572e-01 -1.02065420e+00 -1.17468059e+00 -1.68757868e+00 8.12405229e-01 1.94567107e-02 4.41164523e-01 -2.80762494e-01 -1.23721671e+00 9.26021516e-01 8.49387646e-02 7.28453040e-01 -1.02456892e-03 -4.17661965e-01 -2.15385079e-01 7.67149925e-02 -8.33903372e-01 4.41647738e-01 1.32037294e+00 -7.08171308e-01 -4.21691626e-01 3.57169122e-01 1.03093958e+00 -8.76056492e-01 -6.05002344e-01 4.16234195e-01 2.81991154e-01 -9.05531168e-01 1.00150883e+00 -4.68331099e-01 4.04290378e-01 -5.85247159e-01 3.96926180e-02 -8.54667008e-01 -2.02937305e-01 -8.55442107e-01 -3.97220403e-01 1.29331684e+00 -3.61193866e-02 -1.77407414e-01 9.59747016e-01 6.59712195e-01 -1.56069100e-01 -1.00223398e+00 -8.67120743e-01 -7.19517350e-01 -3.19332331e-01 -5.54835558e-01 2.58845866e-01 7.10007071e-01 -6.84058726e-01 1.91570356e-01 -5.22331774e-01 2.68848985e-01 6.12477303e-01 5.30348182e-01 4.83900368e-01 -8.50087643e-01 -1.16522677e-01 -1.75009906e-01 -5.01347840e-01 -1.60946715e+00 3.67161632e-01 -7.17995822e-01 2.68435985e-01 -1.66304100e+00 3.54034662e-01 -1.52006090e-01 -4.25657719e-01 3.91544431e-01 -3.56145352e-01 3.82193685e-01 2.71696359e-01 3.57638180e-01 -1.20461690e+00 5.93326092e-01 1.43744099e+00 -1.89399585e-01 -5.79318106e-01 -8.91633257e-02 -1.23375989e-01 8.84543777e-01 3.75823438e-01 -5.63050747e-01 -4.58663821e-01 -5.63244998e-01 -2.80990422e-01 3.17933649e-01 4.49463904e-01 -7.69497633e-01 7.28901446e-01 -2.49901548e-01 3.68493974e-01 -9.44036603e-01 2.25894094e-01 -4.99693543e-01 1.72960117e-01 4.09047782e-01 -3.19400221e-01 3.23482901e-02 2.66926456e-03 9.39777792e-01 -4.37021494e-01 1.16330124e-01 9.51071262e-01 -1.29162490e-01 -1.38293540e+00 1.14226997e+00 -1.48499697e-01 3.37728560e-01 1.19394314e+00 -3.60016435e-01 1.67678639e-01 -3.33991170e-01 -9.30618584e-01 6.91859901e-01 4.38401073e-01 7.61883318e-01 8.16685200e-01 -1.16401219e+00 -3.44992250e-01 -4.21866588e-03 -7.93384165e-02 5.18482864e-01 7.22944200e-01 7.67574906e-01 -3.54147166e-01 1.41030490e-01 -3.79039235e-02 -1.04538143e+00 -1.32035565e+00 6.25864327e-01 4.81254429e-01 -1.40302658e-01 -9.89886940e-01 1.10991120e+00 4.19681996e-01 2.83493757e-01 5.45559406e-01 -3.70800465e-01 -2.16146201e-01 1.05707973e-01 6.20322168e-01 3.19523782e-01 -2.45232165e-01 -1.20878863e+00 -3.56964737e-01 8.70470345e-01 -1.15214303e-01 -7.46415854e-02 1.37361169e+00 -5.98514616e-01 -1.09868519e-01 8.17624629e-01 1.39793468e+00 -6.81348801e-01 -2.00711846e+00 -5.66496611e-01 -1.08197778e-01 -4.87068623e-01 2.87040919e-01 -3.36044550e-01 -1.77161717e+00 7.33676791e-01 5.10022521e-01 -1.57016113e-01 1.15288484e+00 -1.29198087e-02 1.37273216e+00 -9.82254073e-02 2.43830293e-01 -1.21154654e+00 1.47234157e-01 5.35950243e-01 5.09484351e-01 -1.25786054e+00 -1.79756120e-01 -5.00699401e-01 -7.38552690e-01 9.56520438e-01 5.58081448e-01 -2.68538922e-01 7.23354697e-01 7.56185949e-02 7.98877180e-02 -2.73111165e-01 -7.45974422e-01 -3.38589609e-01 5.66391528e-01 4.29889143e-01 3.63724981e-03 -2.37381816e-01 1.45725563e-01 4.27813679e-01 3.31199884e-01 9.97538865e-02 1.91209719e-01 9.08061624e-01 -2.43895859e-01 -6.20678425e-01 -1.14593841e-01 2.25674108e-01 -3.24990988e-01 1.65460840e-01 -9.92275104e-02 5.62052608e-01 2.48577908e-01 9.12062705e-01 1.58878475e-01 -2.64337987e-01 1.47901386e-01 -2.40228022e-03 2.96892583e-01 -4.53499019e-01 -4.82583195e-01 4.17967051e-01 -4.10469294e-01 -1.07471478e+00 -7.68903852e-01 -8.57836843e-01 -1.66639698e+00 -2.30212584e-01 -3.56528461e-02 -1.41482055e-01 3.39055002e-01 1.21449244e+00 4.56327617e-01 8.45016837e-01 4.38641161e-01 -1.04664636e+00 1.54435202e-01 -5.24332523e-01 -6.42405093e-01 5.34093380e-01 5.64328730e-01 -4.72538382e-01 -4.60397415e-02 3.90447944e-01]

[9.180350303649902, -0.07871927320957184]

b206e803-24d8-4817-8d5f-d0ec54a2c532

argoverse-2-next-generation-datasets-for-self-1

2301.00493

https://arxiv.org/abs/2301.00493v1

https://arxiv.org/pdf/2301.00493v1.pdf

Argoverse 2: Next Generation Datasets for Self-Driving Perception and Forecasting

We introduce Argoverse 2 (AV2) - a collection of three datasets for perception and forecasting research in the self-driving domain. The annotated Sensor Dataset contains 1,000 sequences of multimodal data, encompassing high-resolution imagery from seven ring cameras, and two stereo cameras in addition to lidar point clouds, and 6-DOF map-aligned pose. Sequences contain 3D cuboid annotations for 26 object categories, all of which are sufficiently-sampled to support training and evaluation of 3D perception models. The Lidar Dataset contains 20,000 sequences of unlabeled lidar point clouds and map-aligned pose. This dataset is the largest ever collection of lidar sensor data and supports self-supervised learning and the emerging task of point cloud forecasting. Finally, the Motion Forecasting Dataset contains 250,000 scenarios mined for interesting and challenging interactions between the autonomous vehicle and other actors in each local scene. Models are tasked with the prediction of future motion for "scored actors" in each scenario and are provided with track histories that capture object location, heading, velocity, and category. In all three datasets, each scenario contains its own HD Map with 3D lane and crosswalk geometry - sourced from data captured in six distinct cities. We believe these datasets will support new and existing machine learning research problems in ways that existing datasets do not. All datasets are released under the CC BY-NC-SA 4.0 license.

['James Hays', 'Peter Carr', 'Deva Ramanan', 'Jhony Kaesemodel Pontes', 'Andrew Hartnett', 'Ratnesh Kumar', 'Bowen Pan', 'Siddhesh Khandelwal', 'Jagjeet Singh', 'John Lambert', 'Tanmay Agarwal', 'William Qi', 'Benjamin Wilson']

2023-01-02

argoverse-2-next-generation-datasets-for-self

https://datasets-benchmarks-proceedings.neurips.cc/paper/2021/hash/4734ba6f3de83d861c3176a6273cac6d-Abstract-round2.html

https://datasets-benchmarks-proceedings.neurips.cc/paper/2021/file/4734ba6f3de83d861c3176a6273cac6d-Paper-round2.pdf

proceedings-of-the-neural-information

['motion-forecasting']

['computer-vision']

[-1.37484923e-01 -2.61974800e-02 -5.49918473e-01 -9.18987989e-01 -7.43012846e-01 -7.70988762e-01 9.41890657e-01 2.35921726e-01 -3.49581182e-01 4.05818105e-01 1.63898796e-01 -2.79948503e-01 1.07105352e-01 -6.95284784e-01 -8.60799789e-01 -4.50661987e-01 -3.89916331e-01 1.02919734e+00 4.03523624e-01 -4.09471929e-01 2.07996130e-01 7.37444818e-01 -2.10673523e+00 1.35909155e-01 4.10680056e-01 1.15059412e+00 4.05641079e-01 7.63837337e-01 5.54124117e-02 5.91361225e-01 1.75726771e-01 -2.90856361e-01 4.87122238e-01 4.78612274e-01 -1.99006170e-01 2.85213679e-01 1.01430476e+00 -2.58581251e-01 -5.01157939e-01 6.67904198e-01 1.27840787e-01 1.11109950e-01 6.10705316e-01 -1.93079627e+00 -1.78269874e-02 1.04071684e-01 -4.08147782e-01 3.14480036e-01 2.96518534e-01 7.18017340e-01 7.67429650e-01 -9.81283724e-01 7.74051309e-01 1.42799127e+00 5.48996687e-01 2.00300843e-01 -1.07112980e+00 -6.07134104e-01 1.06170692e-01 6.05734348e-01 -1.46807921e+00 -6.46408200e-01 6.50567830e-01 -8.00012112e-01 9.08976376e-01 -2.92351749e-02 6.61668777e-01 1.39906657e+00 3.55464891e-02 5.88663399e-01 7.23343670e-01 3.78446102e-01 3.93381387e-01 2.62761384e-01 -2.08720891e-03 6.12892151e-01 2.21979350e-01 4.94712621e-01 -7.61470437e-01 1.62454264e-03 2.00579017e-01 1.64312854e-01 4.44307089e-01 -9.39509213e-01 -1.58191669e+00 7.65478671e-01 3.13635021e-01 -5.20793021e-01 -3.63743037e-01 1.66601658e-01 3.16730499e-01 4.90863435e-02 3.40378821e-01 -2.46707857e-01 -5.11649549e-01 -2.99308330e-01 -5.21387458e-01 7.95441329e-01 5.19952178e-01 1.34574723e+00 1.31894577e+00 4.88871858e-02 3.94375920e-01 5.23439229e-01 3.96913886e-01 1.32233441e+00 -6.19799644e-02 -1.54262578e+00 8.08961153e-01 6.80094302e-01 4.21195120e-01 -1.19438910e+00 -6.23094320e-01 1.02103233e-01 -3.94706398e-01 1.73721001e-01 1.55155241e-01 -4.70946776e-03 -7.12154090e-01 1.39699268e+00 6.49176478e-01 4.22277331e-01 2.35026583e-01 9.69694316e-01 9.22000289e-01 6.65167987e-01 -1.82013046e-02 1.57773226e-01 9.94919062e-01 -6.08748496e-01 -4.34322834e-01 -6.52367830e-01 6.03894114e-01 -2.32873440e-01 7.19547629e-01 1.85447827e-01 -6.65193558e-01 -8.81530106e-01 -9.03172135e-01 4.14350815e-02 -7.01764524e-01 -1.25482738e-01 4.88916248e-01 1.71695217e-01 -8.65582645e-01 -9.85655189e-02 -7.14347601e-01 -4.85215634e-01 5.14947712e-01 -8.18907246e-02 -6.38895631e-01 -4.40072268e-01 -8.80870461e-01 1.15420151e+00 1.92008629e-01 -2.37682909e-02 -1.41689694e+00 -7.50802398e-01 -1.34207571e+00 -6.43438280e-01 2.93091744e-01 -2.63498187e-01 1.17196548e+00 -3.84308919e-02 -6.25975788e-01 1.13277590e+00 -4.47746873e-01 -6.84096515e-01 4.66319978e-01 -6.09461702e-02 -7.24321187e-01 -5.73002920e-02 7.55188584e-01 1.37145972e+00 5.65336764e-01 -1.55619407e+00 -1.33069241e+00 -8.55465889e-01 -2.06735745e-01 4.78823006e-01 3.44156712e-01 -5.85393429e-01 -3.80168408e-01 1.50249422e-01 2.51368880e-01 -1.10228288e+00 -3.60647500e-01 -3.25322077e-02 -2.20761791e-01 -3.01283270e-01 1.15156960e+00 -1.41824782e-01 3.42421561e-01 -2.22952914e+00 -1.60805345e-01 -5.89453615e-02 1.33413553e-01 -2.37304553e-01 -3.01440567e-01 4.41785723e-01 2.55345553e-01 -1.32457793e-01 2.41436884e-02 -5.43965995e-01 1.36105254e-01 6.47751153e-01 -5.90899825e-01 7.34384239e-01 1.35030955e-01 9.07322109e-01 -8.89353573e-01 -4.83901590e-01 6.46099269e-01 8.61684382e-02 -2.35321015e-01 8.63850787e-02 -5.30819833e-01 6.20793402e-01 -3.68129343e-01 8.07355046e-01 9.13443565e-01 3.28380257e-01 -1.81453511e-01 -1.05180718e-01 -5.25689244e-01 2.37073135e-02 -1.32559526e+00 1.69098449e+00 -2.94725802e-02 9.67442453e-01 -1.21691085e-01 -6.90001070e-01 1.24201107e+00 -1.32606268e-01 7.32920706e-01 -8.40611756e-01 -4.87517565e-02 -1.47963699e-03 -5.48073828e-01 -8.88769865e-01 8.76642764e-01 2.42921263e-01 -4.67440546e-01 1.13995718e-02 -3.63111734e-01 -6.25340641e-01 2.83675581e-01 2.28244916e-01 9.52597260e-01 -7.20983371e-02 -1.85068488e-01 9.30957645e-02 1.74661800e-01 8.88370693e-01 6.14219546e-01 6.64417505e-01 -4.48428959e-01 2.98681408e-01 1.28271937e-01 -9.51828420e-01 -1.40034962e+00 -1.36371219e+00 -2.97253847e-01 8.01603317e-01 6.39034271e-01 -1.53273568e-01 1.28236294e-01 -3.31772119e-01 6.67741537e-01 9.85281050e-01 -4.58124250e-01 1.54223546e-01 -3.74259084e-01 -7.90219679e-02 5.51853597e-01 5.00592828e-01 3.89850140e-01 -6.56382203e-01 -8.39884579e-01 -5.95840812e-02 -5.34953713e-01 -1.69508123e+00 1.27292369e-02 -6.36172947e-03 -6.09896779e-01 -1.26801848e+00 1.43328175e-01 -3.78540903e-01 2.27907106e-01 7.90877163e-01 1.08010066e+00 -4.89871830e-01 -2.03986555e-01 9.26434577e-01 -2.68909842e-01 -8.86249244e-01 -2.40036488e-01 -2.41232052e-01 5.12279987e-01 1.05825201e-01 8.93459022e-01 -3.83937955e-01 -3.08770686e-01 6.04910195e-01 -1.19559370e-01 2.41395254e-02 2.94277042e-01 3.36108468e-02 7.41854608e-01 -1.34317011e-01 2.77424663e-01 -1.58254914e-02 -7.80012980e-02 -9.63023782e-01 -7.80203223e-01 -3.60639513e-01 -3.53730917e-01 -4.50353384e-01 -1.72195844e-02 -1.27302915e-01 -8.09363008e-01 5.62194526e-01 4.20843035e-01 -8.56508195e-01 -8.97579491e-01 1.31505460e-01 -9.05866176e-02 3.10706258e-01 8.11292470e-01 1.51124880e-01 1.82920203e-01 -1.48001745e-01 7.15294182e-01 5.70352376e-01 1.21334350e+00 -2.72978932e-01 1.06764913e+00 9.48070884e-01 1.76212609e-01 -1.00942540e+00 -8.55147481e-01 -7.77085304e-01 -9.54980969e-01 -6.70502305e-01 1.08781362e+00 -1.53203738e+00 -8.80266905e-01 3.01781416e-01 -1.01372850e+00 -2.62499124e-01 -4.06559139e-01 6.05654180e-01 -8.53607297e-01 -4.40182872e-02 2.16139823e-01 -1.04698443e+00 5.59774101e-01 -1.05713260e+00 1.39441609e+00 4.84125242e-02 2.11222395e-02 -6.22495055e-01 2.66097039e-01 7.80175984e-01 2.25346461e-02 5.47175467e-01 2.85954148e-01 -3.73626083e-01 -9.47053611e-01 -2.37179935e-01 -6.85150176e-02 -1.45675600e-01 -2.57248670e-01 -1.00269422e-01 -8.83536637e-01 -1.41009822e-01 -4.30564940e-01 -5.62672496e-01 8.64090383e-01 3.33778232e-01 6.75058722e-01 9.98936500e-03 -6.28118515e-01 3.80416453e-01 1.15054357e+00 1.87491745e-01 1.72955677e-01 3.30666900e-01 8.49659979e-01 9.90102828e-01 1.16119897e+00 4.96931434e-01 1.45076728e+00 7.41352558e-01 1.13720536e+00 1.51547879e-01 2.83780962e-01 -5.14053464e-01 2.85821110e-01 1.04474604e-01 1.95726871e-01 -1.52751710e-02 -1.40345812e+00 9.32511032e-01 -1.96538985e+00 -1.25952041e+00 -6.58663809e-01 1.95735407e+00 1.11887850e-01 1.85546786e-01 2.85016060e-01 -3.52571979e-02 3.80562752e-01 4.26557362e-01 -1.03313470e+00 -1.64640155e-02 -2.63398409e-01 -7.68930435e-01 8.94974649e-01 5.23345947e-01 -1.31446302e+00 9.59789455e-01 6.12302780e+00 1.96036026e-01 -8.16781938e-01 -2.66755726e-02 2.39523366e-01 -1.84894949e-01 -2.11212978e-01 1.12532422e-01 -1.36061764e+00 3.87954414e-01 1.26070511e+00 4.21105325e-03 2.00245708e-01 1.11257803e+00 6.75552666e-01 -5.29438317e-01 -1.17270398e+00 1.13900316e+00 1.76821306e-01 -1.67862141e+00 -2.66724348e-01 3.47833335e-01 6.18836522e-01 8.59971344e-01 1.87963769e-01 2.71167159e-01 6.74694896e-01 -8.89807284e-01 1.00538027e+00 6.53284192e-01 7.66606450e-01 -7.84475386e-01 4.60308135e-01 8.95928800e-01 -1.32289898e+00 -5.12267888e-01 -3.69310200e-01 -1.59279615e-01 4.84731823e-01 3.00627023e-01 -1.11666048e+00 3.55227500e-01 1.15038848e+00 1.23333061e+00 -6.67258561e-01 8.50795269e-01 2.95817792e-01 2.37000212e-01 -4.01255429e-01 1.56040266e-01 3.99666131e-01 -2.32761979e-01 8.88565183e-01 9.42798793e-01 1.52932271e-01 5.12944981e-02 6.75406277e-01 6.42246842e-01 3.82620424e-01 -5.63436925e-01 -1.50352263e+00 4.10690814e-01 9.46568370e-01 1.26281953e+00 -1.52153626e-01 -3.69778901e-01 -2.49984369e-01 1.45662546e-01 -2.83081643e-03 2.44226530e-01 -8.84825468e-01 4.03216958e-01 1.22146821e+00 3.91593784e-01 3.46864879e-01 -8.48218441e-01 -3.28167111e-01 -8.24853599e-01 -1.61985740e-01 -4.33921397e-01 1.36126027e-01 -1.09235787e+00 -1.12434435e+00 2.11735651e-01 5.14508069e-01 -1.53968084e+00 -4.19347614e-01 -3.40835392e-01 -3.10345352e-01 4.78561938e-01 -1.60590959e+00 -1.40610993e+00 -7.83603072e-01 7.17371225e-01 7.60307789e-01 -4.78132546e-01 3.44951361e-01 1.08123012e-01 -2.77248234e-01 -2.76001185e-01 -1.19096071e-01 -1.84225380e-01 5.90336204e-01 -8.80484164e-01 6.79449916e-01 3.96922499e-01 1.86583877e-01 -2.44573489e-01 8.26563895e-01 -8.33995104e-01 -1.77225006e+00 -1.64127123e+00 9.44602847e-01 -1.25486243e+00 6.36505306e-01 -5.47278702e-01 -4.39575791e-01 9.22529817e-01 -2.94093817e-01 2.53459215e-01 2.35031098e-01 -1.89467505e-01 -2.60856599e-01 -4.78285789e-01 -1.07516921e+00 3.37484360e-01 1.21604133e+00 -3.85202587e-01 -5.73446035e-01 4.14676070e-01 6.05518401e-01 -5.94187677e-01 -6.22147381e-01 6.48007274e-01 3.35396647e-01 -1.00578439e+00 1.07007825e+00 -5.46500325e-01 1.97792456e-01 -6.12427950e-01 -9.68641877e-01 -1.04119313e+00 -1.83038279e-01 5.47717176e-02 -3.50249335e-02 9.27639186e-01 2.76482463e-01 -3.79533350e-01 9.22864676e-01 5.26004672e-01 -4.11420137e-01 -2.46817410e-01 -1.07964504e+00 -7.34179318e-01 -2.83027947e-01 -1.19560969e+00 7.40722775e-01 6.91912711e-01 -5.30170679e-01 3.34836125e-01 -2.84516454e-01 5.49956858e-01 1.01745319e+00 6.28142655e-02 1.57367384e+00 -1.51361489e+00 6.01288617e-01 -1.53458744e-01 -8.90744090e-01 -1.13727951e+00 4.42181289e-01 -7.98816979e-01 1.39126301e-01 -1.35923171e+00 -1.22694232e-01 -6.68843865e-01 5.53269625e-01 4.35033888e-01 4.37008619e-01 1.11830078e-01 6.11451641e-02 3.26567650e-01 -7.66316175e-01 7.14108407e-01 7.22261071e-01 -3.92515600e-01 -1.29098510e-02 1.18433647e-01 -2.01098040e-01 7.69377351e-01 6.19351208e-01 -2.21978068e-01 -5.86367607e-01 -5.49484134e-01 -2.44194106e-03 1.51913702e-01 9.66674507e-01 -1.17987621e+00 4.34238374e-01 -6.60713732e-01 4.05391902e-01 -1.77773118e+00 9.62298989e-01 -1.04807508e+00 1.87980846e-01 5.14434986e-02 -2.34061584e-01 4.16356534e-01 3.06641102e-01 9.39943671e-01 5.48345968e-03 4.78273302e-01 4.35894907e-01 -1.21095628e-01 -1.63243532e+00 6.65612161e-01 -4.99592632e-01 3.28470841e-02 1.53519678e+00 -5.81109285e-01 -4.12504435e-01 -4.51467484e-01 -5.20033062e-01 9.00628746e-01 4.46575761e-01 1.00238681e+00 7.37504065e-01 -1.50232184e+00 -9.66176212e-01 4.25089568e-01 6.72624171e-01 3.53813499e-01 4.47630703e-01 5.40619731e-01 1.01886941e-02 4.91993427e-01 -3.21490109e-01 -1.39332616e+00 -1.03172231e+00 5.40282965e-01 1.64732039e-01 5.49955964e-01 -7.27026403e-01 3.59510452e-01 -1.53272197e-01 -9.06493366e-01 2.45513782e-01 -6.93668425e-02 -2.81794548e-01 2.21752733e-01 3.65180761e-01 5.25157690e-01 -1.68454096e-01 -1.43401980e+00 -5.52764714e-01 5.56876004e-01 4.80586082e-01 -2.76973933e-01 1.19801354e+00 -7.07118690e-01 2.44525313e-01 1.01124108e+00 1.01130331e+00 -4.13008839e-01 -1.62324512e+00 -2.38252074e-01 9.00730863e-02 -5.13879180e-01 -1.47446081e-01 -3.91040534e-01 -7.29639530e-01 8.29562306e-01 7.98271537e-01 -6.38178736e-02 3.50335181e-01 4.44785744e-01 5.53172290e-01 5.23376942e-01 6.44055128e-01 -1.05752003e+00 2.71294750e-02 7.98669815e-01 7.62351930e-01 -1.81922853e+00 -2.26317182e-01 1.60480430e-03 -1.19314122e+00 6.44314885e-01 8.96552444e-01 1.01265110e-01 5.95237195e-01 2.48305574e-01 2.05541357e-01 -4.35416162e-01 -1.20557106e+00 -5.53805828e-01 9.47297439e-02 1.09323978e+00 -4.98738140e-01 4.11952168e-01 7.37812102e-01 1.49333179e-01 -6.55715287e-01 -3.08828712e-01 4.08592433e-01 6.68975711e-01 -7.34008431e-01 -5.06080329e-01 -4.92055744e-01 3.08825701e-01 7.50369668e-01 5.06393313e-01 -1.64519265e-01 7.47427821e-01 5.47118366e-01 1.34305680e+00 7.01156378e-01 -7.59455740e-01 6.24973357e-01 -2.26014182e-01 3.04961074e-02 -2.57904440e-01 2.13562116e-01 -4.61176872e-01 4.62331742e-01 -8.68642986e-01 -2.78998643e-01 -1.20053184e+00 -1.21627617e+00 -6.32197320e-01 4.27776188e-01 -1.84480190e-01 1.06467390e+00 9.00123239e-01 5.45420408e-01 -9.26503241e-02 9.07430708e-01 -1.44511902e+00 -2.89442152e-01 -6.24717057e-01 -4.35731858e-01 2.39249244e-01 6.78982437e-01 -9.47140336e-01 -1.46283805e-01 2.41497651e-01]

[7.781923294067383, -1.9849220514297485]

8c571041-9f02-4603-9c3d-a72cd527f50d

v4d4d-convolutional-neural-networks-for-video

2002.07442

https://arxiv.org/abs/2002.07442v1

https://arxiv.org/pdf/2002.07442v1.pdf

V4D:4D Convolutional Neural Networks for Video-level Representation Learning

Most existing 3D CNNs for video representation learning are clip-based methods, and thus do not consider video-level temporal evolution of spatio-temporal features. In this paper, we propose Video-level 4D Convolutional Neural Networks, referred as V4D, to model the evolution of long-range spatio-temporal representation with 4D convolutions, and at the same time, to preserve strong 3D spatio-temporal representation with residual connections. Specifically, we design a new 4D residual block able to capture inter-clip interactions, which could enhance the representation power of the original clip-level 3D CNNs. The 4D residual blocks can be easily integrated into the existing 3D CNNs to perform long-range modeling hierarchically. We further introduce the training and inference methods for the proposed V4D. Extensive experiments are conducted on three video recognition benchmarks, where V4D achieves excellent results, surpassing recent 3D CNNs by a large margin.

['Li-Min Wang', 'Weilin Huang', 'Sheng Guo', 'Matthew R. Scott', 'Shiwen Zhang']

2020-02-18

['long-range-modeling']

['natural-language-processing']

[-2.94838697e-01 -4.20448005e-01 -4.83290672e-01 -1.42260075e-01 -1.18864290e-01 -2.95864254e-01 7.38769174e-01 -1.95144415e-01 -3.93762514e-02 1.57173738e-01 3.93648326e-01 -2.46375263e-01 9.26265586e-03 -6.66721761e-01 -1.03503239e+00 -5.49179256e-01 -5.12466431e-01 -2.56453902e-01 4.96245563e-01 -5.28283827e-02 -2.93029994e-02 9.02774513e-01 -1.54497242e+00 5.88056862e-01 2.62478918e-01 1.43551075e+00 -3.87122445e-02 6.16541028e-01 -7.13627934e-02 1.22016490e+00 -3.58224034e-01 2.19177589e-01 2.98887432e-01 -1.76395610e-01 -4.18035716e-01 2.59978801e-01 4.47389543e-01 -5.26041746e-01 -1.19381177e+00 4.57485616e-01 2.13624671e-01 3.81148756e-01 5.09399712e-01 -1.09132111e+00 -9.39934790e-01 2.97120720e-01 -7.03141749e-01 2.87513345e-01 3.62354457e-01 1.91099882e-01 6.80121422e-01 -1.00041890e+00 8.11773658e-01 1.50370955e+00 8.47334266e-01 5.01734018e-01 -1.08170068e+00 -6.71540260e-01 6.31339729e-01 4.07184333e-01 -1.54510128e+00 -1.80421159e-01 1.11399412e+00 -6.54983699e-01 1.40937614e+00 -1.47324763e-02 9.84684169e-01 1.33716404e+00 2.66976506e-01 1.20220506e+00 4.97313559e-01 -2.29222029e-02 -1.27165750e-01 -6.27146006e-01 1.46726564e-01 5.92025936e-01 -1.52942806e-01 1.47977099e-01 -6.82090998e-01 9.91699174e-02 1.27962840e+00 4.95407999e-01 -3.61054093e-01 -4.65803653e-01 -1.13177371e+00 4.62846279e-01 8.04839909e-01 4.58614618e-01 -4.24706548e-01 6.15531564e-01 5.92336953e-01 3.03129613e-01 6.99602664e-01 -1.38563767e-01 -3.80229920e-01 2.69097481e-02 -9.17898595e-01 4.38152969e-01 1.62609577e-01 1.17625356e+00 5.81855297e-01 2.47379705e-01 -6.94683969e-01 7.14754283e-01 1.62864640e-01 9.03010741e-02 2.89838374e-01 -8.90448809e-01 3.70053768e-01 7.13930070e-01 -1.39746860e-01 -1.11219299e+00 -3.02805126e-01 -5.08739054e-01 -1.19871032e+00 1.34374768e-01 -1.45888895e-01 2.75656581e-01 -1.17704082e+00 1.56869078e+00 2.74085887e-02 8.21947992e-01 -1.86668858e-01 8.59454215e-01 1.07181358e+00 1.02883577e+00 1.32262250e-02 -1.05906069e-01 1.01018906e+00 -1.15084136e+00 -6.08040571e-01 2.86116302e-01 6.00250483e-01 -2.95712084e-01 4.76911128e-01 -2.00015102e-02 -1.45450318e+00 -1.21724129e+00 -9.41831470e-01 -2.73535788e-01 -2.77127147e-01 -6.37982041e-02 6.02907777e-01 -1.16645254e-01 -1.20405650e+00 6.29453659e-01 -1.03623867e+00 -8.73797238e-02 6.50580227e-01 1.81668922e-01 -5.30211806e-01 -3.04030091e-01 -1.29244530e+00 5.87293804e-01 4.02270645e-01 2.70168185e-01 -1.29021382e+00 -1.02628994e+00 -1.17732620e+00 1.95277765e-01 4.68814746e-02 -6.52162671e-01 9.52700675e-01 -6.63991749e-01 -1.23485243e+00 7.72899151e-01 -1.88021481e-01 -4.33444798e-01 3.84473771e-01 -1.76203605e-02 -2.69921660e-01 1.98087081e-01 -1.70661271e-01 7.83796370e-01 6.97718740e-01 -1.10319829e+00 -4.38499749e-01 -2.21919209e-01 4.16053772e-01 2.97672823e-02 -2.95679778e-01 -2.43605033e-01 -1.21522641e+00 -1.13775766e+00 3.64020057e-02 -8.09647858e-01 -3.06040913e-01 3.75913829e-01 -3.98649164e-02 -3.50992262e-01 1.12383354e+00 -5.07233500e-01 1.25138152e+00 -2.30625033e+00 4.11026686e-01 -2.15766981e-01 4.07285750e-01 6.20776534e-01 -4.37148929e-01 2.53342599e-01 -2.42189020e-01 1.22921742e-01 -1.00142524e-01 -5.60746849e-01 -1.74631644e-02 2.24657774e-01 -2.63636082e-01 4.12747115e-01 6.58786416e-01 1.27910817e+00 -7.23542094e-01 -2.48347774e-01 4.99505371e-01 9.25284803e-01 -7.31323242e-01 3.86349738e-01 -2.91237682e-01 3.33445996e-01 -5.28745353e-01 6.38657629e-01 8.96557748e-01 -1.73352018e-01 -9.01545137e-02 -3.54488790e-01 -2.27772266e-01 5.47625795e-02 -6.36030316e-01 2.11357450e+00 -4.01822448e-01 8.90965223e-01 -2.36468062e-01 -1.24480498e+00 9.76311803e-01 3.36844444e-01 8.49454343e-01 -1.08862555e+00 3.30472253e-02 -1.70020983e-01 -3.85685623e-01 -4.72305208e-01 3.63221675e-01 3.06044638e-01 2.09255572e-02 -4.22171764e-02 1.70922428e-01 2.86504507e-01 -9.14163515e-02 9.43230242e-02 1.16361141e+00 5.56130230e-01 -2.60787029e-02 -5.22094294e-02 6.72250450e-01 -3.74736905e-01 6.27934337e-01 6.29219592e-01 -1.44611061e-01 7.86518157e-01 6.40645802e-01 -8.48450005e-01 -1.12094605e+00 -7.85196960e-01 -9.29852203e-02 6.35362923e-01 3.56433451e-01 -4.09023106e-01 -2.71734506e-01 -5.41459739e-01 2.54094511e-01 1.76536575e-01 -1.01880157e+00 -3.81948680e-01 -8.30446482e-01 -3.34990233e-01 6.03265345e-01 1.01533890e+00 6.02341115e-01 -8.63503635e-01 -4.38227296e-01 4.77999628e-01 -4.78161592e-03 -1.24815798e+00 -5.01766920e-01 3.11973155e-01 -8.85447800e-01 -9.15423393e-01 -1.11440718e+00 -8.71784687e-01 2.68061191e-01 6.43095791e-01 8.50959241e-01 1.71263918e-01 -1.81679621e-01 2.13158831e-01 -6.68885946e-01 4.03858759e-02 1.43748313e-01 -1.19222149e-01 -1.21941306e-01 3.36882956e-02 3.28575522e-01 -9.46299374e-01 -6.08751595e-01 2.65708923e-01 -9.60404515e-01 1.86560228e-01 4.73679006e-01 8.91498625e-01 7.17801750e-01 -4.72537838e-02 4.02697563e-01 -4.46755290e-01 1.29226614e-02 -6.07700765e-01 -3.34606826e-01 9.29184332e-02 1.13008805e-01 -1.81355640e-01 7.06412315e-01 -7.41637111e-01 -7.93803036e-01 -2.08342634e-02 -2.54242063e-01 -1.54141200e+00 -1.06918655e-01 5.40501237e-01 -1.48168147e-01 -1.57355219e-01 2.36464143e-01 3.30983698e-01 -1.94886878e-01 -6.74023092e-01 1.85813561e-01 1.79968521e-01 4.77691382e-01 -4.49504972e-01 7.28792965e-01 5.47900736e-01 8.14274028e-02 -6.89087093e-01 -8.11170578e-01 -3.82260978e-01 -8.75077665e-01 -2.62941658e-01 1.00575316e+00 -1.28324533e+00 -7.58925319e-01 7.63080299e-01 -1.36680186e+00 -7.09836543e-01 -1.61394194e-01 4.93485391e-01 -5.82281768e-01 2.28837505e-01 -9.11092281e-01 -5.10499954e-01 7.81053826e-02 -1.14966989e+00 1.33539712e+00 3.23152840e-02 2.47606859e-01 -1.06183672e+00 -1.12565495e-02 1.88575499e-02 3.15906495e-01 7.25983620e-01 1.00341940e+00 -1.00550562e-01 -7.56066501e-01 -1.12567075e-01 -4.59421337e-01 3.20544124e-01 -4.55864333e-02 9.13550481e-02 -8.87392700e-01 -2.71555960e-01 -1.81272551e-01 -1.22468002e-01 1.35441852e+00 7.35444367e-01 1.60813761e+00 -7.31818676e-02 -3.54469955e-01 1.12362552e+00 1.29707134e+00 2.66328812e-01 7.37868965e-01 1.65687054e-01 1.03394890e+00 2.48913258e-01 3.68342698e-01 5.01555324e-01 3.39864939e-01 9.80947554e-01 6.19430184e-01 -2.08616912e-01 -3.61314416e-01 -4.24759835e-01 4.39254671e-01 9.85480785e-01 -4.40558076e-01 -4.75517362e-01 -5.33928454e-01 5.62140167e-01 -2.21134830e+00 -1.15296090e+00 -5.62080145e-02 1.55175138e+00 1.95690960e-01 2.43719444e-01 2.38559805e-02 -1.26283662e-02 6.17723286e-01 8.02387595e-01 -6.45514309e-01 -1.00604258e-01 -3.06540638e-01 1.10255070e-01 2.61897862e-01 2.13881016e-01 -1.27758527e+00 9.16374564e-01 5.97217607e+00 7.99997985e-01 -1.14026809e+00 -7.68954307e-02 4.50623333e-01 -3.01345259e-01 -3.14348191e-01 -1.61293671e-01 -5.97043276e-01 2.55873919e-01 5.21166563e-01 -7.32450187e-02 3.58387604e-02 7.47926950e-01 4.61625546e-01 5.58111608e-01 -1.32394373e+00 1.20435655e+00 3.59438732e-02 -1.83850718e+00 5.26274085e-01 1.00991681e-01 8.83162856e-01 -6.67030439e-02 3.98444384e-02 5.60004056e-01 -1.55717239e-01 -1.09993804e+00 8.39678705e-01 7.21963108e-01 9.31527674e-01 -8.19848120e-01 5.84313333e-01 1.52568743e-01 -1.75355256e+00 -1.36317790e-01 -4.21919048e-01 -7.32455552e-02 2.68405497e-01 2.57933766e-01 1.49569556e-01 6.78350270e-01 9.86762166e-01 1.72591257e+00 -3.65756154e-01 9.89820182e-01 1.09219380e-01 3.68047476e-01 8.78223851e-02 2.28864759e-01 8.70253146e-01 1.38887197e-01 5.20940840e-01 1.46383178e+00 4.18721586e-01 2.60639668e-01 2.14346334e-01 7.42978930e-01 -4.27961290e-01 -2.68165320e-01 -7.80800223e-01 7.34100565e-02 7.29820728e-02 7.56880879e-01 -2.67472327e-01 -3.53630424e-01 -7.40446627e-01 1.02840507e+00 3.90743405e-01 5.97428262e-01 -1.11766648e+00 -2.36579642e-01 1.16859388e+00 2.10715055e-01 8.17361951e-01 -5.51544249e-01 3.67077976e-01 -1.32924557e+00 2.39536434e-01 -5.87405026e-01 3.00946712e-01 -8.46477985e-01 -1.20088196e+00 5.93542159e-01 5.26075549e-02 -1.43222725e+00 5.56134805e-02 -7.89126754e-01 -6.51088238e-01 5.94578683e-01 -1.72962201e+00 -1.44364023e+00 -5.58689415e-01 9.70858812e-01 8.56525600e-01 1.72987580e-02 4.82025266e-01 5.90861678e-01 -4.60963428e-01 5.82121313e-01 2.16488373e-02 3.35862964e-01 3.25508416e-01 -7.57416725e-01 7.04032540e-01 6.69455647e-01 -6.48339167e-02 4.23076600e-01 2.51583476e-02 -4.64707822e-01 -1.68079293e+00 -1.58103216e+00 5.89951396e-01 -2.23631352e-01 4.84396279e-01 -5.40269375e-01 -1.23597658e+00 8.23404491e-01 -8.26821402e-02 5.73368788e-01 2.95645416e-01 -8.57958645e-02 -7.56926894e-01 -7.73448870e-02 -6.61950409e-01 5.53965151e-01 1.65199924e+00 -7.96863079e-01 -4.08197761e-01 1.33375943e-01 1.16076231e+00 -3.88776600e-01 -1.08081520e+00 7.25391448e-01 7.69005537e-01 -9.16998982e-01 1.22400677e+00 -7.27168858e-01 6.24542356e-01 -4.17085886e-01 -3.19462389e-01 -9.24152076e-01 -8.35520864e-01 -3.44525427e-01 -7.23133564e-01 9.63665605e-01 -7.84481391e-02 1.60321873e-02 7.03405738e-01 1.59462154e-01 -5.79988599e-01 -9.38883901e-01 -8.96895111e-01 -9.97445941e-01 6.86964318e-02 -7.23820150e-01 5.84737837e-01 9.49839175e-01 -2.93775797e-01 5.25224954e-02 -8.13447416e-01 1.68264598e-01 4.32729483e-01 1.64027765e-01 6.78025603e-01 -8.39843750e-01 -1.17899768e-01 -7.47772694e-01 -1.00539970e+00 -1.83576739e+00 4.04325336e-01 -8.75765622e-01 -1.79531485e-01 -1.36356223e+00 2.08886728e-01 -1.74644649e-01 -6.92072213e-01 3.53020549e-01 1.07670529e-02 4.79200184e-01 2.38108426e-01 1.42951861e-01 -7.08126664e-01 1.04126704e+00 1.50484896e+00 -5.29735744e-01 -2.44960368e-01 -2.56954432e-01 -2.41806805e-01 4.46688920e-01 2.07635492e-01 -7.59390593e-02 -4.90272284e-01 -8.58463228e-01 -2.70908356e-01 2.90911645e-01 6.52437747e-01 -1.07567525e+00 2.93670952e-01 -1.35095716e-01 7.66529918e-01 -9.66598153e-01 5.60699463e-01 -7.51863182e-01 3.22421610e-01 2.01449350e-01 -4.78122801e-01 -1.70578405e-01 3.98734510e-01 7.35229492e-01 -7.09607184e-01 3.69040430e-01 7.54066527e-01 -1.70514271e-01 -1.02276099e+00 1.13510764e+00 -5.36426187e-01 -3.06810439e-01 1.15543258e+00 -2.18946800e-01 4.63979691e-02 -2.53325403e-01 -8.17676902e-01 2.79064208e-01 5.11446655e-01 8.13126147e-01 8.04593027e-01 -1.91667998e+00 -6.28417194e-01 1.62810773e-01 3.20694387e-01 1.57423064e-01 7.99872994e-01 4.74496543e-01 -5.29886425e-01 6.17596567e-01 -4.37869757e-01 -8.57220590e-01 -9.03032959e-01 8.99760365e-01 4.20191884e-01 -1.13884628e-01 -1.01743925e+00 7.20950186e-01 5.52824378e-01 -6.05289303e-02 5.79137027e-01 -5.58807015e-01 -2.26261750e-01 7.33154593e-03 6.41277194e-01 -6.54426068e-02 -2.69935101e-01 -7.93924212e-01 -4.36823487e-01 9.24401820e-01 2.40884461e-02 4.48457271e-01 1.71631229e+00 -3.92384967e-03 1.36765689e-01 4.68650281e-01 1.72377181e+00 -6.52879953e-01 -1.73528194e+00 -4.47212070e-01 -2.74165630e-01 -5.69758654e-01 1.18526891e-01 1.30145838e-02 -1.44009364e+00 1.11804819e+00 2.91698992e-01 2.08687223e-02 1.16441941e+00 -9.70987882e-03 9.64337528e-01 2.68283486e-01 2.36379534e-01 -5.80489099e-01 3.42076659e-01 7.39692628e-01 9.42278564e-01 -9.17997420e-01 -1.51194036e-01 -1.05804190e-01 -3.24429661e-01 1.22391713e+00 8.02813053e-01 -4.54704046e-01 1.02664220e+00 6.07468449e-02 -3.78370106e-01 -2.03117415e-01 -1.17024648e+00 -1.32425696e-01 4.92494613e-01 6.32244706e-01 3.80440772e-01 -2.98244655e-01 1.75256908e-01 4.29191113e-01 7.51526177e-01 8.44223127e-02 1.19763926e-01 8.31617236e-01 3.60622182e-02 -7.33395040e-01 1.83783680e-01 1.21825933e-01 -1.64874524e-01 7.71236941e-02 -4.47270758e-02 8.49754274e-01 4.08991456e-01 4.50298727e-01 3.85008365e-01 -7.30858088e-01 5.31362832e-01 -2.06081584e-01 3.85682166e-01 -2.43051276e-01 -4.41723347e-01 1.59328997e-01 -2.67884046e-01 -8.70648921e-01 -7.55511045e-01 -4.92761672e-01 -9.37023580e-01 -4.47595239e-01 8.61066356e-02 -2.10902587e-01 2.67258316e-01 6.69185281e-01 5.40558636e-01 7.44812250e-01 8.20349514e-01 -1.41351914e+00 1.68326735e-01 -7.39742875e-01 -4.87183720e-01 4.54375595e-01 6.70513868e-01 -8.17691088e-01 -1.69743747e-01 1.47407487e-01]

[8.927339553833008, 0.42889800667762756]

6f21ad73-ba0b-4dfc-90ad-34eaa69a0161

optimal-estimation-of-low-rank-density

1507.05131

http://arxiv.org/abs/1507.05131v4

http://arxiv.org/pdf/1507.05131v4.pdf

Optimal Estimation of Low Rank Density Matrices

The density matrices are positively semi-definite Hermitian matrices of unit trace that describe the state of a quantum system. The goal of the paper is to develop minimax lower bounds on error rates of estimation of low rank density matrices in trace regression models used in quantum state tomography (in particular, in the case of Pauli measurements) with explicit dependence of the bounds on the rank and other complexity parameters. Such bounds are established for several statistically relevant distances, including quantum versions of Kullback-Leibler divergence (relative entropy distance) and of Hellinger distance (so called Bures distance), and Schatten $p$-norm distances. Sharp upper bounds and oracle inequalities for least squares estimator with von Neumann entropy penalization are obtained showing that minimax lower bounds are attained (up to logarithmic factors) for these distances.

['Dong Xia', 'Vladimir Koltchinskii']

2015-07-17

['quantum-state-tomography']

['medical']

[ 2.30567768e-01 3.31958950e-01 5.63987605e-02 -4.00655299e-01 -1.03098488e+00 -6.10770047e-01 3.08184117e-01 1.27131283e-01 -7.85045564e-01 8.89410734e-01 -2.80560348e-02 -3.65034312e-01 -7.76468217e-01 -6.66728497e-01 -4.28958833e-01 -8.59274685e-01 -4.56863225e-01 6.55344605e-01 -1.01663403e-01 -2.55055845e-01 5.34277678e-01 6.07646644e-01 -9.42598641e-01 -6.40625834e-01 7.94183731e-01 8.55847180e-01 -3.93277168e-01 9.08531666e-01 6.52220786e-01 6.66556954e-01 -1.32927060e-01 -5.16956627e-01 5.23523808e-01 -6.58894181e-01 -7.92417347e-01 -1.75980613e-01 7.09518194e-02 -6.89211935e-02 -8.68843436e-01 1.70814073e+00 1.09544076e-01 1.35586843e-01 1.02915931e+00 -8.41234863e-01 -4.24245358e-01 7.55822122e-01 -3.16321045e-01 2.48570770e-01 2.92043149e-01 -2.73572952e-01 1.36496961e+00 -6.92744493e-01 4.53116447e-01 9.18758333e-01 4.84942019e-01 2.31263742e-01 -1.52547503e+00 -4.98962760e-01 -1.06659210e+00 3.60102057e-01 -1.96719325e+00 -5.24287462e-01 1.15012206e-01 -5.66277981e-01 6.95647061e-01 4.29972321e-01 1.37643844e-01 5.57026684e-01 4.31143343e-01 2.64054626e-01 1.18755555e+00 -5.85050046e-01 2.42273033e-01 4.59036022e-01 2.18745589e-01 1.01635051e+00 8.67562175e-01 2.95942217e-01 -3.78543139e-01 -2.84152210e-01 6.23591661e-01 -3.90065491e-01 -2.39941955e-01 -6.73647642e-01 -1.41850626e+00 1.05399144e+00 2.91671585e-02 4.05732721e-01 -2.29572549e-01 2.44736850e-01 1.11424707e-01 4.43596542e-01 -1.43372968e-01 3.21433663e-01 9.83653143e-02 -1.59941554e-01 -5.31907856e-01 -7.61834905e-02 1.29230404e+00 1.07987630e+00 1.04552412e+00 -2.00364128e-01 -1.05639368e-01 2.04530045e-01 4.03995007e-01 1.23284471e+00 -2.55294949e-01 -9.65526283e-01 6.37006342e-01 -5.77884577e-02 3.62293392e-01 -7.53612578e-01 -3.91418964e-01 -2.76333511e-01 -9.20068741e-01 -2.93002605e-01 4.40038651e-01 -1.94043085e-01 -1.92499489e-01 1.71615160e+00 -7.56144105e-03 -1.67246565e-01 2.66866535e-01 7.78354526e-01 -7.22698793e-02 6.59578919e-01 -4.59203154e-01 -4.83927339e-01 6.36638761e-01 -1.37771338e-01 -7.47827649e-01 -1.04529530e-01 9.48522151e-01 -4.10043061e-01 5.54701149e-01 2.87388593e-01 -8.77873182e-01 1.10355906e-01 -1.20496714e+00 2.05917880e-01 2.13644534e-01 -1.02729693e-01 5.16867697e-01 1.18488169e+00 -7.48200774e-01 1.04301834e+00 -6.79312408e-01 -4.08859365e-02 -2.71681905e-01 7.28296518e-01 -6.92803025e-01 6.94184750e-02 -1.11842775e+00 9.42670166e-01 4.72360462e-01 4.13843274e-01 -8.34589660e-01 -1.57867655e-01 -7.74956048e-01 8.54910389e-02 1.42754808e-01 -1.26917213e-01 8.26258719e-01 -3.68681133e-01 -1.42368662e+00 7.51276016e-01 3.11404914e-01 -4.53165442e-01 2.00879112e-01 2.74702966e-01 -3.53147089e-01 1.04905695e-01 -1.65876932e-02 -2.94340461e-01 3.64719123e-01 -4.43018258e-01 -1.04219310e-01 -7.60150075e-01 5.77653795e-02 8.89364928e-02 1.00526445e-01 -1.71576902e-01 2.02514529e-01 4.66569185e-01 5.29590964e-01 -1.27554154e+00 -3.40693325e-01 -3.94480646e-01 -4.56316292e-01 1.30596235e-01 7.58328065e-02 -4.86382723e-01 1.17004454e+00 -2.06054878e+00 3.47103983e-01 7.90146053e-01 -1.32945150e-01 -1.23636201e-01 -6.09893687e-02 7.03267574e-01 1.38894364e-01 -1.98765054e-01 -3.45963955e-01 1.81534573e-01 4.60577637e-01 1.84565976e-01 -8.00709799e-03 1.25930643e+00 -1.28234908e-01 3.46266091e-01 -7.30436921e-01 -2.56114870e-01 1.83544993e-01 1.61474228e-01 -5.87098181e-01 -2.39189610e-01 3.79851580e-01 2.84024298e-01 -4.64863062e-01 -1.14982814e-01 5.57618678e-01 -3.55814427e-01 2.96305299e-01 -1.32049784e-01 -2.49610413e-02 3.53379458e-01 -1.55618513e+00 1.34667230e+00 -3.10257524e-01 5.22513032e-01 2.13669792e-01 -7.99523413e-01 6.34145439e-01 3.26668471e-01 2.26297602e-01 -4.18891817e-01 6.14256680e-01 3.73228699e-01 3.22885871e-01 -2.39108533e-01 5.72903931e-01 -6.98912024e-01 -4.63008016e-01 4.87626046e-01 7.89666101e-02 -1.54987365e-01 4.39584017e-01 3.79668742e-01 1.05364823e+00 -5.32237530e-01 4.44374919e-01 -7.56991684e-01 7.49168396e-01 -5.68945527e-01 3.02517503e-01 8.79668832e-01 -2.07979783e-01 2.72632360e-01 8.31007600e-01 3.04093659e-01 -1.16596818e+00 -1.29019082e+00 -5.65617025e-01 5.74743986e-01 2.76731908e-01 -4.50394183e-01 -5.35891414e-01 -4.47815768e-02 -3.01916637e-02 6.43455207e-01 -5.97040772e-01 -3.87240082e-01 -3.03458422e-02 -1.11009133e+00 6.42267168e-01 -1.62291545e-02 3.60083580e-01 -3.73449355e-01 -8.77159461e-02 -6.65062293e-02 2.58968800e-01 -1.12548888e+00 -3.14998031e-01 3.90442342e-01 -7.94931591e-01 -1.16889644e+00 -2.59164006e-01 -7.20258579e-02 5.37526250e-01 -2.92770743e-01 3.82932812e-01 -7.27968812e-01 -1.45647123e-01 5.29110074e-01 1.02665141e-01 1.39805049e-01 -6.17098987e-01 -1.19230099e-01 4.73309755e-01 1.11191370e-01 2.86277652e-01 -2.27673039e-01 -4.51540112e-01 3.65095317e-01 -7.98475683e-01 -5.46712875e-01 6.77845895e-01 7.94480264e-01 5.67980230e-01 -1.85701579e-01 -1.86761692e-01 -9.32501554e-01 5.70763707e-01 -2.51740843e-01 -1.30381417e+00 1.91607267e-01 -7.49978781e-01 6.83841646e-01 3.72481614e-01 1.77784443e-01 -6.01150632e-01 -3.73843014e-02 2.00106472e-01 8.09656084e-02 4.11898524e-01 4.83250171e-01 7.47907907e-02 -5.16092479e-01 8.70979965e-01 3.18204999e-01 -3.67326468e-01 -7.01383576e-02 2.88395435e-01 7.99980342e-01 3.87327433e-01 -4.66302067e-01 8.98341179e-01 3.68149251e-01 7.63555944e-01 -9.66513574e-01 -1.00748014e+00 -5.54437757e-01 -7.79704094e-01 3.10190320e-01 7.11211979e-01 -7.25998640e-01 -1.06341612e+00 -6.57048449e-02 -5.29556990e-01 -2.62202471e-02 -2.07331926e-01 1.14296806e+00 -9.34892058e-01 7.11495101e-01 -8.54984701e-01 -1.19143486e+00 -1.68924313e-02 -9.12715316e-01 6.37780547e-01 1.21640712e-02 9.12705511e-02 -1.10778093e+00 4.24444735e-01 1.99694857e-01 2.33396694e-01 4.44063451e-03 7.63731837e-01 -5.10154426e-01 -7.44917154e-01 -6.28889978e-01 -4.70071912e-01 5.63979268e-01 -5.40009737e-01 -3.18789035e-01 -5.21605372e-01 -4.01052266e-01 1.24729544e-01 6.06979616e-02 5.36710918e-01 1.39961228e-01 4.90987062e-01 -5.11400640e-01 2.73297578e-02 4.52093482e-01 1.48237574e+00 -5.30289114e-02 7.16687024e-01 -9.39088985e-02 4.18462098e-01 1.26646951e-01 4.06943440e-01 6.92368031e-01 -8.12220294e-03 3.89500082e-01 4.47886549e-02 8.64220381e-01 8.14013720e-01 4.26119799e-03 5.24231315e-01 1.43664932e+00 -6.43868744e-02 1.27807707e-01 -6.76098645e-01 2.32623369e-01 -1.54757380e+00 -1.16260552e+00 -4.34004128e-01 2.87179446e+00 6.49576843e-01 -2.15036515e-02 -1.25816688e-01 2.55085319e-01 6.92858100e-01 -5.13813421e-02 -2.50072390e-01 -4.91134822e-01 -5.10707311e-02 3.95676017e-01 1.34157825e+00 1.11229694e+00 -9.21491981e-01 4.14909452e-01 7.01071930e+00 5.84824562e-01 -4.73030001e-01 2.22431049e-01 -9.77012813e-02 1.14653356e-01 -2.71995604e-01 4.30696189e-01 -6.79409623e-01 3.14875215e-01 1.45143270e+00 -3.67589116e-01 6.93691671e-01 5.90128660e-01 -1.93720743e-01 -3.76325816e-01 -1.12000167e+00 1.36341476e+00 -1.01892836e-01 -9.27791715e-01 -6.33806169e-01 6.50118828e-01 9.46640193e-01 3.39601368e-01 1.12572633e-01 2.20294923e-01 5.76355308e-02 -7.78902709e-01 3.74173224e-01 4.51653808e-01 9.25730050e-01 -9.61500049e-01 8.34289193e-01 3.08444381e-01 -6.86734617e-01 3.74222398e-02 -7.36368299e-01 -2.89220661e-02 5.74452989e-03 7.48158276e-01 -6.00776017e-01 1.66045696e-01 -2.29544625e-01 2.85948664e-01 -9.93057899e-03 1.06748724e+00 -9.57874656e-02 3.95168751e-01 -6.01071179e-01 -3.73945802e-01 4.33614671e-01 -1.11658704e+00 4.62377697e-01 1.07432592e+00 4.59675610e-01 3.56475800e-01 -5.28437078e-01 8.60297143e-01 -1.90777212e-01 2.07290828e-01 -6.17788672e-01 -4.62095112e-01 3.80806893e-01 1.01992595e+00 -3.49934578e-01 3.63656320e-02 -2.18073025e-01 1.00884104e+00 -1.49515808e-01 4.28953059e-02 -6.03565693e-01 -8.41955185e-01 5.77855885e-01 6.73516765e-02 2.76366115e-01 -4.35869455e-01 4.41802554e-02 -1.33444345e+00 -2.08543107e-01 -4.28337246e-01 2.37270832e-01 -2.57122964e-01 -9.92735088e-01 3.57007422e-02 -1.18197098e-01 -8.42679501e-01 -3.26052606e-01 -7.81532824e-01 -1.04326181e-01 9.40642953e-01 -8.35691690e-01 -7.40629658e-02 3.72863770e-01 5.41911781e-01 -8.36958170e-01 -2.58860681e-02 9.23954248e-01 3.72058421e-01 -5.14293790e-01 3.28314573e-01 1.06245148e+00 2.27766365e-01 2.20776379e-01 -1.37979460e+00 -1.15434304e-01 7.80755997e-01 4.25108194e-01 6.98337257e-01 1.02365351e+00 -3.11204433e-01 -1.92023849e+00 -5.40887415e-01 8.04649532e-01 -2.93373764e-01 1.18324089e+00 -1.63316846e-01 -3.02868843e-01 9.27165389e-01 -3.30410242e-01 -5.72552755e-02 1.01637852e+00 4.04318243e-01 -5.04658580e-01 -1.81262046e-01 -1.14204824e+00 1.48458257e-01 5.67232728e-01 -1.19549036e+00 -6.08403012e-02 7.21676707e-01 -2.20385268e-01 -1.69912308e-01 -1.06450140e+00 2.20074758e-01 5.37893474e-01 -1.00955820e+00 5.22624552e-01 -3.06926072e-01 -3.09301227e-01 -2.83691406e-01 -5.54528356e-01 -7.62566149e-01 -3.08315217e-01 -1.10290599e+00 2.04184756e-01 2.09915921e-01 5.98594785e-01 -5.19115329e-01 7.46511042e-01 6.19640648e-01 3.27652216e-01 -2.90748239e-01 -1.29199672e+00 -8.75740886e-01 -3.49839628e-02 -2.90124208e-01 -1.84035659e-01 6.82753980e-01 7.32824981e-01 6.80233002e-01 -5.70862651e-01 2.44339079e-01 1.07507837e+00 -1.49757162e-01 4.86119062e-01 -1.22620690e+00 -6.34318888e-01 -1.11468866e-01 -1.13828194e+00 -1.07634330e+00 7.14502931e-02 -1.17883074e+00 1.45562276e-01 -9.26574588e-01 5.11177361e-01 -3.22108030e-01 -4.55601096e-01 -2.96639591e-01 3.16092342e-01 1.51893884e-01 1.03685863e-01 2.16870848e-02 -1.01719415e+00 5.56006014e-01 8.50664794e-01 1.46159276e-01 2.66255081e-01 2.31651440e-01 -3.47730428e-01 5.59893966e-01 4.32381004e-01 -7.38807380e-01 -1.84475824e-01 -1.42215952e-01 8.75136435e-01 7.12170839e-01 8.54283497e-02 -1.19049978e+00 5.98946549e-02 -1.41084054e-02 -2.09423527e-01 -3.96079093e-01 4.39042598e-01 -3.76115501e-01 3.39127690e-01 5.64043105e-01 -5.47806263e-01 -2.48574689e-01 -5.73674977e-01 7.68104017e-01 -3.22829820e-02 -8.90088201e-01 1.09313226e+00 2.29196846e-01 -1.37039736e-01 4.30657625e-01 -3.76797467e-01 1.28012970e-01 7.80671835e-01 1.22548021e-01 -5.19491620e-02 -6.74205065e-01 -7.78636634e-01 -2.48966500e-01 2.74796695e-01 -4.07497078e-01 3.17087919e-01 -1.18991172e+00 -5.58355451e-01 7.42506236e-02 1.78194851e-01 -7.53990471e-01 1.80868477e-01 1.34933150e+00 -8.57832015e-01 8.76491845e-01 2.03529815e-03 -2.92089075e-01 -9.48287249e-01 2.58410722e-01 4.65427756e-01 -1.52342767e-01 -7.29618222e-02 7.21562326e-01 -1.00765200e-02 -2.35304043e-01 6.57833740e-02 -1.18274204e-01 6.97407246e-01 -3.70125175e-01 5.69086194e-01 7.00498462e-01 -1.22279301e-02 -9.23548698e-01 -3.54189396e-01 3.45899403e-01 1.58544376e-01 -4.46347326e-01 9.90275741e-01 -3.00974637e-01 -4.44657832e-01 6.36261523e-01 1.67417800e+00 2.16499045e-01 -8.31006348e-01 -3.98033649e-01 2.52934337e-01 -3.00840169e-01 -3.79861258e-02 -1.56510577e-01 -4.84893441e-01 1.09961164e+00 7.41663098e-01 2.60715991e-01 3.20591718e-01 3.59436050e-02 5.15131712e-01 1.22534144e+00 7.64761865e-01 -1.09385824e+00 -5.72895885e-01 7.55522132e-01 1.92056105e-01 -1.01221406e+00 9.59905982e-02 -1.63319528e-01 -3.43470246e-01 9.86090124e-01 -3.29737008e-01 -3.13945979e-01 7.72497356e-01 -3.26617844e-02 -5.73843241e-01 -1.20220266e-01 -3.18241537e-01 -2.85252661e-01 2.72316545e-01 3.55873287e-01 3.31618100e-01 4.45935547e-01 -4.95414406e-01 -4.30719294e-02 -2.64243394e-01 -4.88422960e-01 9.66140568e-01 3.14758539e-01 -7.59741426e-01 -8.89383495e-01 -1.29669100e-01 5.27574360e-01 -3.83483678e-01 -1.59880713e-01 -1.44745126e-01 4.71275270e-01 -5.64622164e-01 8.18518758e-01 -3.25038612e-01 -2.04763710e-01 1.56184221e-02 -1.19166628e-01 1.12996054e+00 -5.55045247e-01 1.06275417e-01 -3.99270445e-01 8.22170004e-02 -4.27515179e-01 -1.98466152e-01 -5.90058863e-01 -1.01385581e+00 -7.32972085e-01 -7.74318635e-01 7.51612902e-01 7.60160625e-01 9.93043303e-01 -6.42592683e-02 -4.69702631e-01 7.04218686e-01 -2.48681590e-01 -1.32247543e+00 -9.43737447e-01 -1.40023565e+00 1.60024643e-01 1.19844072e-01 -8.75738338e-02 -7.79308915e-01 -4.46052730e-01]

[5.700873374938965, 4.8992204666137695]

110064b7-f6da-4aab-898b-3f86087a9a6b

cvt-slr-contrastive-visual-textual

2303.05725

https://arxiv.org/abs/2303.05725v4

https://arxiv.org/pdf/2303.05725v4.pdf

CVT-SLR: Contrastive Visual-Textual Transformation for Sign Language Recognition with Variational Alignment

Sign language recognition (SLR) is a weakly supervised task that annotates sign videos as textual glosses. Recent studies show that insufficient training caused by the lack of large-scale available sign datasets becomes the main bottleneck for SLR. Most SLR works thereby adopt pretrained visual modules and develop two mainstream solutions. The multi-stream architectures extend multi-cue visual features, yielding the current SOTA performances but requiring complex designs and might introduce potential noise. Alternatively, the advanced single-cue SLR frameworks using explicit cross-modal alignment between visual and textual modalities are simple and effective, potentially competitive with the multi-cue framework. In this work, we propose a novel contrastive visual-textual transformation for SLR, CVT-SLR, to fully explore the pretrained knowledge of both the visual and language modalities. Based on the single-cue cross-modal alignment framework, we propose a variational autoencoder (VAE) for pretrained contextual knowledge while introducing the complete pretrained language module. The VAE implicitly aligns visual and textual modalities while benefiting from pretrained contextual knowledge as the traditional contextual module. Meanwhile, a contrastive cross-modal alignment algorithm is designed to explicitly enhance the consistency constraints. Extensive experiments on public datasets (PHOENIX-2014 and PHOENIX-2014T) demonstrate that our proposed CVT-SLR consistently outperforms existing single-cue methods and even outperforms SOTA multi-cue methods.

['Stan Z. Li', 'Yidong Chen', 'Jun Xia', 'Ge Wang', 'Siyuan Li', 'Cheng Tan', 'Yile Wang', 'Jiangbin Zheng']

2023-03-10

http://openaccess.thecvf.com//content/CVPR2023/html/Zheng_CVT-SLR_Contrastive_Visual-Textual_Transformation_for_Sign_Language_Recognition_With_Variational_CVPR_2023_paper.html

http://openaccess.thecvf.com//content/CVPR2023/papers/Zheng_CVT-SLR_Contrastive_Visual-Textual_Transformation_for_Sign_Language_Recognition_With_Variational_CVPR_2023_paper.pdf

cvpr-2023-1

['sign-language-recognition']

['computer-vision']

[ 2.74699837e-01 -5.74161887e-01 -5.19139349e-01 -3.63222152e-01 -9.43522215e-01 -3.64906788e-01 7.18608022e-01 -8.95191252e-01 -7.05730259e-01 3.43699306e-01 5.50374985e-01 -1.03131488e-01 -3.95917669e-02 -1.39450729e-01 -7.61302233e-01 -6.28141403e-01 5.84807098e-01 1.20964020e-01 2.14363575e-01 -4.27273735e-02 -7.99630955e-02 1.68244049e-01 -1.78319728e+00 4.48598742e-01 1.10521150e+00 8.59739006e-01 3.53834867e-01 5.32956123e-01 -2.14013815e-01 8.64259779e-01 -2.05064327e-01 -5.64213037e-01 2.96284825e-01 -4.45653528e-01 -3.17927539e-01 1.74554169e-01 1.02992797e+00 -4.94541705e-01 -6.71329677e-01 9.96560514e-01 7.17042029e-01 1.80754825e-01 5.93324721e-01 -1.43818021e+00 -1.08465827e+00 4.01486427e-01 -6.83278441e-01 -2.22686827e-01 1.20819435e-01 5.61715722e-01 1.32715428e+00 -1.18290186e+00 8.82261276e-01 1.20024323e+00 5.45399964e-01 8.71088088e-01 -9.94724870e-01 -6.53259754e-01 5.72677195e-01 5.64562678e-01 -1.23475766e+00 -2.53861785e-01 9.05165136e-01 -3.95989001e-01 7.61333287e-01 2.40266040e-01 7.84101427e-01 1.58029008e+00 -4.19670582e-01 1.44692433e+00 1.48739815e+00 -3.85066748e-01 -2.14966267e-01 -1.00075185e-01 8.59866366e-02 6.70276880e-01 7.75832236e-02 1.52419969e-01 -7.70583987e-01 1.41206503e-01 7.59130180e-01 2.11116135e-01 -3.42511863e-01 -3.95015508e-01 -1.28961968e+00 4.64660764e-01 3.23330492e-01 3.55251193e-01 -3.04767787e-01 1.91787854e-01 4.94470060e-01 1.10860690e-01 -5.55849336e-02 -8.27069581e-02 -3.95709991e-01 2.06686687e-02 -9.16852355e-01 4.78136586e-03 1.67794630e-01 1.14122558e+00 4.37292337e-01 3.41551840e-01 -5.64752698e-01 9.58003104e-01 7.08857059e-01 9.77283120e-01 5.82960486e-01 -9.06183958e-01 6.44901514e-01 6.32900417e-01 -3.68238121e-01 -5.62248588e-01 -8.74234438e-02 -2.85894215e-01 -7.70430684e-01 1.97474241e-01 4.52759743e-01 1.38167262e-01 -1.45006347e+00 1.78260159e+00 1.02075659e-01 4.22002256e-01 9.32284445e-02 1.26678360e+00 1.16636884e+00 2.73805052e-01 3.62226933e-01 7.07607567e-02 1.45326138e+00 -1.10243952e+00 -9.15775478e-01 -1.99299440e-01 2.66105235e-01 -7.57352948e-01 1.51564944e+00 2.16904461e-01 -7.20630884e-01 -4.33401585e-01 -8.28095973e-01 -4.84752864e-01 -3.42913002e-01 6.99324906e-01 6.76994145e-01 2.51374751e-01 -6.65736556e-01 -8.01287293e-02 -8.60056579e-01 -2.87493080e-01 4.00731176e-01 -6.56032562e-02 -5.67666352e-01 -4.64064151e-01 -9.41268682e-01 8.97535861e-01 4.61259335e-02 5.49186885e-01 -6.92706943e-01 -7.27258265e-01 -1.05562437e+00 -4.53639716e-01 6.24533594e-01 -6.77798092e-01 9.62429702e-01 -9.36023951e-01 -1.69286919e+00 8.61467540e-01 -3.82160932e-01 -2.96138767e-02 9.25861299e-01 -4.26884711e-01 -4.33277875e-01 1.36990070e-01 -1.49470910e-01 5.65531552e-01 1.01769543e+00 -1.47059274e+00 -5.32152951e-01 -1.84695095e-01 -1.46999732e-01 2.60428160e-01 -1.40173122e-01 7.67167211e-02 -9.21765685e-01 -8.35258245e-01 1.69115469e-01 -9.01421130e-01 2.44627204e-02 1.16300412e-01 -3.30048233e-01 -3.74638468e-01 8.90352428e-01 -8.96367967e-01 1.04887867e+00 -1.98939407e+00 3.19978863e-01 1.74196020e-01 1.35563806e-01 4.43117172e-01 -4.50766355e-01 5.82680888e-02 7.19389617e-02 -2.42403373e-01 -4.10567045e-01 -4.78631645e-01 3.27068001e-01 5.46443880e-01 -3.58452320e-01 4.02473688e-01 2.27374718e-01 1.28497624e+00 -9.07999516e-01 -7.90570498e-01 4.96415436e-01 7.96543360e-01 -6.57476902e-01 2.18107656e-01 -2.02921540e-01 5.72902501e-01 -1.52223364e-01 1.12387550e+00 5.38499951e-01 -2.80674547e-01 2.38942415e-01 -6.98125720e-01 -6.40104339e-02 -3.66678387e-02 -1.02281058e+00 1.84183991e+00 -2.82390386e-01 8.67115557e-01 -1.07870214e-01 -7.92979002e-01 6.78909600e-01 2.11311489e-01 4.90263522e-01 -8.67050588e-01 3.18973154e-01 4.02508527e-01 -1.40084818e-01 -9.55964506e-01 3.43774050e-01 -1.60684690e-01 3.41626257e-02 1.35957420e-01 3.77741575e-01 2.14570656e-01 1.36419609e-01 2.44345441e-01 7.26053655e-01 6.70391560e-01 -1.16428165e-02 4.20016885e-01 4.72879291e-01 -3.10922742e-01 8.48707974e-01 5.88658631e-01 -6.10534549e-01 6.50224805e-01 7.89944157e-02 2.95777116e-02 -8.45244467e-01 -1.11276758e+00 -6.80778623e-02 1.15842283e+00 1.51125684e-01 -4.09631580e-01 -2.88368851e-01 -9.07421768e-01 1.10694915e-01 4.75254834e-01 -6.50195420e-01 2.23449543e-01 -7.34282494e-01 -4.22783345e-01 7.74086475e-01 8.84578943e-01 6.34091616e-01 -1.03199637e+00 -4.37734067e-01 -3.58529955e-01 -5.03871858e-01 -1.37113988e+00 -8.36263657e-01 -2.66866595e-01 -4.74768132e-01 -1.06049943e+00 -1.12369359e+00 -8.06543529e-01 6.00308955e-01 1.25806972e-01 5.09893000e-01 -5.38051948e-02 -3.25928360e-01 7.18259335e-01 -5.65495849e-01 -6.50667325e-02 -3.99736203e-02 -3.31414908e-01 1.44133465e-02 3.33123147e-01 6.40302658e-01 -4.75217998e-01 -5.39735794e-01 2.26765990e-01 -9.64560330e-01 1.17534399e-01 1.08850992e+00 1.16309583e+00 7.83889532e-01 -8.91236186e-01 3.07661593e-01 -4.92980957e-01 3.57947201e-01 -6.91440627e-02 -3.95155281e-01 7.05700696e-01 -5.54891348e-01 3.30009460e-01 3.03255469e-01 -8.95606279e-01 -1.28135359e+00 -1.09067848e-02 -1.25161335e-01 -9.50861692e-01 -1.06293336e-01 5.70172012e-01 -4.98764247e-01 -1.42357056e-03 2.19567344e-01 5.28235972e-01 -7.75108710e-02 -6.19972348e-01 8.13325047e-01 7.44374514e-01 8.35471570e-01 -6.20468438e-01 8.65856171e-01 5.68163514e-01 -1.82150573e-01 -6.36525691e-01 -6.96940482e-01 -5.03330588e-01 -7.96486378e-01 -5.77040493e-01 8.35099757e-01 -8.02307308e-01 -7.22080410e-01 6.53021276e-01 -8.64674747e-01 -3.99255872e-01 -1.15614578e-01 7.89223254e-01 -4.64831620e-01 7.57565916e-01 -4.15694207e-01 -6.87121809e-01 -1.58697635e-01 -1.20322120e+00 1.07870865e+00 1.76170975e-01 1.36114419e-01 -6.44014180e-01 9.17189941e-02 6.82666957e-01 3.25259209e-01 1.09777436e-01 5.20846546e-01 -3.57550472e-01 -7.21341908e-01 -1.70359276e-02 -5.79308808e-01 6.44201100e-01 1.20696835e-01 1.15401044e-01 -9.72622275e-01 -2.27602869e-01 -6.52341783e-01 -5.22914171e-01 1.18418789e+00 3.79707098e-01 7.99113154e-01 -1.33914158e-01 -7.25180581e-02 7.39884913e-01 1.32570720e+00 -4.06672284e-02 5.52376509e-01 2.87664145e-01 1.16907096e+00 5.04842699e-01 5.30000150e-01 1.06857784e-01 7.81289339e-01 6.65110230e-01 2.14800499e-02 -1.10205226e-02 -6.48763597e-01 -6.18998349e-01 6.03020787e-01 1.29565895e+00 -4.12060231e-01 1.25479445e-01 -7.63364494e-01 6.40646577e-01 -2.08697462e+00 -1.09410727e+00 3.31334285e-02 1.84460950e+00 9.10462081e-01 -4.28218275e-01 9.29044560e-03 1.46746226e-02 5.12289405e-01 3.92051220e-01 -7.47254074e-01 1.25609875e-01 -5.71394920e-01 -1.13410167e-02 6.17176592e-01 4.44695681e-01 -1.03466070e+00 1.18713486e+00 5.60756493e+00 7.56164134e-01 -1.29120624e+00 2.02323258e-01 -4.13796782e-01 -2.49310121e-01 -4.05833781e-01 -5.21691255e-02 -6.06751144e-01 4.08875376e-01 3.27210695e-01 3.16916883e-01 3.53318155e-01 7.84276783e-01 2.03402489e-01 1.86796233e-01 -1.02579176e+00 1.29358435e+00 5.06984532e-01 -9.81053412e-01 3.20872277e-01 7.78333619e-02 7.48799264e-01 2.27447927e-01 2.73522705e-01 5.62693954e-01 3.55932683e-01 -8.05808365e-01 8.69316161e-01 8.54108810e-01 1.09008157e+00 -2.37283528e-01 5.69038272e-01 -6.21068105e-02 -1.49384606e+00 -4.99251038e-02 4.26765345e-02 4.12349582e-01 4.35153753e-01 -2.78380606e-02 -1.49939775e-01 5.95707595e-01 7.73001254e-01 1.10334563e+00 -7.18224525e-01 1.06699026e+00 -5.94187379e-01 6.02329910e-01 -2.71187693e-01 6.36781752e-02 1.94933459e-01 -1.74978852e-01 8.16514313e-01 1.27097845e+00 2.53937878e-02 -7.92234018e-02 1.18682943e-01 7.24907577e-01 1.86064057e-02 1.64697543e-01 -3.92335802e-01 7.77650103e-02 2.82579631e-01 8.94679427e-01 -8.44723266e-03 -4.52354282e-01 -7.43027449e-01 1.07882679e+00 1.97005033e-01 7.78187037e-01 -8.59351516e-01 -2.84757495e-01 6.88406289e-01 -4.18821990e-01 4.62094247e-01 -2.80844748e-01 -2.23107085e-01 -1.79541516e+00 4.83903855e-01 -1.11226773e+00 6.72678590e-01 -9.34091628e-01 -1.64996397e+00 4.07835841e-01 -7.87903890e-02 -1.58564425e+00 -3.81505787e-02 -9.48703706e-01 -2.37967655e-01 6.82482421e-01 -2.08508134e+00 -1.96629477e+00 -4.21040088e-01 1.15248251e+00 5.33529878e-01 -2.17316598e-01 3.52788597e-01 5.41778743e-01 -5.39329648e-01 1.08117509e+00 -8.93968567e-02 5.39151013e-01 1.03946638e+00 -1.08874321e+00 -2.47720271e-01 1.14309025e+00 2.57402122e-01 8.16384435e-01 4.22319502e-01 -7.35413194e-01 -1.53384614e+00 -9.82643783e-01 6.67926192e-01 -5.30570328e-01 8.45983982e-01 -1.35383993e-01 -8.80408585e-01 8.94903541e-01 1.94443733e-01 1.96198434e-01 6.48785770e-01 -3.82727794e-02 -1.01802731e+00 -5.94880469e-02 -7.11296201e-01 8.83664906e-01 1.42932773e+00 -1.08700788e+00 -9.11086082e-01 -1.24891214e-01 4.97038066e-01 -4.08349872e-01 -7.28392482e-01 6.40418887e-01 1.13055170e+00 -4.76528227e-01 8.69334698e-01 -7.80495465e-01 4.26300883e-01 -5.92480779e-01 -4.90154833e-01 -9.37246978e-01 1.74448609e-01 -3.93222123e-01 -3.19537997e-01 1.34692359e+00 2.66170174e-01 -4.55588043e-01 4.76083457e-01 7.30493367e-01 -1.35949120e-01 -4.49772924e-01 -1.03724194e+00 -9.25499499e-01 -1.11691505e-01 -8.95027041e-01 2.46360362e-01 9.98216212e-01 -2.53973126e-01 1.54787943e-01 -8.70197892e-01 1.15106478e-01 9.81420040e-01 2.41738260e-01 9.33430493e-01 -9.01356936e-01 -3.44959915e-01 -5.87086380e-01 -3.66231173e-01 -1.34821808e+00 3.56316358e-01 -1.10925722e+00 1.85911775e-01 -1.60528266e+00 3.44795197e-01 -8.71347114e-02 -5.31781256e-01 8.35058987e-01 -3.72370899e-01 1.95904225e-01 5.26621044e-01 3.22521418e-01 -6.50164664e-01 8.68257284e-01 1.44666314e+00 -3.49855185e-01 -1.93552747e-01 -4.26860809e-01 -3.90793890e-01 7.52614915e-01 2.95373231e-01 -1.02943204e-01 -1.26860291e-01 -6.60191596e-01 -8.50636363e-02 -3.86692882e-01 6.23440564e-01 -5.30540049e-01 5.18808067e-01 -3.16261351e-01 6.33612424e-02 -7.66215444e-01 3.54625016e-01 -8.92483413e-01 -3.44230384e-01 -4.77251746e-02 -4.45016950e-01 -2.92714417e-01 -1.84192266e-02 8.30910504e-01 -4.33006346e-01 2.36277699e-01 6.89162672e-01 1.07543267e-01 -1.19469106e+00 4.51178461e-01 -2.47811064e-01 1.31594181e-01 6.71315014e-01 -2.65837193e-01 -4.00737762e-01 -1.92440405e-01 -6.13551915e-01 4.43855077e-01 1.56720236e-01 8.27710152e-01 8.81244421e-01 -1.66586661e+00 -6.95819378e-01 1.48771837e-01 5.89630842e-01 -2.76750654e-01 6.23381317e-01 1.13937318e+00 -1.43602565e-01 3.89845341e-01 -2.48529434e-01 -7.82402933e-01 -1.50920498e+00 2.84121305e-01 2.64675319e-01 -7.24580660e-02 -9.22257006e-01 7.05766022e-01 8.22287425e-02 -6.40383422e-01 4.99385625e-01 -5.20953536e-01 -1.57889023e-01 7.42839649e-02 4.51458931e-01 1.57914832e-01 -3.87362450e-01 -1.11865866e+00 -4.89839226e-01 9.69060481e-01 1.67704314e-01 -3.90771896e-01 1.32223308e+00 -1.04038931e-01 1.59316629e-01 5.37288427e-01 1.18770051e+00 -1.88849075e-03 -1.36861098e+00 -8.83437991e-01 -6.50490224e-02 -5.80021083e-01 4.98356670e-02 -9.14176822e-01 -1.19618130e+00 1.02653205e+00 5.55542290e-01 -8.04972410e-01 1.26788116e+00 4.78355139e-02 8.10440898e-01 3.82672638e-01 1.58199251e-01 -1.32057571e+00 2.59201765e-01 4.42982614e-01 1.03718781e+00 -1.43952966e+00 -1.67123452e-01 -1.06105931e-01 -1.12619174e+00 8.74715149e-01 7.09099233e-01 -5.27314469e-03 5.19787967e-01 -4.15972341e-03 4.38492984e-01 -4.47948948e-02 -5.17565429e-01 -7.36443520e-01 8.97857547e-01 5.78143835e-01 1.55861244e-01 3.83084826e-02 -3.47664058e-01 8.09002101e-01 1.59944162e-01 2.10183829e-01 3.32202129e-02 9.40596402e-01 1.10988535e-01 -1.14173627e+00 -2.75986254e-01 2.56776631e-01 -4.43337671e-02 -1.45358160e-01 -4.46046531e-01 1.01699531e+00 1.67382076e-01 5.91190279e-01 -2.45906144e-01 -4.45737928e-01 5.08762121e-01 2.55884796e-01 6.81265414e-01 -6.70755357e-02 -2.68074930e-01 2.80069172e-01 -1.04365908e-01 -6.32750809e-01 -9.04537022e-01 -8.96984756e-01 -1.00575590e+00 1.46059781e-01 -2.98083097e-01 -5.75935900e-01 4.84108716e-01 1.24990416e+00 2.11410657e-01 3.87544870e-01 6.16386868e-02 -6.90582097e-01 -3.73029053e-01 -1.01709962e+00 -4.10575032e-01 7.48196959e-01 4.92259204e-01 -9.92156982e-01 -3.24377328e-01 2.36185834e-01]

[9.217232704162598, -6.517561435699463]

29107796-03ee-45db-94b0-bc40d2af8180

in-situ-anomaly-detection-in-additive

2305.02695

https://arxiv.org/abs/2305.02695v1

https://arxiv.org/pdf/2305.02695v1.pdf

In-situ Anomaly Detection in Additive Manufacturing with Graph Neural Networks

Transforming a design into a high-quality product is a challenge in metal additive manufacturing due to rare events which can cause defects to form. Detecting these events in-situ could, however, reduce inspection costs, enable corrective action, and is the first step towards a future of tailored material properties. In this study a model is trained on laser input information to predict nominal laser melting conditions. An anomaly score is then calculated by taking the difference between the predictions and new observations. The model is evaluated on a dataset with known defects achieving an F1 score of 0.821. This study shows that anomaly detection methods are an important tool in developing robust defect detection methods.

