Datasets:

J0nasW

/

paperswithcode

like 1

49c0077b-0098-4edb-82f3-7119a13ea579

ji-yu-zhi-shi-jian-du-de-biao-qian-jiang-zao

https://aclanthology.org/2022.ccl-1.25

https://aclanthology.org/2022.ccl-1.25.pdf

基于知识监督的标签降噪实体对齐(Refined De-noising for Labeled Entity Alignment from Auxiliary Evidence Knowledge)

“大多数现有的实体对齐解决方案都依赖于干净的标记数据来训练模型,很少关注种子噪声。为了解决实体对齐中的噪声问题,本文提出了一个标签降噪框架,在实体对齐中注入辅助知识和附带监督,以纠正标记和引导过程中的种子错误。特别是,考虑到以前基于邻域嵌入方法的弱点,本文应用了一种新的对偶关系注意力匹配编码器来加速知识图谱的结构学习,同时使用辅助知识来弥补结构表征的不足。然后,通过对抗训练来执行弱监督标签降噪。对于误差累积的问题,本文进一步使用对齐精化模块来提高模型的性能。实验结果表明,所提的框架能够轻松应对含噪声环境下的实体对齐问题,在多个真实数据集上的对齐准确性和噪声辨别能力始终优于其他基线方法。”

['Ning Jing', 'Fenglong Su']

ccl-2022-10

['entity-alignment', 'entity-alignment']

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

[-8.16761732e-01 -9.26098645e-01 7.49491453e-01 4.34055597e-01 4.38172370e-01 -1.37447441e+00 1.12779683e-03 5.67168117e-01 2.12330818e-01 1.40728080e+00 1.96407542e-01 -1.58261657e-01 -3.49274665e-01 -1.19254446e+00 -9.24269632e-02 -1.20413029e+00 -1.48025244e-01 1.40303612e+00 5.36212623e-01 -2.71405488e-01 6.63659692e-01 8.91419768e-01 -1.06054056e+00 2.39965022e-01 1.04203236e+00 1.12642646e+00 1.09593475e+00 3.86866242e-01 1.20366253e-01 5.72375059e-01 -1.27864134e+00 2.89307207e-01 -3.02291185e-01 2.27874249e-01 -5.50142288e-01 8.81528437e-01 8.45646083e-01 -6.20068014e-01 -1.37716854e+00 1.15785897e+00 7.74713099e-01 1.63019374e-01 7.36638010e-01 -5.79105258e-01 -8.29679191e-01 5.64064145e-01 1.55330315e-01 1.19015336e-01 -1.44051313e-01 2.96575338e-01 6.73241854e-01 -1.11642206e+00 3.16965997e-01 1.47399294e+00 1.53180826e+00 6.00783348e-01 -1.46536386e+00 -1.28419280e+00 2.68603384e-01 -1.33159552e-02 -1.83446777e+00 4.78148192e-01 4.17869657e-01 -3.24687988e-01 1.04868031e+00 1.34119427e+00 1.53100765e+00 1.78388739e+00 4.80272830e-01 1.25166631e+00 1.55295753e+00 6.92697167e-01 3.65890503e-01 1.88880992e+00 -3.43574345e-01 1.06792882e-01 1.08094609e+00 -6.00545891e-02 7.33813224e-03 9.02023166e-02 1.98506162e-01 4.40617740e-01 -1.07629195e-01 4.22306955e-01 -9.17235017e-01 1.04090393e+00 2.75696367e-01 1.60434604e+00 -2.08167598e-01 -9.70000505e-01 -9.75847661e-01 4.58152503e-01 -1.53145581e-01 1.75778478e-01 4.06414986e-01 1.12551844e+00 -3.02201331e-01 -1.90944016e-01 1.02583122e+00 1.52647901e+00 9.09758627e-01 4.02699590e-01 -7.74721205e-02 7.39905238e-01 1.35148025e+00 1.24211323e+00 1.39558625e+00 -8.12281072e-01 -6.58058763e-01 1.11334550e+00 2.20828459e-01 -1.94695306e+00 -9.86292839e-01 -1.39259326e+00 -9.18136120e-01 -1.16713405e+00 -1.17370628e-01 -6.37894630e-01 -6.82623804e-01 1.33436632e+00 5.54721616e-02 -3.96242887e-01 5.37083149e-01 1.00700700e+00 8.62978816e-01 1.64621627e+00 -2.59889275e-01 -3.79041135e-02 1.67119467e+00 -6.70408905e-01 -7.26409554e-01 -3.45607579e-01 -1.23991024e+00 -1.69841528e+00 8.47845674e-01 -4.26583141e-01 -8.41856182e-01 -3.67469370e-01 -2.27842832e+00 6.57010674e-01 -6.05040908e-01 -1.25811279e-01 1.26609957e+00 1.19121206e+00 -1.69613743e+00 5.32144785e-01 -5.12968004e-01 -1.00694704e+00 1.08263738e-01 3.27132970e-01 1.32793292e-01 1.33230314e-01 -1.05121887e+00 1.15375519e+00 -5.47014654e-01 4.90287751e-01 -2.15163040e+00 -5.99849641e-01 -8.55410919e-02 -1.52446315e-01 8.71701539e-01 -8.60604286e-01 1.23960030e+00 -9.76830423e-01 -8.50538909e-01 6.82836652e-01 -9.09394622e-01 8.48927572e-02 7.58683443e-01 7.09826231e-01 -2.77886838e-01 3.97693098e-01 -4.27534431e-01 1.66270107e-01 6.85916781e-01 -2.51203537e+00 -5.82545161e-01 -1.60594010e+00 2.76186615e-01 -3.02408874e-01 4.11989480e-01 8.36940169e-01 3.48345429e-01 -2.54679561e-01 3.63351882e-01 -1.76416767e+00 -9.94028077e-02 -1.68138874e+00 -5.98730981e-01 2.26403475e-01 1.89354241e+00 1.82094604e-01 1.34692550e+00 -1.97111547e+00 -1.01601267e+00 7.53718734e-01 4.66191262e-01 1.08493721e+00 7.06159547e-02 2.10153890e+00 4.47242349e-01 1.12588799e+00 -1.25732458e+00 5.45880795e-01 -1.84747532e-01 5.21703064e-02 9.70779955e-01 3.34854834e-02 -7.81160966e-02 -1.68870598e-01 -3.15870196e-01 3.45734477e-01 6.50162280e-01 7.75077641e-01 6.59860373e-01 4.08662975e-01 7.49383271e-01 3.75901550e-01 -1.33610189e-01 1.28271234e+00 1.52791345e+00 -7.06110954e-01 9.20931637e-01 2.73290485e-01 -4.77920979e-01 -8.79501283e-01 -2.74013019e+00 6.11226559e-01 -1.40347064e-01 6.99250400e-01 7.23367929e-01 -9.20041502e-01 1.62823594e+00 1.40996671e+00 8.32312822e-01 -4.67907600e-02 5.12633100e-02 8.96166980e-01 2.50316024e-01 -6.52318656e-01 7.46769726e-01 -7.17768312e-01 9.40729380e-01 1.88970551e-01 -7.96241522e-01 -2.33197361e-02 5.62361479e-01 -1.64704487e-01 1.34700751e+00 -1.23238134e+00 -3.13748896e-01 -1.14032149e+00 1.48034167e+00 -4.02936667e-01 2.52515644e-01 3.42348576e-01 -7.12763846e-01 3.61604154e-01 -6.71990454e-01 4.18182015e-01 -6.34882092e-01 -1.75186169e+00 -3.78250122e-01 5.98204136e-01 -2.90311813e-01 -9.59949017e-01 -9.96147037e-01 -3.36238816e-02 -1.37669086e-01 4.34984744e-01 2.00135708e-01 1.47327542e-01 -8.21394801e-01 -1.26931977e+00 -4.43254203e-01 -2.42356539e-01 1.36501110e+00 -1.61508071e+00 -4.18992728e-01 4.92475815e-02 -4.78166252e-01 -7.82469928e-01 -5.02339065e-01 -3.32537830e-01 -1.97420394e+00 -1.53547752e+00 -2.20740419e-02 -1.51197553e+00 1.53307056e+00 1.98352218e-01 7.63646364e-01 7.46312082e-01 -6.00283921e-01 1.04311669e+00 4.42625016e-01 -9.02447045e-01 3.52419913e-01 -1.88255295e-01 1.97477326e-01 1.03755963e+00 6.44914329e-01 -5.12067378e-01 -1.41818428e+00 9.82386053e-01 -1.56598639e+00 -8.20267081e-01 3.90010566e-01 5.55884123e-01 7.22951353e-01 8.03560466e-02 -1.56611018e-02 -8.28417420e-01 1.23671985e+00 6.10211194e-01 -1.02739525e+00 7.95960307e-01 -1.19794083e+00 -8.42396736e-01 9.96617138e-01 -5.87417781e-01 -9.29883897e-01 -9.38781142e-01 -3.70175391e-01 1.39759287e-01 2.78779358e-01 -1.51664525e-01 -3.37137371e-01 2.40912393e-01 -4.95737642e-01 9.28725377e-02 5.31984746e-01 -6.08445466e-01 -3.60176980e-01 1.41751230e+00 2.07175851e-01 -1.15448534e+00 3.78976792e-01 5.85400283e-01 -2.26222184e-02 -5.55117548e-01 1.70014471e-01 -8.44845101e-02 -9.08068120e-01 7.14648068e-01 7.81364441e-01 -1.47047710e+00 -3.83765727e-01 1.06527126e+00 -3.88851702e-01 7.93242872e-01 -6.15466535e-01 1.67213547e+00 -2.83139497e-02 -8.88981670e-02 -4.80264813e-01 -1.10338557e+00 -4.37165648e-01 -2.52307892e+00 3.18181068e-01 2.20490146e+00 7.40932068e-03 -1.01102495e+00 1.54267522e-02 4.42862421e-01 1.11708927e+00 -5.65233268e-02 1.18814456e+00 -1.20124817e+00 -1.01526272e+00 -4.96972471e-01 2.93290555e-01 1.56039333e+00 4.17489171e-01 7.42227733e-01 4.78365235e-02 -2.29030192e-01 2.45318741e-01 8.68275404e-01 5.09632111e-01 6.49246871e-01 1.85844168e-01 -6.82117462e-01 -1.95114002e-01 2.42786720e-01 2.64157534e+00 7.49748230e-01 8.40551674e-01 1.63461530e+00 -9.20610189e-01 1.00673482e-01 1.24121979e-01 5.38549006e-01 -1.27172425e-01 1.05206323e+00 -1.10098004e-01 1.24669170e+00 -8.83545458e-01 -3.37825954e-01 8.42667878e-01 1.60385859e+00 2.40393192e-01 -9.73232388e-02 3.84920090e-01 6.33144617e-01 -2.58467340e+00 -1.83528411e+00 -6.75866827e-02 3.07786989e+00 -5.93200512e-03 -9.48736370e-01 -2.28415504e-01 5.17217040e-01 1.62498534e+00 -3.57165098e-01 -1.73119977e-01 -4.90855604e-01 -4.22611311e-02 1.04607284e+00 5.92645526e-01 2.47561038e-01 -9.80759203e-01 2.96646982e-01 6.28837585e-01 8.71248901e-01 -1.41396427e+00 -1.94175146e-03 -5.65778054e-02 -6.50103807e-01 -9.46919441e-01 3.97351623e-01 -9.03761029e-01 1.69632661e+00 1.10166705e+00 -6.96360230e-01 -1.52335271e-01 3.18833917e-01 9.31324005e-01 -2.19000876e-01 -9.76991355e-01 1.48199809e+00 1.83002666e-01 -1.53132594e+00 7.60220140e-02 2.87674874e-01 9.41711605e-01 -1.31563723e-01 -1.01512678e-01 7.02012897e-01 1.16161041e-01 -1.08823252e+00 5.29738367e-01 -1.15846843e-01 -3.99819106e-01 -1.27315032e+00 1.76576591e+00 8.98759291e-02 -1.67402363e+00 -8.48203778e-01 -1.52117148e-01 -2.38167390e-01 3.40611547e-01 2.35385939e-01 -5.70331395e-01 4.42875475e-01 4.02758539e-01 -1.79076076e-01 -1.21674991e+00 2.09901118e+00 -9.40854192e-01 -1.99411571e-01 -7.60320127e-01 -1.97325659e+00 1.19456851e+00 -7.83888280e-01 9.97213483e-01 1.16058707e+00 2.40221158e-01 6.07134640e-01 5.48184514e-01 6.71949804e-01 -5.15797615e-01 2.03872859e-01 -4.05375361e-01 4.14346457e-01 1.03957176e+00 1.87073934e+00 -1.22506952e+00 -6.28012478e-01 6.60639256e-02 -1.14718807e+00 -3.01601946e-01 -4.23532814e-01 -4.46157038e-01 -8.72020721e-01 4.01830703e-01 2.48012945e-01 -2.01561287e-01 -8.41516376e-01 3.53491545e-01 -4.20524716e-01 5.36167882e-02 -9.29092646e-01 -2.72787899e-01 -6.15671813e-01 -1.29876399e+00 2.47147098e-01 -2.05578253e-01 -1.43372798e+00 1.00325263e+00 -1.26455140e+00 -4.57793415e-01 2.92527318e-01 -1.24058768e-01 -6.99466288e-01 -8.59418213e-02 1.00811875e+00 9.89639640e-01 -1.46887526e-01 8.32795322e-01 1.03593731e+00 -7.11589396e-01 7.54128575e-01 1.57829356e+00 6.04155064e-01 2.68901050e-01 -1.54358196e+00 -1.00165260e+00 2.33856946e-01 4.75968391e-01 4.10307050e-01 -5.07966988e-02 -2.71405071e-01 -2.13862991e+00 -8.39584827e-01 7.84519255e-01 -4.18133467e-01 6.67507827e-01 -5.40282391e-03 -2.92549610e-01 1.32790411e+00 4.80139941e-01 -3.54261875e-01 8.58899891e-01 -2.64032930e-01 3.38887513e-01 -1.08832216e+00 -2.10478497e+00 1.28163564e+00 9.40898061e-01 1.00986868e-01 -6.89250112e-01 4.36761200e-01 -3.82893264e-01 -4.01498318e-01 -1.16926789e+00 8.89514208e-01 1.76304400e+00 -8.31110060e-01 6.98197365e-01 -1.04225218e+00 -5.56888461e-01 -1.70386150e-01 -9.98580039e-01 -1.40238106e+00 3.39947104e-01 -5.47692955e-01 1.04119623e+00 1.40980554e+00 9.71142113e-01 -1.26428258e+00 2.72273391e-01 1.36524963e+00 -5.92920303e-01 -6.66879475e-01 -2.03989005e+00 -1.33399618e+00 -1.65763199e-01 -6.91780671e-02 3.37994426e-01 1.54763721e-02 -2.73307443e-01 5.11116028e-01 -7.56562725e-02 1.56542704e-01 2.49553189e-01 -3.92953932e-01 6.34416640e-01 -9.75007832e-01 1.13098633e+00 -8.68841648e-01 -3.56698424e-01 -1.06856771e-01 -8.78936470e-01 -2.41026551e-01 -1.54601073e+00 -2.04256821e+00 3.84803981e-01 -4.35919970e-01 -3.49383354e-01 -8.73745203e-01 1.55863297e+00 2.73940951e-01 8.37386489e-01 9.16083813e-01 1.59440994e-01 -4.95370477e-01 1.46678925e+00 -1.10487056e+00 -3.08984131e-01 4.40773129e-01 -2.56721944e-01 1.14515185e+00 4.32227284e-01 -1.32919494e-02 -1.38022661e-01 -8.34607422e-01 9.55140516e-02 -1.04417086e+00 -1.18756592e+00 -1.38880491e+00 6.67251050e-01 2.14817137e-01 2.09869623e-01 -1.03412032e-01 5.58181226e-01 -9.95197415e-01 6.65116966e-01 6.04578972e-01 5.50985515e-01 8.59489501e-01 -3.35887700e-01 -4.84144121e-01 -9.73382294e-02 -1.84934568e+00 9.38541651e-01 -7.15263844e-01 -5.33515453e-01 -3.11937124e-01 -6.35815084e-01 -3.29931974e-01 1.86719239e+00 -1.84442699e-01 -6.77031279e-01 7.68112838e-01 -1.24299943e+00 -6.13527834e-01 3.73422891e-01 1.21914113e+00 1.47698712e+00 -1.19092536e+00 -2.11532998e+00 1.43594697e-01 -1.08374143e+00 -1.00134447e-01 6.97217405e-01 9.40717220e-01 -8.48757029e-01 6.58955216e-01 -3.23654056e-01 -6.53559029e-01 -2.26407623e+00 7.97428489e-01 -2.17828378e-01 -5.92348993e-01 -4.35929775e-01 5.80040753e-01 -1.12072265e+00 1.29292667e-01 1.72792524e-01 -2.17441112e-01 -7.46920884e-01 3.64870280e-01 1.09167325e+00 9.06449318e-01 4.26018015e-02 -1.15852427e+00 -9.59152997e-01 1.89605284e+00 -5.30388355e-01 -1.39455199e-01 3.94567192e-01 -6.82706296e-01 -4.95009959e-01 -2.74976015e-01 8.29358160e-01 1.50389886e+00 7.26430789e-02 8.15498352e-01 -4.29296374e-01 -7.61909604e-01 -1.10569918e+00 -1.69774413e+00 -9.43728030e-01 3.77745181e-01 1.02805996e+00 -4.26086277e-01 7.20092952e-01 -6.66233599e-01 1.44285858e+00 3.52392882e-01 1.58599591e+00 -2.21347475e+00 -7.56693065e-01 7.32090235e-01 1.29731321e+00 -1.51027882e+00 2.62553275e-01 -3.34701508e-01 1.54162124e-01 1.63281536e+00 7.89394438e-01 -5.47349989e-01 1.45340776e+00 1.03427507e-01 -2.90350169e-01 4.90610749e-02 -1.54046685e-01 -3.35987091e-01 -6.00062132e-01 1.03615820e+00 8.50824237e-01 2.02040017e-01 -1.58990741e+00 4.99644041e-01 -1.13065743e+00 3.92868102e-01 1.50498009e+00 8.89530420e-01 -8.27719986e-01 -5.80906928e-01 -1.34739137e+00 -3.74974042e-01 -6.94241405e-01 6.68261349e-01 2.31819406e-01 9.97509480e-01 6.60873175e-01 2.73563981e+00 -5.67901433e-01 -1.10317898e+00 1.89054787e+00 4.70726341e-02 7.71776915e-01 -6.31768629e-02 -1.54904735e+00 -4.92379636e-01 9.17495042e-02 7.51058519e-01 -6.44362509e-01 -2.59200782e-01 -7.73233771e-01 -1.10591209e+00 3.12893927e-01 2.14640188e+00 1.66953886e+00 4.06432182e-01 1.21037155e-01 1.00034988e+00 1.35577822e+00 -3.83346826e-01 -1.51543248e+00 -1.15981627e+00 -9.75615740e-01 -3.35386187e-01 -1.71085134e-01 -1.03463662e+00 -1.39814579e+00 -1.25942135e+00]