['Paul A. Hooper', 'Sebastian Larsen']

2023-05-04

['defect-detection']

['computer-vision']

[ 4.46452677e-01 2.63054222e-01 2.16416836e-01 -3.51989567e-01 -6.17713153e-01 -1.18273489e-01 2.27925643e-01 6.28992319e-01 4.25301105e-01 3.84015828e-01 -3.77648205e-01 -5.41187413e-02 -2.54512161e-01 -9.75391686e-01 -7.07477391e-01 -6.00703537e-01 1.65061578e-01 5.35352468e-01 4.42466974e-01 -3.23377877e-01 6.96397901e-01 5.88100255e-01 -1.86486673e+00 6.89280927e-01 8.14264894e-01 1.17910862e+00 3.47335905e-01 6.08052611e-01 1.13274291e-01 3.50459307e-01 -6.21620536e-01 -4.68903668e-02 2.03941658e-01 -3.45372826e-01 -5.48256576e-01 2.90264487e-01 2.39090115e-01 -8.75183940e-02 4.78352040e-01 7.01293945e-01 1.49089932e-01 1.14305735e-01 8.23356628e-01 -7.79716194e-01 -4.50816423e-01 2.58415163e-01 -2.74467528e-01 -2.32745275e-01 5.99339008e-01 3.94303858e-01 9.02366996e-01 -8.92620802e-01 4.86480534e-01 7.93796539e-01 5.89150906e-01 3.91965002e-01 -1.32820714e+00 -6.86877146e-02 -2.83329189e-01 -1.52264694e-02 -8.64693582e-01 -2.48699591e-01 8.36160600e-01 -8.45601678e-01 1.15668440e+00 2.29874909e-01 3.99088889e-01 5.17101824e-01 9.02466059e-01 2.76443183e-01 7.93962836e-01 -6.84117675e-01 4.94677693e-01 -7.83522129e-02 -2.16775954e-01 7.27178276e-01 2.67485857e-01 6.24910295e-01 -3.92438829e-01 1.38653712e-02 5.80795169e-01 -9.90686119e-02 9.49283019e-02 3.58866225e-03 -6.69305146e-01 6.64844215e-01 1.31633162e-01 2.02286109e-01 -7.73366392e-01 -7.00221136e-02 2.95353502e-01 7.57401526e-01 5.72322726e-01 1.19021511e+00 -6.94461226e-01 -1.90008074e-01 -8.72024655e-01 1.64130598e-01 4.14673179e-01 3.51371557e-01 7.67028928e-01 3.47328842e-01 8.63385051e-02 8.84772778e-01 2.39618048e-01 2.60661006e-01 -1.55379716e-02 -9.02239263e-01 -9.79100987e-02 9.13210690e-01 1.41870797e-01 -7.87012935e-01 -2.26969600e-01 -1.09971330e-01 -2.16733158e-01 7.75902689e-01 -9.79306325e-02 -3.60074230e-02 -1.33241653e+00 7.64991343e-01 2.49146938e-01 -3.04480284e-01 9.64865554e-03 3.69477957e-01 3.16621691e-01 5.16123056e-01 -2.23470449e-01 -2.82267839e-01 6.61624491e-01 -4.75234061e-01 -5.71023881e-01 -2.68775612e-01 7.79440045e-01 -1.26919460e+00 8.57693911e-01 9.03483033e-01 -8.63117337e-01 -5.68416774e-01 -1.19581223e+00 6.11178756e-01 -1.91803738e-01 2.17484966e-01 6.87281609e-01 7.26718605e-01 -5.19511878e-01 1.17182374e+00 -8.69764864e-01 -1.32162124e-01 1.74100652e-01 5.27316213e-01 -2.94581234e-01 -8.65452215e-02 -5.76894283e-01 1.05485463e+00 2.80009091e-01 2.35250250e-01 -8.68424356e-01 -9.04672086e-01 -8.27925384e-01 -4.25763309e-01 4.78735328e-01 -1.33067548e-01 1.29912078e+00 -4.29440469e-01 -1.56180739e+00 5.09095609e-01 -6.39916435e-02 -9.75319296e-02 2.27571338e-01 -3.95454168e-01 -6.28512502e-01 -2.62341619e-01 5.68942130e-02 1.06624268e-01 9.58186388e-01 -1.41279829e+00 -7.06287086e-01 -5.56385703e-02 -3.63442659e-01 -7.16279984e-01 2.29586273e-01 5.64986654e-02 2.94680089e-01 -2.83767074e-01 4.90021199e-01 -7.61393785e-01 -4.28564876e-01 -3.62687111e-01 -3.82270157e-01 -1.74638942e-01 6.62212431e-01 -8.12920094e-01 9.80600297e-01 -1.94636786e+00 -2.98046410e-01 5.70407152e-01 -4.31491435e-01 6.07676022e-02 1.42467812e-01 6.78854287e-01 -1.97773084e-01 -1.80210307e-01 -3.66696626e-01 1.28272966e-01 -2.98921496e-01 9.24752653e-02 -2.35527083e-02 1.78948238e-01 8.23141813e-01 4.68932122e-01 -7.73953438e-01 2.42826611e-01 5.32562673e-01 -1.09738976e-01 -6.21703207e-01 2.36760885e-01 -6.98278844e-01 2.95683503e-01 -1.82114542e-01 1.10570419e+00 5.12350082e-01 3.66025031e-01 -1.47031546e-01 -1.52176693e-01 -2.57739276e-01 6.23561367e-02 -9.60183918e-01 1.21591866e+00 -6.95239842e-01 3.70942950e-01 -1.61269352e-01 -9.66739953e-01 1.57529223e+00 3.70129168e-01 7.74561405e-01 -7.10770309e-01 3.15322489e-01 4.67967480e-01 1.89526573e-01 -7.53216326e-01 6.09877527e-01 -3.47173244e-01 -4.49144170e-02 2.01631457e-01 -1.20494530e-01 -8.14966917e-01 2.80703753e-01 -2.84753472e-01 1.28134429e+00 -5.57286516e-02 -1.97022855e-01 -1.98521018e-01 2.96384603e-01 1.92396790e-01 6.85629308e-01 4.32311147e-01 4.03258592e-01 8.12007248e-01 4.25982088e-01 -5.06175458e-01 -1.18352878e+00 -1.04201722e+00 6.29330650e-02 3.16098511e-01 -2.39799961e-01 -4.14327949e-01 -3.33401531e-01 -5.97673416e-01 2.36151427e-01 1.01471519e+00 -4.52513784e-01 -9.71374393e-01 -3.02254349e-01 -4.75772053e-01 -3.02785665e-01 6.51523530e-01 -8.11430905e-03 -1.05000937e+00 -3.79352808e-01 5.03189445e-01 2.85331041e-01 -6.94788516e-01 2.82129832e-02 3.46466929e-01 -9.48585987e-01 -1.25792205e+00 2.66593806e-02 -6.65957570e-01 8.15161407e-01 -2.84049809e-01 1.11795020e+00 4.69892830e-01 -7.04198062e-01 6.71649799e-02 -4.73222256e-01 -6.63361490e-01 -1.13614547e+00 -4.16275471e-01 2.36675069e-01 -2.58475184e-01 3.41538727e-01 -2.55938411e-01 -2.63949662e-01 4.93880153e-01 -6.94021463e-01 -4.24376518e-01 6.37492836e-01 7.74940908e-01 7.07428873e-01 6.97899759e-01 6.01656914e-01 -8.53524983e-01 4.99876559e-01 -2.66958594e-01 -7.21295357e-01 1.92279652e-01 -9.77421641e-01 1.42177954e-01 3.85701030e-01 -1.94782346e-01 -1.22098541e+00 3.83539051e-01 -4.23633009e-01 5.99678326e-03 -3.66889983e-01 6.92442715e-01 -6.38053417e-02 1.07712649e-01 6.33649826e-01 -4.49928790e-01 1.42939910e-01 -6.65016294e-01 -1.70979246e-01 5.44866979e-01 5.01418293e-01 -6.29567742e-01 7.79673994e-01 1.86356865e-02 2.90965915e-01 -8.79760981e-01 -8.47055852e-01 -4.30136949e-01 -7.16431439e-01 -5.83877265e-01 4.14248377e-01 -4.13387328e-01 -2.98320621e-01 4.67661679e-01 -8.42454612e-01 -3.78973097e-01 -5.13268173e-01 3.56785983e-01 -5.55159152e-01 7.25112632e-02 -3.26080769e-01 -9.60320294e-01 -8.74675661e-02 -1.09940445e+00 1.00283635e+00 4.25349250e-02 -6.41246617e-01 -6.95645690e-01 1.51370287e-01 3.38492483e-01 6.54915497e-02 5.01858830e-01 1.15707076e+00 -2.75791347e-01 -3.96837145e-01 -7.11125612e-01 4.84401315e-01 9.73371387e-01 7.00897813e-01 7.34438479e-01 -8.73896599e-01 2.45054718e-03 1.03299148e-01 2.01836322e-02 8.41638565e-01 5.00412166e-01 7.23309755e-01 2.52034217e-01 -3.14600796e-01 -2.68899411e-01 1.35525441e+00 5.53813517e-01 8.04005980e-01 2.28169516e-01 3.03068042e-01 6.55208409e-01 1.46502221e+00 3.55026126e-01 -6.16512239e-01 5.39264679e-01 6.61948144e-01 1.41795889e-01 -1.03606381e-01 -7.15007856e-02 4.30791616e-01 4.01351124e-01 -2.66941130e-01 1.01710662e-01 -1.02319539e+00 6.13070130e-01 -1.51216209e+00 -8.68608057e-01 -5.39712906e-01 2.30207348e+00 4.13157493e-01 8.35196018e-01 -1.35244787e-01 6.25924945e-01 5.95355511e-01 -5.67063689e-01 -2.82339424e-01 -9.78244364e-01 2.56044328e-01 6.12724185e-01 3.84022772e-01 3.78429741e-01 -7.56674886e-01 6.61003292e-01 7.24432659e+00 3.41561556e-01 -1.09242570e+00 -4.16418403e-01 1.96149975e-01 1.85013469e-02 -1.63037091e-01 8.46937224e-02 -3.91761094e-01 3.88663620e-01 1.10042405e+00 2.58381218e-01 -5.07888794e-02 9.11638260e-01 3.91517758e-01 -5.17018557e-01 -1.13874888e+00 2.30446830e-01 -1.68846279e-01 -1.40082991e+00 -1.82827383e-01 -3.61302719e-02 8.46644342e-01 -4.05301064e-01 1.84031017e-02 -6.27982169e-02 3.39567542e-01 -7.30665445e-01 6.06518507e-01 7.11978972e-01 5.22934973e-01 -1.02969813e+00 1.08693242e+00 2.92569287e-02 -8.91411602e-01 -2.70270914e-01 -2.80915707e-01 -3.21928322e-01 4.72504735e-01 1.21754706e+00 -1.46353567e+00 6.99157178e-01 6.83979571e-01 4.84087020e-01 -4.77678657e-01 1.14563870e+00 -1.84470654e-01 5.93594432e-01 -4.45537031e-01 3.04685622e-01 -2.38631010e-01 -1.64224565e-01 3.58568132e-01 5.31567633e-01 5.80831885e-01 -4.69224900e-01 1.48088485e-01 8.98831069e-01 4.47021306e-01 -3.29008877e-01 -6.56661749e-01 -2.35080235e-02 3.74103874e-01 7.60705471e-01 -9.03392792e-01 3.62597436e-01 -1.80688858e-01 7.19812512e-01 -3.45165431e-01 -1.96112946e-01 -4.36389536e-01 -2.90448189e-01 7.32939541e-01 5.07785797e-01 4.63379562e-01 -2.24667132e-01 -4.39920694e-01 -1.27892733e-01 2.36499980e-01 -5.67514002e-01 -3.90427299e-02 -6.46489739e-01 -1.28492558e+00 -2.03684252e-02 -2.60754317e-01 -1.10416901e+00 -2.93799311e-01 -9.77479160e-01 -1.07353282e+00 5.15923083e-01 -9.24132288e-01 -9.34941530e-01 -1.04539461e-01 -3.63292307e-01 7.95536697e-01 -2.82903641e-01 7.08056211e-01 7.91874304e-02 -6.10282242e-01 3.64480019e-01 1.96507588e-01 -3.64341646e-01 6.03023708e-01 -1.21690190e+00 4.24212366e-01 8.20686698e-01 -9.95114967e-02 1.92016140e-01 1.20918906e+00 -1.27461767e+00 -1.06053793e+00 -1.11794710e+00 5.43640792e-01 -4.90317553e-01 4.33968306e-01 5.43596828e-03 -1.07316065e+00 3.09745491e-01 -2.81588703e-01 -3.76117200e-01 5.48880160e-01 2.24629670e-01 1.40976697e-01 -6.93643764e-02 -1.34453213e+00 1.32476717e-01 6.34887695e-01 -2.74670720e-01 -4.58809882e-01 4.00907695e-01 5.43453634e-01 -3.75713080e-01 -1.27203321e+00 7.45116115e-01 3.37577701e-01 -1.04842782e+00 7.29322970e-01 -4.24140960e-01 8.16727281e-01 -2.00121090e-01 2.28572395e-02 -1.40210617e+00 -1.94216311e-01 -5.82702279e-01 1.71286874e-02 1.31998158e+00 8.94176185e-01 -3.37815344e-01 9.37043130e-01 8.12364459e-01 -6.48889840e-01 -8.51642132e-01 -6.01979494e-01 -1.06482399e+00 -2.46778205e-01 -6.06607676e-01 5.69635212e-01 4.03964043e-01 -1.39913291e-01 -9.84183252e-02 -8.93159956e-02 3.85223508e-01 2.99883872e-01 -8.33011940e-02 4.40999210e-01 -1.79095483e+00 -2.40395516e-02 -6.25485405e-02 -6.23905241e-01 -5.79388067e-02 -5.00925966e-02 -4.13297713e-01 5.58825791e-01 -1.39359152e+00 -3.55857134e-01 -6.11720145e-01 -1.18365943e-01 3.68971139e-01 5.14729973e-03 8.52935538e-02 -4.92897838e-01 -2.45400906e-01 2.39206061e-01 3.51668388e-01 8.82438362e-01 -1.92692652e-01 -3.63252103e-01 5.06435990e-01 -3.19090366e-01 6.33430004e-01 1.08103120e+00 -5.21280527e-01 -2.00182498e-01 -2.82618403e-02 4.70256478e-01 -2.52547204e-01 2.15953052e-01 -1.14307451e+00 -5.25369272e-02 -4.00552452e-01 2.08979130e-01 -7.53535032e-01 2.20485836e-01 -9.82808590e-01 2.74456084e-01 5.26145041e-01 -1.92075800e-02 5.30396327e-02 2.17419177e-01 5.75593889e-01 -1.92660376e-01 -7.82060981e-01 6.71133101e-01 -9.69815627e-02 -7.38679469e-01 -1.14811838e-01 -3.42057824e-01 -6.49019241e-01 1.15250790e+00 -5.14875174e-01 1.58061728e-01 2.03263909e-01 -8.10511887e-01 -1.28455907e-01 7.38454700e-01 5.55404603e-01 8.30350280e-01 -1.11619318e+00 -6.10175192e-01 4.14101750e-01 1.68440834e-01 1.67666256e-01 2.20641643e-01 4.71571594e-01 -6.70348525e-01 8.53370130e-02 -1.77652970e-01 -7.47464657e-01 -1.37294853e+00 3.99648815e-01 1.58740833e-01 6.33542193e-04 -4.46328163e-01 8.89870226e-01 -5.70802510e-01 -1.65298894e-01 -4.04658556e-01 -4.58809316e-01 1.59511372e-01 -1.21604525e-01 2.56310880e-01 4.40394849e-01 9.29197490e-01 -2.33657524e-01 -1.56654045e-01 5.67498982e-01 -3.26890945e-01 3.36554617e-01 1.69653392e+00 2.95775533e-01 -1.26774535e-01 5.63905835e-01 6.73655629e-01 -3.65611836e-02 -1.16839647e+00 2.59238243e-01 5.25565743e-01 -6.61310017e-01 2.79246598e-01 -8.78357947e-01 -1.19396269e+00 3.49423826e-01 7.08494484e-01 2.64632285e-01 9.78162766e-01 3.24976444e-01 7.44676530e-01 3.02637756e-01 1.97879896e-01 -1.47652400e+00 3.61326545e-01 2.28445604e-01 1.05515182e+00 -1.36143744e+00 1.69710830e-01 -8.04291666e-01 -4.70280677e-01 1.20527351e+00 7.51446068e-01 -9.42711681e-02 5.93131602e-01 4.58009660e-01 3.26083712e-02 -6.67088628e-01 -6.85090661e-01 3.81231564e-03 2.66999662e-01 8.02203476e-01 4.28170115e-01 -1.26156006e-02 9.27026272e-02 1.68987393e-01 -1.41294569e-01 -1.15391530e-01 6.12657666e-01 1.43054962e+00 -7.61194766e-01 -1.46338701e+00 -3.50674301e-01 9.66225564e-01 -1.60294041e-01 2.97312111e-01 -4.38195765e-01 4.96113896e-01 2.00010389e-01 1.24115765e+00 1.09477915e-01 -6.07960284e-01 9.06370640e-01 1.09636717e-01 5.20807862e-01 -1.03044009e+00 -5.15268385e-01 -3.43443751e-01 3.39136004e-01 -5.39589703e-01 -1.11436687e-01 -8.15542519e-01 -1.17879152e+00 -7.66177922e-02 -1.00093198e+00 -5.98802269e-02 7.56136358e-01 9.61305559e-01 2.91680187e-01 9.08794820e-01 9.49056685e-01 -5.99044085e-01 -3.91060263e-01 -8.36771965e-01 -6.11333370e-01 3.12461644e-01 1.42054319e-01 -1.03470504e+00 -3.79628658e-01 1.73262298e-01]

[7.058096408843994, 2.237840175628662]

83f3b8c5-5c0c-472e-82ba-26835c31c866

fern-leveraging-graph-attention-networks-for

2305.19153

https://arxiv.org/abs/2305.19153v1

https://arxiv.org/pdf/2305.19153v1.pdf

FERN: Leveraging Graph Attention Networks for Failure Evaluation and Robust Network Design

Robust network design, which aims to guarantee network availability under various failure scenarios while optimizing performance/cost objectives, has received significant attention. Existing approaches often rely on model-based mixed-integer optimization that is hard to scale or employ deep learning to solve specific engineering problems yet with limited generalizability. In this paper, we show that failure evaluation provides a common kernel to improve the tractability and scalability of existing solutions. By providing a neural network function approximation of this common kernel using graph attention networks, we develop a unified learning-based framework, FERN, for scalable Failure Evaluation and Robust Network design. FERN represents rich problem inputs as a graph and captures both local and global views by attentively performing feature extraction from the graph. It enables a broad range of robust network design problems, including robust network validation, network upgrade optimization, and fault-tolerant traffic engineering that are discussed in this paper, to be recasted with respect to the common kernel and thus computed efficiently using neural networks and over a small set of critical failure scenarios. Extensive experiments on real-world network topologies show that FERN can efficiently and accurately identify key failure scenarios for both OSPF and optimal routing scheme, and generalizes well to different topologies and input traffic patterns. It can speed up multiple robust network design problems by more than 80x, 200x, 10x, respectively with negligible performance gap.

['Qing Li', 'Mingwei Xu', 'Yuan Yang', 'Nan Geng', 'Tian Lan', 'Vaneet Aggarwal', 'Chenyi Liu']

2023-05-30

['graph-attention']

['graphs']

[-1.60278276e-01 -1.87175155e-01 -6.06454194e-01 -3.69922549e-01 -4.22561139e-01 -4.80056703e-01 -2.08654970e-01 3.14931273e-02 2.83167899e-01 8.08292270e-01 -3.45689535e-01 -7.54330158e-01 -9.13201511e-01 -7.02692091e-01 -6.17973804e-01 -4.34087276e-01 -8.77379239e-01 6.57143414e-01 9.48914215e-02 -2.89985955e-01 2.87209272e-01 1.04624796e+00 -1.02571249e+00 -2.93592453e-01 5.17372191e-01 1.46148360e+00 -3.98530513e-01 6.77532554e-01 1.98836535e-01 9.52332914e-01 -8.20752025e-01 -2.60413848e-02 5.00701070e-01 2.93846399e-01 -1.03452587e+00 2.82503128e-01 3.70725870e-01 -3.19086850e-01 -9.47347403e-01 6.29953623e-01 6.30688906e-01 -2.03672666e-02 5.61826378e-02 -2.09985518e+00 -6.10253274e-01 9.16012466e-01 -4.36792493e-01 2.28144556e-01 -1.65487289e-01 4.81507033e-01 1.17869830e+00 -1.97296619e-01 3.12675983e-01 1.36708009e+00 6.55059040e-01 3.77151787e-01 -1.39042139e+00 -7.87075937e-01 4.67441052e-01 4.03672695e-01 -1.38227761e+00 -6.88024461e-01 9.49783862e-01 1.44255742e-01 9.28915262e-01 2.27278888e-01 9.08848867e-02 6.76023722e-01 4.02569652e-01 3.02631557e-01 6.29294693e-01 6.88501149e-02 2.01124415e-01 -4.13947731e-01 1.21925525e-01 8.70734453e-01 1.54748693e-01 3.33362699e-01 -1.67766120e-02 -8.42411816e-02 8.91122997e-01 -6.66851178e-02 -4.44102556e-01 -4.49762464e-01 -1.08538127e+00 6.06387198e-01 8.15787733e-01 -3.52660745e-01 -1.95662111e-01 8.23109090e-01 7.22554207e-01 9.22958136e-01 4.10357378e-02 4.80924159e-01 -8.17830145e-01 3.00372899e-01 -6.79024100e-01 -5.32968156e-02 1.14927649e+00 1.09698522e+00 8.97897184e-01 7.51364648e-01 5.06083891e-02 7.58157790e-01 2.38433585e-01 4.72419858e-01 -3.74107301e-01 -1.41517341e+00 3.84869576e-01 5.23507059e-01 -2.97474980e-01 -1.35545778e+00 -9.31167483e-01 -6.17354989e-01 -1.01078498e+00 3.42493951e-01 7.43727610e-02 -2.44936332e-01 -6.99380279e-01 1.94337237e+00 1.91569492e-01 5.59264898e-01 -7.91312382e-02 7.12126911e-01 2.83862621e-01 7.17391551e-01 -2.71202594e-01 -3.04687440e-01 8.63654137e-01 -1.11383688e+00 -3.89207512e-01 -9.65924561e-02 5.02263486e-01 -3.43203276e-01 7.25274622e-01 2.57830977e-01 -9.46597159e-01 -2.86772728e-01 -1.19114554e+00 4.06446904e-01 -1.91742346e-01 -3.59670401e-01 6.70018554e-01 7.65493035e-01 -1.50345719e+00 6.74832463e-01 -5.53485692e-01 -4.48186040e-01 4.94286269e-01 8.77654374e-01 -3.50647926e-01 -3.01524103e-01 -1.07958317e+00 5.70333719e-01 4.47293341e-01 3.47926021e-01 -1.34694207e+00 -9.78446424e-01 -9.22398746e-01 3.20814788e-01 1.11750317e+00 -6.48511231e-01 8.09587002e-01 -5.36594152e-01 -1.58342898e+00 1.01286888e-01 4.43264395e-01 -4.87624615e-01 1.86000317e-02 1.61562681e-01 -1.01334560e+00 1.98947296e-01 -1.08254164e-01 2.87922651e-01 1.02236640e+00 -1.27606189e+00 -3.50072712e-01 -3.10787912e-02 4.90807354e-01 -5.20702779e-01 -2.54176825e-01 2.73151010e-01 -4.79675412e-01 -5.55072129e-01 -1.05393669e-02 -8.47532630e-01 -5.64715981e-01 3.17031026e-01 -7.05235958e-01 -1.69182733e-01 1.07319760e+00 -2.51931429e-01 1.30526590e+00 -1.66044104e+00 1.38017520e-01 8.69108677e-01 6.64661825e-01 7.82291219e-02 -6.00958765e-01 5.66704154e-01 -2.22981855e-01 5.14609754e-01 6.74967170e-02 6.21158108e-02 2.52452314e-01 4.46071535e-01 -2.73765385e-01 6.47785544e-01 4.71873611e-01 7.88374841e-01 -6.86380029e-01 -4.27876294e-01 2.86899328e-01 9.68129262e-02 -6.96896553e-01 2.29833767e-01 -1.73161283e-01 8.78979266e-02 -3.01398396e-01 1.00028634e+00 7.35432684e-01 -5.49736500e-01 4.92279500e-01 -5.48321486e-01 2.93113738e-01 -1.37916714e-01 -1.49547505e+00 1.24389315e+00 -7.84102023e-01 4.86093342e-01 5.41915476e-01 -1.39857054e+00 8.41869116e-01 1.47760451e-01 7.93029726e-01 -7.11240113e-01 1.41494751e-01 1.10378221e-01 -8.81336704e-02 -3.91033500e-01 -1.06181568e-02 1.23662665e-01 -2.58522034e-01 7.55305350e-01 9.77233574e-02 3.81480455e-01 2.53992796e-01 2.38975808e-01 1.56752110e+00 -5.24642348e-01 -1.79837823e-01 -3.21032643e-01 6.49993181e-01 -4.45205957e-01 9.35713291e-01 8.92533004e-01 -3.64940405e-01 8.57814923e-02 9.77001071e-01 -6.55224860e-01 -7.50902176e-01 -1.27953446e+00 7.20190331e-02 1.10548556e+00 2.10579842e-01 -2.50331759e-01 -1.81599081e-01 -7.98250198e-01 1.70339301e-01 1.84507146e-01 -2.73310661e-01 -5.85691333e-01 -8.86280775e-01 -6.31531060e-01 6.28818035e-01 4.74096745e-01 1.46128938e-01 -8.40994358e-01 1.57937288e-01 6.01564646e-01 2.65078694e-01 -1.45908070e+00 -5.51250756e-01 2.14640364e-01 -6.31904125e-01 -1.34479499e+00 1.42054021e-01 -9.53295112e-01 5.32029748e-01 2.04301789e-01 1.35794485e+00 6.76858068e-01 -5.59473038e-01 3.18513989e-01 1.63459051e-02 4.20286477e-01 -4.01481569e-01 1.81777716e-01 5.06756723e-01 1.28561512e-01 -6.91380560e-01 -1.03434277e+00 -3.11839759e-01 6.74707055e-01 -8.18867266e-01 -6.09089911e-01 6.99608982e-01 8.79568636e-01 2.90029198e-01 5.86550832e-01 1.17134881e+00 -5.86644351e-01 4.55137938e-01 -8.45141590e-01 -8.21807683e-01 4.07112569e-01 -9.19396877e-01 2.59026021e-01 1.03096497e+00 -2.30990097e-01 -3.03639054e-01 -6.50687695e-01 1.92002043e-01 -8.16537440e-01 1.61006570e-01 6.03582382e-01 -5.39075494e-01 -7.50114262e-01 4.15080249e-01 -2.43385196e-01 3.61146241e-01 -1.38778806e-01 3.17929894e-01 3.16548198e-01 3.75681102e-01 -1.17019808e+00 1.16914678e+00 2.79484421e-01 5.36399007e-01 -4.76620615e-01 -5.28086960e-01 5.48527278e-02 -3.98718528e-02 -3.67625177e-01 -1.03010923e-01 -6.22466266e-01 -1.34031475e+00 2.88712442e-01 -7.63029873e-01 -3.30365241e-01 1.53531041e-03 1.26204723e-02 -5.73474526e-01 4.30264533e-01 -8.61632407e-01 -4.68207479e-01 -3.81513208e-01 -1.32092130e+00 5.76251090e-01 -1.45790711e-01 3.82859290e-01 -1.21100509e+00 -4.13241327e-01 1.32953420e-01 7.73199975e-01 4.38971072e-01 1.30473745e+00 -4.22287673e-01 -9.97084439e-01 -2.45640084e-01 -8.07551086e-01 4.93807614e-01 1.27350941e-01 4.38478708e-01 -2.76599497e-01 -7.76271343e-01 -5.34399331e-01 -4.89829898e-01 5.85738897e-01 1.85906753e-01 1.49158800e+00 -6.25317574e-01 -1.69246316e-01 1.03807044e+00 1.66230989e+00 -1.87228709e-01 5.61936915e-01 6.42226338e-02 8.80189955e-01 5.89120746e-01 1.24884859e-01 5.10455787e-01 5.82970798e-01 4.86760020e-01 1.06913233e+00 -6.73037674e-03 -6.41627908e-02 2.54338652e-01 4.00899619e-01 6.27383053e-01 4.49643761e-01 -5.72875381e-01 -8.94062698e-01 3.41282070e-01 -1.78632855e+00 -7.57293046e-01 3.22762549e-01 1.91960704e+00 3.06734145e-01 5.17090499e-01 1.64268658e-01 4.42773104e-01 9.80138659e-01 2.11990803e-01 -8.13816786e-01 -4.09347743e-01 -4.04787511e-02 2.45390892e-01 7.25088596e-01 5.16620159e-01 -7.60651290e-01 8.37490320e-01 6.62406921e+00 8.03100348e-01 -1.34547734e+00 -3.58928293e-01 5.43129861e-01 -2.33584847e-02 9.09747183e-03 2.18910217e-01 -2.59513944e-01 2.75014371e-01 1.39448678e+00 -1.61644012e-01 8.64625931e-01 6.91602826e-01 4.35905695e-01 7.33015716e-01 -1.12477779e+00 8.04084837e-01 -2.94648409e-01 -1.64758074e+00 1.23944126e-01 5.43302633e-02 5.15883029e-01 -5.54155745e-02 5.11152111e-02 4.72203076e-01 6.08141303e-01 -1.12615299e+00 3.66500199e-01 2.89761215e-01 8.93598914e-01 -9.93165076e-01 6.28208280e-01 -3.31014425e-01 -1.19708920e+00 -7.79329181e-01 -1.51953354e-01 5.09099029e-02 2.56892651e-01 6.12348914e-01 -5.43212771e-01 7.42033005e-01 6.13562763e-01 7.14727521e-01 -4.58527833e-01 1.19794750e+00 2.37948254e-01 5.93457699e-01 -5.11088550e-01 3.76191854e-01 2.29911253e-01 3.80504400e-01 6.26287103e-01 6.25806928e-01 -1.47671541e-02 -4.56060469e-01 8.24567378e-01 7.14008808e-01 -4.99552906e-01 -9.93070453e-02 -3.69119868e-02 -5.47146350e-02 8.51721764e-01 1.33587205e+00 -6.71616375e-01 1.89806938e-01 -2.14476466e-01 3.75696003e-01 5.06707311e-01 7.40006030e-01 -1.11244190e+00 -7.66075373e-01 1.10697603e+00 8.47721323e-02 2.00529099e-01 -3.06219518e-01 -1.22443520e-01 -9.22640920e-01 3.66994478e-02 -9.43275988e-01 5.35204530e-01 -3.83951068e-01 -1.47138596e+00 6.47765577e-01 -2.86418587e-01 -9.55460727e-01 1.55087322e-01 -6.67171776e-01 -9.48690295e-01 4.05657530e-01 -1.81547785e+00 -1.27163672e+00 -2.05701545e-01 7.39335418e-01 2.22160906e-01 -2.95393169e-01 5.13045490e-01 6.84712768e-01 -1.46777141e+00 9.92374182e-01 -2.83630285e-02 5.31398281e-02 5.41692734e-01 -1.07372737e+00 2.33746335e-01 9.38141465e-01 -4.03452635e-01 3.72384816e-01 4.81516153e-01 -2.38284841e-01 -1.97725368e+00 -1.42022157e+00 8.55497867e-02 2.80652605e-02 1.12935865e+00 -5.33320569e-02 -6.56309724e-01 5.75625122e-01 -1.54289916e-01 7.75061250e-01 3.80249172e-01 1.95836872e-02 -6.02305830e-01 -8.19703043e-01 -1.40134835e+00 5.84938288e-01 1.12986457e+00 -4.51252222e-01 4.35533583e-01 4.08192188e-01 1.24276757e+00 -5.67918718e-02 -1.32155025e+00 5.57523847e-01 1.83042869e-01 -4.43730682e-01 1.09126329e+00 -1.16976106e+00 -2.61441946e-01 -5.03926098e-01 -3.17663431e-01 -1.09229326e+00 -5.24226904e-01 -1.18650973e+00 -7.83281803e-01 1.23690796e+00 3.19137901e-01 -7.41595685e-01 7.06904352e-01 2.72406399e-01 -4.23915863e-01 -9.83112097e-01 -1.00507832e+00 -1.04959774e+00 -1.68857023e-01 -6.88448608e-01 1.07121813e+00 8.49926889e-01 -3.44039798e-01 1.87304020e-01 -5.66360354e-01 8.89940500e-01 9.52694714e-01 2.30897367e-01 6.29625201e-01 -1.29126668e+00 -1.01155348e-01 -7.22469211e-01 -7.84188569e-01 -8.61621141e-01 7.65410781e-01 -9.74727511e-01 -2.88259774e-01 -1.28646910e+00 -6.28554761e-01 -1.00421047e+00 -8.78639400e-01 6.09647393e-01 2.90019929e-01 5.15043288e-02 -1.91038139e-02 5.94779290e-03 -9.79692817e-01 4.60817873e-01 9.04696286e-01 -2.78047889e-01 -3.04998960e-02 5.25315143e-02 -1.00980222e+00 1.81156412e-01 1.03908765e+00 -3.65558803e-01 -4.21703041e-01 -5.52949429e-01 9.52051803e-02 5.16540825e-01 4.73357469e-01 -1.22068071e+00 3.07663441e-01 -5.61753571e-01 -8.66364986e-02 -8.42783824e-02 2.55057067e-02 -1.11165726e+00 7.54716843e-02 4.28866714e-01 -2.22823277e-01 3.25961679e-01 1.37845457e-01 8.61518323e-01 2.03094080e-01 4.14078563e-01 8.78225684e-01 3.54282886e-01 -1.06035256e+00 1.10556006e+00 -7.77800456e-02 2.48474717e-01 1.04716969e+00 -3.32333744e-02 -5.52103519e-01 -4.74798977e-01 -7.96529233e-01 1.04972661e+00 3.10262412e-01 5.74220121e-01 8.32599878e-01 -1.41606629e+00 -8.01225007e-01 1.17165610e-01 -1.33251492e-02 -4.02750909e-01 5.59220195e-01 7.79694438e-01 -8.11733365e-01 1.57872707e-01 -4.16223258e-01 -4.80630100e-01 -5.43772101e-01 1.00117266e+00 6.63668752e-01 -3.41821551e-01 -3.56477469e-01 5.00744462e-01 -4.60875988e-01 -7.09884286e-01 5.64533293e-01 -9.45465863e-02 2.75396883e-01 -4.61659610e-01 3.28461498e-01 8.36201429e-01 7.64976442e-02 -4.06169683e-01 -5.44657052e-01 4.06820118e-01 8.94890446e-03 6.10700488e-01 1.37029397e+00 -4.32916790e-01 -5.59737921e-01 -2.46190965e-01 1.40151930e+00 -5.15050888e-01 -1.05028796e+00 -4.08793986e-01 7.96974674e-02 -3.06892961e-01 4.21714216e-01 -6.85501575e-01 -2.07181406e+00 4.77368981e-01 4.65158790e-01 5.01510620e-01 1.33432102e+00 -2.58786023e-01 9.76488590e-01 6.97437346e-01 5.36431730e-01 -8.21596384e-01 2.26128444e-01 4.58786935e-01 7.17969120e-01 -1.08114862e+00 -2.04806685e-01 -6.07067585e-01 3.05356272e-02 1.47323895e+00 1.05279768e+00 -1.74955472e-01 1.02508759e+00 3.58444899e-01 -2.24617094e-01 -2.28776529e-01 -1.16037154e+00 3.34714130e-02 -2.12319389e-01 7.17719555e-01 -3.62846315e-01 -9.80007648e-02 5.24210989e-01 2.20775753e-01 3.29076946e-01 -3.36987466e-01 7.81999409e-01 8.99840474e-01 -5.16600132e-01 -1.09298182e+00 -1.19491212e-01 7.05442011e-01 -3.40669185e-01 1.16616786e-01 2.51726598e-01 6.72186017e-01 -3.33034962e-01 1.18422365e+00 -4.92561571e-02 -7.02971637e-01 3.17586690e-01 -3.87789398e-01 8.46627355e-02 -2.71166980e-01 -3.77825558e-01 -5.30352354e-01 2.38240153e-01 -1.08579075e+00 1.88901126e-01 7.26027042e-02 -1.06963646e+00 -1.06142843e+00 -3.95268947e-01 1.45828456e-01 3.65729719e-01 6.76341355e-01 7.44776309e-01 8.64074707e-01 1.33586514e+00 -6.39994264e-01 -7.41165102e-01 -2.88170487e-01 -7.32653201e-01 -2.26159707e-01 6.89393818e-01 -7.15882063e-01 -4.73206967e-01 -5.47758222e-01]

[5.834531307220459, 1.7324328422546387]

2994df02-6503-4e28-b511-4f69fd3b1900

nose-eyes-and-ears-head-pose-estimation-by

1812.00739

http://arxiv.org/abs/1812.00739v1

http://arxiv.org/pdf/1812.00739v1.pdf

Nose, eyes and ears: Head pose estimation by locating facial keypoints

Monocular head pose estimation requires learning a model that computes the intrinsic Euler angles for pose (yaw, pitch, roll) from an input image of human face. Annotating ground truth head pose angles for images in the wild is difficult and requires ad-hoc fitting procedures (which provides only coarse and approximate annotations). This highlights the need for approaches which can train on data captured in controlled environment and generalize on the images in the wild (with varying appearance and illumination of the face). Most present day deep learning approaches which learn a regression function directly on the input images fail to do so. To this end, we propose to use a higher level representation to regress the head pose while using deep learning architectures. More specifically, we use the uncertainty maps in the form of 2D soft localization heatmap images over five facial keypoints, namely left ear, right ear, left eye, right eye and nose, and pass them through an convolutional neural network to regress the head-pose. We show head pose estimation results on two challenging benchmarks BIWI and AFLW and our approach surpasses the state of the art on both the datasets.

['P. J. Narayanan', 'Vineet Gandhi', 'Aryaman Gupta', 'Kalpit Thakkar']

2018-12-03

['head-pose-estimation']

['computer-vision']

[-2.39781052e-01 5.26549935e-01 2.93257535e-01 -9.05414343e-01 -7.33222008e-01 -4.93198484e-01 5.53101063e-01 -2.73461610e-01 -5.53369701e-01 5.21448672e-01 2.18118951e-01 3.00560206e-01 2.09529579e-01 -2.24737287e-01 -1.02926862e+00 -6.67418718e-01 3.12398828e-04 5.70651114e-01 -1.03421435e-01 -6.62619397e-02 1.79990884e-02 6.70967281e-01 -1.72824693e+00 -2.72557855e-01 1.29785225e-01 1.04767835e+00 -2.70986944e-01 6.28174901e-01 3.88273984e-01 3.03118289e-01 -2.97483176e-01 -4.93712723e-01 2.85695761e-01 4.44845408e-02 -6.10434413e-01 6.59383535e-02 1.09725702e+00 -5.04477203e-01 -1.47656217e-01 9.96482253e-01 7.53428519e-01 -5.30686677e-02 5.47233641e-01 -1.45133173e+00 -1.74890831e-01 8.83760154e-02 -7.11982906e-01 -2.59023845e-01 8.60134244e-01 9.81186926e-02 6.06423497e-01 -9.99489427e-01 7.32773781e-01 1.37934768e+00 9.05012846e-01 7.67863512e-01 -9.64904130e-01 -6.91253126e-01 1.01157896e-01 5.64594120e-02 -1.57094300e+00 -7.64017522e-01 6.31483078e-01 -4.39054996e-01 7.58327007e-01 1.51775658e-01 4.74283159e-01 1.06319118e+00 4.91986163e-02 5.41741312e-01 1.17698908e+00 -3.87785465e-01 1.17213696e-01 5.46879992e-02 -1.13202915e-01 8.28332424e-01 1.94326982e-01 6.90814331e-02 -6.60409153e-01 1.71838924e-02 4.24166232e-01 -2.08629191e-01 -4.30147588e-01 -6.01571321e-01 -7.41543174e-01 5.82492709e-01 7.03318715e-01 -1.66679814e-01 -3.30443203e-01 6.26221523e-02 1.58148974e-01 -2.56743859e-02 4.75060225e-01 6.68264255e-02 -7.14792192e-01 1.21621922e-01 -1.06261277e+00 4.76612687e-01 8.35653245e-01 1.02969813e+00 9.60480809e-01 -2.43823186e-01 -1.50761399e-02 4.77968365e-01 7.85162508e-01 5.27714133e-01 4.05510485e-01 -6.94510579e-01 2.14188591e-01 3.86886954e-01 8.62641111e-02 -8.00984383e-01 -9.15849924e-01 -1.45606890e-01 -3.99811029e-01 3.87229443e-01 5.86511612e-01 -4.39839512e-01 -1.31405210e+00 1.89364970e+00 7.77593732e-01 1.07357986e-01 -2.62540907e-01 1.09457898e+00 9.03539121e-01 1.59665719e-01 -7.83512294e-02 3.55894491e-02 1.46969998e+00 -7.28396058e-01 -6.28809035e-01 -6.07925296e-01 3.61106604e-01 -7.87948012e-01 7.93824196e-01 3.22409183e-01 -1.02291441e+00 -2.94270515e-01 -9.98138428e-01 -3.01563770e-01 -5.48171282e-01 4.25316364e-01 2.66630262e-01 6.79175973e-01 -1.28275311e+00 2.14642167e-01 -1.10048234e+00 -5.30537188e-01 1.66716740e-01 8.04949701e-01 -8.75363529e-01 7.24104121e-02 -7.46770978e-01 1.13354826e+00 -6.07977211e-02 5.82505345e-01 -7.87154496e-01 -6.27086997e-01 -1.39910173e+00 -2.52635986e-01 9.34482962e-02 -5.27905285e-01 1.34054291e+00 -6.54290497e-01 -1.99217737e+00 1.28082991e+00 -1.02483757e-01 -2.37429529e-01 6.95905805e-01 -6.16080225e-01 5.84732816e-02 -2.52319753e-01 -5.70158996e-02 1.05456460e+00 1.07261205e+00 -1.20789719e+00 -3.17073196e-01 -1.02346134e+00 -2.14190245e-01 2.05747578e-02 1.50691822e-01 1.15232319e-01 -5.98431110e-01 -8.63176063e-02 3.94119471e-01 -1.25063217e+00 1.74864396e-01 1.97638795e-01 -5.41739285e-01 -2.50183403e-01 7.71408498e-01 -9.55796599e-01 6.77583992e-01 -1.95878351e+00 2.22136453e-01 1.47074968e-01 -7.26399850e-03 2.77687469e-03 2.00488582e-01 -2.50940174e-01 -2.76655346e-01 -5.32527030e-01 5.32968864e-02 -8.74173224e-01 2.55337864e-01 7.72836581e-02 4.73245122e-02 1.12236917e+00 1.19867921e-01 7.75585115e-01 -5.62406063e-01 -3.65258992e-01 1.53186038e-01 8.99037302e-01 -8.07165384e-01 4.41676915e-01 -1.03667274e-01 6.62037671e-01 -3.12106479e-02 6.56259000e-01 1.01446068e+00 2.76317269e-01 -6.10310622e-02 -4.75865513e-01 -4.80736680e-02 1.96555257e-01 -1.27065933e+00 1.84001279e+00 -4.37316030e-01 6.66401625e-01 4.48490441e-01 -5.77377677e-01 8.40435505e-01 4.13815618e-01 2.39274040e-01 -3.57518137e-01 4.97353286e-01 9.22948420e-02 -2.38635883e-01 -5.19633114e-01 1.22902140e-01 -1.41979411e-01 3.21704894e-02 2.03150228e-01 5.12595713e-01 -3.55879664e-01 -4.38961208e-01 -4.17237371e-01 4.70576763e-01 5.23763835e-01 2.21268401e-01 -1.81803659e-01 7.13201106e-01 -7.20267832e-01 3.08038980e-01 4.02187109e-02 -1.81914300e-01 9.04578984e-01 3.63216490e-01 -5.69571435e-01 -8.91663790e-01 -8.76307964e-01 -2.35521659e-01 1.35176301e+00 -2.63813138e-01 -2.11972162e-01 -1.31766331e+00 -4.88702983e-01 7.16619492e-02 2.04944119e-01 -9.09017563e-01 4.50214259e-02 -6.76154613e-01 -4.85915065e-01 5.15037477e-01 5.00245929e-01 2.17531770e-01 -7.78317928e-01 -6.42577052e-01 -2.89177746e-01 1.39829427e-01 -1.39491785e+00 -4.68045771e-01 1.25001892e-01 -3.49024266e-01 -8.70660603e-01 -7.57883668e-01 -7.66997337e-01 8.50144863e-01 -5.29655278e-01 8.71662676e-01 -2.45778650e-01 -3.50077480e-01 3.65535319e-01 2.46655032e-01 -7.38736928e-01 2.47462600e-01 6.40639616e-03 4.17968690e-01 1.36128828e-01 4.49468374e-01 -4.43914205e-01 -6.37573004e-01 2.43888155e-01 -3.43907863e-01 -2.06222728e-01 3.00807714e-01 4.76287514e-01 3.87624383e-01 -7.01232672e-01 6.67356178e-02 -6.85834527e-01 2.12165803e-01 -2.51929373e-01 -9.17275190e-01 1.60969064e-01 -2.42219940e-01 2.56694466e-01 1.25307947e-01 -1.51271194e-01 -8.19451928e-01 8.41711819e-01 -3.65833849e-01 -3.95991445e-01 -4.14223701e-01 7.87450671e-02 -4.24023569e-01 -1.33494020e-01 6.04027867e-01 -2.57469535e-01 1.05763294e-01 -5.84654987e-01 3.28140587e-01 7.07285762e-01 9.72341895e-01 -4.48869258e-01 7.37256110e-01 5.74699879e-01 2.93448925e-01 -8.03958535e-01 -9.19620693e-01 -3.21489125e-01 -1.25587690e+00 -1.19249962e-01 1.02466488e+00 -1.06058455e+00 -9.98875320e-01 5.41982353e-01 -1.15513968e+00 -6.56426996e-02 3.14198732e-01 4.99576509e-01 -6.18319392e-01 9.17765573e-02 -1.31485656e-01 -7.49160171e-01 -4.55494344e-01 -1.23312891e+00 1.81111729e+00 3.15494448e-01 -2.56768823e-01 -7.77548194e-01 1.12237051e-01 1.90390110e-01 1.80997133e-01 5.88289499e-01 2.78310835e-01 -3.90849262e-01 -4.13438767e-01 -6.62926972e-01 -8.29987898e-02 1.02335803e-01 -2.64070984e-02 7.65752941e-02 -1.56678998e+00 -4.35797155e-01 -1.25615597e-01 -4.46763933e-01 3.98746073e-01 5.98223686e-01 9.75668609e-01 -2.90297806e-01 -2.72316843e-01 1.00668621e+00 9.88823652e-01 -4.45053577e-01 3.52258176e-01 1.47895023e-01 7.16303825e-01 8.86047482e-01 3.15053493e-01 4.40052032e-01 6.81773484e-01 9.57383275e-01 5.32252252e-01 6.73374310e-02 -4.63659875e-02 -3.31638902e-01 4.10107613e-01 3.81844699e-01 -7.37842098e-02 2.76384532e-01 -8.69888961e-01 2.92659551e-01 -1.53853536e+00 -3.25501919e-01 2.38395154e-01 2.30549836e+00 8.12498331e-01 -4.02194727e-03 2.10590780e-01 5.47919497e-02 4.05801147e-01 -1.64506570e-01 -5.17736077e-01 -2.68672466e-01 3.31892729e-01 3.66376430e-01 5.06299376e-01 7.95250773e-01 -1.22315562e+00 9.41090405e-01 6.04491282e+00 -1.38636634e-01 -1.74053979e+00 -2.08628587e-02 4.23071086e-01 -8.72103795e-02 2.47711927e-01 -2.43642569e-01 -1.15794873e+00 1.52910322e-01 9.38647866e-01 2.78550327e-01 3.68196517e-01 8.98409247e-01 -3.86841334e-02 -7.62215182e-02 -1.50011468e+00 1.25294268e+00 3.76855791e-01 -5.73268235e-01 -4.80964869e-01 5.77320829e-02 4.11490202e-01 1.58174664e-01 2.00814918e-01 2.29969382e-01 4.22397628e-02 -1.36785066e+00 8.16288829e-01 7.21115053e-01 8.11139226e-01 -4.80378538e-01 6.88796282e-01 3.56204361e-01 -9.03624892e-01 1.80895492e-01 -1.77882150e-01 1.48225967e-02 -4.23227390e-03 -9.57230926e-02 -1.23865378e+00 1.21128134e-01 7.98221409e-01 3.63834113e-01 -7.66643643e-01 9.77813601e-01 -3.65324229e-01 1.06249213e-01 -8.43829334e-01 2.21437871e-01 -3.90895419e-02 2.66159475e-01 2.76784062e-01 1.11531961e+00 2.82249719e-01 -2.07644820e-01 -5.10373227e-02 5.94060302e-01 -8.83664265e-02 -3.95237394e-02 -5.91839731e-01 5.49759388e-01 1.01892538e-01 1.46220315e+00 -3.37790251e-01 2.63963670e-01 -2.79510021e-01 8.65373671e-01 3.73539031e-01 2.32357427e-01 -6.71439946e-01 -1.35161147e-01 8.21135044e-01 2.63083011e-01 1.41727492e-01 -1.50360867e-01 -1.68313771e-01 -9.90168512e-01 8.79501700e-02 -6.47991300e-01 1.45568162e-01 -8.87705445e-01 -7.75875866e-01 7.57034302e-01 2.16476843e-01 -7.84429371e-01 -7.90951192e-01 -9.79310274e-01 -5.12550712e-01 9.99196351e-01 -1.36352229e+00 -1.47908640e+00 -6.29361093e-01 6.19376779e-01 2.51675040e-01 1.03849366e-01 7.99868464e-01 2.51996249e-01 -4.92047131e-01 1.07789946e+00 -5.22362113e-01 2.65022635e-01 9.93240476e-01 -1.23294210e+00 3.47164869e-01 3.96990567e-01 3.88791300e-02 6.03703618e-01 1.09342396e+00 -1.42063826e-01 -1.50122046e+00 -8.54894102e-01 8.99745047e-01 -9.42723691e-01 1.88133523e-01 -7.29026020e-01 -6.25783145e-01 1.02652228e+00 4.90306839e-02 6.40845239e-01 2.63500631e-01 1.54084954e-02 -2.70289779e-01 -1.55783981e-01 -1.41176808e+00 2.59414256e-01 7.03547895e-01 -6.25485241e-01 -4.43819672e-01 2.48226464e-01 2.79103160e-01 -8.50352526e-01 -5.91461539e-01 4.09700423e-01 9.46404696e-01 -7.57000864e-01 9.66996193e-01 -6.65042818e-01 -6.89122677e-02 -2.63047725e-01 -2.08352789e-01 -1.26554525e+00 2.35406727e-01 -6.77471936e-01 3.71302525e-03 1.01647782e+00 2.45006561e-01 -3.65423352e-01 9.77264404e-01 8.82754624e-01 2.04736739e-01 -6.93358541e-01 -1.12155581e+00 -1.09793492e-01 6.39422238e-02 -2.34225005e-01 8.39700639e-01 4.36464995e-01 -1.63799182e-01 3.25862288e-01 -2.47647166e-01 5.97727597e-01 7.77510405e-01 -3.46632123e-01 1.07126689e+00 -1.38431370e+00 3.07264656e-01 -8.90189633e-02 -8.75445843e-01 -7.06529558e-01 5.74933231e-01 -5.46995401e-01 2.85970241e-01 -1.04247856e+00 -9.97785255e-02 7.90980309e-02 2.03697816e-01 5.79144180e-01 3.22290920e-02 4.13926542e-01 -2.75901966e-02 -2.64274627e-01 -2.93181509e-01 3.20243120e-01 1.01169968e+00 1.64740518e-01 5.72001413e-02 1.14576176e-01 -2.57487535e-01 1.22219193e+00 4.32066083e-01 -2.63860881e-01 -9.00757015e-02 -4.92604405e-01 2.55097568e-01 -7.84593821e-02 4.15656716e-01 -9.69909430e-01 5.47548473e-01 2.36268908e-01 7.64563441e-01 -6.07332468e-01 6.72866166e-01 -9.31068659e-01 -1.86493341e-02 9.92924422e-02 -1.04970522e-01 1.61758915e-01 2.41875485e-01 2.01818794e-01 -9.52292234e-02 -4.08267230e-02 9.80442762e-01 -1.52502805e-02 -3.76960427e-01 4.33421046e-01 2.48469010e-01 -9.38015282e-02 8.81020248e-01 -9.87819955e-02 9.14533511e-02 -4.93644834e-01 -1.05166483e+00 5.97173497e-02 4.54343766e-01 4.29765522e-01 4.76477802e-01 -1.25625658e+00 -6.16213858e-01 8.80181968e-01 1.35587022e-01 2.55703866e-01 -2.33560741e-01 8.91479135e-01 -6.71353042e-01 5.56503236e-01 -3.15441370e-01 -8.59156728e-01 -1.37304664e+00 2.61340916e-01 7.85001397e-01 3.01234037e-01 -2.41541892e-01 1.14614677e+00 1.95808277e-01 -9.30778742e-01 6.88554406e-01 -6.77579522e-01 -2.48181522e-01 2.09226713e-01 4.03149426e-01 4.31092083e-02 4.79173839e-01 -1.28841972e+00 -7.13681042e-01 9.52998698e-01 9.84180048e-02 -1.23109519e-01 1.43673348e+00 -2.38615021e-01 -1.11755140e-01 1.54213786e-01 1.71428514e+00 1.96562726e-02 -1.50178254e+00 -1.48187682e-01 7.47448429e-02 -4.52366114e-01 1.47914425e-01 -5.49764454e-01 -1.08543742e+00 1.05819845e+00 1.02265811e+00 -4.33635741e-01 8.45002413e-01 1.11226425e-01 4.19737220e-01 3.06515276e-01 3.98685038e-01 -1.01365948e+00 -1.46710709e-01 3.92567068e-01 1.16214716e+00 -1.44679856e+00 5.65776750e-02 -2.35513330e-01 -3.93601358e-01 1.11939907e+00 6.73638940e-01 -1.37400478e-01 1.07744658e+00 4.35184896e-01 2.77166635e-01 -2.70816118e-01 -4.29982781e-01 -1.64981872e-01 7.16185331e-01 5.80742478e-01 8.78069282e-01 1.32309571e-01 1.96174562e-01 4.02762324e-01 -8.67251217e-01 7.12925661e-03 1.13447823e-01 7.74125814e-01 -2.45605603e-01 -8.43231380e-01 -7.19189703e-01 8.98940861e-03 -5.60931504e-01 3.80563825e-01 -3.98186713e-01 6.89893901e-01 2.48024285e-01 5.24924755e-01 3.28786708e-02 -1.01340704e-01 5.30696392e-01 2.88259357e-01 9.13060844e-01 -6.39538705e-01 -1.81438774e-01 5.00946306e-02 -3.32787246e-01 -5.90670228e-01 -2.74132609e-01 -8.26227725e-01 -1.13148475e+00 -4.52071242e-02 -2.07159951e-01 -3.59018177e-01 1.13182855e+00 1.04230630e+00 -3.03797219e-02 -8.62607360e-03 4.12463039e-01 -1.51131034e+00 -6.48741364e-01 -1.21692526e+00 -4.16982830e-01 2.90740490e-01 8.89951468e-01 -1.00461125e+00 -3.20375055e-01 1.44943967e-01]

[13.622071266174316, 0.27574923634529114]

3ab96f68-273d-45a0-95ab-f126dd4aae18

cross-modal-hierarchical-modelling-for-fine

2007.15103

https://arxiv.org/abs/2007.15103v2

https://arxiv.org/pdf/2007.15103v2.pdf

Cross-Modal Hierarchical Modelling for Fine-Grained Sketch Based Image Retrieval

Sketch as an image search query is an ideal alternative to text in capturing the fine-grained visual details. Prior successes on fine-grained sketch-based image retrieval (FG-SBIR) have demonstrated the importance of tackling the unique traits of sketches as opposed to photos, e.g., temporal vs. static, strokes vs. pixels, and abstract vs. pixel-perfect. In this paper, we study a further trait of sketches that has been overlooked to date, that is, they are hierarchical in terms of the levels of detail -- a person typically sketches up to various extents of detail to depict an object. This hierarchical structure is often visually distinct. In this paper, we design a novel network that is capable of cultivating sketch-specific hierarchies and exploiting them to match sketch with photo at corresponding hierarchical levels. In particular, features from a sketch and a photo are enriched using cross-modal co-attention, coupled with hierarchical node fusion at every level to form a better embedding space to conduct retrieval. Experiments on common benchmarks show our method to outperform state-of-the-arts by a significant margin.

['Yi-Zhe Song', 'Tao Xiang', 'Ayan Kumar Bhunia', 'Aneeshan Sain', 'Yongxin Yang']

2020-07-29

['sketch-based-image-retrieval']

['computer-vision']

[ 4.13549356e-02 -3.75529826e-01 -3.15495610e-01 -2.37062722e-01 -5.81307113e-01 -6.57169282e-01 9.48521435e-01 -2.52480246e-02 7.65353665e-02 4.15096998e-01 6.33651078e-01 1.67035788e-01 -2.88393408e-01 -7.92686462e-01 -5.63685715e-01 -5.63543975e-01 1.00745484e-01 2.67305933e-02 1.16126254e-01 -2.76352465e-01 6.02432907e-01 9.58225071e-01 -1.61871278e+00 5.92886269e-01 3.04839849e-01 1.03104448e+00 2.06463188e-01 3.65737468e-01 -6.67745948e-01 5.16318440e-01 -4.68874156e-01 -6.08381331e-01 2.37507254e-01 -1.88362867e-01 -5.37703812e-01 3.00342411e-01 1.17917848e+00 -6.22449934e-01 -7.31503725e-01 9.46267128e-01 3.51024598e-01 -9.57400873e-02 8.28970194e-01 -1.27986753e+00 -1.22978151e+00 4.03060466e-01 -7.07163334e-01 -2.80125644e-02 3.99742544e-01 7.73356482e-02 1.35808170e+00 -1.11922371e+00 7.59851038e-01 1.67622256e+00 4.59583908e-01 6.55471206e-01 -1.20096004e+00 -7.04499364e-01 4.64508474e-01 4.62032370e-02 -1.70726836e+00 -3.40653986e-01 1.36408544e+00 -3.09924245e-01 4.10737902e-01 3.54347676e-01 6.95202172e-01 1.25025582e+00 -8.23533162e-02 1.09399927e+00 1.09458280e+00 -2.67568380e-01 -4.05935757e-02 9.24606174e-02 -4.93490919e-02 8.76011372e-01 1.02179863e-01 -9.01434347e-02 -8.05976152e-01 -2.23168492e-01 1.43389177e+00 4.70630080e-01 -2.24807322e-01 -4.69826728e-01 -1.32442486e+00 7.02351034e-01 7.48178959e-01 6.42354667e-01 -4.84580874e-01 3.68926674e-01 3.53255391e-01 1.89199522e-01 2.99056858e-01 4.92541701e-01 1.45028800e-01 2.98198849e-01 -1.17443848e+00 2.97738314e-01 5.85028231e-01 1.14213037e+00 8.12034547e-01 -4.18484434e-02 -5.58705926e-01 9.73370910e-01 1.65249005e-01 1.70646846e-01 2.48831436e-01 -8.34721088e-01 3.10645401e-01 8.66691113e-01 -8.58519524e-02 -1.33207834e+00 2.37504676e-01 -3.24093610e-01 -1.27681673e+00 1.59418091e-01 5.87576674e-03 7.20408976e-01 -9.73233223e-01 1.56905103e+00 6.27974793e-02 7.07889944e-02 -3.60834539e-01 1.05981851e+00 9.95634317e-01 6.15709603e-01 1.34551242e-01 3.64933759e-01 1.60091066e+00 -1.01356220e+00 -6.57754242e-01 4.53012623e-02 -3.05888653e-01 -6.13640130e-01 1.37958360e+00 1.02418035e-01 -1.27830076e+00 -7.91618407e-01 -1.14947462e+00 -2.49492079e-01 -7.29438782e-01 1.08509272e-01 4.91339028e-01 2.55095065e-01 -1.23346055e+00 7.12123096e-01 -2.13782474e-01 -5.18592715e-01 5.79677284e-01 -1.58009678e-01 -5.22684038e-01 -2.95455724e-01 -1.13309991e+00 4.16976601e-01 1.02074817e-01 -1.12438574e-01 -7.75697231e-01 -9.38129663e-01 -6.96521223e-01 3.33660841e-01 3.33255202e-01 -8.60871851e-01 5.76864064e-01 -7.37367570e-01 -1.17257559e+00 8.21702242e-01 -7.49862641e-02 3.80165912e-02 6.20229065e-01 9.97468531e-02 -3.10000628e-01 5.18705070e-01 1.51094288e-01 1.01787257e+00 1.32443237e+00 -1.64095700e+00 -5.07402241e-01 -3.75236154e-01 4.25208002e-01 1.44313797e-01 -7.13527560e-01 -1.71917409e-01 -8.72822642e-01 -9.32218134e-01 1.15907528e-01 -6.75811946e-01 8.38253796e-02 6.85082078e-01 -2.42268294e-01 -6.31192446e-01 9.47082102e-01 -3.57129455e-01 1.51672029e+00 -2.15603495e+00 2.34314188e-01 7.45639279e-02 6.69193745e-01 5.76874651e-02 -4.27699625e-01 1.16414702e+00 1.55511513e-01 4.61251140e-01 5.22903353e-02 -3.40344518e-01 3.37490231e-01 1.66062146e-01 -4.36132520e-01 1.00109994e-01 4.34310645e-01 1.30393553e+00 -8.80811393e-01 -8.23168099e-01 2.28270546e-01 7.51768589e-01 -1.89593151e-01 1.22434370e-01 4.96232323e-02 7.95964450e-02 -7.27053702e-01 9.98755872e-01 5.97219229e-01 -5.75957716e-01 9.56776366e-03 -7.24605799e-01 -1.44953191e-01 -1.72415689e-01 -1.04909909e+00 1.87354898e+00 -3.87815177e-01 5.91159701e-01 1.51476875e-01 -6.12057924e-01 8.31741512e-01 1.90263823e-01 2.65965641e-01 -6.33019209e-01 -2.71583974e-01 -1.39156669e-01 -6.24221027e-01 -1.60428137e-01 7.16812789e-01 3.70453112e-02 -1.73121527e-01 2.45526418e-01 -9.85569358e-02 -2.56388813e-01 7.88668916e-02 4.73579019e-01 8.32359076e-01 -9.87440944e-02 2.69124508e-01 -3.04175735e-01 4.40654665e-01 -3.72605711e-01 7.82772973e-02 9.62361634e-01 -1.03002056e-01 7.43711233e-01 2.42037982e-01 -6.77110612e-01 -1.25487268e+00 -1.05054486e+00 -2.31802501e-02 1.08879173e+00 3.98902923e-01 -5.77032387e-01 -4.12617505e-01 -4.07202959e-01 5.09390533e-01 3.55824605e-02 -8.94188821e-01 7.68190026e-02 -4.38509792e-01 1.73766866e-01 5.70807159e-01 5.73296070e-01 5.68977535e-01 -1.20422781e+00 -4.04622793e-01 5.59557304e-02 1.32729769e-01 -1.10727358e+00 -8.34949970e-01 -4.19321865e-01 -5.34203410e-01 -7.46884465e-01 -1.26925564e+00 -8.07766259e-01 6.83316767e-01 7.12751806e-01 1.45479620e+00 5.22502542e-01 -5.57300270e-01 7.96198368e-01 -3.96159649e-01 -2.98487674e-02 6.98854625e-02 -1.11942224e-01 -2.25221366e-01 3.67342651e-01 -3.18953358e-02 -6.84961200e-01 -9.80198383e-01 1.77358657e-01 -1.35004878e+00 1.54866144e-01 1.02570665e+00 1.00478518e+00 5.50517440e-01 4.26846258e-02 3.38321298e-01 -5.13614953e-01 9.12607551e-01 -2.95441717e-01 -9.36794356e-02 7.21831799e-01 -3.79659921e-01 9.22921151e-02 4.55089480e-01 -5.92799246e-01 -7.06684709e-01 -1.84498459e-01 2.57256180e-01 -9.96336520e-01 -1.86284423e-01 3.89788121e-01 -5.16769625e-02 -3.48173410e-01 1.24003582e-01 4.46279943e-01 -1.89334199e-01 -5.14903426e-01 6.00063086e-01 4.39365864e-01 3.90667439e-01 -9.14742291e-01 9.46282089e-01 5.09075701e-01 1.97967306e-01 -1.09309530e+00 -5.38061976e-01 -4.73169118e-01 -5.72311401e-01 -3.02390277e-01 8.12831938e-01 -8.60617518e-01 -7.81636596e-01 2.52895087e-01 -1.05758822e+00 1.12956271e-01 -7.97683820e-02 -2.85919398e-01 -2.08156168e-01 5.81391811e-01 -7.00725794e-01 -7.70825326e-01 -5.34064949e-01 -9.52210307e-01 1.66815579e+00 1.87559769e-01 2.59878002e-02 -7.07857549e-01 -2.99653977e-01 7.46181458e-02 6.81276500e-01 1.95358723e-01 1.00829971e+00 -4.20741327e-02 -9.54121768e-01 -2.02798098e-01 -8.49653006e-01 5.09177009e-03 2.03836724e-01 7.05561489e-02 -7.80240893e-01 -3.66048187e-01 -5.59607923e-01 -3.85976315e-01 9.87221301e-01 1.87422857e-02 1.46081436e+00 -4.95837927e-01 -3.27648938e-01 3.51503491e-01 1.69460118e+00 -1.63303033e-01 6.56651378e-01 -5.42206038e-03 9.06278014e-01 5.24291754e-01 2.86613047e-01 3.74006480e-01 3.89029473e-01 7.28685081e-01 3.11944216e-01 -1.63087919e-01 -4.98362362e-01 -6.84548020e-01 -1.88105315e-01 6.78321660e-01 7.84569606e-02 -1.14005856e-01 -5.23337245e-01 6.55606508e-01 -1.63608575e+00 -1.13982332e+00 3.67880613e-01 1.91215897e+00 8.20589125e-01 -2.14067385e-01 1.80016868e-02 -1.30295113e-01 8.01799655e-01 6.97270691e-01 -4.15597498e-01 -5.04681133e-02 -1.96340337e-01 4.64320667e-02 2.12308586e-01 1.47615477e-01 -9.55658495e-01 1.03720057e+00 5.60205460e+00 1.19181859e+00 -1.09112096e+00 -3.09958130e-01 3.66823107e-01 1.42285436e-01 -6.51388943e-01 -2.20086977e-01 -5.64172268e-01 4.73699182e-01 4.18144464e-02 1.25833184e-01 7.10871220e-01 6.16757274e-01 -4.35947061e-01 4.59584326e-01 -1.18080127e+00 1.32316232e+00 2.34184027e-01 -1.72582853e+00 7.73538947e-01 6.51189312e-02 7.37135708e-01 -5.51433980e-01 1.83160737e-01 3.31958205e-01 3.19054313e-02 -9.81220067e-01 8.74340534e-01 7.76749313e-01 1.11947191e+00 -4.89006877e-01 1.69939548e-01 1.02432005e-01 -1.75949430e+00 -5.22356443e-02 -4.37314868e-01 2.83970565e-01 -9.43310335e-02 1.45961344e-01 -2.05502614e-01 8.43496442e-01 7.58300364e-01 7.94291377e-01 -6.65218771e-01 7.58440614e-01 1.86722726e-01 -5.10916822e-02 -1.35430684e-02 -2.63130873e-01 6.08469486e-01 -1.96027994e-01 3.29987228e-01 1.33190453e+00 3.15106601e-01 2.23698109e-01 2.54447132e-01 1.16502118e+00 -2.42271096e-01 -1.62347835e-02 -7.97305763e-01 -4.96059299e-01 7.06004679e-01 1.39013362e+00 -5.67804575e-01 -5.39504111e-01 -3.72758389e-01 1.26856720e+00 5.12158334e-01 4.83190238e-01 -4.26409930e-01 -3.84844035e-01 6.76869631e-01 1.45754039e-01 4.85095561e-01 -2.73841947e-01 5.50573878e-02 -1.11772704e+00 9.41390470e-02 -7.20083773e-01 1.85475022e-01 -9.60767448e-01 -1.96324694e+00 6.84121430e-01 -2.59917766e-01 -1.25907648e+00 3.01774114e-01 -4.94003326e-01 -6.05226457e-01 1.01208186e+00 -1.64534378e+00 -1.78804457e+00 -6.98126793e-01 7.63946295e-01 8.36264729e-01 -3.14836390e-02 7.17199445e-01 4.46607620e-01 -1.85863093e-01 7.31606364e-01 -2.83490032e-01 2.73894221e-01 7.11853862e-01 -1.22069442e+00 4.96269763e-01 3.20508391e-01 4.03365761e-01 8.92093062e-01 2.90622801e-01 -6.15625143e-01 -1.86012304e+00 -8.77158046e-01 7.57866740e-01 -3.45092505e-01 7.45206356e-01 -4.54051912e-01 -8.85908186e-01 1.44761011e-01 2.68827885e-01 1.51328504e-01 1.98794439e-01 -3.46446894e-02 -8.94172072e-01 -2.30178937e-01 -9.55984056e-01 9.35379207e-01 1.30270672e+00 -1.17633080e+00 -4.80825871e-01 -5.41217737e-02 7.13751435e-01 -3.87091748e-02 -1.05450082e+00 2.56005466e-01 1.22166324e+00 -8.30887496e-01 1.49704647e+00 -6.40183568e-01 7.10122645e-01 -2.06098855e-01 -3.64481539e-01 -9.74244356e-01 -7.89080679e-01 -4.38675970e-01 -4.57407255e-03 1.49200523e+00 -1.89948410e-01 -2.12913662e-01 7.19766736e-01 2.86911726e-01 8.66515040e-02 -8.95681143e-01 -6.57001853e-01 -6.12795949e-01 -6.07342683e-02 1.14498988e-01 1.00194728e+00 1.14596915e+00 -3.55931014e-01 2.55981892e-01 -6.15291715e-01 3.40422019e-02 7.49368906e-01 4.72074836e-01 7.65524566e-01 -1.38543677e+00 1.97454482e-01 -1.06007600e+00 -5.49157023e-01 -1.12467062e+00 1.55745775e-01 -7.04566061e-01 -3.48418146e-01 -1.69648850e+00 3.66930306e-01 -4.41270292e-01 -5.65713882e-01 2.30035409e-01 -2.98068583e-01 4.66619939e-01 4.90015268e-01 4.21727061e-01 -6.83749139e-01 5.15478373e-01 1.57707691e+00 -5.66645622e-01 2.10418776e-01 -5.24201512e-01 -7.89433002e-01 2.14458987e-01 3.99671495e-01 -2.48054415e-02 -2.89288133e-01 -4.04641569e-01 8.54847059e-02 2.46977642e-01 7.78432488e-01 -7.51918435e-01 4.13116127e-01 -8.05341750e-02 6.64367020e-01 -8.57180655e-01 6.96063399e-01 -1.09943128e+00 1.93791255e-01 5.82983866e-02 -5.75301707e-01 8.08005556e-02 -9.57403611e-03 8.87345254e-01 -4.20559585e-01 1.30360618e-01 3.80744338e-01 -4.45898294e-01 -9.01951969e-01 7.05076933e-01 1.72031924e-01 -2.52304137e-01 5.66672981e-01 -3.28304917e-01 -3.24550718e-01 -4.33411986e-01 -3.91040355e-01 -5.61501384e-02 6.14879310e-01 6.74346030e-01 8.04415882e-01 -1.87113845e+00 -6.72424734e-01 2.35242575e-01 6.28557086e-01 -5.64513922e-01 5.29300570e-01 4.65544224e-01 -1.40903771e-01 6.19000793e-01 -4.95727926e-01 -5.61663032e-01 -1.21686375e+00 8.20136130e-01 -5.83392158e-02 -2.40674824e-01 -7.85941482e-01 7.25651324e-01 6.80635870e-01 -4.48834673e-02 4.50770557e-01 -1.54287010e-01 -5.12742959e-02 2.54849315e-01 5.62961936e-01 1.15242660e-01 -3.49461287e-01 -4.63921130e-01 -1.86384961e-01 1.06717145e+00 -2.87784450e-02 1.29085526e-01 1.09375215e+00 -2.05646753e-01 -2.38297135e-01 4.80104625e-01 1.40604901e+00 2.24508662e-02 -1.32086289e+00 -6.52639687e-01 -9.94058773e-02 -8.09048891e-01 3.75140905e-02 -8.21021616e-01 -1.11903691e+00 9.55748439e-01 3.27086002e-01 4.67514455e-01 9.66630459e-01 1.88399106e-01 6.64431930e-01 3.05255920e-01 5.28605461e-01 -7.73989081e-01 4.09621477e-01 1.07783087e-01 1.42141855e+00 -1.22566760e+00 1.49913684e-01 -1.22395433e-01 -5.71312249e-01 1.18827260e+00 3.01537663e-01 -3.92994732e-01 5.07551491e-01 -1.40238658e-01 -3.06885689e-01 -5.07014453e-01 -6.30108774e-01 -3.00768673e-01 9.26438212e-01 1.50134981e-01 2.25411698e-01 1.63934037e-01 -4.78424206e-02 3.23513091e-01 3.49346638e-01 -1.70200035e-01 -1.20313987e-01 8.46364856e-01 -3.20160121e-01 -1.07063079e+00 -2.62591243e-01 4.39646244e-01 -2.24805623e-01 -2.13783413e-01 -8.77716720e-01 8.81682456e-01 -1.95247099e-01 6.15700901e-01 -3.22985202e-02 -3.23840886e-01 3.15941453e-01 4.75335680e-02 5.74973404e-01 -2.58812219e-01 -5.80313563e-01 -4.91521135e-02 -2.80643553e-01 -6.87570214e-01 -3.20067227e-01 -2.67387033e-01 -4.79635119e-01 -3.30919772e-01 1.27729148e-01 -4.64774966e-02 7.10084856e-01 4.17733282e-01 5.05815268e-01 5.13598919e-01 6.60445094e-01 -1.34824812e+00 -4.62285727e-01 -7.63440907e-01 -6.90785885e-01 5.87116361e-01 5.59024751e-01 -6.99781835e-01 -4.85130191e-01 -2.24367335e-01]

[11.687507629394531, 0.5719165802001953]

2488f3ef-b641-4f93-9f8a-70b72794cafb

learning-adversarial-semantic-embeddings-for

2307.03416

https://arxiv.org/abs/2307.03416v1

https://arxiv.org/pdf/2307.03416v1.pdf

Learning Adversarial Semantic Embeddings for Zero-Shot Recognition in Open Worlds

Zero-Shot Learning (ZSL) focuses on classifying samples of unseen classes with only their side semantic information presented during training. It cannot handle real-life, open-world scenarios where there are test samples of unknown classes for which neither samples (e.g., images) nor their side semantic information is known during training. Open-Set Recognition (OSR) is dedicated to addressing the unknown class issue, but existing OSR methods are not designed to model the semantic information of the unseen classes. To tackle this combined ZSL and OSR problem, we consider the case of "Zero-Shot Open-Set Recognition" (ZS-OSR), where a model is trained under the ZSL setting but it is required to accurately classify samples from the unseen classes while being able to reject samples from the unknown classes during inference. We perform large experiments on combining existing state-of-the-art ZSL and OSR models for the ZS-OSR task on four widely used datasets adapted from the ZSL task, and reveal that ZS-OSR is a non-trivial task as the simply combined solutions perform badly in distinguishing the unseen-class and unknown-class samples. We further introduce a novel approach specifically designed for ZS-OSR, in which our model learns to generate adversarial semantic embeddings of the unknown classes to train an unknowns-informed ZS-OSR classifier. Extensive empirical results show that our method 1) substantially outperforms the combined solutions in detecting the unknown classes while retaining the classification accuracy on the unseen classes and 2) achieves similar superiority under generalized ZS-OSR settings.

['Xin Ning', 'Lei Zhou', 'Jin Zheng', 'Xiao Bai', 'Guansong Pang', 'Tianqi Li']

2023-07-07

['zero-shot-learning', 'open-set-learning']

['methodology', 'miscellaneous']

[ 5.58155298e-01 2.41378963e-01 -2.55642176e-01 -3.03403586e-01 -1.06544721e+00 -5.95613897e-01 5.40079951e-01 -1.21973440e-01 -1.39038429e-01 5.89387298e-01 -2.53284544e-01 -2.25568786e-01 -2.42386032e-02 -1.00050247e+00 -9.26527917e-01 -6.41832948e-01 2.40908727e-01 8.11040640e-01 3.24702978e-01 -2.71174729e-01 3.08425911e-02 2.96414167e-01 -2.08350277e+00 3.29360902e-01 5.98018825e-01 1.15260160e+00 -1.63758248e-01 7.20092416e-01 -1.77903518e-01 5.45255423e-01 -7.24753737e-01 -4.52668428e-01 4.60779101e-01 -1.35923371e-01 -7.20391750e-01 1.94480866e-01 6.28701031e-01 -2.11420551e-01 -5.04372776e-01 1.22595167e+00 5.52970290e-01 3.36735934e-01 8.81866753e-01 -1.56285572e+00 -1.09838533e+00 1.45311147e-01 -1.40751988e-01 1.16488256e-01 3.92799109e-01 2.40526631e-01 9.82985616e-01 -1.01531351e+00 5.98481774e-01 1.28175998e+00 6.48579359e-01 1.20351005e+00 -1.32172382e+00 -7.53938198e-01 1.09952711e-01 3.40066731e-01 -1.45876968e+00 -5.30538917e-01 5.73909879e-01 -3.46989155e-01 5.13325930e-01 4.96518254e-01 7.77855981e-03 1.76860094e+00 -1.84467956e-01 8.74553323e-01 9.73765910e-01 -3.05595905e-01 7.00567842e-01 3.69901627e-01 5.74108958e-01 3.07148039e-01 2.96680480e-01 3.02314162e-01 -2.41014302e-01 -2.09817454e-01 3.34300697e-01 3.23250592e-01 -4.05966729e-01 -5.35220146e-01 -9.19298828e-01 8.75045061e-01 3.15380305e-01 1.20811567e-01 1.41982317e-01 -1.87110946e-01 2.95227826e-01 4.59336430e-01 7.45943546e-01 4.82461780e-01 -5.98203003e-01 2.41545051e-01 -6.94725156e-01 1.50791742e-02 9.25501406e-01 1.04229093e+00 8.44581366e-01 -8.42499267e-03 -3.65871966e-01 9.09610271e-01 -1.36375174e-01 4.20806229e-01 8.95276546e-01 -4.65859979e-01 2.35245883e-01 5.26723921e-01 1.10020757e-01 -6.21744275e-01 2.05353022e-01 -4.09128904e-01 -5.82309723e-01 1.51090041e-01 3.95494729e-01 1.53748468e-01 -1.42458200e+00 1.66883659e+00 4.26817268e-01 7.17423797e-01 6.98293328e-01 7.98874378e-01 1.22485900e+00 6.00603878e-01 -1.74032226e-01 1.87929824e-01 1.33315814e+00 -6.98262155e-01 -5.09551048e-01 -5.54786146e-01 6.04140997e-01 -2.75683820e-01 1.18015063e+00 7.46907070e-02 -3.88371050e-01 -3.86886179e-01 -1.26815569e+00 1.86339933e-02 -1.13344395e+00 -9.48650464e-02 3.41006696e-01 6.86535656e-01 -6.12648904e-01 6.12515032e-01 -3.43658686e-01 -3.63534719e-01 8.37284923e-01 1.83771938e-01 -5.86376011e-01 -7.76094019e-01 -1.26123059e+00 7.26833522e-01 4.08080816e-01 -1.12425648e-01 -1.40593946e+00 -7.79646337e-01 -1.16076791e+00 2.41309106e-01 9.47459996e-01 -4.26383793e-01 1.10713458e+00 -1.16232562e+00 -1.09632730e+00 1.09305644e+00 -4.15597297e-02 -3.51843506e-01 5.86523414e-01 -3.59431580e-02 -6.13815963e-01 7.63765955e-03 3.05107504e-01 3.45829874e-01 1.13046372e+00 -1.49095380e+00 -2.90313035e-01 -6.68362677e-01 1.47962257e-01 -6.42925575e-02 -2.65348852e-01 -3.29727650e-01 -9.96615812e-02 -7.68240809e-01 2.43295863e-01 -6.57436848e-01 -1.18045367e-01 1.92243084e-01 -5.51326334e-01 -2.51711339e-01 1.10462844e+00 -1.24929160e-01 7.55149961e-01 -2.34586954e+00 -3.25535864e-01 -6.97121769e-02 2.19276547e-01 6.67833745e-01 -3.64860654e-01 9.75577608e-02 -4.00803208e-01 5.99575900e-02 -3.37467134e-01 -2.02116832e-01 1.14942968e-01 6.58589721e-01 -8.13898087e-01 3.98892850e-01 4.72238898e-01 8.28729808e-01 -1.19174337e+00 -5.89024983e-02 2.45927244e-01 2.69466072e-01 -2.28876546e-01 5.38853228e-01 -2.90241599e-01 1.03365339e-01 -3.37484211e-01 8.85741830e-01 5.81020176e-01 -2.13980004e-01 -2.66237706e-01 4.08838838e-02 6.62776053e-01 -2.77559012e-01 -1.38179243e+00 1.13287306e+00 -3.70684952e-01 4.08175707e-01 -3.17339748e-01 -1.17769682e+00 9.25515473e-01 4.45329577e-01 -1.43845692e-01 -3.96887958e-01 2.06778213e-01 3.16435486e-01 -3.09742361e-01 -5.66594243e-01 2.24341646e-01 -5.19578755e-01 -2.22051546e-01 9.91815627e-02 5.72259367e-01 6.93551525e-02 -1.46286145e-01 1.08718723e-01 1.25262618e+00 -7.59975016e-02 3.63567293e-01 -8.44543204e-02 4.19439286e-01 -2.11500525e-01 7.20562398e-01 1.15371025e+00 -4.68992651e-01 9.97102857e-01 3.25222880e-01 -5.23490965e-01 -7.73521960e-01 -1.44980741e+00 -3.88670295e-01 8.99311960e-01 4.06593561e-01 3.55536938e-02 -5.89365780e-01 -1.14273763e+00 1.40858084e-01 1.03555405e+00 -9.95381773e-01 -5.80549300e-01 2.75720507e-01 -5.84273517e-01 3.85127872e-01 3.91452253e-01 -1.38374735e-02 -8.94744337e-01 -2.98651099e-01 -5.36422506e-02 -6.82180896e-02 -1.17324734e+00 -1.62875578e-01 2.50184953e-01 -4.47256804e-01 -1.50209999e+00 -6.56103075e-01 -6.59964740e-01 8.71373236e-01 3.23847950e-01 9.71318483e-01 -1.86908543e-01 -7.46295154e-01 5.03647804e-01 -5.96657634e-01 -5.61078429e-01 -3.69040787e-01 -4.56770360e-01 6.47061989e-02 6.18375003e-01 6.97590411e-01 -2.13602334e-01 -1.65797234e-01 6.23880923e-01 -1.13843918e+00 -4.26271230e-01 2.40041509e-01 1.05787075e+00 4.58067745e-01 1.92716286e-01 1.08952439e+00 -1.35108912e+00 1.87230080e-01 -8.56765747e-01 -2.16633171e-01 4.71734703e-01 -4.08887416e-01 -3.19242552e-02 8.84359360e-01 -9.23611164e-01 -6.98733270e-01 -3.02447736e-01 -5.75878620e-02 -1.07970989e+00 -5.58337569e-01 -1.65944308e-01 -5.59932888e-01 2.87545789e-02 7.66593516e-01 3.15184295e-01 -8.27412903e-02 -4.00281072e-01 2.87620068e-01 1.04790413e+00 6.39754117e-01 -6.69528544e-01 1.01663208e+00 6.72946155e-01 -2.82603949e-01 -9.04892445e-01 -1.41572297e+00 -6.66259050e-01 -3.33243579e-01 6.28328975e-03 7.32915223e-01 -9.72751856e-01 -1.17773667e-01 3.39818418e-01 -8.49380612e-01 -7.08586872e-02 -8.43786240e-01 2.02695251e-01 -6.25315726e-01 2.07754657e-01 -1.37098730e-01 -9.99258876e-01 -1.04273684e-01 -1.18897367e+00 1.31387305e+00 -1.05919726e-02 -6.43242672e-02 -9.23533559e-01 -1.59201816e-01 4.08780187e-01 1.43828139e-01 2.36846417e-01 8.99585962e-01 -1.51294756e+00 -3.41104627e-01 -7.14539647e-01 -8.29626173e-02 5.54149806e-01 4.77881171e-02 -4.16003048e-01 -1.48052382e+00 -3.82427990e-01 3.78917903e-02 -7.73488104e-01 8.71283531e-01 -7.91577920e-02 1.44159424e+00 -2.45246187e-01 -2.13148728e-01 4.63383168e-01 1.51097429e+00 5.71368299e-02 7.00007498e-01 -1.40455235e-02 5.76657176e-01 6.45149589e-01 7.28161454e-01 3.00672889e-01 -3.39261778e-02 3.45952094e-01 5.69220066e-01 3.35317627e-02 -2.98729744e-02 -3.91141593e-01 3.12602609e-01 9.90779400e-02 6.27990782e-01 -5.12703061e-01 -5.99563122e-01 7.20024645e-01 -1.78796315e+00 -9.33196843e-01 2.54120201e-01 2.52366757e+00 6.05900764e-01 1.62327752e-01 -4.03942883e-01 3.48704338e-01 1.06276023e+00 1.94485381e-01 -8.16552758e-01 -1.85920879e-01 -1.00956284e-01 5.85696697e-01 3.11060905e-01 2.23541811e-01 -1.27298164e+00 9.38199699e-01 5.63656378e+00 1.17891383e+00 -7.26441622e-01 2.35693157e-01 6.05358899e-01 -3.49873789e-02 -2.53346205e-01 3.08392439e-02 -9.12901878e-01 4.25060004e-01 8.08708251e-01 -3.46362293e-02 4.02766466e-01 1.15588522e+00 -4.42900866e-01 1.64975479e-01 -1.50743461e+00 1.11887109e+00 5.12354970e-01 -1.11560953e+00 3.84528637e-01 -1.88525319e-01 6.15276814e-01 -1.18565828e-01 6.00170270e-02 8.92360985e-01 3.52107853e-01 -9.78872061e-01 6.15223348e-01 3.05077463e-01 1.10006130e+00 -5.19212544e-01 7.90967524e-01 6.57878578e-01 -7.56789625e-01 -4.05373812e-01 -5.96369684e-01 1.69727039e-02 -3.86496484e-01 4.05716509e-01 -6.01336598e-01 5.18334210e-01 6.36712193e-01 7.85834312e-01 -6.47846401e-01 6.75901353e-01 -2.15818465e-01 4.80185211e-01 4.57313424e-03 3.55092376e-01 1.14878602e-01 3.59383337e-02 6.74748480e-01 6.47196889e-01 2.16185093e-01 1.39357686e-01 2.58272707e-01 1.10890090e+00 -1.56733066e-01 -3.02461892e-01 -1.03359020e+00 -1.93012387e-01 3.71740967e-01 1.01650119e+00 -5.86147606e-01 -5.76552868e-01 -3.71667415e-01 9.57509935e-01 3.23752910e-01 3.99086326e-01 -7.02502251e-01 -6.43199027e-01 9.40305054e-01 7.96940327e-02 3.96172255e-01 4.68507260e-01 3.17811482e-02 -1.63644040e+00 -6.62595555e-02 -7.09235847e-01 6.81497753e-01 -8.11754644e-01 -1.88071585e+00 4.53773826e-01 -3.75324816e-01 -1.40884089e+00 -3.13166082e-02 -9.89268184e-01 -5.91879249e-01 6.48934960e-01 -1.61917114e+00 -1.05873537e+00 -1.30133152e-01 6.15810692e-01 7.95822144e-01 -1.82678863e-01 1.13234520e+00 1.84156150e-01 -7.26029396e-01 7.34590352e-01 2.79238015e-01 3.57299298e-01 5.95696986e-01 -1.30170548e+00 2.61402607e-01 8.74821723e-01 1.08926773e-01 3.57802927e-01 6.64521277e-01 -5.42234778e-01 -1.34593153e+00 -1.50466776e+00 6.24229133e-01 -8.33703160e-01 7.32871473e-01 -8.14944744e-01 -1.15410602e+00 8.96384478e-01 -7.24186540e-01 9.20041144e-01 8.84436309e-01 -5.35442010e-02 -8.49057913e-01 4.28334773e-02 -1.49493468e+00 4.14598405e-01 9.83412027e-01 -8.22657824e-01 -1.07935452e+00 2.97412246e-01 9.61366713e-01 -2.33930200e-01 -6.25301182e-01 5.35821319e-01 2.34004930e-01 -6.09342992e-01 1.18706369e+00 -1.14609087e+00 4.65997487e-01 -2.39726767e-01 -5.84616244e-01 -1.27637362e+00 1.00508191e-01 -8.93488303e-02 -2.56709367e-01 1.24684131e+00 1.00269571e-01 -9.00979161e-01 7.05297828e-01 6.07013881e-01 -1.12718724e-01 -6.17551744e-01 -1.10771561e+00 -1.07146311e+00 -2.28996322e-01 -6.66206479e-01 4.84500676e-01 1.20477438e+00 -5.10467410e-01 2.92951167e-01 -2.75672108e-01 4.96658057e-01 7.87466228e-01 2.11009443e-01 7.12200284e-01 -1.46639085e+00 -3.30193609e-01 9.91908684e-02 -8.84627461e-01 -4.50215936e-01 6.21906519e-01 -1.07864940e+00 1.62626699e-01 -1.09725022e+00 2.49424353e-01 -4.41671878e-01 -4.75452036e-01 4.50778186e-01 -2.97079504e-01 3.67628843e-01 2.14447845e-02 -4.97175083e-02 -7.35164881e-01 6.75399899e-01 9.62929845e-01 -3.54252756e-01 1.71284556e-01 1.93268731e-01 -7.72389889e-01 6.41070426e-01 2.83239096e-01 -5.40738285e-01 -6.10858619e-01 -3.35487872e-02 -3.42357427e-01 2.32862830e-02 8.82308006e-01 -1.02748287e+00 -1.37196317e-01 -2.52344906e-02 4.22359765e-01 -2.79211879e-01 4.63791490e-01 -9.71247435e-01 -1.41270623e-01 2.52213597e-01 -4.22028780e-01 -8.93713653e-01 -1.17324598e-01 1.10577273e+00 -2.76214629e-02 -4.93013233e-01 9.43139255e-01 -1.90215930e-01 -8.77037287e-01 6.59064293e-01 -1.77500993e-02 5.69683254e-01 1.37453020e+00 -5.22321045e-01 -5.30807495e-01 -1.17514871e-01 -1.09315932e+00 1.71035737e-01 4.25987720e-01 6.93450809e-01 8.46962214e-01 -1.10681784e+00 -3.60220611e-01 6.98325217e-01 8.62341940e-01 2.86785483e-01 2.97568917e-01 9.93307158e-02 6.86669573e-02 5.59150241e-02 1.59327224e-01 -3.89875203e-01 -9.63984907e-01 1.12685204e+00 2.63046384e-01 -1.70218814e-02 -5.92907727e-01 7.97025263e-01 6.68674707e-01 -8.76921475e-01 3.05942804e-01 4.58550267e-02 -5.35472520e-02 9.69628524e-03 7.80868649e-01 2.75671005e-01 7.58910999e-02 -6.03509903e-01 -1.00809164e-01 3.13394159e-01 -1.17072456e-01 3.90550822e-01 1.28324175e+00 7.49606490e-02 3.39393675e-01 8.57164323e-01 1.50123882e+00 -5.33385694e-01 -1.06642222e+00 -5.38077593e-01 -3.88552360e-02 -7.01432765e-01 -1.15825102e-01 -6.69870079e-01 -6.65178478e-01 1.15897417e+00 6.62267745e-01 5.23376465e-02 7.92355537e-01 1.98451295e-01 7.71881223e-01 2.61650562e-01 6.56038105e-01 -8.89350295e-01 1.49234697e-01 4.48093474e-01 5.03500283e-01 -1.61349618e+00 -5.27301490e-01 -5.50599635e-01 -5.08551300e-01 1.02498460e+00 8.02205324e-01 -2.35111028e-01 5.52682638e-01 -1.40742540e-01 -1.57181665e-01 -2.09924832e-01 -7.54400074e-01 -3.93287897e-01 2.93213308e-01 7.08970666e-01 -3.15781087e-01 2.17133820e-01 4.31907594e-01 9.01471674e-01 6.31426051e-02 -1.08188398e-01 5.10154486e-01 1.02239740e+00 -3.78087491e-01 -6.84517026e-01 -4.72597718e-01 7.15031207e-01 -3.18221062e-01 1.39812948e-02 -3.22195560e-01 6.21614158e-01 1.98799714e-01 8.47649574e-01 2.36532032e-01 -4.67864424e-01 2.65017688e-01 3.71861875e-01 2.20305145e-01 -1.07390547e+00 -6.54197782e-02 -5.44648945e-01 -7.05914795e-02 -6.28661573e-01 1.54733226e-01 -1.66710898e-01 -8.84643018e-01 1.91422462e-01 -4.84629512e-01 7.04237521e-02 3.91381562e-01 1.11976135e+00 2.27458239e-01 5.82285583e-01 8.00001204e-01 -8.29279006e-01 -1.15192521e+00 -5.17866969e-01 -8.61978590e-01 9.43440199e-01 6.92838371e-01 -1.00518870e+00 -9.55136478e-01 -3.17307383e-01]

[9.759100914001465, 2.582951784133911]

d5f64d5e-c0f7-435d-8e9a-f5e6896c7d9e

large-batch-neural-multi-objective-bayesian

2306.01095

https://arxiv.org/abs/2306.01095v2

https://arxiv.org/pdf/2306.01095v2.pdf

Large-Batch, Neural Multi-Objective Bayesian Optimization

Bayesian optimization provides a powerful framework for global optimization of black-box, expensive-to-evaluate functions. However, it has a limited capacity in handling data-intensive problems, especially in multi-objective settings, due to the poor scalability of default Gaussian Process surrogates. We present a novel Bayesian optimization framework specifically tailored to address these limitations. Our method leverages a Bayesian neural networks approach for surrogate modeling. This enables efficient handling of large batches of data, modeling complex problems, and generating the uncertainty of the predictions. In addition, our method incorporates a scalable, uncertainty-aware acquisition strategy based on the well-known, easy-to-deploy NSGA-II. This fully parallelizable strategy promotes efficient exploration of uncharted regions. Our framework allows for effective optimization in data-intensive environments with a minimum number of iterations. We demonstrate the superiority of our method by comparing it with state-of-the-art multi-objective optimizations. We perform our evaluation on two real-world problems - airfoil design and color printing - showcasing the applicability and efficiency of our approach. Code is available at: https://github.com/an-on-ym-ous/lbn_mobo

['Vahid Babaei', 'Hans-Peter Seidel', 'Navid Ansari']

2023-06-01

['efficient-exploration', 'bayesian-optimization']

['methodology', 'methodology']

[-8.09761658e-02 -2.43800297e-01 -1.95747375e-01 -3.21627349e-01 -1.12254798e+00 -5.02057731e-01 2.10249230e-01 -2.36958321e-02 -3.34888548e-01 9.41099286e-01 1.42888166e-02 -5.07355750e-01 -5.79825461e-01 -8.82389903e-01 -7.28541672e-01 -8.69166136e-01 -6.26058653e-02 8.73578668e-01 -7.02040677e-04 -1.63492292e-01 3.05315495e-01 3.85565937e-01 -1.51623344e+00 -2.11930603e-01 1.03757370e+00 9.32160497e-01 3.44634682e-01 6.07663274e-01 3.00173491e-01 2.27364391e-01 -3.34243566e-01 -1.48732454e-01 3.13951045e-01 -8.59877393e-02 -3.94221753e-01 -3.37798119e-01 -9.49919298e-02 -3.34442377e-01 1.13852963e-01 8.69780719e-01 7.52845585e-01 5.63848436e-01 4.72958565e-01 -1.07293534e+00 -2.33718306e-01 3.77575874e-01 -3.55972648e-01 -4.28713262e-02 2.13141769e-01 6.21134400e-01 1.01998961e+00 -6.47109807e-01 2.66985357e-01 1.17437983e+00 5.92046261e-01 1.97601259e-01 -1.33193362e+00 -5.84699512e-01 1.52312443e-01 -2.11144444e-02 -1.39431417e+00 -4.22997087e-01 4.44880873e-01 -3.26664418e-01 9.18255091e-01 4.14053261e-01 6.52310073e-01 8.49978864e-01 2.67988592e-01 7.44109511e-01 1.10219800e+00 -3.01159799e-01 6.44488990e-01 -6.97778538e-02 -1.16618715e-01 7.07999229e-01 1.93088472e-01 6.44776881e-01 -5.75311780e-01 -3.14930022e-01 3.40791613e-01 -1.62672639e-01 -1.54129297e-01 -3.61010671e-01 -8.60233247e-01 9.56928611e-01 5.65734431e-02 -2.49093175e-01 -4.63016450e-01 4.54010755e-01 -3.87740508e-02 -1.78279117e-01 7.01028705e-01 6.61331296e-01 -5.80436170e-01 -7.38089621e-01 -1.14729881e+00 5.18243670e-01 9.52757001e-01 6.21588528e-01 5.34019172e-01 6.30156919e-02 -2.80036330e-01 6.20218277e-01 5.82444906e-01 5.63513637e-01 -2.65743155e-02 -1.08540869e+00 3.44951600e-01 2.92688280e-01 4.02000487e-01 -7.58703589e-01 -6.10757113e-01 -7.32399762e-01 -4.95824426e-01 5.02868533e-01 4.49307591e-01 -4.23114747e-01 -7.39157796e-01 1.55863500e+00 6.45552099e-01 3.08632236e-02 -1.69621587e-01 9.38934982e-01 2.34740451e-01 6.60648406e-01 -1.70055658e-01 -1.21632367e-01 1.14798319e+00 -9.82428849e-01 -4.70259696e-01 -1.43351689e-01 1.70100465e-01 -5.96515000e-01 1.03383577e+00 6.48551285e-01 -1.15364194e+00 -1.32032782e-01 -8.82112801e-01 3.68404180e-01 -2.94198900e-01 1.35882139e-01 8.62584949e-01 1.01450789e+00 -8.42940092e-01 9.04084742e-01 -1.23359561e+00 1.41769638e-02 5.54556429e-01 5.77067137e-01 1.23388641e-01 1.00460842e-01 -7.13424563e-01 8.54656160e-01 5.45657396e-01 2.51615077e-01 -1.11514604e+00 -1.03597283e+00 -6.95371747e-01 1.90409064e-01 9.67337132e-01 -7.65347183e-01 1.19778121e+00 -4.58042622e-01 -2.07513905e+00 7.12025464e-02 -1.96046352e-01 -2.59902179e-01 6.42178953e-01 -4.04160947e-01 -1.06866613e-01 -1.19673170e-01 -3.73166323e-01 3.21099818e-01 6.56473994e-01 -9.66275632e-01 -3.87103498e-01 -2.25858241e-01 2.21707113e-02 1.90138578e-01 -1.28338635e-01 1.24074437e-01 -4.72958624e-01 -5.18713772e-01 -2.67362803e-01 -1.03532565e+00 -6.12239480e-01 -2.03361332e-01 -4.90945369e-01 2.36800507e-01 2.98102230e-01 -6.95732415e-01 1.43399775e+00 -1.80643606e+00 1.83816016e-01 4.00331706e-01 -1.04420856e-01 7.79702440e-02 -4.83506359e-02 5.93254745e-01 3.06991160e-01 1.93742052e-01 -4.85403210e-01 -4.94928211e-01 3.95878851e-01 2.35160396e-01 -5.76470643e-02 4.88970876e-01 1.18724942e-01 8.02415550e-01 -9.02799785e-01 -3.31779599e-01 3.17779750e-01 4.08276975e-01 -8.42397571e-01 3.38123471e-01 -4.96062219e-01 5.81294537e-01 -4.35189396e-01 9.51438546e-01 5.38270533e-01 -3.11501712e-01 1.76987156e-01 5.24660125e-02 -2.49505922e-01 4.34400775e-02 -1.53450465e+00 1.75988770e+00 -5.74764788e-01 1.79998145e-01 2.37021640e-01 -7.93372810e-01 6.28472030e-01 -6.37785420e-02 3.67591470e-01 -2.16375127e-01 2.50365704e-01 2.50561714e-01 -6.40079603e-02 -1.95734620e-01 4.29793060e-01 1.58949077e-01 5.88636510e-02 5.90259016e-01 -3.55795994e-02 -5.85002124e-01 5.91051161e-01 -3.01454723e-01 9.71881986e-01 6.74267590e-01 3.47577125e-01 -3.85792822e-01 8.23312700e-02 -3.84728089e-02 7.34968007e-01 9.27040339e-01 2.92232353e-02 5.64315200e-01 5.26036143e-01 -3.21502924e-01 -7.50831425e-01 -9.32101429e-01 -1.50047481e-01 1.16844225e+00 -7.64231756e-02 -4.29372996e-01 -4.70999330e-01 -3.83143097e-01 2.63715059e-01 1.23090196e+00 -5.59010625e-01 2.59314049e-02 -1.68075740e-01 -1.42552269e+00 2.02642739e-01 3.78185123e-01 3.03414110e-02 -6.18093073e-01 -8.58654141e-01 3.50324273e-01 2.11705536e-01 -6.85824335e-01 -5.50083779e-02 2.56901085e-01 -8.58722031e-01 -9.69134688e-01 -4.50363308e-01 9.70488712e-02 4.64193046e-01 -3.96652400e-01 1.27036786e+00 -2.61536121e-01 -4.30511415e-01 2.91373998e-01 4.73003536e-02 -5.52946150e-01 -4.16484267e-01 3.99817191e-02 5.19934064e-03 -2.66648084e-01 1.75268818e-02 -5.74361265e-01 -5.50145924e-01 5.48657358e-01 -7.09204853e-01 5.19832596e-02 4.09778059e-01 1.06266391e+00 7.89804757e-01 1.08067997e-01 1.54433802e-01 -4.80107725e-01 8.51172864e-01 -5.44606864e-01 -1.56315207e+00 4.76575792e-01 -9.67015803e-01 3.49248767e-01 3.74961674e-01 -2.93179810e-01 -1.03064144e+00 -6.59246668e-02 -1.56375125e-01 -2.95891106e-01 -4.23498899e-02 7.41807222e-01 1.21926285e-01 -2.42079049e-01 2.72192866e-01 -1.40224725e-01 9.09519121e-02 -6.19316399e-01 2.29049802e-01 4.69629079e-01 2.15371236e-01 -1.09993851e+00 7.32572615e-01 3.44009459e-01 3.61931324e-01 -4.68734622e-01 -5.97151279e-01 -1.70544192e-01 -1.72257438e-01 -3.20453554e-01 5.13036668e-01 -6.55676663e-01 -1.15677190e+00 4.52286094e-01 -7.74168015e-01 -6.84327543e-01 -3.68934989e-01 6.04547799e-01 -5.86903393e-01 1.88511714e-01 -2.91933239e-01 -1.20076215e+00 -3.19549054e-01 -1.49602711e+00 1.01784325e+00 3.03180963e-01 -6.65736496e-02 -1.05796337e+00 2.92186469e-01 4.43197161e-01 7.59241581e-01 4.03005838e-01 6.63788676e-01 -5.93946695e-01 -6.98902726e-01 -3.01314145e-02 1.75971925e-01 2.66213864e-01 -3.25554937e-01 3.79331172e-01 -7.75398433e-01 -3.52750480e-01 -2.58369148e-01 -4.26480770e-01 6.25438988e-01 6.27814949e-01 1.39191234e+00 -1.89601019e-01 -1.54881164e-01 8.97076488e-01 1.42519236e+00 3.20495898e-03 1.89419895e-01 4.38516110e-01 1.58726677e-01 4.55233395e-01 8.73497844e-01 1.08559370e+00 3.65056008e-01 8.75186503e-01 7.56354690e-01 -6.09607622e-02 4.13848191e-01 -2.74019912e-02 2.34157965e-01 5.51653445e-01 -2.33297601e-01 -4.89984155e-01 -1.27166212e+00 2.78580993e-01 -2.01601863e+00 -5.08235693e-01 1.65893331e-01 2.38071394e+00 9.75795686e-01 5.46637885e-02 2.48226095e-02 -3.67700398e-01 4.63651866e-01 -1.01878578e-02 -6.61795914e-01 -3.45085323e-01 1.36166170e-01 3.81160885e-01 6.49040639e-01 6.60306394e-01 -1.00554860e+00 6.30747616e-01 6.37461042e+00 1.16622627e+00 -7.45139897e-01 2.17967167e-01 7.54104972e-01 -6.92372262e-01 -3.05158705e-01 1.41981453e-01 -8.61103892e-01 6.88895404e-01 1.20460892e+00 -2.05436796e-01 9.96579349e-01 8.22764099e-01 5.18620729e-01 -4.98982668e-01 -8.82092774e-01 6.08281016e-01 -1.96224540e-01 -1.44090950e+00 -6.16278470e-01 1.44696802e-01 7.85510540e-01 2.65487850e-01 5.69689758e-02 1.62776679e-01 7.24601269e-01 -1.24864125e+00 7.91270494e-01 6.09327912e-01 5.37297487e-01 -9.86909032e-01 7.69897997e-01 2.54501879e-01 -7.24189341e-01 -2.82497793e-01 -2.28070155e-01 -6.40555657e-03 4.59843755e-01 1.02721512e+00 -6.96042836e-01 7.51708269e-01 7.94190407e-01 2.89148271e-01 -3.06841642e-01 1.22905767e+00 -3.60040039e-01 7.03135729e-01 -9.08159554e-01 -2.03211635e-01 2.06108242e-01 -6.04342222e-01 7.44020820e-01 9.54550624e-01 5.01966894e-01 -1.85156047e-01 2.51855820e-01 1.20805097e+00 3.87810767e-01 -9.19443816e-02 -1.66371018e-01 -1.24333113e-01 4.69666123e-01 1.18497574e+00 -4.58308429e-01 1.03172794e-01 -1.88100711e-02 1.74661562e-01 1.81691065e-01 4.30637747e-01 -9.77161109e-01 -3.30177844e-01 6.01857424e-01 -3.20887446e-01 4.37774777e-01 -4.19157386e-01 -3.77433628e-01 -9.27047610e-01 -1.06216716e-02 -1.06172693e+00 3.45877051e-01 -6.37672126e-01 -1.02366805e+00 2.40147576e-01 3.17072392e-01 -8.44264627e-01 -4.62164909e-01 -7.92439699e-01 -4.96093452e-01 9.46376741e-01 -1.52959907e+00 -9.75462556e-01 -1.86135679e-01 1.86770156e-01 1.89188018e-01 -2.14167610e-01 7.63263941e-01 1.04651615e-01 -9.30363119e-01 3.51381689e-01 6.95457280e-01 -5.88451087e-01 4.51083243e-01 -1.32161665e+00 2.20435426e-01 9.84078646e-01 -2.83482909e-01 5.47532499e-01 9.92457807e-01 -5.91684222e-01 -1.81934643e+00 -6.48599327e-01 1.11204721e-01 -3.17827642e-01 8.29650760e-01 -4.26217526e-01 -5.39584756e-01 2.41279572e-01 -1.66444539e-03 -2.57460803e-01 8.36136580e-01 4.43730444e-01 2.20352143e-01 -2.35880658e-01 -1.06317329e+00 5.72921038e-01 5.59886456e-01 -3.34480666e-02 -2.54809469e-01 5.37571907e-01 4.04637158e-01 -7.39354849e-01 -1.08187747e+00 5.57214200e-01 6.85088992e-01 -9.92768168e-01 9.44517493e-01 -4.02395815e-01 3.75531614e-01 -3.84180725e-01 -1.33731127e-01 -1.42108846e+00 1.78942569e-02 -1.08894122e+00 -4.03200746e-01 1.18497157e+00 5.55166245e-01 -9.94895101e-01 6.41407728e-01 8.87434483e-01 -3.97514291e-02 -1.13446069e+00 -1.14749134e+00 -8.88688505e-01 -1.31294101e-01 -6.48650050e-01 9.09907222e-01 3.55180711e-01 -4.27905828e-01 -3.95380735e-01 -4.28342521e-01 3.35178941e-01 7.60438204e-01 2.93594360e-01 8.31471205e-01 -9.51353371e-01 -9.97210681e-01 -5.15134990e-01 7.78068304e-02 -6.55083120e-01 4.65585329e-02 -4.27439809e-01 1.81922421e-01 -1.16025984e+00 4.28478345e-02 -5.73222995e-01 -2.74902940e-01 5.45717299e-01 -2.99262404e-01 -9.61833075e-02 2.03058943e-01 -1.73023403e-01 -5.99879742e-01 8.65496695e-01 9.14949119e-01 5.44218756e-02 -1.96967736e-01 2.66446471e-01 -5.11027277e-01 6.63143396e-01 9.81404841e-01 -7.64422596e-01 -2.23192886e-01 -2.95021981e-01 4.17582870e-01 -4.52371649e-02 1.92498907e-01 -9.63523090e-01 1.23064891e-01 -5.69986999e-01 9.85459462e-02 -5.68431914e-01 5.12483060e-01 -7.17902362e-01 5.07952869e-01 5.03201306e-01 -4.65455316e-02 1.82488367e-01 5.19672632e-01 4.72362638e-01 -9.86712798e-02 -5.01399934e-01 6.03440166e-01 -9.23112705e-02 -4.12742376e-01 1.11674763e-01 -2.63385266e-01 1.48403049e-01 1.02441275e+00 5.85749932e-02 -2.90122837e-01 -4.05722111e-01 -2.81458914e-01 6.82450175e-01 5.78786731e-01 1.04317583e-01 3.19445461e-01 -7.66605556e-01 -7.25543797e-01 9.14396718e-02 -2.04907373e-01 7.58955702e-02 2.04067782e-01 9.17457819e-01 -7.08123863e-01 2.78362095e-01 4.98823039e-02 -6.64798856e-01 -1.01792586e+00 3.10985804e-01 3.57417107e-01 -6.72600448e-01 -7.85142705e-02 8.29467118e-01 -1.81663483e-01 -6.16484165e-01 2.09733546e-01 -3.22571307e-01 3.34407479e-01 -1.68844745e-01 3.09823185e-01 8.76349092e-01 6.46067634e-02 1.68928310e-01 -4.16393608e-01 3.18760514e-01 3.58107120e-01 -3.44579607e-01 1.68656373e+00 6.66867048e-02 -1.92398340e-01 2.88036883e-01 6.72336519e-01 3.08477469e-02 -1.63853097e+00 3.73811096e-01 -1.33528262e-01 -6.39934719e-01 5.27380407e-01 -1.17067599e+00 -8.99532735e-01 6.78077936e-01 5.40031075e-01 -2.80125111e-01 1.17437160e+00 -5.39258420e-01 3.18189025e-01 5.43883443e-01 3.82381946e-01 -1.29948580e+00 -3.22650015e-01 3.74586254e-01 8.46634030e-01 -1.20872533e+00 4.71962959e-01 -1.38427213e-01 -5.01693845e-01 1.15409398e+00 3.36494029e-01 2.46008009e-01 6.42091811e-01 5.13255358e-01 -1.22177251e-01 -1.22684725e-01 -9.24332380e-01 3.67506184e-02 3.55850041e-01 3.29663396e-01 -8.70111510e-02 9.63993445e-02 -1.79035757e-02 5.12806058e-01 7.30255712e-03 8.73930976e-02 2.95141160e-01 1.08470130e+00 -5.17627560e-02 -1.19376600e+00 -5.77119410e-01 3.70099783e-01 -5.28544307e-01 -3.59751463e-01 2.58929521e-01 6.69663429e-01 -3.25531244e-01 1.00483155e+00 -1.59219846e-01 -1.94138521e-03 3.16773355e-02 -1.29171237e-01 3.06223303e-01 -3.39225948e-01 -7.16681778e-01 1.30256861e-01 4.01697457e-01 -1.00859284e+00 -1.72405824e-01 -7.68035531e-01 -5.59694707e-01 -3.21030408e-01 -3.69507313e-01 1.69500485e-01 1.13162243e+00 8.30105841e-01 6.29808545e-01 5.92127085e-01 4.26969677e-01 -1.14315581e+00 -1.04419601e+00 -6.45942509e-01 -4.21750128e-01 -1.99449450e-01 1.11178577e-01 -9.37610924e-01 -4.17416871e-01 -5.79531491e-01]

[6.322909832000732, 3.8473360538482666]

89d43c9f-7f8d-4f07-a8ed-ec4f7d8da13a

shared-tasks-of-the-2015-workshop-on-noisy

https://aclanthology.org/W15-4319

https://aclanthology.org/W15-4319.pdf

Shared Tasks of the 2015 Workshop on Noisy User-generated Text: Twitter Lexical Normalization and Named Entity Recognition

['Young-Bum Kim', 'Wei Xu', 'Timothy Baldwin', 'Bo Han', 'Alan Ritter', 'Marie Catherine de Marneffe']

2015-07-01

ws-2015-7

['lexical-normalization']

['natural-language-processing']

[-8.63703638e-02 1.71006292e-01 -6.22772932e-01 -4.08054382e-01 -8.41685571e-03 -9.08429027e-01 6.55310392e-01 -6.53472245e-01 -2.85945535e-01 1.06888819e+00 -4.63127941e-02 -1.01159286e+00 -3.91567826e-01 -9.63214397e-01 -4.95059669e-01 -6.31337762e-01 -9.79754329e-01 7.25764990e-01 3.30370307e-01 -6.93831444e-01 7.03166842e-01 7.88774848e-01 -1.68942046e+00 7.18545914e-01 7.04417467e-01 8.52217197e-01 2.49141872e-01 1.14950800e+00 -1.95044339e-01 1.55633950e+00 -7.48382092e-01 -5.46825826e-01 3.13719302e-01 -1.23176083e-01 -7.22945035e-01 -1.01074085e-01 9.28529128e-02 -8.59008506e-02 -2.09758401e-01 9.22211111e-01 5.37373662e-01 4.49454933e-02 1.08379531e+00 -1.42548037e+00 -5.91619551e-01 6.10313773e-01 -4.01565880e-02 1.21627934e-01 1.03678203e+00 -5.39447069e-01 1.19919395e+00 -1.13026452e+00 7.20913768e-01 1.26888943e+00 8.66221786e-01 5.44149756e-01 -1.22286928e+00 -1.94712028e-01 -3.26822817e-01 -9.51717794e-02 -1.46558487e+00 -3.25250506e-01 4.25783843e-02 -2.08119690e-01 1.66093647e+00 1.26596653e+00 1.20609856e+00 1.01401424e+00 1.26658809e+00 8.34431887e-01 1.04267764e+00 -5.13792276e-01 3.35295945e-01 3.66983831e-01 1.54683650e-01 6.33519173e-01 8.40953708e-01 5.26628852e-01 -7.06372619e-01 -9.13127720e-01 9.33553874e-01 -2.94925272e-01 1.71355158e-01 -5.05680561e-01 -9.05919552e-01 6.91228509e-01 1.78732842e-01 3.83959889e-01 -1.39880210e-01 9.89067405e-02 1.26390755e-01 5.30987144e-01 -2.58292928e-02 6.47037446e-01 -9.11868811e-01 -1.33165747e-01 -8.71728659e-01 5.10332465e-01 1.25398111e+00 1.52653182e+00 1.24482810e-01 2.94908643e-01 -9.34252143e-02 3.17179203e-01 8.92314315e-01 1.01808000e+00 4.28362608e-01 -1.36146402e+00 -6.87414408e-02 1.72361732e-01 5.01781464e-01 -8.52631688e-01 -6.33224547e-01 -9.64177120e-03 -8.93263519e-01 4.49267089e-01 3.49161088e-01 4.57367361e-01 -8.02827001e-01 5.07305264e-01 4.33481112e-02 -2.34125629e-01 4.53833073e-01 5.55570945e-02 4.99930978e-01 3.76208365e-01 -1.34477139e-01 -5.73289394e-01 1.06082785e+00 -1.36716676e+00 -1.35299087e+00 2.33215362e-01 9.05734658e-01 -1.07320261e+00 4.35900748e-01 5.33875942e-01 -1.55548143e+00 -1.37560293e-01 -1.08699942e+00 1.78573877e-01 -7.27255583e-01 -3.14239264e-01 8.57801437e-01 1.43120694e+00 -1.60129595e+00 9.73287821e-01 -4.91727620e-01 6.59165755e-02 1.20568443e-02 8.24621081e-01 -2.64718989e-03 4.62812334e-01 -1.33193445e+00 1.08501506e+00 2.22979754e-01 -1.21242590e-01 -1.65216476e-01 -2.13068098e-01 -8.23704481e-01 -5.63443303e-01 -4.78693932e-01 -5.29636025e-01 1.44139910e+00 -2.59346128e-01 -1.65295815e+00 9.71794367e-01 -1.42069459e-01 -1.97814897e-01 6.14786744e-01 -1.28011424e-02 -8.31891418e-01 2.42498964e-01 -1.89849049e-01 5.76383233e-01 9.28263724e-01 -1.35132408e+00 -7.59897232e-01 -1.67359829e-01 -1.23336017e-01 2.66287565e-01 -1.25510961e-01 1.89734384e-01 2.11616129e-01 -1.12999000e-01 3.27147305e-01 -7.20919967e-01 -2.53068686e-01 -5.32041907e-01 -1.46512717e-01 -7.10518599e-01 7.70373225e-01 -4.51523662e-01 1.83705616e+00 -1.67618537e+00 -2.06720144e-01 3.98590982e-01 3.57815564e-01 -1.24705513e-03 2.40583986e-01 1.08380008e+00 -2.76906848e-01 7.32199550e-01 4.11965609e-01 -1.10722095e-01 2.24991128e-01 4.85861301e-01 -4.16602850e-01 3.05609167e-01 -1.29282743e-01 1.15307164e+00 -1.16605783e+00 -5.23096442e-01 4.11106765e-01 9.42391157e-02 -4.58732933e-01 4.79237735e-01 2.99364805e-01 1.47170946e-01 -3.56553018e-01 1.39399457e+00 1.15709066e+00 -1.31984919e-01 1.45911396e-01 5.30878425e-01 -3.79135728e-01 3.55090618e-01 -6.96863770e-01 1.05554795e+00 7.08333924e-02 5.00173986e-01 1.02364108e-01 -7.94621468e-01 3.33247900e-01 8.61540735e-01 4.37155962e-01 -9.67555881e-01 -2.26398129e-02 6.92409754e-01 1.12803578e-01 -6.07703328e-01 7.58228302e-01 3.94563079e-02 -3.64872098e-01 6.40070081e-01 -2.37588286e-01 -6.59476995e-01 9.64643434e-02 3.08100313e-01 6.22585893e-01 -5.18246442e-02 5.62923312e-01 -1.05613089e+00 7.86340594e-01 -1.82965681e-01 -1.81299388e-01 1.03415680e+00 -3.09923887e-01 3.19085121e-01 2.99841821e-01 -6.65102363e-01 -6.45341039e-01 -1.12307119e+00 -4.89381433e-01 1.30636716e+00 3.24267983e-01 -4.39044595e-01 -9.54439282e-01 -2.49762803e-01 1.77620783e-01 6.89606130e-01 -5.90509653e-01 3.84124845e-01 -5.03739953e-01 -8.32535863e-01 7.39044368e-01 3.45434904e-01 -5.07752821e-02 -1.33414865e+00 -6.58416986e-01 1.25490099e-01 -2.20292807e-01 -6.63697243e-01 -6.23428151e-02 4.48765576e-01 -1.35989368e+00 -5.18594682e-01 -6.66252747e-02 -8.20914626e-01 5.87345481e-01 2.46782884e-01 1.27047324e+00 5.39230824e-01 -2.31483161e-01 4.26904231e-01 -1.21292919e-01 -4.95818377e-01 -4.59671497e-01 -8.00336525e-02 5.28869390e-01 -5.87835789e-01 5.19427478e-01 -2.50617653e-01 -7.29350567e-01 5.37953973e-01 -6.88540697e-01 1.62748516e-01 1.79803044e-01 1.04410267e+00 1.35816500e-01 -9.34035778e-02 1.22507080e-01 -6.38007045e-01 8.72274399e-01 -1.69219792e-01 -3.78732830e-01 5.77745810e-02 -6.77108407e-01 -3.74140263e-01 3.21430594e-01 -3.25342178e-01 -1.01981449e+00 -4.87835288e-01 -9.82677937e-02 2.45538145e-01 1.11353043e-02 -1.46784872e-01 6.47139177e-02 -5.24923325e-01 8.02199244e-01 9.25758183e-02 1.99174434e-02 -6.80815242e-03 3.01039815e-01 7.09525108e-01 -6.82967342e-03 -6.68678164e-01 8.44880998e-01 4.91470337e-01 7.98524171e-02 -9.57177758e-01 -1.52186140e-01 -2.60129690e-01 -9.51962709e-01 -6.54426932e-01 6.56643391e-01 -6.78531289e-01 -9.10833478e-01 3.91110867e-01 -9.38691139e-01 -3.38627815e-01 -3.91645581e-01 4.25431967e-01 -1.01278400e+00 2.75717527e-02 -3.90154392e-01 -1.27895141e+00 -5.10977268e-01 -1.02017939e+00 9.43384409e-01 5.30070923e-02 -5.10597289e-01 -1.26927447e+00 5.87685481e-02 2.71537274e-01 1.81734428e-01 -1.73075795e-01 6.90226793e-01 -2.38256708e-01 -4.24233019e-01 -1.53791070e-01 2.34436691e-02 -1.39755070e-01 1.70832314e-02 4.95917559e-01 -9.81751978e-01 -5.31145096e-01 6.65065646e-02 -1.92070693e-01 -1.08835101e-01 6.52520418e-01 5.91872573e-01 -2.29931593e-01 -8.56000841e-01 5.40386558e-01 1.38545322e+00 3.85070026e-01 5.32770038e-01 7.28214979e-01 1.41836226e-01 5.53460240e-01 9.17806149e-01 4.63203549e-01 1.30579369e-02 3.28798652e-01 2.40537539e-01 1.49327129e-01 1.11720070e-01 -1.54819340e-01 3.77893507e-01 1.16112018e+00 -8.18235934e-01 -2.69281328e-01 -5.07867396e-01 4.42987174e-01 -1.72482407e+00 -1.40330648e+00 -4.32368398e-01 6.90478683e-01 6.25676990e-01 1.56016424e-01 -1.48347050e-01 3.35214496e-01 4.99015123e-01 -2.03574806e-01 -1.19133167e-01 -1.06291151e+00 -1.43546045e-01 3.15233678e-01 7.37729073e-01 1.00061214e+00 -7.20721722e-01 1.03317809e+00 1.29781246e+01 1.02230716e+00 2.21112028e-01 1.03134915e-01 5.16071796e-01 3.48020852e-01 -4.36954498e-01 -4.56139445e-02 -1.04416132e+00 2.72933897e-02 1.38140702e+00 -4.30666685e-01 6.85999811e-01 5.44219851e-01 3.44648361e-01 -4.23268199e-01 -1.26188684e+00 5.26221812e-01 9.73738134e-02 -1.40886843e+00 -2.83300440e-04 6.85225725e-01 7.73699820e-01 -5.08050561e-01 6.22419357e-01 3.24184299e-01 6.09259963e-01 -1.14389277e+00 8.60300779e-01 2.53660440e-01 1.03040910e+00 -6.05088234e-01 5.67372203e-01 1.68872893e-01 -1.14389896e+00 -2.20873043e-01 -8.77727985e-01 -1.00755692e+00 3.93533185e-02 -1.81779593e-01 -4.29956943e-01 3.48861217e-01 9.58353162e-01 2.99398601e-01 -3.93658698e-01 9.95779395e-01 -4.78694476e-02 1.04875881e-02 -2.71853864e-01 -4.48467314e-01 4.83122796e-01 -3.54241252e-01 4.66730654e-01 1.00164843e+00 2.48499006e-01 3.51035744e-01 -9.84472036e-02 4.01770771e-01 5.45058846e-01 3.29446048e-02 -1.19659424e+00 -1.78908288e-01 2.83276141e-01 9.16795909e-01 -4.83487815e-01 -4.22520459e-01 -2.00212970e-01 8.62069130e-01 -3.55488248e-02 5.01107454e-01 -6.11489356e-01 -4.35615242e-01 9.72222984e-01 -1.27327025e-01 -1.14700586e-01 -3.48497719e-01 -6.23769283e-01 -7.30352640e-01 -5.89872956e-01 -4.54965204e-01 5.93606755e-02 -5.53365827e-01 -1.39813089e+00 5.79277515e-01 -2.27688253e-02 -1.40553558e+00 -6.99901402e-01 -1.27676582e+00 -4.76714373e-01 4.92853165e-01 -1.11898029e+00 -1.10984349e+00 2.50124663e-01 4.52870727e-01 1.64141744e-01 -5.34416080e-01 1.39563632e+00 3.57715860e-02 1.00637585e-01 9.24474537e-01 6.69434488e-01 -7.37814724e-01 5.56605101e-01 -1.27867436e+00 5.68737745e-01 -1.37897313e-01 -4.31265175e-01 9.05828118e-01 6.28349900e-01 -5.39804697e-01 -1.41196322e+00 -3.66917729e-01 1.08350635e+00 -9.83769417e-01 6.55218959e-01 -3.86345625e-01 4.23767231e-02 7.88592756e-01 7.15902448e-01 -6.18741751e-01 8.21781039e-01 -1.83753878e-01 1.80774391e-01 5.75296998e-01 -1.39248300e+00 6.12354755e-01 1.66275799e+00 -4.63594139e-01 -6.25784039e-01 7.60327101e-01 8.13696027e-01 -6.94087505e-01 -1.30082703e+00 3.34633321e-01 8.65424156e-01 -8.75409484e-01 1.61978090e+00 -1.32660246e+00 -4.43697497e-02 2.81152606e-01 -2.61993498e-01 -9.32519078e-01 -5.96193194e-01 -1.23518765e+00 -5.33532679e-01 -5.83747849e-02 5.96577883e-01 -1.13057327e+00 3.42365682e-01 8.74560475e-01 -2.82833427e-01 -6.42737269e-01 -1.06996536e+00 -1.32016802e+00 -3.35779637e-02 -1.45572275e-01 4.90409225e-01 7.63798356e-01 6.83744550e-01 1.09839931e-01 -6.36873543e-02 -1.00294888e-01 5.33176839e-01 9.55312885e-03 4.41501856e-01 -1.34294486e+00 3.86843324e-01 -5.75816095e-01 -3.07655483e-01 -9.45992947e-01 -8.85957032e-02 -8.12076271e-01 -6.53862000e-01 -1.28511906e+00 -8.32044985e-03 -1.91056758e-01 -1.11109078e-01 -1.65725678e-01 3.67937148e-01 2.13746816e-01 1.20859891e-02 1.03788137e-01 -3.71160030e-01 6.18435517e-02 1.29639816e+00 6.91750320e-05 -1.62315920e-01 4.85058486e-01 -4.81304944e-01 7.84440815e-01 8.58408585e-02 -2.99253196e-01 -6.78878546e-01 6.11881316e-02 6.69384480e-01 4.61409837e-02 3.24159935e-02 -7.42885649e-01 5.37211418e-01 -3.75702560e-01 4.78586555e-01 -1.32223868e+00 1.30741090e-01 -9.61415648e-01 6.95283338e-02 9.40189242e-01 2.74610907e-01 1.20424610e-02 8.66204947e-02 5.39500564e-02 -1.44871444e-01 -5.70943117e-01 9.21121240e-01 -4.07591403e-01 -4.92852688e-01 -5.20386267e-03 -1.03226590e+00 8.97834301e-02 9.92593169e-01 -7.84614205e-01 -3.59281451e-01 -4.20183957e-01 -8.29068601e-01 -1.95836127e-02 6.50830388e-01 3.03609259e-02 7.30431557e-01 -1.51530886e+00 -2.30721906e-01 7.18729138e-01 -3.22939813e-01 -3.74400020e-01 -1.70157343e-01 6.58265352e-01 -1.32361674e+00 1.02442718e+00 -5.41665435e-01 -4.55340147e-01 -1.14228773e+00 4.78126436e-01 4.28307921e-01 -2.41845414e-01 -2.15481281e-01 1.11954463e+00 2.71224789e-02 -8.23025763e-01 1.85185194e-01 -9.21545625e-02 -7.54407048e-01 3.55081353e-03 6.88606799e-01 1.05194807e+00 -2.91290224e-01 -6.04341030e-01 -4.56784427e-01 6.65885091e-01 2.32151806e-01 -2.87484169e-01 9.17833567e-01 -2.19243199e-01 -9.89108324e-01 4.28274393e-01 8.38715494e-01 -1.36269778e-01 -3.69319022e-02 4.16855574e-01 1.25943512e-01 -8.02164078e-01 -4.32406247e-01 -3.55811834e-01 -1.85641110e-01 5.54822803e-01 5.34874737e-01 8.99602413e-01 8.57008278e-01 -3.02566767e-01 8.18335712e-01 9.66778398e-01 5.72402716e-01 -1.68019545e+00 -2.13140488e-01 6.89524531e-01 9.12339568e-01 -9.28350806e-01 5.44190466e-01 -7.27165341e-01 -4.14997995e-01 1.32979155e+00 4.68304873e-01 -1.55325383e-01 1.27306652e+00 5.74917436e-01 1.14069022e-02 -3.70670199e-01 -9.44949508e-01 1.12705544e-01 3.75366658e-01 1.11147714e+00 5.06513238e-01 5.07374525e-01 -9.66500878e-01 3.21953118e-01 -7.28706717e-01 -2.34555230e-01 4.90474731e-01 1.41972518e+00 -6.43810987e-01 -1.20391917e+00 -7.23931909e-01 4.61561680e-01 -5.49773455e-01 -1.16372630e-01 -5.28106689e-01 8.46754074e-01 -5.76629937e-02 1.49448860e+00 -1.97535474e-03 -5.03491640e-01 4.11356747e-01 1.41089618e-01 7.21762300e-01 -1.23501487e-01 -9.04846430e-01 4.15413082e-01 3.84890139e-01 -1.23056793e+00 -8.58632207e-01 -1.05834293e+00 -1.40667629e+00 -1.19437599e+00 -5.12782812e-01 1.89310342e-01 3.83317530e-01 3.90289724e-01 -2.06836104e-01 2.85260603e-02 9.80917513e-01 -1.07949340e+00 -3.90341938e-01 -9.39418614e-01 -1.00026262e+00 -7.84516707e-02 2.89751232e-01 -8.00943017e-01 -7.83523321e-01 2.83909619e-01]

[-7.325130462646484, 3.717496156692505]

992420b3-b99e-4cab-8153-720f6af19db8

driver-hand-localization-and-grasp-analysis-a

1802.07854

http://arxiv.org/abs/1802.07854v1

http://arxiv.org/pdf/1802.07854v1.pdf

Driver Hand Localization and Grasp Analysis: A Vision-based Real-time Approach

Extracting hand regions and their grasp information from images robustly in real-time is critical for occupants' safety and in-vehicular infotainment applications. It must however, be noted that naturalistic driving scenes suffer from rapidly changing illumination and occlusion. This is aggravated by the fact that hands are highly deformable objects, and change in appearance frequently. This work addresses the task of accurately localizing driver hands and classifying the grasp state of each hand. We use a fast ConvNet to first detect likely hand regions. Next, a pixel-based skin classifier that takes into account the global illumination changes is used to refine the hand detections and remove false positives. This step generates a pixel-level mask for each hand. Finally, we study each such masked regions and detect if the driver is grasping the wheel, or in some cases a mobile phone. Through evaluation we demonstrate that our method can outperform state-of-the-art pixel based hand detectors, while running faster (at 35 fps) than other deep ConvNet based frameworks even for grasp analysis. Hand mask cues are shown to be crucial when analyzing a set of driver hand gestures (wheel/mobile phone grasp and no-grasp) in naturalistic driving settings. The proposed detection and localization pipeline hence can act as a general framework for real-time hand detection and gesture classification.

['Eshed Ohn-Bar', 'Akshay Rangesh', 'Siddharth', 'Mohan M. Trivedi']

2018-02-22

['hand-detection']

['computer-vision']

[ 2.42598012e-01 -1.96121722e-01 -1.06637768e-01 -3.08435291e-01 -3.84622723e-01 -7.38449991e-01 2.51834124e-01 -4.17655617e-01 -7.34571517e-01 2.89475501e-01 -3.53135288e-01 -3.16057265e-01 2.04811886e-01 -4.50713545e-01 -6.54465914e-01 -1.00479698e+00 2.28355750e-01 2.24486828e-01 7.54377186e-01 -1.41358569e-01 2.89802432e-01 1.24276972e+00 -1.97025859e+00 1.73223719e-01 2.53744394e-01 8.43023956e-01 3.84726971e-01 9.28806961e-01 8.08608085e-02 5.15262663e-01 -6.67860687e-01 -2.61122346e-01 4.79408950e-01 1.03009559e-01 -3.98753256e-01 1.36789963e-01 9.21134412e-01 -8.62160206e-01 -4.55258012e-01 7.56955028e-01 8.19705307e-01 1.09917492e-01 3.28982949e-01 -1.51857555e+00 2.92532265e-01 -2.33725291e-02 -7.33959794e-01 3.31324786e-01 2.62889177e-01 8.39944839e-01 2.95844525e-01 -9.38198626e-01 6.71523511e-01 1.27232635e+00 3.70261759e-01 7.97246933e-01 -6.49195373e-01 -1.10558701e+00 4.69653130e-01 5.30497313e-01 -1.39557981e+00 -6.07921541e-01 6.83702826e-01 -4.22208607e-01 1.03973544e+00 3.72763097e-01 5.39625585e-01 1.26500463e+00 3.24699849e-01 1.06161261e+00 1.30513525e+00 -3.79924923e-01 9.74016860e-02 1.27920091e-01 4.18215632e-01 1.08718741e+00 1.46267384e-01 2.11439773e-01 -6.02269471e-01 2.20225483e-01 6.16107643e-01 4.19122018e-02 -1.32100686e-01 -1.08054303e-01 -1.12226737e+00 8.35730880e-02 3.11446190e-01 1.57905892e-01 -7.79445827e-01 3.60892475e-01 9.82818604e-02 -4.70796451e-02 1.12344526e-01 -3.75470281e-01 -3.05341929e-01 -2.30278760e-01 -1.04622781e+00 3.82932216e-01 6.18955791e-01 1.08722997e+00 5.12480974e-01 -2.04748556e-01 -5.17154992e-01 2.72973746e-01 1.87708780e-01 6.93369746e-01 -2.92453945e-01 -6.73479199e-01 4.57380533e-01 4.52295154e-01 3.73877555e-01 -5.96154153e-01 -6.54527664e-01 -1.37794867e-01 -1.95699349e-01 1.03413367e+00 8.12114120e-01 -6.74339011e-02 -1.39782238e+00 1.23658907e+00 5.85549295e-01 -7.25210160e-02 -3.34034353e-01 1.30576038e+00 8.98253798e-01 -1.41372578e-02 2.76614010e-01 8.92115235e-02 1.69495225e+00 -7.57065475e-01 -8.27828467e-01 -4.47952449e-01 -1.67212903e-01 -6.86817586e-01 8.89164150e-01 5.37759900e-01 -9.26704705e-01 -6.59891844e-01 -9.62946773e-01 1.34065941e-01 -5.97932637e-01 4.84354377e-01 4.90131319e-01 1.21285713e+00 -9.60641682e-01 1.37049794e-01 -1.09046805e+00 -5.04579246e-01 6.07187986e-01 5.91802597e-01 -2.40527138e-01 -2.75465727e-01 -6.52271390e-01 1.16714680e+00 -5.02273701e-02 6.54363692e-01 -8.29516053e-01 -3.97745043e-01 -5.12459457e-01 -2.19639733e-01 5.75660348e-01 -2.38177150e-01 1.01974106e+00 -7.10100651e-01 -1.32337308e+00 9.57260787e-01 -7.21331537e-01 -1.32188305e-01 1.13780236e+00 -2.31281787e-01 -3.09514105e-01 2.93899119e-01 -2.14398831e-01 7.53176630e-01 1.48342538e+00 -1.43880653e+00 -8.33857596e-01 -6.14337504e-01 -2.06405185e-02 9.17220190e-02 1.74908340e-01 5.01867592e-01 -8.45332086e-01 -9.52897742e-02 -1.40237525e-01 -9.23202515e-01 1.50772601e-01 2.04178020e-01 -4.21208620e-01 -6.17077470e-01 1.28464007e+00 -9.60342407e-01 8.26751053e-01 -1.89728463e+00 -1.28079742e-01 4.42318678e-01 3.11922073e-01 4.86755311e-01 -1.80876166e-01 -8.18245932e-02 2.08918527e-01 -6.28208339e-01 -1.10123038e-01 -4.32450593e-01 -5.47113232e-02 2.56506562e-01 -6.08650409e-02 7.03022838e-01 4.14049506e-01 9.92573500e-01 -7.65454054e-01 -5.71598530e-01 6.77568436e-01 7.13463545e-01 2.71426678e-01 1.51323318e-01 9.37296376e-02 3.09731990e-01 -3.20839882e-01 1.28852522e+00 9.67212737e-01 7.28783011e-01 -2.19424486e-01 -4.09452468e-01 -2.89295793e-01 -3.53814751e-01 -1.34390306e+00 1.24133062e+00 -4.22147900e-01 1.02055871e+00 6.47351146e-01 -4.13268924e-01 5.51856875e-01 2.20461577e-01 2.38942131e-01 -5.12151420e-01 5.70405662e-01 5.31906150e-02 1.38416514e-01 -9.15609539e-01 4.23766226e-01 2.09443077e-01 4.75110054e-01 4.23009604e-01 -2.49292180e-01 2.09454000e-02 6.69650435e-02 -9.84522998e-02 1.02073240e+00 1.63732275e-01 -4.45242554e-01 -4.02368307e-02 2.76768804e-01 -1.44085437e-01 1.62891269e-01 7.73230851e-01 -7.97490180e-01 5.31300545e-01 2.42089510e-01 -3.01423311e-01 -6.06226265e-01 -9.81778860e-01 -3.30928192e-02 1.26225650e+00 3.74524117e-01 3.98104668e-01 -1.17253447e+00 -7.61041701e-01 2.17844144e-01 5.67396164e-01 -7.97537625e-01 1.59003526e-01 -7.29706228e-01 -4.31636244e-01 5.22265732e-01 1.04572904e+00 3.67166966e-01 -1.44973993e+00 -1.39219880e+00 2.97153741e-02 -1.18102152e-02 -1.31298876e+00 -4.19403404e-01 2.74389535e-01 -1.57890156e-01 -1.41848886e+00 -9.02727842e-01 -7.72992492e-01 7.35034525e-01 3.34781170e-01 8.08563352e-01 1.59432337e-01 -1.08832848e+00 7.43869960e-01 -1.77823856e-01 -8.14219296e-01 -3.02359164e-01 -1.66370660e-01 1.53894126e-01 2.27794632e-01 7.12119639e-01 1.37323916e-01 -9.27385688e-01 5.57734787e-01 -5.21419346e-01 -2.32786357e-01 6.93098009e-01 4.55334038e-01 3.18673670e-01 -2.23201420e-03 2.98351407e-01 -2.14759678e-01 6.70382977e-01 3.47728401e-01 -6.18798673e-01 5.53969979e-01 -1.98653668e-01 -2.23798126e-01 -2.10708186e-01 -4.14819419e-01 -1.18463933e+00 7.30811536e-01 -1.47342965e-01 -5.69064260e-01 -6.40439987e-01 -5.42384803e-01 -1.73815116e-01 -4.84628469e-01 5.42375326e-01 5.10999486e-02 1.51454642e-01 -1.17924578e-01 3.47022712e-01 1.10259855e+00 6.73006177e-01 -2.41679400e-01 5.93661547e-01 8.65992069e-01 2.24275514e-02 -1.17624128e+00 -1.45222619e-01 -8.59987617e-01 -1.31533086e+00 -1.00165474e+00 1.09806478e+00 -4.86691952e-01 -1.14256442e+00 8.05491269e-01 -1.35985649e+00 -6.18491650e-01 2.05904320e-01 1.78469121e-01 -2.40016192e-01 1.89233899e-01 -1.74493879e-01 -1.50464356e+00 -3.27106804e-01 -1.34319067e+00 1.44509816e+00 2.77607143e-01 -9.81526449e-02 -3.50407809e-01 -7.10309684e-01 3.81807029e-01 3.10063899e-01 3.51668090e-01 2.53786504e-01 1.22324035e-01 -6.19537950e-01 -6.50170386e-01 -4.39436734e-01 2.30353445e-01 1.18542597e-01 1.55055329e-01 -1.62921143e+00 -2.59748518e-01 -4.99829829e-01 -3.70859308e-03 1.10176635e+00 4.43899482e-01 1.06878066e+00 2.67850816e-01 -6.97577417e-01 8.79749432e-02 8.87118280e-01 1.79619521e-01 5.50918341e-01 -1.52742416e-01 8.03447306e-01 1.04734647e+00 7.75965095e-01 1.74025923e-01 7.28827249e-03 5.40582359e-01 5.97632229e-01 -4.41162795e-01 -5.60218751e-01 3.88000935e-01 3.60103667e-01 -2.59845465e-01 -6.58240259e-01 -1.80057362e-02 -9.78280127e-01 6.75106227e-01 -1.45366490e+00 -8.92205477e-01 -3.76541108e-01 1.95177948e+00 4.14996564e-01 1.85332485e-02 6.01549447e-01 3.14196408e-01 6.02566421e-01 -1.86838239e-01 -8.59818101e-01 -1.64670795e-01 3.06069776e-02 6.55607522e-01 1.05952454e+00 5.94262362e-01 -1.16812658e+00 1.13761449e+00 5.61515903e+00 4.82463956e-01 -1.21914041e+00 2.52670079e-01 2.66082942e-01 -2.65484095e-01 3.29725653e-01 -6.17109239e-01 -9.54805851e-01 4.46190089e-02 2.07767010e-01 8.24924529e-01 5.17043591e-01 7.81346321e-01 2.42962345e-01 -4.45899814e-01 -9.50760543e-01 1.18695343e+00 3.64704251e-01 -6.11605883e-01 -5.63779533e-01 -1.44595534e-01 1.86571211e-01 4.75656800e-02 -7.27369729e-03 -2.64357105e-02 -1.02083713e-01 -1.01393771e+00 9.80934918e-01 5.09475470e-01 9.12512124e-01 -6.50255561e-01 5.65053999e-01 2.65977442e-01 -1.41850722e+00 -2.36284420e-01 -1.84057549e-01 3.91937420e-02 6.77768290e-02 1.29498869e-01 -1.03614688e+00 -1.01949193e-01 9.63674068e-01 1.50826558e-01 -6.40641034e-01 7.38115251e-01 -2.58367121e-01 2.13969037e-01 -3.67576838e-01 -2.29254544e-01 2.20742553e-01 4.00987148e-01 5.20516574e-01 1.72581279e+00 -2.92650640e-01 2.51191646e-01 4.02671322e-02 1.01161098e+00 3.41571689e-01 -4.70377207e-01 -3.04219842e-01 2.97480404e-01 1.81222439e-01 1.49750435e+00 -1.17122686e+00 -3.80716026e-01 -3.50370884e-01 1.33304834e+00 -1.70673221e-01 7.43331254e-01 -7.60718644e-01 -6.05458856e-01 9.58855450e-01 3.98923397e-01 3.87029797e-01 -3.94807845e-01 -2.26459354e-01 -5.53005159e-01 5.04420102e-01 -5.00694931e-01 -1.33771673e-01 -6.94255888e-01 -7.58676827e-01 6.41754746e-01 -2.29386650e-02 -8.04762602e-01 -1.51301140e-03 -1.20028102e+00 -5.33176482e-01 1.04854488e+00 -1.76760137e+00 -1.33571589e+00 -1.26257348e+00 8.81229222e-01 8.01972508e-01 6.93771169e-02 4.44268316e-01 1.85221538e-01 -5.10346055e-01 8.35439444e-01 -4.64925170e-01 2.69600153e-01 6.22382462e-01 -1.06704521e+00 5.43230057e-01 9.24358606e-01 -2.43091866e-01 3.17682683e-01 6.27898991e-01 -6.42516971e-01 -1.80340624e+00 -9.70725119e-01 3.45313221e-01 -8.68140161e-01 1.67879254e-01 -6.46540642e-01 -7.50694931e-01 3.41099948e-01 1.85685918e-01 2.07192898e-01 -1.07169479e-01 -3.63366604e-01 -1.02688521e-01 -1.57548487e-01 -1.30546951e+00 4.59501088e-01 1.17339098e+00 -5.19648790e-01 -3.08811605e-01 4.93835688e-01 -1.40312418e-01 -7.78767228e-01 -1.44847751e-01 2.27339417e-01 1.11730516e+00 -1.05154133e+00 1.00093579e+00 -3.43970627e-01 -3.42258781e-01 -4.14200693e-01 2.88022786e-01 -6.59538388e-01 1.03589967e-01 -5.89023292e-01 6.96400367e-03 8.32768559e-01 -5.15534543e-02 -5.06588399e-01 1.12755835e+00 9.97703433e-01 1.07525595e-01 -5.48233628e-01 -1.09222090e+00 -8.39087963e-01 -2.86095172e-01 -7.83343077e-01 2.81374127e-01 3.38962466e-01 -1.49496704e-01 -4.36302990e-01 1.02549314e-01 4.23199266e-01 7.94647455e-01 -1.37870312e-01 9.47367966e-01 -9.67112303e-01 2.01617315e-01 -6.60919011e-01 -5.98525405e-01 -7.27633774e-01 1.86298385e-01 -3.48261446e-01 6.27250671e-01 -1.52611923e+00 1.09424628e-01 -4.27600473e-01 -2.83329338e-01 7.00480461e-01 -2.79133826e-01 5.20986617e-01 1.49838015e-01 -5.83735146e-02 -1.40120804e-01 -3.06554496e-01 1.13243699e+00 -3.11641067e-01 -3.74173015e-01 2.16806188e-01 1.72158793e-01 5.15685320e-01 6.76665246e-01 -8.30644369e-02 1.49081960e-01 -8.83149356e-02 -3.14584404e-01 -1.85488611e-01 9.33212996e-01 -9.67955530e-01 3.89682829e-01 -1.19058721e-01 4.32539165e-01 -9.16264653e-01 4.59321201e-01 -9.89578903e-01 -4.48149562e-01 7.21110761e-01 6.64623007e-02 -1.12809412e-01 5.89100301e-01 4.68856812e-01 1.98788583e-01 4.82675917e-02 7.61973143e-01 -7.42433593e-02 -1.07610357e+00 1.43257022e-01 -6.33766651e-01 -6.89854205e-01 1.38150752e+00 -6.98469162e-01 6.00380041e-02 -6.83241710e-02 -7.08856583e-01 1.76589653e-01 4.76865210e-02 7.93141365e-01 8.84643197e-01 -7.01654434e-01 -6.69299781e-01 6.31591439e-01 9.88244936e-02 6.97894953e-03 3.44002843e-01 1.07027018e+00 -5.36415219e-01 3.61265510e-01 -2.77801305e-01 -8.37845743e-01 -1.98781526e+00 4.08711374e-01 6.04415655e-01 5.48920453e-01 -6.68085396e-01 1.10762811e+00 3.97214247e-03 1.52472809e-01 8.78826678e-01 -8.55003655e-01 6.88065961e-02 5.40471897e-02 6.80089831e-01 6.56931758e-01 4.54363525e-01 -7.37301171e-01 -7.52482116e-01 8.23905885e-01 5.61787412e-02 -7.25004310e-03 7.66115129e-01 9.92709026e-02 2.00912222e-01 -9.21960175e-02 8.55094254e-01 -4.42403480e-02 -1.71964395e+00 1.27917007e-01 -4.04459476e-01 -5.06243646e-01 1.49607897e-01 -1.00300419e+00 -1.28456604e+00 1.17760074e+00 1.23251963e+00 -3.16517591e-01 9.93864715e-01 2.17873514e-01 7.09490478e-01 4.08116877e-01 3.38502944e-01 -1.36701155e+00 -1.80436298e-01 6.81209490e-02 1.06306124e+00 -1.44890559e+00 -1.64416775e-01 -5.98229349e-01 -5.90840816e-01 1.32014310e+00 7.30818689e-01 1.22134864e-01 4.74598557e-01 6.69939995e-01 4.71785694e-01 -3.60024691e-01 9.57449973e-02 -7.05213368e-01 5.18828154e-01 8.78834426e-01 -9.36270356e-02 2.89048016e-01 1.38629377e-01 1.58969715e-01 2.02641194e-03 -6.48576170e-02 7.14232698e-02 1.19392526e+00 -4.83472556e-01 -6.85720980e-01 -7.84835756e-01 4.12404031e-01 -9.65273455e-02 2.80775964e-01 -8.17744970e-01 9.34461117e-01 4.21961874e-01 1.06761420e+00 6.33112863e-02 -3.84323984e-01 5.60604811e-01 3.97895694e-01 8.95631135e-01 -2.53021121e-01 -7.12242663e-01 -1.59949824e-01 -2.22352684e-01 -9.77447033e-01 -4.37015861e-01 -7.21779227e-01 -1.22060788e+00 -1.46845981e-01 -4.98382866e-01 -7.28245735e-01 1.03602767e+00 1.11191642e+00 -2.39323750e-02 6.38590574e-01 1.57540873e-01 -1.46799600e+00 -3.03368300e-01 -7.77726591e-01 -5.91737747e-01 2.33561620e-01 5.99368393e-01 -1.05163717e+00 -6.22256249e-02 2.67788302e-02]

[6.596848964691162, -0.6113993525505066]

a284fffc-0eb1-4453-b08e-c6ebc6e1b5bf

pixel-level-segmentation-based-drivable-road

https://ieeexplore.ieee.org/document/9646953

https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9646953

Pixel Level Segmentation Based Drivable Road Region Detection and Steering Angle Estimation Method for Autonomous Driving on Unstructured Roads

With the recent emergence of deep learning, computer vision-based applications have demonstrated better applicability in accomplishing driving tasks including drivable road region detection, lane keeping and steering control in self-driving cars. Till recently, numerous lane-marking detection based steering control and lane keeping methods have been proposed to perform autonomous driving on urban well-structured roads. But the matter of fact is that these methods are not feasible on roads where lane markings are not available or faded over time which makes drivable road region detection a crucial task. Moreover, it is highly arduous task to estimate steering angle on deteriorated roads using existing road detection and steering angle estimation methods. To the best of our knowledge, there is no standard benchmark available for drivable road region detection and steering angle estimation on unstructured roads. To this end, we present a large-scale dataset for drivable region road detection, comprising of 15,000-pixel level high quality fine annotations. Alongside dataset, we also present an end-to-end drivable road region detection and steering angle estimation method to ensure the autonomous driving on generalized urban, rural, and unstructured road conditions. The proposed method performs pixel-level segmentation to extract drivable road region and quantifies lane interception to estimate the steering angle of self-driving cars. A comprehensive qualitative and quantitative analysis has been carried out to demonstrate the effectiveness of our proposed dataset, road detection and steering angle estimation methods. Our benchmark is available at https://carl-dataset.github.io/index/ .

['Muhammad Akram', 'Adel Sulaiman', 'Faisal Riaz', 'Muhammad Atif Butt', 'Marya Rasib']

2021-12-10

ieee-access-2021-12

['steering-control']

['computer-vision']

[ 1.10545412e-01 2.23533437e-01 -2.79214203e-01 -5.35844445e-01 -5.16717672e-01 -5.37120044e-01 6.98768497e-01 -3.04466903e-01 -4.05327290e-01 5.66369474e-01 -2.21885607e-01 -8.71123016e-01 1.39620453e-01 -9.88433123e-01 -5.81717253e-01 -4.91222382e-01 1.77953586e-01 2.12068960e-01 8.05540383e-01 -6.78394556e-01 3.10884029e-01 8.66545916e-01 -1.95168996e+00 -2.73265570e-01 1.10556054e+00 7.18293726e-01 4.94210690e-01 7.20370591e-01 9.07103643e-02 2.07901657e-01 2.39791825e-01 -2.69020706e-01 4.18201804e-01 3.36432099e-01 -2.63619721e-01 3.54437828e-02 7.96013236e-01 -6.00746393e-01 -7.14184284e-01 7.53512561e-01 4.01856959e-01 3.57761495e-02 7.35196590e-01 -1.19898927e+00 1.31649584e-01 -9.86366719e-02 -6.34837329e-01 3.40905696e-01 -1.17344171e-01 6.00309730e-01 5.30340552e-01 -8.68633091e-01 4.39843416e-01 1.03401518e+00 3.88171524e-01 1.40714839e-01 -6.41739249e-01 -7.51207650e-01 -3.64112388e-03 3.14738095e-01 -1.41370714e+00 -6.94726408e-01 7.81102955e-01 -4.73731965e-01 5.51275671e-01 2.72872802e-02 3.47802669e-01 5.54179370e-01 4.78844494e-01 8.19804847e-01 1.15784395e+00 1.71890166e-02 -1.55898049e-01 2.20195517e-01 4.36388776e-02 8.73312771e-01 2.64987856e-01 5.62175274e-01 8.75256509e-02 7.82618642e-01 5.61752379e-01 -2.36718327e-01 -7.02306405e-02 -5.69712460e-01 -1.14310670e+00 7.03545451e-01 5.54520786e-01 -3.06241214e-01 -2.79198200e-01 1.85353570e-02 4.50119615e-01 2.36180782e-01 2.22308412e-01 -2.45695055e-01 -2.16242507e-01 -2.21408486e-01 -8.20174217e-01 2.46504828e-01 2.46509716e-01 1.10683930e+00 1.19004095e+00 2.01658443e-01 1.62343293e-01 8.24908972e-01 2.41220355e-01 1.02822864e+00 -2.27843672e-01 -1.18038690e+00 6.49171174e-01 3.54443222e-01 1.38710871e-01 -1.07548714e+00 -5.65985382e-01 -3.05683941e-01 -8.35145593e-01 6.92119837e-01 5.02825499e-01 -1.38011754e-01 -9.78323519e-01 1.14443111e+00 4.43691432e-01 2.66210362e-02 -6.12483732e-02 1.02405643e+00 9.38672423e-01 4.99798447e-01 -1.51069120e-01 1.46794856e-01 1.23986936e+00 -8.96140337e-01 -6.09810889e-01 -5.32162309e-01 7.31130898e-01 -7.01762319e-01 1.06328189e+00 2.52702892e-01 -7.36522615e-01 -7.22872257e-01 -1.00496709e+00 -2.24785656e-01 -6.95500076e-01 3.76377672e-01 5.58916926e-01 7.85732090e-01 -9.34797168e-01 -2.10793495e-01 -2.58420706e-01 -1.95402041e-01 6.28152549e-01 1.51038142e-02 -5.70126712e-01 -3.49398881e-01 -1.46663344e+00 1.08007455e+00 1.78114980e-01 3.53423238e-01 -1.07091105e+00 -5.74398160e-01 -8.72698128e-01 -4.74488884e-01 5.90739429e-01 -2.05483362e-01 1.00500298e+00 -3.53502512e-01 -1.38595593e+00 1.07264400e+00 -2.85043269e-01 -5.54325700e-01 9.10198271e-01 -1.44173458e-01 -6.86147034e-01 -1.70273364e-01 2.16017693e-01 8.53832960e-01 7.62175739e-01 -1.38731027e+00 -1.33767736e+00 -2.71808445e-01 -6.78644180e-02 2.84751654e-01 3.63441586e-01 -4.78443116e-01 -5.90281129e-01 4.32166681e-02 -1.13197893e-01 -9.29882169e-01 -1.80164129e-01 7.66561329e-02 -5.86983144e-01 -1.03534125e-01 1.34908819e+00 -4.89253104e-01 1.13375092e+00 -2.03053617e+00 -7.52017200e-01 1.55001581e-01 3.06243360e-01 5.48634827e-01 -2.03046516e-01 4.19859998e-02 2.97760218e-01 -2.21730977e-01 -4.01230752e-01 2.30708867e-01 -1.13309987e-01 4.47483622e-02 -4.01312768e-01 5.26413083e-01 1.74959183e-01 9.56958652e-01 -8.34561765e-01 -5.00444412e-01 8.22436392e-01 5.22881806e-01 5.61299501e-03 -1.26426429e-01 2.97126651e-01 2.81354070e-01 -3.90401185e-01 7.68908322e-01 1.25448811e+00 7.62760878e-01 -4.22178298e-01 -2.12440640e-01 -8.58042836e-01 9.95664150e-02 -1.15654933e+00 8.64347696e-01 -6.65976167e-01 1.34920168e+00 2.16161728e-01 -8.11341703e-01 1.24830222e+00 -2.04574734e-01 1.21266574e-01 -1.28978217e+00 -4.13284404e-03 3.73784542e-01 -4.47851531e-02 -5.28184295e-01 9.12680268e-01 3.43366861e-01 -2.17110977e-01 -1.48323283e-01 -6.80700243e-01 -5.61238885e-01 2.80143529e-01 6.24397621e-02 8.99773002e-01 2.18992941e-02 3.09376679e-02 -4.81902435e-02 8.56853306e-01 2.80601442e-01 3.78792584e-01 5.38210034e-01 -7.79859483e-01 4.29596931e-01 2.44167998e-01 -5.30629158e-01 -1.18820608e+00 -1.34145486e+00 -4.71836835e-01 6.61475599e-01 3.78401428e-01 4.27042276e-01 -5.87537825e-01 -5.07733405e-01 2.78323978e-01 7.18299568e-01 -2.63515890e-01 6.81523234e-02 -7.63006985e-01 -2.20003411e-01 4.35593247e-01 2.92513728e-01 9.72465217e-01 -9.34994996e-01 -6.44498527e-01 2.32920513e-01 -8.58587697e-02 -1.32626879e+00 -3.22553664e-01 -1.18905291e-01 -4.21239823e-01 -1.02383435e+00 -4.65921104e-01 -7.99678624e-01 3.80370557e-01 1.02411652e+00 9.50173020e-01 -4.05614763e-01 -2.71274716e-01 -2.24196315e-02 6.94998503e-02 -6.11555398e-01 -4.19926792e-01 2.16160834e-01 -2.65131921e-01 -1.28131136e-01 2.38541245e-01 -1.67603567e-01 -9.11152422e-01 1.06679046e+00 -3.17905694e-01 2.53028542e-01 9.05963302e-01 3.78689110e-01 6.08689785e-01 2.19022498e-01 8.77342880e-01 -7.82846987e-01 2.62401611e-01 -5.34994125e-01 -1.08122253e+00 -2.32462749e-01 -6.47657514e-01 -3.11113745e-01 5.32862723e-01 2.23044559e-01 -1.07711053e+00 2.73082227e-01 -3.72785717e-01 2.42336374e-02 -7.81988204e-01 -3.71013768e-02 -3.53691638e-01 4.58508842e-02 5.57585955e-01 2.55010635e-01 1.82853580e-01 1.35531068e-01 7.82508314e-01 9.96132910e-01 9.39864695e-01 2.20493570e-01 1.05886495e+00 8.83949578e-01 2.51647860e-01 -1.06354702e+00 -3.81268740e-01 -7.71751046e-01 -7.97390580e-01 -7.77829945e-01 7.03721285e-01 -1.12998009e+00 -7.79176831e-01 6.53457582e-01 -6.96269751e-01 -6.58785582e-01 9.19052288e-02 1.98004723e-01 -6.39575005e-01 4.09854889e-01 6.80449158e-02 -9.62298930e-01 -1.79814756e-01 -1.16750026e+00 9.47396040e-01 2.98550069e-01 2.57500052e-01 -7.55533099e-01 -2.39461526e-01 7.71281421e-01 6.00241721e-01 2.56622791e-01 4.05827850e-01 3.16610783e-01 -8.39938343e-01 -4.15996969e-01 -7.13792145e-01 7.81955197e-02 4.78246287e-02 9.42481905e-02 -9.70045447e-01 1.85137421e-01 -8.40660870e-01 2.42027361e-02 1.17208660e+00 6.42821372e-01 8.55892479e-01 2.23901853e-01 -4.11840022e-01 4.40782547e-01 1.12834001e+00 1.98771670e-01 9.77156579e-01 4.86107916e-01 8.34847867e-01 9.10611928e-01 1.36652529e+00 1.50009915e-02 7.82635570e-01 7.08457828e-01 8.36725354e-01 -3.63046229e-01 -4.73809063e-01 -9.53508243e-02 3.76032799e-01 2.16913790e-01 1.25320971e-01 -3.45696598e-01 -1.09244239e+00 9.78661120e-01 -1.57604432e+00 -1.01162827e+00 -9.63638365e-01 2.33253407e+00 2.87726909e-01 5.63785970e-01 1.90665811e-01 1.64848402e-01 6.86305821e-01 1.45554468e-01 -6.89180553e-01 -6.94751382e-01 -1.28394738e-01 -4.34555888e-01 1.19524240e+00 8.35735142e-01 -1.29205751e+00 1.23317122e+00 5.04579067e+00 1.08640635e+00 -1.18741810e+00 -8.37722495e-02 6.99419737e-01 4.05554563e-01 -4.00221944e-01 -1.06708318e-01 -1.14361143e+00 3.64870906e-01 9.03213859e-01 2.76129037e-01 5.38732633e-02 7.88051128e-01 1.09898877e+00 -5.80654383e-01 -2.47571766e-01 7.15096116e-01 -4.89994615e-01 -1.08615148e+00 -3.68367463e-01 1.34318709e-01 6.65356278e-01 4.48146403e-01 1.78059384e-01 3.06559116e-01 2.43347511e-01 -8.93941581e-01 4.49711829e-01 6.37080908e-01 1.09408092e+00 -1.10497200e+00 7.26105154e-01 4.90731239e-01 -1.50031650e+00 -3.89900394e-02 -2.80567616e-01 2.17116550e-01 4.60603088e-01 8.66953552e-01 -1.18198216e+00 3.14674616e-01 3.05741727e-01 7.96618044e-01 -4.91145641e-01 1.07207406e+00 -3.41342062e-01 4.95847344e-01 -1.80944160e-01 1.50079027e-01 5.39838910e-01 -3.52653325e-01 6.56094015e-01 1.27496815e+00 1.74064532e-01 -3.05494249e-01 -5.58746196e-02 5.95404208e-01 2.05136254e-01 -1.66684054e-02 -1.11369824e+00 4.77371365e-01 4.27509606e-01 1.65748715e+00 -5.70139050e-01 -5.79215363e-02 -4.38741595e-01 3.93045783e-01 -2.10497659e-02 3.37761968e-01 -9.51139569e-01 -6.94571853e-01 8.34743083e-01 8.29244256e-01 1.91197038e-01 -5.01334012e-01 -4.42187548e-01 -4.26138341e-01 -1.58088714e-01 -1.78596959e-01 -2.41634399e-01 -7.34861016e-01 -5.32827973e-01 1.77191481e-01 -1.70233101e-02 -1.25987649e+00 6.41787723e-02 -5.63996077e-01 -7.19803274e-01 7.51084268e-01 -2.29952884e+00 -1.06699562e+00 -8.58123600e-01 2.63463497e-01 9.00163949e-01 -2.83251196e-01 -3.20267230e-02 4.93955255e-01 -6.25030339e-01 3.86828631e-01 1.97833836e-01 -1.02751531e-01 6.89876258e-01 -1.01563895e+00 6.00676417e-01 9.40623224e-01 -7.46765912e-01 -5.99514432e-02 6.58935130e-01 -5.88623345e-01 -1.33484578e+00 -1.50317276e+00 6.83365822e-01 -2.84546226e-01 4.21602637e-01 -3.46447021e-01 -8.06041539e-01 4.75488864e-02 -1.51432261e-01 1.24013498e-01 -2.36237973e-01 -3.99949402e-01 1.82165712e-01 -5.64576149e-01 -1.03387558e+00 6.91972315e-01 1.04963279e+00 -2.78194994e-01 8.32017288e-02 6.51629269e-02 2.01634392e-01 -2.29043767e-01 -4.04739738e-01 5.82663357e-01 6.91857874e-01 -1.07047105e+00 9.87167716e-01 3.28533351e-01 2.65954822e-01 -7.54066765e-01 8.50002840e-02 -1.03244829e+00 -1.86921731e-01 -4.33850348e-01 9.98249352e-02 1.05087507e+00 4.75230336e-01 -8.68640363e-01 9.77374196e-01 1.65024787e-01 -7.66643941e-01 -7.18206882e-01 -9.63703990e-01 -5.12888849e-01 9.60009098e-02 -7.25430965e-01 3.03815633e-01 4.44628447e-01 -6.51094794e-01 1.69196784e-01 -1.50947154e-01 2.75217831e-01 7.66689122e-01 -4.89616767e-02 1.23211873e+00 -1.15094316e+00 8.50342691e-01 -6.28185749e-01 -5.26274323e-01 -1.25806999e+00 2.35975951e-01 -7.28520453e-01 4.38993901e-01 -1.81099355e+00 -2.95012385e-01 -6.64152920e-01 1.26416460e-01 2.27820784e-01 -2.73741689e-02 3.92415762e-01 -3.45986038e-01 7.46430308e-02 -3.48724127e-01 3.56154323e-01 1.51528037e+00 -2.05737546e-01 -1.93996325e-01 2.63763666e-01 -3.18130583e-01 5.02158403e-01 1.29278660e+00 6.84681982e-02 -7.12427378e-01 3.71467322e-02 1.40051141e-01 -4.88117673e-02 4.69530106e-01 -1.02280688e+00 1.55156448e-01 -2.39095122e-01 -1.16305456e-01 -1.42757666e+00 1.24568507e-01 -5.78655124e-01 -2.41162628e-01 3.47509861e-01 4.47296090e-02 -2.53226191e-01 2.93595493e-01 5.94628513e-01 -1.00267403e-01 2.78453976e-01 1.17578089e+00 2.96566427e-01 -1.66814697e+00 3.82224292e-01 -8.92654836e-01 -2.25092564e-02 1.33760107e+00 -8.91293347e-01 -3.61878186e-01 -4.48495358e-01 -1.71715572e-01 6.80775583e-01 4.38089162e-01 6.13414586e-01 9.53969777e-01 -1.04560030e+00 -1.00357974e+00 3.85886520e-01 4.16415989e-01 6.68314695e-02 5.39772570e-01 1.02625263e+00 -6.52711272e-01 8.33061576e-01 -3.53818208e-01 -6.52628422e-01 -1.16528428e+00 3.50910455e-01 3.37527066e-01 2.70187974e-01 -5.85122526e-01 2.62454450e-01 4.05587196e-01 -6.00651503e-01 -2.75364548e-01 -2.51591861e-01 -4.01389182e-01 -4.38125804e-02 1.70391738e-01 7.53464103e-01 9.83127430e-02 -1.17193747e+00 -2.93187529e-01 7.88264692e-01 -1.06672558e-03 1.15296438e-01 7.71581888e-01 -8.92022312e-01 5.75751841e-01 1.65721595e-01 1.01341164e+00 -8.23934972e-02 -1.67326355e+00 3.92582081e-02 -3.07436436e-01 -4.43777651e-01 5.66813648e-01 -4.82180446e-01 -1.14242625e+00 1.10729897e+00 1.00637496e+00 -1.86944846e-03 7.43573725e-01 -2.00513855e-01 1.11114919e+00 3.76464456e-01 2.63311237e-01 -1.49930573e+00 -4.69191462e-01 7.71175265e-01 6.29194558e-01 -1.87115991e+00 -1.28768653e-01 -7.40786552e-01 -7.22400188e-01 1.17063987e+00 7.16152370e-01 1.63743168e-01 6.10763490e-01 2.64313668e-01 2.17258468e-01 -1.54043809e-01 -3.56538624e-01 -7.55106390e-01 3.21489513e-01 9.37498927e-01 1.05054557e-01 3.67878109e-01 -1.42568767e-01 -4.45971817e-01 -1.92200541e-01 -1.39006510e-01 6.84624732e-01 5.06723583e-01 -1.12627804e+00 -6.09357357e-01 -2.06729993e-01 6.36826098e-01 2.12907597e-01 2.21866205e-01 5.80462962e-02 1.01932931e+00 2.27958262e-01 1.26948905e+00 5.43377250e-02 -1.83295488e-01 6.02892995e-01 -3.79914641e-01 -1.95571974e-01 -1.13679498e-01 2.53025085e-01 -3.53485793e-01 4.87157881e-01 -4.54148531e-01 3.94670218e-02 -6.56695962e-01 -1.46995282e+00 -5.67520618e-01 -1.73841611e-01 -5.07498145e-01 8.65553975e-01 6.90818071e-01 3.36118013e-01 1.56589612e-01 1.07416296e+00 -8.11030746e-01 1.19794242e-01 -4.80624765e-01 -4.82663900e-01 9.73626301e-02 4.31806356e-01 -8.47601116e-01 -2.64213502e-01 -2.46160030e-01]

[8.066669464111328, -1.5444262027740479]

9eaf0dcc-c58b-4fe0-9e21-f2eb51181822

sentence-representation-learning-with

2210.08474

https://arxiv.org/abs/2210.08474v2

https://arxiv.org/pdf/2210.08474v2.pdf

Sentence Representation Learning with Generative Objective rather than Contrastive Objective

Though offering amazing contextualized token-level representations, current pre-trained language models take less attention on accurately acquiring sentence-level representation during their self-supervised pre-training. However, contrastive objectives which dominate the current sentence representation learning bring little linguistic interpretability and no performance guarantee on downstream semantic tasks. We instead propose a novel generative self-supervised learning objective based on phrase reconstruction. To overcome the drawbacks of previous generative methods, we carefully model intra-sentence structure by breaking down one sentence into pieces of important phrases. Empirical studies show that our generative learning achieves powerful enough performance improvement and outperforms the current state-of-the-art contrastive methods not only on the STS benchmarks, but also on downstream semantic retrieval and reranking tasks. Our code is available at https://github.com/chengzhipanpan/PaSeR.