[-3.316093683242798, 6.907754421234131]

b4f637b4-b1fd-4dea-a580-8770d64d8de5

node-copying-a-random-graph-model-for

2208.02435

https://arxiv.org/abs/2208.02435v1

https://arxiv.org/pdf/2208.02435v1.pdf

Node Copying: A Random Graph Model for Effective Graph Sampling

There has been an increased interest in applying machine learning techniques on relational structured-data based on an observed graph. Often, this graph is not fully representative of the true relationship amongst nodes. In these settings, building a generative model conditioned on the observed graph allows to take the graph uncertainty into account. Various existing techniques either rely on restrictive assumptions, fail to preserve topological properties within the samples or are prohibitively expensive for larger graphs. In this work, we introduce the node copying model for constructing a distribution over graphs. Sampling of a random graph is carried out by replacing each node's neighbors by those of a randomly sampled similar node. The sampled graphs preserve key characteristics of the graph structure without explicitly targeting them. Additionally, sampling from this model is extremely simple and scales linearly with the nodes. We show the usefulness of the copying model in three tasks. First, in node classification, a Bayesian formulation based on node copying achieves higher accuracy in sparse data settings. Second, we employ our proposed model to mitigate the effect of adversarial attacks on the graph topology. Last, incorporation of the model in a recommendation system setting improves recall over state-of-the-art methods.

['Mark Coates', 'Yanhui Geng', 'Yingxue Zhang', 'Jianing Sun', 'Soumyasundar Pal', 'Florence Regol']

2022-08-04

['graph-sampling']

['graphs']

[ 2.31798500e-01 5.18402040e-01 -3.21625054e-01 -2.89456248e-01 -5.59922218e-01 -7.03330815e-01 7.40148127e-01 5.69779038e-01 5.75474054e-02 6.23516083e-01 1.02576040e-01 -1.61607489e-01 -2.98355490e-01 -1.34831989e+00 -1.05845666e+00 -7.01837122e-01 -3.17349374e-01 6.43330276e-01 4.73864436e-01 8.03129375e-02 6.55953884e-02 6.60164952e-01 -9.92965579e-01 -5.43354489e-02 5.97794712e-01 7.46932030e-01 -5.97047480e-03 5.78537047e-01 1.12537123e-01 7.26735413e-01 -5.97849071e-01 -7.85341442e-01 3.38666797e-01 -1.61153719e-01 -6.42175674e-01 1.33428693e-01 6.27012849e-01 -1.58244908e-01 -7.93576777e-01 1.32078981e+00 1.75112516e-01 5.77253215e-02 7.51526415e-01 -1.44915879e+00 -5.55666029e-01 1.08227205e+00 -4.95196402e-01 -1.06333174e-01 3.81267607e-01 -4.17982846e-01 1.39675272e+00 -3.01768452e-01 6.47954702e-01 1.30880618e+00 6.94330633e-01 5.15820496e-02 -1.67633665e+00 -5.73126197e-01 3.35879028e-01 3.03808618e-02 -1.64274955e+00 -7.61035457e-02 1.07556593e+00 -1.45982966e-01 3.43117505e-01 4.27235931e-01 4.50213611e-01 1.19318390e+00 2.04783499e-01 5.79800069e-01 9.23083901e-01 -3.62466276e-01 3.94773901e-01 2.41093710e-01 2.48082727e-01 7.18841791e-01 7.20455050e-01 2.95584239e-02 -4.37774718e-01 -5.63233435e-01 4.78133768e-01 2.40695029e-01 -3.20663542e-01 -8.27191830e-01 -6.67382836e-01 9.65316653e-01 6.93319440e-01 -2.61662100e-02 -2.08721563e-01 1.04583703e-01 2.41235584e-01 3.01357955e-01 5.65023065e-01 3.16417187e-01 -5.22377454e-02 4.26942348e-01 -1.03326011e+00 1.05754368e-01 1.19509745e+00 1.22641754e+00 8.86262953e-01 -3.43019336e-01 1.08016074e-01 4.38094616e-01 4.00449961e-01 4.59218293e-01 -1.95084035e-01 -6.30699635e-01 2.51559049e-01 5.42610586e-01 -2.73240924e-01 -1.54561269e+00 -2.10943446e-02 -5.09279430e-01 -1.09794080e+00 -9.63949487e-02 5.78607321e-01 2.91693956e-01 -8.71354818e-01 1.88339770e+00 5.78421056e-01 2.04507008e-01 -1.54857650e-01 5.07265806e-01 4.70594376e-01 2.75400311e-01 -1.62550107e-01 -6.19510785e-02 1.12102294e+00 -5.62017977e-01 -4.06454772e-01 -4.92265895e-02 4.14134681e-01 -4.26332325e-01 7.43704915e-01 4.78642315e-01 -6.41124845e-01 -2.42338985e-01 -1.02413487e+00 2.04777777e-01 -4.01301473e-01 -2.23803356e-01 6.65554166e-01 1.04073608e+00 -1.03278339e+00 7.90938556e-01 -1.01960933e+00 -4.65587735e-01 3.63260031e-01 2.98677355e-01 -5.43609619e-01 -4.48510647e-01 -1.13321149e+00 4.94421154e-01 3.01070601e-01 -7.12218732e-02 -8.61834168e-01 -4.57119048e-01 -9.12166476e-01 1.20369151e-01 9.08293545e-01 -5.22069335e-01 6.64032876e-01 -4.89197493e-01 -1.11356270e+00 4.39273924e-01 -9.93605629e-02 -7.57027447e-01 6.00176752e-01 -3.85543481e-02 -2.31635794e-01 2.90817052e-01 -1.33271098e-01 9.97902751e-02 1.05306983e+00 -1.39394319e+00 1.36203077e-02 -4.54935610e-01 3.35004836e-01 -1.77601889e-01 -2.61200905e-01 -5.87110996e-01 -5.66872537e-01 -6.32564843e-01 2.44990081e-01 -1.21556604e+00 -1.88969612e-01 2.30094269e-02 -9.03356731e-01 -4.02965881e-02 7.39434958e-01 -4.43472296e-01 1.26405025e+00 -2.03536153e+00 -2.04611849e-03 8.57811749e-01 5.81316352e-01 3.20637636e-02 1.05341077e-01 8.71485829e-01 1.54845119e-01 2.60817438e-01 -8.27427506e-02 -3.62465650e-01 7.01063275e-02 3.95393312e-01 -4.11392927e-01 7.47846365e-01 -1.20847858e-01 5.49208820e-01 -9.15387511e-01 -4.65364158e-01 1.11459613e-01 3.92390519e-01 -6.15627885e-01 1.42393872e-01 -2.83743262e-01 2.01262623e-01 -4.81461614e-01 4.16549057e-01 8.50404143e-01 -5.40210366e-01 6.16876245e-01 -3.48185360e-01 7.88659692e-01 4.25788641e-01 -1.45367932e+00 1.34515882e+00 -2.13695154e-01 6.47816062e-02 1.94507778e-01 -9.82866824e-01 1.02748728e+00 5.00800423e-02 2.38726676e-01 6.00396432e-02 -4.99321334e-02 -1.58540249e-01 -7.11948872e-02 1.20327502e-01 3.58943164e-01 5.79202026e-02 -1.34210721e-01 5.98225415e-01 5.06985150e-02 -1.19838133e-01 1.84684291e-01 9.16063428e-01 1.46616602e+00 -3.01132858e-01 3.90381962e-01 -2.50912249e-01 2.50757247e-01 -5.02785981e-01 3.89854521e-01 1.20606208e+00 1.94162309e-01 6.10047638e-01 8.56407285e-01 -1.56027704e-01 -7.89035857e-01 -1.28448474e+00 -2.14188062e-02 7.76444972e-01 1.44521683e-01 -7.43371069e-01 -6.31268978e-01 -1.06172729e+00 3.13870251e-01 5.79262435e-01 -7.88320184e-01 -3.97090495e-01 -2.02254772e-01 -5.59442163e-01 5.00915945e-01 3.11575145e-01 2.84491539e-01 -5.11047363e-01 8.66017789e-02 -1.40742790e-02 1.47098051e-02 -1.29457045e+00 -6.18321836e-01 -7.41057396e-02 -7.69666135e-01 -1.34289837e+00 -1.90145075e-02 -3.66393089e-01 9.65463758e-01 2.27705091e-01 1.21718454e+00 2.95334905e-01 1.78419724e-02 5.38574874e-01 -4.50720221e-01 -7.89634436e-02 -6.50241435e-01 4.92119849e-01 -4.07984620e-03 3.41962397e-01 2.60124892e-01 -9.03840959e-01 -2.12182477e-01 2.68187523e-01 -9.83168840e-01 -1.46781206e-01 5.77938497e-01 6.56881750e-01 5.11607528e-01 3.92444491e-01 3.46814573e-01 -1.61915362e+00 4.47163403e-01 -6.44532681e-01 -4.76661950e-01 3.75776261e-01 -7.09342003e-01 3.62834334e-01 7.28427052e-01 -6.23668551e-01 -6.45736516e-01 6.37978464e-02 2.69132704e-01 -4.99359071e-01 -1.11129144e-02 5.46433628e-01 -5.16884029e-01 -1.38294220e-01 5.39507687e-01 5.11922361e-03 1.83777884e-01 -3.99273723e-01 4.98487055e-01 3.76572758e-01 4.47543591e-01 -7.65457213e-01 1.12785649e+00 5.96114814e-01 5.98087609e-01 -8.09055448e-01 -8.50545943e-01 -2.68476248e-01 -6.62942827e-01 4.89954986e-02 1.32359803e-01 -6.51866913e-01 -5.22141635e-01 1.92129090e-01 -6.75324023e-01 -6.14615083e-02 -1.70908108e-01 4.76229578e-01 -1.02413863e-01 7.54086137e-01 -5.53664625e-01 -8.41621041e-01 -1.75509095e-01 -9.54176545e-01 8.91058981e-01 -1.42303899e-01 -2.16588050e-01 -1.28689349e+00 -3.66368480e-02 5.78524843e-02 1.40669897e-01 4.08466607e-01 8.32424998e-01 -1.08656871e+00 -8.67621064e-01 -7.42875993e-01 -1.43120959e-01 2.05061540e-01 2.89226502e-01 3.50519493e-02 -8.08010876e-01 -5.20603180e-01 -1.89385831e-01 -2.18031406e-01 8.58961463e-01 1.92086220e-01 1.18390620e+00 -4.79715556e-01 -6.36136413e-01 3.75867486e-01 1.26205194e+00 -4.97278392e-01 4.34703887e-01 -2.87136108e-01 8.97009134e-01 5.37233651e-01 2.84609944e-01 3.52439970e-01 3.64098608e-01 4.69064057e-01 5.47710061e-01 2.99732894e-01 -3.39010730e-02 -8.51003468e-01 1.42799959e-01 5.75967371e-01 2.47888759e-01 -5.28798044e-01 -6.96826458e-01 4.54041213e-01 -1.66597724e+00 -7.81075060e-01 -5.89718111e-02 2.63201714e+00 7.33892798e-01 4.43456829e-01 8.96329209e-02 1.08147234e-01 9.91295516e-01 3.99010032e-01 -4.89983171e-01 1.58212826e-01 9.58414748e-02 8.28738138e-02 8.20379913e-01 7.41175771e-01 -1.05450881e+00 6.66554928e-01 5.93545628e+00 8.84522617e-01 -8.05784762e-01 -1.00503869e-01 5.01820147e-01 1.04029052e-01 -5.16119540e-01 3.54694754e-01 -8.08575153e-01 4.86862808e-01 8.64279389e-01 -2.01328382e-01 5.65478623e-01 8.55971634e-01 -2.75201708e-01 -1.95224537e-03 -1.37101364e+00 4.89053935e-01 1.06686011e-01 -1.03280902e+00 1.91728055e-01 6.03410661e-01 4.89013553e-01 -9.93568450e-02 -3.69305573e-02 1.09753236e-01 7.59579778e-01 -1.08887899e+00 4.60061371e-01 5.19251406e-01 5.98975539e-01 -7.36954570e-01 6.05897665e-01 3.86235774e-01 -1.25300777e+00 3.14767838e-01 -4.38986510e-01 2.05579132e-01 -7.85331577e-02 1.04815149e+00 -1.09390271e+00 8.99057329e-01 3.93277287e-01 5.60980737e-01 -7.50848532e-01 7.85986602e-01 -3.84626359e-01 9.29939449e-01 -7.89374828e-01 -5.77748232e-02 -1.27626285e-01 -2.54414618e-01 7.47379661e-01 1.06742787e+00 3.27157825e-02 -2.95507014e-01 4.82945681e-01 8.06711853e-01 -4.27100003e-01 -1.31101504e-01 -8.40237558e-01 -1.60354733e-01 8.75406325e-01 1.34370029e+00 -8.58582735e-01 -1.72047615e-01 -2.88591295e-01 5.99652290e-01 6.97091222e-01 2.62186408e-01 -6.84681892e-01 -1.50584698e-01 2.63762087e-01 4.41512585e-01 4.83024240e-01 -2.07848564e-01 -5.03350534e-02 -1.03821516e+00 1.94728449e-01 -7.80073583e-01 5.52570462e-01 -2.17696548e-01 -1.64864683e+00 5.70057571e-01 2.21641317e-01 -9.63267207e-01 -2.25841060e-01 -2.75353879e-01 -5.15657246e-01 7.06413090e-01 -1.03964293e+00 -1.21192467e+00 -2.61377186e-01 5.30810297e-01 -1.41045511e-01 -3.13238688e-02 8.87940407e-01 5.64592406e-02 -3.65183175e-01 7.74722219e-01 9.81305689e-02 2.25015104e-01 6.82386100e-01 -1.48733199e+00 3.42530787e-01 1.06306958e+00 5.09532928e-01 8.54561031e-01 8.26349318e-01 -9.99321640e-01 -1.43485129e+00 -1.18467987e+00 5.84909201e-01 -4.24023867e-01 7.31486678e-01 -8.35011065e-01 -1.12633181e+00 9.38956141e-01 -2.33867690e-01 5.15396714e-01 4.51339394e-01 3.43208343e-01 -8.10184538e-01 -2.90164083e-01 -1.45899284e+00 2.90738493e-01 1.00272512e+00 -6.46108747e-01 -2.08270401e-01 3.30703288e-01 5.44734538e-01 -2.35563353e-01 -1.10461187e+00 1.87051266e-01 4.68537748e-01 -7.70625949e-01 9.50604320e-01 -5.14762580e-01 -2.87425388e-02 -2.73035437e-01 -2.75658816e-01 -1.24166703e+00 -2.72704512e-01 -8.37740242e-01 -5.03411829e-01 1.39133894e+00 3.02900136e-01 -7.45827734e-01 9.92213786e-01 7.19759941e-01 4.04394627e-01 -5.42186677e-01 -7.39481509e-01 -7.47347057e-01 -2.20431969e-01 -3.29863995e-01 6.67430103e-01 9.09633338e-01 -1.93686724e-01 3.57718766e-01 -4.94036466e-01 7.64278710e-01 1.04552126e+00 9.07825902e-02 9.53370810e-01 -1.51117623e+00 -7.29792953e-01 3.55000347e-02 -6.37296617e-01 -1.03379047e+00 1.72460020e-01 -9.36742127e-01 -3.21116537e-01 -1.29441154e+00 2.22341850e-01 -6.86539590e-01 -1.00407392e-01 1.74891725e-01 -1.17394634e-01 2.98632309e-02 -5.24291582e-02 2.84522980e-01 -6.45631373e-01 3.03632259e-01 9.49429631e-01 -2.02452880e-03 1.11814067e-01 3.57549667e-01 -7.54902899e-01 6.30336702e-01 6.55349612e-01 -7.44588256e-01 -6.99246466e-01 2.34511141e-02 2.34116539e-01 -4.26707789e-02 3.62760544e-01 -6.66250288e-01 4.35737908e-01 8.79909098e-02 1.51303738e-01 -4.58372086e-01 3.56701791e-01 -9.55519140e-01 4.90870178e-01 3.22359204e-01 -3.37429434e-01 -2.85554051e-01 -3.52972925e-01 1.32328117e+00 -5.26861148e-03 -3.06487262e-01 7.52565980e-01 -1.34687483e-01 4.23587039e-02 4.59105521e-01 1.61801860e-01 1.38993904e-01 9.15293276e-01 7.96493515e-02 -1.32889047e-01 -7.01723576e-01 -8.15013945e-01 -6.27993718e-02 5.67974210e-01 1.34516135e-01 3.75351965e-01 -1.13677824e+00 -5.10899305e-01 1.85568601e-01 3.68009210e-02 1.76328227e-01 -1.61416054e-01 7.69294918e-01 -3.13361973e-01 4.87678014e-02 3.86704326e-01 -4.84508902e-01 -1.35091460e+00 8.30059230e-01 1.44559950e-01 -6.75846457e-01 -6.02646172e-01 6.31235719e-01 8.10817629e-02 -5.40190518e-01 4.47770268e-01 -1.70202509e-01 9.60345790e-02 -1.66900963e-01 2.18603477e-01 3.70109767e-01 7.15171511e-04 -4.83518243e-01 -3.24261010e-01 9.79441404e-03 -4.03893858e-01 1.36273757e-01 1.23197925e+00 -1.15973353e-01 -1.44181892e-01 4.70186234e-01 1.12673330e+00 5.44473469e-01 -1.06137836e+00 -7.54496992e-01 -4.51731086e-02 -6.72743797e-01 1.32965818e-01 -4.10637051e-01 -1.16812074e+00 5.18702686e-01 -4.82416116e-02 8.28888595e-01 7.37449884e-01 1.84399709e-01 4.60204691e-01 3.38504761e-01 7.38141418e-01 -4.72717136e-01 -2.43852720e-01 1.15122935e-02 5.19162655e-01 -1.13268960e+00 4.65561390e-01 -9.77018297e-01 -3.27964425e-01 9.82439280e-01 2.76655734e-01 -4.29883957e-01 1.00435174e+00 1.22682303e-01 -5.68016946e-01 -2.14383751e-01 -6.61625504e-01 1.11151315e-01 3.80807519e-01 6.93875551e-01 9.78948101e-02 2.26452440e-01 7.56079182e-02 3.95967722e-01 -3.12866777e-01 -4.02697831e-01 6.59305573e-01 7.08385229e-01 -2.00385168e-01 -1.29238629e+00 -2.87417769e-01 8.18185210e-01 -5.62527537e-01 -6.87488541e-02 -5.66921413e-01 7.95362115e-01 -3.55829507e-01 9.00700092e-01 -7.17611387e-02 -5.16442716e-01 8.76732916e-02 -1.92559659e-01 5.10509014e-01 -7.51499653e-01 -1.19405143e-01 -8.53365138e-02 2.18477577e-01 -3.25366139e-01 -2.59206414e-01 -8.16780627e-01 -7.58482277e-01 -6.25792563e-01 -5.59594095e-01 3.54959309e-01 3.59996647e-01 7.86013603e-01 3.88038725e-01 1.94753841e-01 6.81774020e-01 -4.58255142e-01 -9.10820782e-01 -1.01508987e+00 -1.02316535e+00 4.78897601e-01 2.04118252e-01 -6.72423005e-01 -6.93104088e-01 -2.41728365e-01]