['Hai Zhao', 'Bohong Wu']

2022-10-16

['semantic-retrieval']

['natural-language-processing']

[ 2.43160486e-01 4.09964472e-01 -4.48799312e-01 -6.01807058e-01 -1.35629535e+00 -4.01681304e-01 8.04559112e-01 2.32145742e-01 -3.28748912e-01 8.38659823e-01 8.27419102e-01 -3.65487456e-01 5.06350771e-02 -8.85287881e-01 -6.70841217e-01 -5.26678741e-01 4.02844310e-01 6.58100903e-01 -8.80777910e-02 -3.67665470e-01 2.69491941e-01 -2.80622602e-01 -1.14753997e+00 5.25359869e-01 1.11134684e+00 5.48921347e-01 3.75538617e-01 4.40149307e-01 -5.43725550e-01 7.23947167e-01 -5.53746164e-01 -5.56294203e-01 -2.49260753e-01 -6.60729825e-01 -1.06819654e+00 -8.97550285e-02 2.12731019e-01 -6.52736351e-02 -3.59790206e-01 1.10620487e+00 4.19389278e-01 1.61308885e-01 6.59340739e-01 -7.62287855e-01 -1.44948280e+00 1.40053105e+00 -3.78190994e-01 4.30982381e-01 1.72897711e-01 7.24036596e-04 1.60905731e+00 -1.06055224e+00 5.56471109e-01 1.16234982e+00 5.07431269e-01 6.50075495e-01 -1.30963314e+00 -4.30108666e-01 4.42230672e-01 1.90250918e-01 -1.28910267e+00 -4.90642548e-01 8.39810610e-01 -6.61732703e-02 1.19188690e+00 2.10189745e-02 2.95468450e-01 1.42280090e+00 2.69987155e-02 1.10618794e+00 8.76058042e-01 -3.85487914e-01 4.14781906e-02 2.20948700e-02 3.90414327e-01 7.16972709e-01 2.78413594e-01 -3.79051894e-01 -6.07433021e-01 -2.91939098e-02 4.61416125e-01 1.73150818e-03 -1.44300133e-01 8.87711942e-02 -9.98481154e-01 1.09140980e+00 5.33642828e-01 6.91748381e-01 -2.89418489e-01 4.62400436e-01 3.47380966e-01 1.51981935e-01 9.84900534e-01 3.89402747e-01 -4.70452338e-01 -2.19518319e-01 -1.08023047e+00 -4.34931368e-02 4.73167777e-01 9.85758185e-01 7.21480131e-01 1.02978133e-01 -5.32337129e-01 9.71755803e-01 3.94255191e-01 3.63085151e-01 7.26609290e-01 -6.33521974e-01 7.24303782e-01 4.71833795e-01 -3.85242939e-01 -6.52195990e-01 -3.18974592e-02 -8.40381384e-01 -7.32189000e-01 -7.68445969e-01 5.84856533e-02 -5.42421360e-03 -9.05100763e-01 1.73778999e+00 -1.69044226e-01 -8.30220897e-03 2.61035472e-01 6.49410665e-01 1.02293193e+00 8.99887621e-01 5.12883186e-01 -9.90879983e-02 1.30627811e+00 -1.37973034e+00 -6.86031699e-01 -6.27710223e-01 8.29001844e-01 -6.04089677e-01 1.47936666e+00 2.50295689e-03 -1.18567169e+00 -5.57422578e-01 -7.99669445e-01 -4.54275221e-01 -1.98220745e-01 2.60641366e-01 8.41466129e-01 4.15317029e-01 -1.13207316e+00 5.14836371e-01 -8.48744631e-01 -1.88576922e-01 6.87324762e-01 -5.32477945e-02 -1.13060571e-01 -1.52592868e-01 -1.40619004e+00 7.47506738e-01 5.18609464e-01 -2.87977681e-02 -7.77569532e-01 -9.13311064e-01 -9.38858211e-01 4.20727640e-01 3.13634962e-01 -1.15809643e+00 1.43364668e+00 -8.54251444e-01 -1.48305809e+00 1.00138557e+00 -5.51474452e-01 -5.72376013e-01 1.56616718e-01 -6.24931574e-01 -5.97702898e-02 1.17182493e-01 3.43350619e-01 7.44571328e-01 6.29734099e-01 -1.20144665e+00 -3.31396073e-01 -1.45635709e-01 1.48924455e-01 3.82036984e-01 -6.28553689e-01 -1.15346305e-01 -3.55969429e-01 -8.40773165e-01 3.42767090e-02 -5.71534276e-01 -3.40570062e-01 -5.80514669e-01 -5.49591780e-01 -7.06864893e-01 3.23722810e-01 -6.44469261e-01 1.29195774e+00 -1.94940722e+00 1.94830030e-01 -3.44438434e-01 1.17319234e-01 2.90076166e-01 -3.22219938e-01 7.19186366e-01 -1.11103766e-02 3.66599351e-01 -4.37765867e-01 -8.18597972e-01 1.81607842e-01 2.48230979e-01 -6.80845082e-01 -2.85205245e-02 3.78067374e-01 1.35039198e+00 -1.26986694e+00 -4.54094529e-01 -7.16596320e-02 3.99664760e-01 -6.45872831e-01 8.25624168e-02 -4.23990726e-01 2.89829969e-01 -7.22550452e-01 4.06806380e-01 2.92948753e-01 -5.21516562e-01 3.29825133e-01 1.30352736e-01 2.97825247e-01 1.10129082e+00 -2.12550253e-01 2.29878998e+00 -8.11931252e-01 3.63786042e-01 -5.07485032e-01 -1.39880741e+00 8.17927957e-01 2.42990196e-01 2.32166827e-01 -6.49538457e-01 8.84148851e-02 1.94985166e-01 -2.70877659e-01 -2.79043227e-01 6.62976861e-01 -3.58450919e-01 -3.77171487e-01 5.69072187e-01 3.24637890e-01 -6.37146160e-02 2.34233558e-01 5.72196901e-01 1.06056798e+00 3.24728817e-01 2.84887731e-01 -4.75861281e-01 4.18548852e-01 3.75675894e-02 4.77017492e-01 6.83856606e-01 1.91726297e-01 8.06253970e-01 4.65612203e-01 3.75172608e-02 -9.10758913e-01 -1.32469487e+00 7.37127140e-02 1.33899701e+00 -6.72069436e-04 -8.10072541e-01 -7.31879175e-01 -9.62164521e-01 -1.31080061e-01 1.15624857e+00 -4.52893645e-01 -4.43538189e-01 -4.77963448e-01 -8.20522845e-01 5.71152508e-01 8.11352193e-01 4.35007125e-01 -1.07802069e+00 6.55121505e-02 2.05920950e-01 -4.93151963e-01 -1.05800664e+00 -5.56097925e-01 7.08111562e-03 -1.01720822e+00 -5.24432957e-01 -6.09161377e-01 -9.19462740e-01 8.09975684e-01 2.33323246e-01 1.45979869e+00 3.29180919e-02 -4.45108041e-02 2.61198908e-01 -6.10150576e-01 -1.95072085e-01 -3.90905112e-01 5.58413088e-01 -2.85803348e-01 -3.41586411e-01 4.96912777e-01 -5.97554028e-01 -5.57601988e-01 -3.17819983e-01 -7.98459291e-01 1.88161507e-01 6.78885758e-01 1.01457357e+00 6.70067310e-01 -1.87464699e-01 9.37173665e-01 -1.12943518e+00 8.02763700e-01 -7.12623537e-01 -1.46762468e-02 3.68921280e-01 -7.05863714e-01 4.53352451e-01 6.74490750e-01 -1.38050631e-01 -1.36086917e+00 -3.89247239e-01 -4.67976958e-01 -1.55407533e-01 -1.53245673e-01 6.96239710e-01 -1.06760211e-01 8.02924216e-01 4.76313323e-01 5.70145071e-01 -2.44719744e-01 -6.01373076e-01 6.83306754e-01 5.11746168e-01 3.13496023e-01 -7.75966883e-01 7.58818865e-01 2.56334037e-01 -4.06428695e-01 -6.42728865e-01 -1.61914241e+00 -4.11274225e-01 -5.46312392e-01 3.11660945e-01 6.94770277e-01 -1.04780436e+00 -1.14215523e-01 9.24159191e-04 -1.28220606e+00 -1.93274245e-01 -4.51910526e-01 1.40785605e-01 -4.84608382e-01 5.14488935e-01 -8.14360559e-01 -4.64396745e-01 -7.77311385e-01 -6.75918221e-01 1.28035772e+00 4.45504226e-02 -4.00676131e-01 -1.26059139e+00 1.47086635e-01 7.49634683e-01 5.04081309e-01 -3.60173196e-01 9.78726149e-01 -7.89253056e-01 -4.89116699e-01 8.21143910e-02 -1.35459706e-01 4.85412240e-01 1.87024504e-01 -3.55507791e-01 -9.36847329e-01 -8.75726864e-02 -6.34465665e-02 -3.69426548e-01 1.43632114e+00 2.33506203e-01 1.28266788e+00 -7.00587988e-01 -2.20310688e-01 4.05699134e-01 1.34719849e+00 -2.65356153e-01 5.50194144e-01 1.97115123e-01 7.33465910e-01 5.39886892e-01 4.36482251e-01 1.17275551e-01 6.72928631e-01 3.50258708e-01 5.17277345e-02 8.47258046e-02 -3.13645005e-01 -7.99302280e-01 5.86683035e-01 1.11859882e+00 7.23639205e-02 -5.02781272e-01 -7.92737722e-01 7.16941416e-01 -2.06427860e+00 -1.01501834e+00 -1.70493126e-03 1.71516764e+00 1.16721213e+00 2.25209624e-01 -2.71413565e-01 -1.45631388e-01 5.49382687e-01 4.44459826e-01 -2.47171462e-01 -3.98524106e-01 -1.01335458e-01 5.33558786e-01 1.65075690e-01 5.41177750e-01 -1.05335820e+00 1.41882658e+00 5.93319845e+00 1.15903807e+00 -7.94652939e-01 4.51126546e-01 7.86928356e-01 -1.13752753e-01 -9.31153655e-01 2.11133048e-01 -1.02336931e+00 4.57179815e-01 1.04235160e+00 -3.15775990e-01 -4.38149013e-02 9.14186120e-01 1.16480924e-01 9.75676030e-02 -1.04862547e+00 7.24845409e-01 2.72460908e-01 -1.52449942e+00 3.86332542e-01 -2.11991265e-01 7.65518188e-01 1.21273324e-01 1.09253913e-01 6.26987278e-01 3.61671716e-01 -9.95735228e-01 5.57060003e-01 4.59410489e-01 4.40898120e-01 -5.82988977e-01 6.80262327e-01 3.71760368e-01 -9.22598600e-01 2.56761760e-01 -4.45653915e-01 -1.38437495e-01 4.00939167e-01 8.20295513e-01 -6.91977799e-01 6.93837106e-01 3.36081624e-01 1.09321058e+00 -5.83511174e-01 4.15713400e-01 -9.05559778e-01 1.07293510e+00 5.53065874e-02 -2.61169821e-01 4.82617646e-01 -3.78836878e-02 4.94436949e-01 1.43029034e+00 2.47192711e-01 4.85623851e-02 7.39764720e-02 1.12790847e+00 -3.54312360e-01 1.42348990e-01 -4.74189579e-01 -4.04384434e-01 3.53591949e-01 1.12990296e+00 -5.47668517e-01 -5.73917150e-01 -3.10579181e-01 1.02487493e+00 5.87207019e-01 3.86432350e-01 -6.65594637e-01 -1.92091420e-01 6.90013647e-01 1.05197519e-01 2.85508066e-01 -3.49969178e-01 -4.29169148e-01 -1.51417267e+00 1.24800593e-01 -4.53775764e-01 4.25496548e-01 -5.87878287e-01 -1.60013282e+00 5.61466157e-01 2.63808556e-02 -7.17260718e-01 -4.87259030e-01 -4.19342041e-01 -7.46746242e-01 7.53738225e-01 -1.66839588e+00 -1.41986179e+00 8.36063698e-02 2.12749109e-01 1.04573250e+00 -7.98253417e-02 9.22606230e-01 1.33596882e-01 -4.93467152e-01 6.31133258e-01 1.34140566e-01 1.93318695e-01 4.49105144e-01 -1.29063118e+00 5.96064627e-01 8.93096030e-01 5.33067942e-01 8.60392988e-01 6.29451990e-01 -4.98103440e-01 -1.17774212e+00 -1.07457244e+00 1.44240689e+00 -5.38214445e-01 8.52340877e-01 -5.45851529e-01 -9.83567655e-01 8.00202310e-01 5.59090853e-01 -2.41248384e-01 7.72338927e-01 4.95447040e-01 -4.01601762e-01 9.52725187e-02 -5.90235710e-01 5.17415762e-01 1.43255508e+00 -6.93540156e-01 -1.10375381e+00 6.35913253e-01 1.06308854e+00 -2.29412243e-02 -6.04788244e-01 3.14118028e-01 2.39556674e-02 -6.04638755e-01 1.04334307e+00 -8.91729891e-01 9.77766037e-01 1.59132540e-01 1.49629274e-02 -1.46289861e+00 -4.41910446e-01 -4.58542913e-01 -8.31966847e-02 1.51484644e+00 7.62170255e-01 -6.68686926e-01 9.05669034e-01 1.65834337e-01 -4.93507475e-01 -9.48181868e-01 -6.22312129e-01 -8.52853000e-01 3.97414565e-01 -3.82170498e-01 2.42317423e-01 8.71075034e-01 2.38256842e-01 9.79364455e-01 -2.60713231e-02 -1.95015609e-01 7.01316118e-01 3.44391912e-01 3.37763876e-01 -7.67662048e-01 -5.15803039e-01 -6.17801905e-01 -1.00560665e-01 -1.26723719e+00 6.82919562e-01 -1.47951436e+00 1.50886774e-01 -1.97276270e+00 5.66874027e-01 -3.41358125e-01 -4.55957204e-01 6.91732705e-01 -5.45801222e-01 1.81638047e-01 -6.58307970e-02 3.19843553e-02 -7.72851944e-01 1.06341493e+00 1.10224617e+00 -2.29347274e-01 1.95403755e-01 -1.41719148e-01 -1.16090000e+00 5.79828203e-01 1.02202237e+00 -5.73554873e-01 -6.89700067e-01 -9.50468600e-01 2.77305067e-01 -2.34498248e-01 3.37106079e-01 -5.02877831e-01 8.07839110e-02 -1.43678570e-02 2.29692515e-02 -4.29916561e-01 2.56343156e-01 -3.75632159e-02 -3.43507022e-01 3.69812697e-01 -8.47525537e-01 -6.51630610e-02 3.35003249e-02 5.30758500e-01 -4.94672686e-01 -4.35677975e-01 3.14945221e-01 -3.08558434e-01 -6.20352924e-01 2.62440711e-01 -3.26797575e-01 4.09235001e-01 4.92528230e-01 1.82921410e-01 -5.67322016e-01 -4.27186370e-01 -7.04445601e-01 1.69116259e-01 3.25150639e-02 7.46074677e-01 7.20638812e-01 -1.21978569e+00 -9.44036424e-01 -6.14995398e-02 1.07217439e-01 1.71359051e-02 2.74970710e-01 5.37438095e-01 -1.49016485e-01 8.74276936e-01 2.28492722e-01 -3.48894417e-01 -9.26867545e-01 3.01105708e-01 -1.46908760e-01 -6.12610042e-01 -6.37434542e-01 1.17189264e+00 4.75354016e-01 -3.08560133e-01 -1.06890470e-01 -2.43400574e-01 -8.50939527e-02 -1.12334706e-01 2.44760051e-01 -3.07152383e-02 8.49425886e-03 -3.92544091e-01 -2.43932724e-01 2.21270189e-01 -2.75183171e-01 -5.39159663e-02 1.50105870e+00 -1.77806258e-01 -1.61852896e-01 4.04854804e-01 1.42554951e+00 -1.78944349e-01 -8.15644860e-01 -4.48444098e-01 2.48327538e-01 -1.82383209e-01 9.65037346e-02 -6.58173800e-01 -7.54312336e-01 1.09958196e+00 -2.15943381e-01 -5.86193241e-02 9.92184997e-01 4.25147146e-01 1.11573684e+00 4.99169946e-01 3.56627256e-01 -9.85048473e-01 3.95377249e-01 7.43756711e-01 8.22020590e-01 -1.12918556e+00 -1.11486785e-01 -4.92586583e-01 -7.70244539e-01 7.13893116e-01 4.52083290e-01 -3.98561001e-01 5.14150679e-01 5.33827916e-02 -2.13897616e-01 -1.66478410e-01 -1.01394570e+00 -3.02264124e-01 4.32085901e-01 2.72036493e-01 9.06067789e-01 7.93070346e-03 -6.93944991e-01 8.86709273e-01 -4.63082820e-01 -2.75698155e-01 1.68589905e-01 7.60277390e-01 -5.62606990e-01 -1.34576166e+00 2.92376965e-01 3.69248778e-01 -5.19501448e-01 -8.09631288e-01 -3.43375474e-01 3.44467461e-01 -3.40586066e-01 7.63402581e-01 9.31115076e-02 -8.98035541e-02 -6.93699047e-02 2.89008111e-01 4.50566888e-01 -1.13574517e+00 -5.05565643e-01 -2.73839720e-02 3.18731666e-01 -4.26489085e-01 -3.83331984e-01 -6.77024782e-01 -1.38543177e+00 -8.11977834e-02 -4.93823774e-02 5.20450771e-01 4.91116166e-01 1.13583195e+00 5.81102610e-01 7.66890287e-01 4.16283488e-01 -5.03034830e-01 -7.26537466e-01 -1.18997753e+00 -1.91697687e-01 3.38490129e-01 -9.52492282e-02 -3.70168716e-01 -2.75466233e-01 1.36273742e-01]

[10.973440170288086, 8.776704788208008]

d003363e-1bbf-4b43-a6f4-4b70133230f9

a-black-box-approach-for-non-stationary-multi

2306.07465

https://arxiv.org/abs/2306.07465v1

https://arxiv.org/pdf/2306.07465v1.pdf

A Black-box Approach for Non-stationary Multi-agent Reinforcement Learning

We investigate learning the equilibria in non-stationary multi-agent systems and address the challenges that differentiate multi-agent learning from single-agent learning. Specifically, we focus on games with bandit feedback, where testing an equilibrium can result in substantial regret even when the gap to be tested is small, and the existence of multiple optimal solutions (equilibria) in stationary games poses extra challenges. To overcome these obstacles, we propose a versatile black-box approach applicable to a broad spectrum of problems, such as general-sum games, potential games, and Markov games, when equipped with appropriate learning and testing oracles for stationary environments. Our algorithms can achieve $\widetilde{O}\left(\Delta^{1/4}T^{3/4}\right)$ regret when the degree of nonstationarity, as measured by total variation $\Delta$, is known, and $\widetilde{O}\left(\Delta^{1/5}T^{4/5}\right)$ regret when $\Delta$ is unknown, where $T$ is the number of rounds. Meanwhile, our algorithm inherits the favorable dependence on number of agents from the oracles. As a side contribution that may be independent of interest, we show how to test for various types of equilibria by a black-box reduction to single-agent learning, which includes Nash equilibria, correlated equilibria, and coarse correlated equilibria.

['Simon S. Du', 'Maryam Fazel', 'Zhihan Xiong', 'Qiwen Cui', 'Haozhe Jiang']

2023-06-12

['multi-agent-reinforcement-learning']

['methodology']

[-1.27858311e-01 1.14423648e-01 -1.50795847e-01 2.74704754e-01 -1.13436699e+00 -9.01830494e-01 -2.50759840e-01 2.95889433e-02 -7.19365716e-01 1.23142815e+00 -4.16005820e-01 -5.94297290e-01 -9.86683846e-01 -8.67292225e-01 -9.29552555e-01 -1.07607222e+00 -6.35652959e-01 5.36180377e-01 1.25620496e-02 -2.74674088e-01 -4.83304672e-02 -9.28095505e-02 -1.17935205e+00 -4.27564561e-01 1.03535664e+00 1.28315306e+00 6.98118191e-03 8.30309331e-01 2.27756783e-01 8.21129799e-01 -5.94577193e-01 -5.44187069e-01 6.49331093e-01 -6.66928828e-01 -4.74536806e-01 -3.49979922e-02 -3.78441632e-01 -4.12877768e-01 -2.37024933e-01 1.62739420e+00 5.58827400e-01 1.26213744e-01 3.83454442e-01 -1.27853477e+00 -1.91834196e-01 7.05043018e-01 -7.03838468e-01 2.76415527e-01 1.47137478e-01 3.41638476e-01 1.26845050e+00 3.82833965e-02 1.96452096e-01 8.93964827e-01 5.80840051e-01 4.75627810e-01 -9.92513061e-01 -7.97451317e-01 5.24177730e-01 -1.00300312e-01 -1.10514104e+00 -1.00573771e-01 2.93057323e-01 -3.01511765e-01 4.47678119e-01 4.83643621e-01 6.28072083e-01 5.99229634e-01 1.16638906e-01 8.37252796e-01 1.25703287e+00 -3.52637947e-01 6.15161717e-01 -1.02216460e-01 -1.06919751e-01 6.71172798e-01 5.47119915e-01 2.88957745e-01 -1.63911968e-01 -1.61379948e-01 7.96066225e-01 -3.41045447e-02 -2.63745010e-01 -1.32490575e-01 -6.89605415e-01 9.96043026e-01 -1.71002477e-01 -1.91321939e-01 -4.72580582e-01 1.64761513e-01 7.71101937e-02 8.23180437e-01 3.37084889e-01 4.97761756e-01 -4.99749929e-01 -4.50478107e-01 -5.49312413e-01 2.11583570e-01 9.46483731e-01 1.00220239e+00 9.17549312e-01 1.48490056e-01 -8.94145072e-02 4.59029287e-01 8.37794617e-02 8.83408129e-01 1.21012300e-01 -1.28377950e+00 8.25218976e-01 3.30259472e-01 7.03515828e-01 -4.64482993e-01 -3.47588599e-01 -7.17221439e-01 -9.04472768e-01 1.47091821e-01 6.76570654e-01 -1.05884469e+00 -4.22522247e-01 2.11471677e+00 2.38296866e-01 6.40516654e-02 2.82722950e-01 8.26001048e-01 1.98323205e-01 5.10268331e-01 -6.70408070e-01 -8.95955861e-01 9.27492738e-01 -6.81024134e-01 -2.82381266e-01 -4.73097324e-01 4.30274397e-01 -4.02347088e-01 9.54920411e-01 1.93172917e-01 -1.45692301e+00 2.65686035e-01 -8.36543322e-01 9.19031501e-01 1.36483014e-01 -4.93994594e-01 5.60057461e-01 9.73618984e-01 -8.99194241e-01 2.91667998e-01 -8.37637722e-01 1.84749007e-01 1.64817318e-01 7.14284539e-01 1.02140643e-01 -6.86455518e-02 -1.05844831e+00 3.95092309e-01 1.45904869e-01 8.11460838e-02 -1.04816759e+00 -5.77859342e-01 -6.03539765e-01 1.86048433e-01 1.29406929e+00 -4.67184156e-01 1.45796514e+00 -9.17016327e-01 -1.66734123e+00 2.01712534e-01 9.53236148e-02 -4.08775896e-01 9.07715857e-01 1.20803908e-01 -4.65556420e-02 -2.15766001e-02 3.27370614e-01 -2.90571690e-01 3.89375210e-01 -8.06636631e-01 -8.20974767e-01 -4.18513328e-01 7.35780895e-01 5.14648795e-01 -2.75743324e-02 -1.92343965e-01 -1.41939968e-01 -2.00007781e-01 -1.16034389e-01 -1.13421655e+00 -5.97461283e-01 -7.13727474e-01 -2.46045679e-01 1.89675108e-01 2.17169479e-01 -1.90900028e-01 1.05175924e+00 -1.98633313e+00 -6.75859908e-03 5.13471007e-01 6.33692592e-02 1.45091151e-03 -7.18302652e-02 5.13286710e-01 3.93870384e-01 2.72629261e-01 2.58365348e-02 8.76048487e-03 2.16376200e-01 2.31968001e-01 7.84498528e-02 4.39759165e-01 -3.99459958e-01 6.55580580e-01 -8.27508688e-01 -1.23259155e-02 -2.20271960e-01 -3.90190154e-01 -6.74493611e-01 6.99187517e-02 -3.38356376e-01 3.25620055e-01 -8.14776659e-01 4.48255390e-01 5.55776238e-01 -5.66729665e-01 3.70758295e-01 7.13021517e-01 -1.99154496e-01 -1.95213966e-02 -1.87064755e+00 9.78027523e-01 -2.47684628e-01 7.23815709e-02 5.37124038e-01 -1.26324689e+00 2.69618690e-01 3.33704025e-01 6.82647407e-01 -7.23532259e-01 1.54864699e-01 2.10627586e-01 2.60968089e-01 -3.25851917e-01 1.11778669e-01 -4.12275404e-01 -2.27213323e-01 8.66229892e-01 -1.46549150e-01 1.63390726e-01 4.48984504e-01 -5.20993061e-02 1.42642915e+00 -4.66694295e-01 1.73743829e-01 -2.69451052e-01 1.30266920e-01 -2.38596633e-01 9.39458191e-01 1.48027527e+00 -2.23611861e-01 1.28500372e-01 1.29884672e+00 -1.03417657e-01 -7.78809905e-01 -9.49350536e-01 4.30964082e-01 1.11018658e+00 4.80376989e-01 -1.29286721e-01 -5.83384514e-01 -3.97835642e-01 -6.02933913e-02 2.45017231e-01 -6.42994225e-01 1.31525453e-02 -8.25915486e-02 -1.19072151e+00 3.05786222e-01 1.43354908e-01 7.35842109e-01 -7.38090515e-01 -6.32106900e-01 2.13193357e-01 -2.07020536e-01 -8.31425250e-01 -6.27789676e-01 4.01335776e-01 -4.90078896e-01 -1.20462048e+00 -6.57631755e-01 -2.22867966e-01 5.78181922e-01 3.19948763e-01 7.71254659e-01 -1.43342897e-01 1.77871004e-01 6.66789412e-01 -1.46259233e-01 -7.20318377e-01 -1.32983923e-01 8.62237811e-02 1.76245585e-01 2.46485267e-02 -1.52195647e-01 -3.82761031e-01 -7.81200349e-01 3.74301940e-01 -8.82832825e-01 -2.91735321e-01 6.16828620e-01 9.39965427e-01 5.28144360e-01 3.18106979e-01 6.09267652e-01 -9.14858460e-01 7.92841554e-01 -3.66275012e-01 -1.26256394e+00 3.76006216e-01 -5.28979421e-01 1.09004043e-01 6.57954037e-01 -5.32824874e-01 -6.81305110e-01 -2.71111995e-01 2.03934342e-01 -1.49157628e-01 3.50755543e-01 8.36148739e-01 -2.81618033e-02 -2.51594819e-02 4.29696262e-01 3.74606162e-01 2.09691584e-01 3.72450985e-03 6.54736310e-02 3.95259589e-01 1.38955072e-01 -7.18877852e-01 6.42054141e-01 2.05503911e-01 1.36438221e-01 -5.25770843e-01 -8.85473311e-01 -1.37045234e-01 8.52924064e-02 -2.03787670e-01 3.04407090e-01 -9.54171181e-01 -1.64358568e+00 3.70353222e-01 -4.84540284e-01 -6.99509621e-01 -4.05251056e-01 6.94401324e-01 -8.28770936e-01 2.40212113e-01 -4.98540789e-01 -1.47860658e+00 -1.36302382e-01 -1.13282800e+00 2.77025551e-01 5.86910963e-01 3.62166226e-01 -8.87698114e-01 2.81786527e-02 3.64615262e-01 3.46480250e-01 2.13740766e-02 7.78550625e-01 -6.20028615e-01 -9.46749628e-01 -2.45773271e-01 5.74480444e-02 2.77390718e-01 8.83388892e-02 -4.18441713e-01 -2.26681948e-01 -8.14726591e-01 1.18128983e-02 -3.58220160e-01 4.71697927e-01 6.95105016e-01 7.24772811e-01 -9.28845465e-01 2.55831759e-02 3.07471991e-01 1.37522411e+00 6.53871894e-01 3.58080953e-01 4.25277412e-01 -1.44030005e-01 -8.20039064e-02 3.41858029e-01 9.82064605e-01 4.66335952e-01 2.34032869e-01 7.38233685e-01 1.86023295e-01 8.77002418e-01 7.55041167e-02 5.60752034e-01 3.53042245e-01 -1.64745286e-01 -3.20663095e-01 -5.66558659e-01 4.66872871e-01 -1.95198715e+00 -1.19578135e+00 4.30025667e-01 2.75513887e+00 9.50355232e-01 3.59086186e-01 4.60919350e-01 -1.28817514e-01 7.21416652e-01 7.71974325e-02 -1.01316917e+00 -2.21727803e-01 -2.30535865e-01 3.43627989e-01 7.20647991e-01 5.99960685e-01 -7.47914314e-01 6.62961245e-01 5.55233288e+00 8.07064354e-01 -1.06614542e+00 1.71430051e-01 6.23394549e-01 -5.03114223e-01 -1.00757249e-01 9.30107981e-02 -4.78672683e-01 6.46946788e-01 7.52145290e-01 -5.00377357e-01 8.94099176e-01 7.40107954e-01 2.50618726e-01 -5.73594809e-01 -9.34773505e-01 8.16725373e-01 -2.68533230e-01 -1.16242158e+00 -7.06584215e-01 4.36978012e-01 1.15787208e+00 -7.08875619e-03 2.43969008e-01 4.08120841e-01 1.03101146e+00 -8.04153800e-01 5.72835803e-01 8.18190202e-02 6.09551430e-01 -8.65083873e-01 9.65146065e-01 7.97745109e-01 -1.03032947e+00 -6.05024338e-01 -2.13567972e-01 -5.15953958e-01 -1.78127110e-01 3.58160168e-01 -3.97803754e-01 8.15487802e-01 7.21037388e-01 -8.45893249e-02 2.18932688e-01 1.28413332e+00 -1.28122903e-02 4.10379469e-01 -6.07615113e-01 -2.84554183e-01 3.78934085e-01 -5.64130604e-01 5.26739120e-01 5.91033757e-01 5.19784868e-01 4.92539853e-01 5.87051332e-01 4.93854612e-01 -1.73918337e-01 3.96849029e-02 -2.32098147e-01 8.00114870e-03 5.06622314e-01 8.43444169e-01 -6.92322612e-01 -1.06456034e-01 -3.85802180e-01 5.81694543e-01 1.21845067e-01 5.97015560e-01 -8.45743954e-01 -4.18170422e-01 8.57451379e-01 -1.93579659e-01 4.88505393e-01 -1.24840185e-01 -4.93476316e-02 -1.23141360e+00 3.29310238e-01 -8.27615142e-01 7.42309391e-01 -1.90532625e-01 -1.12128115e+00 1.21243089e-01 -1.92269310e-01 -1.07246745e+00 -5.97049415e-01 -2.73125768e-01 -4.64777559e-01 5.73485434e-01 -1.10836399e+00 -3.32067817e-01 1.38341174e-01 8.50514889e-01 4.34523858e-02 -2.59911239e-01 6.39801979e-01 1.00927994e-01 -7.31303096e-01 7.57918000e-01 8.90526652e-01 2.36346632e-01 7.78467357e-02 -1.24109054e+00 -4.72932458e-01 8.26060832e-01 -1.04778685e-01 1.66910872e-01 7.21887767e-01 -2.76478112e-01 -1.54093754e+00 -8.54954004e-01 1.80566832e-02 8.23489651e-02 8.18546295e-01 -1.55192792e-01 -3.78235519e-01 7.52080023e-01 -2.58519743e-02 8.50715414e-02 5.35083890e-01 1.84926108e-01 -1.63331872e-03 -4.21518385e-01 -1.06407905e+00 7.38186419e-01 8.55691135e-01 -2.92774826e-01 1.31999657e-01 2.57880449e-01 5.15961528e-01 -7.44386673e-01 -7.19087541e-01 1.89975142e-01 3.85453969e-01 -1.00140953e+00 4.20396179e-01 -7.05573320e-01 -1.05849351e-03 4.56767194e-02 -2.20002607e-01 -1.40481997e+00 -7.20146671e-02 -1.40687525e+00 4.02448624e-02 7.39218235e-01 8.04145217e-01 -9.89787340e-01 1.08560300e+00 6.72806501e-01 1.36247605e-01 -4.83132184e-01 -1.56095982e+00 -1.03214431e+00 3.76131952e-01 -3.84711802e-01 3.65072519e-01 5.67013383e-01 8.98805410e-02 9.43499245e-03 -6.46967530e-01 4.44094062e-01 6.35796785e-01 2.49261960e-01 6.54327393e-01 -9.95573223e-01 -9.20798421e-01 -4.58134413e-01 5.89650199e-02 -1.18064880e+00 -1.57344297e-01 -2.36734435e-01 5.67588694e-02 -1.11316264e+00 5.36863804e-01 -6.48048997e-01 -5.74529052e-01 3.59252661e-01 -4.59732376e-02 -1.03398442e-01 4.11543190e-01 -5.83207747e-03 -1.24937606e+00 4.48341161e-01 1.11283326e+00 -2.35565081e-01 -4.23077255e-01 7.05886602e-01 -1.07353199e+00 5.02300858e-01 6.58594131e-01 -4.60432321e-01 -4.51452732e-01 -3.84486943e-01 6.38303995e-01 9.13263798e-01 1.40074760e-01 -7.91644037e-01 4.29508150e-01 -7.49462724e-01 -2.97159642e-01 -7.58082122e-02 2.89765477e-01 -4.12557900e-01 3.13747823e-01 6.54607534e-01 -2.72862732e-01 1.58024266e-01 -5.10446243e-02 8.02192867e-01 6.14707023e-02 -4.35184479e-01 5.97571790e-01 -4.52673048e-01 -3.21658403e-02 4.44215119e-01 -6.01288140e-01 5.95342398e-01 1.19068038e+00 -8.27249587e-02 -4.98471141e-01 -1.21361661e+00 -7.24903643e-01 6.96696520e-01 1.13265388e-01 -2.15472370e-01 1.11059830e-01 -8.97763550e-01 -5.92809319e-01 -4.72806469e-02 -2.89040476e-01 6.80327341e-02 5.59490025e-01 9.91978705e-01 -6.79486245e-02 2.97791183e-01 1.31059825e-01 -3.86584967e-01 -8.62801015e-01 3.44168693e-01 7.59145916e-01 -4.93630588e-01 -2.64696982e-02 7.09705055e-01 3.51993218e-02 1.43238110e-02 2.72052109e-01 -2.15947807e-01 5.39789259e-01 -8.98504779e-02 3.48139167e-01 4.97588187e-01 -2.70895567e-02 4.19912674e-02 -8.71773511e-02 2.30931312e-01 -1.26534894e-01 -4.51111704e-01 1.35322833e+00 -3.74683380e-01 -3.46589349e-02 3.28717142e-01 5.30261695e-01 4.66979146e-02 -1.63739729e+00 -4.73700494e-01 -9.92193446e-02 -1.81424379e-01 -3.34370703e-01 -9.32013035e-01 -1.13429904e+00 4.27195638e-01 4.55277145e-01 7.96479940e-01 9.77676988e-01 1.20736472e-02 3.14174503e-01 5.16960442e-01 7.72500515e-01 -1.21766567e+00 1.44504786e-01 8.36620867e-01 3.27153683e-01 -1.29921222e+00 -2.52274722e-01 1.56255830e-02 -6.31033719e-01 6.21946692e-01 5.38956523e-01 -8.35973956e-03 6.69025302e-01 3.06683302e-01 -1.81621805e-01 9.02610272e-02 -9.46686387e-01 -5.27122915e-01 -2.71351159e-01 3.26112241e-01 -1.39658019e-01 3.68016779e-01 -2.80090481e-01 1.01368546e+00 -2.13097513e-01 -1.65433079e-01 9.00243163e-01 8.76263678e-01 -4.45635617e-01 -9.60573018e-01 -3.50457668e-01 6.77102864e-01 -8.50445151e-01 -5.47416098e-02 2.96849310e-02 9.54412103e-01 -1.03625499e-01 1.21788871e+00 -6.53449297e-02 2.86349673e-02 2.78709590e-01 -2.14605495e-01 4.08474743e-01 -2.85502255e-01 -2.87484974e-01 3.87191415e-01 -4.71544266e-02 -2.76028574e-01 -3.29623312e-01 -6.96231961e-01 -1.00689387e+00 -5.09229839e-01 -3.43645364e-01 6.34752691e-01 3.55570577e-02 1.31125629e+00 2.39186615e-01 1.66556403e-01 8.45942736e-01 -2.28268027e-01 -1.09748816e+00 -7.73278892e-01 -9.74978387e-01 -6.85921386e-02 5.26993036e-01 -5.79663157e-01 -4.53372329e-01 -6.21783495e-01]

[4.347800254821777, 3.131152629852295]

20a202f3-a1bb-4076-ab97-a5a9c7ece3f2

siamese-detr

2303.18144

https://arxiv.org/abs/2303.18144v1

https://arxiv.org/pdf/2303.18144v1.pdf

Siamese DETR

Recent self-supervised methods are mainly designed for representation learning with the base model, e.g., ResNets or ViTs. They cannot be easily transferred to DETR, with task-specific Transformer modules. In this work, we present Siamese DETR, a Siamese self-supervised pretraining approach for the Transformer architecture in DETR. We consider learning view-invariant and detection-oriented representations simultaneously through two complementary tasks, i.e., localization and discrimination, in a novel multi-view learning framework. Two self-supervised pretext tasks are designed: (i) Multi-View Region Detection aims at learning to localize regions-of-interest between augmented views of the input, and (ii) Multi-View Semantic Discrimination attempts to improve object-level discrimination for each region. The proposed Siamese DETR achieves state-of-the-art transfer performance on COCO and PASCAL VOC detection using different DETR variants in all setups. Code is available at https://github.com/Zx55/SiameseDETR.

['Lu Sheng', 'Chen Change Loy', 'Jing Shao', 'Kun Wang', 'Jianing Teng', 'Wei Li', 'Gengshi Huang', 'Zeren Chen']

2023-03-31

http://openaccess.thecvf.com//content/CVPR2023/html/Huang_Siamese_DETR_CVPR_2023_paper.html

http://openaccess.thecvf.com//content/CVPR2023/papers/Huang_Siamese_DETR_CVPR_2023_paper.pdf

cvpr-2023-1

['multi-view-learning']

['computer-vision']

[ 8.76246989e-02 -2.07389399e-01 -2.32737452e-01 -2.38672510e-01 -1.31258273e+00 -7.61016548e-01 6.95677996e-01 -2.36965165e-01 -4.99272972e-01 3.79716754e-01 4.78275493e-02 4.70524132e-02 4.74715114e-01 -5.09425044e-01 -9.07689035e-01 -7.10572004e-01 3.36398304e-01 5.61635375e-01 4.84110385e-01 -2.74244249e-01 -1.40006952e-02 5.39019406e-01 -1.51423597e+00 9.33291256e-01 4.30044174e-01 1.13559091e+00 4.07093078e-01 7.26943076e-01 2.91277230e-01 8.98983955e-01 -4.06240702e-01 -2.10316807e-01 2.28873342e-01 -4.70162421e-01 -8.31175268e-01 -3.61943692e-02 1.04724777e+00 -1.52254969e-01 -3.48160833e-01 8.76295209e-01 6.79908514e-01 -5.55574000e-02 9.79520917e-01 -1.00969744e+00 -7.76108980e-01 4.46292579e-01 -8.64110589e-01 6.41610801e-01 1.90968364e-01 7.25080520e-02 1.16698885e+00 -1.53477609e+00 6.38037801e-01 1.18610072e+00 5.92778027e-01 6.91129506e-01 -1.49728942e+00 -5.79741180e-01 1.27891049e-01 3.77368301e-01 -1.32364500e+00 -5.21853268e-01 7.21848547e-01 -3.65306616e-01 1.10904121e+00 1.76159471e-01 4.88951176e-01 1.55617988e+00 2.51229793e-01 1.47166753e+00 1.50816154e+00 -2.38476172e-01 -7.45299384e-02 3.74546856e-01 -9.20216814e-02 6.84638441e-01 -6.13460988e-02 1.53375357e-01 -6.17160022e-01 3.36451977e-01 6.00953996e-01 1.92392945e-01 -1.05569869e-01 -7.90362835e-01 -1.25181007e+00 7.63173819e-01 8.63468170e-01 3.43854398e-01 -9.32924375e-02 5.57464734e-02 6.33429885e-01 2.59284616e-01 7.07218468e-01 1.31782070e-01 -5.64055920e-01 4.56771046e-01 -1.01121557e+00 -1.15716055e-01 3.07749987e-01 9.94263113e-01 7.74604142e-01 1.74110323e-01 -3.13973457e-01 8.91161382e-01 1.53961837e-01 5.69288790e-01 7.11288691e-01 -4.53034878e-01 6.66719079e-01 4.86493140e-01 -3.36535811e-01 -3.04399133e-01 -2.77180225e-01 -9.35580671e-01 -8.89196336e-01 4.15442109e-01 3.94353896e-01 6.15322515e-02 -1.18684864e+00 1.67816758e+00 2.17164040e-01 -4.50490275e-03 2.45924383e-01 8.02154958e-01 1.16894591e+00 3.87699664e-01 4.40432839e-02 3.25382531e-01 1.50268936e+00 -1.61353016e+00 -2.06578702e-01 -6.20187998e-01 5.27471662e-01 -8.22197855e-01 1.08657897e+00 2.41079181e-01 -1.24336588e+00 -9.03312206e-01 -8.64536285e-01 -4.51063544e-01 -6.96394026e-01 7.21533895e-01 2.85987258e-01 2.33891964e-01 -1.06360066e+00 3.20528388e-01 -8.19384217e-01 -3.71547401e-01 7.24948823e-01 -1.25760166e-02 -6.66447699e-01 -2.25081638e-01 -9.37209845e-01 8.30927014e-01 2.58558899e-01 1.63176451e-02 -1.55671263e+00 -7.64507234e-01 -1.07132781e+00 7.73289502e-02 2.71262914e-01 -8.54106367e-01 1.09189045e+00 -1.12216353e+00 -1.00413430e+00 1.64845061e+00 -6.73710555e-02 -5.70712566e-01 8.07775736e-01 -3.67857724e-01 -3.20912004e-01 2.97783703e-01 5.29074371e-01 7.11686373e-01 1.16444743e+00 -1.42139649e+00 -6.77968025e-01 -6.96587443e-01 -1.06462613e-02 3.34774494e-01 -1.25143096e-01 4.14366573e-02 -3.87137800e-01 -8.18153799e-01 -1.95889324e-02 -8.22713435e-01 5.56799173e-02 8.01301971e-02 -5.11098742e-01 -2.33349428e-01 9.61645484e-01 -6.33919835e-01 4.51618701e-01 -2.24369884e+00 3.19917679e-01 -3.06413680e-01 2.90490240e-01 2.45956853e-01 -3.52085084e-01 2.77715147e-01 -2.97813088e-01 -3.83111000e-01 -1.77264795e-01 -7.02928722e-01 -2.03135610e-01 -1.61323637e-01 -2.83562332e-01 8.67023468e-01 2.31208950e-01 1.04790580e+00 -7.88727105e-01 -5.90201437e-01 4.07251745e-01 4.77814436e-01 -3.80731344e-01 5.10921538e-01 6.24894798e-02 4.18099821e-01 -2.00382143e-01 7.20809579e-01 7.13417292e-01 -3.70463938e-01 -1.22169271e-01 -5.87220252e-01 -8.13300386e-02 6.71102330e-02 -8.32206428e-01 2.06358242e+00 -8.86806726e-01 6.32461309e-01 2.45917439e-01 -1.15917456e+00 5.76368332e-01 7.04963952e-02 1.99337915e-01 -7.15219021e-01 5.33796959e-02 2.16683358e-01 -2.70170182e-01 -1.42586619e-01 2.86008656e-01 -9.23018157e-03 -5.34344241e-02 3.15200448e-01 8.06464493e-01 2.04419699e-02 1.70789972e-01 1.82027921e-01 7.64312685e-01 4.73464370e-01 6.12596810e-01 -2.35082567e-01 6.85903132e-01 -9.92951170e-02 2.73756564e-01 6.33690476e-01 -1.90234542e-01 1.13000774e+00 1.92119986e-01 -2.58727908e-01 -6.92723691e-01 -1.50140178e+00 -2.57282048e-01 1.53028369e+00 1.39132410e-01 -8.93685371e-02 -5.61126769e-01 -1.11357248e+00 1.32320717e-01 5.15442669e-01 -9.75519776e-01 -1.58608377e-01 -4.45291340e-01 -5.41329682e-01 4.23306286e-01 1.02655101e+00 6.14807725e-01 -9.07254398e-01 -6.60337448e-01 -1.50750950e-01 -6.72362447e-02 -1.26027203e+00 -7.12855875e-01 6.64707839e-01 -7.18981087e-01 -1.12738276e+00 -1.11963344e+00 -9.99926209e-01 6.76368475e-01 7.10162342e-01 1.25296199e+00 -4.32489842e-01 -4.02828097e-01 6.06935501e-01 -1.98406279e-01 -3.20496738e-01 -1.25479728e-01 2.26251960e-01 -2.12134838e-01 1.58121690e-01 1.84241101e-01 -4.67064947e-01 -9.48597610e-01 6.55630112e-01 -7.26853073e-01 1.16718344e-01 9.73873258e-01 1.10686553e+00 9.64758158e-01 -7.24297643e-01 3.16208750e-01 -1.05578220e+00 -2.83977925e-03 -3.67806166e-01 -2.51953840e-01 4.84082311e-01 -2.80857086e-01 -1.24919429e-01 8.52453768e-01 -2.79181451e-01 -1.07527471e+00 2.95699894e-01 -1.85509622e-01 -8.32451820e-01 -4.67676252e-01 -9.00214463e-02 -2.11802781e-01 -1.88582703e-01 9.80228186e-01 5.92732787e-01 -1.54987693e-01 -4.85830963e-01 7.40368307e-01 5.66293776e-01 6.29143596e-01 -2.85637319e-01 6.61149323e-01 7.51788497e-01 -4.03337002e-01 -5.56815624e-01 -1.27786684e+00 -8.90624285e-01 -9.44771290e-01 -1.61031976e-01 1.12337565e+00 -1.53620780e+00 -1.72014624e-01 3.72491658e-01 -8.48960817e-01 -4.33946043e-01 -5.95949948e-01 3.83313388e-01 -9.49507296e-01 3.61129083e-02 -5.71992457e-01 -4.53565717e-01 -4.67859387e-01 -1.06504703e+00 1.57041907e+00 1.61913425e-01 2.01426446e-01 -1.00109041e+00 3.50806043e-02 7.79137731e-01 3.58628064e-01 1.62076965e-01 3.13715667e-01 -7.63603985e-01 -3.02055508e-01 -1.33579880e-01 -5.10964453e-01 6.91295326e-01 -7.72116706e-03 -4.17397916e-01 -1.69823539e+00 -8.62657130e-01 -1.81956589e-01 -9.98321831e-01 1.47730374e+00 2.66772419e-01 1.30160618e+00 1.18288631e-03 -4.31074411e-01 1.03693867e+00 1.45818019e+00 -2.28474110e-01 4.07390177e-01 9.27365869e-02 8.73814225e-01 3.39309573e-01 5.39492249e-01 1.65719148e-02 3.93596351e-01 8.42228830e-01 5.39316535e-01 -6.18472576e-01 -6.94824457e-01 -2.92293966e-01 5.30016065e-01 4.59355682e-01 -9.48457420e-02 -1.60030141e-01 -7.93282747e-01 7.44270563e-01 -1.66291153e+00 -9.02684808e-01 2.34782487e-01 1.94921100e+00 4.23205256e-01 1.01086095e-01 3.24311733e-01 -2.02166826e-01 6.42529011e-01 4.26318854e-01 -7.62488246e-01 -7.34926239e-02 -2.74634898e-01 2.13245019e-01 5.62547505e-01 5.33712022e-02 -1.57792199e+00 1.09147966e+00 5.07064581e+00 1.06710303e+00 -1.25371933e+00 5.70722401e-01 8.50872099e-01 -9.28619280e-02 9.31700245e-02 -4.04964566e-01 -9.15317357e-01 -1.27092097e-02 5.64942062e-01 1.97670355e-01 5.62381335e-02 1.23806322e+00 -3.15854102e-01 1.30262643e-01 -1.21354115e+00 1.02892470e+00 5.69117963e-01 -1.27128172e+00 2.54287478e-02 -2.33282849e-01 8.21006536e-01 4.70202684e-01 3.95722777e-01 5.66574097e-01 8.38711560e-02 -8.76466990e-01 9.04764771e-01 1.42747775e-01 1.08269429e+00 -4.40499663e-01 5.64506590e-01 1.78521588e-01 -1.46745062e+00 -1.18843600e-01 -3.51065695e-01 5.64173937e-01 -1.20691784e-01 1.56210914e-01 -4.49726194e-01 6.61683261e-01 9.18686330e-01 1.26753879e+00 -8.54908943e-01 7.75282204e-01 -3.63086283e-01 4.32029754e-01 3.47203277e-02 4.30858225e-01 2.26562530e-01 1.20411187e-01 5.14045119e-01 1.54698062e+00 -4.70808856e-02 -6.11975074e-01 1.56045005e-01 9.57882524e-01 -2.19312280e-01 2.08829213e-02 -6.95968747e-01 3.08463454e-01 -3.84422019e-02 1.54164505e+00 -7.33237624e-01 -3.56098711e-01 -6.50913835e-01 1.38348138e+00 7.06687450e-01 4.57727909e-01 -7.50533044e-01 -1.85231641e-01 4.76128787e-01 2.88583785e-01 6.64486885e-01 1.64628886e-02 -6.75118789e-02 -1.49568057e+00 -1.71992406e-01 -9.06385720e-01 7.83370554e-01 -7.04665899e-01 -1.47867596e+00 7.21061051e-01 -2.01846994e-02 -1.52078307e+00 -2.10519925e-01 -9.00189519e-01 -7.91699171e-01 7.61203527e-01 -1.82888579e+00 -1.71251309e+00 -3.42225075e-01 8.71151388e-01 8.92759144e-01 -4.84936953e-01 7.23215520e-01 3.14661145e-01 -5.44745564e-01 9.13262963e-01 2.50981778e-01 2.66512185e-01 1.04442406e+00 -1.57787645e+00 4.07603264e-01 9.51320887e-01 5.77610373e-01 3.10291439e-01 2.18909323e-01 -1.32155895e-01 -1.33148253e+00 -1.46468842e+00 3.43603730e-01 -5.18972516e-01 5.55076897e-01 -8.26363206e-01 -7.00204134e-01 7.86532998e-01 3.35121542e-01 8.11800659e-01 5.41241586e-01 -3.14036533e-02 -8.40852976e-01 -3.39209259e-01 -1.10340822e+00 1.96017459e-01 1.13583505e+00 -9.52522933e-01 -4.40392524e-01 5.56899548e-01 3.78582716e-01 -6.48070455e-01 -6.86991692e-01 3.33108038e-01 5.23670018e-01 -1.10660934e+00 1.44446707e+00 -8.44304562e-01 4.25771445e-01 -2.07741305e-01 -3.12720805e-01 -1.37074471e+00 -3.47057253e-01 -4.72249463e-02 -2.05000594e-01 1.09084189e+00 4.40497249e-01 -5.49229205e-01 7.43734002e-01 -5.03957927e-01 -2.32091084e-01 -8.86064529e-01 -1.01966763e+00 -8.02447677e-01 1.16889074e-01 -1.24778785e-01 -1.45209908e-01 8.13699245e-01 -6.01262271e-01 7.69137263e-01 -2.65231282e-01 1.20151162e-01 8.89040351e-01 5.60698986e-01 7.38237083e-01 -7.34622955e-01 -5.59691906e-01 -2.96861172e-01 -4.63552296e-01 -1.01120639e+00 2.34043181e-01 -1.22112954e+00 -1.23153239e-01 -1.62694955e+00 4.87446636e-01 -1.14134550e-01 -5.62800288e-01 4.58210140e-01 -3.64829823e-02 7.23032653e-01 2.41310015e-01 2.51019001e-01 -9.02768731e-01 6.22187436e-01 1.27050781e+00 -4.40655917e-01 1.38566479e-01 1.70247674e-01 -6.09625578e-01 6.74720228e-01 8.91896605e-01 -3.44435602e-01 -5.05251586e-01 -5.24582207e-01 -3.14424127e-01 1.89005584e-02 6.24457896e-01 -1.20825076e+00 2.89809760e-02 2.87434161e-01 8.06912422e-01 -7.66922414e-01 5.14929116e-01 -5.74543834e-01 -3.55614960e-01 4.37064797e-01 -4.65847790e-01 9.81515199e-02 3.50249857e-01 7.69888639e-01 -3.89551193e-01 1.31256923e-01 1.27760267e+00 -3.74744803e-01 -9.23151791e-01 2.75868684e-01 -7.58317634e-02 4.90331501e-01 1.11265135e+00 -1.52470186e-01 -4.68044192e-01 -1.25096038e-01 -1.01075280e+00 1.54271141e-01 2.45181754e-01 5.15742898e-01 7.51914203e-01 -1.16688204e+00 -8.97466958e-01 1.80019543e-01 6.05624080e-01 -7.42813349e-02 5.14281034e-01 9.28099513e-01 -2.69018531e-01 3.20829928e-01 -4.50800449e-01 -8.16194236e-01 -1.50178885e+00 6.53017759e-01 6.41531348e-01 -4.39069301e-01 -6.79328024e-01 1.21533918e+00 9.84977484e-01 -3.76400471e-01 2.61444300e-01 -1.89946771e-01 -2.87075490e-01 2.45067477e-01 5.52267432e-01 1.42613336e-01 1.61798552e-01 -9.45406020e-01 -5.02148032e-01 7.40603685e-01 -3.37929726e-01 1.78183928e-01 1.29438472e+00 -5.84702007e-02 1.58909783e-01 4.77154255e-01 1.54480672e+00 -1.97996736e-01 -1.23415959e+00 -4.12785381e-01 -4.38195944e-01 -2.81030744e-01 1.31638989e-01 -7.66913295e-01 -1.30795383e+00 1.25882769e+00 1.11932635e+00 -3.13072592e-01 1.10332072e+00 3.84374082e-01 5.90317369e-01 2.09885538e-01 2.33237475e-01 -9.28527951e-01 3.40145499e-01 3.24900210e-01 1.14150786e+00 -1.53767788e+00 -6.95123896e-02 -2.50515670e-01 -7.96906948e-01 8.87296855e-01 9.76087987e-01 -2.71704853e-01 5.04229367e-01 1.36719331e-01 7.68629536e-02 -2.12257311e-01 -6.69557512e-01 -7.18085229e-01 6.90756679e-01 7.35411882e-01 6.11780405e-01 4.09173444e-02 2.19883993e-01 3.03528607e-01 2.57913530e-01 -5.23007154e-01 1.57567188e-01 7.10479736e-01 -2.51213282e-01 -6.90334141e-01 -2.28612259e-01 4.69771266e-01 -3.65159363e-01 -7.23777488e-02 -4.76806968e-01 7.64901400e-01 7.64786303e-02 5.44239402e-01 3.12960036e-02 -3.02890301e-01 5.30424774e-01 -8.47000480e-02 6.01549149e-01 -8.63227785e-01 -7.14720666e-01 1.11262336e-01 -7.38364458e-02 -7.13853180e-01 -4.32072908e-01 -7.61066973e-01 -7.96648860e-01 3.04742187e-01 -1.71822347e-02 -8.61306489e-02 3.07874918e-01 5.84541678e-01 2.47628853e-01 5.85295498e-01 5.45924306e-01 -1.09567785e+00 -6.28187478e-01 -8.76100779e-01 -5.07605195e-01 4.46365982e-01 5.33149004e-01 -7.44164348e-01 -2.07925230e-01 -1.75742075e-01]

[9.733621597290039, 1.8356924057006836]

1092ca7f-6c3c-4de3-80b1-c79c0d09492d

differentiable-instruction-optimization-for

2306.10098

https://arxiv.org/abs/2306.10098v1

https://arxiv.org/pdf/2306.10098v1.pdf

Differentiable Instruction Optimization for Cross-Task Generalization

Instruction tuning has been attracting much attention to achieve generalization ability across a wide variety of tasks. Although various types of instructions have been manually created for instruction tuning, it is still unclear what kind of instruction is optimal to obtain cross-task generalization ability. This work presents instruction optimization, which optimizes training instructions with respect to generalization ability. Rather than manually tuning instructions, we introduce learnable instructions and optimize them with gradient descent by leveraging bilevel optimization. Experimental results show that the learned instruction enhances the diversity of instructions and improves the generalization ability compared to using only manually created instructions.

['Ichiro Sakata', 'Junichiro Mori', 'Masaru Isonuma']

2023-06-16

['bilevel-optimization']

['methodology']

[ 9.37934443e-02 -2.79513627e-01 -5.04356802e-01 -7.85759568e-01 -5.51619053e-01 -6.82619691e-01 1.48310393e-01 1.37854934e-01 -6.29394531e-01 6.60444975e-01 1.44878447e-01 -6.42348707e-01 -8.42312351e-02 -6.50215089e-01 -6.04243457e-01 -3.47523540e-01 -1.01136580e-01 2.46990934e-01 1.15244314e-01 -3.42744172e-01 5.48348308e-01 2.40763679e-01 -1.77718461e+00 3.03110957e-01 1.58242381e+00 6.38041854e-01 3.49296242e-01 6.30121410e-01 -2.43603140e-01 1.41576067e-01 -7.98706174e-01 -1.93718538e-01 2.56471872e-01 -4.56360251e-01 -6.51149154e-01 -2.53851652e-01 5.68039596e-01 -3.59848678e-01 1.80314347e-01 9.00013745e-01 3.15554291e-01 4.68719333e-01 6.69678926e-01 -1.02095735e+00 -1.07289922e+00 5.43226659e-01 -1.60839766e-01 4.99138564e-01 4.45855439e-01 7.38029405e-02 1.04454839e+00 -4.84344184e-01 1.34663701e-01 1.16499782e+00 3.21614593e-01 7.46294856e-01 -1.21565235e+00 -8.48092496e-01 2.79441953e-01 9.52442214e-02 -1.18787718e+00 7.55981058e-02 5.52013099e-01 -3.77801627e-01 1.06133008e+00 1.62145078e-01 3.37716490e-01 8.81156266e-01 5.72685063e-01 9.13938284e-01 1.35308719e+00 -4.74671215e-01 9.94977802e-02 2.97479868e-01 5.90626776e-01 7.10040748e-01 3.77521634e-01 5.30355215e-01 -5.62796831e-01 1.23525910e-01 7.70575345e-01 -1.99676886e-01 -3.96900058e-01 -3.42093110e-01 -9.69170988e-01 1.11707211e+00 5.72679996e-01 2.26147741e-01 -1.55729964e-01 2.84426641e-02 3.81980836e-01 5.21430314e-01 -2.45279409e-02 1.19887877e+00 -9.43261981e-01 -4.49120939e-01 -3.83862376e-01 1.89206794e-01 7.95697987e-01 8.10124695e-01 9.81226325e-01 5.33270299e-01 -4.98759061e-01 9.28971887e-01 -6.40729219e-02 5.40766180e-01 9.46690023e-01 -7.14814067e-01 5.42835474e-01 8.85530710e-01 -1.69673130e-01 -7.96672523e-01 -4.01904821e-01 -7.47786582e-01 -2.49174714e-01 2.87051767e-01 3.10274094e-01 -3.44006836e-01 -8.91310990e-01 2.00643039e+00 2.86511728e-03 -1.07228786e-01 -1.07515909e-01 7.76065946e-01 7.11063921e-01 4.29635465e-01 1.52750403e-01 -1.20813996e-02 1.09206367e+00 -9.73076701e-01 -7.37435758e-01 -4.50226158e-01 9.67262864e-01 -5.88977277e-01 1.99563682e+00 5.40164530e-01 -7.28655279e-01 -7.16373563e-01 -1.15284741e+00 1.74905971e-01 -5.73029518e-01 -1.72431653e-04 8.27480614e-01 1.06603205e+00 -8.61778855e-01 5.01210809e-01 -7.11216092e-01 8.17638040e-02 2.22073406e-01 7.71200061e-01 -2.07834229e-01 2.04323046e-02 -1.18946135e+00 9.03203070e-01 8.49109292e-01 -5.56822419e-01 -8.52536678e-01 -7.80458152e-01 -1.06388128e+00 2.55248964e-01 4.75239038e-01 -5.22584796e-01 1.33901930e+00 -8.79728138e-01 -1.83669388e+00 2.95199543e-01 -3.57518345e-02 -2.19720706e-01 -1.36148810e-01 -5.70606112e-01 -2.51721889e-01 -5.16732514e-01 -1.19979076e-01 8.68001819e-01 7.73656487e-01 -1.07699597e+00 -6.99236870e-01 -1.37661889e-01 3.05229962e-01 4.59857047e-01 -8.45396757e-01 -3.29390734e-01 -2.51919299e-01 -9.21153128e-01 -1.62739903e-01 -1.09049249e+00 -2.52375398e-02 -7.21135199e-01 -1.04302302e-01 -2.99734980e-01 7.25514770e-01 -3.12590569e-01 1.64917433e+00 -1.86006272e+00 3.21398526e-02 1.68446779e-01 7.93451294e-02 2.81440139e-01 -4.70696747e-01 -5.02364263e-02 -4.93162572e-02 2.46364146e-01 -1.07831992e-01 1.73440114e-01 1.73867255e-01 4.52675998e-01 -1.40406907e-01 -1.41197309e-01 -7.55989775e-02 9.85952914e-01 -9.48552847e-01 1.46520250e-02 2.65748445e-02 2.04430416e-01 -1.11591434e+00 4.40805614e-01 -4.25348133e-01 5.99904001e-01 -5.91163397e-01 4.39316630e-01 2.10182846e-01 -1.13003574e-01 -2.36278340e-01 8.92115384e-02 1.45605892e-01 4.59405482e-01 -1.00595665e+00 1.40342081e+00 -8.50699961e-01 4.00046527e-01 -4.62325990e-01 -7.56827354e-01 1.07266653e+00 -9.22520161e-02 3.18645723e-02 -9.38051224e-01 2.88145214e-01 1.02128327e-01 4.47480738e-01 -6.48121297e-01 5.94631732e-01 2.87657864e-02 -2.77540624e-01 5.75006664e-01 -7.49222487e-02 -4.01272714e-01 3.51436406e-01 -2.10753173e-01 8.45741093e-01 2.32392952e-01 3.62715214e-01 -4.54975814e-01 5.69038093e-01 7.41113499e-02 3.77005845e-01 6.95980132e-01 -1.65925533e-01 1.48770913e-01 9.44635794e-02 -4.37055081e-01 -5.88656604e-01 -9.15697575e-01 -1.32566184e-01 1.84701359e+00 2.01066341e-05 -4.55824643e-01 -8.97958219e-01 -9.41625476e-01 1.46746621e-01 1.15736544e+00 -6.97593808e-01 -5.72827578e-01 -7.27594554e-01 -7.71063864e-01 2.21288174e-01 6.71774447e-01 6.18145108e-01 -1.07206762e+00 -6.17660105e-01 -4.87195365e-02 1.12997748e-01 -8.01289022e-01 -1.13640654e+00 4.43212271e-01 -1.09240305e+00 -8.60046804e-01 -9.63790417e-02 -7.57144034e-01 7.93920636e-01 3.14524770e-01 1.44304204e+00 2.22070619e-01 -1.16317146e-01 3.66223514e-01 -3.30923498e-01 -3.69388103e-01 -4.85646993e-01 5.06116092e-01 2.01992527e-01 -3.96461040e-01 6.37473464e-01 -2.69652188e-01 -4.99834776e-01 5.00529349e-01 -1.08800137e+00 -4.91007686e-01 9.56615746e-01 9.18754458e-01 4.99018341e-01 2.12865353e-01 7.88599908e-01 -9.37379241e-01 1.49623096e+00 -3.59952807e-01 -5.80704153e-01 2.74471909e-01 -1.08974528e+00 7.85629749e-01 7.24211097e-01 -7.10967600e-01 -9.87043858e-01 -2.89278645e-02 3.35644297e-02 -1.52618736e-01 -2.21349448e-01 5.63073754e-01 3.71185131e-03 -3.26214403e-01 9.54439402e-01 1.57907769e-01 -1.84698313e-01 -3.08467388e-01 4.78755325e-01 5.44660866e-01 2.35081241e-01 -1.06130981e+00 4.62749511e-01 -2.91931540e-01 -5.52833855e-01 -4.97615725e-01 -9.81349349e-01 -1.17985278e-01 -4.82750267e-01 1.55963330e-02 9.30176437e-01 -3.98234516e-01 -6.95747674e-01 -6.36651516e-02 -7.16832697e-01 -7.68562198e-01 -8.71265456e-02 7.81240106e-01 -4.06618834e-01 -9.35993865e-02 -3.16094875e-01 -1.30590171e-01 -2.85032421e-01 -1.78676820e+00 8.17856073e-01 3.82551223e-01 -6.46905124e-01 -1.27759731e+00 1.00933969e-01 1.73694313e-01 6.11940444e-01 -2.52312094e-01 1.13121498e+00 -7.52189159e-01 -5.24713814e-01 -1.74814373e-01 2.21234653e-02 3.96068364e-01 5.94117522e-01 -3.59219074e-01 -6.94172144e-01 -4.16541159e-01 5.60290925e-02 -3.36639643e-01 4.41224635e-01 4.12063986e-01 1.50748932e+00 -4.09238905e-01 -1.74064666e-01 8.54429066e-01 1.14768052e+00 4.61160630e-01 3.42677116e-01 6.74702823e-01 5.75911820e-01 2.70171314e-01 4.79511172e-01 8.49902704e-02 4.51055139e-01 7.33949780e-01 2.54711211e-01 3.15080315e-01 8.54870826e-02 -1.34687662e-01 4.23720479e-01 7.55221546e-01 1.26403630e-01 8.67552087e-02 -8.65170658e-01 2.23905072e-01 -1.25009906e+00 -4.00498748e-01 2.47307092e-01 2.49422264e+00 1.14207208e+00 2.75983155e-01 7.72087052e-02 -1.19981177e-01 2.30481863e-01 -2.13292867e-01 -8.03609967e-01 -8.80403578e-01 2.37520233e-01 2.90563166e-01 3.85384053e-01 8.21074188e-01 -8.77763689e-01 1.07503164e+00 7.93615627e+00 7.67491937e-01 -1.40705884e+00 -2.16844067e-01 4.15750831e-01 -6.66530570e-03 -3.98965210e-01 -2.20360115e-01 -1.34811032e+00 3.48441601e-01 8.83217990e-01 -2.77394950e-01 5.88076890e-01 9.19308364e-01 6.32133260e-02 -1.92732234e-02 -1.23814142e+00 6.65865362e-01 1.39098898e-01 -8.54368925e-01 3.01182747e-01 -1.79965660e-01 1.01440561e+00 -4.80864406e-01 3.52887005e-01 1.06926835e+00 5.99216104e-01 -1.24404991e+00 1.74070328e-01 1.04758121e-01 3.81660849e-01 -7.62398481e-01 5.30372202e-01 3.91322285e-01 -8.47907186e-01 -2.05141887e-01 -1.94784403e-01 -1.57191187e-01 -1.49247229e-01 3.61779064e-01 -1.05178654e+00 3.61402519e-02 5.28394163e-01 7.65926763e-02 -9.86263812e-01 8.85068476e-01 -6.66424870e-01 6.55393004e-01 -3.11798230e-02 -3.98968428e-01 2.23106429e-01 -1.84377357e-01 2.86454052e-01 1.22978044e+00 3.30217242e-01 2.07469687e-02 5.45607686e-01 7.24893153e-01 -1.16565689e-01 3.39095771e-01 -5.54207385e-01 -3.96116786e-02 4.97663707e-01 8.54685843e-01 -5.27312696e-01 -1.54177308e-01 -4.47755069e-01 7.88069069e-01 5.66609681e-01 4.17222172e-01 -7.99569130e-01 -5.27572036e-01 7.12139428e-01 -1.63850039e-02 1.63028240e-01 -5.77796400e-01 -6.42849803e-01 -9.63389933e-01 -2.12518126e-01 -1.10818875e+00 4.91943151e-01 -5.52935600e-01 -9.77547884e-01 7.18519807e-01 3.77610594e-01 -9.52047884e-01 -4.10751402e-01 -8.56801629e-01 -7.15788901e-01 8.06364298e-01 -1.19521201e+00 -3.59276295e-01 -4.88987267e-01 3.09525073e-01 7.23566294e-01 -3.76915574e-01 7.19864964e-01 1.22923844e-01 -6.43584788e-01 1.19592345e+00 -1.23961158e-01 -3.26872081e-01 8.61325383e-01 -1.47914183e+00 -1.16530024e-02 4.95245963e-01 -3.62011343e-02 1.17182863e+00 8.72235656e-01 -6.49215043e-01 -1.43176699e+00 -9.64451134e-01 4.05702502e-01 -3.88731182e-01 5.05352676e-01 -1.04796544e-01 -1.20135868e+00 7.07898676e-01 4.12032157e-01 -3.78971785e-01 9.56932008e-01 2.07559705e-01 -4.32097614e-01 -1.17797263e-01 -9.45732951e-01 8.35095644e-01 8.50481868e-01 -2.86724448e-01 -7.78068841e-01 1.71553656e-01 8.88401508e-01 -6.01735890e-01 -9.52295959e-01 4.55251575e-01 3.53275567e-01 -6.95613027e-01 1.00736141e+00 -6.90046430e-01 1.67826965e-01 -1.52182981e-01 -1.18529558e-01 -1.90144753e+00 -3.13840568e-01 -3.13919067e-01 1.13634532e-02 8.11548352e-01 6.50310755e-01 -7.42393792e-01 8.88754904e-01 6.80581570e-01 -4.35388684e-01 -1.00513840e+00 -1.43960223e-01 -1.14639044e+00 4.16756302e-01 -2.19696358e-01 8.38957071e-01 8.45736265e-01 -2.05892380e-02 7.25793183e-01 5.05429879e-02 2.91150436e-02 2.38617212e-01 2.09252745e-01 8.91143024e-01 -9.07420099e-01 -5.15263379e-01 -9.26022410e-01 -8.36146623e-02 -1.37054169e+00 3.08365196e-01 -1.01045752e+00 1.92321807e-01 -1.03136611e+00 -9.97361988e-02 -5.16005397e-01 -1.55083030e-01 5.99649906e-01 -8.54812503e-01 -1.07237816e-01 -1.10294931e-01 -1.17054492e-01 -5.27519166e-01 6.25518918e-01 1.46020353e+00 3.75359599e-03 -7.20452905e-01 1.12463079e-01 -9.92387295e-01 4.24504220e-01 1.18014002e+00 -2.93647498e-01 -8.39939058e-01 -7.31954992e-01 1.32329300e-01 -3.19283634e-01 -3.54346067e-01 -1.00810409e+00 -3.91350985e-02 -4.09635514e-01 8.26305524e-02 -2.28175595e-01 -2.24457905e-02 -5.60317397e-01 -2.83561856e-01 5.33731520e-01 -6.81936383e-01 4.83458161e-01 7.60175765e-01 4.13016140e-01 -2.33966962e-01 -5.39021313e-01 6.85851276e-01 9.76461396e-02 -9.49520409e-01 1.83150306e-01 -3.14494431e-01 1.29340261e-01 1.11071050e+00 -2.47714654e-01 -1.70182705e-01 -3.63902062e-01 -6.07345760e-01 3.67183685e-01 2.70315260e-01 7.37566948e-01 6.15492225e-01 -1.27007306e+00 -6.76635504e-01 5.05952597e-01 1.55555040e-01 -3.88667703e-01 -1.83250099e-01 3.46002817e-01 -2.41620570e-01 7.16303825e-01 -2.45253935e-01 -6.27625167e-01 -1.28988361e+00 7.55464792e-01 2.34593764e-01 -3.29958916e-01 -1.69344261e-01 8.53980839e-01 3.60496849e-01 -9.52941239e-01 4.32050467e-01 -6.44491494e-01 -3.47128183e-01 -4.15863246e-01 5.47874570e-01 2.28227809e-01 1.46393895e-01 -4.77113351e-02 -8.72137723e-04 7.01799750e-01 -2.45040596e-01 1.44064784e-01 9.55715954e-01 1.02926306e-01 3.02832782e-01 3.23431820e-01 1.40538847e+00 -4.50893771e-03 -1.15256047e+00 -2.77969167e-02 1.48435578e-01 -6.27547979e-01 2.16012839e-02 -9.31026578e-01 -1.03105104e+00 8.64900649e-01 7.30288625e-01 1.61683694e-01 1.09028387e+00 -4.71529961e-01 6.15230262e-01 8.49272609e-01 3.95741314e-01 -9.00822520e-01 5.97130001e-01 9.74666119e-01 8.87685776e-01 -1.50487757e+00 -1.00452505e-01 -3.35333318e-01 -6.62828088e-01 1.12666047e+00 1.14473271e+00 -1.59928992e-01 5.26073992e-01 3.24650884e-01 -1.31104039e-02 1.25836879e-02 -5.41134655e-01 -1.43441617e-01 6.28654957e-01 6.19955301e-01 9.32204306e-01 1.49944708e-01 -4.94393498e-01 4.65937495e-01 -7.01198697e-01 3.83572616e-02 2.49969959e-01 1.00327492e+00 -7.67692089e-01 -1.41076899e+00 -4.99776542e-01 8.43094647e-01 -2.44539738e-01 -2.41683602e-01 -1.72569990e-01 8.14134240e-01 -7.31080845e-02 7.90186405e-01 -2.00768784e-01 -6.77726686e-01 5.68714380e-01 1.96721464e-01 5.89179695e-01 -9.98439670e-01 -7.66273558e-01 -5.59832990e-01 -1.07767098e-01 -4.08828288e-01 1.98771983e-01 -2.29986146e-01 -1.29481208e+00 -1.81116924e-01 -5.09428680e-01 3.71301621e-01 6.07949555e-01 1.00472450e+00 3.22087646e-01 9.22508299e-01 4.71931905e-01 -4.91473734e-01 -8.59540462e-01 -8.43429625e-01 -3.87575431e-03 3.67178619e-01 1.20332427e-01 -9.77729738e-01 -3.57918322e-01 4.07563103e-03]

[10.652546882629395, 8.352051734924316]

945e2a78-c1dc-4afe-91a3-886511b5bcd0

background-aware-3d-point-cloud

2111.07248

https://arxiv.org/abs/2111.07248v1

https://arxiv.org/pdf/2111.07248v1.pdf

Background-Aware 3D Point Cloud Segmentationwith Dynamic Point Feature Aggregation

With the proliferation of Lidar sensors and 3D vision cameras, 3D point cloud analysis has attracted significant attention in recent years. After the success of the pioneer work PointNet, deep learning-based methods have been increasingly applied to various tasks, including 3D point cloud segmentation and 3D object classification. In this paper, we propose a novel 3D point cloud learning network, referred to as Dynamic Point Feature Aggregation Network (DPFA-Net), by selectively performing the neighborhood feature aggregation with dynamic pooling and an attention mechanism. DPFA-Net has two variants for semantic segmentation and classification of 3D point clouds. As the core module of the DPFA-Net, we propose a Feature Aggregation layer, in which features of the dynamic neighborhood of each point are aggregated via a self-attention mechanism. In contrast to other segmentation models, which aggregate features from fixed neighborhoods, our approach can aggregate features from different neighbors in different layers providing a more selective and broader view to the query points, and focusing more on the relevant features in a local neighborhood. In addition, to further improve the performance of the proposed semantic segmentation model, we present two novel approaches, namely Two-Stage BF-Net and BF-Regularization to exploit the background-foreground information. Experimental results show that the proposed DPFA-Net achieves the state-of-the-art overall accuracy score for semantic segmentation on the S3DIS dataset, and provides a consistently satisfactory performance across different tasks of semantic segmentation, part segmentation, and 3D object classification. It is also computationally more efficient compared to other methods.

['Senem Velipasalar', 'Burak Kakillioglu', 'Jiajing Chen']

2021-11-14

['3d-object-classification', 'point-cloud-segmentation']

['computer-vision', 'computer-vision']

[-1.08120635e-01 -2.65659332e-01 -1.01186015e-01 -4.36925650e-01 -4.12794530e-01 -2.83271670e-01 5.00863492e-01 2.15225443e-01 -3.34722757e-01 -8.30316916e-03 -3.12595546e-01 -8.26490298e-02 5.66304429e-03 -9.06885505e-01 -8.02777231e-01 -5.88755071e-01 1.25998229e-01 4.70178545e-01 7.11712062e-01 1.40296608e-01 2.47980326e-01 1.15177417e+00 -1.70622754e+00 5.58964238e-02 1.04777193e+00 1.44134402e+00 5.19263685e-01 4.62769903e-02 -6.34496152e-01 1.37171924e-01 -3.74886245e-01 -4.49498966e-02 5.04078925e-01 9.40282568e-02 -6.39501393e-01 3.04300457e-01 5.60263157e-01 -2.89452016e-01 -1.89392790e-01 1.08807147e+00 4.63004708e-01 2.43265897e-01 4.02156234e-01 -1.27672124e+00 -5.62686980e-01 1.89874768e-01 -8.30735147e-01 2.10685641e-01 4.23039049e-02 1.55060604e-01 9.90776956e-01 -1.23011076e+00 3.01388234e-01 1.54913926e+00 6.36777639e-01 3.87229979e-01 -1.09840989e+00 -7.95809388e-01 5.93058288e-01 1.79438353e-01 -1.54731536e+00 1.63155153e-01 1.06165767e+00 -3.65289927e-01 1.03548419e+00 3.22758295e-02 9.71068442e-01 5.57873487e-01 -1.34228483e-01 1.20900130e+00 8.13303947e-01 -1.07836954e-01 2.90134490e-01 -2.36989111e-01 2.84012109e-01 6.21259928e-01 2.31964648e-01 -1.71659932e-01 -3.29508603e-01 -5.89211509e-02 1.04129958e+00 5.30942559e-01 -5.45507483e-02 -7.36546755e-01 -1.02825427e+00 8.13687205e-01 9.96598303e-01 1.80499703e-01 -5.72234571e-01 2.25665241e-01 1.28271416e-01 -1.70350924e-01 8.86669338e-01 6.87384233e-02 -5.19292057e-01 3.45355064e-01 -9.57602620e-01 5.06469548e-01 3.89666528e-01 1.01765954e+00 1.05448401e+00 -2.44711697e-01 -1.12003595e-01 8.98018122e-01 5.39363444e-01 5.42935193e-01 1.63326770e-01 -9.00055408e-01 3.84813249e-01 1.24595737e+00 -5.66793755e-02 -8.80299091e-01 -4.36550885e-01 -4.26016390e-01 -6.30039275e-01 4.19058084e-01 3.61885540e-02 2.79392630e-01 -1.29012203e+00 1.35579133e+00 6.51701272e-01 5.40461898e-01 -2.89742947e-01 1.09425950e+00 1.14968503e+00 5.86506367e-01 3.63820419e-02 3.22252572e-01 1.11318076e+00 -7.68450618e-01 -1.34677589e-01 -1.88711315e-01 2.31966615e-01 -4.75976467e-01 9.84733880e-01 -3.83099876e-02 -1.00029147e+00 -6.79391861e-01 -8.30348074e-01 -2.78276205e-01 -4.61519659e-01 -1.48957253e-01 7.04564989e-01 2.92890340e-01 -9.37052548e-01 5.60898185e-01 -1.06068718e+00 -4.11249310e-01 1.16352737e+00 5.63343227e-01 -7.87987113e-02 -6.68328106e-02 -6.19384885e-01 5.22815704e-01 3.75234216e-01 1.20203324e-01 -5.81053257e-01 -8.53839576e-01 -7.87837923e-01 2.22372741e-01 4.07701761e-01 -8.91784489e-01 1.01872206e+00 -6.27016485e-01 -1.35150802e+00 1.09265196e+00 -2.60827988e-01 -3.51922840e-01 2.95319676e-01 -4.75569248e-01 1.08151652e-01 2.19485551e-01 3.03063065e-01 1.05356455e+00 6.87624693e-01 -1.39296162e+00 -9.15711224e-01 -7.52559900e-01 4.46258672e-02 2.54025757e-01 6.89230189e-02 -1.34306952e-01 -8.64227295e-01 -4.22391295e-01 6.77778065e-01 -7.56756604e-01 -4.15688872e-01 1.64370522e-01 -3.69538665e-01 -7.42379367e-01 1.26191509e+00 -4.62161824e-02 6.77704036e-01 -2.21828103e+00 1.02413744e-01 2.70516336e-01 3.44077647e-01 4.82830644e-01 -5.97591586e-02 -1.28064170e-01 8.79395753e-02 2.25857079e-01 -4.53168660e-01 -5.19505739e-01 4.73078191e-02 2.30455220e-01 -1.75454497e-01 3.67214233e-01 6.06513977e-01 1.13931000e+00 -8.09790552e-01 -4.54699397e-01 6.56145275e-01 5.71701884e-01 -6.98225737e-01 7.28119351e-03 -4.00564015e-01 3.64885598e-01 -8.50840092e-01 1.05052269e+00 1.05326962e+00 -3.26120228e-01 -6.19442821e-01 -9.90311056e-02 -2.22543016e-01 7.05499947e-02 -1.06679904e+00 1.90340519e+00 -2.08044827e-01 3.73264104e-01 1.21772021e-01 -9.53077078e-01 1.17321658e+00 -1.34886935e-01 8.18114877e-01 -5.40156543e-01 1.91687405e-01 3.59955132e-01 -2.70962059e-01 -1.95882797e-01 3.14021051e-01 3.32500897e-02 6.29849918e-03 6.26420276e-03 1.71129122e-01 -4.74336207e-01 -1.10685296e-01 -1.06962219e-01 8.66731286e-01 2.11090282e-01 -1.19750619e-01 -1.65707022e-01 7.04947889e-01 5.20709530e-02 6.42990947e-01 6.77685261e-01 -2.80006468e-01 6.97373331e-01 1.86605170e-01 -5.49982667e-01 -6.94307983e-01 -1.12509429e+00 -3.10006559e-01 6.12654090e-01 8.03638279e-01 -1.72098041e-01 -6.73918903e-01 -8.03694308e-01 5.91229618e-01 4.01086777e-01 -2.99792111e-01 -7.48688355e-02 -6.66219711e-01 -5.24632692e-01 1.30532771e-01 6.92722619e-01 9.28365707e-01 -1.12049353e+00 -6.35749936e-01 1.22760810e-01 1.57850683e-01 -1.24893463e+00 -2.26102620e-01 2.62974501e-01 -1.30205655e+00 -9.11346555e-01 -6.91761672e-01 -8.68272364e-01 5.95836282e-01 6.46690309e-01 9.96653259e-01 1.20034084e-01 -1.24337554e-01 4.16305602e-01 -3.35984200e-01 -6.94539189e-01 2.85962373e-01 2.93254077e-01 -1.76274270e-01 1.52219564e-01 8.83907676e-01 -5.53953707e-01 -6.38578475e-01 2.95225650e-01 -8.29344571e-01 -1.63418695e-01 6.31626070e-01 5.69197714e-01 1.09904373e+00 -1.21226363e-01 2.76438296e-01 -6.59131944e-01 3.35642219e-01 -3.53951544e-01 -7.61086285e-01 -1.44017294e-01 -1.77095458e-01 -3.48518014e-01 2.97785997e-01 -4.27156426e-02 -7.51229107e-01 2.56757945e-01 -3.10020894e-01 -1.01289654e+00 -4.21683699e-01 3.15806448e-01 -4.91935849e-01 -3.81305993e-01 1.59000307e-01 1.02631889e-01 -3.93827595e-02 -7.70335853e-01 3.13784480e-01 5.91661155e-01 3.20918500e-01 -4.16519612e-01 9.15574491e-01 7.68211782e-01 8.41241926e-02 -9.07075822e-01 -8.98021221e-01 -8.06671917e-01 -9.76715565e-01 -1.32869184e-01 1.07053506e+00 -9.78804350e-01 -7.38328218e-01 7.48659790e-01 -1.31688976e+00 -1.46139964e-01 -4.76914078e-01 2.93731749e-01 -5.03491700e-01 3.25387090e-01 -2.66043693e-01 -6.72032475e-01 -3.24620545e-01 -1.29295266e+00 1.60234523e+00 4.74987030e-01 2.03789070e-01 -7.64053166e-01 -3.22748452e-01 2.39697784e-01 2.11657640e-02 3.36767942e-01 9.07226205e-01 -6.56757057e-01 -1.16093731e+00 -2.38710523e-01 -6.53272390e-01 4.44862962e-01 1.39951900e-01 -2.01739311e-01 -9.21811938e-01 -7.32438415e-02 2.02746820e-02 7.54704252e-02 1.19639838e+00 6.31590247e-01 1.54138196e+00 3.14528614e-01 -6.04586303e-01 9.51424479e-01 1.47986567e+00 1.62267536e-01 2.78804034e-01 2.77513236e-01 1.05054939e+00 3.19970012e-01 5.94489634e-01 2.74940521e-01 3.38412732e-01 5.34605384e-01 8.82684171e-01 -3.43801886e-01 -1.50703266e-02 -1.31191984e-01 -9.27232727e-02 4.59033340e-01 -1.40884370e-01 -2.93862075e-02 -9.72955465e-01 5.56076169e-01 -1.92050338e+00 -6.23064876e-01 -6.39060661e-02 2.05309868e+00 1.28962457e-01 2.50923604e-01 1.18866086e-01 -5.40713258e-02 8.37261379e-01 3.15173239e-01 -9.12267625e-01 -6.20221086e-02 -1.46307290e-01 4.90821779e-01 4.65018392e-01 2.09110051e-01 -1.37828469e+00 1.24075270e+00 5.41018867e+00 8.92399728e-01 -1.11242807e+00 2.41233804e-03 3.55273932e-01 -8.93649980e-02 -1.37745142e-01 -1.27359286e-01 -1.01528978e+00 4.89026040e-01 1.31287441e-01 1.68735161e-01 1.45524982e-02 9.82289612e-01 2.31096193e-01 -7.63622345e-03 -9.53145206e-01 1.08504415e+00 -1.05184354e-01 -1.39105344e+00 3.70119870e-01 1.18807241e-01 5.99266827e-01 4.92142111e-01 -4.82962169e-02 1.25315040e-01 3.02068330e-02 -7.34342158e-01 7.84278095e-01 3.69052887e-01 4.11843956e-01 -9.79898572e-01 7.30965257e-01 4.54264939e-01 -1.34692454e+00 -2.24251032e-01 -5.76502085e-01 8.87901410e-02 1.41952828e-01 8.46216023e-01 -4.33890224e-01 5.61985075e-01 1.12471104e+00 1.04575443e+00 -4.41583484e-01 1.36931121e+00 -1.52892426e-01 3.09462577e-01 -6.09288216e-01 -1.07453287e-01 6.19394183e-01 -3.50180894e-01 7.49051809e-01 9.65009630e-01 2.63705522e-01 2.14172453e-01 4.96029258e-01 1.28749311e+00 -1.76568612e-01 -3.65443248e-03 -5.15573263e-01 1.33749500e-01 5.20421982e-01 1.14259028e+00 -1.00013065e+00 -2.83443898e-01 -5.17117381e-01 6.90887451e-01 2.97959924e-01 2.87404299e-01 -6.29715323e-01 -5.04145861e-01 1.03088498e+00 1.36385068e-01 8.12844813e-01 -5.11257648e-01 -4.76804137e-01 -9.95805025e-01 7.08927512e-02 -2.16983348e-01 9.08550173e-02 -6.60999119e-01 -1.33260834e+00 4.20280010e-01 7.59088024e-02 -1.17740381e+00 4.40176308e-01 -7.19322443e-01 -6.96038008e-01 9.68786001e-01 -1.84981275e+00 -1.27452695e+00 -5.33837616e-01 5.51557600e-01 6.16335392e-01 -7.30471220e-03 2.67660439e-01 2.66938239e-01 -4.20042396e-01 5.63252307e-02 -1.70055583e-01 1.38174832e-01 1.25945076e-01 -1.20787799e+00 6.62580132e-01 7.37401962e-01 1.34374619e-01 4.83396471e-01 -2.61128396e-02 -7.05871820e-01 -1.08853161e+00 -1.37299562e+00 6.23349726e-01 -3.21484625e-01 2.68965870e-01 -5.51944315e-01 -1.09784222e+00 4.66803223e-01 -3.04071814e-01 2.26487488e-01 3.94587308e-01 -1.21754140e-01 -3.63639593e-02 -2.34224677e-01 -1.17678630e+00 3.56362700e-01 1.41590619e+00 -3.70175034e-01 -6.96045637e-01 3.10701489e-01 1.08267331e+00 -5.19553185e-01 -5.60983062e-01 7.28213131e-01 9.40402746e-02 -9.93421853e-01 1.25385392e+00 -2.67394871e-01 1.81990802e-01 -5.49746335e-01 -1.83800414e-01 -1.04731107e+00 -5.38177311e-01 -2.35465407e-01 -7.50382766e-02 1.09513140e+00 1.43360505e-02 -6.07698917e-01 1.14185560e+00 3.09418261e-01 -6.62033021e-01 -9.46914256e-01 -1.14378834e+00 -7.67506599e-01 7.23965615e-02 -6.09885931e-01 1.03621507e+00 5.99022150e-01 -7.65349150e-01 1.21250957e-01 3.34503233e-01 3.43661398e-01 7.00856984e-01 6.21971726e-01 8.05324733e-01 -1.68622112e+00 2.59745985e-01 -8.17268252e-01 -7.45615065e-01 -1.61189771e+00 3.45433861e-01 -1.24017513e+00 -1.63583476e-02 -1.73387110e+00 -4.20208126e-02 -5.94983399e-01 -3.33588034e-01 4.32897925e-01 -2.30730698e-01 2.68069685e-01 4.02610570e-01 3.07143360e-01 -5.68829894e-01 7.49297976e-01 1.34640932e+00 -2.74500936e-01 -5.58666527e-01 2.96648771e-01 -4.24907207e-01 9.05328214e-01 5.23179114e-01 -2.93663114e-01 -7.89721608e-02 -6.79929793e-01 -3.22851241e-01 -5.63717246e-01 7.26247549e-01 -1.11125112e+00 2.68619001e-01 5.20324111e-02 5.91275215e-01 -1.37703049e+00 4.67718571e-01 -1.02585363e+00 -2.42183149e-01 3.11474115e-01 2.21734509e-01 -2.43449554e-01 3.38456899e-01 6.14059508e-01 -3.50388676e-01 -3.19333784e-02 7.75029540e-01 -3.98084432e-01 -1.07219684e+00 8.44056129e-01 -1.44877899e-02 -2.17256799e-01 1.08328950e+00 -6.58917010e-01 7.18989819e-02 3.33472461e-01 -6.78693354e-01 5.69055915e-01 5.93093455e-01 5.64224541e-01 9.04433370e-01 -1.39863968e+00 -4.46886033e-01 4.49272186e-01 3.86166610e-02 8.70007277e-01 2.62488246e-01 5.84651530e-01 -5.38602948e-01 4.10058379e-01 -9.82462838e-02 -1.34129226e+00 -9.49095905e-01 4.42416310e-01 3.05618912e-01 1.58754840e-01 -9.11699831e-01 1.06541872e+00 4.77508247e-01 -6.05312407e-01 2.03253910e-01 -7.62085319e-01 -1.66103885e-01 -3.73313725e-02 4.58023772e-02 3.08539212e-01 1.18289948e-01 -7.14681923e-01 -6.23318911e-01 1.14366651e+00 3.73441093e-02 3.79684895e-01 1.51794875e+00 -3.23198810e-02 -1.48671821e-01 4.34909672e-01 1.09970510e+00 -3.46347421e-01 -1.42517626e+00 -4.75752205e-01 4.35797013e-02 -7.34312534e-01 3.16979498e-01 -4.81658638e-01 -1.40603888e+00 1.01457620e+00 5.87356508e-01 3.02139521e-01 1.14352083e+00 3.41569066e-01 9.54072893e-01 1.19583368e-01 4.33498442e-01 -7.91802883e-01 -1.00592896e-01 7.48493612e-01 6.62106693e-01 -1.29094815e+00 -2.94407383e-02 -7.72872567e-01 -1.96359381e-01 8.77189100e-01 7.64084458e-01 -5.24171650e-01 8.61673951e-01 -1.61531776e-01 -7.06062689e-02 -4.91966456e-01 -2.35923916e-01 -4.88210112e-01 4.16116625e-01 5.98441899e-01 -7.51799671e-03 -9.92600694e-02 1.06152341e-01 6.18661582e-01 4.25150841e-02 -7.70649463e-02 -1.17308363e-01 8.97497118e-01 -6.64867103e-01 -8.88374805e-01 -3.55112016e-01 5.14831662e-01 -1.56474277e-01 1.99353248e-01 -5.46198785e-01 8.97526979e-01 4.26671743e-01 5.83722413e-01 5.61468482e-01 -2.69830495e-01 6.36012971e-01 4.34832275e-02 3.73398006e-01 -8.65098536e-01 -6.29174709e-01 1.72433347e-01 -6.38810694e-01 -7.89587677e-01 -7.20627964e-01 -7.60218263e-01 -1.51879644e+00 -3.17747369e-02 -5.48311055e-01 -1.09716758e-01 7.22634256e-01 9.35821235e-01 6.80924296e-01 4.82169271e-01 6.38542950e-01 -1.49800122e+00 -1.65532008e-01 -6.03353858e-01 -6.28736198e-01 2.88938254e-01 3.11297208e-01 -9.44218874e-01 -2.73248553e-01 -4.26371098e-01]

[7.95474910736084, -3.357167959213257]

e67e7117-97a4-487c-b7c1-7dde6b4072a5

language-model-pre-training-with-sparse

2210.12582

https://arxiv.org/abs/2210.12582v2

https://arxiv.org/pdf/2210.12582v2.pdf

Language Model Pre-Training with Sparse Latent Typing

Modern large-scale Pre-trained Language Models (PLMs) have achieved tremendous success on a wide range of downstream tasks. However, most of the LM pre-training objectives only focus on text reconstruction, but have not sought to learn latent-level interpretable representations of sentences. In this paper, we manage to push the language models to obtain a deeper understanding of sentences by proposing a new pre-training objective, Sparse Latent Typing, which enables the model to sparsely extract sentence-level keywords with diverse latent types. Experimental results show that our model is able to learn interpretable latent type categories in a self-supervised manner without using any external knowledge. Besides, the language model pre-trained with such an objective also significantly improves Information Extraction related downstream tasks in both supervised and few-shot settings. Our code is publicly available at: https://github.com/renll/SparseLT.