[6.97433614730835, 5.543204307556152]

0ab5c369-d664-4eda-985e-b42bd4bde5d6

adaptive-graph-based-feature-normalization

2207.11123

https://arxiv.org/abs/2207.11123v1

https://arxiv.org/pdf/2207.11123v1.pdf

Adaptive Graph-Based Feature Normalization for Facial Expression Recognition

Facial Expression Recognition (FER) suffers from data uncertainties caused by ambiguous facial images and annotators' subjectiveness, resulting in excursive semantic and feature covariate shifting problem. Existing works usually correct mislabeled data by estimating noise distribution, or guide network training with knowledge learned from clean data, neglecting the associative relations of expressions. In this work, we propose an Adaptive Graph-based Feature Normalization (AGFN) method to protect FER models from data uncertainties by normalizing feature distributions with the association of expressions. Specifically, we propose a Poisson graph generator to adaptively construct topological graphs for samples in each mini-batches via a sampling process, and correspondingly design a coordinate descent strategy to optimize proposed network. Our method outperforms state-of-the-art works with accuracies of 91.84% and 91.11% on the benchmark datasets FERPlus and RAF-DB, respectively, and when the percentage of mislabeled data increases (e.g., to 20%), our network surpasses existing works significantly by 3.38% and 4.52%.

['Yujie Xiong', 'Qingqing Wang', 'Yangtao Du']

2022-07-22

['facial-expression-recognition']

['computer-vision']

[ 2.17708528e-01 2.10110545e-01 8.22720379e-02 -7.99778521e-01 -2.28404805e-01 -3.20701033e-01 2.06641182e-01 -1.78340212e-01 -2.51849085e-01 8.22468817e-01 -6.44201115e-02 2.70565659e-01 -2.38684174e-02 -7.73847103e-01 -6.40644193e-01 -7.03712583e-01 4.73200642e-02 3.45825404e-01 -8.26278329e-02 -2.41185635e-01 2.09928788e-02 5.82640529e-01 -1.62918913e+00 -4.10007052e-02 9.58115816e-01 1.27244484e+00 -5.07225215e-01 9.86760259e-02 -4.31457460e-01 7.99960971e-01 -6.91718757e-01 -9.51792955e-01 9.64703634e-02 -3.59121054e-01 -4.57813680e-01 3.61900121e-01 3.99590969e-01 -5.20990677e-02 -3.53168398e-01 1.53318024e+00 5.63081920e-01 4.58888188e-02 7.11262107e-01 -1.70070934e+00 -8.84022593e-01 5.43381691e-01 -1.16296041e+00 -1.15177192e-01 -1.60479665e-01 -1.70142323e-01 4.32383478e-01 -1.05691314e+00 5.30776858e-01 1.51954484e+00 7.34045923e-01 8.38459849e-01 -1.17124879e+00 -1.18323743e+00 2.74441421e-01 2.91340530e-01 -1.97976017e+00 -6.57197714e-01 1.05217278e+00 -2.69044578e-01 1.80031791e-01 2.24461854e-01 3.51283342e-01 1.11781108e+00 -1.36558026e-01 4.41515923e-01 1.11365891e+00 -1.58344179e-01 3.15357268e-01 3.07915341e-02 1.07702287e-02 7.61243224e-01 4.33190078e-01 -3.02306622e-01 -6.41391814e-01 -3.34051579e-01 6.38196468e-01 -1.92520216e-01 -1.06830917e-01 -1.13250017e-01 -4.38632727e-01 5.55271924e-01 1.76514462e-01 3.31639801e-03 -4.21403229e-01 7.12235346e-02 4.35164750e-01 1.34632111e-01 7.08880186e-01 -6.84914067e-02 -3.64573568e-01 8.44341144e-02 -6.94083393e-01 -1.64716598e-02 5.24107516e-01 1.07646453e+00 8.91131163e-01 3.55369896e-01 -1.34747386e-01 1.02410579e+00 3.59253228e-01 6.11451745e-01 2.63100624e-01 -7.60423541e-01 1.83464125e-01 5.51070213e-01 -5.41902669e-02 -1.50370908e+00 -3.42792928e-01 -3.79380047e-01 -1.26332951e+00 -1.98327720e-01 3.96778673e-01 -4.00423646e-01 -1.08509362e+00 2.05892467e+00 4.52352881e-01 4.43017781e-01 -7.36091193e-03 9.97041643e-01 8.16492021e-01 3.36053014e-01 4.04472888e-01 -5.41694760e-01 1.21790802e+00 -4.72627163e-01 -9.49760914e-01 6.42732531e-02 3.53677392e-01 -5.99342346e-01 1.05549848e+00 6.35763824e-01 -7.00127542e-01 -3.20804536e-01 -7.73724616e-01 3.28152448e-01 -4.29927707e-02 4.34852064e-01 7.36906111e-01 8.18847537e-01 -8.91782761e-01 4.14015502e-01 -5.30601323e-01 -1.59197316e-01 9.89396930e-01 4.90125269e-01 -4.86448199e-01 -8.67825374e-02 -1.21175599e+00 3.38068724e-01 2.04491705e-01 3.97019118e-01 -6.90156698e-01 -7.40523636e-01 -7.36585617e-01 -1.07915707e-01 5.11066198e-01 -3.68136048e-01 7.90506423e-01 -1.38663018e+00 -1.61493349e+00 7.59231925e-01 -1.05226703e-01 -2.27741227e-01 5.46783149e-01 -2.61473227e-02 -6.20619595e-01 -3.02768834e-02 -2.64634579e-01 6.48995697e-01 9.75392401e-01 -1.21666908e+00 -2.32150868e-01 -7.64066935e-01 -3.06097746e-01 6.86362311e-02 -5.98861933e-01 5.06039411e-02 -3.85266036e-01 -5.62933266e-01 1.98374823e-01 -7.94641197e-01 -1.88674212e-01 1.45535201e-01 -4.68397409e-01 -2.96329349e-01 7.03753054e-01 -4.65732902e-01 1.26060545e+00 -2.23822236e+00 -1.76348209e-01 5.17904580e-01 4.49285448e-01 2.17734769e-01 -1.50002837e-01 -2.34990194e-01 -2.18914926e-01 3.82561594e-01 -1.74500167e-01 -3.73854756e-01 -8.07330534e-02 2.61661083e-01 -9.02558863e-03 5.35206676e-01 3.30558926e-01 5.94478309e-01 -6.85047448e-01 -6.62717283e-01 -2.01516181e-01 5.35301149e-01 -3.52467984e-01 3.56525145e-02 -1.03616364e-01 3.09638649e-01 -5.24311483e-01 8.00779343e-01 1.28076375e+00 -2.71055847e-01 2.93980479e-01 -6.07541382e-01 4.53300238e-01 -5.41593373e-01 -1.44698787e+00 1.38493979e+00 -4.35801409e-02 1.69903949e-01 1.75256625e-01 -1.03000069e+00 1.55657363e+00 1.15594938e-02 4.45575982e-01 -5.23572624e-01 5.61340630e-01 -2.45204959e-02 -2.30719879e-01 -4.38691914e-01 2.26428926e-01 -1.56968877e-01 4.41471972e-02 5.61602525e-02 1.91197678e-01 3.42040151e-01 -8.59404281e-02 8.62361044e-02 8.81264985e-01 -1.69398427e-01 -3.74164172e-02 -3.54308337e-01 7.59464324e-01 -5.32207251e-01 9.37942147e-01 3.44781280e-01 -2.34255612e-01 5.62587857e-01 9.14815426e-01 -5.29655516e-01 -8.02589893e-01 -8.76834393e-01 -3.34222019e-02 1.00364375e+00 1.68295309e-01 -3.29093248e-01 -9.68477249e-01 -9.26583588e-01 -6.66718232e-03 5.80703437e-01 -8.34761918e-01 -3.99926275e-01 -9.93841738e-02 -1.02600718e+00 7.52829671e-01 3.43731493e-01 7.83288419e-01 -7.23125637e-01 3.63965750e-01 -6.74243122e-02 4.34447974e-02 -1.12217057e+00 -3.13693315e-01 -2.82960594e-01 -5.41797459e-01 -1.08080876e+00 -4.76336211e-01 -5.13209462e-01 1.16040158e+00 -1.01951018e-01 8.86841536e-01 7.66405836e-02 -1.43141955e-01 9.17691365e-02 -2.41635591e-01 -3.99344385e-01 -7.41729811e-02 4.31151642e-03 2.20856518e-01 7.47061849e-01 5.95980763e-01 -8.20080519e-01 -4.84620005e-01 5.08372784e-01 -9.90364134e-01 -1.06239438e-01 4.05983150e-01 8.75609219e-01 7.69891858e-01 1.85866877e-02 7.30166495e-01 -1.01994419e+00 7.57412791e-01 -5.91226757e-01 -6.84392810e-01 4.27485377e-01 -8.63931119e-01 1.23695068e-01 5.62160611e-01 -5.25713563e-01 -1.20892024e+00 2.35589519e-01 1.21616542e-01 -9.13788974e-01 -1.95941865e-01 2.07330629e-01 -4.68770355e-01 -3.52386683e-01 8.14423323e-01 -8.42161030e-02 2.44159862e-01 -2.02329606e-01 3.54760259e-01 6.20336533e-01 5.38938582e-01 -7.46551394e-01 5.59717894e-01 3.30541909e-01 2.64192969e-01 -4.62973863e-01 -5.72796106e-01 2.08037734e-01 -3.44587266e-01 -4.41425055e-01 5.10291994e-01 -8.65915120e-01 -8.48314464e-01 7.25995541e-01 -1.05191553e+00 3.21448267e-01 -1.34888515e-01 2.58876652e-01 -1.27614409e-01 3.66733402e-01 -4.16028529e-01 -9.50558364e-01 -3.53638113e-01 -9.18688357e-01 9.75368440e-01 5.99764168e-01 1.19898997e-01 -4.79034662e-01 -3.77820015e-01 1.41800404e-01 4.15647417e-01 5.84135175e-01 4.68433470e-01 -7.39999831e-01 -2.01362327e-01 -1.12704262e-01 -5.19586563e-01 5.28465867e-01 2.81004757e-01 2.78809190e-01 -1.06916296e+00 1.35079086e-01 -2.00959608e-01 -4.89210844e-01 5.17998040e-01 5.95606677e-02 1.60689616e+00 -5.09747148e-01 -9.63888913e-02 6.06817961e-01 1.13508344e+00 -8.84643476e-03 7.43350625e-01 -2.16880441e-01 7.80387402e-01 7.36092806e-01 5.65096676e-01 8.52398217e-01 3.93791884e-01 1.92516774e-01 5.84985912e-01 4.93318886e-02 4.42549661e-02 -2.76876986e-01 -8.26175883e-02 5.67339540e-01 -4.30368632e-02 -2.41268262e-01 -8.12785268e-01 2.15327740e-01 -1.85526025e+00 -4.43660110e-01 2.41943188e-02 1.89362466e+00 9.87819195e-01 -2.60445904e-02 -1.24257244e-01 -1.35032907e-01 1.17937815e+00 9.03722569e-02 -7.42245018e-01 -6.67105317e-02 -3.64694804e-01 2.69503236e-01 5.76615453e-01 2.91831940e-01 -9.36888456e-01 1.11817646e+00 5.20305777e+00 1.22778690e+00 -1.15749049e+00 -7.08174380e-03 1.25251997e+00 9.98031627e-03 -3.50058824e-01 -2.19704598e-01 -6.80659354e-01 5.06294847e-01 6.34272277e-01 -3.14686745e-01 5.98972440e-01 9.94482756e-01 5.49435280e-02 2.11400554e-01 -5.78758240e-01 1.45409954e+00 3.75311047e-01 -9.14498031e-01 1.87409073e-01 -1.59142271e-01 6.09596193e-01 -4.51348901e-01 8.53415579e-02 6.58886209e-02 3.16406012e-01 -1.18268406e+00 4.36487138e-01 9.83442426e-01 8.94175351e-01 -1.02455771e+00 8.49537611e-01 1.11990906e-02 -1.02814150e+00 1.92944452e-01 -4.98318881e-01 1.93645716e-01 -9.25065055e-02 9.28235292e-01 -5.64023972e-01 5.44319391e-01 6.65514290e-01 4.81450438e-01 -6.87951624e-01 5.97859263e-01 -1.93512127e-01 7.34871566e-01 -5.88600278e-01 -1.63267374e-01 -2.81294942e-01 -2.29480997e-01 1.65679365e-01 7.54757106e-01 3.54716212e-01 3.22133154e-01 -2.21277736e-02 7.73600876e-01 -5.42040050e-01 3.85090083e-01 -3.45809877e-01 -4.10099030e-02 7.26948142e-01 1.52957392e+00 -5.39665997e-01 -1.90488309e-01 -5.04499897e-02 8.03948343e-01 4.02155519e-01 4.27710444e-01 -1.01457512e+00 -4.61354464e-01 5.20224750e-01 -1.34680811e-02 1.33580174e-02 1.51864262e-02 -1.95581540e-01 -9.87902701e-01 1.43522367e-01 -7.41188884e-01 3.69524032e-01 -9.12053764e-01 -1.60965085e+00 8.33515882e-01 -1.73412114e-01 -9.10767078e-01 8.01637322e-02 -4.08567339e-01 -2.78674781e-01 7.07516789e-01 -1.28845334e+00 -1.00367272e+00 -7.36407280e-01 6.87400997e-01 -1.04103103e-01 -1.92653239e-01 5.95555484e-01 5.14627397e-01 -7.47724533e-01 1.16937101e+00 -2.72918046e-01 2.76077151e-01 8.09445977e-01 -7.30419815e-01 -1.24843353e-02 5.41157901e-01 -6.16497509e-02 3.64066511e-01 6.00496292e-01 -5.94115317e-01 -1.25247550e+00 -1.39348650e+00 5.81987500e-01 4.68158200e-02 4.93997216e-01 -4.86800373e-01 -9.85715330e-01 4.13816541e-01 -1.37353063e-01 6.29047394e-01 6.56779230e-01 -1.34624839e-01 -4.78122324e-01 -7.03239501e-01 -1.59363270e+00 7.26600349e-01 1.45749056e+00 -1.85559809e-01 -3.89489322e-03 2.44903952e-01 5.69713473e-01 -5.77619255e-01 -9.42543864e-01 7.14981139e-01 4.02878761e-01 -7.31432974e-01 4.25043046e-01 -6.83688462e-01 2.33559478e-02 -5.05352676e-01 -1.45096868e-01 -1.19890630e+00 -3.06050450e-01 -8.71124744e-01 1.16548896e-01 1.60617423e+00 3.42725515e-01 -7.40117311e-01 9.56409752e-01 8.46980572e-01 2.13633075e-01 -7.43674815e-01 -9.39535677e-01 -4.34539586e-01 -3.92083436e-01 -3.86757582e-01 1.09217429e+00 1.09049201e+00 -3.89628023e-01 1.93238616e-01 -4.83295918e-01 3.02490473e-01 7.79612780e-01 -4.90954250e-01 7.72699773e-01 -1.24774432e+00 2.20986336e-01 -2.94091552e-01 -7.12083995e-01 -5.49739540e-01 4.19372857e-01 -6.30352259e-01 -1.23837031e-01 -7.56928205e-01 -7.20812902e-02 -4.13539052e-01 -3.73858809e-01 6.81241632e-01 -1.46418735e-01 2.79835582e-01 -9.06913206e-02 -3.04614045e-02 -7.83792138e-01 9.41331863e-01 1.12782562e+00 -5.61176576e-02 7.66179338e-02 -3.14018428e-01 -8.89421642e-01 8.15559328e-01 9.34951007e-01 -4.62496549e-01 -6.11034632e-01 -1.94392100e-01 2.71075606e-01 -1.96411252e-01 2.78577447e-01 -7.79895782e-01 1.65940642e-01 -2.42774904e-01 6.37497783e-01 -1.89956546e-01 1.31674027e-02 -7.50544429e-01 3.58736426e-01 -2.19962716e-01 -2.47643128e-01 -2.97491848e-02 2.71137655e-01 7.08015621e-01 -1.56601146e-01 3.58186029e-02 7.02908516e-01 2.25120872e-01 -5.39733827e-01 7.39954472e-01 -9.83220991e-03 2.34931663e-01 9.38025057e-01 -3.51917520e-02 -3.49776834e-01 -4.10013288e-01 -9.14367020e-01 2.60088027e-01 2.02425882e-01 4.10965681e-01 6.21053338e-01 -1.69623995e+00 -6.67794108e-01 3.57320517e-01 1.74606323e-01 9.05712619e-02 4.42403942e-01 6.81276500e-01 -2.65002251e-01 -3.68484616e-01 -1.22422822e-01 -5.20772815e-01 -1.28869915e+00 2.29546979e-01 4.99893397e-01 1.59785628e-01 6.39986843e-02 9.94609416e-01 1.52723892e-02 -3.55719835e-01 2.74548799e-01 -3.00879050e-02 -3.17894518e-01 5.54389730e-02 3.81366789e-01 4.32880461e-01 1.04171015e-01 -7.09461570e-01 -4.41185147e-01 6.21676743e-01 -1.65317327e-01 2.12132528e-01 1.12302375e+00 -8.39070529e-02 -2.80571491e-01 3.83123346e-02 1.11579061e+00 -1.39352486e-01 -1.14443099e+00 -3.33461374e-01 -2.32599720e-01 -4.84518945e-01 -1.58552065e-01 -7.44002163e-01 -1.51422715e+00 5.06039262e-01 7.52303064e-01 -4.34728675e-02 1.51579869e+00 7.12335948e-03 3.33928198e-01 2.03625962e-01 4.31930959e-01 -1.06369817e+00 9.13017429e-03 3.28583181e-01 8.32480073e-01 -9.66026247e-01 -7.79189989e-02 -7.03206003e-01 -8.86023581e-01 1.11375844e+00 1.03884649e+00 -1.79495439e-01 8.37240756e-01 2.34685406e-01 6.19615726e-02 -2.77978987e-01 -5.94899297e-01 1.83561668e-01 1.41146407e-01 6.80500209e-01 1.05274819e-01 6.66454956e-02 -4.18945611e-01 1.26940894e+00 -2.10681036e-01 3.38474691e-01 2.61963814e-01 5.34567356e-01 -2.95019090e-01 -8.73964548e-01 -2.56243318e-01 5.23068905e-01 -4.53042835e-01 3.40508744e-02 -4.19738352e-01 6.90154433e-01 3.52256745e-01 7.78526545e-01 1.22893438e-01 -7.40768671e-01 4.47956145e-01 6.79989979e-02 3.62821341e-01 -1.03850313e-01 -6.92812353e-02 -3.79353762e-02 1.65374115e-01 -5.89766383e-01 -5.19068420e-01 -6.12286806e-01 -1.30222952e+00 -4.05960739e-01 -4.76553857e-01 2.66264588e-01 4.34344828e-01 7.39715457e-01 7.99526334e-01 3.81416708e-01 8.65221620e-01 -2.42649451e-01 -5.71625710e-01 -9.52499628e-01 -8.01230431e-01 5.05111516e-01 -2.10035130e-01 -8.89048398e-01 -5.60651958e-01 5.54371998e-02]

[13.599101066589355, 1.6315535306930542]

feefe2dc-292b-4d1b-ba95-1c51d4781f8a

independent-sign-language-recognition-with-3d

2012.05698

https://arxiv.org/abs/2012.05698v1

https://arxiv.org/pdf/2012.05698v1.pdf

Independent Sign Language Recognition with 3D Body, Hands, and Face Reconstruction

Independent Sign Language Recognition is a complex visual recognition problem that combines several challenging tasks of Computer Vision due to the necessity to exploit and fuse information from hand gestures, body features and facial expressions. While many state-of-the-art works have managed to deeply elaborate on these features independently, to the best of our knowledge, no work has adequately combined all three information channels to efficiently recognize Sign Language. In this work, we employ SMPL-X, a contemporary parametric model that enables joint extraction of 3D body shape, face and hands information from a single image. We use this holistic 3D reconstruction for SLR, demonstrating that it leads to higher accuracy than recognition from raw RGB images and their optical flow fed into the state-of-the-art I3D-type network for 3D action recognition and from 2D Openpose skeletons fed into a Recurrent Neural Network. Finally, a set of experiments on the body, face and hand features showed that neglecting any of these, significantly reduces the classification accuracy, proving the importance of jointly modeling body shape, facial expression and hand pose for Sign Language Recognition.

['Petros Maragos', 'Georgios Pavlakos', 'Agelos Kratimenos']

2020-11-24

['3d-human-action-recognition', 'face-reconstruction']

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

[ 6.23266697e-02 -2.07238436e-01 -2.23348260e-01 -2.52048194e-01 -2.75334656e-01 -2.95440197e-01 6.24095619e-01 -1.01310742e+00 -6.34868860e-01 3.22853655e-01 3.86991292e-01 -8.33217427e-03 8.00823942e-02 -7.98151121e-02 -3.36912304e-01 -8.39424431e-01 1.43870279e-01 5.10266721e-01 2.94875801e-01 -2.01420084e-01 2.02158820e-02 9.67834413e-01 -2.04716396e+00 1.28913075e-02 2.43078381e-01 1.05994070e+00 -3.46561402e-01 7.47195899e-01 -1.52988389e-01 7.00291634e-01 -2.47318402e-01 -1.93264619e-01 3.60209972e-01 -5.55712044e-01 -5.80333173e-01 1.99296981e-01 7.92386174e-01 -7.90029168e-01 -3.92850101e-01 5.23831725e-01 9.28161740e-01 -1.39964119e-01 7.62322128e-01 -1.06620955e+00 -2.80934274e-01 9.30587500e-02 -4.19595480e-01 -2.07867011e-01 5.12598932e-01 4.96361434e-01 7.62233734e-01 -7.93520212e-01 9.51720536e-01 1.21109807e+00 5.63282669e-01 9.17286217e-01 -8.49673748e-01 -5.12106597e-01 2.20025480e-01 1.97038710e-01 -8.68872523e-01 -5.10745525e-01 9.91007984e-01 -5.08013964e-01 9.68381882e-01 1.09526522e-01 1.18930387e+00 1.35807645e+00 -5.07500827e-01 1.17183995e+00 1.35361385e+00 -7.09389508e-01 -2.13344797e-01 -2.19770536e-01 1.72312185e-01 1.00571573e+00 4.29567210e-02 3.48235697e-01 -8.21325660e-01 2.76929200e-01 9.83645380e-01 -1.63005784e-01 -3.45302612e-01 -6.79226875e-01 -1.07714450e+00 2.94785082e-01 1.71333820e-01 4.47194219e-01 -4.87741739e-01 3.96990061e-01 2.10241914e-01 3.25518876e-01 1.08179875e-01 -2.84215063e-01 -4.27597493e-01 -4.71705288e-01 -9.26677704e-01 1.00644037e-01 8.91153693e-01 4.92255181e-01 3.55683744e-01 1.57303974e-01 -1.09818242e-02 6.08020365e-01 8.33572268e-01 9.37903643e-01 5.96440494e-01 -7.60586202e-01 1.95918411e-01 8.46291006e-01 -2.25473374e-01 -5.60203731e-01 -7.09050477e-01 -6.61628395e-02 -6.31156981e-01 9.17095482e-01 9.63919938e-01 -1.73095930e-02 -1.37439716e+00 1.57723391e+00 2.90039539e-01 -1.46009356e-01 -2.65451223e-02 1.49967551e+00 1.06534708e+00 -1.82627812e-01 -3.96798626e-02 2.29300261e-01 1.24890852e+00 -7.91917324e-01 -4.27120984e-01 -2.82120723e-02 2.43339121e-01 -6.03504419e-01 7.21906185e-01 5.87279975e-01 -1.11221814e+00 -4.52324688e-01 -7.71342099e-01 -2.69908309e-01 -4.42515522e-01 5.47433734e-01 7.27305114e-01 8.10898423e-01 -9.45110261e-01 4.05500650e-01 -1.08212471e+00 -7.17418194e-01 4.00405616e-01 6.62245810e-01 -8.15872669e-01 -7.17486721e-03 -7.20910251e-01 1.12503433e+00 -1.99270919e-01 5.20912528e-01 -4.67191577e-01 -1.84383914e-01 -8.56135249e-01 -5.24327397e-01 2.31802955e-01 -9.86479759e-01 9.32760239e-01 -9.11768079e-01 -2.13874125e+00 1.32253325e+00 -3.21258217e-01 -1.41212374e-01 8.91294777e-01 -4.11951900e-01 3.60140274e-03 3.04098368e-01 -6.62218153e-01 6.01678550e-01 1.22587299e+00 -1.02588105e+00 -1.97829932e-01 -1.00431502e+00 -2.76245207e-01 1.61172628e-01 -1.63042784e-01 3.04938227e-01 -5.31731904e-01 -3.05263877e-01 3.47863287e-01 -9.61117864e-01 3.88644412e-02 4.88761187e-01 -1.32594764e-01 -1.72140285e-01 8.42848897e-01 -9.11497116e-01 5.68592966e-01 -1.86224496e+00 7.96087623e-01 3.25323373e-01 -5.01552597e-02 6.46196246e-01 -1.66664600e-01 1.38107866e-01 8.47174376e-02 -2.80221671e-01 -1.18050314e-01 -5.39852679e-01 2.70911217e-01 4.69320208e-01 -1.12148032e-01 8.47853005e-01 2.01885462e-01 1.23588872e+00 -5.61210334e-01 -4.50857997e-01 7.73651540e-01 1.07313275e+00 -2.38181874e-01 1.80934012e-01 -1.61922276e-02 7.33852386e-01 -3.74465913e-01 1.02564049e+00 3.11753035e-01 8.80997106e-02 5.32498956e-02 -1.57100484e-01 -7.84416683e-03 -3.13945226e-02 -1.19052792e+00 1.97255945e+00 -3.97639841e-01 4.94888723e-01 3.23938489e-01 -9.21601176e-01 8.38055670e-01 5.41185081e-01 6.55253232e-01 -6.68211401e-01 6.22697771e-01 5.89146614e-01 -4.75944877e-02 -7.66179800e-01 -2.82437801e-01 -2.30349049e-01 2.78910458e-01 5.07618904e-01 3.89953405e-01 -2.07380369e-01 -3.41171250e-02 -5.64720869e-01 8.40891719e-01 8.57001007e-01 6.93483576e-02 5.27048409e-01 8.12904894e-01 -3.75052601e-01 9.21883248e-03 4.23751682e-01 -4.47394609e-01 7.01413214e-01 2.46496245e-01 -4.59549695e-01 -5.38309097e-01 -8.66155624e-01 1.85377032e-01 1.03374732e+00 -2.83097953e-01 1.32387936e-01 -5.44158459e-01 -7.85102129e-01 2.76600540e-01 5.01470454e-02 -6.28558397e-01 2.84775198e-01 -7.88394392e-01 -3.96190703e-01 7.42421448e-01 6.08884752e-01 4.28452343e-01 -1.26352131e+00 -1.30243242e+00 -6.46279827e-02 2.65800744e-01 -1.15076959e+00 -7.13404939e-02 4.56381142e-02 -8.59202147e-01 -1.26507866e+00 -1.28671694e+00 -5.19324899e-01 4.01617020e-01 -3.18603128e-01 5.55584252e-01 5.98274283e-02 -6.54314280e-01 9.88854527e-01 -5.22364020e-01 -1.93020731e-01 -2.87344962e-01 -1.48737222e-01 2.52725817e-02 2.17671543e-01 4.29599404e-01 -9.60721433e-01 -2.69770741e-01 2.63966501e-01 -6.04377389e-01 -8.93391594e-02 7.85653591e-01 8.30631912e-01 3.09136093e-01 -9.21673775e-01 -2.55626649e-01 5.82923591e-02 1.41388163e-01 4.26505119e-01 -4.77068037e-01 2.45042115e-01 -1.39287293e-01 3.03290606e-01 1.11384682e-01 -3.24279040e-01 -8.99759948e-01 6.38062596e-01 -4.18295354e-01 -5.75450242e-01 -6.20270729e-01 2.49141082e-02 -1.44632697e-01 -4.35875028e-01 3.22446972e-01 5.78729868e-01 6.55077040e-01 -7.80538797e-01 7.29502559e-01 6.93579376e-01 7.70290792e-01 -2.52368122e-01 6.73396409e-01 8.84217322e-01 4.55303669e-01 -1.09825671e+00 -5.61412096e-01 -8.04908991e-01 -1.54065120e+00 -5.41898012e-01 9.23857987e-01 -6.09282970e-01 -9.95942652e-01 1.13203502e+00 -1.12335002e+00 -2.77458370e-01 -2.77555168e-01 6.41036391e-01 -8.85813534e-01 6.95678592e-01 -3.01917762e-01 -1.12678778e+00 -4.79811430e-01 -1.07071257e+00 1.49731624e+00 -4.05659564e-02 -2.12791234e-01 -6.53009057e-01 8.15442502e-02 5.00205159e-01 2.98263550e-01 5.42068064e-01 2.43502870e-01 -1.99493006e-01 -5.28944314e-01 -4.18014824e-01 -1.26442641e-01 5.50118566e-01 5.81692681e-02 7.95017183e-02 -1.11222708e+00 -7.39317527e-03 -2.91414648e-01 -5.77180982e-01 1.23105776e+00 3.83269370e-01 5.07749021e-01 1.33502707e-01 -3.95599119e-02 8.24880779e-01 9.94731843e-01 -2.08202034e-01 4.36349183e-01 4.26061749e-02 7.96918869e-01 7.18030274e-01 1.03982408e-02 5.16608894e-01 1.87727481e-01 9.75609541e-01 4.57679659e-01 -2.16384470e-01 -7.89511323e-01 -1.29438609e-01 6.27908051e-01 6.39375210e-01 -9.50367272e-01 4.00293708e-01 -8.77031803e-01 2.56508917e-01 -1.78249943e+00 -9.51485753e-01 -7.90228173e-02 1.84727430e+00 6.08599424e-01 -2.54115403e-01 3.67743045e-01 5.40066779e-01 5.23287505e-02 8.23495388e-02 -5.82191586e-01 -8.20392594e-02 -4.08537835e-01 4.46470350e-01 4.05094028e-01 3.75797987e-01 -1.10877883e+00 1.03117609e+00 5.85717630e+00 1.92954987e-01 -1.52121174e+00 -2.02265054e-01 -2.65543073e-01 -2.78897226e-01 2.44623512e-01 -3.28602016e-01 -7.55374134e-01 -2.81520542e-02 4.29323822e-01 5.72022855e-01 4.98763561e-01 7.06771433e-01 1.32890239e-01 8.12957529e-03 -1.03736269e+00 1.16623187e+00 6.15157008e-01 -6.50823832e-01 3.40196155e-02 1.96814775e-01 1.85899824e-01 3.19727778e-01 -8.43820721e-02 -6.85888603e-02 4.82347347e-02 -1.05375218e+00 8.39994967e-01 1.00118124e+00 8.14917803e-01 -1.42458379e-01 5.73767781e-01 3.89411002e-01 -9.38951254e-01 -1.69981584e-01 3.15820575e-01 -2.24000826e-01 4.13916379e-01 5.01591899e-02 -5.50621867e-01 4.02198941e-01 5.88215292e-01 9.19851780e-01 -4.74784940e-01 7.92468190e-01 -7.16151714e-01 4.63291973e-01 -7.52199054e-01 -1.83571242e-02 1.31125301e-01 -2.35048328e-02 7.40203798e-01 1.15671277e+00 1.03614993e-01 -2.97991540e-02 -6.93095997e-02 7.38058269e-01 4.63482976e-01 6.86599240e-02 -5.91247499e-01 -8.39682519e-02 -5.54872692e-01 9.74850714e-01 -4.66262996e-01 -1.30472407e-01 -4.07145888e-01 1.16569197e+00 1.00328930e-01 2.40031943e-01 -4.17780370e-01 1.81091465e-02 7.85219252e-01 -9.95949507e-02 6.96822226e-01 -6.55553997e-01 -2.04659253e-01 -1.40332532e+00 3.33781004e-01 -5.55753589e-01 1.56695724e-01 -7.65233040e-01 -1.07728958e+00 3.09101731e-01 -3.45343292e-01 -1.04021883e+00 -5.87852895e-01 -1.34092844e+00 -2.40530938e-01 9.03435528e-01 -1.87628686e+00 -1.69716513e+00 -4.61370885e-01 9.15896952e-01 1.67595763e-02 -1.00369774e-01 9.94319260e-01 1.98054403e-01 -3.34331721e-01 5.68562448e-01 -3.57969552e-01 5.27348578e-01 5.63392103e-01 -9.95851696e-01 2.40250647e-01 6.36435568e-01 5.15812993e-01 3.99068028e-01 2.83844203e-01 -2.88487911e-01 -1.81866062e+00 -3.65170211e-01 9.20067012e-01 -7.44036973e-01 3.58644605e-01 -1.84348553e-01 -3.68568271e-01 4.23825949e-01 -3.13746750e-01 3.24220479e-01 3.67666006e-01 -1.53777435e-01 -6.32203758e-01 4.93816994e-02 -8.74570847e-01 4.51770574e-01 1.49113309e+00 -5.48445702e-01 -8.57245207e-01 3.33956606e-03 -1.13333188e-01 -4.94417936e-01 -7.80697703e-01 4.04176503e-01 1.40141165e+00 -1.33845484e+00 1.12424004e+00 -7.07181692e-01 1.69241473e-01 -1.89053610e-01 -1.81950957e-01 -8.23189914e-01 4.16961402e-01 -4.18302268e-01 -2.82965004e-01 9.02220488e-01 -3.06693930e-02 -4.95705813e-01 9.86942589e-01 4.67563272e-01 2.11952657e-01 -5.39336383e-01 -1.33369815e+00 -7.12943375e-01 -1.19382195e-01 -7.51438737e-01 2.81267583e-01 3.32165807e-01 -5.72112836e-02 5.83813302e-02 -5.79005420e-01 -2.51844108e-01 6.44675434e-01 4.37497437e-01 1.19558609e+00 -1.49229956e+00 -2.54233658e-01 -7.78339803e-01 -8.63099098e-01 -1.19344759e+00 3.23983490e-01 -9.20432925e-01 -1.13584690e-01 -1.68137980e+00 -1.92839965e-01 4.24984396e-02 -1.02227293e-02 9.50686574e-01 3.64087343e-01 5.91331422e-01 5.23654997e-01 8.42954516e-02 -3.29452515e-01 5.42321086e-01 1.46782529e+00 -6.79686889e-02 -1.45929426e-01 4.90705371e-02 1.53442957e-02 9.16992605e-01 2.57925361e-01 7.48101398e-02 2.21920133e-01 -3.46111864e-01 -1.14105701e-01 8.16974565e-02 9.64363158e-01 -1.05136657e+00 1.85432434e-01 2.56456614e-01 4.76427734e-01 -5.76631129e-01 7.55764067e-01 -1.11486125e+00 -1.56176761e-01 5.74733138e-01 3.58722359e-02 -5.57336986e-01 5.50550707e-02 1.70656756e-01 -2.52312124e-01 1.89672068e-01 5.57513118e-01 -3.59807223e-01 -1.03801274e+00 2.13837683e-01 -2.75121033e-01 -2.88936555e-01 7.03217745e-01 -7.09455132e-01 2.26024225e-01 -2.89402336e-01 -1.04073298e+00 -1.45135045e-01 3.37467104e-01 5.11524439e-01 5.90979159e-01 -1.02857840e+00 -7.26519763e-01 7.39058495e-01 -7.75991753e-02 -2.23914877e-01 1.02124259e-01 1.21963990e+00 -3.50457281e-01 6.41219199e-01 -5.83788514e-01 -8.86044562e-01 -1.80670869e+00 -1.83530468e-02 5.88562548e-01 -1.71627656e-01 -8.15780282e-01 8.64857972e-01 -5.74801326e-01 -6.58449352e-01 5.53289175e-01 -5.99972188e-01 -2.50993490e-01 3.10014516e-01 3.42694908e-01 3.62672955e-01 -2.98590045e-02 -1.07926929e+00 -4.98552918e-01 1.54465055e+00 6.01615131e-01 -3.81763816e-01 1.40653741e+00 2.06228673e-01 -1.53327554e-01 3.79047185e-01 1.08995438e+00 -2.39144549e-01 -1.22929311e+00 -2.17078760e-01 -2.51276493e-01 -2.89720535e-01 1.47714801e-02 -1.14099944e+00 -1.12678945e+00 1.32513487e+00 6.67491436e-01 -4.50791419e-01 1.25981522e+00 -3.24089255e-04 6.49331808e-01 4.09864366e-01 4.98208344e-01 -8.80502224e-01 -2.02308685e-01 6.44390583e-01 1.12470794e+00 -1.22463131e+00 4.94262986e-02 -1.77441224e-01 -4.25470918e-01 1.42658234e+00 1.54833272e-01 -1.69731025e-02 7.93675363e-01 1.72661915e-01 4.83009309e-01 -1.85590550e-01 -8.53456780e-02 -7.81873345e-01 6.68846965e-01 7.33762145e-01 2.71596760e-01 -6.24959990e-02 -2.22968146e-01 4.30322409e-01 -1.66467965e-01 5.28450787e-01 -7.33667687e-02 9.89477277e-01 -2.77657527e-02 -1.32618952e+00 -4.35806662e-01 3.05006765e-02 -1.49755448e-01 2.34115019e-01 -7.36611485e-01 1.00909984e+00 3.23032707e-01 4.77211654e-01 -3.08576256e-01 -2.28664592e-01 5.67925036e-01 5.81955969e-01 1.20869732e+00 -1.55720696e-01 -6.55313432e-01 -7.79733434e-02 -4.62168828e-02 -8.54656041e-01 -1.01029921e+00 -8.77931535e-01 -1.26565981e+00 3.02064896e-01 -5.87753803e-02 -7.21746266e-01 1.14326930e+00 1.07219708e+00 3.18056308e-02 1.87833682e-01 -4.44820561e-02 -1.52358556e+00 -8.42640221e-01 -9.06700969e-01 -7.40166426e-01 5.54879189e-01 6.67137861e-01 -1.00628746e+00 -4.00913149e-01 3.40104140e-02]