['Heng Ji', 'ChengXiang Zhai', 'Clare R. Voss', 'Han Wang', 'Zixuan Zhang', 'Liliang Ren']

2022-10-23

['few-shot-ner']

['natural-language-processing']

[ 2.16689408e-01 5.36073208e-01 -6.82266712e-01 -4.58095312e-01 -1.04329967e+00 -4.28923905e-01 6.56227171e-01 1.91383690e-01 -1.69077978e-01 5.32948732e-01 7.41720557e-01 -4.75605726e-01 2.53969818e-01 -5.82713723e-01 -7.39295423e-01 -3.56252313e-01 3.03503394e-01 4.66033518e-01 -2.94499755e-01 8.90213624e-02 1.40170395e-01 -1.41681373e-01 -1.17458558e+00 7.25507498e-01 8.15821946e-01 5.25814056e-01 5.46139598e-01 5.12236297e-01 -5.48066676e-01 1.04509234e+00 -1.92628011e-01 -5.36706209e-01 -1.62922367e-01 -4.17570323e-01 -9.75288451e-01 1.78993911e-01 8.57035071e-02 -2.29450807e-01 -4.76967901e-01 9.17563856e-01 1.45050049e-01 -7.30278343e-02 6.86255813e-01 -7.58311033e-01 -6.95646405e-01 1.33489764e+00 -2.86076218e-01 6.65791929e-02 1.92624912e-01 1.04848355e-01 1.51563716e+00 -1.20372856e+00 5.60479999e-01 1.44350553e+00 4.87266332e-01 5.86836874e-01 -1.40346813e+00 -5.18257797e-01 4.57991809e-01 1.81939021e-01 -1.13941705e+00 -7.38749921e-01 9.17996943e-01 -1.92733720e-01 1.12803829e+00 2.32074112e-01 2.11606145e-01 1.53839040e+00 3.97607908e-02 1.32921278e+00 8.54123235e-01 -6.91477656e-01 9.17988420e-02 3.39681923e-01 3.75565886e-01 8.75912249e-01 1.62464693e-01 -6.01050556e-01 -6.66377127e-01 -1.03976401e-02 2.90646136e-01 1.29655913e-01 -2.03847900e-01 -2.72803083e-02 -1.11742616e+00 1.00276113e+00 1.10589772e-01 5.42695224e-01 -1.59845874e-01 1.88902900e-01 4.56594110e-01 1.13349810e-01 9.92064834e-01 5.67430615e-01 -6.31762862e-01 -2.01711640e-01 -9.64303255e-01 -2.79482365e-01 8.35457325e-01 1.02746224e+00 7.53208280e-01 5.21399565e-02 -4.07973498e-01 9.38426077e-01 4.44221705e-01 3.60413909e-01 7.11829424e-01 -8.63767564e-01 8.73571277e-01 6.27128899e-01 -2.86523104e-01 -8.47758830e-01 -1.06395312e-01 -5.50901234e-01 -1.00364423e+00 -6.77854896e-01 2.20479779e-02 -6.53284341e-02 -8.82294834e-01 1.61277437e+00 -2.27039352e-01 1.51824296e-01 1.77575514e-01 4.36101675e-01 8.66013646e-01 9.94080186e-01 9.34176967e-02 -2.26042211e-01 1.40181124e+00 -1.31742954e+00 -1.00415719e+00 -7.35162199e-01 1.02783966e+00 -6.15031838e-01 1.41763902e+00 2.70495921e-01 -1.07357001e+00 -4.20531899e-01 -8.18557680e-01 -5.24663031e-01 -1.70165569e-01 5.11053443e-01 7.59416580e-01 2.70633936e-01 -7.41820633e-01 3.05568337e-01 -1.08587217e+00 -9.19851661e-02 5.82570970e-01 1.13382123e-01 -1.80148825e-01 -1.29136816e-01 -1.12246394e+00 6.76423073e-01 3.33290458e-01 1.22792132e-01 -1.03613627e+00 -6.16283178e-01 -1.01758420e+00 3.35026175e-01 5.09108543e-01 -9.50506747e-01 1.38318956e+00 -8.59220684e-01 -1.55820572e+00 8.30224633e-01 -1.04389143e+00 -5.50074399e-01 1.49342909e-01 -4.43663269e-01 -2.82396059e-02 -2.37205345e-02 2.57176578e-01 4.82270122e-01 7.49718249e-01 -1.34341204e+00 -3.01443219e-01 -1.84037581e-01 1.09356187e-01 1.68948412e-01 -8.22038829e-01 -1.12164058e-01 -6.88079119e-01 -8.13348830e-01 5.82168363e-02 -6.13974988e-01 -3.16494286e-01 -3.50185007e-01 -7.70888031e-01 -4.95538622e-01 4.24616992e-01 -6.89311206e-01 1.52640712e+00 -1.86413360e+00 4.65509444e-01 -3.11725914e-01 3.86770517e-01 2.35023320e-01 -3.69863629e-01 6.35240078e-01 8.97716209e-02 4.27464575e-01 -4.19347882e-01 -1.25090492e+00 3.53356376e-02 2.94169188e-01 -9.25818264e-01 1.04433835e-01 1.64688468e-01 1.24619186e+00 -1.03323793e+00 -4.60801899e-01 1.00680172e-01 2.26494864e-01 -6.26740873e-01 3.30673307e-01 -5.93405962e-01 2.95663446e-01 -6.26662254e-01 3.27309668e-01 2.69415766e-01 -6.06019676e-01 3.30056161e-01 4.57676947e-02 2.11442206e-02 9.60775197e-01 -5.54571211e-01 2.07575631e+00 -9.72138345e-01 7.67623842e-01 -1.49570748e-01 -1.35405838e+00 5.70783794e-01 4.02045220e-01 2.41430312e-01 -2.10131302e-01 3.19013223e-02 6.91925734e-02 -2.98132032e-01 -4.95230019e-01 4.70668554e-01 -1.75789461e-01 -1.54029950e-01 6.93647087e-01 2.47461244e-01 2.18627261e-04 7.31658340e-02 6.02087259e-01 1.03915119e+00 -2.67278422e-02 2.06736058e-01 8.34624842e-03 5.08832633e-01 -8.44798982e-02 4.82960790e-01 8.83369505e-01 5.20261705e-01 5.08651793e-01 5.72582603e-01 -1.14075951e-01 -8.95043790e-01 -9.35448587e-01 -5.12770563e-02 1.06635308e+00 -8.16878974e-02 -1.01332855e+00 -5.11644959e-01 -6.98628008e-01 -1.31991431e-01 1.10595393e+00 -3.23603541e-01 -2.31211931e-01 -4.63290811e-01 -5.58751941e-01 6.23845756e-01 4.04235542e-01 1.89658687e-01 -9.39339161e-01 2.46211737e-01 1.87893331e-01 -6.45747006e-01 -1.23489571e+00 -3.30960751e-01 3.05967063e-01 -1.10864770e+00 -4.81762856e-01 -3.57200056e-01 -8.72136474e-01 1.01675797e+00 2.67906070e-01 1.10780370e+00 -1.64089389e-02 -7.51085812e-03 1.88842416e-01 -4.76169020e-01 -3.21034938e-01 -6.02840424e-01 5.09852052e-01 4.62195836e-02 7.87106529e-02 5.59831917e-01 -5.19690990e-01 -1.78705126e-01 -1.20823339e-01 -8.04725289e-01 5.02187669e-01 7.42583394e-01 8.34182799e-01 4.63966072e-01 -1.28692508e-01 4.74237561e-01 -1.22889853e+00 6.72765017e-01 -7.07194209e-01 -1.49848893e-01 3.11897486e-01 -5.88613987e-01 5.45971453e-01 8.94691169e-01 -4.17118758e-01 -1.18961966e+00 -1.81335494e-01 -3.52447182e-01 -3.04935366e-01 -1.55912086e-01 9.42052662e-01 -3.10885996e-01 7.43827522e-01 2.80638695e-01 5.40892959e-01 -2.86107957e-01 -9.94059801e-01 5.74683666e-01 8.90700221e-01 2.09154218e-01 -5.73936939e-01 9.67542827e-01 5.08454859e-01 -3.44139606e-01 -8.53855550e-01 -1.66687799e+00 -6.05206490e-01 -7.79241800e-01 2.85192132e-01 6.48919702e-01 -1.23870599e+00 -1.31091222e-01 1.99762002e-01 -1.35930502e+00 -4.35902447e-01 -2.23893508e-01 2.92398721e-01 -3.81054223e-01 6.01525486e-01 -7.29653418e-01 -7.95849621e-01 -4.88210291e-01 -8.93989861e-01 1.20513272e+00 -2.83032715e-01 -4.45631564e-01 -1.30116034e+00 -5.44231310e-02 6.82130158e-01 2.00945482e-01 -5.37756622e-01 9.85518515e-01 -6.07683778e-01 -7.91255832e-01 -1.59882843e-01 2.34113876e-02 3.26589316e-01 3.25155854e-01 -4.20798272e-01 -1.11191559e+00 -1.38437122e-01 1.54202729e-01 -4.64946270e-01 1.38988996e+00 2.60809124e-01 1.54129016e+00 -8.05561602e-01 -4.49582666e-01 7.61895239e-01 1.00872743e+00 -4.40250546e-01 2.79875129e-01 -2.82972176e-02 9.15840626e-01 4.94356185e-01 3.73236328e-01 3.06064337e-01 6.06549025e-01 5.74995041e-01 5.70096299e-02 -9.54654589e-02 -4.31877412e-02 -7.96236753e-01 6.49643958e-01 1.37546861e+00 1.44204766e-01 -4.55876410e-01 -8.62026751e-01 5.09184539e-01 -2.10576797e+00 -9.73810971e-01 -2.76054181e-02 1.65548563e+00 1.30151761e+00 3.06705534e-01 -4.00038838e-01 -1.14744239e-01 2.81931937e-01 3.15184474e-01 -5.97757339e-01 -2.82582343e-01 -3.26264165e-02 4.87777553e-02 2.99784303e-01 8.08709383e-01 -1.07160139e+00 1.29448676e+00 5.39998770e+00 1.08218598e+00 -9.18695331e-01 3.07567894e-01 8.63308787e-01 -4.13597934e-02 -9.85068500e-01 3.28795522e-01 -1.20424330e+00 4.52287495e-01 1.10082614e+00 -3.85731965e-01 3.12438726e-01 7.67739236e-01 5.65201581e-01 2.19235435e-01 -1.19054866e+00 7.81141937e-01 1.47375211e-01 -1.65057325e+00 3.82569611e-01 -1.55341467e-02 6.21360719e-01 1.29272446e-01 -1.84366200e-02 6.02262199e-01 2.52357244e-01 -1.14006019e+00 6.74195409e-01 5.56017578e-01 7.59766757e-01 -3.08034211e-01 5.55597842e-01 1.01897824e+00 -1.07245123e+00 -1.78067237e-02 -4.75296885e-01 -2.87653297e-01 3.31736505e-01 8.10273707e-01 -9.31204081e-01 3.57486635e-01 1.64370030e-01 1.22782111e+00 -5.19583046e-01 4.26939636e-01 -7.70493627e-01 1.17436242e+00 -1.56921640e-01 -1.87740386e-01 2.57023096e-01 -6.45399094e-03 6.42675638e-01 1.35910213e+00 1.12465564e-02 -1.21640526e-02 2.71441519e-01 1.12978530e+00 -4.22633082e-01 2.82133464e-02 -6.52516127e-01 -5.11278033e-01 3.63488078e-01 1.27449918e+00 -3.82638395e-01 -6.16802216e-01 -3.67260337e-01 1.09095335e+00 6.36426508e-01 3.81637394e-01 -5.95528781e-01 -1.15247302e-01 5.97933888e-01 1.35472313e-01 1.44842297e-01 -4.68405575e-01 -6.23098075e-01 -1.96113908e+00 1.89758278e-02 -7.41128981e-01 2.16912270e-01 -5.72193027e-01 -1.21961403e+00 4.27329391e-01 1.37632834e-02 -8.98523629e-01 -5.40027201e-01 -4.83665437e-01 -5.58592856e-01 8.94586742e-01 -1.58531630e+00 -1.39871430e+00 2.30744600e-01 2.00997218e-01 1.22347605e+00 -1.70546144e-01 9.86747921e-01 -1.57471355e-02 -7.17960000e-01 6.83097839e-01 1.83715388e-01 2.86307126e-01 4.79710340e-01 -1.18050647e+00 4.82548445e-01 1.05490184e+00 5.69510698e-01 1.11034930e+00 6.32513523e-01 -5.41250348e-01 -1.65390110e+00 -1.25587082e+00 1.38255227e+00 -7.64156222e-01 8.22589278e-01 -8.30426693e-01 -8.51397514e-01 1.02717257e+00 3.17394972e-01 -3.33891690e-01 7.78834105e-01 6.01951122e-01 -3.28690469e-01 4.49539088e-02 -3.44505668e-01 6.05864406e-01 1.10926783e+00 -8.77570927e-01 -7.55167723e-01 6.54040039e-01 9.66752648e-01 -5.38382865e-02 -5.58382690e-01 2.23088264e-01 1.44932851e-01 -2.93888152e-01 9.92276490e-01 -8.72254252e-01 9.42576051e-01 2.06207439e-01 8.73498023e-02 -1.20485926e+00 -3.37528467e-01 -5.89161456e-01 -6.38092875e-01 1.40623057e+00 7.58660257e-01 -5.40991008e-01 7.26798236e-01 4.03645158e-01 -3.33139956e-01 -9.53785002e-01 -5.32495856e-01 -5.78237772e-01 6.64961711e-02 -6.42662942e-01 8.40993449e-02 7.93208301e-01 2.26182967e-01 9.66981113e-01 -5.69564998e-01 1.05093300e-01 5.88222802e-01 4.38141227e-01 6.99176371e-01 -9.52983141e-01 -5.47690272e-01 -3.97261858e-01 1.99294731e-01 -1.73349774e+00 7.62945294e-01 -1.46835721e+00 9.70111564e-02 -1.77462685e+00 6.07226372e-01 -3.37315828e-01 -2.21422270e-01 8.62977207e-01 -3.85536820e-01 -5.55734076e-02 -2.06540134e-02 3.71293724e-01 -8.24660838e-01 8.12778771e-01 1.00311482e+00 -3.70651007e-01 -5.48387598e-03 2.29087785e-01 -1.01091206e+00 9.80659604e-01 9.81902838e-01 -6.18329227e-01 -4.70188707e-01 -8.92627180e-01 3.41641337e-01 -1.45736054e-01 1.23252891e-01 -3.71858627e-01 1.76968589e-01 -5.18184677e-02 5.05415462e-02 -5.75071812e-01 4.61377949e-01 -3.96412075e-01 -3.00099134e-01 3.33349943e-01 -9.44553614e-01 -3.34043026e-01 -9.08740051e-03 6.20592594e-01 -2.74441540e-01 -4.90229517e-01 3.42558265e-01 -3.63478839e-01 -5.79390347e-01 3.08100879e-01 -5.43407857e-01 2.30849236e-01 4.39339727e-01 2.38153234e-01 -1.15582153e-01 -5.95164835e-01 -6.42568171e-01 2.30471238e-01 2.76627958e-01 5.88933527e-01 6.36668742e-01 -1.05566835e+00 -6.41588032e-01 1.35221362e-01 8.56702104e-02 1.19348414e-01 -1.05219074e-01 7.06851423e-01 9.00745839e-02 9.27708209e-01 5.41442990e-01 -4.32378978e-01 -1.12998986e+00 3.98289710e-01 -8.85137394e-02 -7.37708330e-01 -8.68642032e-01 9.96783555e-01 3.39569151e-01 -4.65357572e-01 3.67209107e-01 -5.39648712e-01 -6.65538982e-02 1.81385279e-02 4.60789740e-01 -1.14016578e-01 -1.70222268e-01 -3.74260068e-01 -1.49189249e-01 1.68576345e-01 -1.78420573e-01 -7.01929405e-02 1.63339961e+00 -3.56315970e-01 -2.42028028e-01 9.57816780e-01 1.37849748e+00 1.60040244e-01 -8.67570400e-01 -6.10043049e-01 2.99082041e-01 -2.68734008e-01 2.48247802e-01 -5.68387687e-01 -6.21667802e-01 1.28872991e+00 -3.00251991e-01 1.61669776e-02 8.89570594e-01 3.56596917e-01 1.01095366e+00 7.98368633e-01 2.60453790e-01 -7.81372786e-01 2.75917321e-01 9.08504367e-01 7.59057462e-01 -1.24145091e+00 -6.17704764e-02 -3.90812635e-01 -6.09201670e-01 1.05355215e+00 3.46962839e-01 1.98200986e-01 5.90876162e-01 3.39662164e-01 -2.09613785e-01 2.28939876e-02 -1.29263508e+00 -1.00742914e-01 5.13993621e-01 1.51913362e-02 7.86052227e-01 2.99756620e-02 -1.18546464e-01 9.17972207e-01 -2.05543473e-01 -8.48300457e-02 3.59151691e-01 5.97969770e-01 -5.76572955e-01 -1.39363587e+00 1.35729000e-01 6.11429393e-01 -4.99922633e-01 -6.54672861e-01 -2.90467680e-01 9.59528089e-02 -2.73832381e-01 8.51850510e-01 -5.73514663e-02 -2.59454399e-01 -2.04181254e-01 2.94497877e-01 1.19760156e-01 -1.24205673e+00 -9.55818370e-02 9.95862707e-02 3.54406565e-01 -3.93818140e-01 -2.23331571e-01 -4.86520797e-01 -1.25093412e+00 -8.07427019e-02 -2.31519923e-01 3.14716935e-01 4.79966402e-01 1.13619864e+00 5.35361528e-01 4.86670196e-01 5.65577924e-01 -7.55454898e-01 -7.39662945e-01 -1.26945937e+00 -1.06063619e-01 2.73694545e-01 3.29706788e-01 -2.25949883e-01 -5.72098970e-01 3.56152892e-01]

[10.987235069274902, 8.824398040771484]

24999f9b-8623-45f0-af4c-85c8ec761fcd

automatic-speech-recognition-of-non-native

2306.16710

https://arxiv.org/abs/2306.16710v1

https://arxiv.org/pdf/2306.16710v1.pdf

Automatic Speech Recognition of Non-Native Child Speech for Language Learning Applications

Voicebots have provided a new avenue for supporting the development of language skills, particularly within the context of second language learning. Voicebots, though, have largely been geared towards native adult speakers. We sought to assess the performance of two state-of-the-art ASR systems, Wav2Vec2.0 and Whisper AI, with a view to developing a voicebot that can support children acquiring a foreign language. We evaluated their performance on read and extemporaneous speech of native and non-native Dutch children. We also investigated the utility of using ASR technology to provide insight into the children's pronunciation and fluency. The results show that recent, pre-trained ASR transformer-based models achieve acceptable performance from which detailed feedback on phoneme pronunciation quality can be extracted, despite the challenging nature of child and non-native speech.

['Helmer Strik', 'Catia Cucchiarini', 'Cristian Tejedor-Garcia', 'Yu Bai', 'Simone Wills']

2023-06-29

['speech-recognition']

['speech']

[-1.80757791e-01 2.84062445e-01 -1.40060067e-01 -3.65748972e-01 -8.91539574e-01 -8.92785847e-01 4.73801702e-01 2.26552591e-01 -6.13020957e-01 1.60929531e-01 8.23420882e-01 -6.01710618e-01 3.13144803e-01 -3.55599999e-01 -4.11215007e-01 -9.87567380e-02 4.02518809e-01 4.88000512e-01 4.18655351e-02 -6.35863900e-01 -2.47371987e-01 4.16791171e-01 -1.56458092e+00 4.08089072e-01 9.53626573e-01 2.61760473e-01 3.55876952e-01 9.92945373e-01 -3.03103864e-01 1.08571911e+00 -8.74656796e-01 -5.90234756e-01 -2.17507884e-01 -3.32518697e-01 -8.66569877e-01 -4.80030209e-01 5.47872961e-01 -5.59503734e-01 -6.11200273e-01 8.22583795e-01 7.27245688e-01 2.01135769e-01 3.66271853e-01 -6.13077164e-01 -9.32012320e-01 1.34856462e+00 4.07288641e-01 4.60066557e-01 6.99694812e-01 4.02445048e-01 8.57736766e-01 -6.84298873e-01 3.81932139e-01 1.27331984e+00 3.89646232e-01 1.21774387e+00 -1.23043430e+00 -7.42886782e-01 8.38367119e-02 6.13283440e-02 -9.40807045e-01 -1.14234817e+00 5.99993885e-01 -5.23612201e-01 1.34604025e+00 9.91698503e-02 9.44752812e-01 1.38147581e+00 -4.15235877e-01 1.12036061e+00 1.22063863e+00 -6.49291039e-01 -1.33300662e-01 3.91806722e-01 6.79418221e-02 4.73139822e-01 -3.43208879e-01 1.96821243e-01 -1.23389363e+00 4.72070754e-01 4.14064348e-01 -7.18967438e-01 -2.98184961e-01 3.46720159e-01 -1.14082730e+00 7.48224258e-01 3.35218236e-02 8.38457584e-01 -1.29453674e-01 1.18405759e-01 4.44206774e-01 7.20530450e-01 5.15532970e-01 7.79182374e-01 -7.43241549e-01 -1.15348947e+00 -7.61088610e-01 2.12751562e-03 7.39037812e-01 1.15208614e+00 -1.56798184e-01 6.13484561e-01 -3.22687387e-01 1.40809846e+00 4.28589195e-01 6.20935500e-01 9.07291412e-01 -6.91589713e-01 6.01864159e-01 2.53250062e-01 -4.45199072e-01 -7.60491714e-02 8.50507542e-02 -1.80125207e-01 7.78972656e-02 1.37603050e-02 5.86965382e-01 -1.04975402e-01 -9.82039630e-01 1.96094692e+00 2.51693904e-01 -2.19275877e-01 2.53168166e-01 2.54612714e-01 1.42196357e+00 7.25122213e-01 1.39604956e-01 -1.34528011e-01 8.04257393e-01 -1.03766799e+00 -8.76732409e-01 -1.53025478e-01 6.11954808e-01 -8.78240824e-01 1.41809285e+00 4.84241724e-01 -1.61163080e+00 -4.48276281e-01 -7.13504791e-01 -4.82170768e-02 -5.03859222e-01 -2.54449457e-01 5.66359758e-01 1.32394660e+00 -1.42578375e+00 5.86990178e-01 -1.03903973e+00 -3.01109314e-01 1.58813313e-01 2.82890141e-01 -5.64619899e-01 7.48853832e-02 -1.03112495e+00 1.17843151e+00 4.41771597e-02 -4.52219218e-01 -1.05830204e+00 -1.17580783e+00 -1.01658654e+00 -6.89033195e-02 -3.20833802e-01 3.23445141e-01 2.04878402e+00 -7.73795009e-01 -2.33663511e+00 1.18965578e+00 2.46745855e-01 -1.10989302e-01 2.77785122e-01 -4.62596416e-01 -5.81231177e-01 5.47814146e-02 7.11581036e-02 6.74314320e-01 3.52258384e-01 -4.23971951e-01 -7.31258154e-01 -3.02912056e-01 -5.18367112e-01 4.14035857e-01 -4.61251825e-01 7.13051140e-01 8.98201689e-02 -6.92785800e-01 -1.69648007e-01 -5.51831722e-01 1.74473122e-01 -4.58837301e-01 -1.11630477e-01 -6.09707654e-01 1.82823330e-01 -1.05131936e+00 1.13430393e+00 -1.90721726e+00 -6.35715276e-02 -1.73807129e-01 -2.26448610e-01 7.84379423e-01 -4.54451501e-01 4.38047558e-01 1.11755729e-01 2.06534490e-01 4.41014588e-01 -1.22977369e-01 -2.30068229e-02 -1.28816351e-01 -9.99731757e-03 2.10149571e-01 2.73870558e-01 8.99841785e-01 -1.15150845e+00 -2.48515718e-02 4.56625253e-01 6.63880527e-01 -5.14566481e-01 8.90310705e-01 7.23924041e-02 6.04156375e-01 2.07573101e-01 4.82613742e-01 -2.48449966e-01 7.55978167e-01 -2.10653573e-01 6.12938404e-01 -5.56712151e-01 1.41590691e+00 -5.81320405e-01 1.71747077e+00 -8.35119069e-01 9.08388913e-01 3.85138273e-01 -5.73613465e-01 8.11635077e-01 8.69289339e-01 1.71098873e-01 -6.45320177e-01 3.21185112e-01 5.12751281e-01 7.01457620e-01 -4.76183444e-01 2.69231588e-01 -7.20605776e-02 3.70657474e-01 1.72058865e-01 5.41793644e-01 -8.38822246e-01 -4.89935465e-02 -2.45423838e-01 1.25702429e+00 1.45595083e-02 5.55885732e-02 -4.95813787e-01 1.64829805e-01 -8.93981457e-02 2.27011051e-02 6.90976501e-01 -2.19718635e-01 4.97775674e-01 -3.23748559e-01 4.86697294e-02 -7.37313807e-01 -1.41683972e+00 1.37935191e-01 1.59385991e+00 -8.04099321e-01 -3.10881227e-01 -9.76034284e-01 -3.09319407e-01 -1.91942200e-01 1.36302233e+00 -1.38470098e-01 -1.76251113e-01 -6.36394799e-01 2.94308186e-01 8.29822004e-01 5.79695821e-01 -2.78908968e-01 -1.55322504e+00 -5.21580815e-01 7.96314716e-01 1.76133186e-01 -1.12120450e+00 -6.52297020e-01 1.23076655e-01 -4.31420058e-01 -6.39957368e-01 -8.91115069e-01 -9.46375191e-01 -2.01695517e-01 5.01016974e-02 1.06069958e+00 -9.53588039e-02 -1.10885002e-01 9.25290287e-01 -5.38481832e-01 -8.39827240e-01 -1.12049198e+00 2.01983586e-01 3.70472014e-01 -7.24956691e-01 6.69594228e-01 -8.43319297e-01 -6.46664351e-02 -1.89599350e-01 -3.28869462e-01 9.50148422e-03 2.00371251e-01 5.90999722e-01 -3.65649676e-03 -4.83385891e-01 5.67654848e-01 -5.67005396e-01 7.04836965e-01 -2.51828104e-01 -3.71313483e-01 1.71166494e-01 -3.87813717e-01 -1.90375865e-01 6.82115078e-01 -9.34882760e-01 -9.12784696e-01 -7.23518953e-02 -7.40452290e-01 -2.97174066e-01 -5.71740568e-01 2.05312297e-01 -2.04031765e-01 1.98007926e-01 5.04768848e-01 2.27126479e-01 1.91691995e-01 -6.82568550e-01 7.89127588e-01 1.31393921e+00 1.01739407e+00 -7.44017720e-01 5.10416329e-01 -5.18981457e-01 -8.15901101e-01 -1.49207604e+00 -6.48947895e-01 -4.82984722e-01 -5.91693878e-01 -2.72132665e-01 7.55293548e-01 -1.01395130e+00 -7.17628121e-01 8.56324434e-01 -1.31967020e+00 -8.51064444e-01 -5.40027201e-01 8.09796572e-01 -6.67840838e-01 -5.94701946e-01 -8.29755247e-01 -8.99540186e-01 -5.62674224e-01 -1.29164386e+00 4.19739246e-01 4.87660527e-01 -6.47216022e-01 -9.52612102e-01 4.04363006e-01 6.23229623e-01 7.32370853e-01 -5.54234862e-01 8.36053908e-01 -1.36870265e+00 -1.21805809e-01 1.85067952e-01 4.90444571e-01 7.37378895e-01 3.21539789e-01 9.61586684e-02 -1.26494312e+00 -1.14369132e-01 -3.31857473e-01 -6.85180306e-01 2.67959293e-02 2.22960562e-01 4.21533585e-01 -3.90913814e-01 3.33525509e-01 4.39444900e-01 6.88990951e-01 4.91075367e-01 1.71346188e-01 6.38581486e-03 7.39754438e-01 8.86105299e-01 2.09450483e-01 -2.07092449e-01 4.82548505e-01 5.99863946e-01 -2.46087760e-01 3.42732310e-01 -6.33547604e-01 -8.45492482e-01 7.70753324e-01 1.78851533e+00 1.59974307e-01 -1.76892996e-01 -1.08820117e+00 1.18994236e+00 -9.64626253e-01 -7.52692640e-01 -1.72779504e-02 2.08177328e+00 1.49037468e+00 -6.42894115e-03 3.22923869e-01 2.16760755e-01 5.90212226e-01 1.48379048e-02 -1.62622079e-01 -1.13060904e+00 3.39999646e-01 9.78831232e-01 1.23842828e-01 6.45261884e-01 -3.77801746e-01 1.76964796e+00 6.67855740e+00 8.48678410e-01 -1.30267012e+00 1.75766006e-01 1.90993294e-01 -7.53101781e-02 -5.92450917e-01 -6.63153589e-01 -8.81172955e-01 3.00664395e-01 1.43263066e+00 -1.42629281e-01 9.19168293e-01 6.53174937e-01 9.88311172e-02 2.90532500e-01 -1.35181999e+00 7.52251625e-01 -1.60130367e-01 -8.90354753e-01 -3.37513506e-01 -5.58954060e-01 4.82836723e-01 5.49838901e-01 3.31353880e-02 6.25156581e-01 8.38366032e-01 -1.41854298e+00 1.13707590e+00 -5.88875562e-02 1.17153943e+00 -6.63877606e-01 8.61579850e-02 2.27330789e-01 -9.28829610e-01 2.95349926e-01 1.54911056e-01 -3.55305701e-01 -2.72600412e-01 -2.13137835e-01 -1.30482352e+00 -4.51714098e-01 5.75759828e-01 3.33453029e-01 -3.77330959e-01 7.36822069e-01 -6.66410744e-01 1.57435799e+00 -4.72225219e-01 -6.90496147e-01 1.57901391e-01 2.39113197e-01 5.77250600e-01 1.34589446e+00 3.27837437e-01 3.47091436e-01 -2.44525298e-01 6.01206362e-01 1.70885757e-01 7.11393297e-01 -9.21616673e-01 -7.17657447e-01 9.89080131e-01 8.09519470e-01 -4.41679060e-01 1.12010747e-01 -7.19222605e-01 6.35336697e-01 5.94577670e-01 -4.64505292e-02 2.24379357e-02 -3.03987235e-01 1.09363842e+00 6.44992292e-02 -2.53254492e-02 -3.74693900e-01 -2.27925122e-01 -8.65019917e-01 -4.00611043e-01 -1.41112924e+00 -1.34433091e-01 -5.62516749e-01 -9.65466559e-01 4.55410153e-01 -2.72603422e-01 -4.59535897e-01 -5.96667945e-01 -8.21383119e-01 -8.98905694e-01 9.82350528e-01 -9.01147842e-01 -1.15857017e+00 2.06978649e-01 3.75230223e-01 1.09260452e+00 -6.45831287e-01 9.39725399e-01 1.73586067e-02 -3.06088477e-01 1.00313258e+00 -5.79145662e-02 2.56894916e-01 3.30425829e-01 -1.50461912e+00 7.63128936e-01 7.17267632e-01 6.60077333e-01 3.66157055e-01 8.86740386e-01 -4.72757339e-01 -1.31617236e+00 -6.96663439e-01 9.22850907e-01 -5.92116475e-01 9.11601961e-01 -4.88799453e-01 -8.13572228e-01 7.20831931e-01 5.07475197e-01 -3.96051913e-01 7.40531802e-01 2.37512782e-01 -1.97074130e-01 -1.93384197e-02 -1.09727108e+00 6.10838413e-01 1.23870456e+00 -8.72975409e-01 -9.75151181e-01 1.40756741e-02 1.32756507e+00 -4.07036722e-01 -9.10051227e-01 3.52393776e-01 5.80634356e-01 -6.95404112e-01 6.65443659e-01 -8.07450294e-01 4.39469218e-01 5.14086902e-01 -1.79387674e-01 -2.05676675e+00 -8.27892795e-02 -1.17997205e+00 2.66182184e-01 1.89630866e+00 5.31377971e-01 -4.11254853e-01 4.48899657e-01 6.98493183e-01 -5.25181711e-01 -5.92055917e-01 -9.86938000e-01 -8.23562145e-01 5.59079528e-01 -8.01120400e-01 5.65794230e-01 7.88620055e-01 2.78216690e-01 4.56317127e-01 2.80669153e-01 -5.23479097e-02 -1.21584825e-01 -7.26222396e-01 5.50280392e-01 -1.02635169e+00 -1.14679761e-01 -7.00428426e-01 -4.82900113e-01 -4.82500076e-01 3.79999459e-01 -1.00972116e+00 4.88193035e-01 -1.17576051e+00 -4.56489801e-01 -2.51205415e-01 -1.87779427e-01 5.66399992e-01 -1.80236772e-01 -1.71336398e-01 3.43366921e-01 -6.23522878e-01 -2.03368021e-03 5.63745916e-01 1.00402677e+00 -2.99475202e-03 -6.32977843e-01 4.09148782e-01 -6.67186499e-01 7.74991274e-01 7.31849313e-01 -3.79053622e-01 -6.79201245e-01 -5.79740345e-01 -3.58651221e-01 1.71203092e-01 -4.07966614e-01 -1.00935829e+00 1.58878565e-01 -2.06832379e-01 -7.66241029e-02 -1.98365673e-01 1.91897422e-01 -3.25147927e-01 -4.48051155e-01 1.60492927e-01 -8.11218262e-01 1.66709751e-01 4.63155717e-01 -6.45621955e-01 -7.42029920e-02 -4.33359653e-01 9.04263854e-01 -2.35364847e-02 -3.35090369e-01 3.26865204e-02 -9.84879136e-01 7.15731144e-01 5.90760112e-01 1.86079264e-01 -1.81564823e-01 -6.00837350e-01 -4.26939189e-01 1.18688829e-01 2.34623224e-01 8.38927984e-01 5.49750805e-01 -1.27802026e+00 -7.20710993e-01 5.24729729e-01 7.78193995e-02 -9.52593759e-02 -2.77627647e-01 2.63693422e-01 -3.86750787e-01 5.36975026e-01 -2.03229621e-01 -1.74817115e-01 -1.62818229e+00 2.13663787e-01 2.42230490e-01 3.40339482e-01 -3.70274842e-01 1.55651855e+00 -2.23676994e-01 -7.79915929e-01 6.87967300e-01 -3.89145434e-01 -3.82237852e-01 3.35294716e-02 7.21244752e-01 5.73278487e-01 4.10855263e-02 -6.46704376e-01 -1.95749611e-01 7.63631761e-02 -1.92166492e-01 -8.11161935e-01 1.48811030e+00 2.21816555e-01 2.34819382e-01 9.34915841e-01 9.98531282e-01 6.75071776e-01 -7.11865306e-01 -3.86672243e-02 -7.04285502e-02 -1.73581645e-01 6.15631975e-02 -9.44797516e-01 -7.67232239e-01 1.26581299e+00 4.67125565e-01 1.96630508e-01 6.84457004e-01 2.22090513e-01 7.43726671e-01 2.17561886e-01 1.55179545e-01 -1.28147471e+00 -5.28366305e-02 9.08043802e-01 7.87264943e-01 -9.91886795e-01 -7.61070967e-01 -3.12604994e-01 -4.35102344e-01 9.19805348e-01 9.04242933e-01 1.95637196e-01 4.51312542e-01 3.76312286e-01 5.23210406e-01 1.56145826e-01 -8.70509684e-01 -4.66358662e-01 5.76347589e-01 8.41454625e-01 1.19426239e+00 5.47421873e-01 -1.01985827e-01 7.98944056e-01 -1.18614995e+00 -2.36389413e-01 6.23762071e-01 7.43762255e-01 -4.77052212e-01 -1.01572967e+00 -1.92385241e-01 2.67788768e-01 -7.55942345e-01 -5.48579454e-01 -6.43316448e-01 6.71831667e-01 -7.41669014e-02 1.33674777e+00 -1.94789678e-01 -4.84032989e-01 6.12031817e-01 4.51912224e-01 8.35471153e-01 -1.45867264e+00 -9.97238576e-01 -2.14091733e-01 3.27047050e-01 -4.93796110e-01 2.61936814e-01 -7.53989935e-01 -9.51451361e-01 9.74140465e-02 -3.33418489e-01 2.76276439e-01 9.60794091e-01 1.02947712e+00 -4.25451458e-01 5.38092911e-01 5.05273283e-01 -4.13579434e-01 -7.10393250e-01 -1.25682712e+00 -3.45969319e-01 -2.27116331e-01 5.77607036e-01 -1.25979453e-01 -2.13892058e-01 -3.63721341e-01]

[14.308941841125488, 6.820274353027344]

7b3ca433-5a2e-4437-8a92-489b6f2c8f7d

efficient-bayesian-inference-using-physics

2304.12541

https://arxiv.org/abs/2304.12541v2

https://arxiv.org/pdf/2304.12541v2.pdf

Efficient Bayesian inference using physics-informed invertible neural networks for inverse problems

In the paper, we propose a novel approach for solving Bayesian inverse problems with physics-informed invertible neural networks (PI-INN). The architecture of PI-INN consists of two sub-networks: an invertible neural network (INN) and a neural basis network (NB-Net). The invertible map between the parametric input and the INN output with the aid of NB-Net is constructed to provide a tractable estimation of the posterior distribution, which enables efficient sampling and accurate density evaluation. Furthermore, the loss function of PI-INN includes two components: a residual-based physics-informed loss term and a new independence loss term. The presented independence loss term can Gaussianize the random latent variables and ensure statistical independence between two parts of INN output by effectively utilizing the estimated density function. Several numerical experiments are presented to demonstrate the efficiency and accuracy of the proposed PI-INN, including inverse kinematics, inverse problems of the 1-d and 2-d diffusion equations, and seismic traveltime tomography.

['Hao Wu', 'Xintong Wang', 'Xiaofei Guan']

2023-04-25

['bayesian-inference']

['methodology']

[ 4.31819819e-02 2.75642842e-01 2.43910566e-01 -3.81538093e-01 -5.25939763e-01 2.46648461e-01 4.05964196e-01 -7.39903331e-01 -3.78686309e-01 9.50255990e-01 6.85140043e-02 -1.93668634e-01 -6.10412896e-01 -1.10167789e+00 -9.08429563e-01 -1.11739624e+00 -9.05144215e-02 7.83084929e-01 2.03948930e-01 1.68618351e-01 -2.71395519e-02 6.50188684e-01 -1.02078795e+00 -3.91897559e-01 9.09350753e-01 1.30097139e+00 1.77292272e-01 2.94850945e-01 5.92965782e-02 9.99416173e-01 -2.23810881e-01 7.91923329e-02 1.10826075e-01 -3.91082734e-01 -3.37799907e-01 -4.64066148e-01 -2.37703979e-01 -8.20365429e-01 -7.59231567e-01 1.07528543e+00 5.39163351e-01 7.61506677e-01 1.20825624e+00 -1.13754320e+00 -7.96189964e-01 5.87570727e-01 -4.74567503e-01 6.87306188e-03 -3.50086510e-01 -1.43170496e-02 5.06976187e-01 -9.17176664e-01 2.50259757e-01 1.44102812e+00 9.53833163e-01 1.88282847e-01 -1.06230986e+00 -8.10267925e-01 -2.22530901e-01 2.21438125e-01 -1.69181931e+00 -2.61897624e-01 9.31038201e-01 -5.17280340e-01 7.71453381e-01 -2.25425899e-01 5.82528591e-01 9.92851317e-01 4.72123176e-01 5.54997683e-01 8.87985528e-01 -2.02893525e-01 5.46566546e-01 7.35514462e-02 8.86225104e-02 6.83578253e-01 9.71462280e-02 4.10071194e-01 -4.27735209e-01 -2.39145815e-01 1.41862583e+00 -5.48424050e-02 -3.70209813e-01 -2.77741253e-01 -6.47129953e-01 9.46558595e-01 6.36472762e-01 -8.68854299e-02 -5.80132365e-01 4.05812413e-01 6.93739951e-02 -3.74358773e-01 4.99983788e-01 -1.56051606e-01 8.17726329e-02 1.75851628e-01 -9.15421009e-01 2.45752335e-01 7.33040452e-01 8.90336275e-01 7.62664080e-01 5.09447634e-01 -2.86311693e-02 8.03158104e-01 9.18170750e-01 8.86231065e-01 3.32013935e-01 -1.21359539e+00 3.30345243e-01 -1.20069854e-01 1.06509589e-01 -1.09233761e+00 -2.03455985e-01 -4.37188953e-01 -1.39257145e+00 3.34368855e-01 2.87199914e-01 -2.18024164e-01 -9.06596839e-01 2.05384040e+00 3.52505803e-01 4.07965422e-01 -4.69758399e-02 9.08459723e-01 6.17097080e-01 1.02911997e+00 -1.90475240e-01 6.55541047e-02 1.01116002e+00 -5.40174127e-01 -8.98624897e-01 -1.56887889e-01 -1.12887904e-01 -1.64807543e-01 7.71300256e-01 2.17100263e-01 -1.33563387e+00 -4.61360842e-01 -1.12967336e+00 -1.74795151e-01 1.61017567e-01 3.47369015e-02 2.80010253e-01 5.98916449e-02 -9.08998191e-01 9.04763043e-01 -1.23508143e+00 3.45555991e-01 3.00616354e-01 3.24798435e-01 -8.79794806e-02 -8.54003988e-03 -1.44780040e+00 7.07400143e-01 2.86285818e-01 6.09770000e-01 -1.06069791e+00 -7.42230058e-01 -9.24712658e-01 3.33910525e-01 8.62689763e-02 -8.05525362e-01 1.02893329e+00 -3.79611820e-01 -1.92822218e+00 4.76812012e-02 5.41198626e-02 -2.55163163e-01 5.65085232e-01 -2.50547320e-01 -1.54377311e-01 2.47180983e-01 2.74922788e-01 4.11774904e-01 9.40687120e-01 -1.21736753e+00 -7.46569335e-02 -3.21789235e-01 -3.62711072e-01 2.41968017e-02 -1.71247482e-01 -6.01155460e-01 -4.94406968e-01 -7.33349919e-01 4.10396338e-01 -6.62318707e-01 -1.41626686e-01 3.10201973e-01 -6.86830819e-01 1.00776693e-02 7.51546323e-01 -1.09082139e+00 9.53332663e-01 -2.21700144e+00 2.00956270e-01 5.42884588e-01 1.79113090e-01 -1.27558619e-01 1.29495293e-01 6.83566704e-02 -1.04195951e-02 -3.13227504e-01 -8.15841973e-01 -2.54130810e-01 7.31762797e-02 3.16091686e-01 -4.66737390e-01 6.33100748e-01 9.65243503e-02 6.34318590e-01 -6.22375906e-01 -3.81747901e-01 1.68640271e-01 1.02718794e+00 -4.09137040e-01 2.99363434e-01 -6.99394345e-02 4.73528504e-01 -7.64335334e-01 3.49981308e-01 8.94087553e-01 -2.40357652e-01 -2.90505677e-01 -3.35387468e-01 -9.99271795e-02 1.24007419e-01 -1.27990246e+00 1.36608410e+00 -5.27668059e-01 3.67228955e-01 4.10519630e-01 -1.01291418e+00 1.01232195e+00 3.17816496e-01 2.00024962e-01 -6.03885174e-01 2.48626918e-01 2.53627211e-01 -2.55321153e-02 -3.60626638e-01 2.67030001e-01 -6.20959640e-01 3.51925492e-01 5.36292493e-01 1.10557497e-01 -1.89371303e-01 -2.08455235e-01 6.31920174e-02 8.18880081e-01 3.16227734e-01 -6.49102405e-02 -4.92641985e-01 6.30317271e-01 -4.32627559e-01 6.86682820e-01 8.48913550e-01 6.79797828e-02 5.43897271e-01 2.86886632e-01 -3.53694916e-01 -1.09148526e+00 -1.46083403e+00 -2.69593358e-01 3.18023980e-01 1.57529816e-01 5.06635785e-01 -6.57125294e-01 -6.62705451e-02 5.10950424e-02 1.12228477e+00 -5.30911922e-01 -4.14059341e-01 -7.72091687e-01 -7.46256113e-01 5.06936252e-01 8.07434261e-01 8.81274104e-01 -1.01195025e+00 -1.58839002e-01 2.12308496e-01 -2.69108355e-01 -7.75368214e-01 -3.13573509e-01 2.95981973e-01 -9.02717292e-01 -5.48924327e-01 -8.00805748e-01 -3.91227037e-01 6.12836897e-01 -4.39837635e-01 6.83933079e-01 -5.78797758e-01 3.11139412e-02 1.14246361e-01 4.13447201e-01 -2.30067268e-01 -3.41401130e-01 -1.56259939e-01 4.82852310e-02 1.36794418e-01 -1.16787024e-01 -1.18449962e+00 -6.17727757e-01 3.29950333e-01 -7.17960656e-01 -5.24198674e-02 7.07237482e-01 8.65559816e-01 7.80494869e-01 2.74359584e-01 3.46201330e-01 -3.28893661e-01 4.70688552e-01 -6.41078413e-01 -8.65491331e-01 4.84687835e-02 -2.66421825e-01 3.56054157e-01 5.67752063e-01 -4.90994960e-01 -1.53966403e+00 -2.56748408e-01 -4.43720996e-01 -6.83861613e-01 2.08783552e-01 5.52072585e-01 -2.69900680e-01 2.06950288e-02 2.75938869e-01 3.03922683e-01 1.76075250e-01 -6.13444746e-01 2.60761559e-01 5.42642593e-01 9.76287842e-01 -7.86717236e-01 7.32050180e-01 6.62561953e-01 3.36867481e-01 -8.63638401e-01 -6.98014975e-01 -5.65387234e-02 -3.04156095e-01 -1.39914230e-01 9.67423558e-01 -8.33602071e-01 -8.08774233e-01 7.41095185e-01 -1.22024310e+00 -4.47946638e-01 -6.73080444e-01 8.71090233e-01 -7.32660055e-01 2.39021987e-01 -9.61356640e-01 -1.07093143e+00 -3.67803752e-01 -1.23846018e+00 8.31689119e-01 2.74757653e-01 7.65150487e-02 -1.11252987e+00 7.92962387e-02 -5.24411909e-02 5.08141041e-01 1.88845962e-01 1.20349264e+00 -2.25868419e-01 -7.77501047e-01 -1.96622729e-01 -4.20193940e-01 6.41277373e-01 -1.96502745e-01 -1.54894769e-01 -9.68763173e-01 -6.76935464e-02 6.97095633e-01 -1.93855882e-01 9.04052436e-01 1.06655025e+00 1.10272121e+00 -2.86260754e-01 -2.58402228e-01 9.04287696e-01 1.32120275e+00 4.04286027e-01 8.14888179e-01 6.31283447e-02 6.49694979e-01 3.91691655e-01 6.06809147e-02 3.77320021e-01 2.20538363e-01 3.10880780e-01 3.65158677e-01 3.71264033e-02 -3.57694514e-02 -1.25425398e-01 2.07594588e-01 1.13124478e+00 -2.68769920e-01 -4.59988177e-01 -7.85740197e-01 3.62944186e-01 -1.70468831e+00 -7.66927481e-01 -6.05391748e-02 2.06972742e+00 5.59414089e-01 2.28594318e-02 -4.64820623e-01 -1.49040744e-01 8.14398646e-01 -8.12936425e-02 -9.86014843e-01 4.77484018e-02 1.01287335e-01 6.57385811e-02 3.04984778e-01 6.90201879e-01 -9.20753956e-01 1.63838521e-01 6.64342880e+00 1.15183175e+00 -8.62005293e-01 2.47380942e-01 7.38102674e-01 1.96000695e-01 -4.24194604e-01 -1.82787493e-01 -8.25620651e-01 5.18852711e-01 7.72176683e-01 3.62609290e-02 3.68884325e-01 8.46604466e-01 3.12081784e-01 -1.99791357e-01 -7.65775442e-01 5.51929712e-01 -2.17722625e-01 -1.23610330e+00 9.48367491e-02 1.45375460e-01 4.42400157e-01 1.13446601e-01 2.08554476e-01 1.53594494e-01 5.22260964e-01 -8.96281302e-01 1.03108633e+00 1.06850874e+00 8.15751016e-01 -7.55933046e-01 6.70798182e-01 4.54136163e-01 -9.67708766e-01 1.19927572e-02 -4.42939907e-01 -4.19925898e-02 7.74659753e-01 9.90149379e-01 -2.47254714e-01 4.62648481e-01 8.21914196e-01 6.39407039e-01 4.30297941e-01 7.66209841e-01 -4.31262434e-01 6.43053293e-01 -7.71002591e-01 2.29707420e-01 3.39242756e-01 -8.52912664e-01 7.78831959e-01 7.33920455e-01 5.87026536e-01 1.18525431e-01 -1.31082699e-01 1.74318755e+00 3.75188887e-02 -4.77182031e-01 -2.78402567e-01 7.98287690e-02 4.98188257e-01 9.20252562e-01 -5.52389681e-01 -2.10051924e-01 -9.43054184e-02 5.07337630e-01 2.51910806e-01 7.80893922e-01 -1.05704343e+00 -4.84620452e-01 4.13222313e-01 2.60075685e-02 3.82398516e-01 -2.64904380e-01 -1.45685926e-01 -6.31420195e-01 5.62398694e-03 -1.33147970e-01 1.15286587e-02 -1.04476368e+00 -1.54548609e+00 4.95679826e-01 4.63527828e-01 -7.46453404e-01 -2.29663804e-01 -7.18227386e-01 -6.97166502e-01 1.34013927e+00 -1.27227616e+00 -8.19146752e-01 -2.69550830e-01 4.09885228e-01 -6.37597144e-02 -1.52405933e-01 4.68676597e-01 3.85319740e-01 -6.09138012e-01 1.77329466e-01 6.54153883e-01 4.50290404e-02 1.83883727e-01 -9.93379295e-01 2.45256692e-01 6.30456984e-01 -5.40480435e-01 6.83071315e-01 5.94158411e-01 -7.88909018e-01 -1.01732707e+00 -1.01758349e+00 2.36278102e-01 1.78932607e-01 7.11508155e-01 -2.07342103e-01 -1.18487537e+00 7.52737105e-01 -2.64014810e-01 8.26019868e-02 1.81019366e-01 -3.85254890e-01 -8.28865469e-02 -6.67933300e-02 -1.18739390e+00 3.02974433e-01 6.61171436e-01 -5.55568576e-01 -4.19717193e-01 2.40842208e-01 7.59881735e-01 -4.34516072e-01 -9.87727821e-01 6.31239831e-01 7.39346027e-01 -8.53975117e-01 1.42835903e+00 4.08415943e-02 7.04807460e-01 -2.17233166e-01 -2.32311636e-01 -9.67529535e-01 -3.62254709e-01 -3.31352890e-01 -3.28674704e-01 1.05322814e+00 8.04887861e-02 -8.50835562e-01 6.61768138e-01 6.28189743e-01 -2.35797733e-01 -7.38835931e-01 -1.25712919e+00 -1.08613586e+00 3.38791132e-01 -5.52356958e-01 4.70816106e-01 4.75690424e-01 -7.10198820e-01 1.89700082e-01 -4.78574216e-01 3.89056563e-01 9.47656870e-01 -4.54996943e-01 2.07556874e-01 -9.81810093e-01 -7.05911398e-01 -3.36611152e-01 -7.41687790e-02 -1.41089535e+00 1.09363556e-01 -6.10757351e-01 5.66746593e-01 -1.54118800e+00 2.10995004e-01 -6.34048283e-01 -2.68966079e-01 6.00232072e-02 1.63197502e-01 7.61558935e-02 -3.20541918e-01 3.79354298e-01 1.57322839e-01 1.23575819e+00 1.37130320e+00 -8.27350169e-02 -9.38643813e-02 1.17588453e-01 -5.00431918e-02 9.99361753e-01 4.87219512e-01 -7.31519759e-01 -6.03920281e-01 -4.16098177e-01 -8.51804912e-02 3.43922257e-01 6.92421734e-01 -1.01572418e+00 3.85242611e-01 -1.32896183e-02 3.47836614e-01 -8.72647703e-01 6.63750470e-01 -6.15318596e-01 4.22661424e-01 4.01434630e-01 -8.70711356e-02 -2.40741655e-01 9.68223736e-02 7.52348304e-01 -2.24322230e-01 -4.27892387e-01 8.55735481e-01 -5.01362048e-02 -2.54543751e-01 3.74316067e-01 -4.51830000e-01 -7.59794936e-02 4.81843799e-01 2.20786571e-03 -8.07723403e-02 -6.15217447e-01 -7.77011871e-01 -3.87684405e-02 -1.33179307e-01 -3.66495043e-01 7.03479767e-01 -1.37846649e+00 -4.16151047e-01 3.57373774e-01 -5.84232152e-01 4.27228093e-01 6.34660602e-01 9.35238481e-01 -6.89999461e-01 1.72768533e-01 -3.37642282e-02 -7.13358104e-01 -2.25155607e-01 1.55498639e-01 7.03019202e-01 -4.02442575e-01 -9.76832449e-01 1.13664234e+00 6.00297928e-01 -8.03184211e-01 2.98911870e-01 -3.58969420e-01 1.88130111e-01 -5.31173289e-01 4.19719279e-01 8.11886311e-01 -3.24712098e-01 -6.45869732e-01 -1.16089016e-01 4.47290421e-01 2.56795734e-01 -5.00689447e-01 1.48106182e+00 -9.94217098e-02 -4.14903134e-01 5.22917986e-01 1.18068695e+00 -5.23271203e-01 -1.68473232e+00 -3.02628666e-01 -5.45810282e-01 2.85071488e-02 4.78840113e-01 -6.07675970e-01 -1.15878987e+00 1.07963574e+00 4.47538078e-01 -1.70193270e-01 9.68437374e-01 -1.81288421e-01 8.03951621e-01 4.53525633e-01 8.86931643e-02 -8.92581701e-01 2.42495947e-02 7.35046685e-01 1.04772830e+00 -6.97783828e-01 1.66441172e-01 -3.08443815e-01 -4.28182706e-02 9.68940735e-01 4.58003044e-01 -3.12856108e-01 1.18300247e+00 4.77859497e-01 -1.88723832e-01 -1.83744341e-01 -1.00821406e-01 4.85262901e-01 3.42400849e-01 5.09715497e-01 -7.32798949e-02 -3.04005325e-01 1.19122773e-01 8.12530816e-01 -6.79524168e-02 2.90223621e-02 8.31202045e-02 8.12031925e-01 -3.19546402e-01 -5.00669181e-01 -5.19817829e-01 3.08957815e-01 1.19295493e-02 -1.22516051e-01 3.90201211e-01 9.53038692e-01 -2.09568683e-02 3.47757518e-01 3.37245435e-01 -8.73125941e-02 1.02871917e-01 -8.56234357e-02 3.29137802e-01 -9.33451578e-02 1.75071254e-01 2.88776577e-01 -3.33794326e-01 -4.70119715e-01 -2.13437960e-01 -5.14007926e-01 -1.38058901e+00 -3.56625289e-01 -5.10694504e-01 1.92069247e-01 6.79038525e-01 1.21600246e+00 9.45939794e-02 7.88521886e-01 1.29424706e-01 -1.17031908e+00 -8.18798780e-01 -1.09220099e+00 -1.07097340e+00 -1.47128016e-01 3.33507538e-01 -1.03224766e+00 -5.91118038e-01 -1.30577311e-01]

[6.953646183013916, 3.590203046798706]

abef3298-ecee-4839-b448-4af57e4dcb67

implicit-ray-transformers-for-multi-view

2303.08401

https://arxiv.org/abs/2303.08401v1

https://arxiv.org/pdf/2303.08401v1.pdf

Implicit Ray-Transformers for Multi-view Remote Sensing Image Segmentation

The mainstream CNN-based remote sensing (RS) image semantic segmentation approaches typically rely on massive labeled training data. Such a paradigm struggles with the problem of RS multi-view scene segmentation with limited labeled views due to the lack of considering 3D information within the scene. In this paper, we propose ''Implicit Ray-Transformer (IRT)'' based on Implicit Neural Representation (INR), for RS scene semantic segmentation with sparse labels (such as 4-6 labels per 100 images). We explore a new way of introducing multi-view 3D structure priors to the task for accurate and view-consistent semantic segmentation. The proposed method includes a two-stage learning process. In the first stage, we optimize a neural field to encode the color and 3D structure of the remote sensing scene based on multi-view images. In the second stage, we design a Ray Transformer to leverage the relations between the neural field 3D features and 2D texture features for learning better semantic representations. Different from previous methods that only consider 3D prior or 2D features, we incorporate additional 2D texture information and 3D prior by broadcasting CNN features to different point features along the sampled ray. To verify the effectiveness of the proposed method, we construct a challenging dataset containing six synthetic sub-datasets collected from the Carla platform and three real sub-datasets from Google Maps. Experiments show that the proposed method outperforms the CNN-based methods and the state-of-the-art INR-based segmentation methods in quantitative and qualitative metrics.

['Zhengxia Zou', 'Zhenwei Shi', 'Chenyang Liu', 'Hao Chen', 'Zipeng Qi']

2023-03-15

['scene-segmentation']

['computer-vision']

[ 4.41457987e-01 -8.05715621e-02 -6.05394430e-02 -6.70543909e-01 -8.97378385e-01 -6.72519743e-01 4.49269205e-01 -1.93919957e-01 -2.59173781e-01 3.01105052e-01 -2.27371417e-02 -1.26901716e-01 -2.63127804e-01 -1.21272027e+00 -9.25832629e-01 -6.08905733e-01 2.88062572e-01 4.71107811e-01 4.16123092e-01 -1.21336646e-01 2.80921370e-01 6.33045912e-01 -1.53144932e+00 1.94539160e-01 8.34322512e-01 1.21052790e+00 4.63596433e-01 2.86193073e-01 -2.68882692e-01 5.93848348e-01 6.84421137e-02 1.22672366e-02 7.41177738e-01 -2.09126443e-01 -9.66875494e-01 4.84048843e-01 4.91434276e-01 -4.21167225e-01 9.01709199e-02 1.11005008e+00 3.51573497e-01 3.03824067e-01 6.09338403e-01 -8.29591990e-01 -5.02466142e-01 1.50056601e-01 -9.98332202e-01 -8.91108960e-02 1.09599411e-01 -1.89425066e-01 9.83057320e-01 -9.12483275e-01 5.03925085e-01 1.20304847e+00 8.14392507e-01 3.04613322e-01 -9.83381867e-01 -5.15424192e-01 4.13076818e-01 -3.82108726e-02 -1.35545194e+00 -1.01273760e-01 1.16110682e+00 -6.15906894e-01 5.84603012e-01 1.91201493e-01 6.93683326e-01 6.11446857e-01 -2.29049981e-01 5.89022517e-01 1.65981162e+00 -2.13576630e-01 3.13145876e-01 -8.98981169e-02 6.78677633e-02 8.40746939e-01 -1.62956402e-01 7.87922665e-02 -3.10766608e-01 1.19961239e-01 1.04197538e+00 4.10270989e-01 -2.01334462e-01 -2.56171405e-01 -9.98668790e-01 9.75336552e-01 6.81234837e-01 -6.43586963e-02 -2.89825976e-01 1.73056513e-01 7.16276541e-02 -1.94307208e-01 1.05979586e+00 1.20955803e-01 -7.22060621e-01 5.16391039e-01 -9.26277459e-01 3.52988429e-02 3.66930395e-01 7.14090645e-01 1.44977009e+00 -4.58872914e-02 2.16710612e-01 1.09122419e+00 5.14009833e-01 6.79603577e-01 -3.79120186e-02 -1.12417066e+00 3.82693440e-01 8.21550369e-01 5.08061498e-02 -8.45963836e-01 -4.31778044e-01 -2.91720539e-01 -7.31285155e-01 1.46912351e-01 1.66267887e-01 1.01075426e-01 -1.34198177e+00 1.29708099e+00 6.41629994e-01 1.02883294e-01 1.28372032e-02 1.01918495e+00 9.19567883e-01 7.13894367e-01 -2.05532685e-01 1.57184631e-01 1.23025644e+00 -1.12529588e+00 -1.77464783e-01 -4.26779240e-01 3.06041569e-01 -6.65220976e-01 1.13126731e+00 3.17303896e-01 -7.16902673e-01 -5.30528247e-01 -8.04335356e-01 -1.05250433e-01 -5.04562378e-01 2.64726043e-01 8.82617235e-01 6.27516508e-01 -8.12592089e-01 3.78735155e-01 -7.53161132e-01 -3.50304008e-01 6.60331726e-01 9.95789468e-03 -2.76552349e-01 -4.78788704e-01 -1.01012576e+00 3.87227565e-01 2.70453244e-01 4.21701074e-01 -1.14722800e+00 -7.58757532e-01 -9.35190856e-01 -1.83590025e-01 5.16532123e-01 -6.53506279e-01 8.81070554e-01 -9.57163215e-01 -1.47762513e+00 1.12528956e+00 2.55761091e-02 2.39310056e-01 4.56461191e-01 -2.80484885e-01 1.06536694e-01 3.98040265e-01 3.69446158e-01 8.03194582e-01 5.52139103e-01 -1.89584720e+00 -6.40816271e-01 -7.22329617e-01 5.06299734e-01 5.26744246e-01 2.43374124e-01 -2.67779857e-01 -4.65856761e-01 -4.79218602e-01 6.58870459e-01 -8.07012677e-01 -4.54939574e-01 1.83433563e-01 -5.60731530e-01 1.28060490e-01 7.59334087e-01 -6.16046011e-01 2.81529337e-01 -1.98770857e+00 -1.69015322e-02 1.85573682e-01 2.12963056e-02 -9.60549265e-02 5.22220097e-02 1.62349731e-01 1.42346621e-01 2.49322250e-01 -5.77621043e-01 -2.97107190e-01 -2.59251624e-01 4.24985141e-01 -1.36390075e-01 7.38486588e-01 6.35546967e-02 6.68843389e-01 -8.64697278e-01 -3.94597530e-01 4.58138168e-01 4.72231030e-01 -5.77607393e-01 3.66117388e-01 -4.67254281e-01 8.27896953e-01 -8.50558758e-01 8.38851810e-01 1.11450768e+00 -4.23349082e-01 -1.85318530e-01 -3.52934629e-01 -3.61929774e-01 1.61425211e-02 -1.01626992e+00 1.90457070e+00 -6.91441000e-01 1.99915946e-01 1.68360010e-01 -1.30202174e+00 1.00353169e+00 7.58110965e-03 7.29613364e-01 -7.11219490e-01 2.52587181e-02 8.19365829e-02 -7.21887171e-01 -5.94509065e-01 2.48111621e-01 -3.24892938e-01 1.07854009e-01 3.99077326e-01 -1.75944135e-01 -6.68234527e-01 -2.75019765e-01 -1.14269279e-01 4.74568188e-01 5.46411693e-01 1.76233845e-03 -1.27836749e-01 5.67097664e-01 1.10182777e-01 7.01348424e-01 5.50714612e-01 2.10829899e-01 9.65842605e-01 1.71388552e-01 -5.28205574e-01 -9.59738910e-01 -9.48759615e-01 -2.03817382e-01 9.37331319e-01 5.34857690e-01 1.24244675e-01 -6.45810604e-01 -7.64246881e-01 -2.34156489e-01 2.81106502e-01 -7.31993437e-01 4.29171771e-01 -3.24854493e-01 -8.83081734e-01 2.85742611e-01 5.09984553e-01 1.15786183e+00 -6.07532144e-01 -7.79839158e-01 -1.21954270e-01 -3.92754972e-01 -1.41052020e+00 -1.50298581e-01 2.14367494e-01 -9.70391810e-01 -1.13862586e+00 -6.03366315e-01 -5.41482449e-01 7.76228547e-01 7.05438614e-01 9.28225398e-01 -1.78403020e-01 -5.14299907e-02 4.58220541e-01 -6.80607259e-01 -8.20479244e-02 1.36705071e-01 7.63616115e-02 -5.96103430e-01 1.86828718e-01 -8.47344026e-02 -7.24758565e-01 -7.88359284e-01 4.31597561e-01 -1.12479079e+00 5.57025611e-01 6.10812068e-01 6.52658582e-01 9.27605271e-01 8.61434489e-02 3.04621786e-01 -1.26912105e+00 -3.21306854e-01 -4.86114800e-01 -7.43729591e-01 2.62755036e-01 -5.15851140e-01 -1.93887651e-01 4.40614372e-01 9.82074514e-02 -1.29799473e+00 3.81811887e-01 -2.80448943e-01 -3.86011511e-01 -4.92628276e-01 6.65540695e-01 -2.60035455e-01 -2.30226159e-01 3.85963470e-01 2.62158871e-01 -2.57335186e-01 -5.89514852e-01 5.10950208e-01 6.61727667e-01 1.36351243e-01 -6.40156865e-01 6.74862742e-01 1.09138811e+00 -1.01399854e-01 -7.93298542e-01 -1.50416970e+00 -8.05700481e-01 -7.59732544e-01 -4.22501653e-01 1.24946606e+00 -1.26489532e+00 -3.00508678e-01 6.68923378e-01 -1.03815746e+00 -4.83867913e-01 -1.35260031e-01 4.74421144e-01 -8.03898573e-01 3.72866154e-01 -4.09423202e-01 -7.38355458e-01 -2.56703794e-01 -1.21297657e+00 1.69414675e+00 1.81451365e-01 7.09663749e-01 -9.94443834e-01 -5.51045239e-02 9.12038743e-01 3.03335547e-01 4.96838123e-01 8.32400918e-01 -1.70456067e-01 -8.37636411e-01 3.69898021e-01 -6.49805903e-01 5.74822426e-01 1.44534707e-01 -3.13028187e-01 -1.34039211e+00 7.28893504e-02 1.79628164e-01 -4.79656518e-01 9.47401106e-01 5.42580664e-01 1.36323190e+00 -8.07605311e-02 -1.53732330e-01 1.12169921e+00 1.91473293e+00 -1.56805348e-02 5.68824768e-01 3.40944976e-01 1.20488584e+00 8.93783808e-01 5.85990727e-01 3.79908562e-01 7.79450655e-01 3.95458728e-01 7.50435889e-01 -4.56287682e-01 1.08625432e-02 -2.66097695e-01 3.99903394e-02 7.40223289e-01 -3.93741339e-01 -2.36343056e-01 -9.09545898e-01 4.81179237e-01 -1.63188827e+00 -6.58232629e-01 -1.81046426e-01 1.88010597e+00 5.00722051e-01 -2.12103471e-01 -2.75293976e-01 5.47543578e-02 6.07015371e-01 4.51380759e-01 -6.49695635e-01 1.59922659e-01 -7.04971403e-02 1.38140872e-01 8.82082820e-01 5.88500261e-01 -1.17529094e+00 1.11138487e+00 5.19920540e+00 7.75241137e-01 -1.32169116e+00 2.44997710e-01 9.45691347e-01 3.67619276e-01 -6.83018148e-01 1.88272789e-01 -7.03219652e-01 7.66628161e-02 2.90366650e-01 6.67348802e-01 5.49339294e-01 6.69690251e-01 3.75331730e-01 -4.18979347e-01 -6.93860412e-01 9.77497101e-01 1.16859347e-01 -1.37219536e+00 1.41730428e-01 3.98074575e-02 1.08029377e+00 5.11067688e-01 -1.07749470e-01 -1.60523579e-01 3.41293424e-01 -9.46915805e-01 9.30615664e-01 6.37746871e-01 9.60737407e-01 -5.58838189e-01 4.89192814e-01 3.93270016e-01 -1.38640404e+00 -2.84388885e-02 -2.29928523e-01 9.33511555e-02 2.14306682e-01 8.37459445e-01 -3.88059944e-01 9.66301084e-01 9.52822506e-01 1.13683844e+00 -5.55036724e-01 6.31296635e-01 -1.68318838e-01 5.43997824e-01 -3.74923229e-01 4.55052346e-01 4.93933022e-01 -5.93188643e-01 1.55372694e-01 7.32125461e-01 2.54288435e-01 3.11996073e-01 3.99681240e-01 1.08189273e+00 1.11896873e-01 -5.87140061e-02 -6.47468865e-01 2.03521267e-01 1.86185047e-01 1.39264345e+00 -1.22954607e+00 -1.09075330e-01 -4.69503522e-01 8.31294298e-01 2.08127707e-01 6.31061494e-01 -7.48020172e-01 8.06785524e-02 1.90647058e-02 1.99995518e-01 4.62671667e-01 -4.13438112e-01 -3.77306789e-01 -1.17539155e+00 -6.16542250e-02 -5.38422346e-01 1.76233143e-01 -1.14603150e+00 -1.33958709e+00 4.67424124e-01 2.71513343e-01 -1.25575161e+00 2.71650493e-01 -5.90601385e-01 -3.58138323e-01 8.43178928e-01 -2.11718488e+00 -1.86451888e+00 -6.86604798e-01 5.48583210e-01 7.15368748e-01 3.50400031e-01 5.73466480e-01 3.06495726e-01 -3.12575996e-01 -2.07408622e-01 7.82488510e-02 9.16226432e-02 2.59603649e-01 -1.07377613e+00 1.88576058e-01 6.60544574e-01 -1.14075594e-01 1.47949785e-01 1.28825441e-01 -4.76757914e-01 -1.17634559e+00 -1.46390772e+00 2.15843201e-01 -1.54977947e-01 2.21739829e-01 -3.29242349e-01 -6.87050164e-01 6.23678207e-01 -1.20729931e-01 4.27060902e-01 6.47190571e-01 -2.13845387e-01 -2.54609495e-01 -1.52182415e-01 -1.23001075e+00 3.33896950e-02 1.19155097e+00 -6.64780974e-01 -2.84742922e-01 3.93035144e-01 6.93130374e-01 -4.43599284e-01 -9.64616597e-01 7.85771072e-01 4.59280759e-01 -1.17378879e+00 1.13446999e+00 -6.59616217e-02 7.40034163e-01 -4.38459069e-01 -6.55946016e-01 -1.21964300e+00 3.46459001e-02 2.56098844e-02 6.09144509e-01 1.10031259e+00 1.44501626e-01 -5.87756217e-01 8.36539924e-01 4.33434024e-02 -4.49397326e-01 -8.71412933e-01 -8.15415978e-01 -4.02362108e-01 2.94135138e-03 -5.48368394e-01 5.27362406e-01 1.12617230e+00 -1.00371075e+00 3.51497144e-01 -1.11632228e-01 5.40781081e-01 7.07049131e-01 5.96829295e-01 6.17327332e-01 -1.35895264e+00 -1.92357585e-01 -1.79798156e-01 -5.72236143e-02 -1.27428842e+00 1.52346298e-01 -8.78630519e-01 2.26526648e-01 -1.87275374e+00 3.36174935e-01 -9.23951387e-01 1.03274383e-01 3.14435780e-01 -3.02760582e-03 3.97659808e-01 -1.44315004e-01 1.75944731e-01 -5.72895408e-01 9.29061532e-01 1.45989168e+00 -2.22054273e-01 5.98132685e-02 -1.04893863e-01 -5.16291082e-01 9.69473541e-01 6.69235885e-01 -4.86673683e-01 -5.56400597e-01 -8.49067092e-01 5.14408112e-01 1.28938138e-01 6.68895423e-01 -7.38308549e-01 5.33303879e-02 -4.74511504e-01 2.90129542e-01 -9.75187421e-01 5.22414327e-01 -8.73181403e-01 6.73944801e-02 -1.81131084e-02 -7.09798634e-02 -2.70847797e-01 -1.17668077e-01 7.74881423e-01 -2.67181814e-01 -1.52868629e-01 8.80225122e-01 -6.51045144e-01 -8.58518720e-01 5.92526913e-01 -2.88193226e-02 1.71458617e-01 7.64149845e-01 -3.99419039e-01 -2.46717662e-01 -1.27847791e-01 -5.72757840e-01 2.38914117e-01 6.05802655e-01 8.25716555e-02 7.79157937e-01 -1.02481318e+00 -5.75024724e-01 3.04882199e-01 2.33384460e-01 9.00050461e-01 5.64400911e-01 6.83006465e-01 -9.23214018e-01 7.00473636e-02 -2.21011564e-01 -9.68149126e-01 -6.72701359e-01 -4.25224379e-03 5.96073687e-01 -5.36097810e-02 -6.42989397e-01 8.28292429e-01 5.94764709e-01 -1.10102046e+00 -2.15453804e-01 -4.51820552e-01 -3.62659156e-01 -4.26062681e-02 -2.18001641e-02 2.94705600e-01 5.72228841e-02 -7.99959838e-01 -1.98695242e-01 1.18158662e+00 2.34927163e-01 -2.92666465e-01 1.74379909e+00 -4.13204938e-01 -2.69636631e-01 6.57437086e-01 1.14655554e+00 -7.57266283e-02 -1.59227479e+00 -1.91272601e-01 -3.74672890e-01 -7.41696477e-01 4.52656329e-01 -5.89317024e-01 -1.55386806e+00 1.03672814e+00 7.04298735e-01 2.20273975e-02 1.12043619e+00 9.35083628e-02 7.74048865e-01 1.84185222e-01 5.25372386e-01 -9.34961438e-01 1.79408714e-01 6.04076266e-01 5.39522827e-01 -1.51068962e+00 2.13226393e-01 -7.99362540e-01 -4.92105931e-01 9.57195759e-01 6.81837380e-01 -2.17238098e-01 9.35537279e-01 -2.44409770e-01 2.94388264e-01 -6.61096036e-01 -1.80260047e-01 -3.37828189e-01 2.81055242e-01 5.57125628e-01 1.95885241e-01 -2.87450459e-02 1.67875096e-01 3.16974670e-01 9.32224244e-02 -1.16088346e-01 4.22200114e-01 8.28368068e-01 -3.36003333e-01 -7.04572499e-01 -3.66635770e-01 3.08432221e-01 -2.96343952e-01 -6.01067655e-02 -1.06499434e-01 5.99519491e-01 5.07834136e-01 8.28569829e-01 1.27289116e-01 -4.01217312e-01 1.68213755e-01 -4.51740742e-01 5.37400723e-01 -8.35755706e-01 -2.83168197e-01 3.41127455e-01 -1.76618710e-01 -5.78991413e-01 -1.20162165e+00 -5.45610726e-01 -1.22110558e+00 3.52621526e-02 -3.86856079e-01 -1.01760454e-01 8.13523889e-01 1.13504624e+00 4.53148112e-02 4.04834390e-01 1.01988173e+00 -1.20416343e+00 -1.35852203e-01 -6.72543943e-01 -8.20618629e-01 1.92781419e-01 3.13873559e-01 -9.30148900e-01 -4.03184682e-01 1.59885302e-01]

[8.889982223510742, -2.831312894821167]

076b913d-ac5e-4e96-bb7a-bb411606c791

nlg-evaluation-metrics-beyond-correlation

2305.08566

https://arxiv.org/abs/2305.08566v4

https://arxiv.org/pdf/2305.08566v4.pdf

NLG Evaluation Metrics Beyond Correlation Analysis: An Empirical Metric Preference Checklist

In this study, we analyze automatic evaluation metrics for Natural Language Generation (NLG), specifically task-agnostic metrics and human-aligned metrics. Task-agnostic metrics, such as Perplexity, BLEU, BERTScore, are cost-effective and highly adaptable to diverse NLG tasks, yet they have a weak correlation with human. Human-aligned metrics (CTC, CtrlEval, UniEval) improves correlation level by incorporating desirable human-like qualities as training objective. However, their effectiveness at discerning system-level performance and quality of system outputs remain unclear. We present metric preference checklist as a framework to assess the effectiveness of automatic metrics in three NLG tasks: Text Summarization, Dialogue Response Generation, and Controlled Generation. Our proposed framework provides access: (i) for verifying whether automatic metrics are faithful to human preference, regardless of their correlation level to human; and (ii) for inspecting the strengths and limitations of NLG systems via pairwise evaluation. We show that automatic metrics provide a better guidance than human on discriminating system-level performance in Text Summarization and Controlled Generation tasks. We also show that multi-aspect human-aligned metric (UniEval) is not necessarily dominant over single-aspect human-aligned metrics (CTC, CtrlEval) and task-agnostic metrics (BLEU, BERTScore), particularly in Controlled Generation tasks.

['Mykola Pechenizkiy', 'Vlado Menkovski', 'Meng Fang', "Iftitahu Ni'mah"]

2023-05-15

['dialogue-generation', 'response-generation', 'text-summarization', 'controllable-language-modelling', 'dialogue-generation']

['natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'natural-language-processing', 'speech']

[ 1.49422780e-01 2.71084726e-01 1.23101035e-02 -3.17146063e-01 -1.11263180e+00 -1.01423073e+00 1.07759356e+00 5.12475967e-01 -5.38041115e-01 1.02207398e+00 5.53568363e-01 -2.71787345e-01 -3.50553304e-01 -3.49890888e-01 7.74356797e-02 -2.90902317e-01 1.57899484e-01 7.21249938e-01 -6.26035705e-02 -6.10865831e-01 8.28746021e-01 1.48907200e-01 -1.21300566e+00 2.62586087e-01 1.43772709e+00 6.70765221e-01 1.01646826e-01 1.21276975e+00 5.93457930e-02 7.59844542e-01 -1.20651019e+00 -5.58315933e-01 -1.58865869e-01 -7.54422486e-01 -1.21008539e+00 -3.00624013e-01 4.37355816e-01 -5.88475540e-02 2.38895908e-01 8.19860697e-01 9.36770916e-01 -7.63766840e-02 1.15061057e+00 -1.28506982e+00 -7.63679087e-01 7.49512672e-01 -4.11995873e-02 2.14115456e-01 1.03248501e+00 5.42571723e-01 1.26899421e+00 -6.23536170e-01 6.00961924e-01 1.36131954e+00 5.89525402e-01 5.34312785e-01 -9.58939314e-01 -2.34868675e-01 -7.00053200e-02 -3.28134239e-01 -9.21754003e-01 -3.45627874e-01 1.95973724e-01 -7.60636806e-01 1.11410451e+00 6.36687458e-01 1.39723867e-01 1.17404759e+00 3.67985368e-01 7.55800903e-01 1.14275265e+00 -4.55148876e-01 4.67549302e-02 3.19201887e-01 4.20300424e-01 4.29078937e-01 4.39477623e-01 -2.16740146e-01 -5.40119588e-01 -2.90154666e-01 4.00096714e-01 -8.12453449e-01 -5.67753434e-01 2.63027668e-01 -1.80696547e+00 8.05005968e-01 -7.25302845e-02 6.42463088e-01 -3.34330678e-01 -1.60529092e-01 5.32014132e-01 5.13803542e-01 3.74428451e-01 1.34168363e+00 -5.60040593e-01 -6.81885064e-01 -9.08362925e-01 4.74543869e-01 1.18859076e+00 1.26027155e+00 5.29729724e-01 1.35523930e-01 -1.24282563e+00 8.38469446e-01 -1.34092271e-01 5.22751451e-01 1.16652238e+00 -9.68143106e-01 6.49831831e-01 8.36075306e-01 3.65440071e-01 -9.59193587e-01 -6.14800990e-01 -1.91299096e-01 -9.27626729e-01 -2.03938469e-01 3.62081409e-01 -3.03172678e-01 -6.38353646e-01 1.73636413e+00 -2.07729787e-01 -1.09010541e+00 1.61009222e-01 8.08835864e-01 1.10099244e+00 5.59600353e-01 8.92660245e-02 -4.91797179e-01 1.34807849e+00 -1.02110803e+00 -7.16938794e-01 -1.22303359e-01 8.40359747e-01 -1.12667465e+00 1.52184510e+00 3.19059402e-01 -1.28667474e+00 -7.90797234e-01 -9.48282599e-01 3.30411047e-02 -4.13250327e-01 3.66521508e-01 2.59652466e-01 8.39610219e-01 -1.20062506e+00 7.47159362e-01 -2.16450021e-01 -6.05349660e-01 -4.99452472e-01 1.30138576e-01 -2.10473716e-01 5.51630914e-01 -1.27001643e+00 1.27292073e+00 5.10904074e-01 -3.16614151e-01 -5.62227190e-01 -2.40888774e-01 -6.83356106e-01 -7.63975754e-02 -1.49266882e-04 -8.70158613e-01 1.60101855e+00 -6.66091740e-01 -1.48268974e+00 7.43227303e-01 1.68992296e-01 -3.44935983e-01 9.00333166e-01 -1.38004854e-01 -5.66396117e-01 8.06400031e-02 3.50556076e-01 5.19030869e-01 3.80103409e-01 -1.04315579e+00 -6.68270707e-01 9.91051644e-03 1.38530806e-01 4.81051713e-01 -1.55155733e-01 3.82612869e-02 1.86765507e-01 -6.27547622e-01 -1.55841246e-01 -6.02366328e-01 2.45787185e-02 -6.34902239e-01 -5.30419052e-01 -6.46976352e-01 4.43537652e-01 -7.49842942e-01 1.74741817e+00 -1.49976170e+00 3.62602179e-03 -1.53384089e-01 3.29740107e-01 5.62250495e-01 -3.49194705e-01 9.27615523e-01 3.23832154e-01 6.02236986e-01 -2.68023223e-01 5.72547652e-02 3.03028107e-01 -2.65027642e-01 1.55180857e-01 -1.39599994e-01 3.08579803e-01 9.91444051e-01 -1.30835140e+00 -7.05907881e-01 -1.32467076e-01 -1.82535738e-01 -3.46943527e-01 4.08006757e-01 -1.35560811e-01 2.15433002e-01 -3.81588340e-01 5.14106154e-01 9.02749896e-02 -1.80303082e-01 -8.05029273e-02 -2.38055393e-01 -2.13203534e-01 3.83410066e-01 -8.21107686e-01 1.27903783e+00 -5.48985600e-01 7.40999758e-01 -3.49780351e-01 -3.98178518e-01 1.13911271e+00 5.52151978e-01 -1.70124203e-01 -6.68716073e-01 1.42244950e-01 4.36422527e-01 2.89261281e-01 -4.87290293e-01 1.10907066e+00 1.99976131e-01 -4.55366582e-01 7.34918475e-01 3.74788433e-01 -4.50458914e-01 7.96002388e-01 5.60552478e-01 1.13284981e+00 -1.68723121e-01 8.97898912e-01 -5.00517547e-01 7.72360325e-01 -1.19203992e-01 2.09315106e-01 1.09385204e+00 -3.01624298e-01 8.06169033e-01 7.74932206e-01 4.29094620e-02 -9.71441269e-01 -8.66324723e-01 2.13860527e-01 1.07908857e+00 -5.47268465e-02 -7.00607359e-01 -8.85261536e-01 -8.47712696e-01 -2.68659830e-01 1.06042993e+00 -4.41946834e-01 -1.52338028e-01 -3.26575130e-01 -6.33676648e-01 8.77497911e-01 3.60820204e-01 5.30928910e-01 -1.10798359e+00 -6.34039164e-01 2.80373067e-01 -6.90864801e-01 -9.91076171e-01 -8.62303555e-01 -1.72279999e-01 -7.07874179e-01 -1.04450309e+00 -8.43043208e-01 -4.94070858e-01 3.75736356e-01 1.27529502e-01 1.57067585e+00 1.14638999e-01 1.48695707e-01 6.89291775e-01 -6.45762503e-01 -2.98480988e-01 -9.69708025e-01 3.61720324e-01 1.25702262e-01 -3.93307954e-01 6.58326447e-02 -2.04559684e-01 -8.12422633e-01 6.63486719e-01 -7.08723187e-01 -6.05230778e-02 7.69608915e-01 8.41274738e-01 -5.88396154e-02 -7.75169969e-01 1.05364728e+00 -9.95484233e-01 1.79292822e+00 -9.20473859e-02 7.49918967e-02 5.61366141e-01 -1.19866037e+00 7.09159598e-02 7.01724589e-01 -3.89423937e-01 -9.34044480e-01 -7.07604587e-01 1.84570804e-01 3.39695811e-01 -4.83823121e-02 5.76523840e-01 -3.58543731e-02 2.38833159e-01 1.29791212e+00 9.33991224e-02 -2.97723830e-01 -1.66072518e-01 4.00380313e-01 9.80805516e-01 5.14430165e-01 -7.11807549e-01 5.79448283e-01 -4.47357625e-01 -4.56109047e-01 -7.31811404e-01 -6.85220242e-01 -5.98895013e-01 -6.90138340e-01 -2.99287766e-01 8.67972970e-01 -6.19304776e-01 -5.54505825e-01 3.03788811e-01 -1.42554295e+00 -1.58555984e-01 -2.87543952e-01 4.16014612e-01 -7.57519484e-01 6.22486115e-01 -4.42716539e-01 -9.87963021e-01 -1.02928722e+00 -8.64533246e-01 1.20360053e+00 2.14303628e-01 -1.08356547e+00 -1.13051713e+00 2.16100246e-01 5.52889287e-01 5.17709494e-01 4.34627324e-01 9.20291543e-01 -1.15766311e+00 1.59034654e-01 -3.76717806e-01 -3.42700660e-01 5.08098781e-01 2.35721409e-01 1.79091066e-01 -7.63772786e-01 -2.15782881e-01 -1.08481243e-01 -4.13156301e-01 4.40019816e-01 8.25894624e-02 5.55883467e-01 -8.21678936e-01 8.77451375e-02 -1.45537198e-01 8.89480114e-01 1.58682942e-01 5.38070917e-01 2.21375287e-01 4.42204088e-01 6.62933230e-01 8.92961442e-01 3.96434635e-01 2.96730399e-01 5.21697342e-01 -1.98921278e-01 1.30744055e-01 -1.00205719e-01 -3.70357603e-01 5.49957931e-01 1.14576960e+00 -3.67414534e-01 -7.92701304e-01 -8.80296350e-01 4.88339603e-01 -1.85163713e+00 -9.42134380e-01 -4.75170970e-01 2.21851134e+00 9.53039944e-01 2.08646774e-01 3.06109637e-01 1.36082903e-01 8.54166806e-01 1.63695946e-01 -4.29357253e-02 -9.62354779e-01 -2.89556384e-01 -1.06788434e-01 8.63805339e-02 4.33380395e-01 -7.13620245e-01 8.98850977e-01 6.67315102e+00 9.98945773e-01 -7.12177157e-01 5.74731603e-02 5.35475016e-01 3.28581870e-01 -4.19155389e-01 -4.63749235e-03 -5.61038256e-01 4.63539302e-01 1.04522371e+00 -8.43095541e-01 -1.64423157e-02 6.36967480e-01 3.24587077e-01 -7.40252137e-02 -1.50969982e+00 7.64643908e-01 2.64646024e-01 -9.11351740e-01 4.77838099e-01 -6.10349998e-02 9.52585280e-01 -4.22816306e-01 -1.95954695e-01 5.34215510e-01 3.38167131e-01 -9.40998316e-01 6.98565066e-01 3.92647713e-01 7.05330551e-01 -4.86077517e-01 1.08962536e+00 4.41396952e-01 -7.46021807e-01 3.44585925e-01 -3.13000083e-01 -7.53518417e-02 2.50580966e-01 4.38466847e-01 -1.15272141e+00 8.18962634e-01 8.22639763e-02 1.80185035e-01 -7.76910245e-01 8.13718736e-01 -4.20185745e-01 3.98426563e-01 2.76807815e-01 -4.92295116e-01 3.95385861e-01 -1.22904845e-01 7.56662130e-01 1.70667708e+00 4.28722918e-01 3.35699953e-02 -3.72046158e-02 7.34081089e-01 1.33857951e-01 4.42474157e-01 -5.60161173e-01 -3.27553153e-01 5.04363060e-01 1.44074166e+00 -4.64541256e-01 -3.57917607e-01 1.94094792e-01 1.03688061e+00 1.07114017e-01 1.51027411e-01 -6.14813149e-01 -6.67589903e-01 1.51897252e-01 -6.38560504e-02 -4.24801171e-01 -1.23014160e-01 -4.24472272e-01 -9.28160310e-01 1.05054053e-02 -1.25502574e+00 4.32712644e-01 -7.48400271e-01 -1.29525244e+00 1.03431964e+00 6.63043335e-02 -1.62166321e+00 -7.20008314e-01 -5.30617356e-01 -9.55907583e-01 8.36297870e-01 -8.61924887e-01 -9.75218892e-01 -6.47172928e-01 3.34770530e-01 7.18470812e-01 -4.89893168e-01 8.30040693e-01 -9.24132392e-02 -4.39525425e-01 8.54961157e-01 -1.39484897e-01 -1.29251644e-01 1.12548280e+00 -1.65059853e+00 5.02724290e-01 7.41642058e-01 -8.03464875e-02 7.37220585e-01 9.27099764e-01 -5.69847643e-01 -7.09288359e-01 -9.25838411e-01 1.26899743e+00 -7.86657631e-01 4.92204905e-01 2.12267116e-02 -5.51380873e-01 1.50522009e-01 3.74208868e-01 -1.02756524e+00 6.46533132e-01 8.67993012e-02 -1.09315664e-01 1.85085684e-01 -1.12478852e+00 6.01738930e-01 9.14007068e-01 -2.57792324e-01 -8.08312714e-01 6.57768548e-01 8.62082422e-01 -1.79614067e-01 -9.38148499e-01 4.09695745e-01 4.37278718e-01 -1.02879739e+00 4.12667841e-01 -6.13085628e-01 5.96663594e-01 -1.15078323e-01 4.28475253e-02 -1.60820448e+00 -4.24951434e-01 -1.04922509e+00 1.12356596e-01 1.49858022e+00 9.02815402e-01 -7.55997062e-01 1.67183846e-01 6.68139756e-01 -3.57998699e-01 -5.74667811e-01 -5.12563348e-01 -1.06156242e+00 2.26921976e-01 9.50184241e-02 3.88311148e-01 1.00396597e+00 3.87029439e-01 8.52707505e-01 -3.58876109e-01 -3.86555761e-01 1.35232851e-01 -2.37936527e-01 1.07849789e+00 -1.23654878e+00 -7.02323392e-02 -8.46618891e-01 -3.80466640e-01 -6.95760787e-01 -2.17506513e-01 -6.76804483e-01 1.38395980e-01 -1.95451784e+00 2.16831073e-01 4.35645506e-02 3.21995430e-02 2.54134327e-01 -4.89840776e-01 6.64585410e-03 2.84781456e-01 2.87677586e-01 -6.74117446e-01 4.81050789e-01 1.29715037e+00 -1.94510773e-01 -3.01025987e-01 -1.24175707e-02 -9.69915688e-01 6.10442996e-01 8.93667400e-01 -2.06153005e-01 -4.15403813e-01 -3.91374290e-01 2.23660693e-01 8.53687897e-02 4.98897322e-02 -1.12722909e+00 3.19508165e-01 -2.33105630e-01 8.22165385e-02 -1.86666876e-01 -1.83446750e-01 5.00888005e-02 -2.25816444e-01 4.05077189e-01 -5.04796624e-01 4.47316915e-01 -1.85515523e-01 2.72158027e-01 -4.14868206e-01 -6.17577136e-01 4.87709731e-01 -2.49531284e-01 -3.01931322e-01 -8.79611522e-02 -3.93621117e-01 5.54207444e-01 5.30072987e-01 -5.73814929e-01 -7.20175564e-01 -8.58868897e-01 -3.32580507e-01 4.56634730e-01 2.09567279e-01 5.55355191e-01 3.63588303e-01 -1.09226656e+00 -1.12008286e+00 -4.40866381e-01 4.08761889e-01 -4.82755095e-01 2.01711282e-02 8.41269672e-01 -7.29240716e-01 8.01980495e-01 -1.92799345e-01 -4.96458292e-01 -1.14336479e+00 3.53811421e-02 1.01725630e-01 -7.87098169e-01 1.69918537e-01 5.63908041e-01 2.71934450e-01 -5.47122359e-01 1.70969233e-01 -3.08478594e-01 -2.26152986e-01 3.40591431e-01 3.55047077e-01 8.51945519e-01 2.66576529e-01 -5.14663875e-01 -1.76177308e-01 3.72261792e-01 -1.02336653e-01 -4.33669060e-01 6.30364299e-01 -7.74903074e-02 -3.31664807e-03 5.08706570e-01 8.08566928e-01 -9.61837098e-02 -4.69793409e-01 2.23712862e-01 2.30068728e-01 -9.06957313e-02 -3.47501993e-01 -1.35424066e+00 -2.56461680e-01 7.60567307e-01 2.32555807e-01 6.39476240e-01 8.16763520e-01 -3.61645967e-01 5.28469801e-01 5.37688911e-01 4.27249432e-01 -1.47606969e+00 4.17808443e-01 7.66013563e-01 1.40385640e+00 -1.30398595e+00 -7.92713929e-03 -1.70175314e-01 -1.19604576e+00 1.14912868e+00 9.16819453e-01 3.76076013e-01 -6.42525330e-02 -4.20788497e-01 7.52161667e-02 -1.29386544e-01 -9.08746302e-01 -1.51908681e-01 7.84150898e-01 6.93617880e-01 1.08153880e+00 2.20127329e-01 -1.21407950e+00 6.90130651e-01 -8.78767192e-01 -4.57178026e-01 6.27086639e-01 5.85842669e-01 -6.58226073e-01 -8.74670267e-01 -1.14216097e-01 5.45962155e-01 -1.81922749e-01 -7.23692030e-02 -1.14879251e+00 8.81611764e-01 -3.35928649e-01 1.52879512e+00 -3.53830189e-01 -4.40556914e-01 6.30539298e-01 7.40437135e-02 3.75401050e-01 -7.50439942e-01 -1.15860379e+00 -2.88999557e-01 7.59231806e-01 -2.55902231e-01 -2.00870693e-01 -2.58757800e-01 -7.37677455e-01 -2.87653297e-01 -5.53927481e-01 6.53681874e-01 4.41094875e-01 8.11248481e-01 2.16441020e-01 2.88937449e-01 7.14194775e-01 -7.23534167e-01 -9.75630820e-01 -1.74495542e+00 -2.82957107e-01 4.63572353e-01 -7.58516742e-03 -2.07294211e-01 -6.98783159e-01 -1.64119825e-01]

[11.84183406829834, 9.033109664916992]

edc6a116-ff78-4d00-964d-0ce5fbd85312

from-wide-to-deep-dimension-lifting-network

2303.12816

https://arxiv.org/abs/2303.12816v2

https://arxiv.org/pdf/2303.12816v2.pdf

From Wide to Deep: Dimension Lifting Network for Parameter-efficient Knowledge Graph Embedding

Knowledge graph embedding (KGE) that maps entities and relations into vector representations is essential for downstream tasks. Conventional KGE methods require relatively high-dimensional entity representations to preserve the structural information of knowledge graph, but lead to oversized model parameters. Recent methods reduce model parameters by adopting low-dimensional entity representations, while developing techniques (e.g., knowledge distillation) to compensate for the reduced dimension. However, such operations produce degraded model accuracy and limited reduction of model parameters. Specifically, we view the concatenation of all entity representations as an embedding layer, and then conventional KGE methods that adopt high-dimensional entity representations equal to enlarging the width of the embedding layer to gain expressiveness. To achieve parameter efficiency without sacrificing accuracy, we instead increase the depth and propose a deeper embedding network for entity representations, i.e., a narrow embedding layer and a multi-layer dimension lifting network (LiftNet). Experiments on three public datasets show that the proposed method (implemented based on TransE and DistMult) with 4-dimensional entity representations achieves more accurate link prediction results than counterpart parameter-efficient KGE methods and strong KGE baselines, including TransE and DistMult with 512-dimensional entity representations.

['Tom Luan', 'Jiong Jin', 'He Zhang', 'Di wu', 'Longxiang Gao', 'Yong Xiang', 'Borui Cai']

2023-03-22

['knowledge-graph-embedding']

['graphs']

[-2.83304900e-01 4.20184553e-01 -4.65870649e-01 -1.00647867e-01 -1.61713079e-01 -4.40213501e-01 3.82032692e-01 2.33463660e-01 -4.46039706e-01 5.01493752e-01 3.98670137e-01 -4.22423780e-01 -3.25712293e-01 -1.29883838e+00 -7.09276438e-01 -3.15790445e-01 -2.27989897e-01 2.65641898e-01 2.39465848e-01 -1.88892528e-01 -2.02486962e-01 2.83303797e-01 -1.02217519e+00 -1.68699443e-01 9.09470141e-01 8.70883465e-01 -3.36028993e-01 3.14395577e-01 -3.73544723e-01 5.77800274e-01 -2.65289009e-01 -8.72934222e-01 2.14163274e-01 9.14075002e-02 -8.93309116e-01 -5.46301186e-01 3.88903469e-01 -2.92066813e-01 -9.94422972e-01 7.82328725e-01 4.97392356e-01 2.19856605e-01 6.04553521e-01 -1.44494903e+00 -1.26943040e+00 8.68322134e-01 -4.70728606e-01 6.03379011e-02 4.30853711e-03 -1.52526423e-01 1.39440310e+00 -1.11549175e+00 5.30415595e-01 1.21262026e+00 1.01990092e+00 3.64137441e-01 -1.54844534e+00 -6.74434662e-01 3.68873715e-01 3.72141838e-01 -1.81958687e+00 -2.79682517e-01 7.13511050e-01 -3.87961239e-01 1.32997453e+00 2.14985739e-02 4.91149276e-01 9.23426747e-01 -2.16272324e-01 4.81546342e-01 3.07627469e-01 -9.02131423e-02 -9.06522050e-02 1.46724895e-01 4.60281283e-01 1.02461028e+00 8.99011731e-01 -2.29330614e-01 -3.64733011e-01 -3.35209876e-01 7.51328528e-01 1.75885648e-01 -4.14865255e-01 -6.74036086e-01 -1.22414196e+00 9.33451116e-01 9.00013983e-01 1.33687168e-01 -5.51086366e-01 5.07186830e-01 6.90591037e-01 3.39269698e-01 4.04941916e-01 5.48725724e-01 -5.47656775e-01 1.73824176e-01 -4.29778427e-01 6.46913871e-02 7.92813003e-01 1.06144619e+00 1.01764095e+00 1.03141904e-01 -2.46782824e-01 8.83079469e-01 1.91737115e-01 1.01519212e-01 4.45738524e-01 -4.92971420e-01 8.24924886e-01 1.16720200e+00 -1.28219396e-01 -1.55846286e+00 -3.85838896e-01 -4.62524146e-01 -1.00975835e+00 -4.79238600e-01 -1.82580158e-01 -1.75485134e-01 -7.63314009e-01 1.89684045e+00 4.55652297e-01 4.54712212e-01 3.46528769e-01 7.20924497e-01 9.44895983e-01 7.21600771e-01 3.34697574e-01 2.74900973e-01 1.35358453e+00 -1.26515961e+00 -6.11367464e-01 -3.81693728e-02 1.28284419e+00 -1.57285795e-01 9.38468397e-01 -3.27235252e-01 -7.80866444e-01 -3.32569003e-01 -1.02133036e+00 -4.65199888e-01 -9.77875352e-01 1.32009581e-01 1.16221321e+00 6.08711541e-01 -9.18362916e-01 6.11165702e-01 -6.90639555e-01 -2.24990666e-01 4.01868522e-01 5.55299938e-01 -7.67076254e-01 -1.10535719e-01 -1.68336415e+00 7.99024582e-01 9.78707194e-01 3.31226289e-01 -1.33340478e-01 -1.03176355e+00 -1.35573232e+00 5.29172480e-01 5.10984004e-01 -1.00201190e+00 3.90940338e-01 -7.83764496e-02 -1.21378446e+00 3.60189945e-01 1.52745500e-01 -5.16436577e-01 -3.71372774e-02 -3.27759653e-01 -5.42606890e-01 -1.27948793e-02 -3.12431425e-01 5.61892927e-01 4.85970557e-01 -9.50509846e-01 -2.82165527e-01 -1.50152326e-01 4.12659854e-01 5.92676103e-02 -1.07280326e+00 -5.15347004e-01 -7.43348777e-01 -6.64710462e-01 8.07491764e-02 -9.45112884e-01 -2.34136716e-01 -7.03904033e-02 -5.36873043e-01 -3.66701216e-01 8.11181307e-01 -7.23109126e-01 1.79866481e+00 -2.04503226e+00 3.31240714e-01 4.03439760e-01 7.94636786e-01 5.45205772e-01 -2.93248206e-01 4.92631704e-01 -5.38098477e-02 2.91074097e-01 -6.56732172e-02 -2.64528930e-01 2.90603161e-01 4.45310116e-01 -1.39480695e-01 2.41179749e-01 2.59039998e-01 1.40833521e+00 -9.32026982e-01 -5.65390944e-01 1.10691831e-01 7.46881425e-01 -9.70132411e-01 -1.00633405e-01 3.56571116e-02 -4.12869275e-01 -5.95529914e-01 4.76685882e-01 6.52613282e-01 -6.14583194e-01 4.89311963e-01 -7.75934458e-01 3.52984250e-01 4.25878495e-01 -1.28195488e+00 1.50274241e+00 -7.27534592e-01 2.13075101e-01 -2.42987156e-01 -9.85728383e-01 9.07265425e-01 1.89082086e-01 4.02140826e-01 -2.95467615e-01 -1.93037521e-02 2.15766311e-01 -1.31019533e-01 -1.03748173e-01 8.17493379e-01 2.54993021e-01 -2.38768250e-01 1.49273351e-01 1.69364899e-01 4.95042562e-01 2.20262446e-02 6.70611084e-01 1.06638026e+00 -1.26655608e-01 1.89740494e-01 2.70343907e-02 4.40605849e-01 -4.08905923e-01 7.17470229e-01 3.42102289e-01 9.01795030e-02 -1.75057620e-01 7.25590408e-01 -5.32329500e-01 -8.72366607e-01 -8.28752756e-01 7.69034475e-02 9.64127004e-01 2.77525276e-01 -1.17193615e+00 -3.42356950e-01 -9.28590775e-01 5.96997738e-01 5.86095214e-01 -7.82639802e-01 -7.66324520e-01 -6.28783405e-01 -8.46416235e-01 7.54543722e-01 8.72006178e-01 4.47073400e-01 -5.73445499e-01 8.63944069e-02 3.10736775e-01 -6.25983637e-04 -1.17990661e+00 -5.24818897e-01 3.93365361e-02 -7.35139787e-01 -9.35758710e-01 -4.04406846e-01 -8.09026480e-01 6.32061541e-01 1.72870710e-01 9.22939956e-01 1.31840289e-01 -9.22602192e-02 2.30291754e-01 -4.10722941e-01 1.89551055e-01 5.45741320e-02 5.04782140e-01 1.88633054e-01 -1.17229983e-01 4.66282398e-01 -8.00347567e-01 -7.29088545e-01 1.44843146e-01 -8.02624106e-01 4.15034965e-02 7.01570332e-01 1.07082319e+00 5.59518397e-01 -7.91134611e-02 8.17999065e-01 -1.05393028e+00 6.24024928e-01 -6.48573756e-01 -3.95647347e-01 5.47057867e-01 -1.14123607e+00 3.93382549e-01 7.31704772e-01 -4.88000810e-01 -6.46441579e-01 -3.38473558e-01 7.63007253e-02 -8.29373538e-01 6.26312017e-01 8.12327981e-01 -4.78995256e-02 -2.22793981e-01 4.55000281e-01 1.34475648e-01 -3.09655547e-01 -5.81383944e-01 9.70866084e-01 4.00390893e-01 3.65049183e-01 -5.38132668e-01 9.96840119e-01 3.48960253e-04 5.85723035e-02 -2.86466867e-01 -6.54253542e-01 -3.17765385e-01 -5.34668326e-01 5.98711252e-01 6.06564403e-01 -1.03417122e+00 -7.59165823e-01 -9.45073813e-02 -9.77331638e-01 5.57183772e-02 -3.39844286e-01 6.52374983e-01 -5.43828979e-02 2.97163934e-01 -7.72491217e-01 -2.95138657e-01 -6.44849479e-01 -6.78454280e-01 8.24572146e-01 -6.18660003e-02 2.13352740e-02 -1.13324666e+00 3.77210304e-02 2.08182126e-01 6.03909731e-01 2.18906254e-01 1.41483688e+00 -7.03408360e-01 -4.79520649e-01 -2.87175447e-01 -6.07065499e-01 2.96671540e-01 1.48063958e-01 -3.86776984e-01 -6.90003157e-01 -3.40027899e-01 -9.32457983e-01 -2.64146179e-01 1.22924757e+00 -1.29970357e-01 1.18769991e+00 -5.99659562e-01 -7.24605620e-01 1.01988769e+00 1.56808996e+00 -3.63035947e-01 5.30301094e-01 2.58754462e-01 1.29550457e+00 3.03452462e-01 6.29735589e-02 2.90835619e-01 8.47860396e-01 7.33306706e-01 1.62794963e-01 -1.11862421e-01 -3.37136388e-01 -7.59830773e-01 1.19080625e-01 1.23259318e+00 -1.34601057e-01 -1.78418159e-01 -8.10686171e-01 7.82478094e-01 -1.90931952e+00 -6.34109616e-01 1.98840126e-01 1.95923185e+00 8.40247214e-01 6.27005324e-02 -3.80851924e-02 4.91797924e-02 6.27089977e-01 3.46717089e-01 -6.55774355e-01 -3.39455932e-01 -5.60901947e-02 6.12108000e-02 7.44436026e-01 3.55580658e-01 -9.83791411e-01 1.33726025e+00 5.37369537e+00 8.15049708e-01 -8.74996960e-01 1.39226779e-01 -6.27987459e-02 6.92700818e-02 -4.93857324e-01 7.48858321e-03 -8.42030823e-01 3.42338264e-01 1.05891144e+00 -5.54546535e-01 3.28485429e-01 9.76669073e-01 -4.60725963e-01 6.77680194e-01 -1.12527466e+00 9.95514929e-01 -2.46041223e-01 -1.53593099e+00 3.39075148e-01 6.27108067e-02 4.70472664e-01 -1.12283640e-01 -2.14855999e-01 1.01973641e+00 4.01017815e-01 -9.24275994e-01 9.38352644e-02 3.25321347e-01 9.21491623e-01 -8.06789458e-01 7.65437067e-01 -1.92771092e-01 -1.66324675e+00 -9.45172012e-02 -6.36661470e-01 2.13653937e-01 1.30502895e-01 4.75409895e-01 -8.33681643e-01 9.52439070e-01 4.53255743e-01 8.63008499e-01 -6.91734433e-01 6.41980648e-01 -1.21034749e-01 4.93652523e-01 -3.54381949e-01 1.93718001e-01 2.11386770e-01 -1.76541477e-01 2.37543419e-01 1.31888556e+00 2.11210176e-01 1.17682464e-01 1.11853808e-01 7.39368618e-01 -7.20927179e-01 1.35065153e-01 -6.64701521e-01 -4.28862303e-01 9.11921144e-01 1.26972342e+00 -1.49873987e-01 -3.93171966e-01 -7.45383024e-01 1.10229671e+00 9.01384354e-01 3.92102689e-01 -8.56709898e-01 -8.77642989e-01 1.03892756e+00 -8.03711489e-02 6.06559873e-01 -2.81696647e-01 2.09473178e-01 -1.38016963e+00 8.25595856e-02 -3.99568409e-01 5.48989713e-01 -2.32369646e-01 -1.30485272e+00 4.99065042e-01 -1.57761231e-01 -8.39408696e-01 -8.52601305e-02 -5.03659844e-01 -3.01765949e-01 9.40822899e-01 -1.84470212e+00 -1.37603664e+00 -1.51275724e-01 6.75159097e-01 -1.58171520e-01 -1.39764538e-02 1.09879994e+00 6.56979442e-01 -1.02119255e+00 1.21446586e+00 1.15701057e-01 4.49929327e-01 5.14644563e-01 -1.18883789e+00 5.61254382e-01 3.90655279e-01 7.62847960e-02 1.18397510e+00 5.95610514e-02 -5.62197387e-01 -1.78238368e+00 -1.49130404e+00 1.01477325e+00 -2.95725971e-01 8.83701861e-01 -5.67629933e-01 -1.22901893e+00 1.06979120e+00 -3.05704296e-01 6.54883265e-01 7.92886198e-01 6.09031379e-01 -8.76568675e-01 -1.19346857e-01 -9.76472378e-01 6.49342597e-01 1.37319326e+00 -7.95653760e-01 -6.77322447e-01 1.51894301e-01 1.39647162e+00 -2.82098576e-02 -1.62834692e+00 4.83707756e-01 6.96395218e-01 -2.96103984e-01 1.30403864e+00 -9.85734820e-01 1.34612203e-01 -3.06678802e-01 -2.23403782e-01 -1.33790267e+00 -6.98655009e-01 -4.84968781e-01 -8.94422054e-01 1.26176810e+00 6.89009428e-01 -9.33696032e-01 8.06749701e-01 5.99290788e-01 7.74230622e-03 -1.20934129e+00 -6.63326085e-01 -1.09256530e+00 7.29018003e-02 6.10260814e-02 1.01689005e+00 1.33939755e+00 1.49866238e-01 7.14151323e-01 -4.32445943e-01 4.61484969e-01 4.44988698e-01 2.06065834e-01 7.56928921e-01 -1.31904209e+00 -6.78163096e-02 -3.19263905e-01 -7.79975414e-01 -1.01990879e+00 4.26475942e-01 -1.20986581e+00 -6.92107320e-01 -1.80707717e+00 7.65974522e-02 -8.44382107e-01 -7.61205435e-01 9.11393464e-01 -3.87864918e-01 -1.93223916e-02 1.17415972e-01 2.15964735e-01 -5.71878850e-01 1.05502284e+00 9.47886288e-01 -1.03539377e-01 -2.65433878e-01 -6.10131502e-01 -8.39254797e-01 5.06196558e-01 5.17691731e-01 -3.52988303e-01 -6.94253981e-01 -6.68080091e-01 4.88741577e-01 -3.04277360e-01 3.14110279e-01 -5.85033536e-01 4.27646160e-01 1.47672772e-01 2.92987585e-01 -2.51253724e-01 4.23768222e-01 -8.27973843e-01 1.49410978e-01 2.90835112e-01 -2.26645246e-01 9.20496695e-03 1.68736994e-01 9.46100950e-01 -3.34247351e-01 1.15329958e-01 2.85352051e-01 3.13307703e-01 -9.95244503e-01 6.98424697e-01 4.80106860e-01 -1.79032654e-01 9.22079146e-01 -1.39472052e-01 -5.77708721e-01 7.95727298e-02 -6.84157789e-01 4.97790277e-01 2.97718614e-01 4.80542421e-01 6.71010196e-01 -1.79547799e+00 -5.09579360e-01 1.19535111e-01 4.25918728e-01 -1.52723091e-02 2.86972046e-01 8.75461638e-01 -2.27271512e-01 5.53773403e-01 8.04372057e-02 8.54080543e-02 -9.57502663e-01 7.84441352e-01 2.20343292e-01 -6.93237066e-01 -9.41453636e-01 8.81254494e-01 2.22847521e-01 -7.57387459e-01 1.04341365e-01 -3.08191150e-01 -2.19790369e-01 2.10144028e-01 2.77263850e-01 5.69502413e-01 1.03762887e-01 -3.74340713e-01 -5.34089804e-01 5.90497971e-01 -3.17411989e-01 6.04640245e-01 1.42376816e+00 -6.64003715e-02 -6.40159920e-02 -1.40093520e-01 1.51851952e+00 3.88131812e-02 -8.46277654e-01 -6.81757689e-01 3.55455689e-02 -3.70932549e-01 4.74266380e-01 -4.81199831e-01 -1.27270651e+00 6.77780449e-01 1.89659059e-01 4.60917130e-02 9.21631455e-01 4.16700132e-02 1.19933176e+00 8.36073995e-01 4.29213226e-01 -7.95133889e-01 -2.79025793e-01 4.40915555e-01 6.62833869e-01 -1.04381764e+00 1.41352311e-01 -6.24381602e-01 -4.55822885e-01 8.64690900e-01 7.76140332e-01 8.53362214e-03 7.91879416e-01 -6.86913952e-02 -6.05780602e-01 -4.35616493e-01 -8.63000393e-01 -2.37207919e-01 4.76226419e-01 4.37345862e-01 2.88074523e-01 2.16153175e-01 -2.53098935e-01 9.87639904e-01 -9.25972834e-02 -2.19713766e-02 1.68530867e-01 7.17827678e-01 -1.66227371e-01 -1.10151625e+00 2.94931531e-01 6.22794688e-01 -1.08851306e-01 -4.40560251e-01 -1.68272838e-01 9.85413015e-01 1.10959105e-01 4.16877896e-01 -5.64170852e-02 -8.72185707e-01 6.97645903e-01 2.53929764e-01 2.43205160e-01 -6.11019433e-01 -4.24754202e-01 -5.89405775e-01 2.98647761e-01 -6.10070705e-01 4.11099009e-02 -3.69268768e-02 -1.21613348e+00 -7.02948332e-01 -6.03585541e-01 1.93468034e-01 3.41624111e-01 3.49308938e-01 9.88726616e-01 6.41937494e-01 4.39136684e-01 -4.11503255e-01 -5.60141504e-01 -8.45667303e-01 -7.73736715e-01 3.59367847e-01 6.28429353e-02 -9.50252831e-01 -1.94498137e-01 -2.24261746e-01]

[8.734603881835938, 7.870014190673828]

c012cf84-617d-4eee-a475-4da92fc14e07

text2struct-a-machine-learning-pipeline-for

2212.09044

https://arxiv.org/abs/2212.09044v3

https://arxiv.org/pdf/2212.09044v3.pdf

Text2Struct: A Machine Learning Pipeline for Mining Structured Data from Text

Many analysis and prediction tasks require the extraction of structured data from unstructured texts. However, an annotation scheme and a training dataset have not been available for training machine learning models to mine structured data from text without special templates and patterns. To solve it, this paper presents an end-to-end machine learning pipeline, Text2Struct, including a text annotation scheme, training data processing, and machine learning implementation. We formulated the mining problem as the extraction of metrics and units associated with numerals in the text. Text2Struct was trained and evaluated using an annotated text dataset collected from abstracts of medical publications regarding thrombectomy. In terms of prediction performance, a dice coefficient of 0.82 was achieved on the test dataset. By random sampling, most predicted relations between numerals and entities were well matched to the ground-truth annotations. These results show that Text2Struct is viable for the mining of structured data from text without special templates or patterns. It is anticipated to further improve the pipeline by expanding the dataset and investigating other machine learning models. A code demonstration can be found at: https://github.com/zcc861007/CourseProject

['Bo Yang', 'Chaochao Zhou']

2022-12-18

['text-annotation']

['natural-language-processing']

[ 2.74203598e-01 7.23828197e-01 -1.28260955e-01 -6.09015405e-01 -8.67027402e-01 -5.22688210e-01 3.75678599e-01 8.27110052e-01 -4.86309528e-01 8.14772904e-01 1.68986440e-01 -4.42551911e-01 -1.85357794e-01 -6.99783146e-01 -3.89736772e-01 -2.27934912e-01 8.41910392e-02 9.31327403e-01 5.66692725e-02 2.06196606e-01 3.53701442e-01 1.45496637e-01 -1.35242808e+00 8.92379940e-01 9.96528864e-01 8.13708365e-01 3.35232019e-01 5.41863739e-01 -4.15460736e-01 9.17325377e-01 -6.47284508e-01 -5.01961946e-01 1.14869006e-01 -3.21873218e-01 -1.12949669e+00 -1.36863902e-01 2.08711829e-02 8.69728327e-02 1.54260844e-01 6.34562731e-01 4.74611610e-01 -3.29404354e-01 7.81090438e-01 -9.34340656e-01 6.00909553e-02 8.18917930e-01 -1.84914321e-01 -1.24499924e-01 4.43385333e-01 -1.39184207e-01 1.15147233e+00 -1.11421061e+00 1.00962055e+00 6.46298647e-01 7.19008863e-01 2.79323041e-01 -8.97071958e-01 -5.34893990e-01 -5.03539026e-01 2.62388941e-02 -1.36021447e+00 -3.29103619e-01 2.56685525e-01 -5.73218644e-01 1.27621984e+00 3.64079028e-01 5.52931726e-01 8.15431237e-01 1.67082161e-01 7.45339513e-01 8.65880728e-01 -8.15358162e-01 1.55028775e-01 5.65790713e-01 1.47315815e-01 9.33620512e-01 5.08136988e-01 -2.23115861e-01 -5.00382125e-01 -2.28267834e-01 1.25669315e-01 -2.84644663e-01 -2.75039244e-02 -9.71093401e-02 -1.17272377e+00 7.41018355e-01 -7.15871528e-02 3.30918998e-01 -3.17742944e-01 -6.90784216e-01 7.01960742e-01 2.42072865e-01 5.29968739e-01 8.62174332e-01 -1.00475347e+00 -3.41880083e-01 -1.08005726e+00 -7.07787797e-02 1.28134108e+00 1.07255161e+00 6.13164723e-01 -5.24703205e-01 1.26785189e-01 6.73351228e-01 2.71079019e-02 6.71027899e-02 6.04418993e-01 -6.40842617e-01 7.17275977e-01 1.25199008e+00 7.24839373e-03 -8.33380461e-01 -8.51081073e-01 -3.09719086e-01 -6.05657995e-01 -1.26066700e-01 6.78327203e-01 -4.84251767e-01 -5.45861065e-01 1.00377595e+00 4.00870383e-01 -3.86708558e-01 1.79868057e-01 5.91039121e-01 1.40036595e+00 2.69975007e-01 1.48755506e-01 -3.22551847e-01 1.43102324e+00 -5.09469032e-01 -7.54551768e-01 5.92021421e-02 1.61127293e+00 -9.88786459e-01 9.63806689e-01 5.51770627e-01 -9.47566092e-01 -2.58245707e-01 -8.42842758e-01 -3.42204452e-01 -6.00518286e-01 5.03245294e-01 4.79881793e-01 5.39541543e-01 -6.09973490e-01 7.66679227e-01 -9.48458791e-01 -4.67767000e-01 5.83677828e-01 4.19219971e-01 -6.10228300e-01 7.50607252e-02 -1.00442016e+00 1.04389012e+00 7.18955219e-01 -1.33605242e-01 -1.41658382e-02 -8.51367414e-01 -9.00280058e-01 2.83471085e-02 3.53976816e-01 -6.39923215e-01 1.08289576e+00 -7.12601721e-01 -9.68325317e-01 1.21985567e+00 -9.51936468e-02 -3.93996686e-01 4.20099020e-01 -9.62444097e-02 -1.47610813e-01 1.16664715e-01 2.46161222e-01 3.47278744e-01 2.82204896e-01 -7.39850044e-01 -8.08386445e-01 -4.17354226e-01 -5.27528226e-01 3.54456790e-02 -5.44112384e-01 1.30524486e-01 -3.26988459e-01 -5.82963288e-01 -1.46881655e-01 -7.75435865e-01 -2.04294950e-01 -2.11809605e-01 -4.42245007e-01 -4.24321115e-01 5.29734731e-01 -9.08824623e-01 1.43291140e+00 -1.85938251e+00 -2.10429773e-01 3.29290479e-01 4.24112350e-01 2.07734644e-01 2.00607762e-01 6.41557336e-01 -1.01491168e-01 2.74370313e-01 -1.67963400e-01 -1.21885642e-01 -1.78243175e-01 9.31842178e-02 -5.97600304e-02 2.83585280e-01 2.47977719e-01 8.14228475e-01 -8.87647986e-01 -9.33409154e-01 3.91159244e-02 2.50014961e-01 -5.02142429e-01 2.52519220e-01 -3.66074204e-01 2.13081419e-01 -5.32367408e-01 4.58899349e-01 2.63559401e-01 -4.62524295e-01 4.79972482e-01 -1.23455018e-01 -9.53171477e-02 6.87387109e-01 -1.02312231e+00 1.60595906e+00 -2.65002519e-01 6.57322884e-01 -1.09729797e-01 -9.57784355e-01 1.13253343e+00 3.77587408e-01 9.34638023e-01 -3.95564556e-01 1.95885524e-01 3.91872346e-01 2.17189398e-02 -9.15316105e-01 4.17123318e-01 8.37751105e-02 -9.30980891e-02 5.09101987e-01 6.29285425e-02 -2.40176633e-01 6.00860059e-01 2.41454527e-01 1.49450898e+00 2.04186097e-01 5.44315815e-01 -2.66131371e-01 3.41720909e-01 6.63720489e-01 5.96579731e-01 5.11814654e-01 3.24269623e-01 5.11292517e-01 6.01107180e-01 -6.05255008e-01 -1.15131629e+00 -3.51283908e-01 -3.33083987e-01 7.91643739e-01 -5.78732550e-01 -1.29149723e+00 -6.48360312e-01 -9.32153225e-01 -1.44547135e-01 7.28159130e-01 -6.38168454e-01 2.05558017e-01 -3.40482563e-01 -6.70675695e-01 6.05655372e-01 3.08922976e-01 1.44099668e-01 -1.08252823e+00 -8.43063176e-01 2.50765622e-01 -3.74148667e-01 -1.22035348e+00 -1.07510246e-01 4.66236532e-01 -9.34277534e-01 -1.44818735e+00 -9.63761881e-02 -7.93463230e-01 7.54321337e-01 -3.67702782e-01 1.17438650e+00 2.51195788e-01 -6.30101800e-01 -1.10745482e-01 -5.44299364e-01 -9.24224973e-01 -4.63539213e-01 5.78471959e-01 -3.64820600e-01 -5.11300206e-01 7.69553244e-01 -1.39517725e-01 -4.96091485e-01 3.46652359e-01 -9.44964051e-01 4.25834894e-01 6.16113961e-01 8.03742111e-01 6.61963224e-01 1.28565803e-01 4.02416259e-01 -1.17978191e+00 4.80774462e-01 -4.94425893e-01 -5.12441814e-01 1.15086779e-01 -9.41748798e-01 8.29229206e-02 7.13104486e-01 -7.18049612e-03 -8.84698153e-01 3.97851974e-01 -2.09493637e-01 1.53598770e-01 -2.64797330e-01 1.03246105e+00 -1.27048254e-01 6.39266133e-01 8.51139724e-01 -1.40253901e-01 1.46967366e-01 -5.55930853e-01 1.84639115e-02 1.15078866e+00 3.33232954e-02 -4.14901137e-01 4.49213892e-01 1.94994286e-01 2.97015179e-02 -8.95557940e-01 -9.96156216e-01 -6.24844849e-01 -9.28303957e-01 1.02375709e-01 7.00831711e-01 -7.34815538e-01 -3.33191901e-01 -1.31676421e-01 -8.64342153e-01 -4.23508078e-01 -4.22882020e-01 7.36247182e-01 -3.84419620e-01 2.83727020e-01 -3.85803223e-01 -2.92696923e-01 -7.99295187e-01 -7.84131825e-01 9.79416013e-01 -1.03818946e-01 -9.59964156e-01 -9.85408664e-01 5.75742796e-02 6.87103212e-01 -6.01908332e-03 2.79215068e-01 9.61970627e-01 -1.33884871e+00 4.33396846e-02 -6.34281635e-01 -1.89807899e-02 -3.09016258e-02 2.73072034e-01 2.20350266e-01 -6.72450423e-01 -5.10867424e-02 -2.23523095e-01 -3.54024589e-01 5.17294228e-01 5.72564453e-02 1.23373985e+00 -4.01073009e-01 -5.60360670e-01 4.46176052e-01 9.29835498e-01 9.31732059e-02 4.53656077e-01 5.95587611e-01 5.46270192e-01 1.01667547e+00 8.05329323e-01 6.65497422e-01 3.61457765e-01 4.26935613e-01 2.60577966e-02 -1.32164126e-02 1.96695119e-01 -1.91896081e-01 3.70997116e-02 5.80953538e-01 1.00542292e-01 -3.92933302e-02 -1.42504919e+00 4.37577963e-01 -1.63581538e+00 -7.56768584e-01 -5.63586771e-01 2.00026464e+00 1.08987296e+00 4.02729928e-01 7.05090314e-02 3.46661240e-01 3.92891794e-01 -6.81031644e-01 -3.95343661e-01 -3.04508537e-01 4.48920578e-02 3.81550431e-01 2.16272309e-01 2.55425543e-01 -1.01979184e+00 9.52156484e-01 5.64148474e+00 6.91533208e-01 -8.31783295e-01 -1.11696184e-01 5.15068293e-01 -2.17472628e-01 4.57493588e-03 1.04068771e-01 -1.11751235e+00 3.19472164e-01 1.27857244e+00 -2.43922740e-01 -2.76456892e-01 7.92172253e-01 5.85153461e-01 -3.39490026e-02 -9.97722030e-01 6.14948034e-01 -2.35442176e-01 -1.52091742e+00 -1.56281456e-01 1.82469442e-01 3.79568636e-01 8.19210410e-02 -5.50439596e-01 2.80986190e-01 3.02336067e-01 -1.06703091e+00 3.00857574e-01 2.74796337e-01 8.54596913e-01 -4.12993878e-01 1.07002115e+00 4.40863788e-01 -8.92291427e-01 1.09080911e-01 -2.74783313e-01 -9.17998478e-02 -2.66314149e-01 9.29271638e-01 -1.64267671e+00 7.56420910e-01 7.21614718e-01 1.00563562e+00 -8.22700560e-01 9.99634266e-01 -2.22074062e-01 7.46192396e-01 -3.51942480e-01 -1.83749467e-01 -8.42648223e-02 -2.58721057e-02 2.19077364e-01 1.56068623e+00 2.09159300e-01 1.84245750e-01 2.20027506e-01 4.80954468e-01 -8.15905035e-02 7.29031682e-01 -4.18785006e-01 -2.17165589e-01 3.38765234e-01 1.56601310e+00 -8.79487813e-01 -4.03860390e-01 -5.37141979e-01 4.22322303e-01 3.10574472e-01 -8.06145146e-02 -4.59040314e-01 -7.30577886e-01 6.01879135e-02 4.72098321e-01 1.80222780e-01 5.47910631e-02 -7.98010468e-01 -1.14628136e+00 1.60756305e-01 -9.54893231e-01 9.10675526e-01 -7.09763229e-01 -1.15286911e+00 6.98655605e-01 7.59056509e-02 -1.29174221e+00 -3.41521889e-01 -7.74726808e-01 -2.29489252e-01 7.60931671e-01 -9.33565438e-01 -1.02489591e+00 -3.60906988e-01 3.86566281e-01 4.74379510e-01 -3.95112187e-01 1.14415371e+00 -1.24865184e-02 -5.39904833e-01 4.56389457e-01 7.29536340e-02 5.25969744e-01 9.44016695e-01 -1.33625352e+00 2.01703254e-02 4.05236661e-01 1.36169801e-02 6.09851718e-01 6.14552379e-01 -9.59340870e-01 -1.25619769e+00 -1.17669439e+00 1.25809395e+00 -7.06479549e-01 7.02439249e-01 -2.38894612e-01 -9.46940303e-01 5.70642650e-01 2.43703976e-01 -3.50677937e-01 1.41916299e+00 1.87971950e-01 -4.78551537e-02 2.28670478e-01 -1.08754408e+00 2.21035019e-01 7.16530681e-01 -1.41941920e-01 -7.97631323e-01 5.32409191e-01 2.56574959e-01 -4.77719784e-01 -1.48059654e+00 3.92086446e-01 5.22784770e-01 -3.58183295e-01 3.90752941e-01 -7.90863693e-01 8.30490530e-01 -9.82630476e-02 1.10046692e-01 -9.97664213e-01 6.65546507e-02 -5.98409772e-01 -1.12189971e-01 1.11734366e+00 1.08948314e+00 -3.40264022e-01 1.04095685e+00 9.44766521e-01 -2.55248189e-01 -1.01015115e+00 -6.98086083e-01 -3.87010098e-01 2.42705703e-01 -5.11480927e-01 1.56298995e-01 1.06059051e+00 7.79217958e-01 7.00656593e-01 9.67392698e-02 -1.30536661e-01 3.65130723e-01 2.55295992e-01 7.90138006e-01 -1.45289409e+00 2.89880130e-02 -3.51069570e-01 -2.46657103e-01 -4.33943659e-01 1.23545036e-01 -1.33719027e+00 -2.57113427e-01 -1.97615647e+00 2.49289677e-01 -4.96405065e-01 2.30828822e-01 9.51543689e-01 -1.51774839e-01 -2.05128863e-01 -2.54422605e-01 4.56708908e-01 -4.94712889e-01 1.44541785e-01 9.43608046e-01 4.74213995e-02 -3.90619695e-01 7.00999633e-04 -7.42582142e-01 6.63065791e-01 1.20814538e+00 -7.29101241e-01 -2.06386045e-01 -9.46414396e-02 3.67668003e-01 -6.39353171e-02 -2.22906888e-01 -5.88462174e-01 3.67903680e-01 -2.69361604e-02 5.97936809e-01 -8.74968290e-01 -9.59153473e-02 -8.37885976e-01 1.60299689e-01 5.41788936e-01 -5.58972836e-01 1.12582810e-01 3.53603870e-01 6.85165599e-02 -9.26057342e-03 -4.77624416e-01 3.10431898e-01 -3.63084435e-01 -2.66868085e-01 -3.39928009e-02 -6.40955865e-01 1.47430941e-01 9.04910326e-01 -2.16555864e-01 -3.34654212e-01 1.30303353e-01 -9.63597834e-01 3.45555902e-01 3.05479586e-01 2.42660165e-01 5.50996721e-01 -7.48391926e-01 -9.56015646e-01 1.19033657e-01 2.68395394e-01 2.30656669e-01 -4.00306255e-01 9.13428843e-01 -6.21456146e-01 7.40413964e-01 -1.88356102e-01 -4.52641577e-01 -1.61802971e+00 4.11285460e-01 1.37770951e-01 -7.54539549e-01 -7.41186202e-01 3.51652890e-01 -3.25141251e-01 -7.50066042e-01 1.97865814e-01 -2.19096303e-01 -3.62920374e-01 3.12855273e-01 5.18849313e-01 2.00171009e-01 5.30145586e-01 -2.33468831e-01 -2.84396827e-01 5.28781787e-02 -2.21870571e-01 -1.07869841e-02 1.74253321e+00 1.37209192e-01 -2.76536852e-01 2.24712968e-01 9.77518380e-01 -2.44415775e-02 -4.49826390e-01 -1.36919722e-01 6.67022824e-01 -2.44537905e-01 -9.91120040e-02 -1.03069508e+00 -6.76151395e-01 5.81220746e-01 1.36528760e-01 1.53871387e-01 1.00051785e+00 -5.67420665e-03 3.17825526e-01 6.54751122e-01 4.96861562e-02 -1.12891662e+00 -1.15197003e-01 5.42103469e-01 7.13339210e-01 -1.38407564e+00 1.65351972e-01 -5.82043648e-01 -5.51892877e-01 1.37892151e+00 5.46070576e-01 5.21711290e-01 8.43184769e-01 6.50857449e-01 4.47563976e-02 -5.20656288e-01 -9.99067545e-01 -4.55192290e-02 3.47420514e-01 4.28413600e-01 1.03931105e+00 -1.76617250e-01 -6.54120147e-01 7.48500645e-01 -5.18017590e-01 2.40733996e-01 5.10951340e-01 1.08592844e+00 -2.98463196e-01 -1.30650377e+00 -2.17884481e-01 1.04728580e+00 -8.53565693e-01 -2.61388332e-01 -8.15161824e-01 7.02662706e-01 -4.62432392e-02 1.07659698e+00 -1.21191032e-01 -2.88221359e-01 4.27462757e-01 2.85252690e-01 1.70547739e-01 -1.03899193e+00 -8.26202333e-01 1.17802538e-01 6.10070705e-01 -2.75740027e-01 -3.20024192e-01 -8.64394367e-01 -1.50930512e+00 -9.46102664e-02 -3.91350091e-01 6.35131717e-01 6.31289124e-01 9.85496342e-01 5.89984298e-01 3.89435589e-01 2.44729996e-01 -3.22085172e-01 -1.77379057e-01 -1.17832923e+00 -2.06356958e-01 3.55349839e-01 -2.41651431e-01 -2.28183031e-01 -1.25485212e-01 5.05053163e-01]

[8.71609878540039, 8.554275512695312]

6e03daa1-9ccc-45c3-9d29-2f850d1b849b

causal-identification-with-subjective

2212.14622

https://arxiv.org/abs/2212.14622v2

https://arxiv.org/pdf/2212.14622v2.pdf

Causal identification with subjective outcomes

Survey questions often elicit responses on ordered scales for which the definitions of the categories are subjective, possibly varying by individual. This paper clarifies what is learned when these subjective reports are used as an outcome in regression-based causal inference. When a continuous treatment variable is statistically independent of both i) potential outcomes; and ii) heterogeneity in reporting styles, a nonparametric regression of integer category numbers on that variable uncovers a positively-weighted linear combination of causal responses among individuals who are on the margin between adjacent response categories. Though the weights do not integrate to one, the ratio of local regression derivatives with respect to two such explanatory variables identifies the relative magnitudes of convex averages of their effects. When results are extended to discrete treatment variables, different weighting schemes apply to different regressors, making comparisons of magnitude less informative. I obtain a partial identification result for comparing the effects of a discrete treatment variable to those of another treatment variable when there are many categories and individual reporting functions are linear. I also provide results for identification using instrumental variables.

['Leonard Goff']

2022-12-30

['unity', 'causal-identification']

['computer-vision', 'reasoning']

[ 2.96298325e-01 6.36563301e-02 -1.20962954e+00 -6.21309757e-01 -9.41414714e-01 -9.13859129e-01 6.46423876e-01 3.24395746e-01 -4.45585608e-01 1.11302519e+00 1.02024698e+00 -5.61419010e-01 -4.48004335e-01 -8.19372654e-01 -4.72131580e-01 -5.51099062e-01 -2.20566499e-03 2.23697439e-01 -2.98127174e-01 3.23137522e-01 3.61577779e-01 1.91425756e-01 -8.40319514e-01 -1.71563238e-01 7.64815688e-01 5.67293823e-01 -3.21555048e-01 4.43623513e-01 -8.34317058e-02 7.59934962e-01 -6.32884324e-01 -1.82945088e-01 5.27072214e-02 -3.15400958e-01 -3.40537608e-01 2.23821759e-01 4.34993982e-01 -1.53265029e-01 1.24496862e-01 9.95530903e-01 -1.29947364e-02 9.89911705e-02 1.37368429e+00 -1.18328762e+00 -1.26464403e+00 7.42679656e-01 -8.06777716e-01 7.82686770e-02 6.93651915e-01 -1.59785122e-01 1.11774480e+00 -6.61958873e-01 6.14244461e-01 1.50786412e+00 6.50594413e-01 2.51603720e-04 -1.80535150e+00 -8.67452919e-01 3.05239916e-01 -3.02074373e-01 -1.11993480e+00 -4.07902896e-01 5.24110675e-01 -9.54656780e-01 9.10380259e-02 3.99306387e-01 2.02723686e-02 1.03200674e+00 2.12605804e-01 -2.29081392e-01 1.45453012e+00 -3.49719018e-01 4.12518114e-01 4.35817063e-01 5.02589464e-01 1.39289215e-01 7.24221885e-01 3.46707016e-01 -3.65458041e-01 -6.19313180e-01 8.01106453e-01 1.96208641e-01 -2.51880974e-01 5.33868186e-02 -1.16378605e+00 1.44789219e+00 2.32671410e-01 1.86731756e-01 -6.69620514e-01 1.89699158e-01 1.58790737e-01 4.71717268e-01 6.64004505e-01 4.40664202e-01 -5.52452326e-01 1.91845119e-01 -7.11917579e-01 3.67840439e-01 8.70108545e-01 3.93823475e-01 6.63744092e-01 -9.32658464e-03 -2.24637449e-01 7.75632739e-01 -7.53999874e-02 7.21877217e-01 3.19963843e-01 -1.24521291e+00 4.92907166e-01 6.10184908e-01 7.16780424e-01 -1.09345639e+00 -6.21272683e-01 -7.67612606e-02 -6.10394001e-01 2.28398845e-01 9.93833840e-01 -7.28439629e-01 -3.69449526e-01 2.11554122e+00 1.23577617e-01 -4.31046426e-01 -2.80316651e-01 8.40241373e-01 3.41994435e-01 5.31744838e-01 6.53641760e-01 -7.69832134e-01 1.40207219e+00 -1.59778565e-01 -8.38774025e-01 -2.60541618e-01 6.90618157e-01 -5.30009210e-01 9.50468183e-01 1.21733993e-02 -1.04378533e+00 -2.71812081e-01 -2.97635227e-01 2.11022049e-01 -1.56672254e-01 -2.05057353e-01 8.06789517e-01 6.93844914e-01 -7.62091875e-01 3.41558844e-01 -2.96322525e-01 -2.11192995e-01 -1.00083292e-01 3.50259990e-01 -3.48036617e-01 2.69064099e-01 -1.14880836e+00 7.79798269e-01 -3.75109375e-01 -5.86749852e-01 -1.80575937e-01 -8.89963269e-01 -6.91950500e-01 1.42056137e-01 4.08738077e-01 -5.76265633e-01 1.04142082e+00 -1.28220022e+00 -9.54210162e-01 5.63539803e-01 -3.20343524e-01 -5.82997724e-02 3.12016487e-01 3.42438638e-01 -4.63328779e-01 -2.64234930e-01 7.66594708e-01 6.60160556e-02 5.57827175e-01 -9.76556540e-01 -7.83893704e-01 -6.32771909e-01 1.18047036e-01 1.86759815e-01 -1.78286538e-01 5.65912664e-01 6.84296072e-01 -7.67841816e-01 -6.58894479e-02 -7.04189837e-01 -2.61836767e-01 -1.18756801e-01 -1.98715508e-01 -5.06036818e-01 1.69827014e-01 -5.38335562e-01 1.39623487e+00 -1.90899861e+00 -2.22822186e-02 1.77261040e-01 1.10842340e-01 -8.72192323e-01 6.25491366e-02 2.81498611e-01 -5.88718593e-01 5.46512842e-01 -1.66010410e-01 7.50282556e-02 2.64428258e-01 -1.16200134e-01 -4.11758065e-01 9.90757644e-01 1.17396891e-01 5.37753403e-01 -4.95364428e-01 -2.63071150e-01 -2.27650013e-02 -2.32726574e-01 -5.51759183e-01 -1.46426827e-01 3.68777901e-01 1.29456609e-01 -4.13475960e-01 5.01337171e-01 4.18552041e-01 -8.15394986e-03 3.83837730e-01 1.93802357e-01 -6.15263045e-01 6.59852564e-01 -1.18993330e+00 5.87894499e-01 -2.76300192e-01 5.34456611e-01 1.70700103e-01 -1.13236630e+00 8.40272307e-01 4.93891418e-01 3.18296045e-01 -3.90516400e-01 -1.27263710e-01 1.53137341e-01 2.99408566e-02 -2.59222746e-01 4.17205513e-01 -8.24455917e-01 -7.23145723e-01 3.36742610e-01 -3.23300868e-01 9.93827134e-02 5.32225743e-02 -6.77648783e-02 1.10790610e+00 -6.11040533e-01 6.74275160e-01 -5.43396533e-01 2.78607327e-02 2.85857730e-02 8.36168528e-01 8.65477979e-01 -1.19166467e-02 2.65143126e-01 1.14593709e+00 1.17316581e-02 -9.38253701e-01 -1.16515493e+00 -4.89122689e-01 1.22425604e+00 -9.24600959e-02 4.53634799e-01 -1.32029355e-01 -3.84698033e-01 5.80866098e-01 1.16019738e+00 -9.87247646e-01 8.98582637e-02 3.40394396e-03 -8.63906205e-01 6.19447567e-02 5.67839324e-01 -2.06806421e-01 -5.03992319e-01 -2.45736003e-01 8.68410543e-02 -1.93395130e-02 -5.44758856e-01 -4.94970560e-01 1.26232326e-01 -8.21414411e-01 -9.83174324e-01 -5.62417507e-01 -3.95644128e-01 5.72131991e-01 2.08905324e-01 8.66554618e-01 -2.76554316e-01 4.43896264e-01 4.54085410e-01 -7.72875100e-02 -4.22232866e-01 -1.83191031e-01 -1.92580938e-01 1.51110187e-01 -1.91115662e-01 5.00749409e-01 -5.68795681e-01 -4.34236109e-01 1.25447676e-01 -5.55133700e-01 -5.06661832e-01 4.39656898e-02 7.40239561e-01 7.34193549e-02 -1.04111902e-01 1.05401027e+00 -1.03153944e+00 6.59659922e-01 -1.10268497e+00 -6.65097773e-01 2.06732545e-02 -4.12834287e-01 -1.22342028e-01 4.40902352e-01 -8.52236867e-01 -1.31890309e+00 -4.35138822e-01 7.13745773e-01 2.69403696e-01 -3.47561121e-01 9.82071280e-01 1.38146251e-01 2.93161720e-01 9.20327365e-01 -6.46434188e-01 -2.37016395e-01 -3.71847838e-01 1.65563107e-01 5.28509140e-01 2.79483318e-01 -4.81545359e-01 8.10701311e-01 2.46001050e-01 -1.39442042e-01 -4.38666821e-01 -6.92883313e-01 -2.66154885e-01 -7.29649514e-02 1.10462524e-01 9.82088089e-01 -1.01010132e+00 -8.28866482e-01 -1.29785970e-01 -8.65157723e-01 -3.27700794e-01 -3.38469863e-01 1.19369233e+00 -3.66590172e-01 -1.87488377e-01 -4.80546236e-01 -1.22764039e+00 4.35080171e-01 -1.00503004e+00 5.99052012e-01 1.58713922e-01 -9.39560831e-01 -1.36587524e+00 3.11257597e-02 2.08199143e-01 2.29203835e-01 3.09954017e-01 1.02062011e+00 -7.77411163e-01 -6.82745278e-02 -5.75335264e-01 -3.07328939e-01 -3.97821605e-01 3.43711823e-01 2.08901942e-01 -4.34207439e-01 -1.31924525e-02 1.78821817e-01 -7.79631659e-02 4.90470380e-01 1.18692338e+00 9.02023375e-01 -8.20335627e-01 -3.76023650e-01 1.98160291e-01 1.34411478e+00 4.20502305e-01 1.96611658e-01 1.19267046e-01 3.66882741e-01 9.79909301e-01 3.64420295e-01 4.97135967e-01 4.10153896e-01 4.68009681e-01 7.16377050e-02 -2.83405244e-01 5.63336968e-01 -1.58904433e-01 5.35793006e-01 1.30859524e-01 -2.64484473e-02 -6.57922626e-02 -5.31223536e-01 7.70092249e-01 -1.46188045e+00 -1.31677437e+00 -5.73410749e-01 2.52677989e+00 8.99368644e-01 1.11152738e-01 4.58334714e-01 -1.62629098e-01 9.90313351e-01 1.05604522e-01 -3.23401779e-01 -6.14903569e-01 -9.22541227e-03 1.75040677e-01 1.18970680e+00 9.17163849e-01 -8.46798420e-01 2.63175666e-01 7.54309654e+00 5.02732456e-01 -8.28175545e-01 6.25964627e-03 8.36084604e-01 -1.33803666e-01 -7.76365399e-01 4.86096054e-01 -4.41541642e-01 4.65809137e-01 1.01712513e+00 -6.34390771e-01 2.01535180e-01 6.16061985e-01 8.65757167e-01 -5.60254276e-01 -1.03505921e+00 1.57408975e-02 -6.01685107e-01 -9.25709665e-01 -5.19006550e-01 5.17285168e-01 1.04621279e+00 -6.35543466e-01 3.19327533e-01 1.46266684e-01 9.64419901e-01 -1.12784076e+00 9.06335771e-01 4.60850328e-01 9.65881586e-01 -6.62908494e-01 5.22916138e-01 2.91167617e-01 -5.77363968e-01 -4.57769006e-01 -3.22277308e-01 -8.73109460e-01 7.50760883e-02 6.65881097e-01 -6.44008934e-01 -7.99328610e-02 1.28239051e-01 1.40964329e-01 -6.87067807e-02 5.48234940e-01 -2.25265354e-01 1.15774190e+00 -1.13673478e-01 5.76665923e-02 5.20033156e-03 -4.92025197e-01 4.24618900e-01 9.43111837e-01 4.25102562e-01 5.09513557e-01 -4.13910896e-02 1.24255264e+00 -8.29591304e-02 9.81972143e-02 -7.72543550e-01 -1.15265980e-01 9.01010334e-01 8.85622621e-01 -5.58277011e-01 -3.21470529e-01 -9.89502192e-01 1.06541164e-01 2.07510114e-01 7.48251140e-01 -6.69510007e-01 -4.96218987e-02 7.66488731e-01 4.32232678e-01 -2.55932122e-01 -1.14411034e-01 -9.86820996e-01 -9.13475096e-01 -2.81755805e-01 -5.11415899e-01 5.53191185e-01 -6.01948380e-01 -1.43003392e+00 -6.18656933e-01 3.92063320e-01 -5.98101556e-01 -4.00619835e-01 -2.81480402e-01 -6.97674990e-01 1.40419865e+00 -8.93699706e-01 -5.21368265e-01 3.11996341e-01 2.85649121e-01 1.39703423e-01 1.14745036e-01 9.40255821e-01 -2.90979687e-02 -5.10465503e-01 5.17551422e-01 1.70978457e-01 8.70378241e-02 8.61741543e-01 -1.41013062e+00 -1.23296320e-01 5.03519595e-01 -3.51049870e-01 8.82895172e-01 9.40495193e-01 -8.52415502e-01 -7.97073781e-01 -7.24095941e-01 1.26638567e+00 -3.55877429e-01 1.10267425e+00 1.24049045e-01 -7.92429566e-01 9.99428630e-01 1.70967504e-02 -5.13118327e-01 8.16274822e-01 7.56419420e-01 -3.90112817e-01 6.10544905e-02 -1.17339861e+00 7.22115636e-01 5.56570411e-01 -6.32669747e-01 -5.97769678e-01 2.61062264e-01 8.18265140e-01 3.16558689e-01 -1.16992831e+00 4.37239073e-02 5.27175307e-01 -9.28555608e-01 6.52631760e-01 -7.27977157e-01 5.75306535e-01 6.66522188e-03 -3.17064166e-01 -1.31615627e+00 -7.89258718e-01 -5.27983271e-02 8.71505022e-01 1.31169140e+00 5.42079985e-01 -1.05425322e+00 4.61548954e-01 1.22660661e+00 2.90155739e-01 -2.10559532e-01 -9.08013940e-01 -3.35596293e-01 4.70206350e-01 -1.93619221e-01 3.36016774e-01 1.57769728e+00 3.96672666e-01 5.06176233e-01 -3.91544513e-02 2.96598021e-02 6.97479844e-01 1.90880939e-01 4.06728864e-01 -1.50129187e+00 -1.37846082e-01 -4.07280087e-01 -1.11981668e-01 -4.60427254e-01 2.64869332e-01 -4.50934172e-01 -5.69246948e-01 -1.13818038e+00 5.93496025e-01 -4.20199186e-01 1.55912399e-01 4.05436784e-01 -1.68042704e-01 -1.47534594e-01 -1.50121152e-02 1.24951050e-01 1.20651178e-01 -2.53754482e-02 7.45588958e-01 1.64351150e-01 -4.79272246e-01 1.28698751e-01 -1.22363210e+00 1.00964856e+00 7.54021466e-01 -7.13678658e-01 -1.54016227e-01 1.09357201e-01 2.45213747e-01 9.91976798e-01 6.24424338e-01 -3.23223591e-01 2.97164880e-02 -1.15346789e+00 4.58304912e-01 -1.65484920e-01 -7.00712996e-03 -7.12261796e-01 6.41443610e-01 1.84516296e-01 -1.02224851e+00 3.60676982e-02 -1.95900053e-01 4.24158990e-01 8.19984227e-02 -3.43778074e-01 5.78507006e-01 -1.83466841e-02 1.13620274e-01 -1.25911966e-01 -8.29769850e-01 1.07019007e-01 7.08165109e-01 -1.69278309e-01 -3.37614030e-01 -9.28547502e-01 -6.51732504e-01 -1.04362229e-02 6.98868632e-01 2.44786385e-02 4.97229323e-02 -1.57521880e+00 -8.37455630e-01 -4.96171057e-01 -1.31790608e-01 -7.45579243e-01 5.06458282e-02 9.55653369e-01 3.45384985e-01 5.07800817e-01 3.86537500e-02 1.75101861e-01 -1.06953394e+00 6.94316566e-01 2.70246297e-01 -1.92374811e-01 1.20597824e-01 5.98724663e-01 9.63820994e-01 -1.40569702e-01 -1.52899161e-01 -1.83797285e-01 -1.54031143e-01 7.10289657e-01 3.89650196e-01 7.41180003e-01 -6.00484490e-01 -5.47815144e-01 -2.84562528e-01 2.94291228e-01 3.29229891e-01 -5.99392176e-01 1.21743345e+00 -4.34007287e-01 -2.57982582e-01 9.60730433e-01 1.19802499e+00 4.57228869e-01 -1.06380606e+00 -2.21183524e-01 1.02609523e-01 -3.95501673e-01 -7.28679821e-02 -7.13703275e-01 -7.97703147e-01 1.64717197e-01 9.66263935e-02 6.54462039e-01 9.11931992e-01 -2.66856831e-02 -3.17513227e-01 -3.63258839e-01 2.07024664e-01 -9.36216950e-01 -3.22127879e-01 -3.88124050e-03 9.86822367e-01 -1.14729679e+00 1.91998780e-01 -4.69549865e-01 -5.04358590e-01 7.63865530e-01 -2.57446598e-02 -4.67143893e-01 3.11188668e-01 6.82156384e-02 -2.06522569e-01 -1.17008619e-01 -7.56164968e-01 1.69866532e-01 2.17355758e-01 5.23220599e-01 9.51052785e-01 7.18784690e-01 -1.36239552e+00 7.33430386e-01 -3.42060000e-01 -2.86296636e-01 9.17782545e-01 2.68807858e-01 -4.62137967e-01 -6.69779837e-01 -1.06783080e+00 8.48264694e-01 -7.81382799e-01 -1.00294702e-01 -4.04956490e-01 1.12710953e+00 -1.35849997e-01 1.33386350e+00 5.21409094e-01 -1.46478070e-02 3.49826097e-01 1.09069385e-01 -2.34137140e-02 -5.54542959e-01 -3.16992044e-01 1.15459643e-01 4.51107860e-01 -1.51010007e-01 -5.28803587e-01 -1.10576725e+00 -1.02076137e+00 -7.64148533e-01 -4.75890607e-01 1.64824396e-01 3.30485493e-01 8.15193892e-01 -3.83307695e-01 2.42164940e-01 9.16702151e-01 -6.61192834e-01 -9.54482794e-01 -9.76461411e-01 -1.06815815e+00 2.97293872e-01 5.64706087e-01 -8.68805826e-01 -9.01885092e-01 -9.50984135e-02]

[8.003931045532227, 5.275940418243408]

2714194b-6c0a-4f31-b17b-604cf462a6eb

machine-translation-by-projecting-text-into

2305.12371

https://arxiv.org/abs/2305.12371v1

https://arxiv.org/pdf/2305.12371v1.pdf

Machine Translation by Projecting Text into the Same Phonetic-Orthographic Space Using a Common Encoding

The use of subword embedding has proved to be a major innovation in Neural Machine Translation (NMT). It helps NMT to learn better context vectors for Low Resource Languages (LRLs) so as to predict the target words by better modelling the morphologies of the two languages and also the morphosyntax transfer. Even so, their performance for translation in Indian language to Indian language scenario is still not as good as for resource-rich languages. One reason for this is the relative morphological richness of Indian languages, while another is that most of them fall into the extremely low resource or zero-shot categories. Since most major Indian languages use Indic or Brahmi origin scripts, the text written in them is highly phonetic in nature and phonetically similar in terms of abstract letters and their arrangements. We use these characteristics of Indian languages and their scripts to propose an approach based on common multilingual Latin-based encodings (WX notation) that take advantage of language similarity while addressing the morphological complexity issue in NMT. These multilingual Latin-based encodings in NMT, together with Byte Pair Embedding (BPE) allow us to better exploit their phonetic and orthographic as well as lexical similarities to improve the translation quality by projecting different but similar languages on the same orthographic-phonetic character space. We verify the proposed approach by demonstrating experiments on similar language pairs (Gujarati-Hindi, Marathi-Hindi, Nepali-Hindi, Maithili-Hindi, Punjabi-Hindi, and Urdu-Hindi) under low resource conditions. The proposed approach shows an improvement in a majority of cases, in one case as much as ~10 BLEU points compared to baseline techniques for similar language pairs. We also get up to ~1 BLEU points improvement on distant and zero-shot language pairs.

['Anil Kumar Singh', 'Ajay Pratap', 'Shantipriya Parida', 'Amit Kumar']

2023-05-21

['nmt']

['computer-code']

[-4.04002406e-02 -3.53491217e-01 -2.71687925e-01 -2.35706806e-01 -9.01040494e-01 -8.33215952e-01 7.74516404e-01 8.75455290e-02 -5.99246144e-01 1.01301527e+00 4.61119890e-01 -7.46955752e-01 1.53530553e-01 -8.39379907e-01 -7.55987525e-01 -5.71068347e-01 3.06749612e-01 7.95695722e-01 -1.10565729e-01 -6.54647827e-01 2.37688869e-01 4.81939346e-01 -1.13521266e+00 1.67772278e-01 1.01952636e+00 7.94084892e-02 6.40246153e-01 4.26601142e-01 -5.71460664e-01 1.67260125e-01 -1.60018712e-01 -5.45889378e-01 4.84654725e-01 -6.96074128e-01 -7.60065615e-01 -4.16146427e-01 3.80624861e-01 9.10246670e-02 -3.18225026e-01 9.78667140e-01 6.63349271e-01 -7.43887126e-02 8.64093840e-01 -5.30998468e-01 -9.46233630e-01 7.92176187e-01 -7.40512729e-01 2.60249496e-01 2.30698749e-01 -5.67009719e-03 1.09465182e+00 -1.09279883e+00 7.19007552e-01 1.38911319e+00 6.70976579e-01 3.24281693e-01 -1.12097812e+00 -5.09607852e-01 -3.40959460e-01 2.85816133e-01 -1.59653306e+00 -4.33054030e-01 4.93840635e-01 -2.56593853e-01 1.16687608e+00 4.33157235e-01 4.08910871e-01 9.79964375e-01 3.97815317e-01 4.44875211e-01 1.50321889e+00 -8.40860605e-01 -9.51658189e-02 5.43893576e-01 -1.05438001e-01 5.21737993e-01 2.22536564e-01 -4.79206815e-02 -2.46415600e-01 8.49359557e-02 6.46127582e-01 -1.30851477e-01 -4.17780317e-02 1.16021015e-01 -1.35629165e+00 9.76234019e-01 2.11383685e-01 8.27411056e-01 -2.60065943e-01 -1.81351617e-01 4.40236390e-01 4.13672745e-01 2.10697442e-01 1.60344154e-01 -3.50964576e-01 -2.77964026e-01 -8.39815259e-01 9.38985422e-02 5.04495263e-01 1.03337300e+00 8.73747170e-01 3.94837320e-01 1.50999412e-01 1.31606209e+00 3.26494366e-01 7.87828088e-01 9.76776302e-01 -1.48640573e-01 7.49948502e-01 1.94497272e-01 -2.51163661e-01 -8.63740861e-01 7.29085654e-02 -2.53268749e-01 -7.40089774e-01 -1.03852414e-01 3.64193946e-01 1.29830360e-01 -8.21352184e-01 1.82900965e+00 2.07656734e-02 -2.53620863e-01 4.84906226e-01 5.70942760e-01 4.23643559e-01 1.29291487e+00 -1.41516268e-01 -3.36987525e-01 1.49531281e+00 -1.02764761e+00 -4.79052395e-01 -2.98531592e-01 5.31181991e-01 -1.27037930e+00 1.57500637e+00 -9.83189568e-02 -1.07174945e+00 -7.33558893e-01 -9.78115559e-01 -4.87100929e-02 -6.23951554e-01 1.20289065e-01 2.28922829e-01 7.87348986e-01 -8.60436857e-01 6.48787737e-01 -4.90772128e-01 -8.10235977e-01 -7.66705871e-02 3.04637432e-01 -4.98935193e-01 -1.39760375e-01 -1.22522080e+00 1.07820678e+00 6.21975124e-01 -1.60354733e-01 -4.53710347e-01 -3.76911849e-01 -7.73502409e-01 -9.14532170e-02 -1.61318123e-01 -1.12914756e-01 6.61922395e-01 -9.74156976e-01 -1.59697211e+00 8.88785839e-01 -2.25170597e-01 -3.38177741e-01 4.52566773e-01 1.01030298e-01 -5.22642434e-01 -2.76122719e-01 3.00122071e-02 5.21268547e-01 4.09455478e-01 -1.00174344e+00 -5.97773075e-01 -4.50234562e-01 -3.06984991e-01 3.82956415e-01 -4.71522003e-01 5.08233845e-01 -4.84082013e-01 -1.02765489e+00 2.97351554e-02 -1.12318814e+00 -7.06086382e-02 -6.03754044e-01 -1.58165082e-01 1.10706761e-01 6.34781599e-01 -1.10552061e+00 1.04710817e+00 -1.83918631e+00 2.23747179e-01 1.50932167e-02 -6.66572213e-01 4.47062045e-01 -3.30808938e-01 7.67936409e-01 -4.78314385e-02 1.63187787e-01 -3.48592430e-01 5.43487212e-03 -1.02370344e-01 8.23452353e-01 -1.85701385e-01 3.64367783e-01 1.72964752e-01 9.88250494e-01 -6.43509686e-01 -4.25551474e-01 2.02860475e-01 5.44347286e-01 -2.60339051e-01 -2.65504215e-02 2.62636662e-01 5.33448815e-01 2.40278374e-02 4.23820347e-01 5.69905758e-01 7.26235926e-01 2.73524672e-01 3.06777395e-02 -4.35371131e-01 5.47658622e-01 -1.06246781e+00 1.52074027e+00 -9.44449246e-01 4.74042952e-01 -3.97402316e-01 -1.04983330e+00 1.05729878e+00 4.26888704e-01 1.49562238e-02 -8.23836327e-01 9.52007398e-02 8.80368590e-01 4.21968669e-01 -2.69569755e-01 5.11134744e-01 -4.40483481e-01 -1.09937645e-01 4.88316715e-01 9.96697396e-02 -1.86107248e-01 1.68213442e-01 -3.46451223e-01 4.34750795e-01 2.67085761e-01 6.28662825e-01 -6.28383100e-01 7.39113212e-01 -1.11592956e-01 6.35456741e-01 4.09451663e-01 7.76296929e-02 6.63734257e-01 -3.09911510e-03 -2.93066978e-01 -1.54632628e+00 -1.04193103e+00 -8.49431381e-02 9.48776603e-01 -2.68226922e-01 -1.29062742e-01 -6.33468688e-01 -3.20311695e-01 -3.12539637e-01 9.37999785e-01 -2.51134545e-01 1.09356865e-01 -1.25779545e+00 -7.97865629e-01 7.11523056e-01 3.22949916e-01 2.97468662e-01 -1.19013858e+00 -7.01139942e-02 4.93680686e-01 -1.99816272e-01 -1.04306090e+00 -5.12400568e-01 3.15134227e-01 -9.99771476e-01 -2.88877666e-01 -1.10396719e+00 -1.17301679e+00 3.15743417e-01 1.44137159e-01 8.21307182e-01 -4.74597037e-01 -9.90483984e-02 -2.59611934e-01 -4.22208637e-01 -3.98806900e-01 -9.27881539e-01 1.51289657e-01 5.16780317e-01 -8.42777044e-02 4.07412827e-01 -6.12956583e-01 -2.84981206e-02 1.91717774e-01 -8.07178795e-01 2.64129490e-02 6.98618293e-01 7.33638048e-01 6.56216264e-01 -2.44170919e-01 3.92315596e-01 -8.10706675e-01 3.16649050e-01 -4.67628688e-01 -3.88124228e-01 3.69783998e-01 -5.20300984e-01 2.11814612e-01 1.07829189e+00 -4.07094300e-01 -9.13045943e-01 -1.23311296e-01 -5.43506145e-01 -4.09769965e-03 -8.79024044e-02 4.22337413e-01 -3.89823645e-01 4.72341180e-02 4.89681005e-01 7.07942426e-01 -2.48583987e-01 -7.93101370e-01 3.98623109e-01 1.02383089e+00 5.89192212e-01 -6.86143458e-01 1.00235760e+00 2.74355058e-02 -3.61030757e-01 -1.10276139e+00 8.08708966e-02 -2.89229423e-01 -8.61727834e-01 3.10696721e-01 9.18891013e-01 -7.88291156e-01 4.14231308e-02 3.17673206e-01 -1.25640869e+00 5.17041534e-02 -1.66750431e-01 7.64171660e-01 -4.68041331e-01 4.88705128e-01 -8.56699467e-01 -7.08054543e-01 -4.39647079e-01 -1.34754145e+00 6.20793343e-01 9.61643904e-02 -1.33059472e-01 -1.05739260e+00 2.30658174e-01 1.77423984e-01 5.32186925e-01 -2.66337264e-02 1.67994463e+00 -8.01421702e-01 -2.37409860e-01 8.32427219e-02 -1.17462493e-01 5.30796289e-01 4.24637854e-01 -1.87272266e-01 -6.52451634e-01 -4.00526822e-01 -8.03088248e-02 4.27747928e-02 4.75742906e-01 4.32446338e-02 1.97827697e-01 -3.26431751e-01 1.42666504e-01 5.06054521e-01 1.84958315e+00 4.80937272e-01 6.73326254e-01 4.19490099e-01 7.90848136e-01 5.18694878e-01 4.11290824e-01 -7.31200576e-02 2.82693535e-01 9.92449343e-01 -1.95927531e-01 -6.61643744e-02 -3.45524967e-01 -2.92577863e-01 9.33633268e-01 1.89806592e+00 -1.39506087e-01 -7.44111836e-02 -1.10392928e+00 8.86364043e-01 -1.41812468e+00 -6.28559053e-01 -2.68979669e-01 2.44180322e+00 1.00522017e+00 3.64002064e-02 8.64037722e-02 1.04580313e-01 6.42776310e-01 -3.64324227e-02 -1.55878696e-03 -1.27037704e+00 -3.74449968e-01 6.51259482e-01 6.65620267e-01 6.52637780e-01 -4.63192075e-01 1.36046398e+00 5.09008741e+00 1.07603824e+00 -1.28500283e+00 3.68692100e-01 3.37575346e-01 3.90412092e-01 -5.03687918e-01 1.34355158e-01 -7.10795939e-01 5.40810645e-01 1.34557033e+00 -2.06129834e-01 6.28874004e-01 4.49931830e-01 2.70719618e-01 3.17566276e-01 -9.57150221e-01 9.75501359e-01 1.15742292e-02 -1.18642569e+00 2.98341900e-01 2.18919367e-01 7.70175755e-01 4.05996621e-01 5.15178069e-02 3.50779116e-01 -1.97313819e-03 -1.02034831e+00 7.40824461e-01 -1.08166784e-01 1.06422448e+00 -9.82730210e-01 8.21642995e-01 3.22012573e-01 -1.22696853e+00 2.90578812e-01 -8.51364493e-01 6.15220964e-02 1.03827700e-01 2.52795592e-02 -8.18428278e-01 7.95743525e-01 4.01690543e-01 3.13515365e-01 -3.97653610e-01 5.59548676e-01 -1.62614703e-01 9.16193545e-01 -3.39852095e-01 -8.21016654e-02 6.74093246e-01 -5.45152366e-01 5.77237785e-01 1.55612373e+00 7.07275152e-01 -2.94252276e-01 -2.54624598e-02 5.01178324e-01 1.13464519e-01 1.01613319e+00 -8.29380274e-01 -2.69375920e-01 2.80602783e-01 7.79724360e-01 -7.07461417e-01 -3.21202189e-01 -5.67251861e-01 1.30807209e+00 1.74306348e-01 1.22900784e-01 -8.04031193e-01 -4.53280330e-01 7.23669469e-01 1.18648643e-02 3.34080577e-01 -4.94491041e-01 -1.93231881e-01 -1.01798630e+00 4.14912999e-02 -1.22832155e+00 -9.11092460e-02 -3.36961180e-01 -1.08352697e+00 1.01839757e+00 -1.63033769e-01 -1.17019498e+00 -4.15169835e-01 -7.04434574e-01 -4.71130937e-01 1.32360160e+00 -1.42597330e+00 -1.36856329e+00 5.10651529e-01 4.78412896e-01 9.54970121e-01 -6.64803922e-01 1.14617026e+00 6.23564720e-01 -3.95754337e-01 8.03420722e-01 7.82188773e-01 8.17800835e-02 6.81839824e-01 -1.18655574e+00 8.15918148e-01 9.58950877e-01 7.64932394e-01 6.87251270e-01 6.41383827e-01 -5.01245022e-01 -1.43647003e+00 -1.07197297e+00 1.48640287e+00 -2.56346494e-01 6.12149417e-01 -5.12044430e-01 -9.08178985e-01 5.82300067e-01 4.83585358e-01 -5.07928431e-01 8.59934568e-01 -3.81202996e-02 -3.28446865e-01 2.44677067e-02 -9.33405161e-01 1.00409174e+00 8.46026599e-01 -6.73950255e-01 -8.47097039e-01 3.17645639e-01 6.04627132e-01 1.17181249e-01 -6.51282609e-01 9.31138247e-02 6.60193801e-01 -7.56587446e-01 7.78387308e-01 -6.27887368e-01 4.53365296e-01 -2.35241011e-01 -6.70331180e-01 -1.49391854e+00 -3.12681079e-01 -5.53983152e-01 5.77987373e-01 1.49596250e+00 5.47079980e-01 -6.67309463e-01 3.26496482e-01 -2.98666388e-01 -9.49132666e-02 -5.38188517e-01 -1.04711473e+00 -1.11614096e+00 6.01212382e-01 -3.59999508e-01 4.68289942e-01 1.12398851e+00 -3.94884437e-01 5.34824908e-01 -7.12869883e-01 -1.05089545e-01 2.11505905e-01 4.48190793e-02 6.66083872e-01 -9.10719633e-01 -5.97030818e-01 -4.15693253e-01 -4.73393798e-01 -5.88734388e-01 1.02577575e-01 -1.42008507e+00 -9.94583368e-02 -1.30832624e+00 1.14800721e-01 -5.88214517e-01 -4.20894951e-01 3.88974816e-01 -1.26338437e-01 6.42656326e-01 4.77122903e-01 3.01789880e-01 4.32036221e-01 2.97481984e-01 8.42743576e-01 -8.37165769e-03 -2.12811738e-01 -2.52876252e-01 -4.81414497e-01 5.95092535e-01 7.33927131e-01 -6.68517768e-01 -1.92851990e-01 -8.89243424e-01 6.00641780e-02 -1.14114128e-01 -4.05639350e-01 -1.04016256e+00 -3.10106486e-01 -3.10516268e-01 3.16585675e-02 -3.44488174e-01 1.38280690e-01 -8.61400843e-01 2.54301339e-01 5.68167448e-01 -1.11494027e-01 6.10077977e-01 3.32746625e-01 1.87788174e-01 -1.82300925e-01 -3.89834166e-01 9.92710888e-01 -2.54210085e-01 -7.27830291e-01 1.04273960e-01 -6.02923751e-01 -7.85053000e-02 8.47146749e-01 -6.16549134e-01 2.26157114e-01 -1.43868357e-01 -1.20123684e-01 -4.03311968e-01 5.85941195e-01 7.63987422e-01 3.12384129e-01 -1.48126721e+00 -1.16674232e+00 4.28106576e-01 2.24390611e-01 -8.39153767e-01 -3.55962217e-02 6.94432914e-01 -8.62716794e-01 5.17210662e-01 -8.34800184e-01 -2.86710382e-01 -1.28422701e+00 3.94741833e-01 -5.39979115e-02 -2.95017362e-01 -3.70277375e-01 6.51126862e-01 6.89843297e-02 -8.67707372e-01 -9.35396850e-02 -1.19082078e-01 -1.22583807e-01 -9.22210142e-02 2.70053089e-01 2.26606861e-01 1.23350330e-01 -1.35015440e+00 -3.77766967e-01 9.03286219e-01 -1.39599755e-01 -3.51124942e-01 1.26601887e+00 -1.42007902e-01 -3.02078098e-01 6.84431911e-01 1.44833791e+00 7.84544766e-01 -1.91142678e-01 -2.67526388e-01 1.69596583e-01 -4.58468080e-01 -3.36422652e-01 -6.75967455e-01 -5.63403606e-01 1.25490892e+00 8.07030141e-01 -4.23251331e-01 8.62797022e-01 -2.87680984e-01 1.19906855e+00 2.41120562e-01 4.58556235e-01 -1.11680269e+00 -5.31003594e-01 8.89342129e-01 6.23551190e-01 -1.03012574e+00 -3.99681181e-01 -4.93932366e-02 -7.52598822e-01 1.22125101e+00 1.08776040e-01 -3.43498774e-02 1.49058685e-01 2.74075530e-02 3.56186122e-01 4.51509356e-01 -2.88875639e-01 -2.64107466e-01 2.12805510e-01 5.70111275e-01 7.55174875e-01 3.26871008e-01 -1.03328419e+00 2.23270088e-01 -5.12756646e-01 -5.34762204e-01 4.78792787e-01 7.76852787e-01 -3.39587599e-01 -1.74482346e+00 -4.79815781e-01 1.21749274e-01 -6.21060491e-01 -4.40190971e-01 -2.12561548e-01 8.72863948e-01 2.83479124e-01 6.57190561e-01 -4.11152579e-02 -4.21527535e-01 1.59197927e-01 4.09122825e-01 4.26742315e-01 -7.07727492e-01 -7.48651981e-01 3.10611397e-01 -9.20405015e-02 -9.41035748e-02 2.65262485e-03 -4.81265366e-01 -1.11744523e+00 -4.13614362e-01 -7.31926784e-02 2.27136388e-01 9.76048291e-01 8.79892826e-01 -6.17290922e-02 1.20199859e-01 7.73101866e-01 -5.56969941e-01 -5.50984323e-01 -1.00491762e+00 -5.18531203e-01 3.22122902e-01 -2.66435236e-01 -1.98505893e-01 9.00458172e-02 -7.72468299e-02]

[11.282066345214844, 10.221285820007324]

4f2aa9ce-df66-487e-aba1-7ae5c65f7c1f

saliency-based-multi-view-mixed-language

https://aclanthology.org/2021.findings-emnlp.55

https://aclanthology.org/2021.findings-emnlp.55.pdf

Saliency-based Multi-View Mixed Language Training for Zero-shot Cross-lingual Classification

Recent multilingual pre-trained models, like XLM-RoBERTa (XLM-R), have been demonstrated effective in many cross-lingual tasks. However, there are still gaps between the contextualized representations of similar words in different languages. To solve this problem, we propose a novel framework named Multi-View Mixed Language Training (MVMLT), which leverages code-switched data with multi-view learning to fine-tune XLM-R. MVMLT uses gradient-based saliency to extract keywords which are the most relevant to downstream tasks and replaces them with the corresponding words in the target language dynamically. Furthermore, MVMLT utilizes multi-view learning to encourage contextualized embeddings to align into a more refined language-invariant space. Extensive experiments with four languages show that our model achieves state-of-the-art results on zero-shot cross-lingual sentiment classification and dialogue state tracking tasks, demonstrating the effectiveness of our proposed model.