[9.148006439208984, -6.456523418426514]

80e39430-9a81-4bf0-bfb8-269c07d4d4d2

semantic-embedded-unsupervised-spectral

2108.06659

https://arxiv.org/abs/2108.06659v2

https://arxiv.org/pdf/2108.06659v2.pdf

Semantic-embedded Unsupervised Spectral Reconstruction from Single RGB Images in the Wild

This paper investigates the problem of reconstructing hyperspectral (HS) images from single RGB images captured by commercial cameras, \textbf{without} using paired HS and RGB images during training. To tackle this challenge, we propose a new lightweight and end-to-end learning-based framework. Specifically, on the basis of the intrinsic imaging degradation model of RGB images from HS images, we progressively spread the differences between input RGB images and re-projected RGB images from recovered HS images via effective unsupervised camera spectral response function estimation. To enable the learning without paired ground-truth HS images as supervision, we adopt the adversarial learning manner and boost it with a simple yet effective $\mathcal{L}_1$ gradient clipping scheme. Besides, we embed the semantic information of input RGB images to locally regularize the unsupervised learning, which is expected to promote pixels with identical semantics to have consistent spectral signatures. In addition to conducting quantitative experiments over two widely-used datasets for HS image reconstruction from synthetic RGB images, we also evaluate our method by applying recovered HS images from real RGB images to HS-based visual tracking. Extensive results show that our method significantly outperforms state-of-the-art unsupervised methods and even exceeds the latest supervised method under some settings. The source code is public available at https://github.com/zbzhzhy/Unsupervised-Spectral-Reconstruction.

['Qingfu Zhang', 'Huanqiang Zeng', 'Junhui Hou', 'Hui Liu', 'Zhiyu Zhu']

2021-08-15

http://openaccess.thecvf.com//content/ICCV2021/html/Zhu_Semantic-Embedded_Unsupervised_Spectral_Reconstruction_From_Single_RGB_Images_in_the_ICCV_2021_paper.html

http://openaccess.thecvf.com//content/ICCV2021/papers/Zhu_Semantic-Embedded_Unsupervised_Spectral_Reconstruction_From_Single_RGB_Images_in_the_ICCV_2021_paper.pdf

iccv-2021-1

['spectral-reconstruction']

['computer-vision']

[ 8.50814402e-01 -8.66369382e-02 4.71987613e-02 -2.89970458e-01 -1.03247344e+00 -6.95765078e-01 1.21713422e-01 -5.53140640e-01 -3.47550839e-01 7.90044904e-01 -1.95139959e-01 -1.42924264e-01 3.77734415e-02 -7.90176511e-01 -1.17850828e+00 -1.16927755e+00 5.26366830e-01 -1.29682377e-01 -6.16847686e-02 -4.09209207e-02 -1.27454624e-01 4.18144494e-01 -1.48445880e+00 8.67424347e-03 8.75507712e-01 1.19328356e+00 3.55982661e-01 5.83263218e-01 1.91260561e-01 8.77585411e-01 -1.42601758e-01 -1.59473717e-01 8.25363457e-01 -5.12221277e-01 -5.87113976e-01 6.10496283e-01 4.63615835e-01 -4.93662775e-01 -5.82658410e-01 1.48229754e+00 5.39237976e-01 2.74606556e-01 3.33272874e-01 -1.20951426e+00 -6.45190179e-01 2.15076745e-01 -6.33263648e-01 -3.44839811e-01 3.73223633e-01 7.32338846e-01 7.37134516e-01 -6.79354727e-01 4.85339165e-01 5.60250938e-01 7.76799202e-01 3.75719488e-01 -1.41562402e+00 -7.70338595e-01 -1.46414906e-01 2.74078199e-03 -1.53735244e+00 -4.74164248e-01 1.12768614e+00 -1.41947344e-01 1.62764341e-01 5.40814817e-01 6.52941883e-01 1.09173572e+00 -4.88220215e-01 6.59928799e-01 1.81762969e+00 -5.59142470e-01 1.17635094e-01 5.38497604e-02 -5.06771922e-01 6.85051143e-01 -1.54082224e-01 3.63090307e-01 -4.82230276e-01 1.16094753e-01 8.75310898e-01 3.66515726e-01 -8.18288267e-01 -2.91883230e-01 -1.17592096e+00 5.10483325e-01 7.70538270e-01 -1.18675046e-01 -3.80479217e-01 2.11188886e-02 -2.39126205e-01 6.26124293e-02 2.03911051e-01 -6.31387308e-02 -3.37343812e-01 5.44944227e-01 -9.29325819e-01 -3.14770818e-01 3.92416477e-01 9.48267460e-01 1.20040798e+00 1.58718720e-01 2.15396836e-01 5.89855194e-01 4.25973058e-01 1.12602484e+00 3.09073180e-01 -1.20374632e+00 3.59003156e-01 4.05413181e-01 2.11382240e-01 -7.84538269e-01 -1.02825090e-01 -2.19509423e-01 -1.01524615e+00 3.25126261e-01 3.38099808e-01 9.99471620e-02 -9.73236501e-01 1.53773367e+00 2.45695651e-01 3.30588222e-01 1.93091244e-01 1.08056056e+00 5.62545419e-01 7.76123464e-01 -1.56778526e-02 -4.25742686e-01 6.27149045e-01 -8.16388965e-01 -4.63130563e-01 -2.90388256e-01 -6.53361529e-02 -5.86641729e-01 1.50969493e+00 4.19683546e-01 -9.00328100e-01 -5.46045840e-01 -1.00649381e+00 4.36362699e-02 -3.70341718e-01 3.66372019e-01 4.82431978e-01 5.54152846e-01 -8.26436758e-01 5.50306082e-01 -8.25232267e-01 -3.00680310e-01 4.27791327e-01 1.37752637e-01 -4.13538009e-01 -3.44425052e-01 -1.09656680e+00 3.92736554e-01 5.99582672e-01 1.84608832e-01 -1.09685075e+00 -6.55474544e-01 -6.64419532e-01 -5.00483692e-01 4.15496528e-01 -3.74190062e-01 8.33585203e-01 -1.46280956e+00 -1.64796150e+00 9.95498419e-01 1.37568727e-01 -9.80790779e-02 5.12116134e-01 6.49773628e-02 -4.06856388e-01 5.07860541e-01 -4.66683097e-02 3.87773693e-01 1.18958974e+00 -1.66494441e+00 -3.04841429e-01 -3.84626567e-01 -1.39890283e-01 3.17492425e-01 -4.48540211e-01 -3.06665003e-01 -5.72079360e-01 -5.19140244e-01 3.04978222e-01 -1.10295463e+00 -8.10038075e-02 3.82547081e-01 -7.17900991e-01 8.05631638e-01 6.73368812e-01 -9.91015375e-01 6.55631900e-01 -2.11727357e+00 3.18698138e-01 2.93299556e-01 -1.23933621e-01 2.06670970e-01 7.67990425e-02 2.03837663e-01 -1.04338787e-01 -2.25564659e-01 -1.00359106e+00 -3.40458483e-01 -7.05154985e-02 2.62126803e-01 -4.18126911e-01 8.71042073e-01 -1.37872100e-01 7.78990090e-01 -9.61658955e-01 -2.27018431e-01 6.01571262e-01 5.17108977e-01 5.77290766e-02 5.60851097e-01 -3.06033671e-01 1.01705146e+00 -1.41916707e-01 1.06674612e+00 8.95715833e-01 -2.56069392e-01 2.50911444e-01 -6.22210979e-01 -9.27877650e-02 -3.80018473e-01 -1.19412720e+00 1.98368561e+00 -3.68082941e-01 3.56056869e-01 2.43392199e-01 -1.00512862e+00 6.65546954e-01 -4.62651849e-02 5.35327554e-01 -7.08622098e-01 -1.60098001e-02 1.76694974e-01 -7.26819098e-01 -5.02242804e-01 1.56772379e-02 -3.65130723e-01 1.07524455e-01 3.39868277e-01 -6.05202802e-02 -6.36814713e-01 -4.11687225e-01 -1.02395795e-01 7.13689268e-01 5.73187292e-01 7.45038986e-02 2.41123706e-01 6.07101202e-01 -8.18808898e-02 4.17555302e-01 5.37028909e-01 -1.36434287e-01 1.00843310e+00 -1.64363921e-01 -1.59193233e-01 -1.05523646e+00 -1.35497940e+00 -4.64934632e-02 8.91739011e-01 4.55395341e-01 2.90468067e-01 -6.41156316e-01 -5.93788743e-01 -1.37147512e-02 6.46506846e-01 -5.29668152e-01 -2.29114387e-02 2.36298963e-02 -8.74202251e-01 6.57661080e-01 2.93611586e-01 9.43888724e-01 -9.46406186e-01 -4.91097033e-01 -2.19219700e-01 -2.75552779e-01 -1.18779266e+00 -1.96351573e-01 3.16490024e-01 -6.61906779e-01 -1.19667876e+00 -6.18504763e-01 -4.55607951e-01 7.04305291e-01 4.70870823e-01 7.30135739e-01 -9.86279398e-02 -4.66212958e-01 6.59740746e-01 -3.15254211e-01 -2.69025452e-02 -2.02528670e-01 -3.08076054e-01 -1.30631896e-02 3.97202879e-01 5.05734840e-03 -5.57601392e-01 -8.32908332e-01 2.53019333e-01 -1.54289985e+00 2.11568117e-01 6.07714236e-01 7.41788447e-01 1.01564515e+00 3.83358717e-01 -1.04450598e-01 -7.37379432e-01 -2.46072367e-01 -2.65547872e-01 -7.75353074e-01 4.23428118e-01 -5.41285038e-01 -1.87887311e-01 8.20958078e-01 -2.17459887e-01 -1.19665861e+00 7.40875006e-01 4.92315553e-02 -8.54282260e-01 -3.35963011e-01 1.83377236e-01 -4.16585892e-01 -5.91473877e-01 5.57207644e-01 8.53526771e-01 1.42181739e-01 -2.49302909e-01 6.77144945e-01 7.08616138e-01 9.97104585e-01 -4.37547505e-01 1.32751310e+00 9.40079153e-01 -1.32352814e-01 -8.86192024e-01 -9.93074417e-01 -5.59013665e-01 -6.02560103e-01 -4.69963193e-01 8.03567886e-01 -1.24445534e+00 -4.19482142e-01 7.48620033e-01 -6.68921828e-01 -7.62127817e-01 -2.74797797e-01 2.22203106e-01 -6.83691978e-01 5.96891224e-01 -5.26019752e-01 -7.85772920e-01 -2.08862364e-01 -1.06705141e+00 1.26732063e+00 3.67275029e-01 7.07117617e-01 -9.39741969e-01 -1.01466775e-01 5.98974645e-01 2.86839128e-01 6.28051758e-01 3.01081598e-01 1.79765940e-01 -7.89123416e-01 -5.24120480e-02 -3.69316220e-01 9.33269680e-01 3.25858742e-01 -7.05637857e-02 -1.24785590e+00 -3.57455403e-01 7.32359439e-02 -6.17527544e-01 8.76411378e-01 1.42708600e-01 1.45304072e+00 -3.25027406e-01 6.64241910e-02 1.34960532e+00 2.03058839e+00 -2.47808933e-01 9.46796477e-01 5.72499692e-01 1.03995121e+00 5.17902315e-01 5.54977596e-01 4.50807601e-01 1.99552134e-01 3.27537000e-01 8.84161711e-01 -4.50013757e-01 -1.48444399e-01 -2.72468537e-01 5.58615863e-01 4.06020015e-01 -1.83669776e-01 -5.21171838e-02 -7.06391037e-01 1.38828501e-01 -1.66245878e+00 -9.00978327e-01 -6.48606941e-02 2.47692895e+00 9.92520928e-01 -3.12843144e-01 -8.64607170e-02 1.37574375e-01 7.81831205e-01 3.48536402e-01 -7.95184195e-01 6.79074943e-01 -5.67884266e-01 2.77517170e-01 1.12385964e+00 4.56522346e-01 -1.16555703e+00 9.85299766e-01 5.12153053e+00 6.24768674e-01 -1.37038589e+00 1.95036054e-01 6.54704332e-01 5.78337535e-02 -3.42880785e-01 5.44052944e-02 -2.87931412e-01 5.50972462e-01 5.29905617e-01 3.35074514e-01 1.09073997e+00 4.76188004e-01 1.54698193e-01 -8.97867009e-02 -4.71768647e-01 1.12726521e+00 1.63526326e-01 -9.69053447e-01 -1.45644173e-01 -1.26767278e-01 9.69422162e-01 7.69283250e-02 2.82734215e-01 -2.18285382e-01 2.68036544e-01 -8.32327306e-01 8.29256654e-01 8.54004443e-01 1.21369016e+00 -4.24009591e-01 3.83069545e-01 2.55759954e-01 -1.13286209e+00 -2.11961418e-01 -5.07355809e-01 2.93244481e-01 -8.72414783e-02 5.23812711e-01 -2.71442980e-01 8.51725817e-01 9.26382780e-01 9.19786692e-01 -6.56933308e-01 7.04748571e-01 -7.27867186e-01 6.04753971e-01 -2.35385358e-01 7.05892444e-01 -1.57906618e-02 -6.64343953e-01 2.46899351e-01 9.40728009e-01 3.80676270e-01 3.45510036e-01 2.19251633e-01 9.79550779e-01 -1.14813820e-01 -2.11674780e-01 -6.96193099e-01 -8.52790102e-02 3.04927170e-01 1.39928186e+00 -5.62235832e-01 -4.84892875e-02 -3.34410667e-01 1.54301298e+00 -5.59711196e-02 6.56757712e-01 -1.10237026e+00 -2.29228199e-01 2.13244081e-01 -2.05495325e-03 1.82019919e-01 -1.25071406e-01 -3.10564917e-02 -1.22796440e+00 -1.63718071e-02 -8.49487066e-01 1.97550252e-01 -1.48085403e+00 -1.24138141e+00 4.73324507e-01 -2.22709313e-01 -1.60557723e+00 1.92111939e-01 -6.98598504e-01 -4.07371223e-01 7.69808590e-01 -1.87292969e+00 -1.60286891e+00 -1.02668703e+00 1.06080782e+00 1.10942326e-01 6.97039515e-02 6.66124582e-01 3.00330035e-02 -6.25982523e-01 1.90004155e-01 5.04180610e-01 -5.39908670e-02 7.96795785e-01 -1.25953114e+00 -2.07346588e-01 1.02343690e+00 -6.48028329e-02 2.92234033e-01 5.18982887e-01 -4.32202071e-01 -1.67562795e+00 -1.45248151e+00 -2.29601145e-01 -3.10030818e-01 7.30425954e-01 -4.66295555e-02 -7.23382711e-01 6.88809156e-01 2.23726168e-01 4.68226105e-01 5.78805327e-01 -9.59452212e-01 -4.23845917e-01 -5.00976682e-01 -1.27720273e+00 2.68479347e-01 9.51860785e-01 -9.40002143e-01 -1.46591738e-01 6.05554581e-01 7.23332822e-01 -3.23857993e-01 -9.60528493e-01 4.83443797e-01 4.25911784e-01 -1.14719248e+00 1.30009043e+00 -5.96926026e-02 4.14689988e-01 -7.39008188e-01 -7.01617181e-01 -1.22866821e+00 1.70734480e-01 -4.81344670e-01 1.07553430e-01 1.11700773e+00 1.84745967e-01 -5.23679376e-01 7.67428517e-01 4.95381147e-01 -3.58395390e-02 -1.68851569e-01 -4.13254827e-01 -6.62975490e-01 -2.36610577e-01 -2.94044703e-01 5.08494318e-01 1.11476696e+00 -6.37037218e-01 -2.92831540e-01 -5.14759362e-01 7.70240366e-01 1.33967543e+00 3.60938519e-01 7.92835534e-01 -6.86166704e-01 -5.25927246e-01 -9.44554359e-02 -4.37945016e-02 -5.87992728e-01 2.66691595e-01 -7.48236477e-01 3.56593877e-01 -1.22855735e+00 1.88802287e-01 -3.41613293e-01 -3.97028118e-01 6.33175373e-01 -1.87423408e-01 8.36278439e-01 2.08148211e-01 4.58852410e-01 -6.32619083e-01 6.71903789e-01 1.11905491e+00 -4.21411097e-01 -2.34538130e-02 -1.08095273e-01 -3.80606532e-01 5.55520058e-01 8.95392478e-01 -3.26591015e-01 -3.05982947e-01 -3.53242695e-01 5.83965816e-02 9.87241939e-02 8.53605807e-01 -1.03228974e+00 2.96955377e-01 -4.01093334e-01 6.30919993e-01 -1.24528602e-01 2.82538682e-01 -1.04483521e+00 4.87965494e-01 2.78477073e-01 -1.69666633e-01 -5.16586304e-01 6.69228956e-02 6.81577206e-01 -5.30338027e-02 -5.88887110e-02 9.98004735e-01 -4.15486336e-01 -9.18398380e-01 3.72554004e-01 1.75007597e-01 -3.46703708e-01 9.73969996e-01 -1.36283711e-01 -2.81787395e-01 -4.14711863e-01 -6.78872764e-01 -2.01280624e-01 1.07425559e+00 -2.67829657e-01 7.00781941e-01 -1.09968626e+00 -3.75836104e-01 2.76599377e-01 3.24956685e-01 1.90886185e-01 3.13908577e-01 8.15734029e-01 -8.05914342e-01 -1.89327747e-01 -3.33245009e-01 -6.08886719e-01 -8.86186600e-01 5.21412730e-01 4.83774632e-01 2.09410846e-01 -4.69566762e-01 7.28330910e-01 -4.62954305e-03 -8.08607817e-01 3.64279747e-01 1.25596218e-03 4.30572122e-01 -3.45497459e-01 2.88689733e-01 2.56065160e-01 -6.83883531e-03 -7.48626709e-01 -1.68536529e-01 6.09641731e-01 6.75907075e-01 -1.30437687e-01 1.50003135e+00 -4.30687070e-01 -2.66229004e-01 3.64623129e-01 1.26412559e+00 5.80119938e-02 -1.90042818e+00 -3.07946414e-01 -4.24932718e-01 -7.33120263e-01 2.98831314e-01 -8.53476763e-01 -1.54724371e+00 6.30100369e-01 9.27461863e-01 2.19793916e-02 1.77876699e+00 -2.30237454e-01 5.32612145e-01 2.02503532e-01 4.92563158e-01 -9.47694898e-01 1.53477758e-01 -6.41592294e-02 6.41069531e-01 -1.57490528e+00 2.31789380e-01 -3.18662703e-01 -7.54387975e-01 9.86151993e-01 3.29883933e-01 4.94931303e-02 2.09867388e-01 -1.00350663e-01 1.52101114e-01 -2.78551155e-03 1.38272837e-01 -5.37945092e-01 3.30493189e-02 6.91317081e-01 -8.59220549e-02 9.89060327e-02 4.64065135e-01 1.55185953e-01 1.91031873e-01 2.24368237e-02 3.86934727e-01 8.30705881e-01 -2.40098923e-01 -9.14001346e-01 -6.58634365e-01 1.33775603e-02 -6.52409196e-02 -1.62437841e-01 -3.67172003e-01 7.53880382e-01 1.95629716e-01 7.29944348e-01 -4.60105777e-01 -5.77354729e-01 2.07476243e-01 4.45123054e-02 5.97373605e-01 -2.54291952e-01 -9.62679312e-02 2.97673680e-02 -4.73392695e-01 -7.13971972e-01 -8.70572686e-01 -6.47569776e-01 -1.06867158e+00 -1.03355125e-01 -2.24198923e-01 -2.46556774e-01 8.08686256e-01 6.93445504e-01 -8.68219733e-02 2.96439618e-01 1.13303244e+00 -1.28589892e+00 -5.66460431e-01 -6.91438138e-01 -9.39929664e-01 6.86094642e-01 4.91118968e-01 -3.89070898e-01 -8.12641442e-01 5.23493290e-01]

[10.299968719482422, -2.3907926082611084]

a59535c1-e5bc-4d41-ac5f-fec0e47244af

convolutional-random-walk-networks-for

1605.07681

http://arxiv.org/abs/1605.07681v3

http://arxiv.org/pdf/1605.07681v3.pdf

Convolutional Random Walk Networks for Semantic Image Segmentation

Most current semantic segmentation methods rely on fully convolutional networks (FCNs). However, their use of large receptive fields and many pooling layers cause low spatial resolution inside the deep layers. This leads to predictions with poor localization around the boundaries. Prior work has attempted to address this issue by post-processing predictions with CRFs or MRFs. But such models often fail to capture semantic relationships between objects, which causes spatially disjoint predictions. To overcome these problems, recent methods integrated CRFs or MRFs into an FCN framework. The downside of these new models is that they have much higher complexity than traditional FCNs, which renders training and testing more challenging. In this work we introduce a simple, yet effective Convolutional Random Walk Network (RWN) that addresses the issues of poor boundary localization and spatially fragmented predictions with very little increase in model complexity. Our proposed RWN jointly optimizes the objectives of pixelwise affinity and semantic segmentation. It combines these two objectives via a novel random walk layer that enforces consistent spatial grouping in the deep layers of the network. Our RWN is implemented using standard convolution and matrix multiplication. This allows an easy integration into existing FCN frameworks and it enables end-to-end training of the whole network via standard back-propagation. Our implementation of RWN requires just $131$ additional parameters compared to the traditional FCNs, and yet it consistently produces an improvement over the FCNs on semantic segmentation and scene labeling.

['Jianbo Shi', 'Lorenzo Torresani', 'Gedas Bertasius', 'Stella X. Yu']

2016-05-24

convolutional-random-walk-networks-for-1

http://openaccess.thecvf.com/content_cvpr_2017/html/Bertasius_Convolutional_Random_Walk_CVPR_2017_paper.html

http://openaccess.thecvf.com/content_cvpr_2017/papers/Bertasius_Convolutional_Random_Walk_CVPR_2017_paper.pdf

cvpr-2017-7

['scene-labeling']

['computer-vision']