['Jian Liu', 'Jinan Xu', 'Yufeng Chen', 'Dong Jing', 'Hui Huang', 'Siyu Lai']

findings-emnlp-2021-11

['multi-view-learning']

['computer-vision']

[-2.79877841e-01 -2.42232054e-01 -7.61257529e-01 -4.74673033e-01 -1.06353593e+00 -7.14637697e-01 8.47893059e-01 -3.83897088e-02 -3.76310706e-01 4.23222125e-01 8.59968483e-01 -2.65127450e-01 6.28764093e-01 -8.38373136e-03 -4.71237957e-01 -2.57905185e-01 3.56826276e-01 2.81709522e-01 8.56467187e-02 -8.00755084e-01 3.12890798e-01 -1.92001075e-01 -1.06029069e+00 6.63803935e-01 1.09845781e+00 2.13210285e-01 5.99634945e-01 4.53620553e-01 -4.65958446e-01 5.39443851e-01 -3.90384883e-01 -5.72550476e-01 -1.00977182e-01 -3.75917315e-01 -7.52911508e-01 -1.29866496e-01 4.09731030e-01 6.57569543e-02 -4.82328050e-03 9.66886580e-01 6.01186037e-01 1.42375484e-01 4.71149743e-01 -9.81419146e-01 -1.05242407e+00 8.35508108e-01 -9.44320142e-01 1.74764678e-01 3.83280724e-01 -4.09493633e-02 1.33507562e+00 -1.44194353e+00 8.70563924e-01 1.69661689e+00 4.62441236e-01 1.04021764e+00 -1.00725353e+00 -6.57855272e-01 8.25093865e-01 -2.00205352e-02 -8.72291863e-01 -7.05814242e-01 5.71829498e-01 -1.83950841e-01 1.26949513e+00 -2.96041127e-02 3.75892252e-01 1.44725895e+00 3.14765543e-01 1.25988400e+00 1.19198096e+00 -6.28264844e-01 -1.49514213e-01 5.49796581e-01 1.01506576e-01 8.34741414e-01 -2.54823323e-02 -3.83181930e-01 -1.05620527e+00 1.14868702e-02 -8.30902457e-02 -1.30472153e-01 -2.28522032e-01 -5.13794005e-01 -1.34794807e+00 1.10273957e+00 2.61755791e-02 3.98069203e-01 1.42620161e-01 -1.98352754e-01 1.00005198e+00 8.73218700e-02 9.21060860e-01 6.22820795e-01 -9.96874332e-01 -1.12278379e-01 -7.43107855e-01 -2.27724612e-01 5.51747501e-01 1.05873358e+00 5.27317286e-01 4.65941206e-02 -4.46644515e-01 1.23257327e+00 6.01935804e-01 6.10525131e-01 9.21584368e-01 -5.55613041e-01 9.31946576e-01 8.99552107e-01 -2.31664747e-01 -1.77264839e-01 -1.74048722e-01 -2.71028906e-01 -4.08946365e-01 -1.44893602e-01 2.91802771e-02 -2.77384579e-01 -6.46049798e-01 1.82305193e+00 2.41878107e-01 -2.05613688e-01 7.33988643e-01 6.04404390e-01 8.62006307e-01 8.55637968e-01 1.08069204e-01 -6.39858283e-03 1.46547341e+00 -1.67020464e+00 -7.84817755e-01 -7.26846695e-01 1.04170597e+00 -8.98741305e-01 1.54703426e+00 5.37367798e-02 -7.23669767e-01 -4.70631450e-01 -1.00867200e+00 -4.02136326e-01 -5.91208696e-01 4.36968505e-01 4.71534997e-01 3.64485145e-01 -8.17619145e-01 1.04273029e-01 -6.21037662e-01 -4.56241310e-01 1.54050514e-01 -1.89396054e-01 -3.09409350e-01 -1.71193212e-01 -1.56120992e+00 1.13323295e+00 -1.09100249e-02 -1.27570212e-01 -9.95678425e-01 -5.55272341e-01 -1.27764463e+00 -2.09150538e-01 2.96983451e-01 -1.33527830e-01 1.24666154e+00 -7.45032132e-01 -1.62127125e+00 1.16365504e+00 -7.15518415e-01 -1.73704609e-01 1.82065114e-01 -6.40051961e-01 -5.92976391e-01 -1.54647067e-01 5.80053568e-01 6.66881323e-01 7.86042392e-01 -1.22327375e+00 -5.79224110e-01 -2.83944190e-01 -1.95466205e-02 6.48646474e-01 -5.29483438e-01 2.31264144e-01 -7.65194654e-01 -4.97325152e-01 -3.04043442e-01 -9.66828406e-01 -1.79052755e-01 -3.40756834e-01 -4.05108690e-01 -6.54625595e-01 8.00843358e-01 -6.85632706e-01 1.34742022e+00 -2.04524851e+00 5.56593716e-01 -5.45010149e-01 -1.42751187e-01 4.57813859e-01 -3.10924053e-01 5.16247451e-01 7.43107572e-02 -3.98281515e-02 2.82838549e-02 -7.95091331e-01 -2.40486920e-01 3.64757106e-02 -4.86247033e-01 3.97363275e-01 5.41777723e-02 9.27022934e-01 -1.26038384e+00 -4.73760992e-01 6.89655542e-02 2.80701071e-01 -4.66918945e-01 2.38747567e-01 -3.41275662e-01 5.39829791e-01 -4.16817367e-01 4.86770868e-01 2.64054805e-01 -3.22258204e-01 2.93152153e-01 -6.90114051e-02 -4.01832610e-01 5.55852830e-01 -4.08762813e-01 2.46409464e+00 -1.08060908e+00 6.13228440e-01 1.10484496e-01 -5.84907591e-01 9.15576279e-01 3.35490614e-01 6.06658123e-02 -6.21584237e-01 -1.82349645e-02 1.66463807e-01 -3.11262578e-01 -5.48533499e-01 8.41533542e-01 1.19894892e-01 -5.64632058e-01 5.82399011e-01 4.23305690e-01 -5.51631162e-03 2.96391901e-02 5.45745134e-01 3.43795300e-01 5.56394935e-01 5.39342105e-01 -3.95799845e-01 8.54362071e-01 -1.17235519e-01 6.11976206e-01 4.02169943e-01 -4.50054795e-01 1.75203532e-01 4.51670259e-01 3.46830599e-02 -7.80943096e-01 -8.12665939e-01 2.41417587e-02 1.88024592e+00 1.48561567e-01 -5.42656481e-01 -6.80394113e-01 -1.17422485e+00 -1.58616975e-01 1.12401867e+00 -6.19159162e-01 -2.79812962e-01 -5.05137980e-01 -4.66854244e-01 3.41079235e-01 2.98098773e-01 2.73350328e-01 -9.15190578e-01 -3.38655889e-01 1.06555514e-01 -4.11763221e-01 -1.18392599e+00 -1.16364408e+00 2.18098551e-01 -6.74556851e-01 -7.30753958e-01 -7.14679241e-01 -1.04866076e+00 5.85202217e-01 6.36275768e-01 1.18883717e+00 -3.67475182e-01 3.99656594e-03 3.80039275e-01 -5.10490894e-01 -2.22687855e-01 -6.38190031e-01 4.73024905e-01 1.50298446e-01 6.35493100e-02 6.33380413e-01 1.54825464e-01 -4.57820743e-01 4.78790402e-02 -2.85208315e-01 1.70790508e-01 3.28817606e-01 9.31054115e-01 3.75655770e-01 -9.94472206e-01 7.80265391e-01 -1.15223777e+00 9.44151461e-01 -4.25009727e-01 -4.14266914e-01 7.65656650e-01 -7.12575316e-01 3.66248518e-01 6.97215259e-01 -4.14010614e-01 -1.21647108e+00 -7.78262466e-02 6.40181154e-02 -3.57559830e-01 3.04812253e-01 4.32556450e-01 -7.87088647e-02 4.20910060e-01 4.48419958e-01 4.82871264e-01 -1.21514238e-01 -4.51134145e-01 9.78921175e-01 8.61464739e-01 2.05049977e-01 -4.00993586e-01 2.85892218e-01 4.11115259e-01 -8.00483227e-01 -5.99645138e-01 -1.41234148e+00 -7.56741285e-01 -7.06623614e-01 -1.28706381e-01 1.08912122e+00 -1.62435126e+00 -1.97788492e-01 3.70161444e-01 -1.23807085e+00 -1.53485283e-01 2.15466768e-02 3.23577195e-01 -2.39184037e-01 2.73332149e-01 -5.58329105e-01 -7.25779712e-01 -6.43405855e-01 -1.42551351e+00 1.35784662e+00 2.74934739e-01 -3.34321469e-01 -1.34561622e+00 3.82679135e-01 5.31907856e-01 2.89795995e-01 -6.21294081e-01 8.47001374e-01 -9.32072699e-01 5.78502938e-02 1.44412979e-01 7.35938847e-02 3.13639134e-01 2.56535918e-01 -1.04295433e-01 -1.05086803e+00 -7.20470369e-01 -8.47938880e-02 -7.12786615e-01 9.48231161e-01 3.14516485e-01 6.64652884e-01 6.72433339e-03 -5.90021670e-01 4.20901775e-01 1.12314498e+00 -3.28761488e-02 1.71152223e-02 3.84186298e-01 7.92069495e-01 7.52391815e-01 9.05542195e-01 2.88417190e-01 8.42581570e-01 7.13383257e-01 1.09742105e-01 -6.41663447e-02 -1.11479335e-01 -5.32010257e-01 1.00901401e+00 1.62209547e+00 6.58480406e-01 -2.31593743e-01 -6.01043999e-01 8.53544414e-01 -1.80743814e+00 -7.74364471e-01 1.65984526e-01 1.79984260e+00 1.03433478e+00 1.42782196e-01 -1.14525579e-01 -7.39814699e-01 8.68216097e-01 6.65393054e-01 -8.43191206e-01 -7.15044916e-01 -1.09857388e-01 -1.34023637e-01 3.33007962e-01 7.25719213e-01 -1.14044142e+00 1.61529279e+00 5.89959383e+00 7.90187836e-01 -1.20729601e+00 6.98589683e-01 4.53402221e-01 -1.61222801e-01 -5.84453344e-01 5.93301803e-02 -1.50033104e+00 2.67139524e-01 8.48734617e-01 -2.52878815e-01 2.65600145e-01 1.07330179e+00 1.59689769e-01 5.02802581e-02 -8.40300143e-01 6.67857349e-01 6.13954365e-01 -1.18557155e+00 -8.91101919e-03 -3.81121337e-01 1.07993889e+00 5.25934279e-01 2.44855285e-01 7.91546226e-01 6.70079887e-01 -5.46071708e-01 5.00763118e-01 1.12185717e-01 8.02579403e-01 -6.55804574e-01 3.48308921e-01 2.37230897e-01 -1.21694577e+00 2.68151194e-01 -2.41551235e-01 4.03205723e-01 2.88919628e-01 8.13180432e-02 -8.62592518e-01 3.28117996e-01 4.49931353e-01 1.17376709e+00 -7.25315154e-01 1.88673526e-01 -4.69608128e-01 4.44533795e-01 3.81552696e-01 -4.54865038e-01 4.64138180e-01 -1.11817032e-01 4.78689134e-01 1.39318883e+00 1.05560519e-01 -7.56021321e-01 2.08943561e-01 6.27464712e-01 -2.32995316e-01 4.48920399e-01 -7.77338326e-01 -5.35870790e-02 5.21751702e-01 1.46029592e+00 -3.08517188e-01 -4.78259176e-01 -1.02713239e+00 1.29191530e+00 6.86243355e-01 2.15671405e-01 -6.75875306e-01 -2.57619083e-01 8.92886698e-01 -5.72043121e-01 2.84815013e-01 -1.94966018e-01 3.58743295e-02 -1.79369116e+00 -3.25666487e-01 -1.01955223e+00 2.75871992e-01 -6.67043984e-01 -1.29758668e+00 8.19056869e-01 -3.29761952e-01 -1.38147962e+00 -2.75826275e-01 -4.45803970e-01 -4.23386872e-01 9.53807712e-01 -1.81481504e+00 -1.44312060e+00 3.93655390e-01 4.90370929e-01 1.45645344e+00 -7.52202094e-01 1.05116975e+00 3.25794294e-02 -6.91296875e-01 7.91037798e-01 3.93961936e-01 1.11149587e-01 1.30537033e+00 -1.27789247e+00 7.23530173e-01 9.23684359e-01 2.70776182e-01 7.94513822e-01 4.61040288e-01 -6.80333614e-01 -1.42338967e+00 -1.28795695e+00 1.12996078e+00 -6.96114421e-01 7.90878594e-01 -8.10133755e-01 -8.24284613e-01 7.37242877e-01 8.35258067e-01 -1.03975661e-01 8.41358423e-01 3.83327127e-01 -7.18221307e-01 1.42941967e-01 -6.73051536e-01 8.19372177e-01 6.89779580e-01 -1.04034591e+00 -8.02833378e-01 4.67473388e-01 1.15890229e+00 -4.18527037e-01 -5.45044482e-01 9.46427435e-02 3.68661672e-01 -7.44423330e-01 7.53505051e-01 -8.90529573e-01 5.62697232e-01 -2.97715217e-02 -1.79724440e-01 -1.86765504e+00 2.16574036e-02 -5.42556643e-01 -7.95131251e-02 1.44675314e+00 7.10635483e-01 -4.19986665e-01 2.40967169e-01 -9.84310955e-02 -2.99443811e-01 -8.31142843e-01 -9.76079464e-01 -4.63326275e-01 2.80893028e-01 -1.34133235e-01 1.20935902e-01 1.22542727e+00 4.52884704e-01 1.17531073e+00 -6.20559931e-01 8.35393835e-03 2.98012912e-01 3.97082865e-01 5.31374753e-01 -7.03039944e-01 -2.00733393e-01 -3.28595400e-01 3.59846324e-01 -1.36759031e+00 7.40022242e-01 -1.26542425e+00 2.29146019e-01 -1.42746580e+00 3.67650181e-01 -2.08660766e-01 -3.20091099e-01 2.82556623e-01 -5.86821675e-01 -3.39668453e-01 6.70176372e-02 1.09196946e-01 -1.05711234e+00 8.58220160e-01 1.32451248e+00 -2.28336766e-01 -2.32795075e-01 -1.70979023e-01 -9.23626423e-01 6.56907618e-01 5.43598533e-01 -3.42553914e-01 -3.69241118e-01 -6.25161469e-01 1.11004673e-01 -2.18638871e-02 -4.38078910e-01 -3.63138616e-01 9.32642519e-02 -7.42728561e-02 -1.19137749e-01 -7.92792380e-01 3.68090868e-01 -3.72832775e-01 -1.01419008e+00 4.69023585e-01 -8.48770320e-01 2.82130390e-01 1.79125428e-01 6.40079021e-01 -2.21368372e-01 -3.45528781e-01 7.46226609e-01 -2.10310996e-01 -1.08608365e+00 4.38436270e-02 -3.86293143e-01 5.70673585e-01 9.28357422e-01 5.91756105e-01 -4.43884552e-01 -3.54990959e-01 -3.97726417e-01 6.30090177e-01 2.31734052e-01 1.25329781e+00 3.54725242e-01 -1.28939772e+00 -6.44329190e-01 1.24426827e-01 7.16972411e-01 -3.73456925e-01 1.46741360e-01 5.63655138e-01 9.05565917e-02 7.62920380e-01 2.93791950e-01 -6.50194049e-01 -1.35277271e+00 5.74803293e-01 1.73247397e-01 -4.91283983e-01 -4.02251333e-01 1.15896940e+00 2.92649955e-01 -1.09000492e+00 5.58055341e-02 -1.66981265e-01 -5.54600596e-01 3.42362851e-01 4.56038654e-01 -8.19022655e-02 -1.30716115e-01 -8.58794689e-01 -5.02443969e-01 5.55402100e-01 -6.32724643e-01 -2.26614013e-01 1.15786242e+00 -4.94763792e-01 -1.12722255e-01 9.73291695e-01 1.34813929e+00 3.72895539e-01 -1.28957140e+00 -5.58040857e-01 2.03503102e-01 -3.95095907e-02 -3.83863412e-02 -6.64422989e-01 -8.85902345e-01 1.19114470e+00 6.29054546e-01 -3.59279543e-01 5.59431851e-01 2.04280764e-01 8.77319634e-01 2.10212663e-01 3.16233605e-01 -1.44222105e+00 4.78794456e-01 9.96753037e-01 8.22996318e-01 -1.53193748e+00 -8.00210461e-02 -8.46624225e-02 -1.33170736e+00 8.74313593e-01 1.07563841e+00 1.70605242e-01 4.52079117e-01 -7.17550591e-02 5.58200479e-01 -1.07325733e-01 -1.14943612e+00 -6.93954900e-02 1.77780345e-01 2.04759046e-01 9.87816036e-01 3.92344333e-02 -2.55116314e-01 4.43225682e-01 1.18711501e-01 -6.06281400e-01 3.37645620e-01 8.65551770e-01 -3.35640103e-01 -1.29465973e+00 1.81204733e-02 1.36417106e-01 -3.80407900e-01 -4.57289785e-01 -2.40991279e-01 5.35031199e-01 -2.84707695e-01 9.35961366e-01 -1.65839240e-01 -3.85895699e-01 3.34412873e-01 5.32792747e-01 1.69386461e-01 -1.09334242e+00 -7.31879294e-01 2.63562769e-01 8.86304602e-02 -4.16952014e-01 -5.00984132e-01 -8.69920313e-01 -1.11418140e+00 2.86578059e-01 -5.50251901e-01 1.58216283e-01 7.53497303e-01 1.13502634e+00 6.19728029e-01 6.68156028e-01 7.65195251e-01 -5.49257636e-01 -7.08537698e-01 -1.11943042e+00 -2.31359184e-01 3.15932035e-01 3.66105139e-01 -5.07020772e-01 -1.87546790e-01 -1.12229101e-01]

[11.273503303527832, 9.758744239807129]

80fd57fc-41a6-4d49-9a0e-3ccecc0b4f91

probabilistic-3d-surface-reconstruction-from

2010.02041

https://arxiv.org/abs/2010.02041v1

https://arxiv.org/pdf/2010.02041v1.pdf

Probabilistic 3D surface reconstruction from sparse MRI information

Surface reconstruction from magnetic resonance (MR) imaging data is indispensable in medical image analysis and clinical research. A reliable and effective reconstruction tool should: be fast in prediction of accurate well localised and high resolution models, evaluate prediction uncertainty, work with as little input data as possible. Current deep learning state of the art (SOTA) 3D reconstruction methods, however, often only produce shapes of limited variability positioned in a canonical position or lack uncertainty evaluation. In this paper, we present a novel probabilistic deep learning approach for concurrent 3D surface reconstruction from sparse 2D MR image data and aleatoric uncertainty prediction. Our method is capable of reconstructing large surface meshes from three quasi-orthogonal MR imaging slices from limited training sets whilst modelling the location of each mesh vertex through a Gaussian distribution. Prior shape information is encoded using a built-in linear principal component analysis (PCA) model. Extensive experiments on cardiac MR data show that our probabilistic approach successfully assesses prediction uncertainty while at the same time qualitatively and quantitatively outperforms SOTA methods in shape prediction. Compared to SOTA, we are capable of properly localising and orientating the prediction via the use of a spatially aware neural network.

['Ender Konukoglu', 'Marc Pollefeys', 'Andrew King', 'Esther Puyol-Antón', 'Matthew Lee', 'Sarah Parisot', 'Katarína Tóthová']

2020-10-05

['probabilistic-deep-learning']

['computer-vision']

[ 2.93552428e-01 4.55508143e-01 2.77066499e-01 -5.51953197e-01 -1.36365807e+00 -3.88763435e-02 5.52826583e-01 2.19327614e-01 -2.36929134e-01 5.37000418e-01 2.04272121e-01 -1.17352381e-01 -5.65607250e-01 -5.32393456e-01 -7.73794115e-01 -7.71357656e-01 -3.52203429e-01 1.63459623e+00 3.40574890e-01 2.54676402e-01 6.32664785e-02 7.72114038e-01 -9.89805937e-01 4.47816223e-01 4.58823502e-01 1.16291142e+00 3.25204402e-01 4.62788045e-01 6.67238012e-02 4.80714500e-01 1.19744152e-01 1.34456873e-01 -5.51879592e-02 7.65052736e-02 -9.89922583e-01 1.83077548e-02 1.77444413e-01 -3.63795459e-02 1.05233956e-02 8.10033143e-01 7.29880214e-01 -1.09650336e-01 1.18168783e+00 -2.54653394e-01 -5.73159233e-02 5.07752955e-01 -4.99163568e-01 3.66792202e-01 1.81836203e-01 -2.18392298e-01 4.40408647e-01 -1.07116759e+00 7.45197833e-01 8.51901293e-01 1.09456336e+00 3.33997101e-01 -1.76838470e+00 -1.70755833e-02 -4.74718720e-01 -1.77453866e-03 -1.48110723e+00 -3.71163100e-01 1.07873642e+00 -7.78398097e-01 9.04384553e-01 2.49791682e-01 3.32379311e-01 9.45608616e-01 8.45301509e-01 6.97039664e-01 1.48082876e+00 -2.29134038e-01 3.14305782e-01 -2.22568020e-01 -4.55755651e-01 6.15845740e-01 -1.63136452e-01 3.40010554e-01 -3.21267694e-01 -5.98568022e-01 1.22386253e+00 -1.90327212e-01 -2.28082597e-01 -9.07611787e-01 -1.53225279e+00 5.35921156e-01 2.74714381e-01 2.98122168e-01 -1.01907265e+00 2.31697887e-01 3.27702940e-01 -2.88872153e-01 6.36082649e-01 2.85708457e-01 -5.63184083e-01 -6.42134920e-02 -1.35536921e+00 1.43982723e-01 5.73375762e-01 4.12902206e-01 2.91725218e-01 1.97797804e-03 -7.27005824e-02 7.27698267e-01 6.73679054e-01 5.06539822e-01 4.05554295e-01 -1.03572333e+00 -1.10418148e-01 7.23723695e-02 5.01657575e-02 -9.28132534e-01 -9.42692041e-01 -5.10517776e-01 -1.09882879e+00 4.06912088e-01 1.04507498e-01 7.72309080e-02 -8.98061693e-01 1.11525810e+00 4.80877787e-01 3.80540192e-01 -3.90747070e-01 9.15858388e-01 6.72677934e-01 2.82341719e-01 1.14396848e-02 -2.87103176e-01 1.32660174e+00 -1.70717254e-01 -5.14150977e-01 -7.75816739e-02 3.01865220e-01 -8.39274704e-01 4.05306786e-01 6.95197046e-01 -1.32148409e+00 -3.41439992e-01 -7.34840155e-01 3.56989443e-01 4.68709081e-01 1.28452957e-01 2.17744276e-01 5.16610920e-01 -1.07678211e+00 1.15094721e+00 -1.59750450e+00 4.71381098e-01 7.92413890e-01 4.69261169e-01 -5.51327229e-01 -2.46398188e-02 -8.30360174e-01 1.28812897e+00 1.62689447e-01 1.10108621e-01 -8.70056212e-01 -9.53166008e-01 -1.03842270e+00 -5.08096457e-01 -6.27403930e-02 -7.68198609e-01 1.03555799e+00 -5.79464436e-01 -1.61300492e+00 9.60444987e-01 -5.54175228e-02 -5.39811671e-01 6.35730684e-01 -1.88024007e-02 -1.35994837e-01 4.11204696e-01 2.24128775e-02 5.83375812e-01 9.01894331e-01 -1.49538410e+00 1.44518346e-01 -5.75253367e-01 -9.34147954e-01 1.15717106e-01 8.37849915e-01 -2.86204293e-02 7.76274055e-02 -5.83117664e-01 9.46878016e-01 -8.11957240e-01 -7.32531786e-01 -8.06826651e-02 -1.57077670e-01 4.22752686e-02 2.43408233e-01 -9.78445649e-01 7.33447492e-01 -1.56987333e+00 3.87596041e-01 5.96028090e-01 4.53962862e-01 -2.23806217e-01 4.28001463e-01 -1.80804655e-01 -2.08893970e-01 -1.00043088e-01 -9.17967677e-01 -4.86588955e-01 -1.22476824e-01 3.90346080e-01 9.28494483e-02 1.11546898e+00 2.04246461e-01 8.89009535e-01 -8.58478844e-01 -7.80340552e-01 5.56018889e-01 9.56722856e-01 -5.05869508e-01 1.38691306e-01 -1.23367019e-01 9.64928627e-01 -3.99725974e-01 4.79232132e-01 6.18084490e-01 -2.57516742e-01 2.64562458e-01 -4.33570772e-01 1.03873640e-01 2.04749107e-01 -1.28533983e+00 2.01165748e+00 -5.43551028e-01 2.63890892e-01 1.90248117e-01 -1.08548617e+00 9.90130782e-01 6.80342197e-01 1.05023098e+00 -5.00549793e-01 1.44642219e-01 4.78686184e-01 -2.02015489e-01 -3.61703604e-01 -1.08014666e-01 -8.23140264e-01 1.88742787e-01 4.81690258e-01 4.69649024e-02 -5.14802337e-01 -8.86724293e-01 -2.04459682e-01 9.56640184e-01 6.24924898e-01 1.63050026e-01 -5.84969103e-01 3.93496692e-01 -1.23810388e-01 3.16368163e-01 5.13078868e-01 -3.61433834e-01 1.24968684e+00 2.89686114e-01 -8.06727827e-01 -1.32578444e+00 -1.30594468e+00 -9.93246615e-01 3.38160157e-01 -2.45649070e-01 1.73753709e-01 -5.65991700e-01 -4.98200297e-01 -3.21128011e-01 7.76108861e-01 -7.27249265e-01 2.72800952e-01 -7.77021229e-01 -9.93025899e-01 2.25408509e-01 3.76319945e-01 -3.29824090e-01 -1.11023200e+00 -6.97869897e-01 5.55884421e-01 -2.01607898e-01 -8.44310164e-01 -5.04348939e-03 3.32699090e-01 -1.38568568e+00 -8.49222839e-01 -7.34017611e-01 -4.95744467e-01 5.53328812e-01 -5.73033988e-01 1.38758671e+00 -3.54586273e-01 -4.28488016e-01 5.48802912e-01 2.77539231e-02 -8.16736817e-02 -6.28249824e-01 -4.96113658e-01 2.59011179e-01 -2.96097070e-01 1.62501663e-01 -9.56388652e-01 -6.88160658e-01 2.20038369e-01 -7.52303183e-01 1.37237953e-02 6.56961262e-01 1.03241563e+00 1.45751822e+00 -3.39214392e-02 3.03789437e-01 -9.30322766e-01 2.52217025e-01 -6.44194365e-01 -2.70993084e-01 -1.89628061e-02 -6.46374643e-01 3.99996668e-01 5.66881374e-02 -1.56791452e-02 -1.04997730e+00 4.18188453e-01 -6.74431503e-01 -5.96106887e-01 -5.61920404e-01 5.56630194e-01 3.43547165e-01 -2.43147686e-01 1.00553620e+00 3.63599092e-01 4.08230215e-01 -6.28551960e-01 1.29454762e-01 2.28526354e-01 4.39969867e-01 -9.75120842e-01 2.72243470e-02 7.98305392e-01 4.95224059e-01 -5.75444221e-01 -3.42093050e-01 -3.93522501e-01 -1.23640466e+00 -3.00913870e-01 6.87651277e-01 -5.61819553e-01 -2.87203610e-01 1.52127132e-01 -1.10139048e+00 -1.34148493e-01 -3.03513378e-01 6.90424025e-01 -1.14321554e+00 7.12181687e-01 -5.40386438e-01 -7.88495183e-01 -4.88043904e-01 -1.55655789e+00 1.52087736e+00 -5.98675847e-01 -3.75702262e-01 -1.18018925e+00 4.27574188e-01 3.10154319e-01 4.85207379e-01 6.32202148e-01 7.95035303e-01 -6.19441986e-01 -3.11769515e-01 -2.32438892e-01 5.38857188e-03 2.17329800e-01 -1.25980064e-01 -4.83404905e-01 -8.89533877e-01 6.43389970e-02 5.33248603e-01 -1.31814361e-01 6.48735464e-01 1.15755939e+00 1.35764587e+00 8.17276835e-02 -3.20428818e-01 5.99419534e-01 1.40018547e+00 -2.54583508e-01 5.12828112e-01 9.23269242e-02 5.79822659e-01 6.53691769e-01 2.06914902e-01 4.63297814e-01 1.89840257e-01 7.23628283e-01 5.23364365e-01 3.15441966e-01 -8.75467956e-02 1.02892652e-01 -1.10916056e-01 9.43688989e-01 -2.18746230e-01 6.27551198e-01 -1.53315258e+00 4.92187053e-01 -1.68287230e+00 -5.93576014e-01 -2.67723948e-01 2.18936372e+00 9.97335494e-01 1.57922357e-01 -2.53346354e-01 -1.26786437e-02 4.22009587e-01 -1.30589992e-01 -2.45913357e-01 -1.22213781e-01 1.52339607e-01 4.76831853e-01 4.46844488e-01 7.37680852e-01 -1.11369181e+00 2.70782411e-01 6.36166430e+00 9.65533972e-01 -9.52335835e-01 4.67759967e-01 9.91974652e-01 2.76014488e-02 -5.33553779e-01 -3.90343755e-01 -1.88794345e-01 2.15737104e-01 1.06226695e+00 4.91790444e-01 1.93924010e-01 6.92869365e-01 2.30940297e-01 -1.63825080e-01 -1.01301658e+00 1.04590631e+00 3.59610133e-02 -1.63869178e+00 -1.09108560e-01 -1.20750345e-01 6.64577365e-01 6.01683199e-01 -4.54392955e-02 -1.40038043e-01 -2.29905937e-02 -1.48556828e+00 7.14685082e-01 1.19969571e+00 7.79561937e-01 -6.90365791e-01 9.46998358e-01 6.19522631e-01 -6.23976707e-01 6.56615615e-01 -2.23766133e-01 3.74613494e-01 5.72906256e-01 8.81022036e-01 -9.49719727e-01 5.97944975e-01 6.59256756e-01 4.05113876e-01 4.45198454e-02 1.03676426e+00 2.80290395e-02 5.19784093e-01 -6.85674250e-01 6.01608694e-01 1.23964794e-01 -5.76764718e-02 8.43423128e-01 1.04790032e+00 2.69547164e-01 2.53671974e-01 1.38861835e-01 9.36758757e-01 4.46145505e-01 8.24778825e-02 -3.74371648e-01 6.17402077e-01 6.79126829e-02 1.05719757e+00 -7.97796190e-01 2.49291155e-02 1.32816508e-01 6.49549782e-01 2.52865642e-01 -1.32015631e-01 -4.86203343e-01 6.28546178e-01 6.93047494e-02 6.38493657e-01 1.61915183e-01 -3.62298608e-01 -7.70366967e-01 -6.99682415e-01 -2.24504266e-02 -5.30654788e-01 1.75098270e-01 -8.23089361e-01 -1.52754736e+00 1.02147532e+00 2.37935677e-01 -8.26888204e-01 -3.30142498e-01 -7.37730145e-01 -2.23173797e-01 1.10809219e+00 -1.36003375e+00 -1.13806927e+00 4.17214543e-01 2.03813612e-01 1.94374382e-01 4.92612049e-02 1.15547884e+00 1.05272613e-01 1.09949455e-01 -1.32369369e-01 2.56766647e-01 -2.11825833e-01 3.58369201e-01 -1.53474867e+00 8.51751193e-02 2.41770402e-01 2.75846571e-01 3.54213923e-01 8.84886742e-01 -9.51709151e-01 -1.18594170e+00 -8.74009848e-01 6.46798372e-01 -8.42142940e-01 5.92762291e-01 2.02703103e-01 -1.15290308e+00 4.86636132e-01 -2.96921819e-01 7.53270030e-01 6.87565506e-01 4.71063577e-05 2.25523099e-01 4.30973113e-01 -1.55673552e+00 -5.08018583e-03 5.07470906e-01 -4.22798991e-01 -8.44493449e-01 4.13272142e-01 2.02685118e-01 -9.27201629e-01 -1.54866183e+00 8.32755446e-01 5.34016669e-01 -8.84240091e-01 1.33679652e+00 -3.59294266e-01 4.68736231e-01 -1.00653067e-01 -2.45657206e-01 -1.19759810e+00 -4.34786290e-01 -3.42481911e-01 -3.16997826e-01 2.42661461e-01 2.60163963e-01 -1.71366036e-01 1.09411490e+00 7.83463836e-01 -4.25875455e-01 -1.33529150e+00 -1.65430570e+00 -3.66079360e-01 4.31214541e-01 -1.04358530e+00 2.21556097e-01 7.02744186e-01 -3.30299020e-01 -2.87114382e-01 -2.72109777e-01 5.34775555e-01 1.23395705e+00 -8.31383392e-02 -1.02999561e-01 -1.39153457e+00 -3.43180239e-01 -5.00728488e-01 -4.91078079e-01 -5.56692243e-01 6.74598143e-02 -1.02506983e+00 3.26124907e-01 -1.65838718e+00 1.00399032e-01 -8.56722116e-01 -3.38278353e-01 -5.46241812e-02 2.02473521e-01 3.20301324e-01 -5.85404515e-01 4.73981172e-01 -4.32724267e-01 5.00642598e-01 1.22780728e+00 2.46687576e-01 2.36360222e-01 1.87474325e-01 4.42103762e-03 1.02714288e+00 5.90554833e-01 -7.87421346e-01 -3.62931676e-02 -2.52452195e-01 3.88181597e-01 5.59843183e-01 6.47030175e-01 -8.55805635e-01 1.90985858e-01 8.22465718e-02 6.95574641e-01 -8.37142706e-01 3.22617650e-01 -1.04566181e+00 5.61931849e-01 2.91890442e-01 -1.01592183e-01 -1.90185547e-01 2.87883013e-01 7.76067853e-01 2.92981043e-02 -3.78735363e-01 1.03816795e+00 -4.52040434e-01 -3.35434675e-01 5.57179034e-01 -3.98809731e-01 -1.56189024e-01 6.30004823e-01 -1.26874745e-01 6.71481371e-01 7.73787573e-02 -1.47962022e+00 -2.41678223e-01 2.51171201e-01 -1.76964059e-01 9.85751033e-01 -1.38531792e+00 -1.01087153e+00 2.07528472e-01 -1.35207921e-01 4.90210116e-01 7.36279130e-01 1.08665252e+00 -8.25195134e-01 5.07074416e-01 -1.75602898e-01 -1.25528085e+00 -6.97789609e-01 3.14374596e-01 7.66972840e-01 -4.67547476e-01 -1.00867987e+00 9.68189061e-01 -1.16717033e-01 -8.45032752e-01 -1.98097870e-01 -2.53184885e-01 -3.24624211e-01 -3.12144846e-01 2.77378678e-01 2.06167370e-01 4.36389714e-01 -1.00128436e+00 -5.10354519e-01 5.56023657e-01 3.08821797e-01 -2.50820935e-01 1.74502969e+00 -1.38918415e-01 -1.53631359e-01 4.85157460e-01 9.09020841e-01 -1.60366669e-01 -1.51365185e+00 -5.15303850e-01 1.51646718e-01 -1.44988507e-01 8.06035459e-01 -9.40233827e-01 -9.12948072e-01 1.02492642e+00 7.47900486e-01 -3.21094304e-01 6.51659191e-01 2.42109358e-01 5.17465353e-01 -8.77227411e-02 5.58864713e-01 -8.08085918e-01 -3.03829134e-01 1.80788979e-01 1.17579687e+00 -1.51159906e+00 3.77947628e-01 -9.71892253e-02 -7.99213529e-01 1.25274467e+00 -1.20642193e-01 -3.89533013e-01 1.28801966e+00 4.81798142e-01 -1.77426770e-01 -7.25274503e-01 -5.45195758e-01 3.73577565e-01 8.04550827e-01 7.54879951e-01 5.84057927e-01 4.50049609e-01 1.19194806e-01 8.09929550e-01 1.61772698e-01 -1.07772328e-01 1.25647023e-01 6.79129362e-01 -4.85419780e-01 -9.86107290e-01 -6.68903947e-01 5.79394996e-01 -8.29197228e-01 -1.70692772e-01 4.42327529e-01 2.96346575e-01 -1.96722336e-02 1.88427139e-02 8.99215639e-02 7.33244196e-02 2.43410300e-02 7.32078403e-02 8.83901000e-01 -6.40477240e-01 -2.80854911e-01 3.79257709e-01 -2.17202902e-02 -6.56743705e-01 -2.55027115e-01 -1.23265362e+00 -1.36305439e+00 2.18986243e-01 -5.92917018e-02 -1.95183381e-02 1.02863944e+00 1.06543028e+00 2.83127427e-01 5.70024788e-01 2.75097609e-01 -1.40367746e+00 -5.90591133e-01 -8.82720709e-01 -6.52805924e-01 2.98001356e-02 1.69737503e-01 -9.33108389e-01 -3.13281864e-02 -8.48201513e-02]

[13.796443939208984, -2.3009612560272217]

497cc863-1c6b-4d61-ae9e-53c27c341168

extreme-acceleration-of-graph-neural-network

2211.13853

https://arxiv.org/abs/2211.13853v1

https://arxiv.org/pdf/2211.13853v1.pdf

Extreme Acceleration of Graph Neural Network-based Prediction Models for Quantum Chemistry

Molecular property calculations are the bedrock of chemical physics. High-fidelity \textit{ab initio} modeling techniques for computing the molecular properties can be prohibitively expensive, and motivate the development of machine-learning models that make the same predictions more efficiently. Training graph neural networks over large molecular databases introduces unique computational challenges such as the need to process millions of small graphs with variable size and support communication patterns that are distinct from learning over large graphs such as social networks. This paper demonstrates a novel hardware-software co-design approach to scale up the training of graph neural networks for molecular property prediction. We introduce an algorithm to coalesce the batches of molecular graphs into fixed size packs to eliminate redundant computation and memory associated with alternative padding techniques and improve throughput via minimizing communication. We demonstrate the effectiveness of our co-design approach by providing an implementation of a well-established molecular property prediction model on the Graphcore Intelligence Processing Units (IPU). We evaluate the training performance on multiple molecular graph databases with varying degrees of graph counts, sizes and sparsity. We demonstrate that such a co-design approach can reduce the training time of such molecular property prediction models from days to less than two hours, opening new possibilities for AI-driven scientific discovery.

['Sutanay Choudhury', 'Sotiris Xantheas', 'Ang Li', 'Tom Murray', 'Mario Michael Krell', 'Jenna Bilbrey', 'Jesun Firoz', 'Hatem Helal']

2022-11-25

['molecular-property-prediction']

['miscellaneous']

[ 4.03299987e-01 1.59856811e-01 -4.50985610e-01 -4.39885259e-01 -3.23932737e-01 -4.67760175e-01 2.47244053e-02 1.06470346e+00 -3.35820228e-01 8.79423440e-01 -3.68727058e-01 -1.05468166e+00 -2.32322961e-01 -1.38237536e+00 -1.15297616e+00 -4.87408876e-01 -7.07256138e-01 7.05872536e-01 2.31172591e-01 -5.45580983e-02 1.88736603e-01 9.48825777e-01 -1.39038122e+00 3.11062664e-01 8.50238204e-01 8.36043894e-01 4.05384004e-02 7.69775808e-01 8.86926055e-02 9.08242285e-01 -3.16913098e-01 -3.61728191e-01 2.28175789e-01 -1.30758375e-01 -6.51634872e-01 -3.72480452e-01 7.09003866e-01 9.84172821e-02 -5.23426175e-01 9.91876960e-01 5.17469287e-01 3.40076685e-01 3.70998949e-01 -1.03344607e+00 -3.78131032e-01 7.71655262e-01 -2.90368199e-01 2.67921329e-01 2.43252054e-01 2.11356014e-01 1.00763297e+00 -3.46035391e-01 7.18235075e-01 8.50131750e-01 9.10318911e-01 2.19254151e-01 -1.61078477e+00 -8.19027722e-01 -1.32235467e-01 3.21263820e-01 -1.50568914e+00 -3.92255872e-01 4.72376943e-01 -2.41116107e-01 1.81473255e+00 2.30930105e-01 8.65574300e-01 4.45154816e-01 6.60443306e-01 -3.36993784e-02 6.48385942e-01 -2.21661404e-01 5.35847068e-01 -3.26525152e-01 4.45687145e-01 1.17914844e+00 9.57862377e-01 3.22106220e-02 -7.46523440e-01 -6.06454968e-01 3.78356010e-01 7.80062601e-02 -5.61199337e-02 -2.48226151e-01 -8.93893957e-01 9.48877633e-01 7.31915355e-01 -7.64165968e-02 -4.21568781e-01 5.42398393e-01 4.75665212e-01 3.10328692e-01 5.57749987e-01 9.07383323e-01 -5.94887197e-01 1.73229322e-01 -9.32683349e-01 -1.60870608e-02 1.25556219e+00 9.49439287e-01 1.05460763e+00 1.97506994e-01 3.96226853e-01 1.88280389e-01 8.32705870e-02 1.42082006e-01 7.00153038e-02 -4.23384517e-01 2.33778566e-01 7.21831858e-01 -5.18153131e-01 -1.10377681e+00 -7.93456912e-01 -3.75662506e-01 -8.68102431e-01 -2.60765627e-02 8.23898241e-02 -3.13872755e-01 -8.89072239e-01 1.32998836e+00 6.13559544e-01 1.23115383e-01 -9.28016976e-02 3.88150722e-01 9.06038940e-01 9.37748015e-01 4.31266427e-01 -2.50009775e-01 1.03867292e+00 -8.54138851e-01 -1.34476900e-01 -8.73344466e-02 1.25715649e+00 -4.68177080e-01 5.66626787e-01 3.94622922e-01 -9.57957804e-01 -2.65439034e-01 -1.40815270e+00 -1.04640514e-01 -6.33534133e-01 -4.47782785e-01 1.53595042e+00 8.90694559e-01 -1.17202914e+00 1.53759670e+00 -1.00748003e+00 -1.87175833e-02 5.07423282e-01 1.09325933e+00 -5.10899842e-01 -1.71234772e-01 -7.78991997e-01 7.46526539e-01 8.37568402e-01 -2.18831941e-01 -5.35583258e-01 -9.95221972e-01 -7.48287976e-01 2.92030066e-01 3.97788048e-01 -7.25449681e-01 6.17698073e-01 -8.56786370e-01 -1.29179549e+00 2.90172994e-01 -2.63941456e-02 -6.13004863e-01 -3.68349969e-01 4.74396259e-01 -4.16262984e-01 1.18386410e-01 -4.66039389e-01 3.30608964e-01 4.57048208e-01 -4.58650261e-01 -2.90134232e-02 -4.09892946e-01 -1.90844223e-01 1.24599189e-02 -5.20007670e-01 -1.57481208e-01 -8.33669081e-02 -1.68650582e-01 -1.24020547e-01 -9.66274023e-01 -7.25381851e-01 -2.96319544e-01 -3.69803846e-01 -1.12399399e-01 2.43150219e-01 -2.51852363e-01 1.08976912e+00 -1.60082579e+00 5.78240938e-02 8.24066997e-01 7.72397399e-01 3.96227986e-02 -2.40403980e-01 6.42050087e-01 -3.12055796e-01 2.15424120e-01 5.10392301e-02 1.45059094e-01 -4.01379853e-01 1.71832949e-01 1.17060311e-01 5.72078824e-01 -3.15100476e-02 8.85694385e-01 -7.24435270e-01 -3.44855905e-01 -1.61804736e-01 4.40839231e-01 -9.23545301e-01 1.14086471e-01 -6.34787023e-01 -1.04241326e-01 -5.29172778e-01 7.11281836e-01 5.12983084e-01 -1.04723120e+00 8.64790499e-01 -2.53698051e-01 -1.10846482e-01 6.74735069e-01 -7.56286263e-01 1.46177495e+00 4.60739024e-02 2.19910339e-01 -7.56214783e-02 -8.96373689e-01 7.73344517e-01 3.69202122e-02 6.09693766e-01 -4.57530767e-01 1.51001304e-01 2.58444875e-01 4.63684440e-01 9.44362879e-02 4.82223094e-01 -1.52205527e-01 2.41685897e-01 5.50765455e-01 2.14231893e-01 -2.02842146e-01 5.81668556e-01 2.47964159e-01 1.52590811e+00 -2.13460758e-01 2.91714489e-01 -4.79171455e-01 1.06614999e-01 5.22089779e-01 1.90164968e-01 6.56137943e-01 1.70889124e-01 -1.94873244e-01 5.19275308e-01 -9.79401588e-01 -1.21878529e+00 -5.17228186e-01 1.60612822e-01 1.35514700e+00 -1.97016105e-01 -9.94261682e-01 -6.22817695e-01 -4.75922346e-01 2.34689847e-01 2.14049488e-01 -3.15718174e-01 -3.03633779e-01 -3.85053962e-01 -1.19024873e+00 3.09887469e-01 2.39990428e-01 -1.59020081e-01 -8.67330790e-01 -5.04162535e-02 4.30531830e-01 8.48908246e-01 -9.27167594e-01 -2.17520639e-01 8.76618385e-01 -1.12621844e+00 -1.05960226e+00 1.67942673e-01 -7.43281841e-01 9.15939987e-01 2.94091314e-01 1.27578509e+00 5.80455542e-01 -5.29767275e-01 -2.54203200e-01 -2.84065809e-02 -4.26231474e-01 -5.96215427e-01 4.95067626e-01 2.38225952e-01 -5.48466802e-01 4.04010057e-01 -1.11967587e+00 -5.97790420e-01 -1.25679940e-01 -6.92775786e-01 2.64345288e-01 6.32958174e-01 6.40456140e-01 9.37359869e-01 2.10690439e-01 2.99679995e-01 -1.24162245e+00 6.39566839e-01 -6.07117236e-01 -1.04026520e+00 2.20059857e-01 -1.10028517e+00 3.97652149e-01 1.01639223e+00 -4.52658027e-01 -1.68454289e-01 4.17975128e-01 -2.77779792e-02 -3.20816576e-01 2.58036017e-01 9.60365593e-01 1.20254554e-01 -9.07286644e-01 8.17627251e-01 1.58624798e-02 1.28784359e-01 -1.84937835e-01 3.95733774e-01 2.22052783e-01 -3.52796502e-02 -6.76061511e-01 5.12517631e-01 -6.18994310e-02 7.55884767e-01 -9.11855280e-01 -5.24001539e-01 -2.38304675e-01 -3.63359302e-01 3.36217768e-02 6.18881047e-01 -9.15562570e-01 -1.41587114e+00 -8.78782570e-02 -1.08523738e+00 -4.01795655e-01 -6.01360798e-02 4.29935873e-01 -1.57793254e-01 4.22472745e-01 -7.94586182e-01 -1.63848534e-01 -9.47687030e-01 -1.14575052e+00 6.24285221e-01 1.23600900e-01 -3.00100088e-01 -9.14745331e-01 1.40713915e-01 2.57171035e-01 4.50270444e-01 2.14930683e-01 1.37327254e+00 -9.59359288e-01 -9.99430895e-01 -2.62003601e-01 -3.52878630e-01 -1.10321060e-01 -7.20629767e-02 -1.42682672e-01 -5.45238495e-01 -5.55321455e-01 -6.33702457e-01 -3.71956944e-01 7.27701902e-01 2.13634640e-01 1.34001446e+00 -4.06949729e-01 -5.15786231e-01 1.01171517e+00 1.53937650e+00 1.08753704e-01 3.62407178e-01 -5.62503673e-02 1.19754887e+00 2.29584679e-01 -8.05666670e-02 3.05733293e-01 5.46859130e-02 2.68940657e-01 3.48170251e-01 -1.39393777e-01 1.95816442e-01 -3.17487568e-01 2.26355612e-01 1.12699223e+00 -4.11515474e-01 -3.33529294e-01 -1.06579494e+00 -1.04738317e-01 -1.50093055e+00 -8.81159544e-01 -2.01566234e-01 2.18962002e+00 1.08572960e+00 1.83340818e-01 1.22115240e-02 -1.82338610e-01 3.81572455e-01 3.01740244e-02 -7.14925885e-01 -6.73981071e-01 3.29536706e-01 9.17955101e-01 1.28089345e+00 6.04735851e-01 -8.14357102e-01 1.12980545e+00 6.89122391e+00 9.34522212e-01 -1.26875341e+00 -5.68212755e-02 7.48506367e-01 -2.11001992e-01 -2.68157482e-01 2.04153985e-01 -9.00765538e-01 1.57940358e-01 1.87076306e+00 -4.02801365e-01 8.19945574e-01 1.09538889e+00 -2.26279628e-02 3.16292867e-02 -1.32114136e+00 8.65779579e-01 -1.17307648e-01 -2.13978148e+00 3.26100379e-01 3.71520579e-01 5.59940696e-01 5.98865211e-01 -3.96095842e-01 1.54934600e-02 6.61906064e-01 -1.40347707e+00 1.43322572e-01 1.16724290e-01 8.72530997e-01 -9.67453539e-01 2.92829961e-01 6.70477226e-02 -1.21226251e+00 2.86936343e-01 -6.70396864e-01 -3.40225309e-01 -2.69617945e-01 7.26068020e-01 -1.22219837e+00 4.47756082e-01 2.85408258e-01 6.51671469e-01 -5.90018332e-01 8.35396826e-01 2.74908960e-01 5.77142954e-01 -5.60202360e-01 -6.78163052e-01 -1.62559201e-03 -4.58924055e-01 2.89773643e-02 9.98257756e-01 1.84765235e-01 9.62165967e-02 4.24271166e-01 6.52941346e-01 -5.31024575e-01 3.78110647e-01 -7.82086372e-01 -7.47332275e-01 4.36717659e-01 1.15478492e+00 -9.48189080e-01 -4.70287681e-01 -3.43131751e-01 5.53105354e-01 8.26855361e-01 -8.70701578e-03 -6.14426196e-01 -3.92169774e-01 4.93156314e-01 3.54635388e-01 1.70182154e-01 -6.50677800e-01 -1.94666058e-01 -9.43915546e-01 -3.57417315e-01 -9.73115623e-01 3.86627734e-01 -4.48596239e-01 -1.17512918e+00 6.00000381e-01 -4.16106284e-01 -6.24053717e-01 4.95929420e-02 -9.47448373e-01 -3.23493242e-01 7.68137157e-01 -1.42581534e+00 -8.26922417e-01 -1.12351887e-01 4.79675412e-01 -2.01949716e-01 -1.16246149e-01 1.09557545e+00 3.60432535e-01 -7.08417416e-01 4.20345157e-01 1.23718113e-01 -3.44032168e-01 2.84303159e-01 -9.86411870e-01 6.43818617e-01 5.05887210e-01 2.16728806e-01 9.25972641e-01 5.86501658e-01 -9.29521084e-01 -2.45297480e+00 -1.17074919e+00 6.99579597e-01 -1.95636988e-01 1.03213561e+00 -4.88038808e-01 -8.83657336e-01 5.03733635e-01 -1.58282980e-01 2.56929874e-01 1.12959063e+00 5.23427784e-01 -4.77243841e-01 -1.72382325e-01 -8.16617668e-01 4.29691583e-01 1.11227083e+00 -5.40468514e-01 2.30861828e-01 1.02798605e+00 9.12122965e-01 -3.07840794e-01 -1.37238765e+00 2.01171592e-01 3.53148490e-01 -4.98327136e-01 9.82473135e-01 -1.02735388e+00 3.39643031e-01 -3.66282836e-02 1.38899134e-02 -1.02182150e+00 -3.54876161e-01 -1.01705074e+00 -2.64040887e-01 1.69966415e-01 9.22547579e-01 -7.13232994e-01 1.30118978e+00 8.19589734e-01 -1.30406186e-01 -8.12455654e-01 -7.68336415e-01 -6.31954491e-01 -1.69845119e-01 -2.05268115e-01 6.45522058e-01 9.85918820e-01 3.24206412e-01 7.86341786e-01 -1.20666303e-01 2.06479356e-01 5.27121007e-01 2.81308293e-01 7.40737379e-01 -1.38735414e+00 -8.41006398e-01 -2.82602489e-01 -6.75462067e-01 -6.51122987e-01 1.22423209e-01 -1.48333120e+00 -4.20769900e-01 -1.18805015e+00 3.27093542e-01 -5.22428691e-01 -2.03862518e-01 7.70929694e-01 1.05906688e-01 3.80942300e-02 -7.57211447e-02 8.07835534e-02 -5.81495404e-01 1.92147523e-01 1.07975864e+00 -1.72235429e-01 -2.59776860e-01 -4.57820296e-01 -5.40440142e-01 3.43749642e-01 7.23751128e-01 -6.79311752e-01 -4.40567404e-01 -1.22594759e-01 7.52844274e-01 1.27650529e-01 -4.27462682e-02 -1.13891923e+00 5.63573360e-01 -2.74270475e-01 4.28615183e-01 -2.51510262e-01 2.40178645e-01 -7.04178810e-01 3.70208710e-01 6.59252882e-01 -2.08240002e-01 3.19066375e-01 4.90174025e-01 7.12176085e-01 2.06167147e-01 3.91685218e-02 6.17320657e-01 -1.64004296e-01 -2.73978591e-01 8.49748850e-01 -2.62074023e-01 -4.26109552e-01 1.00848532e+00 -1.34124964e-01 -4.83621627e-01 4.74407151e-02 -6.41102195e-01 5.31397946e-03 5.35482943e-01 -2.41276741e-01 5.61467230e-01 -7.26270914e-01 -2.44154483e-01 1.53947234e-01 -1.21502392e-01 -1.99168637e-01 4.07021642e-02 5.59059799e-01 -1.19139159e+00 3.48056018e-01 -2.32073233e-01 -1.70532733e-01 -1.50022399e+00 8.65333140e-01 3.78713310e-01 -5.47555864e-01 -4.35405821e-01 8.22924793e-01 -3.60881716e-01 -2.46981740e-01 -1.74959451e-01 -1.51096001e-01 2.49519080e-01 -2.60907710e-01 2.98750520e-01 1.29540682e-01 5.39426208e-01 -1.36527240e-01 -2.80699223e-01 1.20262131e-02 -4.31500494e-01 6.42969966e-01 1.73766696e+00 7.63466716e-01 -6.16214037e-01 -1.44183993e-01 1.25999856e+00 -1.39662758e-01 -6.22057199e-01 1.39575422e-01 5.79913184e-02 2.20900476e-02 3.73761117e-01 -4.31882590e-01 -9.92024302e-01 6.09260619e-01 2.91196972e-01 2.35697418e-01 7.84563959e-01 -1.88831255e-01 8.11545610e-01 1.17560685e+00 5.61304688e-01 -1.03061318e+00 -3.04462284e-01 4.90617424e-01 1.46977454e-01 -9.56686378e-01 8.59853089e-01 -8.01967323e-01 9.85214561e-02 1.28408515e+00 4.82870132e-01 -1.54847175e-01 7.75565326e-01 3.50874126e-01 -7.62843370e-01 -8.77453625e-01 -9.99013484e-01 1.54514238e-01 2.42245749e-01 3.11702251e-01 6.94871128e-01 4.67043668e-01 -1.83030903e-01 1.88788995e-01 -2.71644264e-01 7.59968460e-02 4.74108219e-01 1.18392551e+00 -5.86333752e-01 -1.39058053e+00 1.80791453e-01 9.77384627e-01 -3.12534660e-01 -6.65474474e-01 -4.82796788e-01 5.63224256e-01 -2.95645177e-01 7.09234476e-01 -4.71225977e-02 -6.54510617e-01 -3.82064283e-02 2.44330987e-02 6.60129488e-01 -9.45733666e-01 -7.82102048e-01 -3.29681814e-01 6.50327206e-01 -7.18943655e-01 -1.52661026e-01 5.05219959e-02 -1.34002972e+00 -9.57673967e-01 -3.82443458e-01 3.64027828e-01 8.17472100e-01 6.17156386e-01 9.98581827e-01 5.02066731e-01 3.44284534e-01 -1.01593828e+00 -4.23006475e-01 -5.81792712e-01 -5.42890191e-01 9.44068842e-03 -9.00878012e-02 -1.95127800e-01 -9.82879251e-02 -1.83918521e-01]

[5.494052886962891, 5.7152581214904785]

04a09f61-582f-47b2-9be0-844f0aaa4a13

multi-lingual-wikipedia-summarization-and

https://aclanthology.org/W19-8904

https://aclanthology.org/W19-8904.pdf

Multi-lingual Wikipedia Summarization and Title Generation On Low Resource Corpus

MultiLing 2019 Headline Generation Task on Wikipedia Corpus raised a critical and practical problem: multilingual task on low resource corpus. In this paper we proposed QDAS extractive summarization model enhanced by sentence2vec and try to apply transfer learning based on large multilingual pre-trained language model for Wikipedia Headline Generation task. We treat it as sequence labeling task and develop two schemes to handle with it. Experimental results have shown that large pre-trained model can effectively utilize learned knowledge to extract certain phrase using low resource supervised data.

['Lei LI', 'Zuying Huang', 'Yinan Liu', 'Wei Liu']

2019-09-01

ranlp-2019-9

['headline-generation']

['natural-language-processing']

[-2.51138825e-02 5.54362416e-01 -2.56823540e-01 -4.54703160e-03 -1.42368352e+00 -6.22705460e-01 7.97371030e-01 6.16989173e-02 -8.06761384e-01 1.68422246e+00 1.13511872e+00 -1.79677442e-01 4.51809466e-01 -7.10212946e-01 -8.20163012e-01 -3.29351351e-02 1.43002898e-01 6.17941201e-01 9.24332999e-03 -1.04636896e+00 4.46748972e-01 -8.39454755e-02 -8.00929010e-01 4.66412872e-01 1.29629147e+00 2.96927858e-02 6.41585112e-01 1.04221070e+00 -3.14694822e-01 1.13013828e+00 -1.02069557e+00 -4.99461263e-01 1.22665837e-02 -6.61190331e-01 -1.26435161e+00 2.74426863e-02 4.13840771e-01 -3.15011889e-01 -4.52326566e-01 1.06044030e+00 9.93165731e-01 8.47947374e-02 1.19880176e+00 -7.23890960e-01 -1.25988913e+00 1.23383713e+00 -5.14899373e-01 5.29877841e-01 5.51298797e-01 -1.55392736e-01 1.11052907e+00 -8.34243715e-01 1.16818297e+00 1.19261551e+00 3.71813178e-01 7.78603673e-01 -7.33866751e-01 -2.67877907e-01 -2.10899994e-01 1.88617215e-01 -1.09014940e+00 -3.91146451e-01 6.18029296e-01 -1.75174579e-01 1.72355402e+00 1.65482294e-02 2.22644776e-01 9.50797975e-01 4.18217778e-01 1.08570743e+00 7.46560872e-01 -6.05908334e-01 -3.33858818e-01 3.22004020e-01 3.19964826e-01 6.05191290e-01 3.11994851e-01 -5.94921708e-01 -3.54643315e-01 1.05326995e-01 3.64592254e-01 -8.92006218e-01 -3.31090450e-01 3.78546983e-01 -1.42981768e+00 1.07161391e+00 -3.55149098e-02 3.58848512e-01 -3.42265159e-01 -2.10787237e-01 1.03946698e+00 6.38304770e-01 7.73710847e-01 8.38434279e-01 -7.58044302e-01 -1.28874242e-01 -8.26848924e-01 2.68482178e-01 9.55513477e-01 1.43465078e+00 7.38759458e-01 3.39834869e-01 -6.12703383e-01 1.30331445e+00 -1.54896319e-01 8.08044195e-01 1.15332997e+00 -3.23545545e-01 1.08445358e+00 2.60974556e-01 -1.05145220e-02 -6.89546645e-01 -3.49676281e-01 -2.23876491e-01 -8.60242307e-01 -5.24254084e-01 -3.11339915e-01 -9.65257823e-01 -6.12785280e-01 1.39369977e+00 -1.32978469e-01 -2.93757021e-01 8.78046572e-01 4.51613456e-01 1.23381519e+00 1.36139643e+00 -1.02098621e-01 -5.38848996e-01 1.08069003e+00 -1.57459950e+00 -9.87500608e-01 8.06057081e-02 1.03071213e+00 -9.71158385e-01 8.64535570e-01 2.18859985e-01 -1.29813516e+00 -6.82287991e-01 -1.31000566e+00 -5.16636968e-01 -5.38178682e-01 4.51010883e-01 3.39082360e-01 1.28863811e-01 -1.21837485e+00 4.48380381e-01 -2.42140427e-01 -6.97632372e-01 -6.40493482e-02 4.68800105e-02 -6.49426520e-01 1.22860726e-02 -1.51218605e+00 1.29264832e+00 1.17665529e+00 -4.80926223e-02 -7.11946070e-01 -3.90635371e-01 -1.21159744e+00 -2.34374538e-01 1.65050641e-01 -9.29961801e-01 1.42200840e+00 -5.59374034e-01 -1.62775469e+00 7.05628395e-01 3.05956062e-02 -8.07181239e-01 2.35162303e-01 -3.82923871e-01 -6.60061896e-01 1.94728389e-01 5.83405435e-01 6.45709693e-01 4.71018195e-01 -8.53168786e-01 -8.06679428e-01 4.07746695e-02 -2.04198778e-01 7.33924508e-01 -5.15781164e-01 5.80948479e-02 -1.50031552e-01 -7.01141477e-01 -8.22510064e-01 -6.52870417e-01 -3.06515783e-01 -1.23027110e+00 -6.86605334e-01 -5.70206583e-01 4.55105573e-01 -1.24138308e+00 1.35280907e+00 -1.25705659e+00 1.56445518e-01 -5.60290635e-01 -2.56223023e-01 7.04621553e-01 -6.42564595e-01 1.13238466e+00 3.34443390e-01 2.17614755e-01 -1.56692699e-01 5.19311950e-02 -3.92689370e-02 -9.40668657e-02 -3.98744822e-01 -5.67703061e-02 4.81163383e-01 1.02421188e+00 -1.23623240e+00 -7.80535758e-01 -1.57836378e-01 1.06021464e-01 -2.85134017e-01 4.92609650e-01 -3.18411648e-01 1.40855119e-01 -4.62338507e-01 6.22583218e-02 6.13068879e-01 3.61602664e-01 -2.94493251e-02 -2.31293067e-01 -3.75591397e-01 3.85155201e-01 -6.02403224e-01 2.17716551e+00 -8.16383421e-01 6.33488774e-01 -5.36014855e-01 -8.54665339e-01 1.01504421e+00 5.29391587e-01 -5.17328233e-02 -6.32796288e-01 1.61185548e-01 1.96727574e-01 -8.69994089e-02 -1.03406107e+00 1.17282701e+00 9.75795835e-03 -5.79277158e-01 5.04406929e-01 9.04964924e-01 -4.48644876e-01 7.84585655e-01 7.08332956e-01 9.43593740e-01 1.47525176e-01 6.86141014e-01 -3.72882098e-01 8.54644299e-01 2.92026192e-01 3.00495714e-01 5.94502389e-01 1.13619320e-01 4.81749892e-01 2.56916463e-01 1.27097098e-02 -1.20960176e+00 -7.98216403e-01 1.66305333e-01 1.09891176e+00 -1.97852120e-01 -5.08335471e-01 -7.68522024e-01 -9.87226725e-01 -4.35617208e-01 9.12700534e-01 -3.47264409e-01 -1.08432874e-01 -8.13022852e-01 -8.56565833e-01 7.74879813e-01 4.06355202e-01 5.98884523e-01 -1.35211933e+00 1.08943909e-01 6.58320844e-01 -3.06255072e-01 -1.13556433e+00 -8.82003248e-01 -1.63804591e-01 -4.69199181e-01 -8.20769429e-01 -1.09214723e+00 -1.52918935e+00 3.79319578e-01 7.60895982e-02 1.17912924e+00 -4.88588452e-01 -2.08956346e-01 1.68402195e-01 -6.19537890e-01 -5.94600081e-01 -7.09458351e-01 6.89954996e-01 4.15640958e-02 -5.26002228e-01 2.45516405e-01 -2.03655556e-01 -1.11123785e-01 -7.04204798e-01 -8.19558501e-01 1.96980640e-01 9.10731554e-01 9.79062557e-01 1.29858553e-01 -3.88728708e-01 1.58685362e+00 -9.57058609e-01 1.62214983e+00 -5.84895253e-01 -4.66222242e-02 7.50455618e-01 -4.37858373e-01 6.21643364e-01 8.70646775e-01 -3.58173437e-02 -1.45841980e+00 -4.01751697e-01 -3.85621786e-01 4.36770767e-01 1.98678404e-01 1.05178595e+00 -1.85096174e-01 4.96778846e-01 5.43858588e-01 6.48778677e-01 -1.99901372e-01 -4.72804010e-01 9.77987289e-01 1.11377800e+00 5.99014401e-01 -5.28439164e-01 6.00203633e-01 -2.69487262e-01 -5.99793613e-01 -1.14954221e+00 -8.30810845e-01 -4.59299147e-01 -9.17165756e-01 7.83558935e-03 1.06429517e+00 -1.15630877e+00 -5.68791218e-02 2.05372095e-01 -1.56915534e+00 -1.20731547e-01 -1.56669274e-01 4.48721439e-01 -4.81564224e-01 5.92173636e-01 -9.39226091e-01 -3.59305680e-01 -1.24794269e+00 -5.60373247e-01 8.32581401e-01 1.84894428e-01 -4.66674417e-02 -1.39000428e+00 9.18896735e-01 2.05597192e-01 2.63549656e-01 1.23735726e-01 6.91273570e-01 -9.66248274e-01 8.22042078e-02 -2.68343508e-01 -1.13673471e-01 8.13369036e-01 3.01750749e-01 -7.60038123e-02 -4.47137058e-01 -2.48222724e-01 -2.93248951e-01 -8.38922501e-01 9.88314152e-01 1.02195167e-03 4.01094288e-01 -7.68448949e-01 4.23588715e-02 -1.27950367e-02 1.50413418e+00 -2.38657892e-01 4.84825879e-01 2.21731901e-01 8.67146373e-01 5.00202000e-01 7.25184739e-01 1.85766593e-01 7.13682353e-01 9.85082388e-02 -5.15021324e-01 1.95380040e-02 -4.43162471e-01 -6.17208242e-01 6.57925010e-01 2.22073984e+00 1.62008077e-01 -5.40618837e-01 -6.43117070e-01 7.26110637e-01 -1.96026921e+00 -1.09353125e+00 -1.03238307e-01 1.64724874e+00 1.51850438e+00 2.15460621e-02 -4.34239469e-02 -5.59346795e-01 7.27687180e-01 1.70060709e-01 -8.63991231e-02 -9.04099762e-01 -4.39854354e-01 2.00346828e-01 6.91752732e-01 7.71242082e-01 -9.63582039e-01 1.42614365e+00 5.82006550e+00 1.22346425e+00 -9.10227060e-01 2.41380960e-01 5.02269454e-02 3.37545872e-01 -3.98032457e-01 -3.55503023e-01 -1.09803939e+00 3.05325747e-01 1.13160908e+00 -1.23254025e+00 -1.89528182e-01 5.63367903e-01 1.43291190e-01 1.09072991e-01 -7.93479919e-01 7.52446055e-01 7.80135572e-01 -1.38537931e+00 7.14945078e-01 -4.00050521e-01 1.20841682e+00 2.14117020e-01 -3.23367119e-01 9.73988652e-01 5.31271815e-01 -8.54589641e-01 1.73073202e-01 3.31881046e-01 7.36286461e-01 -9.81607080e-01 1.00326884e+00 5.01450062e-01 -1.05289996e+00 4.95044291e-01 -9.20652926e-01 -1.18526123e-01 5.13575137e-01 9.63489264e-02 -1.30878139e+00 1.01914430e+00 -2.60481954e-01 1.01106286e+00 -6.13652945e-01 8.11391294e-01 -4.31873918e-01 5.50135434e-01 3.02680492e-01 -3.51051450e-01 6.37427092e-01 -2.54514068e-01 5.18176138e-01 1.83771729e+00 3.01287532e-01 -2.55246073e-01 3.07182372e-01 2.04016134e-01 -4.98664796e-01 8.04068863e-01 -1.02022564e+00 -1.67394549e-01 1.61859319e-01 1.43313861e+00 -1.67634383e-01 -7.48958766e-01 -4.17179942e-01 1.36155355e+00 7.20761061e-01 2.13471234e-01 -5.94936967e-01 -9.58330989e-01 -1.20805033e-01 -5.03641486e-01 1.89205199e-01 -2.48900965e-01 2.84162670e-01 -1.54086065e+00 -2.24585086e-01 -6.94882572e-01 2.67944068e-01 -6.33275151e-01 -1.42376935e+00 9.84936655e-01 -1.07479163e-01 -1.27026868e+00 -5.61680675e-01 -2.82339245e-01 -7.89357722e-01 7.87551582e-01 -1.76479006e+00 -1.64028180e+00 3.61649752e-01 5.79759777e-01 1.36970794e+00 -6.22189581e-01 9.19525146e-01 1.89491138e-01 -6.68406188e-01 5.53019822e-01 5.60523629e-01 3.72962266e-01 1.23668778e+00 -1.63410878e+00 3.83590549e-01 8.09329152e-01 2.96392348e-02 4.37345803e-01 7.26485372e-01 -8.58585715e-01 -1.06086838e+00 -1.48311341e+00 1.46335161e+00 -1.79607704e-01 9.82959032e-01 -1.42443925e-01 -5.46788275e-01 8.10568690e-01 1.50804484e+00 -7.56392062e-01 8.39626074e-01 -1.07793212e-01 9.38023999e-02 1.26535505e-01 -8.40967178e-01 6.58081174e-01 6.67913675e-01 -4.39763159e-01 -1.32597899e+00 8.06019485e-01 1.14106786e+00 -1.65081605e-01 -1.02710545e+00 1.97514504e-01 -6.63219690e-02 5.49716875e-02 4.55220640e-01 -9.27254379e-01 8.41183782e-01 -3.86158787e-02 2.02211469e-01 -2.06585145e+00 -1.08108312e-01 -7.11282730e-01 2.98407534e-03 1.72831655e+00 9.11042929e-01 -3.23212087e-01 1.48533270e-01 8.11697170e-03 -5.34483373e-01 -3.98104042e-01 -4.40806538e-01 -8.64493012e-01 7.07705319e-01 3.42401922e-01 2.78315067e-01 8.67938399e-01 5.83381176e-01 1.50019181e+00 -6.95132971e-01 -3.15198034e-01 3.54270935e-01 -3.23764920e-01 8.40351820e-01 -6.39874399e-01 1.13525458e-01 -1.21462323e-01 -1.02030113e-01 -1.11952400e+00 7.17528462e-01 -1.33944595e+00 1.15773261e-01 -1.99736011e+00 4.48888928e-01 2.94075549e-01 -8.67260173e-02 -1.32049723e-02 -5.43419957e-01 6.13467842e-02 6.51563853e-02 -1.98307410e-01 -1.03386736e+00 9.90554035e-01 1.35557544e+00 -5.36058128e-01 1.01072714e-02 -4.47855473e-01 -7.18830228e-01 2.69808263e-01 9.38036561e-01 -2.61392325e-01 -6.32319450e-01 -7.18062937e-01 1.25897005e-01 4.55117106e-01 -5.07166088e-01 -8.69371355e-01 3.09506834e-01 1.08571187e-01 -2.11862344e-02 -6.75364792e-01 -9.21799913e-02 3.39267962e-02 -7.46306658e-01 2.75697827e-01 -8.86493087e-01 3.25168014e-01 2.30597228e-01 3.90508533e-01 -5.33333063e-01 -7.23183632e-01 2.70372212e-01 -4.99630719e-01 -9.64110851e-01 2.08981574e-01 -7.29927003e-01 6.52416587e-01 8.89122069e-01 4.39640433e-01 -5.92233241e-01 -5.22128999e-01 -4.96156126e-01 6.08268082e-01 -4.40038927e-02 7.19631493e-01 5.91767967e-01 -1.35201752e+00 -1.57830167e+00 -2.87197679e-01 2.32466817e-01 -3.41305852e-01 1.74508333e-01 2.47879654e-01 -8.31129253e-01 9.09987867e-01 -5.11481702e-01 1.50327459e-01 -1.04978120e+00 6.36796594e-01 -1.70166820e-01 -7.10703135e-01 -3.07348609e-01 8.00787568e-01 -2.36906335e-01 -7.23817110e-01 -1.45561546e-01 -3.83631252e-02 -8.06846380e-01 2.90064633e-01 7.52151608e-01 4.40819472e-01 1.42321646e-01 -8.11936975e-01 1.81313381e-01 3.15426826e-01 -6.38374031e-01 -3.28440249e-01 1.26380193e+00 -4.23246473e-01 -3.31621051e-01 5.32273829e-01 1.80534995e+00 9.87256393e-02 -3.94676536e-01 -2.09550187e-01 2.11441755e-01 2.99559772e-01 -2.44226024e-01 -8.01917136e-01 -5.16745210e-01 7.76100516e-01 5.35521097e-02 -7.25976005e-02 4.64487016e-01 -2.94422269e-01 1.36719859e+00 8.50206137e-01 2.29465276e-01 -1.70257711e+00 7.50945061e-02 1.01487803e+00 1.27227199e+00 -1.42500293e+00 -6.50693700e-02 1.50099382e-01 -1.11285329e+00 1.25718021e+00 8.65784526e-01 -4.81245607e-01 4.35440332e-01 -4.35528196e-02 2.33723879e-01 4.48091105e-02 -8.83645654e-01 -2.23514840e-01 2.41109684e-01 7.84492552e-01 8.02323937e-01 4.57581691e-03 -1.11934888e+00 6.70668483e-01 -6.26930952e-01 -5.64274490e-02 1.17527294e+00 8.34672093e-01 -6.62331760e-01 -1.22480524e+00 1.98929146e-01 5.31436086e-01 -5.93082786e-01 -5.45187414e-01 -3.16758841e-01 6.24302506e-01 -2.53501922e-01 9.50100780e-01 -1.05440021e-01 -2.60470271e-01 3.03176612e-01 -1.37866614e-02 5.70725620e-01 -1.08616221e+00 -6.50592923e-01 -1.48272293e-03 7.25286782e-01 6.48816824e-02 -4.03574079e-01 -3.92046243e-01 -9.45742011e-01 1.18407220e-01 -3.17107230e-01 7.17292011e-01 4.11352873e-01 7.99992979e-01 1.50436893e-01 6.12093747e-01 7.13985622e-01 -7.01740623e-01 -8.07086647e-01 -1.62876177e+00 -6.30819619e-01 4.00560021e-01 3.09597075e-01 2.37978458e-01 8.96876678e-02 3.55346292e-01]

[12.29757022857666, 9.474344253540039]

84205f9a-5c3e-47a0-a6f9-d75e4a0c3f19

capture-learning-and-synthesis-of-3d-speaking

1905.03079

https://arxiv.org/abs/1905.03079v1

https://arxiv.org/pdf/1905.03079v1.pdf

Capture, Learning, and Synthesis of 3D Speaking Styles

Audio-driven 3D facial animation has been widely explored, but achieving realistic, human-like performance is still unsolved. This is due to the lack of available 3D datasets, models, and standard evaluation metrics. To address this, we introduce a unique 4D face dataset with about 29 minutes of 4D scans captured at 60 fps and synchronized audio from 12 speakers. We then train a neural network on our dataset that factors identity from facial motion. The learned model, VOCA (Voice Operated Character Animation) takes any speech signal as input - even speech in languages other than English - and realistically animates a wide range of adult faces. Conditioning on subject labels during training allows the model to learn a variety of realistic speaking styles. VOCA also provides animator controls to alter speaking style, identity-dependent facial shape, and pose (i.e. head, jaw, and eyeball rotations) during animation. To our knowledge, VOCA is the only realistic 3D facial animation model that is readily applicable to unseen subjects without retargeting. This makes VOCA suitable for tasks like in-game video, virtual reality avatars, or any scenario in which the speaker, speech, or language is not known in advance. We make the dataset and model available for research purposes at http://voca.is.tue.mpg.de.

['Daniel Cudeiro', 'Michael J. Black', 'Cassidy Laidlaw', 'Anurag Ranjan', 'Timo Bolkart']

2019-05-08

capture-learning-and-synthesis-of-3d-speaking-1

http://openaccess.thecvf.com/content_CVPR_2019/html/Cudeiro_Capture_Learning_and_Synthesis_of_3D_Speaking_Styles_CVPR_2019_paper.html

http://openaccess.thecvf.com/content_CVPR_2019/papers/Cudeiro_Capture_Learning_and_Synthesis_of_3D_Speaking_Styles_CVPR_2019_paper.pdf

cvpr-2019-6

['3d-face-animation', 'talking-face-generation']

['computer-vision', 'computer-vision']

[-6.98942021e-02 1.66251794e-01 5.88009283e-02 -3.89026105e-01 -6.65110648e-01 -6.12317264e-01 4.43180889e-01 -6.64758027e-01 -1.71628013e-01 1.92760631e-01 1.25048637e-01 -1.14790410e-01 5.09398460e-01 -1.54931769e-01 -6.07315183e-01 -4.55017745e-01 -3.76113266e-01 5.66671252e-01 -1.12040155e-01 -2.29287446e-01 -3.15296918e-01 9.56582010e-01 -1.62924719e+00 1.98570248e-02 6.65155472e-03 6.72344744e-01 -2.28170380e-01 1.01587355e+00 2.62167901e-01 3.74931455e-01 -8.39028835e-01 -3.31693858e-01 3.97069126e-01 -4.74238575e-01 -5.60834050e-01 2.66179144e-01 7.47793555e-01 -6.51171386e-01 -3.83843839e-01 7.32580185e-01 8.77851784e-01 3.26888889e-01 4.26659137e-01 -1.37118983e+00 -4.31614399e-01 2.28633329e-01 -4.85287428e-01 3.44369491e-03 6.53293848e-01 4.04307902e-01 5.62644720e-01 -8.74534249e-01 7.61190772e-01 1.67952490e+00 4.44333553e-01 1.41421974e+00 -1.14751697e+00 -9.03480828e-01 1.00138508e-01 -1.83463857e-01 -1.40628076e+00 -8.21948707e-01 9.37197208e-01 -4.38240647e-01 4.79429394e-01 3.01250398e-01 9.92726564e-01 1.70119953e+00 -2.00009167e-01 5.87090433e-01 1.02098966e+00 -3.11139554e-01 1.21070564e-01 9.20263156e-02 -4.39818650e-01 6.75891161e-01 -4.41776544e-01 2.61857450e-01 -6.68255448e-01 -1.34672701e-01 1.10064209e+00 -6.89753592e-01 -3.36171597e-01 -3.30951899e-01 -9.30244446e-01 6.53891206e-01 -1.40469670e-01 -1.19981632e-01 -1.07632227e-01 3.03920746e-01 3.21875811e-01 3.25241178e-01 3.86650294e-01 3.22767794e-01 -3.10663968e-01 -3.52830678e-01 -7.23222911e-01 4.97332007e-01 7.69840419e-01 7.21416712e-01 2.37436756e-01 7.27173507e-01 3.02004725e-01 8.78421307e-01 3.10812622e-01 7.33019173e-01 4.25710708e-01 -1.48789799e+00 -2.25659311e-01 -7.54058287e-02 -1.75236449e-01 -9.65031445e-01 -6.37838006e-01 -1.13879420e-01 -4.85853046e-01 6.43944919e-01 6.27743006e-01 -3.37711930e-01 -8.73312652e-01 2.20852971e+00 6.73385203e-01 4.83677953e-01 -1.77752480e-01 1.17752266e+00 1.15644264e+00 3.95661116e-01 -5.71421459e-02 -1.30009547e-01 1.36211061e+00 -4.54223901e-01 -5.59751272e-01 -1.44120306e-01 1.62885860e-01 -8.72159660e-01 1.14582634e+00 4.46643442e-01 -1.27332270e+00 -5.41174233e-01 -6.61654651e-01 -4.65201819e-03 8.89992900e-03 -1.89612567e-01 5.26784062e-01 1.05772901e+00 -1.35264111e+00 1.71311930e-01 -9.39975321e-01 -3.83197993e-01 2.99135745e-01 6.70402527e-01 -7.10756242e-01 2.95560300e-01 -1.19431150e+00 7.51938105e-01 -2.39462256e-01 -3.30214687e-02 -1.13601756e+00 -7.51204252e-01 -1.06309831e+00 -3.11973691e-01 1.27225816e-01 -6.65086925e-01 1.68530536e+00 -1.32538784e+00 -2.14373016e+00 1.30503452e+00 -4.37299237e-02 -1.94280028e-01 6.05312586e-01 6.57227635e-02 -4.82073098e-01 3.01626116e-01 -2.82674819e-01 1.20376194e+00 1.22849822e+00 -1.17608988e+00 1.32784128e-01 -4.48743820e-01 9.75486711e-02 3.01431984e-01 -9.52253193e-02 3.95882040e-01 -6.12230361e-01 -7.00171471e-01 -2.85365403e-01 -1.28847957e+00 1.38838008e-01 5.72036207e-01 -1.93886250e-01 1.57080755e-01 8.87806594e-01 -6.30027115e-01 5.70220709e-01 -2.34255219e+00 9.00511220e-02 -5.27247880e-03 1.81033731e-01 2.97540426e-01 -4.18307334e-01 -1.02638833e-01 -3.74470294e-01 4.17372212e-02 1.07469782e-01 -5.59243977e-01 -1.43035114e-01 -1.66918132e-02 2.27204897e-02 6.27657473e-01 1.64594859e-01 5.29578745e-01 -6.39927328e-01 -4.31808263e-01 1.25119030e-01 1.04917395e+00 -8.56455028e-01 1.34229198e-01 -1.81593969e-01 9.64307308e-01 -2.33425826e-01 7.81243920e-01 5.89977682e-01 1.93659127e-01 -4.60564010e-02 -6.52918010e-04 1.76565856e-01 3.67512852e-02 -1.19806242e+00 1.64420259e+00 -5.53608477e-01 8.85274947e-01 4.73509759e-01 -5.38656592e-01 9.80848491e-01 5.52092612e-01 6.16457462e-01 -3.32821190e-01 4.81245965e-01 -5.09051792e-02 1.34202391e-01 -4.86256510e-01 7.23924562e-02 -2.97997981e-01 -1.15765063e-02 4.57789183e-01 4.98647504e-02 -4.17363435e-01 -2.35077724e-01 -6.35532141e-02 7.25939095e-01 1.38447568e-01 2.46925261e-02 -2.15988839e-03 2.98439860e-01 -3.23487312e-01 4.50975895e-01 1.77201092e-01 -5.23324251e-01 8.51652563e-01 5.29460371e-01 -2.71389961e-01 -9.15339470e-01 -1.06496119e+00 4.37323004e-02 1.20454693e+00 -2.06035465e-01 -1.16073012e-01 -9.72543538e-01 -1.36959314e-01 -7.31797889e-02 5.28137505e-01 -6.56644821e-01 -1.85008183e-01 -7.43238270e-01 -1.68133438e-01 7.30365753e-01 2.11631984e-01 1.99267715e-01 -1.04926801e+00 -4.42271054e-01 -4.91009355e-02 1.52378539e-02 -1.29425526e+00 -8.47673118e-01 -4.34739113e-01 -4.82322365e-01 -9.21557069e-01 -8.39303017e-01 -7.75144517e-01 4.70986515e-01 -9.88955703e-03 1.03349674e+00 -1.64390966e-01 -4.49160069e-01 7.58970678e-01 -8.78845807e-04 -4.85650301e-01 -7.91793048e-01 -3.12178582e-01 5.32062232e-01 1.20669022e-01 3.47492397e-02 -6.79544091e-01 -4.63627726e-01 3.15667629e-01 -4.26019192e-01 -5.38406707e-03 -4.58366126e-02 5.76410234e-01 4.08867031e-01 -4.13235575e-01 3.97606552e-01 -7.24837482e-01 4.71429348e-01 -1.46103695e-01 -5.95283628e-01 -3.05880755e-01 1.96615875e-01 -4.57066864e-01 3.78087848e-01 -8.62215221e-01 -8.94706428e-01 3.23226750e-01 -5.04925668e-01 -1.03602052e+00 -4.48228687e-01 -2.41072532e-02 -3.77765805e-01 -2.40775302e-01 8.11662376e-01 -1.25728035e-02 5.49409509e-01 -3.38982046e-01 3.14508498e-01 6.29448891e-01 9.39537466e-01 -5.62219322e-01 7.55794287e-01 4.48668122e-01 6.62207603e-02 -1.25826550e+00 -3.11513096e-01 1.34437278e-01 -4.99459088e-01 -5.80335677e-01 8.51451755e-01 -1.12593615e+00 -1.18784702e+00 6.41947031e-01 -1.10168648e+00 -6.03437126e-01 -2.69369632e-01 6.50446713e-01 -7.38698602e-01 1.18262783e-01 -4.64522839e-01 -8.10291290e-01 -2.06034735e-01 -1.35120153e+00 1.08335268e+00 1.50935486e-01 -6.66826963e-01 -7.89255440e-01 -1.45121776e-02 3.72945249e-01 1.92506313e-01 5.17984688e-01 5.53942084e-01 -4.59799260e-01 -2.81455338e-01 -1.15828611e-01 3.83740395e-01 1.97584078e-01 2.07117811e-01 4.97836262e-01 -1.19790590e+00 -3.11372548e-01 -1.73413098e-01 -4.46135342e-01 1.18369728e-01 5.47144234e-01 1.06356514e+00 -2.53841549e-01 9.89164189e-02 8.01669657e-01 4.16462868e-01 2.94336259e-01 1.74296215e-01 -7.47709945e-02 6.30753577e-01 7.36205757e-01 2.26021752e-01 5.05648077e-01 3.02075863e-01 1.00400603e+00 4.90678668e-01 2.29941420e-02 -3.92651558e-01 -1.83101252e-01 5.86074293e-01 6.15467548e-01 -6.94952011e-02 -1.86766326e-01 -7.78392136e-01 3.05479109e-01 -9.33335006e-01 -1.05259740e+00 3.40561718e-01 2.20221186e+00 8.01836073e-01 1.89253520e-02 5.91965497e-01 -5.19506782e-02 8.37984324e-01 9.58996639e-02 -7.33338654e-01 -5.99457204e-01 -2.46770983e-03 4.08139020e-01 -1.17764108e-01 8.04009020e-01 -1.05431187e+00 1.09293950e+00 6.10064411e+00 4.00411069e-01 -1.66235113e+00 -1.42096095e-02 5.46598256e-01 -5.52807271e-01 -2.40988925e-01 -3.99281949e-01 -6.02771699e-01 2.42552698e-01 8.81313980e-01 -1.61798656e-01 6.74122930e-01 8.29700589e-01 5.53540647e-01 4.11131144e-01 -1.11235988e+00 1.14709413e+00 3.25162023e-01 -9.61657584e-01 -1.03090229e-02 9.11962520e-03 3.19223195e-01 -2.79007405e-01 5.17363012e-01 2.58642077e-01 2.68722117e-01 -1.21926820e+00 8.43286753e-01 2.48172596e-01 1.34361088e+00 -6.52266026e-01 -5.47584482e-02 1.99122101e-01 -8.64984155e-01 3.55327129e-01 1.24102376e-01 1.54878676e-01 2.21657559e-01 -2.19749063e-01 -8.68853807e-01 -3.00756961e-01 5.77698231e-01 3.89618367e-01 -2.11026385e-01 5.79587042e-01 -1.66040659e-01 6.63546383e-01 -4.96681869e-01 1.39256477e-01 -8.56770501e-02 1.68737337e-01 8.21905255e-01 9.13564086e-01 2.88548827e-01 2.79976159e-01 -4.70269881e-02 4.13343579e-01 -1.80425018e-01 2.00424820e-01 -7.30368018e-01 7.62661779e-03 5.38793862e-01 1.05911100e+00 -3.35777283e-01 2.15057999e-01 -3.09290737e-01 8.45560730e-01 -4.66209948e-02 3.24558169e-01 -7.77601957e-01 -1.08284270e-03 1.24209201e+00 3.62654656e-01 -3.82466726e-02 -3.89097780e-01 2.49193192e-01 -1.01906049e+00 -4.31487381e-01 -1.37714350e+00 1.62769482e-01 -9.64525044e-01 -8.74304950e-01 8.28782082e-01 1.16758481e-01 -1.10996878e+00 -7.16443658e-01 -6.04866266e-01 -4.58760411e-01 7.31540918e-01 -9.33753729e-01 -1.11642897e+00 -3.62941176e-01 8.00864875e-01 6.60714090e-01 -3.58108848e-01 9.19240475e-01 2.82963425e-01 -6.18847966e-01 1.01283729e+00 -6.18168056e-01 3.04272622e-01 9.81969774e-01 -8.27297986e-01 6.67551219e-01 3.34576964e-01 2.42801666e-01 4.76722389e-01 9.01532590e-01 -2.51023710e-01 -1.48465121e+00 -8.89563441e-01 5.32337844e-01 -5.13876259e-01 4.73672718e-01 -5.76317489e-01 -8.45928013e-01 7.30880976e-01 -1.84575003e-02 2.70801127e-01 8.54759216e-01 -1.79826841e-01 -3.56982589e-01 -8.18246305e-02 -1.20376027e+00 9.43693638e-01 1.06558669e+00 -5.21645904e-01 1.76439795e-03 3.43605161e-01 5.68951726e-01 -9.80370224e-01 -7.79772162e-01 1.60046145e-01 8.98563802e-01 -9.09356534e-01 1.00461125e+00 -6.57069921e-01 5.39921708e-02 -9.09108892e-02 -1.67871073e-01 -1.28879392e+00 3.85094546e-02 -1.11300504e+00 -5.03529375e-03 1.09132922e+00 1.08393751e-01 -4.58132118e-01 9.89309907e-01 6.61522806e-01 1.58418138e-02 -4.52747047e-01 -1.07105911e+00 -6.16020203e-01 1.82105005e-01 -6.05153620e-01 6.59852386e-01 9.53175783e-01 -2.63569802e-01 2.95770466e-01 -5.90999842e-01 3.03269386e-01 5.11018395e-01 -1.21296503e-01 1.23844957e+00 -1.15641487e+00 -3.96628469e-01 -5.50637364e-01 -5.31459570e-01 -8.71194124e-01 6.85854971e-01 -7.14301229e-01 -2.37206385e-01 -6.85296059e-01 -2.93991089e-01 -3.02533448e-01 4.86387521e-01 5.85675836e-01 1.36643931e-01 3.49320740e-01 2.46609658e-01 -1.76297035e-02 1.67876005e-01 5.45763373e-01 1.51709569e+00 6.76889047e-02 -3.00317556e-01 3.04233193e-01 -4.42271054e-01 9.75926101e-01 8.21327209e-01 -2.46489987e-01 -6.21611834e-01 -4.13958460e-01 -4.27308559e-01 5.17741263e-01 5.91943681e-01 -8.06600511e-01 -3.38918157e-02 -1.29770860e-01 3.72124642e-01 3.94636653e-02 1.00959468e+00 -6.01935387e-01 3.23224485e-01 2.31394634e-01 -3.53291154e-01 1.54219165e-01 4.62891102e-01 1.25663340e-01 -9.87445004e-03 1.44620821e-01 1.07234204e+00 -1.20824486e-01 -4.82196957e-01 6.13390684e-01 -5.64427376e-01 3.06049019e-01 8.49163949e-01 -2.80883193e-01 8.85958299e-02 -1.07907271e+00 -1.14852667e+00 -1.44957110e-01 5.62880158e-01 5.66616595e-01 5.22927284e-01 -1.27643979e+00 -8.68919075e-01 6.19293809e-01 -1.09107099e-01 -1.14953533e-01 4.46803510e-01 3.64945948e-01 -6.48132980e-01 -7.70876780e-02 -3.98712933e-01 -6.40862703e-01 -1.73146069e+00 3.12748075e-01 7.01505244e-01 4.96365577e-01 -4.91556942e-01 9.62511003e-01 2.26785630e-01 -4.57805842e-01 4.04923409e-01 1.05643816e-01 -1.44818306e-01 -3.05040702e-02 4.72632825e-01 8.52903947e-02 -2.03585982e-01 -1.21370327e+00 -4.39995289e-01 6.82184219e-01 1.29339695e-01 -4.42973644e-01 9.90430534e-01 6.42440319e-02 4.08684552e-01 3.63459289e-01 1.22927117e+00 2.48452976e-01 -1.42795563e+00 1.38684541e-01 -8.22012663e-01 -4.53156292e-01 -9.14880857e-02 -3.65855336e-01 -1.37127674e+00 1.11184323e+00 6.58201516e-01 -2.87176341e-01 1.07314456e+00 -3.71949710e-02 5.23736954e-01 1.36413649e-01 2.18566969e-01 -5.67533791e-01 3.19005907e-01 3.26138467e-01 1.11663234e+00 -1.03067362e+00 -1.59277678e-01 -3.95735025e-01 -8.14091086e-01 1.01994741e+00 8.18643272e-01 7.10751936e-02 7.94112623e-01 5.10163307e-01 6.94134176e-01 -1.69337839e-02 -7.17530072e-01 3.04359533e-02 1.18188739e-01 9.56952333e-01 3.97741288e-01 -5.74128367e-02 3.67450058e-01 2.55612135e-01 -8.00524473e-01 -1.81940317e-01 5.46676636e-01 6.07089579e-01 -1.19763099e-01 -9.17046130e-01 -5.06006539e-01 -1.37799814e-01 -5.26082933e-01 2.97597259e-01 -4.82908905e-01 1.03731859e+00 -4.22535613e-02 7.22093225e-01 2.38528118e-01 -3.11345339e-01 4.29262996e-01 1.14481486e-01 6.72733009e-01 -6.45954430e-01 -4.10947651e-01 2.28788540e-01 6.59410888e-03 -4.53009278e-01 -1.98543295e-01 -9.83418047e-01 -1.22975123e+00 -5.22192895e-01 2.14703660e-02 -1.60330147e-01 7.66443014e-01 3.92395467e-01 3.40322942e-01 1.42078042e-01 8.15447152e-01 -1.20107722e+00 -2.88185775e-01 -6.82142675e-01 -5.56416988e-01 2.27024376e-01 6.32288992e-01 -7.78740764e-01 -4.39455211e-01 1.65012076e-01]

[13.166370391845703, -0.40191134810447693]

32e2adab-dad2-478c-9789-1243982067ba

self-organizing-intelligent-matter-a-1

2101.07627

https://arxiv.org/abs/2101.07627v1

https://arxiv.org/pdf/2101.07627v1.pdf

Self-Organizing Intelligent Matter: A blueprint for an AI generating algorithm

We propose an artificial life framework aimed at facilitating the emergence of intelligent organisms. In this framework there is no explicit notion of an agent: instead there is an environment made of atomic elements. These elements contain neural operations and interact through exchanges of information and through physics-like rules contained in the environment. We discuss how an evolutionary process can lead to the emergence of different organisms made of many such atomic elements which can coexist and thrive in the environment. We discuss how this forms the basis of a general AI generating algorithm. We provide a simplified implementation of such system and discuss what advances need to be made to scale it up further.

['Frederic Besse', 'Karol Gregor']

2021-01-19

self-organizing-intelligent-matter-a

https://openreview.net/forum?id=160xFQdp7HR

https://openreview.net/pdf?id=160xFQdp7HR

['artificial-life']

['miscellaneous']

[ 1.54140681e-01 4.07525122e-01 4.96399999e-01 5.95710415e-04 7.53086925e-01 -4.04522270e-01 1.16593230e+00 6.48316042e-03 -4.03298885e-01 1.02043164e+00 -1.51706571e-02 -1.37220874e-01 -8.24905038e-02 -1.34563732e+00 -7.17379093e-01 -8.96927357e-01 -3.60153139e-01 5.89080870e-01 3.59664887e-01 -8.93282235e-01 2.54104406e-01 3.80131960e-01 -1.94750667e+00 -6.70180097e-02 9.16696966e-01 3.60483855e-01 4.04754281e-01 9.63250935e-01 -8.96203071e-02 1.03646421e+00 -7.34401762e-01 -2.10731193e-01 3.27251315e-01 -1.00207663e+00 -8.57856929e-01 8.05741251e-02 -3.84328008e-01 2.52818525e-01 -1.12334877e-01 7.40652859e-01 1.52723685e-01 3.05788517e-01 7.33593166e-01 -1.03536558e+00 -9.30632234e-01 8.29047322e-01 1.63122788e-01 1.87046006e-02 4.12052602e-01 3.39908272e-01 4.97345209e-01 -3.73691589e-01 7.82952845e-01 1.11565495e+00 4.88068223e-01 8.46455514e-01 -9.67102170e-01 -5.04149199e-02 -2.55019099e-01 4.41399254e-02 -1.14459968e+00 -3.61597955e-01 4.64939505e-01 -2.12350622e-01 1.21516073e+00 3.61135900e-01 1.44172418e+00 8.49286795e-01 7.67361701e-01 3.58587384e-01 8.28585148e-01 -1.09177303e+00 6.38918459e-01 2.22805202e-01 9.87084135e-02 6.69326484e-01 6.77136183e-01 3.67802531e-01 -7.01959074e-01 -8.22553858e-02 9.78820920e-01 -5.25652051e-01 -6.94467872e-02 -1.85445234e-01 -1.15633118e+00 4.66701001e-01 3.67230862e-01 1.05645311e+00 -5.62568367e-01 4.54510212e-01 -5.31855822e-02 4.38710183e-01 -1.10643595e-01 1.03823507e+00 -3.02116543e-01 -7.90001303e-02 -1.96574271e-01 2.97942877e-01 1.10547256e+00 3.81618053e-01 7.02126682e-01 1.27980337e-01 6.18062913e-01 3.33489120e-01 3.18871379e-01 9.07896310e-02 7.02404201e-01 -1.17997527e+00 -5.42338073e-01 7.95993626e-01 -2.65734224e-03 -9.60714757e-01 -2.54344791e-01 -5.00945687e-01 -3.82351846e-01 4.27545249e-01 -1.59872230e-02 -2.75355250e-01 -5.24372041e-01 1.75201178e+00 4.01337475e-01 2.65484959e-01 4.26781237e-01 3.55931312e-01 5.18553615e-01 8.22247148e-01 -3.00180048e-01 -2.44112447e-01 1.08235741e+00 -9.08827543e-01 -7.52606332e-01 1.72375619e-01 3.93792421e-01 -1.18847914e-01 4.45035666e-01 5.76994538e-01 -1.59316170e+00 -5.07754385e-01 -1.32481945e+00 3.34034473e-01 -7.66068697e-01 -7.10425436e-01 1.21986663e+00 7.51863420e-01 -1.62348950e+00 6.33710027e-01 -7.44084835e-01 -7.28611708e-01 -2.92308658e-01 5.41085660e-01 -1.50447309e-01 9.28879499e-01 -1.17297590e+00 1.24034274e+00 7.82012343e-01 -5.99404648e-02 -6.86272979e-01 -1.74978405e-01 -4.65646684e-01 -2.27840245e-01 1.51428074e-01 -1.44103670e+00 1.10986674e+00 -1.32868171e+00 -1.63944221e+00 9.77127254e-01 -8.05641562e-02 -5.95913172e-01 6.26745671e-02 3.07306230e-01 -3.90192837e-01 -9.02332664e-02 -4.73871112e-01 5.66637874e-01 2.15851754e-01 -1.70127356e+00 -6.61470294e-01 -2.40170285e-01 2.80428171e-01 3.03883404e-01 -3.41300011e-01 1.17193468e-01 -1.13964714e-01 -5.36387742e-01 -4.81895767e-02 -1.01870155e+00 -4.61493254e-01 -3.86767417e-01 1.05476946e-01 -4.16942745e-01 3.54221612e-01 6.13744482e-02 9.44476128e-01 -1.86007166e+00 5.47775984e-01 2.36476839e-01 4.39144939e-01 4.40734364e-02 -3.84585671e-02 9.58658159e-01 2.06580281e-01 2.21291915e-01 -2.58985430e-01 3.57750431e-02 2.41312869e-02 4.66799825e-01 6.06458783e-02 6.64096549e-02 -8.09343159e-02 9.93557453e-01 -9.63779986e-01 -2.48443902e-01 4.39025462e-02 5.55550754e-01 -5.38487852e-01 9.18303728e-02 -4.15712059e-01 5.70612013e-01 -6.83576941e-01 2.33454019e-01 4.47136760e-02 -1.21617846e-01 2.89943293e-02 6.95363224e-01 -5.99639237e-01 1.57233283e-01 -9.17762518e-01 1.56998110e+00 -2.16164500e-01 4.95501965e-01 1.17463320e-01 -1.12093198e+00 9.38280702e-01 4.13302094e-01 2.73342073e-01 -3.76932949e-01 5.99393129e-01 2.30480522e-01 7.87857234e-01 -3.84644806e-01 3.72239321e-01 -8.67993906e-02 6.28290251e-02 8.10814202e-01 8.68530422e-02 -6.05484009e-01 4.12688017e-01 1.32841781e-01 1.32692945e+00 9.34468210e-02 6.30338550e-01 -5.38705289e-01 7.83597052e-01 3.77949215e-02 5.04471421e-01 9.71977353e-01 -9.17768106e-02 -1.99694246e-01 -2.00401291e-01 -9.25708473e-01 -1.15540397e+00 -9.61924076e-01 -2.24785417e-01 8.10993135e-01 3.17746609e-01 -4.58608359e-01 -1.06528044e+00 2.94590056e-01 -3.78527910e-01 9.52298164e-01 -1.02752376e+00 -1.71542645e-01 -4.42461401e-01 -1.19951618e+00 2.54139096e-01 -2.46905893e-01 6.80113554e-01 -1.80426717e+00 -1.38550889e+00 4.03422385e-01 4.25284147e-01 -6.98526323e-01 4.32920754e-01 3.46755862e-01 -9.61832523e-01 -3.30542684e-01 2.63830530e-03 -7.42164612e-01 5.91540754e-01 -1.43423915e-01 1.27595651e+00 7.95235455e-01 -5.18209815e-01 4.72785383e-01 -4.65461850e-01 -7.85717189e-01 -1.11100042e+00 -3.86762649e-01 4.61876631e-01 -2.45467469e-01 1.11313455e-01 -1.25759768e+00 -3.42391193e-01 -2.34172121e-01 -1.08010960e+00 3.37247103e-01 3.31640720e-01 4.95443106e-01 3.49795632e-02 6.48743510e-01 4.40019816e-01 -4.41449076e-01 7.46600270e-01 -3.63128603e-01 -2.52095014e-01 4.85451132e-01 -3.19278419e-01 1.82893455e-01 5.64824164e-01 -1.81731418e-01 -1.06348014e+00 -9.49384570e-02 1.12435587e-01 6.71845317e-01 -3.13400269e-01 2.26170644e-01 -8.66357535e-02 -3.86024177e-01 7.13611305e-01 4.95931476e-01 -7.61294216e-02 -2.18640044e-01 4.85141218e-01 6.87915444e-01 6.54420972e-01 -7.42431402e-01 5.69291949e-01 3.64857525e-01 9.21090171e-02 -9.64205146e-01 1.01094000e-01 4.21949089e-01 -7.04088330e-01 -4.81898427e-01 8.58574212e-01 -2.60060400e-01 -6.16599977e-01 6.98901176e-01 -1.25196576e+00 -2.49788478e-01 -6.78560734e-01 2.76765257e-01 -7.84839332e-01 1.48681523e-02 -5.96544862e-01 -1.09249818e+00 -2.57802218e-01 -7.87232995e-01 3.08178395e-01 6.69715703e-01 -4.44978952e-01 -1.13881743e+00 7.27735996e-01 1.72191530e-01 5.43680131e-01 2.01600015e-01 7.59767056e-01 -5.12228191e-01 -4.72663909e-01 1.53692484e-01 7.29240716e-01 -1.34211168e-01 2.76892334e-01 6.20219231e-01 -5.86056113e-01 -3.51937898e-02 4.29958850e-01 5.59220463e-03 4.72027183e-01 4.21341173e-02 4.78001803e-01 -2.35254720e-01 -5.85883260e-01 3.62055153e-01 1.36426735e+00 1.09448087e+00 6.86246753e-01 6.09501839e-01 1.53689921e-01 9.66912448e-01 -2.73271233e-01 3.24865460e-01 2.68770605e-01 4.58056569e-01 3.14030439e-01 1.58822224e-01 1.27674624e-01 2.49171644e-01 2.46068552e-01 1.38603032e+00 -8.32332075e-01 -3.99628222e-01 -1.10371196e+00 3.49325895e-01 -1.87280011e+00 -1.16170859e+00 -1.20833762e-01 1.84167218e+00 9.95720565e-01 -1.44417083e-03 3.73481028e-02 3.00607979e-02 6.02674246e-01 -5.89770675e-01 -5.80253720e-01 -9.66735959e-01 -4.10563737e-01 3.25234175e-01 -1.21862419e-01 4.66061294e-01 -5.47835827e-01 1.08318555e+00 8.19575977e+00 2.39003897e-01 -7.71020591e-01 -7.95004666e-02 3.01271379e-01 1.35782108e-01 -3.78133357e-01 1.52954638e-01 -3.56796741e-01 4.40771580e-01 1.26715088e+00 -4.62521702e-01 8.10495913e-01 2.15956569e-01 1.14872836e-01 -4.86169904e-01 -9.30784464e-01 5.29062688e-01 -2.02712342e-02 -1.44331586e+00 2.41512269e-01 3.22958231e-01 9.44594264e-01 -9.09932181e-02 -3.25144023e-01 -2.80824423e-01 9.66308355e-01 -1.08053827e+00 7.05136895e-01 8.74774694e-01 -2.79716343e-01 -6.68674052e-01 2.20757082e-01 7.06166387e-01 -8.56332600e-01 -1.70878589e-01 -2.48183787e-01 -6.48326457e-01 2.35830605e-01 3.33973765e-01 -3.31052423e-01 5.89187682e-01 6.64944172e-01 2.57658660e-01 -4.58901942e-01 1.04785240e+00 -6.64072484e-02 1.15628220e-01 -4.20011938e-01 -6.17537200e-01 8.45358521e-02 -6.51655555e-01 8.13803613e-01 8.70330095e-01 4.35233653e-01 5.00005841e-01 -3.78974855e-01 1.12613964e+00 3.71790290e-01 7.30922893e-02 -8.45062613e-01 -6.13851547e-02 4.09672469e-01 1.06617928e+00 -1.00124204e+00 -5.92970788e-01 -1.57127589e-01 9.42689717e-01 4.84778583e-02 -6.78392826e-04 -7.15894818e-01 -2.35681221e-01 4.97099310e-01 -2.21741170e-01 -1.86605416e-02 -4.26516533e-01 -1.59789175e-01 -1.04197705e+00 -5.17664015e-01 -1.11929548e+00 -3.49194586e-01 -8.34724605e-01 -1.09509957e+00 9.64320958e-01 -1.46467283e-01 -2.31127962e-01 -3.90209258e-01 -4.63920087e-01 -9.39684629e-01 4.29276288e-01 -4.60397750e-01 -7.63102293e-01 -2.12089539e-01 4.19336647e-01 1.42352790e-01 -5.55571258e-01 1.13391745e+00 -5.29751956e-01 -5.66283882e-01 -1.02201551e-01 1.66400567e-01 -3.79121989e-01 -1.62709311e-01 -1.17508948e+00 5.65931678e-01 8.51767838e-01 3.86498988e-01 9.84878123e-01 1.10775685e+00 -6.66533172e-01 -1.50263190e+00 -2.35983312e-01 8.32528889e-01 -5.71957469e-01 6.62262440e-01 -4.52835530e-01 -6.21671081e-01 3.90748084e-01 9.62580144e-01 -6.09838665e-01 6.56130910e-01 -1.04875371e-01 3.13675642e-01 1.76932573e-01 -9.74087358e-01 9.45722103e-01 1.44595194e+00 -3.23649645e-02 -1.14366150e+00 1.29664153e-01 7.72770345e-01 3.19646180e-01 -6.42583013e-01 5.21146297e-01 6.64920211e-01 -1.23517442e+00 7.87801206e-01 -4.68497992e-01 5.28727770e-02 -5.16995311e-01 1.28715366e-01 -1.21215355e+00 -4.90826756e-01 -9.90515292e-01 -6.32084161e-03 9.19932961e-01 3.69904906e-01 -1.23626745e+00 3.72133583e-01 7.17861414e-01 3.01931258e-02 -4.02434498e-01 -6.45716429e-01 -7.38624394e-01 2.92999387e-01 -1.06259696e-01 8.35909605e-01 8.60272825e-01 5.49125493e-01 2.75466025e-01 1.88896000e-01 -3.80542219e-01 5.63196540e-01 -5.07781878e-02 5.78580976e-01 -1.30772984e+00 -7.76228786e-01 -6.08905494e-01 -5.99114895e-01 -4.14024979e-01 3.48071158e-02 -6.60456598e-01 1.18588954e-01 -1.39810610e+00 2.79060125e-01 -4.74098235e-01 -2.94921935e-01 2.11494774e-01 3.07611763e-01 1.18164785e-01 1.93834677e-01 3.81129205e-01 -4.82668638e-01 5.24733424e-01 1.06733489e+00 4.67755824e-01 -2.82893986e-01 -5.83984137e-01 -7.86994994e-01 9.55465376e-01 1.01375079e+00 -2.78035700e-01 -4.51004982e-01 -1.94858611e-01 6.92136645e-01 -2.30993912e-01 3.20496000e-02 -1.53342545e+00 3.75794888e-01 -2.30521157e-01 2.40762889e-01 4.73722704e-02 3.40938061e-01 -6.90596163e-01 8.83280218e-01 1.00991225e+00 -1.47134677e-01 3.25826824e-01 -7.15179667e-02 1.51313901e-01 5.29704764e-02 -5.66430151e-01 5.70697844e-01 -5.44979870e-01 -5.24279952e-01 -3.36548984e-01 -1.07831645e+00 -5.68194747e-01 1.48890233e+00 -4.90990967e-01 -1.80071950e-01 -1.91002220e-01 -7.80565083e-01 1.64514557e-02 1.08421278e+00 1.86330035e-01 1.87622577e-01 -8.11868250e-01 -6.30501568e-01 1.79803953e-01 -4.12914753e-01 -3.72014016e-01 -3.45279127e-01 2.58998662e-01 -9.96236682e-01 4.57590759e-01 -8.88931394e-01 -7.47354701e-02 -1.06857145e+00 5.75118005e-01 7.47250795e-01 1.11237496e-01 -4.08726066e-01 1.03898728e+00 4.83277082e-01 -2.86861151e-01 -1.49875760e-01 -2.36212965e-02 -4.98147726e-01 -3.58669728e-01 5.98462701e-01 -4.86986488e-02 -3.61793160e-01 -6.22034252e-01 -3.61936480e-01 5.77839434e-01 3.96841645e-01 -5.00068247e-01 1.53199208e+00 -5.51707260e-02 -9.72817719e-01 8.70404959e-01 2.54719377e-01 5.58346063e-02 -6.09305203e-01 3.99192870e-01 -1.30248815e-01 3.80836502e-02 -1.88575134e-01 -7.31566727e-01 -6.70462668e-01 3.24965984e-01 1.86789021e-01 7.86659777e-01 1.22586262e+00 9.97854979e-04 5.63400984e-01 5.83094180e-01 9.12606776e-01 -1.03024197e+00 -1.10159434e-01 5.63344300e-01 8.77609372e-01 -3.97445619e-01 -2.11478427e-01 -1.57192752e-01 -6.12590052e-02 1.11576760e+00 4.77577686e-01 -5.49904883e-01 6.11601710e-01 6.85261428e-01 -3.11610192e-01 -2.48352617e-01 -1.44829869e+00 -3.22809905e-01 -1.16528548e-01 7.33954370e-01 5.64560413e-01 -1.09418973e-01 -9.32706118e-01 2.88003623e-01 -6.15683854e-01 2.20166519e-02 6.72248781e-01 1.19631839e+00 -7.58177817e-01 -1.57567060e+00 -4.24896002e-01 -1.39691874e-01 -2.07654491e-01 7.13270381e-02 -9.51344430e-01 6.49009287e-01 9.39814985e-01 9.35603023e-01 1.04948059e-01 -4.33099329e-01 -2.96503276e-01 -2.91545540e-02 1.05324566e+00 -5.56854546e-01 -8.85337591e-01 -5.37469923e-01 3.25651728e-02 -1.68282911e-01 -7.62458384e-01 -8.55311811e-01 -1.61124647e+00 -5.65103352e-01 -1.64846227e-01 6.02933645e-01 6.85765207e-01 7.33087599e-01 2.85191298e-01 5.63009202e-01 3.01841855e-01 -7.76174307e-01 2.09140614e-01 -4.63990271e-01 -4.29934651e-01 4.20218669e-02 -3.55116092e-02 -5.38392305e-01 -3.67251426e-01 3.53424013e-01]

[5.582370758056641, 4.135457515716553]

b59de970-a8d0-4de2-995f-f062c2015dce

2d-driven-3d-object-detection-in-rgb-d-images

http://openaccess.thecvf.com/content_iccv_2017/html/Lahoud_2D-Driven_3D_Object_ICCV_2017_paper.html

http://openaccess.thecvf.com/content_ICCV_2017/papers/Lahoud_2D-Driven_3D_Object_ICCV_2017_paper.pdf

2D-Driven 3D Object Detection in RGB-D Images

In this paper, we present a technique that places 3D bounding boxes around objects in an RGB-D scene. Our approach makes best use of the 2D information to quickly reduce the search space in 3D, benefiting from state-of-the-art 2D object detection techniques. We then use the 3D information to orient, place, and score bounding boxes around objects. We independently estimate the orientation for every object, using previous techniques that utilize normal information. Object locations and sizes in 3D are learned using a multilayer perceptron (MLP). In the final step, we refine our detections based on object class relations within a scene. When compared to state-of-the-art detection methods that operate almost entirely in the sparse 3D domain, extensive experiments on the well-known SUN RGB-D dataset show that our proposed method is much faster (4.1s per image) in detecting 3D objects in RGB-D images and performs better (3 mAP higher) than the state-of-the-art method that is 4.7 times slower and comparably to the method that is two orders of magnitude slower. This work hints at the idea that 2D-driven object detection in 3D should be further explored, especially in cases where the 3D input is sparse.

['Jean Lahoud', 'Bernard Ghanem']

2017-10-01

iccv-2017-10

['object-detection-in-indoor-scenes']

['computer-vision']

[ 2.12000504e-01 4.55553271e-02 1.57253057e-01 -2.33995497e-01 -3.06312561e-01 -5.77535152e-01 5.70907652e-01 4.53641832e-01 -7.52047181e-01 -7.47619867e-02 -3.65679562e-01 -3.62879038e-01 8.45322087e-02 -7.38231957e-01 -7.72493184e-01 -5.43721259e-01 -3.04470062e-01 7.84591377e-01 1.03810644e+00 3.17065567e-02 6.30163729e-01 1.23273516e+00 -1.79531932e+00 1.76332414e-01 -8.38522017e-02 1.31284678e+00 3.95215988e-01 9.70268250e-01 -2.11763754e-01 3.97373378e-01 -5.50611973e-01 2.88807362e-01 6.49250567e-01 1.41897634e-01 -5.14567256e-01 3.84563506e-01 6.87551796e-01 -3.83593112e-01 -2.56411135e-01 7.05075264e-01 4.43253726e-01 -6.48606345e-02 6.47903383e-01 -9.36280608e-01 -2.78952032e-01 -9.07259956e-02 -1.02522635e+00 3.48319739e-01 4.14953023e-01 -2.43600998e-02 5.83020806e-01 -1.14286554e+00 3.85577112e-01 1.35674357e+00 7.35622406e-01 1.72088698e-01 -1.09249425e+00 -3.07196110e-01 2.59157836e-01 -7.32982755e-02 -1.47848892e+00 7.50574246e-02 7.97637403e-01 -5.09974897e-01 1.45688593e+00 3.23663443e-01 7.64957845e-01 2.89797038e-01 -2.10560426e-01 7.97311842e-01 1.21828699e+00 -9.67275262e-01 2.43542358e-01 3.25994462e-01 1.05489783e-01 8.24156225e-01 3.36869121e-01 9.83517766e-02 -5.12899518e-01 -1.29977530e-02 1.11078882e+00 1.90767869e-01 2.21945390e-01 -7.83473849e-01 -1.38028383e+00 6.20615900e-01 8.22675526e-01 9.75790322e-02 -4.68749315e-01 1.73874840e-01 8.29303116e-02 -2.03979492e-01 4.25130785e-01 2.69304484e-01 -3.86383742e-01 1.11391947e-01 -8.52250576e-01 2.98836529e-01 6.59520626e-01 8.62248421e-01 8.28526080e-01 -4.93581682e-01 1.56746820e-01 4.95021611e-01 5.02969265e-01 5.07492363e-01 3.91513258e-02 -8.35294485e-01 2.41601706e-01 1.02730274e+00 2.18284905e-01 -7.79854715e-01 -7.09011197e-01 -1.67995960e-01 -4.90610540e-01 9.04473722e-01 4.00498688e-01 4.09029990e-01 -1.16008854e+00 8.08006287e-01 7.20481098e-01 -3.11108083e-01 -3.23573202e-01 1.13246906e+00 7.29283512e-01 4.34345335e-01 -4.74145681e-01 3.74918282e-01 1.38549578e+00 -7.60235488e-01 1.42386436e-01 -4.96405840e-01 2.21249729e-01 -9.09344196e-01 6.17679954e-01 4.83297795e-01 -1.04942405e+00 -8.10651660e-01 -1.14618289e+00 -7.54107088e-02 -6.61746323e-01 4.34852272e-01 7.09134102e-01 7.59208202e-01 -9.31750000e-01 3.00184786e-01 -8.99950266e-01 -6.67955875e-01 4.69796598e-01 5.30396104e-01 -2.37029165e-01 1.27579078e-01 -3.64789933e-01 1.16707444e+00 7.28624344e-01 8.12642649e-02 -7.02749014e-01 -4.40333605e-01 -6.95551336e-01 -3.03720206e-01 4.95504886e-01 -3.98464173e-01 1.25526881e+00 -2.73115903e-01 -1.34168863e+00 1.41272962e+00 -1.27584234e-01 -1.84412882e-01 3.97383332e-01 -3.63453090e-01 2.14764923e-01 2.55871296e-01 -9.62735489e-02 9.02930260e-01 8.16415012e-01 -1.25961113e+00 -1.15484047e+00 -7.24156260e-01 2.96064436e-01 3.07918578e-01 -5.46777770e-02 4.17301618e-03 -7.91022778e-01 -7.37562776e-02 7.61134684e-01 -8.72568727e-01 -3.67593139e-01 5.04426777e-01 -4.89333749e-01 -4.45357293e-01 8.34347367e-01 -6.44292012e-02 6.78339183e-01 -2.05700016e+00 4.84871864e-02 2.69834638e-01 1.44022435e-01 1.78948656e-01 3.08630258e-01 2.13900611e-01 1.22963727e-01 -1.51961267e-01 1.99832837e-03 -5.35690188e-01 -3.62077355e-02 8.07044730e-02 -1.22385308e-01 8.43970954e-01 4.19009864e-01 2.95558989e-01 -7.02201366e-01 -5.29447198e-01 6.35173798e-01 5.12367547e-01 -4.18213069e-01 1.87765121e-01 -8.71772021e-02 -3.90123867e-04 -5.01248658e-01 1.04059708e+00 8.32313180e-01 -2.53452629e-01 -2.69676417e-01 -2.93111920e-01 -6.01027548e-01 2.67739803e-01 -1.53774691e+00 1.56312096e+00 -2.52548397e-01 7.11502850e-01 -7.86378086e-02 -9.20387685e-01 1.25906682e+00 -1.62072286e-01 3.27071697e-01 -4.68618155e-01 1.33899292e-02 2.57547468e-01 -3.55969459e-01 -2.76516795e-01 5.93758762e-01 2.31143072e-01 5.07133007e-02 5.68156004e-01 2.08599512e-02 -4.73374277e-01 2.02019393e-01 -8.41151699e-02 1.09961879e+00 2.98178673e-01 5.02948165e-01 4.81739417e-02 4.15542096e-01 2.48513684e-01 -1.15787163e-01 1.23038030e+00 -2.41755933e-01 6.16469502e-01 2.93388039e-01 -7.29725420e-01 -1.12798488e+00 -1.02403569e+00 -2.80520737e-01 8.73870373e-01 3.77501726e-01 -2.92997390e-01 -3.80833894e-01 -6.98923469e-01 3.31644714e-01 2.06527770e-01 -5.81129313e-01 4.19158012e-01 -6.71934128e-01 -7.53788710e-01 2.21172273e-01 6.44349515e-01 4.11906242e-01 -7.41391480e-01 -1.51820838e+00 3.69313732e-02 5.56517899e-01 -1.07863033e+00 2.87399054e-01 8.80910158e-01 -1.17618394e+00 -1.06468856e+00 -7.67129779e-01 -6.85997248e-01 1.00778389e+00 6.94973886e-01 1.11353600e+00 -9.02520195e-02 -6.49743915e-01 4.66903597e-01 -4.60035771e-01 -8.33635926e-01 -2.76038256e-02 -1.49363443e-01 1.52852878e-01 -4.75320548e-01 8.25884104e-01 -2.12914065e-01 -5.50564945e-01 3.04758251e-01 -6.04166269e-01 7.33554885e-02 1.00230467e+00 3.46528649e-01 8.22292984e-01 5.01515046e-02 -2.93148279e-01 -4.78508890e-01 -9.44376066e-02 2.25722836e-03 -1.10406697e+00 -1.61286015e-02 -3.94287169e-01 5.35222739e-02 -1.05768777e-02 -4.89651591e-01 -6.40562892e-01 8.40422094e-01 1.86679795e-01 -4.81783748e-01 -7.68578231e-01 -2.85414875e-01 3.90456319e-01 -5.18139303e-01 8.05726409e-01 5.98950572e-02 -2.72707462e-01 -7.54082561e-01 3.31451654e-01 7.70310640e-01 3.42679948e-01 -7.99973011e-02 9.24027801e-01 9.52493906e-01 2.90496796e-01 -9.09431517e-01 -8.44731987e-01 -9.80898738e-01 -1.23755682e+00 -3.53869587e-01 6.50890768e-01 -7.55293846e-01 -7.65054405e-01 2.79749662e-01 -1.47341204e+00 -1.13939635e-01 -2.68189520e-01 5.52960992e-01 -4.23051417e-01 7.89128095e-02 -4.13975388e-01 -1.36139035e+00 4.05265801e-02 -8.75197291e-01 1.59901106e+00 3.12684983e-01 1.03672773e-01 -4.27650899e-01 8.76071304e-02 -3.07248738e-02 1.65371582e-01 1.71977609e-01 4.99465495e-01 -3.80138874e-01 -8.49146247e-01 -4.90965933e-01 -5.60549498e-01 8.23206604e-02 -1.48284629e-01 -1.46020472e-01 -1.07533097e+00 1.45892948e-01 4.70314873e-03 -7.55739585e-02 9.42192197e-01 3.98877978e-01 1.07477880e+00 1.96258664e-01 -5.66435039e-01 3.15807462e-01 1.45886445e+00 2.52643917e-02 1.47759587e-01 7.47901201e-01 5.87974548e-01 6.48352504e-01 8.07283759e-01 5.98021805e-01 2.46321276e-01 7.20790625e-01 8.32838595e-01 -5.48064113e-01 -1.05200328e-01 8.89400244e-02 -1.41637996e-02 7.43014812e-02 -2.96406746e-01 1.75721303e-01 -9.44733679e-01 4.07326698e-01 -1.75868177e+00 -5.59942245e-01 -2.52608895e-01 2.08117318e+00 6.24593258e-01 7.71025360e-01 4.45829004e-01 4.24244136e-01 5.84104478e-01 3.74683104e-02 -3.50638807e-01 -2.60664970e-01 1.15244031e-01 6.34459406e-02 1.10705888e+00 3.16108108e-01 -1.36128056e+00 7.24464476e-01 6.93221474e+00 3.64317358e-01 -1.02176237e+00 -2.34123230e-01 3.65709484e-01 -4.15392131e-01 4.54235315e-01 -1.79250576e-02 -1.32898545e+00 5.11483140e-02 1.96343765e-01 6.62158906e-01 7.69030750e-02 1.24562597e+00 8.85648057e-02 -6.60530746e-01 -1.32752872e+00 1.14334965e+00 3.84077013e-01 -1.16168380e+00 -4.40686852e-01 -1.03910929e-02 4.19235110e-01 3.35680991e-01 -1.97097078e-01 -9.04923379e-02 -4.24741842e-02 -6.28590107e-01 1.11248744e+00 4.10987467e-01 3.56345773e-01 -5.23908019e-01 7.25064397e-01 6.65644526e-01 -1.12940168e+00 -2.45899893e-03 -7.66955793e-01 -4.22436088e-01 5.49042895e-02 9.31743145e-01 -1.37881207e+00 -8.40051174e-02 1.30716026e+00 3.48339260e-01 -9.41002369e-01 1.37510085e+00 -3.70928049e-01 1.96657315e-01 -8.78944933e-01 -5.00188589e-01 3.60715985e-01 1.23187058e-01 5.40502787e-01 1.52098536e+00 2.34811276e-01 1.36862993e-01 2.27157250e-01 7.99885213e-01 2.45311692e-01 -2.84838587e-01 -6.34083152e-01 3.80779296e-01 4.66368765e-01 1.28357255e+00 -1.29549718e+00 -3.59125406e-01 -4.45232242e-01 8.79783988e-01 2.59320080e-01 -1.91421583e-02 -4.71126050e-01 -5.83764136e-01 1.01157904e-01 1.75463870e-01 9.84139025e-01 -6.07564926e-01 -3.60011011e-01 -4.34892595e-01 1.53907984e-01 -3.01812410e-01 2.12424248e-01 -1.01503229e+00 -1.13065875e+00 5.66661656e-01 2.07492545e-01 -1.22375369e+00 -1.31759822e-01 -1.30467951e+00 -2.53115326e-01 7.15563595e-01 -1.59464741e+00 -1.08468437e+00 -4.72530663e-01 2.41839767e-01 4.12432522e-01 1.47618473e-01 7.45329678e-01 8.91591311e-02 -1.80630952e-01 -1.38273701e-01 -1.56170949e-01 5.73098809e-02 5.63175976e-01 -1.41508687e+00 6.16211951e-01 8.46919417e-01 5.24228692e-01 2.80832201e-01 6.06283188e-01 -4.45619315e-01 -1.61865461e+00 -6.28920674e-01 7.43649006e-01 -8.37401330e-01 4.23454762e-01 -7.20356941e-01 -5.41056633e-01 3.94051015e-01 -2.40959808e-01 2.75773525e-01 4.49060112e-01 2.05024034e-01 -3.29576582e-01 -1.78045705e-01 -1.19403005e+00 1.51183069e-01 1.08355021e+00 -4.13356572e-01 -7.58456826e-01 4.79672700e-01 5.35693824e-01 -7.89297998e-01 -5.51834643e-01 3.97516102e-01 5.46636164e-01 -1.26142883e+00 1.37881982e+00 -8.79271701e-02 -1.31643802e-01 -7.15050936e-01 -1.88296735e-01 -7.80992389e-01 -3.20994496e-01 -2.53494024e-01 -4.22996908e-01 6.04754031e-01 2.60217011e-01 -1.68637201e-01 1.03852785e+00 1.42047822e-01 -5.23328222e-02 -5.48266709e-01 -9.60179687e-01 -6.51703715e-01 -6.69606090e-01 -7.00761437e-01 1.67292744e-01 3.09082896e-01 -3.64589930e-01 1.28360346e-01 1.80948570e-01 5.07282615e-01 8.06884050e-01 5.61228991e-01 9.11046386e-01 -1.36755204e+00 -3.35569888e-01 -5.57646573e-01 -7.61281610e-01 -1.41416597e+00 -5.03491879e-01 -4.00276423e-01 2.64140844e-01 -1.39174652e+00 8.20526555e-02 -6.26709819e-01 -1.95183322e-01 5.27568996e-01 7.98069686e-02 8.85581076e-01 1.69052392e-01 1.39801249e-01 -8.48668754e-01 -2.72168778e-02 7.60009825e-01 2.28512362e-02 -2.39356771e-01 -1.54457828e-02 -3.17322761e-01 1.04944885e+00 5.96701801e-01 -6.02146804e-01 3.67496192e-01 -4.34429288e-01 1.24515951e-01 -5.18587172e-01 5.96619725e-01 -1.21747744e+00 4.88891542e-01 7.34181181e-02 9.15423453e-01 -1.60080719e+00 6.63904011e-01 -8.88112903e-01 -5.49268961e-01 5.93353868e-01 8.94288495e-02 -2.02328160e-01 4.29997683e-01 3.09111059e-01 9.28303376e-02 -3.92105043e-01 6.12025201e-01 -4.92088050e-01 -1.03372526e+00 -8.99246894e-03 -3.27649742e-01 -5.93167007e-01 1.12892282e+00 -6.43865466e-01 4.30294387e-02 1.66192874e-01 -5.37401974e-01 -9.12374407e-02 4.95374739e-01 2.16710523e-01 7.92492270e-01 -1.04509318e+00 -5.29887438e-01 4.75240409e-01 2.34875917e-01 4.20491248e-01 -1.80454612e-01 6.27475858e-01 -7.54322350e-01 6.75278962e-01 -2.13129818e-01 -1.31604195e+00 -1.34317529e+00 6.45240724e-01 1.30851686e-01 1.33547947e-01 -6.70214474e-01 1.12736094e+00 -8.42849016e-02 -3.51768643e-01 6.95582926e-01 -5.87566793e-01 2.22695228e-02 -1.00222386e-01 6.18402064e-01 3.81611317e-01 3.37809563e-01 -3.12953442e-01 -7.23746061e-01 1.00740123e+00 1.10765323e-01 -1.51439816e-01 1.43621624e+00 3.23616080e-02 1.88848120e-03 4.73846763e-01 7.73921490e-01 2.00103186e-02 -1.48655856e+00 -2.56222337e-01 1.06448382e-01 -8.63303244e-01 2.56217360e-01 -6.30428076e-01 -5.68133473e-01 9.10047889e-01 1.02476072e+00 5.97244680e-01 1.15076530e+00 5.65468848e-01 5.20991199e-02 7.63474643e-01 3.63155335e-01 -9.60524738e-01 2.72955328e-01 5.47494590e-01 6.65691793e-01 -1.48512566e+00 5.80610275e-01 -4.00865763e-01 -1.22934796e-01 1.33884621e+00 5.34690857e-01 -4.54246104e-01 4.77065295e-01 5.04714489e-01 -9.22643393e-02 -3.15848529e-01 -2.07600087e-01 -5.28691828e-01 4.11437690e-01 7.18633115e-01 1.41985923e-01 -1.21716037e-01 3.54365110e-01 -9.52171907e-02 1.19111888e-01 -4.26735044e-01 2.50562549e-01 1.27704108e+00 -9.39988375e-01 -9.16856885e-01 -1.10868633e+00 2.27372691e-01 -8.71804133e-02 2.02733591e-01 -6.63655519e-01 1.02158070e+00 3.62525672e-01 6.69463277e-01 4.45274353e-01 -1.10807441e-01 5.47954738e-01 -1.24459192e-01 9.25408781e-01 -7.96505570e-01 -3.50337833e-01 1.76928416e-01 -1.93982020e-01 -9.42856073e-01 -7.91683853e-01 -4.96972203e-01 -1.14523220e+00 8.56964365e-02 -5.11637092e-01 -4.20378506e-01 1.28433394e+00 6.65740609e-01 2.16444597e-01 2.92153835e-01 4.37589824e-01 -1.70801306e+00 -3.25772673e-01 -8.10705543e-01 -4.57746625e-01 -1.58379346e-01 4.46478277e-01 -8.03551614e-01 -3.49382669e-01 -1.03248566e-01]

[7.655828475952148, -2.611741304397583]

451eecc4-587c-4a22-a3c3-48e4b5da24e5

wavelet-based-unsupervised-label-to-image-1

2305.09647

https://arxiv.org/abs/2305.09647v1

https://arxiv.org/pdf/2305.09647v1.pdf

Wavelet-based Unsupervised Label-to-Image Translation

Semantic Image Synthesis (SIS) is a subclass of image-to-image translation where a semantic layout is used to generate a photorealistic image. State-of-the-art conditional Generative Adversarial Networks (GANs) need a huge amount of paired data to accomplish this task while generic unpaired image-to-image translation frameworks underperform in comparison, because they color-code semantic layouts and learn correspondences in appearance instead of semantic content. Starting from the assumption that a high quality generated image should be segmented back to its semantic layout, we propose a new Unsupervised paradigm for SIS (USIS) that makes use of a self-supervised segmentation loss and whole image wavelet based discrimination. Furthermore, in order to match the high-frequency distribution of real images, a novel generator architecture in the wavelet domain is proposed. We test our methodology on 3 challenging datasets and demonstrate its ability to bridge the performance gap between paired and unpaired models.

['Bin Yang', 'Shuai Zhang', 'Karim Armanious', 'Mohamed Abdelsamad', 'George Eskandar']

2023-05-16

wavelet-based-unsupervised-label-to-image

https://ieeexplore.ieee.org/document/9746759

https://arxiv.org/pdf/2109.14715.pdf

ieee-international-conference-on-acoustics-14

['unsupervised-image-to-image-translation', 'image-to-image-translation', 'multimodal-unsupervised-image-to-image', 'image-to-image-translation']

['computer-vision', 'computer-vision', 'computer-vision', 'miscellaneous']

[ 1.05743206e+00 3.67765665e-01 1.30433708e-01 -3.61742646e-01 -9.14825261e-01 -8.63859117e-01 8.55870068e-01 -3.39453638e-01 -2.39362732e-01 7.22004533e-01 -1.63390011e-01 -2.48432644e-02 1.42037392e-01 -1.09284222e+00 -1.05923676e+00 -7.67946005e-01 5.16451776e-01 5.75689495e-01 1.35126993e-01 -3.77405912e-01 2.82119345e-02 4.24364448e-01 -1.41365302e+00 2.57015586e-01 9.94734824e-01 8.57201159e-01 4.20035049e-02 6.77239001e-01 -2.17203811e-01 7.04218864e-01 -8.57018828e-01 -5.72915375e-01 6.44610465e-01 -1.20570111e+00 -7.23494947e-01 1.44788191e-01 7.18904674e-01 6.54732659e-02 -1.14887729e-01 1.27971971e+00 4.48015481e-01 -2.14996293e-01 9.00092542e-01 -1.53117478e+00 -8.47626567e-01 3.19702625e-01 -3.15515786e-01 -3.74834239e-01 2.40511328e-01 2.54641742e-01 7.53717005e-01 -6.38190746e-01 9.31834698e-01 1.10201800e+00 6.24574780e-01 8.01229656e-01 -1.71943104e+00 -5.33598125e-01 -4.25998598e-01 -1.14588641e-01 -1.08559239e+00 -2.55770862e-01 1.17957604e+00 -4.07890290e-01 2.37175032e-01 2.70027578e-01 6.27027690e-01 1.54178357e+00 2.56131236e-02 5.21105409e-01 1.68031275e+00 -6.14200890e-01 4.28096026e-01 2.71464765e-01 -7.48833477e-01 5.99227548e-01 -3.60277817e-02 4.42152530e-01 -4.31668073e-01 2.50432342e-01 9.05301392e-01 -3.19574177e-01 -2.02520221e-01 -5.99018872e-01 -1.07080293e+00 9.47649658e-01 5.58414578e-01 3.91449392e-01 -1.11639760e-01 3.35289508e-01 9.36413109e-02 3.50092858e-01 3.86313975e-01 6.87085211e-01 5.84911630e-02 2.41554096e-01 -1.29002345e+00 2.80254185e-01 6.54188395e-01 9.05337811e-01 8.68277550e-01 5.44375896e-01 -3.74821752e-01 6.59435570e-01 -1.60871089e-01 5.81151187e-01 4.94914472e-01 -9.64509368e-01 1.16387501e-01 4.78593141e-01 -1.78101450e-01 -8.24409306e-01 1.56984150e-01 -5.35595059e-01 -9.16707814e-01 6.06170952e-01 5.45439720e-01 7.50829056e-02 -1.21550250e+00 1.86159241e+00 1.79916874e-01 2.21129864e-01 2.98631132e-01 7.62325048e-01 6.14469469e-01 4.25930887e-01 6.50810972e-02 9.35158283e-02 1.06140244e+00 -8.31951022e-01 -3.93244535e-01 -2.26294577e-01 2.42861360e-02 -7.56805718e-01 1.25728226e+00 1.55517071e-01 -1.38231039e+00 -8.41418862e-01 -1.14152861e+00 -4.13534008e-02 -5.62707245e-01 6.47033602e-02 1.14208914e-01 8.31622601e-01 -1.17306733e+00 7.04504609e-01 -4.30319220e-01 -1.97002903e-01 6.38376415e-01 1.42668530e-01 -4.13823634e-01 8.95794854e-02 -1.09134781e+00 6.45603180e-01 4.45682496e-01 -3.23008090e-01 -9.53122199e-01 -9.72549021e-01 -8.59545052e-01 -7.60510787e-02 -5.12381271e-02 -9.74027336e-01 6.97344124e-01 -1.73840892e+00 -1.78947258e+00 1.41040587e+00 3.19768727e-01 -6.48957610e-01 9.85002637e-01 3.92248213e-01 -1.18452646e-01 5.53749919e-01 1.60518438e-01 1.07972360e+00 1.65517795e+00 -1.72506034e+00 -2.73090512e-01 -2.18767270e-01 -2.07570672e-01 -1.76082656e-01 -1.32799730e-01 -2.53913283e-01 -1.52148783e-01 -1.19193983e+00 -6.78120032e-02 -8.88630271e-01 -4.77254763e-02 1.62018370e-02 -5.15863657e-01 4.52153027e-01 9.13186193e-01 -6.91130459e-01 4.29080129e-01 -2.12024212e+00 4.95966524e-01 3.16868216e-01 1.05100341e-01 1.24335654e-01 -3.67568165e-01 4.06717449e-01 -3.35406244e-01 8.92597809e-02 -7.01757610e-01 -5.32730937e-01 3.74194495e-02 2.40153626e-01 -4.62325007e-01 3.50548625e-01 4.96735841e-01 1.25429404e+00 -8.05523932e-01 -5.00631869e-01 2.21284479e-01 5.21022916e-01 -5.78477204e-01 4.50584203e-01 -2.81243622e-01 9.18581545e-01 -4.76815999e-02 3.77819657e-01 6.99652851e-01 4.26114462e-02 -3.54443341e-02 -2.66869932e-01 2.78059751e-01 -3.50296021e-01 -6.60165489e-01 1.84490097e+00 -5.87811649e-01 5.48331499e-01 -6.74900040e-02 -1.34800076e+00 1.09893382e+00 9.48162749e-02 4.84997302e-01 -9.65950310e-01 1.98730752e-01 3.34994525e-01 -2.19096720e-01 -1.47081271e-01 1.62066534e-01 -5.03025770e-01 -1.20457001e-01 2.04390988e-01 3.38843137e-01 -9.60880160e-01 1.32835969e-01 9.37675610e-02 9.03497219e-01 4.74193603e-01 -1.76760778e-01 -2.98712999e-01 5.15893221e-01 1.40030131e-01 3.52429599e-01 7.20878899e-01 2.42971778e-01 1.34050369e+00 5.16897798e-01 -2.61948615e-01 -1.58327556e+00 -1.40343726e+00 1.67759404e-01 4.74149197e-01 1.27261141e-02 1.71823934e-01 -1.41339600e+00 -7.85600781e-01 -2.41414264e-01 8.55300725e-01 -8.80654633e-01 -4.42435861e-01 -4.06082481e-01 -4.20635134e-01 7.70226538e-01 3.37512821e-01 6.38263464e-01 -1.17204690e+00 -6.75763309e-01 9.97217149e-02 -1.77631244e-01 -1.37800658e+00 -3.52299690e-01 1.42063841e-01 -5.20192862e-01 -1.08434379e+00 -1.01493526e+00 -9.30283129e-01 9.67656791e-01 -1.42420113e-01 1.30324471e+00 -2.18974218e-01 -6.45796239e-01 5.66961467e-01 -4.55417156e-01 -3.25221539e-01 -1.03747022e+00 -1.36999324e-01 -3.09844255e-01 3.63828391e-01 -2.39110425e-01 -8.37704182e-01 -6.90202355e-01 1.56747296e-01 -1.49290574e+00 3.72403800e-01 6.56370997e-01 1.11366403e+00 5.61331391e-01 5.67591935e-02 3.87205094e-01 -1.02502775e+00 4.19378936e-01 -2.27151122e-02 -6.52096868e-01 2.85168767e-01 -4.66448277e-01 1.27239957e-01 8.69017363e-01 -4.74004626e-01 -1.01409602e+00 2.62258381e-01 -1.42033532e-01 -7.36871779e-01 -2.97305584e-01 -2.60292329e-02 -1.31407320e-01 -5.01919270e-01 7.64373779e-01 6.43656671e-01 3.24353933e-01 -1.47005230e-01 5.53474486e-01 1.30930319e-01 9.99220431e-01 -6.13593042e-01 1.40464485e+00 6.46001637e-01 1.83501869e-01 -5.90763330e-01 -5.30256808e-01 1.87983707e-01 -5.95544457e-01 -1.24432921e-01 1.17006004e+00 -7.83362329e-01 -3.69694293e-01 5.14003038e-01 -1.06360650e+00 -6.31471813e-01 -9.45371747e-01 -6.62606657e-02 -1.05452752e+00 2.86182135e-01 -4.11959559e-01 -4.87488031e-01 -2.33432382e-01 -1.09001219e+00 1.36053181e+00 6.82181716e-02 -5.35369106e-02 -9.28029239e-01 -2.93112006e-02 4.07527953e-01 5.06509423e-01 7.90951610e-01 1.07361257e+00 -3.03839058e-01 -3.63847673e-01 -4.35341336e-02 -1.98733881e-01 8.27714562e-01 -3.80025581e-02 -1.32644787e-01 -8.56624842e-01 -3.33170623e-01 3.76215950e-02 -3.73670787e-01 7.84586430e-01 2.05090493e-01 1.05618370e+00 -2.92145252e-01 9.43293124e-02 8.12098265e-01 1.78658330e+00 9.42341760e-02 1.05483675e+00 1.04869150e-01 7.17252672e-01 8.09441149e-01 1.29380226e-01 -8.38364195e-03 -1.01360574e-01 8.09694827e-01 4.10954952e-01 -6.12564623e-01 -7.89998233e-01 -7.54541874e-01 2.89757043e-01 2.70373911e-01 1.25160947e-01 -1.75718576e-01 -5.41856229e-01 3.79519314e-01 -1.37782872e+00 -8.50247502e-01 4.83052991e-02 2.15484071e+00 7.70044863e-01 9.14293304e-02 1.20948926e-01 3.79460722e-01 5.51317513e-01 -1.30370870e-01 -3.66221845e-01 -2.51864791e-01 -4.49865758e-01 8.69167984e-01 5.60236514e-01 3.01906943e-01 -9.05602813e-01 1.04671979e+00 6.19308949e+00 1.20999467e+00 -1.24548292e+00 1.96722835e-01 8.68918419e-01 3.36556047e-01 -5.18235326e-01 1.17795160e-02 -1.40808597e-01 6.58168495e-01 6.67326987e-01 1.98498800e-01 5.11774480e-01 6.38489723e-01 -1.86124677e-03 5.17435931e-03 -8.52367580e-01 9.73478198e-01 3.25863987e-01 -1.43539822e+00 3.41979086e-01 -7.32795447e-02 1.06573522e+00 -5.24890900e-01 3.55015188e-01 -1.10393673e-01 2.88995653e-01 -1.27613235e+00 1.10132229e+00 4.03473884e-01 1.20287776e+00 -6.14794850e-01 4.42975044e-01 -5.33417426e-03 -8.50193560e-01 6.99700862e-02 -6.19088262e-02 3.97506714e-01 9.48847011e-02 4.15100098e-01 -6.62058532e-01 7.01893628e-01 4.68686312e-01 3.99696827e-01 -7.44603932e-01 5.43381393e-01 -4.96569186e-01 4.47434127e-01 -2.97956020e-02 5.24808705e-01 2.69178927e-01 -5.82712710e-01 5.08701444e-01 8.63072157e-01 4.34937656e-01 -3.89416605e-01 -2.83738542e-02 1.39265931e+00 -6.26944527e-02 -2.27646483e-03 -6.83375180e-01 -1.53013527e-01 1.29979784e-02 9.87772644e-01 -9.91053998e-01 -2.05994412e-01 -1.37981502e-02 1.57346821e+00 -1.06489487e-01 4.37543929e-01 -8.97571385e-01 -3.15169066e-01 2.55279392e-01 2.03702599e-01 3.68722379e-01 -7.23761320e-02 -3.07900488e-01 -9.24103260e-01 -3.07872109e-02 -9.61822331e-01 -6.33492544e-02 -9.94124711e-01 -1.22893345e+00 7.19733477e-01 -2.43330047e-01 -1.39370060e+00 -3.01869184e-01 -4.00379539e-01 -6.84668481e-01 6.28055334e-01 -1.54061449e+00 -1.81862020e+00 -5.32352209e-01 6.88876510e-01 4.62879658e-01 -2.70251781e-01 9.20578182e-01 1.29204899e-01 -1.04813781e-02 6.72080576e-01 7.10617080e-02 3.14492524e-01 5.95916927e-01 -1.38406277e+00 4.27368641e-01 8.98344517e-01 3.68692636e-01 9.98051539e-02 8.82522821e-01 -3.71378094e-01 -1.23199189e+00 -1.16009641e+00 4.05794501e-01 -3.63566071e-01 2.73687035e-01 -5.74601054e-01 -4.95397329e-01 2.56565005e-01 3.86836708e-01 1.56068385e-01 3.53443056e-01 -7.59452999e-01 -5.05026102e-01 -1.47122666e-01 -1.45764875e+00 5.44654429e-01 1.04278183e+00 -5.16829491e-01 -3.59377414e-01 2.17732728e-01 6.34915471e-01 -1.91974014e-01 -7.65360892e-01 4.42481726e-01 3.90666664e-01 -1.23374104e+00 1.21398532e+00 -4.15027320e-01 8.17152321e-01 -4.40061986e-01 2.60615256e-02 -1.48724329e+00 1.34856507e-01 -8.26764643e-01 5.62102735e-01 1.24085164e+00 1.84170112e-01 -4.96594131e-01 9.48576927e-01 4.43042777e-02 -6.84881285e-02 -3.41283649e-01 -8.25959563e-01 -9.28758264e-01 1.90507457e-01 -4.68323231e-02 4.14292186e-01 7.94760108e-01 -5.55132210e-01 2.66486138e-01 -3.75122219e-01 -2.04944476e-01 8.76714170e-01 2.89525598e-01 8.83945704e-01 -9.83514845e-01 -4.43646342e-01 -5.36738753e-01 -6.28304839e-01 -4.06623155e-01 3.94237310e-01 -8.75143051e-01 -3.83022837e-02 -1.05115259e+00 -7.49469325e-02 -4.59003091e-01 1.17475517e-01 3.06903958e-01 2.21776426e-01 1.08384442e+00 3.86788666e-01 5.84067702e-02 -7.34529868e-02 7.85197914e-01 1.45041037e+00 -4.70416188e-01 8.98839980e-02 -1.88009158e-01 -3.91184866e-01 3.53332549e-01 6.67316914e-01 -5.60041904e-01 -5.07593989e-01 -1.05331197e-01 1.14403218e-01 1.34539884e-02 8.51425231e-01 -1.25856459e+00 -1.01972230e-01 1.08558081e-01 4.89032000e-01 5.05073294e-02 4.20453191e-01 -9.14700329e-01 5.44372261e-01 5.73474467e-01 -4.25170481e-01 -3.18750262e-01 -2.26992220e-02 5.74982643e-01 -5.60957909e-01 -1.05851896e-01 1.24467766e+00 -2.28932574e-01 -3.53426099e-01 1.44774228e-01 -2.13741902e-02 2.45052695e-01 1.14649141e+00 -5.59471130e-01 3.07437102e-03 -4.66777176e-01 -6.14755094e-01 -4.71856564e-01 8.60989511e-01 3.92359048e-01 5.91977537e-01 -1.32550991e+00 -7.58235693e-01 4.58819270e-01 1.93506345e-01 -1.73659235e-01 2.79194236e-01 6.11104071e-01 -6.78801358e-01 2.20946763e-02 -5.85241020e-01 -6.04059160e-01 -9.88916934e-01 5.51841319e-01 2.95732558e-01 -9.88383144e-02 -5.71526349e-01 7.10132897e-01 3.81918371e-01 -3.90186459e-01 -2.05240965e-01 5.61144724e-02 2.12304175e-01 -1.03003554e-01 1.16519861e-01 -7.76422620e-02 8.55670497e-02 -7.04430044e-01 1.38612226e-01 7.82674789e-01 5.35998225e-01 -2.80138701e-01 1.17468631e+00 2.42114469e-01 -1.27868131e-01 -8.84341076e-02 1.37099445e+00 3.15084644e-02 -1.45444310e+00 5.17279766e-02 -3.55632395e-01 -5.90268493e-01 -1.63475201e-01 -7.44817555e-01 -1.34727454e+00 8.35637331e-01 7.10025907e-01 1.98558748e-01 1.36588752e+00 -2.24723414e-01 8.24763238e-01 -3.75206470e-01 4.38851416e-01 -1.07413435e+00 4.14114654e-01 -1.26372874e-01 9.82922614e-01 -1.12112546e+00 -3.39321554e-01 -5.94117165e-01 -6.50353014e-01 1.05445504e+00 2.05127656e-01 -4.89164889e-01 1.75632581e-01 2.89066881e-01 3.36809531e-02 -5.00625111e-02 -1.29826386e-02 -3.86592954e-01 2.62442946e-01 9.53923643e-01 -7.89584145e-02 -3.87616940e-02 -1.20218188e-01 1.07639112e-01 -3.79849672e-01 -2.01329291e-01 3.52410078e-01 6.36849940e-01 2.34328777e-01 -1.42909682e+00 -3.32440078e-01 -1.85505338e-02 -3.48814875e-01 -5.39819635e-02 -5.50785422e-01 7.57523298e-01 2.16271490e-01 7.83189178e-01 -1.75004955e-02 -2.97605157e-01 8.09943676e-02 1.08230226e-01 8.04724574e-01 -4.10468578e-01 -6.86904728e-01 3.77294533e-02 -3.83500606e-01 -5.85654378e-01 -5.33526599e-01 -3.17467630e-01 -8.31161082e-01 -4.80227247e-02 1.49004027e-01 7.58402701e-03 8.49798918e-01 7.33240902e-01 2.70901114e-01 7.37367034e-01 7.25444257e-01 -8.84247482e-01 -2.77208060e-01 -6.03915930e-01 -5.82557499e-01 1.02084064e+00 1.02529943e-01 -5.44022977e-01 -4.00476247e-01 5.29464543e-01]

[11.697796821594238, -0.46125540137290955]

5ec2ff1c-aca3-41ac-842a-e2fa855539dc

robust-neural-circuit-reconstruction-from

1811.11356

https://arxiv.org/abs/1811.11356v4

https://arxiv.org/pdf/1811.11356v4.pdf

Robust neural circuit reconstruction from serial electron microscopy with convolutional recurrent networks

Recent successes in deep learning have started to impact neuroscience. Of particular significance are claims that current segmentation algorithms achieve "super-human" accuracy in an area known as connectomics. However, as we will show, these algorithms do not effectively generalize beyond the particular source and brain tissues used for training -- severely limiting their usability by the broader neuroscience community. To fill this gap, we describe a novel connectomics challenge for source- and tissue-agnostic reconstruction of neurons (STAR), which favors broad generalization over fitting specific datasets. We first demonstrate that current state-of-the-art approaches to neuron segmentation perform poorly on the challenge. We further describe a novel convolutional recurrent neural network module that combines short-range horizontal connections within a processing stage and long-range top-down connections between stages. The resulting architecture establishes the state of the art on the STAR challenge and represents a significant step towards widely usable and fully-automated connectomics analysis.

['David Berson', 'Junkyung Kim', 'Drew Linsley', 'Thomas Serre']

2018-11-28

['contour-detection']

['computer-vision']

[ 2.15802312e-01 9.53949690e-02 7.12057576e-02 -3.22947472e-01 -5.11069834e-01 -5.88877320e-01 3.56959671e-01 9.28518772e-02 -6.49847627e-01 5.82598269e-01 -2.73406208e-02 -3.53816867e-01 1.13577796e-02 -2.73897827e-01 -7.91685581e-01 -5.39349914e-01 -1.30872890e-01 5.12345135e-01 1.31530121e-01 -1.34130001e-01 2.16272160e-01 7.94629872e-01 -1.04395139e+00 3.93910259e-01 5.19518077e-01 7.74642527e-01 -8.29002075e-03 4.49603558e-01 -1.94646761e-01 2.72197485e-01 -3.61705571e-01 -3.30732971e-01 2.96027571e-01 -4.63622242e-01 -1.08206165e+00 -2.87931412e-01 6.28847420e-01 -1.20831050e-01 -1.63169622e-01 9.27632511e-01 6.69418037e-01 -2.88434714e-01 5.22303343e-01 -9.81053531e-01 -4.85417932e-01 9.04660702e-01 -4.97381777e-01 8.03742886e-01 -2.55313307e-01 2.89848357e-01 8.57262850e-01 -4.24697548e-01 9.75161076e-01 8.89756501e-01 1.12856615e+00 8.61077249e-01 -1.72169530e+00 -7.05748022e-01 4.63417292e-01 -2.63368636e-01 -1.04198444e+00 -5.44822097e-01 3.73078704e-01 -7.86164224e-01 1.36239946e+00 -9.80233550e-02 1.17230451e+00 1.25153530e+00 2.62818813e-01 7.08264530e-01 1.21316040e+00 1.93538681e-01 1.68737054e-01 -4.12495106e-01 3.04074407e-01 6.29327297e-01 4.08671498e-01 -3.35914999e-01 -1.83587357e-01 2.27439001e-01 1.18909562e+00 -9.02327523e-02 -3.18146758e-02 -2.85853565e-01 -1.57954323e+00 4.75289971e-01 7.45662749e-01 5.34487188e-01 -1.85556218e-01 4.83958513e-01 7.36930072e-01 4.94528450e-02 6.37425184e-01 5.92866838e-01 -6.99591696e-01 1.96898416e-01 -1.48561323e+00 3.38619232e-01 5.75798750e-01 7.24185705e-01 3.93109649e-01 1.00900792e-01 8.74252338e-03 8.21599960e-01 3.03645402e-01 -7.52497688e-02 5.57605565e-01 -1.07798505e+00 1.17517151e-01 4.95317906e-01 -4.30963844e-01 -5.76309025e-01 -1.10447562e+00 -9.42995787e-01 -7.90575802e-01 4.34730738e-01 8.42390478e-01 -2.65310198e-01 -9.26721632e-01 1.80624723e+00 1.80020615e-01 1.07466996e-01 -3.74790996e-01 9.11120415e-01 7.66764104e-01 -1.36083188e-02 3.93921137e-01 2.08772168e-01 1.25399458e+00 -7.69051254e-01 -2.13834479e-01 -3.81972790e-01 6.10502303e-01 -3.83058459e-01 8.21147442e-01 4.65757400e-01 -1.33672905e+00 -1.74256101e-01 -1.04739344e+00 -3.55152011e-01 -6.89185798e-01 -1.12527944e-02 9.79384124e-01 4.56556886e-01 -1.42192900e+00 9.26970065e-01 -1.03981698e+00 -7.57892668e-01 1.25367951e+00 6.08019888e-01 -4.73949105e-01 4.02428180e-01 -6.00228131e-01 9.50986624e-01 8.70266780e-02 1.21443503e-01 -1.03383493e+00 -1.23135436e+00 -3.54173779e-01 1.19523071e-02 -6.02425039e-02 -1.17687702e+00 1.17570937e+00 -8.71371448e-01 -1.12626112e+00 1.60237586e+00 6.86798394e-02 -8.15931082e-01 5.65985501e-01 -1.23434864e-01 6.77951500e-02 2.83796221e-01 1.75478220e-01 1.23624027e+00 4.74423259e-01 -1.05107164e+00 -1.89698502e-01 -6.59211099e-01 -2.95381278e-01 -1.14601783e-01 -3.54112382e-03 2.68847078e-01 -2.41785988e-01 -6.23402297e-01 3.11138183e-01 -7.05664635e-01 -5.65723002e-01 3.95323753e-01 -6.14135742e-01 6.77668825e-02 3.07441115e-01 -6.28940821e-01 5.22257745e-01 -1.72724330e+00 2.20375881e-01 -1.35155125e-02 7.54601717e-01 1.77336767e-01 -2.09459737e-01 1.09523378e-01 -4.41032499e-01 4.65764314e-01 -3.64977479e-01 -4.03930187e-01 -2.16396168e-01 -1.27587765e-01 -5.56035899e-02 7.87376702e-01 2.76909351e-01 1.16374052e+00 -8.73511493e-01 -3.80823582e-01 3.32704931e-02 5.32740355e-01 -3.77737999e-01 -3.33666325e-01 -1.70777991e-01 6.75860167e-01 -3.14305037e-01 5.28698146e-01 4.52782750e-01 -3.45757872e-01 5.14906049e-02 -1.86281204e-01 -2.64170915e-01 2.61335403e-01 -3.59856993e-01 2.29596949e+00 -6.84378073e-02 7.08428562e-01 6.15707040e-01 -1.34629440e+00 5.57975113e-01 2.21187159e-01 8.93414259e-01 -3.66131455e-01 4.34041500e-01 2.92010754e-01 3.79368812e-01 -2.89810270e-01 -2.09032953e-01 -2.47419789e-01 3.54488045e-01 3.62732738e-01 6.01492584e-01 -1.21259294e-01 1.67968094e-01 1.64133593e-01 1.16774881e+00 5.16434014e-01 3.15306000e-02 -6.36062562e-01 3.19264606e-02 1.84410334e-01 5.61944962e-01 4.20369625e-01 -5.88761508e-01 8.70762289e-01 6.31223619e-01 -3.51483971e-01 -1.41061509e+00 -9.99568403e-01 -2.13064358e-01 9.66119230e-01 -3.90553325e-01 4.97913398e-02 -1.34013760e+00 -5.77966869e-01 -1.04242884e-01 -3.02426517e-04 -8.72198343e-01 3.09064776e-01 -4.56549048e-01 -1.06534040e+00 1.09441948e+00 5.71914971e-01 3.40586543e-01 -1.10985720e+00 -7.49214888e-01 3.77622157e-01 2.45167628e-01 -1.16300547e+00 -7.84335509e-02 4.84122157e-01 -1.23322594e+00 -1.08405626e+00 -1.11205971e+00 -8.56645405e-01 5.98772943e-01 5.75809367e-02 1.06846046e+00 7.59654194e-02 -6.24799430e-01 7.02601150e-02 3.44581641e-02 -3.37953418e-01 1.43559754e-01 4.99442339e-01 -1.67892978e-01 -4.26674515e-01 2.73911804e-01 -1.01888216e+00 -8.27998936e-01 -4.48511429e-02 -6.39158726e-01 1.42011568e-01 4.49330419e-01 6.11872911e-01 7.83915758e-01 -6.89545572e-01 8.22117150e-01 -1.11119020e+00 5.80574214e-01 -5.42502522e-01 -4.10057306e-01 2.63278503e-02 -3.82948190e-01 -1.87541559e-01 5.34776449e-01 -2.35433653e-01 -5.43223619e-01 5.24175353e-02 -5.91902077e-01 -2.72704214e-01 -3.70731503e-01 4.55170631e-01 8.35836828e-02 -3.70586753e-01 6.37965202e-01 -1.12389885e-01 1.32973894e-01 -4.87420678e-01 3.68000627e-01 -5.61001599e-02 7.87048340e-01 -5.61651886e-01 3.03670347e-01 7.25273609e-01 3.01480860e-01 -5.37353575e-01 -6.89545810e-01 -4.17933732e-01 -1.09405816e+00 -1.21787928e-01 1.13107777e+00 -7.02384412e-01 -6.15620077e-01 4.37923044e-01 -1.07478416e+00 -8.96236598e-01 -2.81344861e-01 2.58037180e-01 -6.88743412e-01 1.19910009e-01 -9.50798273e-01 -4.09769624e-01 -6.95184886e-01 -1.25174904e+00 9.79192257e-01 1.90826774e-01 -3.92849892e-01 -1.09458792e+00 1.41327128e-01 3.95130545e-01 4.33165431e-01 4.45071071e-01 9.18657601e-01 -7.79683411e-01 -3.56941313e-01 8.53034183e-02 -4.05647904e-01 2.15879530e-02 -3.72280508e-01 1.93218186e-01 -1.16320539e+00 1.24558611e-02 -3.25446576e-01 -4.42666531e-01 1.09081459e+00 8.34291875e-01 1.31714511e+00 2.69117922e-01 -5.97491205e-01 1.07206166e+00 1.25383794e+00 -3.32413346e-01 4.96237606e-01 4.16217297e-01 7.15276301e-01 8.61771584e-01 -3.28043133e-01 -1.44494295e-01 2.08992958e-01 3.31769496e-01 3.22880507e-01 -5.84672391e-01 -5.62474191e-01 2.68039927e-02 4.77786502e-03 6.47460163e-01 -1.76141813e-01 1.81604028e-01 -1.10743570e+00 7.21695483e-01 -1.66201544e+00 -9.47338283e-01 -3.06936294e-01 1.79214013e+00 8.24425876e-01 -9.12766904e-04 4.44574684e-01 -8.85246918e-02 5.06232262e-01 -1.77039549e-01 -9.47873354e-01 -4.37708884e-01 -1.24936998e-01 4.99011546e-01 5.38554311e-01 1.30427256e-01 -1.06414413e+00 1.30931628e+00 7.75346136e+00 6.87208652e-01 -1.24648869e+00 3.13848764e-01 9.44710314e-01 -2.57704407e-01 -2.10359260e-01 -2.98685104e-01 -5.84899545e-01 1.01785183e-01 9.72927392e-01 2.75000662e-01 6.98462784e-01 6.22774839e-01 3.75543684e-01 1.40025452e-01 -1.12111557e+00 8.77861679e-01 -1.77069023e-01 -1.51467073e+00 -2.80904680e-01 1.31218061e-01 5.20970523e-01 1.01615477e+00 8.72452036e-02 -2.32574344e-02 3.17414373e-01 -1.38455343e+00 6.91300094e-01 5.51033258e-01 8.12801540e-01 -4.13672179e-01 3.00433844e-01 1.32204160e-01 -8.63765299e-01 2.81493217e-01 -5.23012161e-01 1.52416229e-01 1.99231505e-01 6.08253717e-01 -5.58977544e-01 1.68250903e-01 8.02970111e-01 8.33374798e-01 -6.18811131e-01 1.44658136e+00 2.21084128e-03 5.52503645e-01 -3.50706100e-01 2.31666327e-01 2.15373039e-01 -1.87834978e-01 3.96985739e-01 1.65066087e+00 2.60832906e-02 -3.37260701e-02 -7.90825486e-02 1.43563235e+00 -4.48587537e-01 4.65858169e-03 -5.80432236e-01 -3.38443846e-01 7.22773448e-02 1.72228754e+00 -1.55837953e+00 -3.00522298e-01 -2.70940304e-01 7.64479339e-01 7.22792327e-01 4.69812483e-01 -4.46089357e-01 -5.86646311e-02 8.72430325e-01 -1.02566564e-02 2.27363214e-01 -3.94815445e-01 -1.13258684e+00 -1.21048772e+00 -5.23855507e-01 -6.92644417e-01 1.20009221e-01 -7.20358312e-01 -1.50816786e+00 4.15958852e-01 -3.29096675e-01 -4.16623861e-01 1.62711918e-01 -9.05769169e-01 -7.12885380e-01 9.06841099e-01 -1.27539623e+00 -1.29769754e+00 -6.59202859e-02 3.08143198e-01 3.94279957e-01 -2.13741735e-02 8.36976528e-01 3.09412837e-01 -9.35934305e-01 3.40266138e-01 -1.32348552e-01 1.99649408e-01 3.68048340e-01 -1.12298417e+00 9.00281131e-01 7.29881525e-01 1.93650767e-01 1.05205238e+00 5.55246174e-01 -6.70460403e-01 -1.02862990e+00 -1.00361502e+00 5.21101236e-01 -3.29181522e-01 8.85804594e-01 -6.55536592e-01 -6.68414056e-01 1.01986849e+00 1.94456816e-01 3.13927323e-01 7.51846790e-01 3.45976025e-01 -5.46845973e-01 1.36471480e-01 -1.33545637e+00 6.78239524e-01 1.32535112e+00 -4.82347727e-01 -4.45844889e-01 2.46670917e-01 3.84468317e-01 -1.30433813e-01 -9.70534086e-01 2.49533460e-01 7.68612862e-01 -8.82494926e-01 1.11986864e+00 -8.13907146e-01 4.43179756e-01 -9.75085795e-02 3.10259908e-01 -1.19491565e+00 -3.76970083e-01 -5.29326618e-01 1.37461796e-01 1.01492405e+00 6.61290288e-01 -4.90867734e-01 1.09471405e+00 5.33630729e-01 -5.78633904e-01 -1.16975081e+00 -9.23129499e-01 -4.77122158e-01 8.39120686e-01 -4.24367100e-01 2.66335845e-01 9.46698666e-01 2.83963293e-01 4.69269991e-01 2.41668895e-01 -4.16999102e-01 6.16106093e-01 -1.56331733e-01 4.60177898e-01 -1.33184123e+00 -7.76779354e-02 -1.20330667e+00 -4.74595785e-01 -8.94884527e-01 8.55215788e-02 -1.26752245e+00 -6.46596923e-02 -1.81200516e+00 3.50474030e-01 -2.65345275e-01 -3.25907856e-01 6.18669987e-01 1.25078157e-01 7.11224377e-01 1.42677082e-02 1.73377991e-01 -4.19122845e-01 6.88906983e-02 1.29993618e+00 4.03592773e-02 5.34880422e-02 -3.58274072e-01 -9.96242106e-01 7.52175629e-01 1.05671012e+00 -2.86131293e-01 -2.51368523e-01 -6.59656763e-01 2.79159278e-01 -4.25198585e-01 5.14130771e-01 -1.14250147e+00 1.21808521e-01 2.66716361e-01 8.08305264e-01 -3.06302279e-01 2.53869087e-01 -3.91402841e-01 -1.44722998e-01 4.10213262e-01 -6.01878464e-01 -3.82879041e-02 4.15144652e-01 1.11529194e-01 3.44429404e-01 -1.87366698e-02 8.48238468e-01 -5.61190248e-01 -2.32061252e-01 5.52765429e-01 -5.80532134e-01 1.77532986e-01 6.87956035e-01 -3.32976252e-01 -5.01791239e-01 2.28372514e-02 -9.91466522e-01 9.69441533e-02 5.63059390e-01 -2.00573757e-01 2.87446171e-01 -7.17736185e-01 -6.67387307e-01 -1.40867725e-01 -3.44694853e-01 -9.96971969e-03 2.10607871e-01 1.23177445e+00 -6.66285932e-01 3.58093500e-01 -6.30476773e-01 -4.73952085e-01 -8.53958905e-01 3.86053175e-01 6.64238334e-01 -1.62270680e-01 -8.44723642e-01 1.21385610e+00 2.35179305e-01 -5.35339177e-01 -4.42505209e-03 -6.56158984e-01 -2.34916031e-01 -1.74769908e-02 2.63997644e-01 3.05261493e-01 9.99186411e-02 -5.62748194e-01 -3.04689109e-01 4.86722231e-01 -8.70726034e-02 -2.17738286e-01 1.61646175e+00 -9.36169550e-02 -3.16719234e-01 6.37486458e-01 1.06927979e+00 -3.93496037e-01 -1.44120550e+00 2.65758514e-01 2.96101198e-02 2.86769092e-01 8.78797770e-02 -1.13309824e+00 -1.37590265e+00 1.35526955e+00 4.38391894e-01 8.18656459e-02 7.29111493e-01 -4.38758135e-02 9.53035057e-01 1.06426954e-01 2.45639130e-01 -9.58585918e-01 -3.48146468e-01 5.32558501e-01 6.29064023e-01 -9.03573036e-01 1.00490041e-02 -3.35902065e-01 -3.12976003e-01 1.18053138e+00 5.88119030e-01 -4.77085501e-01 5.28280020e-01 4.48962122e-01 1.79636240e-01 -6.50717437e-01 -4.28453475e-01 -2.95645952e-01 1.92884162e-01 8.39446485e-01 9.49184716e-01 -1.66278899e-01 -2.15170264e-01 6.91539109e-01 -3.12746495e-01 1.59761444e-01 2.20948696e-01 5.98873794e-01 -4.73915726e-01 -9.03954983e-01 7.85941109e-02 7.59390831e-01 -8.25005531e-01 -3.03258151e-01 -7.42063105e-01 7.16560841e-01 2.41843030e-01 5.87471664e-01 5.95014207e-02 -1.25707254e-01 -8.13849643e-02 2.76234299e-01 9.21230137e-01 -7.46827543e-01 -1.21176255e+00 4.75766152e-01 -4.77885501e-03 -8.51602137e-01 -3.12078029e-01 -7.94739246e-01 -1.58170986e+00 -2.36653686e-01 7.28410110e-02 -3.22174639e-01 8.77726853e-01 1.15576065e+00 4.48172957e-01 7.58268118e-01 -3.37957770e-01 -1.31381667e+00 -6.53835535e-02 -9.96450841e-01 -5.48708737e-01 1.32488936e-01 2.11027279e-01 -3.58374327e-01 8.13322589e-02 1.83734998e-01]

[14.250567436218262, -2.9973933696746826]