[ 2.62180328e-01 2.44906768e-01 -9.56180841e-02 -7.04056561e-01 -3.01656336e-01 -3.65472734e-01 4.90109265e-01 9.24025178e-02 -6.88315630e-01 6.11132979e-01 -1.56615332e-01 -1.62391722e-01 -6.35255035e-03 -1.01697123e+00 -8.57850611e-01 -4.29108411e-01 1.26092359e-01 3.80694270e-01 1.05344415e+00 1.02668345e-01 1.72288299e-01 6.29153669e-01 -1.37904072e+00 3.55355740e-01 9.91232514e-01 1.23625004e+00 6.65339649e-01 3.62854123e-01 -5.05608439e-01 8.51712584e-01 -3.40031803e-01 -2.51223445e-01 1.46819413e-01 2.04683095e-03 -1.15461493e+00 -2.83540934e-01 5.98152518e-01 -9.26816612e-02 -5.96180372e-02 9.55325007e-01 3.29995006e-01 3.70123178e-01 3.83617848e-01 -1.00652695e+00 -6.47146523e-01 6.38514459e-01 -5.56935132e-01 -9.11325514e-02 -2.58943111e-01 -1.46321669e-01 9.77873206e-01 -7.88385689e-01 4.34482008e-01 1.19522071e+00 1.18635213e+00 5.10449290e-01 -1.38562465e+00 -4.96976823e-01 3.83576512e-01 -8.60035643e-02 -1.58286822e+00 -1.76203579e-01 4.06686902e-01 -2.35414296e-01 1.36170518e+00 7.62627646e-03 5.10463119e-01 5.48300624e-01 -4.86060083e-02 7.59116232e-01 9.13591027e-01 -1.97270602e-01 2.35600471e-01 1.23930303e-03 2.33096987e-01 7.56519973e-01 1.42871767e-01 -2.74116844e-01 -2.62106359e-01 2.12975815e-01 1.01676071e+00 2.00026080e-01 -5.05126379e-02 -3.26189488e-01 -8.49970043e-01 8.19240034e-01 1.05632746e+00 3.90522361e-01 -1.52648434e-01 3.88829976e-01 5.37798814e-02 -3.80192727e-01 4.07210320e-01 4.07545358e-01 -6.68924153e-01 3.00992250e-01 -1.30335653e+00 2.87907541e-01 4.99472499e-01 6.97050095e-01 1.06455779e+00 -3.13507646e-01 -1.84855163e-01 1.04454660e+00 5.38383722e-01 1.31282762e-01 3.83108616e-01 -1.08766675e+00 2.68945664e-01 7.62095332e-01 -7.74054974e-02 -1.10091245e+00 -7.00185001e-01 -5.26954889e-01 -6.89396679e-01 3.20008248e-01 3.81044894e-01 -1.08234994e-02 -1.45852876e+00 1.82118297e+00 2.63893336e-01 3.50199223e-01 -2.26298720e-01 9.35072005e-01 8.65441620e-01 4.86069202e-01 5.53739846e-01 4.37305838e-01 9.66332614e-01 -1.43565190e+00 -4.55381632e-01 -6.85839653e-01 6.05266690e-01 -8.16627085e-01 1.04178190e+00 3.78115289e-02 -1.10173643e+00 -6.15473926e-01 -9.76128936e-01 -4.30706918e-01 -6.81001306e-01 9.71157327e-02 9.86830652e-01 5.67696154e-01 -1.55347455e+00 7.93097615e-01 -9.35704172e-01 -3.55162829e-01 9.04859424e-01 8.09558868e-01 -1.13453880e-01 -6.45654127e-02 -9.48934019e-01 9.35975611e-01 5.55989802e-01 3.51464331e-01 -3.28883678e-01 -5.99065423e-01 -1.04001975e+00 1.57485068e-01 2.46766627e-01 -6.18883193e-01 1.20662451e+00 -9.92265761e-01 -1.43004072e+00 6.80861890e-01 -2.49879926e-01 -4.54725087e-01 3.69388998e-01 -4.40593183e-01 -7.87631720e-02 -1.28257260e-01 3.21144611e-01 1.49894059e+00 3.20209146e-01 -1.21571064e+00 -5.28293014e-01 -1.26489893e-01 -9.54739079e-02 1.05159730e-01 9.05100908e-03 -1.65752247e-01 -7.63722003e-01 -5.70084691e-01 4.63167161e-01 -7.10841954e-01 -6.63256049e-01 2.95647662e-02 -5.76139390e-01 -2.38523498e-01 8.51619065e-01 -2.64435798e-01 1.02487016e+00 -1.96544945e+00 -2.47458026e-01 2.62394547e-01 1.47783309e-01 5.27604818e-01 -1.90187588e-01 -2.06684158e-03 1.05296329e-01 2.38332748e-01 -5.43754458e-01 -5.62793434e-01 -5.03940359e-02 3.60063881e-01 -1.80801049e-01 3.23239416e-01 3.94124031e-01 1.11740756e+00 -6.52421117e-01 -6.41469359e-01 3.32805872e-01 6.82736158e-01 -7.56392121e-01 -2.29436792e-02 -4.68096554e-01 2.44126096e-01 -2.26497948e-01 4.17807102e-01 1.01024175e+00 -5.38217843e-01 1.86989978e-02 -1.40158609e-01 -2.95783550e-01 4.22131568e-01 -1.04682171e+00 1.87405658e+00 -2.98436373e-01 4.94870961e-01 -7.90017918e-02 -1.11711764e+00 8.91844451e-01 -1.37764394e-01 3.44238579e-01 -7.24918365e-01 1.36540726e-01 3.22755098e-01 -2.46002585e-01 -5.36530651e-02 4.46240038e-01 -1.63970277e-01 3.64801288e-02 1.28235072e-01 3.00276786e-01 1.59171876e-02 8.36429521e-02 6.90618679e-02 9.57520485e-01 5.31214297e-01 -9.18475389e-02 -2.48241678e-01 2.98959374e-01 1.05623737e-01 8.55907917e-01 8.49253953e-01 -5.52048422e-02 9.55922127e-01 1.90047905e-01 -5.72084308e-01 -7.35422134e-01 -1.07582557e+00 -2.47482434e-01 1.05893576e+00 4.68845755e-01 -2.69615412e-01 -9.66095686e-01 -7.86470473e-01 -8.50085616e-02 4.78157341e-01 -4.93862301e-01 1.68401495e-01 -7.08844066e-01 -7.50540972e-01 5.50820708e-01 8.62992346e-01 9.53973293e-01 -1.29811561e+00 -5.51583886e-01 3.19014490e-01 -7.69028962e-02 -1.30126131e+00 -1.33568227e-01 3.36296201e-01 -9.04544353e-01 -8.45450282e-01 -6.49062932e-01 -1.00770998e+00 7.52840519e-01 1.25343680e-01 1.09975183e+00 9.98608023e-02 -3.25074255e-01 -2.89195746e-01 -1.25146165e-01 -1.77399397e-01 9.82695445e-02 3.53245974e-01 -3.56560171e-01 -2.38305524e-01 4.47094202e-01 -4.87890303e-01 -7.60249794e-01 4.69479650e-01 -8.58748376e-01 3.21479350e-01 5.82891107e-01 6.33502007e-01 8.70638311e-01 9.22512710e-02 4.64494228e-01 -1.18590224e+00 1.55701518e-01 -2.52994537e-01 -7.41280138e-01 2.08706394e-01 -5.54090440e-01 4.96408902e-02 5.46128392e-01 -3.53404656e-02 -1.18605793e+00 3.98307472e-01 -6.20094299e-01 -9.84966531e-02 -4.44806278e-01 3.18317056e-01 6.65724725e-02 -2.47032881e-01 4.57901150e-01 -2.82057613e-01 -2.01317549e-01 -6.15944982e-01 3.99870634e-01 4.02234524e-01 5.17557800e-01 -2.67368913e-01 5.19411147e-01 4.30681139e-01 -2.02506676e-01 -5.98448157e-01 -1.18779564e+00 -4.47133064e-01 -7.84798980e-01 1.72036067e-01 1.19467258e+00 -7.49251425e-01 -5.65002382e-01 6.05794013e-01 -1.21697867e+00 -7.61614501e-01 -2.78562635e-01 3.81995916e-01 -5.11045456e-01 1.69495299e-01 -7.36481488e-01 -4.87742573e-01 -1.59983575e-01 -1.02179980e+00 9.96986806e-01 6.13527775e-01 -2.34478220e-01 -9.95122313e-01 -9.52683240e-02 4.23863471e-01 6.77036166e-01 1.34836748e-01 8.01359951e-01 -4.31094676e-01 -7.18158305e-01 6.54532835e-02 -8.26823235e-01 2.18345702e-01 -4.89819087e-02 -1.39485300e-01 -1.21788764e+00 8.53647292e-02 -4.50183362e-01 -2.77810812e-01 1.29446387e+00 7.98469245e-01 1.49672616e+00 -2.24647466e-02 -6.11329377e-01 7.93329597e-01 1.76252818e+00 -6.70177788e-02 7.04177856e-01 4.12860572e-01 8.86925578e-01 5.91407061e-01 4.65327531e-01 -1.00402278e-03 3.99308622e-01 4.96132016e-01 5.18094301e-01 -6.62123322e-01 -2.93846458e-01 -1.34841740e-01 -1.20435946e-01 2.75340706e-01 1.22290179e-02 -3.24523836e-01 -1.02160597e+00 6.21169567e-01 -2.14975715e+00 -6.19204164e-01 -1.85937032e-01 1.80946505e+00 7.18080103e-01 1.86120763e-01 -1.67808428e-01 -1.37284994e-01 6.60358429e-01 7.22887665e-02 -5.00870705e-01 -4.93155509e-01 -1.14715442e-01 6.24884307e-01 8.10306787e-01 5.30734122e-01 -1.42069495e+00 1.44820893e+00 6.43860722e+00 8.48698258e-01 -1.14145541e+00 1.43919080e-01 7.72547483e-01 -2.00599823e-02 -8.91849175e-02 -1.24951880e-02 -1.01403141e+00 2.37915069e-01 6.39989972e-01 6.70376062e-01 1.60809442e-01 7.77412653e-01 1.34981707e-01 -4.16907638e-01 -6.91648602e-01 6.08256221e-01 -3.82225573e-01 -1.53905046e+00 -9.55818072e-02 -1.20624073e-01 7.86764026e-01 5.59921980e-01 -5.74817732e-02 2.19107434e-01 7.48243511e-01 -1.51360440e+00 7.42458165e-01 3.65052789e-01 6.92028582e-01 -6.53890371e-01 9.65539277e-01 3.38632852e-01 -1.15892565e+00 -1.05731018e-01 -5.61819673e-01 -1.14804536e-01 2.42025182e-01 6.87467337e-01 -6.82710230e-01 1.35894611e-01 9.56344545e-01 4.75755364e-01 -5.05088031e-01 1.20272076e+00 -1.96321160e-01 4.89391625e-01 -5.31629860e-01 1.98629975e-01 6.80096626e-01 9.81990173e-02 -1.73508584e-01 1.43914187e+00 1.45911351e-01 -3.97605374e-02 3.86443317e-01 1.13514864e+00 -1.04198158e-01 5.16620912e-02 -2.83680528e-01 2.31868252e-01 2.27262050e-01 1.24942410e+00 -1.37183750e+00 -1.47940978e-01 -4.34105515e-01 9.07341182e-01 7.89158344e-01 3.69260848e-01 -9.41965401e-01 -4.53471392e-01 5.54158449e-01 2.65911222e-01 6.19741142e-01 -1.99982524e-01 -6.99337244e-01 -7.24031448e-01 -1.55865416e-01 -1.08861014e-01 1.19089149e-01 -4.85648125e-01 -1.24410498e+00 6.40315831e-01 -2.90034175e-01 -5.13612568e-01 2.00960040e-01 -6.20347619e-01 -5.19658983e-01 9.79273617e-01 -1.87409914e+00 -1.29372203e+00 -3.03032368e-01 5.30237377e-01 4.35339659e-01 4.21394527e-01 8.77205133e-01 4.18323308e-01 -6.17540777e-01 4.89647359e-01 -2.90191621e-01 2.69544870e-01 4.81097937e-01 -1.38791871e+00 5.04088581e-01 7.95576870e-01 6.10793941e-02 6.07207716e-01 3.46854359e-01 -6.63932621e-01 -3.56469333e-01 -1.48429978e+00 1.11293375e+00 -6.12652116e-02 2.17185736e-01 -4.44519043e-01 -1.05036378e+00 5.68521202e-01 1.44034743e-01 4.37806129e-01 5.80833495e-01 1.89714164e-01 -2.09088624e-01 7.25132376e-02 -1.19624150e+00 4.00183588e-01 1.15498447e+00 -1.95265681e-01 -3.20368290e-01 2.39406466e-01 8.51990283e-01 -2.60856926e-01 -5.22391081e-01 6.40094876e-01 3.02634388e-01 -1.19806731e+00 1.02505577e+00 -8.97075832e-02 3.65573674e-01 -4.94178385e-01 -1.59620829e-02 -9.91083205e-01 -5.23811519e-01 -1.65801033e-01 2.78671652e-01 1.15059090e+00 7.43760467e-01 -6.62161052e-01 1.04320014e+00 8.04078639e-01 -2.48465523e-01 -9.97615039e-01 -5.76177776e-01 -5.57484150e-01 1.04667045e-01 -4.78078157e-01 5.08432508e-01 8.66552174e-01 -4.57314402e-01 2.34392092e-01 -9.33206826e-03 1.80442572e-01 4.02943105e-01 1.74361035e-01 2.63992369e-01 -1.33167577e+00 -1.71883315e-01 -5.84069550e-01 -3.51306617e-01 -1.25622833e+00 1.22397378e-01 -9.20136631e-01 4.40749854e-01 -2.00850630e+00 1.57914534e-01 -1.09854722e+00 -4.16598022e-01 8.59583795e-01 -9.16763395e-02 7.37867713e-01 6.31199852e-02 8.52013156e-02 -8.25096607e-01 3.27962995e-01 1.09508526e+00 8.09493288e-02 -3.54452133e-01 -1.04968332e-01 -5.03764331e-01 1.05655968e+00 9.54850972e-01 -4.51883167e-01 -4.12756056e-01 -7.29282677e-01 9.75709781e-02 -5.23682773e-01 5.40228367e-01 -1.22819340e+00 3.70893240e-01 5.57323061e-02 7.48851120e-01 -6.80958092e-01 1.84249163e-01 -7.04414368e-01 -5.85161634e-02 1.37311563e-01 -1.92025721e-01 -2.49354780e-01 3.79792601e-01 4.35285300e-01 -3.50380927e-01 -1.91053748e-01 9.50540841e-01 -3.38262379e-01 -9.51418221e-01 2.76015252e-01 -3.23088169e-01 -1.32186025e-01 9.27175820e-01 -4.46039259e-01 -1.85275763e-01 1.44524425e-01 -7.97506392e-01 2.89670795e-01 5.67194462e-01 2.22956762e-01 4.84769642e-01 -8.95182908e-01 -1.15630552e-01 1.80136457e-01 -4.13506210e-01 7.79048562e-01 2.65674561e-01 7.78270245e-01 -8.81865084e-01 4.83097345e-01 -2.00471103e-01 -6.97275281e-01 -9.77866232e-01 1.89980835e-01 4.98590410e-01 -4.16994512e-01 -5.77355742e-01 1.38655722e+00 4.46523130e-01 -8.40383232e-01 1.94003388e-01 -4.05851215e-01 -3.52283031e-01 -2.44092837e-01 1.19310908e-01 1.10201962e-01 6.24948665e-02 -6.89044893e-01 -4.70818281e-01 6.90380573e-01 -1.59260333e-01 1.04078613e-01 1.45833623e+00 -1.37696892e-01 -1.61016405e-01 -2.37615872e-02 1.12452078e+00 -3.17431003e-01 -1.56512749e+00 -2.19988629e-01 1.94638774e-01 -2.30474934e-01 2.20564455e-01 -1.02480590e+00 -1.24627948e+00 1.04155755e+00 4.91413862e-01 -9.23119672e-03 1.10575151e+00 1.07013777e-01 1.06038857e+00 6.03321753e-02 2.06859753e-01 -1.16872525e+00 -2.19836801e-01 6.58450007e-01 2.32985094e-01 -1.19907665e+00 -8.54022875e-02 -7.67504930e-01 -4.28876132e-01 8.82768273e-01 9.07579064e-01 -2.52634048e-01 7.26853907e-01 4.49875087e-01 1.60867229e-01 -2.94865012e-01 -3.08142424e-01 -4.03499544e-01 2.12611496e-01 3.97010803e-01 5.36048293e-01 6.95713386e-02 -1.98453575e-01 5.29093087e-01 -1.15879647e-01 1.05855748e-01 1.64444093e-02 9.03329015e-01 -6.29711688e-01 -1.15814269e+00 -1.43482368e-02 3.05299044e-01 -7.05257177e-01 -2.14356706e-01 -1.94480211e-01 5.90112567e-01 6.54435098e-01 8.14043105e-01 3.46157283e-01 -2.03244314e-01 9.58593413e-02 2.57053599e-02 2.18130887e-01 -7.54233956e-01 -6.03182316e-01 1.10387757e-01 -1.69404835e-01 -8.69731724e-01 -5.44107139e-01 -3.52145582e-01 -1.90913379e+00 -7.88227096e-02 -5.07035851e-01 -7.05105066e-02 6.99789047e-01 1.05701697e+00 4.47303385e-01 5.36622345e-01 6.04626760e-02 -8.52771521e-01 3.44633982e-02 -7.53658533e-01 -5.43385148e-01 1.63810506e-01 1.51500598e-01 -7.06240237e-01 1.78448141e-01 -1.36041731e-01]

[9.534236907958984, 0.36487188935279846]

fdb8170c-49c9-4915-a86b-10903f70ff1c

the-dreem-headband-as-an-alternative-to

https://doi.org/10.1101/662734

https://www.biorxiv.org/content/biorxiv/early/2019/06/10/662734.full-text.pdf

The Dreem Headband as an Alternative to Polysomnography for EEG Signal Acquisition and Sleep Staging

Despite the central role of sleep in our lives and the high prevalence of sleep disorders, sleep is still poorly understood. The development of ambulatory technologies capable of monitoring brain activity during sleep longitudinally is critical to advancing sleep science and facilitating the diagnosis of sleep disorders. We introduced the Dreem headband (DH) as an affordable, comfortable, and user-friendly alternative to polysomnography (PSG). The purpose of this study was to assess the signal acquisition of the DH and the performance of its embedded automatic sleep staging algorithms compared to the gold-standard clinical PSG scored by 5 sleep experts. Thirty-one subjects completed an over-night sleep study at a sleep center while wearing both a PSG and the DH simultaneously. We assessed 1) the EEG signal quality between the DH and the PSG, 2) the heart rate, breathing frequency, and respiration rate variability (RRV) agreement between the DH and the PSG, and 3) the performance of the DH’s automatic sleep staging according to AASM guidelines vs. PSG sleep experts manual scoring. Results demonstrate a strong correlation between the EEG signals acquired by the DH and those from the PSG, and the signals acquired by the DH enable monitoring of alpha (r= 0.71 ± 0.13), beta (r= 0.71 ± 0.18), delta (r = 0.76 ± 0.14), and theta (r = 0.61 ± 0.12) frequencies during sleep. The mean absolute error for heart rate, breathing frequency and RRV was 1.2 ± 0.5 bpm, 0.3 ± 0.2 cpm and 3.2 ± 0.6 %, respectively. Automatic Sleep Staging reached an overall accuracy of 83.5 ± 6.4% (F1 score : 83.8 ± 6.3) for the DH to be compared with an average of 86.4 ± 8.0% (F1 score: 86.3 ± 7.4) for the five sleep experts. These results demonstrate the capacity of the DH to both precisely monitor sleep-related physiological signals and process them accurately into sleep stages. This device paves the way for high-quality, large-scale, longitudinal sleep studies.

['Fabien Sauvet', 'Pascal Van Beers', 'Mathias Guillard', 'Pierrick J. Arnal', 'Albert Bou Hernandez', 'Mounir Chennaoui', 'Michael E. Ballard', 'Mason Harris', 'Hugo Jourde', 'Antoine Guillot', 'Valentin Thorey']

2019-06-10

biorxiv-2019-6

['sleep-stage-detection', 'sleep-quality-prediction', 'sleep-staging']

['medical', 'medical', 'medical']

[-1.51363596e-01 -2.83964664e-01 2.39245221e-01 -3.74318004e-01 -3.61796260e-01 -4.39056724e-01 -1.81486011e-01 2.05628306e-01 -6.29806161e-01 9.83644426e-01 2.37264801e-02 -2.27845103e-01 1.59881487e-01 -2.89990187e-01 -7.60809854e-02 -6.76539004e-01 -3.88589203e-01 1.31179184e-01 9.33641791e-02 1.33777857e-01 -7.12930039e-02 1.55234084e-01 -1.26074779e+00 -3.24079126e-01 1.20222676e+00 1.21050048e+00 2.78336048e-01 6.77136779e-01 4.68846172e-01 3.29240076e-02 -1.07003248e+00 7.96967521e-02 5.47250472e-02 -7.21935034e-01 -2.24352881e-01 -3.14906687e-01 -1.88042538e-03 -2.27214441e-01 7.94827119e-02 7.10978746e-01 8.24275851e-01 1.23133369e-01 -6.12991024e-03 -1.02484345e+00 4.79706712e-02 1.77994102e-01 -9.92137045e-02 9.44054544e-01 4.51114565e-01 2.77766526e-01 4.01808202e-01 -2.04740658e-01 -1.57782808e-03 1.45761687e-02 8.14800918e-01 6.10105276e-01 -1.48629498e+00 -1.06903636e+00 -6.01069331e-01 3.44090670e-01 -1.58420646e+00 -9.83487666e-01 3.73253584e-01 -4.54364002e-01 1.12645197e+00 4.31617022e-01 1.67349708e+00 6.00855827e-01 8.45673382e-01 -4.35770094e-01 1.49601412e+00 -3.37509885e-02 5.60953498e-01 3.32613111e-01 2.23138586e-01 3.38346124e-01 4.39572722e-01 -1.44632876e-01 -6.86212778e-01 -2.91698482e-02 4.31235820e-01 5.82180806e-02 -7.59700239e-01 5.62033892e-01 -9.19965327e-01 4.36606854e-01 3.77993174e-02 5.94027817e-01 -4.49875504e-01 -1.48002505e-01 1.83744952e-01 1.67066231e-01 4.00538057e-01 7.11419523e-01 -2.54923761e-01 -7.68562973e-01 -1.49911821e+00 -4.47848141e-01 7.69451559e-01 5.85212529e-01 2.83398151e-01 1.81121267e-02 -7.04480559e-02 7.74584055e-01 4.87455070e-01 7.69421458e-01 8.61690462e-01 -1.03088284e+00 -2.37249896e-01 4.41720545e-01 3.25203598e-01 -7.80388057e-01 -1.08906567e+00 -6.49913490e-01 -5.67850888e-01 -3.12084705e-01 2.79105216e-01 4.02904861e-02 -2.31057167e-01 1.54594922e+00 -1.24469772e-01 -7.04720169e-02 -3.15852195e-01 8.59870732e-01 7.02338934e-01 4.24477220e-01 -3.15735303e-02 -9.52544749e-01 1.71169329e+00 -3.38335574e-01 -9.61544394e-01 -4.80748802e-01 1.00722335e-01 -5.85869670e-01 1.13188636e+00 4.99481857e-01 -1.37422192e+00 -6.25811338e-01 -1.25687027e+00 2.37720877e-01 2.54424691e-01 1.84473589e-01 -6.41798824e-02 1.12411642e+00 -1.47182500e+00 4.75865394e-01 -1.55265129e+00 -5.33833265e-01 2.42692381e-01 6.95053816e-01 -2.14483067e-01 3.75470906e-01 -9.90676761e-01 1.04821455e+00 -2.51505315e-01 2.29893193e-01 -5.61127067e-01 -7.25618362e-01 -3.69579226e-01 1.07646681e-01 -3.57435524e-01 -5.79422951e-01 1.05341852e+00 -3.24562222e-01 -1.60937059e+00 8.62565100e-01 -5.19257843e-01 -3.03048074e-01 1.12981517e-02 -1.76628366e-01 -8.48908246e-01 4.80034024e-01 3.92359823e-01 -2.99579687e-02 4.86545712e-01 -3.43225330e-01 -2.00021327e-01 -6.93772793e-01 -3.83560210e-01 -1.25299603e-01 -2.78125145e-02 1.53854592e-02 1.37975201e-01 -8.10442269e-02 3.74980569e-01 -8.76735628e-01 2.35592231e-01 -2.04277456e-01 -2.01660879e-02 2.93738037e-01 1.20350197e-02 -8.56531441e-01 1.39159703e+00 -2.20211768e+00 -3.03867668e-01 1.42752111e-01 6.89070463e-01 2.06290446e-02 6.43384337e-01 8.27687383e-02 1.95139974e-01 -8.91974419e-02 -1.61508858e-01 -4.46608126e-01 -2.82432169e-01 -2.67361943e-02 3.07438165e-01 1.06299829e+00 -5.40982664e-01 6.43922210e-01 -8.89773130e-01 3.38479169e-02 3.61401707e-01 5.40567994e-01 -1.68611795e-01 5.42188406e-01 7.88081706e-01 6.98309481e-01 3.38943511e-01 4.59025919e-01 2.92381644e-01 -1.32425845e-01 1.03332847e-01 -3.84734720e-01 -6.10010862e-01 8.00625145e-01 -6.79147243e-01 1.57710326e+00 -3.60911667e-01 1.04915118e+00 1.52697042e-01 -4.45165128e-01 1.18223166e+00 5.96628010e-01 4.69550997e-01 -1.01787031e+00 2.60780722e-01 3.14874530e-01 5.17881274e-01 -5.83080173e-01 1.37519941e-01 -9.35946047e-01 2.98952401e-01 7.14141428e-01 8.35191458e-02 -1.88001603e-01 6.36930764e-02 2.75953524e-02 1.22232687e+00 -5.98775566e-01 5.92842340e-01 -7.26100147e-01 3.40222448e-01 -5.76162577e-01 4.40815479e-01 4.54956532e-01 -5.99650681e-01 4.62234259e-01 2.29427636e-01 -9.41805169e-03 -3.13046664e-01 -1.14501464e+00 -5.12989342e-01 3.02316338e-01 -3.97518836e-02 -8.35933566e-01 -6.95624530e-01 8.43097121e-02 -2.51851976e-01 8.68207455e-01 -3.91902655e-01 -6.60751939e-01 3.68098617e-02 -8.24215055e-01 3.68063420e-01 5.68344533e-01 4.50460285e-01 -7.67912388e-01 -1.22865498e+00 4.03965086e-01 -3.89016181e-01 -1.01555443e+00 -3.75462979e-01 3.34966749e-01 -8.80763888e-01 -8.79542589e-01 -2.60446191e-01 4.95737940e-02 3.63373548e-01 2.07587123e-01 1.06697953e+00 -1.89705342e-02 -4.42042172e-01 5.60540557e-01 -2.47496352e-01 -3.37412566e-01 -4.94708717e-02 -1.05206020e-01 6.72109723e-01 -8.00380334e-02 4.96927261e-01 -1.27759612e+00 -1.13965106e+00 5.14729977e-01 -1.57129273e-01 -2.40956172e-01 1.70016214e-01 1.12245306e-01 4.81834620e-01 -3.18565130e-01 5.67373753e-01 1.32518308e-02 6.49530768e-01 -4.06279325e-01 -5.68083584e-01 -2.74246007e-01 -1.14605820e+00 -5.52235305e-01 5.68424106e-01 -5.12145087e-02 -3.00935239e-01 -5.10732710e-01 -3.57251987e-02 1.22335479e-01 -1.00665815e-01 -2.36556828e-02 5.28028095e-03 1.13624349e-01 6.89850807e-01 2.58167535e-01 1.70887291e-01 -1.65915653e-01 -3.61411542e-01 8.14218521e-01 8.25811028e-01 1.40040563e-02 4.04293805e-01 3.91927272e-01 -2.54565209e-01 -1.19058430e+00 -6.12061203e-01 -6.05085552e-01 -5.53478420e-01 -2.93912798e-01 1.33881104e+00 -9.06008005e-01 -8.76163363e-01 1.09080799e-01 -5.91223061e-01 -5.07747114e-01 -2.88105905e-01 1.05386329e+00 -3.22867572e-01 2.00350299e-01 -2.57342815e-01 -9.25554752e-01 -1.06375098e+00 -1.05907285e+00 5.22095323e-01 3.34334761e-01 -1.11143267e+00 -7.75766194e-01 1.68406397e-01 4.87079412e-01 7.52948761e-01 2.52562135e-01 2.70679981e-01 -2.68647790e-01 2.12905947e-02 -1.80208549e-01 4.04567361e-01 4.66751158e-01 6.47859216e-01 -3.32319856e-01 -1.13203549e+00 -2.09800348e-01 9.04320121e-01 2.19956130e-01 -5.82461245e-02 6.11076355e-01 5.56053340e-01 3.74323688e-02 3.06329839e-02 6.94848120e-01 1.13322651e+00 7.96739042e-01 6.89317405e-01 2.88711518e-01 1.02093421e-01 1.29312038e-01 -4.85940464e-02 3.60821068e-01 2.90917456e-01 5.41105628e-01 1.34848848e-01 2.43520647e-01 2.68694252e-01 3.16608578e-01 6.32957101e-01 1.01839030e+00 -3.21471304e-01 2.52013594e-01 -7.40017474e-01 2.21893981e-01 -9.91153598e-01 -6.65646553e-01 -6.04612648e-01 2.52636456e+00 7.17548132e-01 3.58508348e-01 5.48687160e-01 2.89064884e-01 4.77515966e-01 -2.39442021e-01 -4.32270348e-01 -5.96602559e-01 3.13939363e-01 7.12556362e-01 2.65782982e-01 2.34912530e-01 -3.34849566e-01 9.60542411e-02 5.63565350e+00 -2.73901671e-01 -1.14907110e+00 2.98228115e-01 -2.65062582e-02 -7.77205348e-01 2.41294473e-01 -2.45066464e-01 -4.91908967e-01 1.21496010e+00 1.92753303e+00 -5.71328461e-01 7.54989147e-01 4.53291774e-01 8.38911176e-01 -6.92677557e-01 -1.09021866e+00 1.38248205e+00 6.27908856e-02 -8.99662256e-01 -1.06808877e+00 1.36305153e-01 9.35614333e-02 2.16559991e-01 -4.36419129e-01 -1.80932790e-01 -8.40407431e-01 -9.81665552e-01 7.40259945e-01 7.60919809e-01 1.22162282e+00 -3.11836571e-01 7.89198220e-01 1.24808662e-01 -1.14569581e+00 1.64725229e-01 -9.22155976e-02 -4.03358668e-01 3.25357109e-01 8.76123011e-01 -6.05215192e-01 8.03498030e-02 8.74997854e-01 7.46369421e-01 -4.96297956e-01 1.16958797e+00 -5.96807718e-01 1.05244315e+00 -5.96201539e-01 -4.42055007e-03 -4.52709019e-01 -3.89053375e-01 4.37144578e-01 7.08929062e-01 3.41963947e-01 4.48333025e-01 -5.57719469e-01 1.23854315e+00 2.50564277e-01 -3.00188959e-01 -1.51280567e-01 -5.13664000e-02 5.11620641e-01 1.28163779e+00 -9.36206639e-01 -4.47958149e-02 -5.43209910e-01 4.82879251e-01 -4.15523976e-01 5.48105873e-02 -6.48475826e-01 -4.96629179e-01 7.37613738e-01 4.15323466e-01 -5.08391440e-01 -1.73019469e-01 -5.76436102e-01 -1.03783512e+00 2.52873749e-01 -5.70573211e-01 8.97662193e-02 -1.01221359e+00 -8.36798072e-01 7.65219808e-01 -1.65056661e-01 -1.21738720e+00 9.78672653e-02 -1.26237897e-02 -8.40264678e-01 1.16837525e+00 -9.53041017e-01 1.59680828e-01 -8.08188021e-01 5.00722528e-01 1.48371860e-01 3.74200612e-01 9.93604600e-01 4.01453912e-01 -7.07451940e-01 3.47322494e-01 -2.80249596e-01 -5.39276898e-01 6.00634277e-01 -1.32414818e+00 1.00537829e-01 6.23380840e-01 -3.74801636e-01 1.03164423e+00 5.94139457e-01 -3.70057255e-01 -1.26116717e+00 -6.67846799e-01 9.62144732e-01 -5.71258485e-01 3.98649544e-01 -4.38325644e-01 -7.46840715e-01 4.36704040e-01 -5.34687517e-03 -3.31998229e-01 1.34428108e+00 -1.37213498e-01 4.32591498e-01 -6.82800472e-01 -1.40456498e+00 1.29101546e-02 4.97674793e-01 -7.18800366e-01 -8.65482032e-01 2.97171734e-02 9.40863639e-02 -3.44504386e-01 -1.31486905e+00 -4.90240147e-03 1.03428769e+00 -1.27174747e+00 3.99507463e-01 2.14686617e-01 -1.86841637e-01 -3.14022839e-01 7.70652294e-02 -1.25015497e+00 -3.69995415e-01 -1.04703045e+00 7.69041628e-02 1.04603648e+00 1.16509043e-01 -1.04281735e+00 2.88693756e-01 1.04563665e+00 -5.54784834e-01 -5.16520381e-01 -9.25626397e-01 -6.11113608e-01 -5.20626962e-01 -4.16769505e-01 2.30282009e-01 4.05133665e-01 4.93914008e-01 5.08629441e-01 2.21057441e-02 1.56104609e-01 2.73139447e-01 -1.32881537e-01 2.26775885e-01 -1.37738752e+00 -1.68024212e-01 -8.09775069e-02 -5.75748920e-01 -3.73566270e-01 -4.52246457e-01 -9.47603345e-01 -2.77864814e-01 -1.80632067e+00 -6.45591095e-02 6.11837842e-02 -4.31402266e-01 2.36568555e-01 -4.57891189e-02 3.75312269e-01 -1.12497032e-01 1.04171805e-01 -7.46691525e-02 3.68651748e-01 5.60124397e-01 3.82621855e-01 -7.67591596e-01 2.56199211e-01 -8.15245688e-01 3.98673713e-01 1.20826781e+00 -6.36521757e-01 -6.52674317e-01 -3.02401446e-02 1.76621318e-01 1.13481387e-01 2.01443344e-01 -1.35486817e+00 1.82402924e-01 5.73068559e-01 6.09233439e-01 -1.42470419e-01 4.53414828e-01 -7.72464275e-01 6.25474274e-01 6.36118948e-01 2.89861828e-01 2.65052825e-01 3.98694664e-01 -6.99153394e-02 2.79635668e-01 2.08310708e-01 9.55684483e-01 -2.84458697e-01 1.26484996e-02 -2.28889175e-02 -7.37850785e-01 3.07780206e-01 6.42177403e-01 -6.84789717e-01 -5.10809302e-01 -3.66416872e-01 -9.58659410e-01 7.53874630e-02 4.96183306e-01 -1.23782322e-01 4.77080554e-01 -8.19385290e-01 -1.15281485e-01 7.77860224e-01 -6.28597066e-02 -2.55454779e-01 2.25868180e-01 1.97706091e+00 -3.46386224e-01 3.57320577e-01 -4.37631607e-01 -8.12260449e-01 -1.06447792e+00 -1.83555216e-01 7.32957721e-01 3.72261554e-01 -8.48344266e-01 7.19097912e-01 -3.38700652e-01 7.80245245e-01 3.53666209e-02 -5.49362063e-01 2.19795164e-02 1.05335973e-01 7.94492900e-01 8.55071843e-01 4.07515109e-01 -2.17434749e-01 -7.00394154e-01 5.90778768e-01 3.91286492e-01 -1.34100720e-01 1.12502253e+00 -5.39114892e-01 -5.69129348e-01 1.04091239e+00 9.27054644e-01 2.64380366e-01 -6.76669240e-01 8.19745958e-01 -3.15598130e-01 -3.63917470e-01 3.14294770e-02 -9.72797692e-01 -8.98972929e-01 8.74814272e-01 1.11220682e+00 6.41288519e-01 1.46380520e+00 -8.41570571e-02 8.73015940e-01 -1.87050179e-01 5.95672250e-01 -8.75120878e-01 -3.60836118e-01 -3.44567031e-01 6.09513819e-01 -3.92114758e-01 1.86867744e-01 1.22991517e-01 -2.32389793e-01 1.11769831e+00 3.23423952e-01 3.87812816e-02 5.73026776e-01 1.81480303e-01 1.21377409e-01 -3.57011825e-01 -5.40885568e-01 1.41272679e-01 1.53448358e-01 4.88328457e-01 5.91064572e-01 2.67706841e-01 -7.42978871e-01 8.78384233e-01 -8.48430693e-01 2.92981297e-01 7.13412523e-01 6.61629975e-01 -5.06159186e-01 -5.59637666e-01 -1.74776301e-01 9.48784769e-01 -5.62031031e-01 7.52413869e-02 -7.79095218e-02 5.83511949e-01 1.30230352e-01 1.67404652e+00 3.05068612e-01 -3.98110300e-01 5.82718253e-01 4.88255709e-01 4.97848034e-01 -9.14238930e-01 -7.15964019e-01 2.08668739e-01 3.37359309e-02 -8.92992795e-01 -5.95925152e-01 -7.00273454e-01 -1.16707098e+00 1.55141996e-02 -1.02829263e-01 7.36358702e-01 8.33106816e-01 7.70722151e-01 4.96639550e-01 5.76238811e-01 3.81051838e-01 -5.18831134e-01 1.30686089e-01 -1.07006407e+00 -1.16741455e+00 -2.55807430e-01 5.51559269e-01 -6.25038624e-01 -7.86702573e-01 -8.52690488e-02]

[13.521849632263184, 3.429280996322632]

5ee8fb85-8700-4f98-a1a3-7ec6a61cef68

defect-detection-on-semiconductor-wafers-by

2111.03727

https://arxiv.org/abs/2111.03727v1

https://arxiv.org/pdf/2111.03727v1.pdf

Defect Detection on Semiconductor Wafers by Distribution Analysis

A method for object classification that is based on distribution analysis is proposed. In addition, a method for finding relevant features and the unification of this algorithm with another classification algorithm is proposed. The presented classification algorithm has been applied successfully to real-world measurement data from wafer fabrication of close to hundred thousand chips of several product types. The presented algorithm prefers finding the best rater in a low-dimensional search space over finding a good rater in a high-dimensional search space. Our approach is interesting in that it is fast (quasi-linear) and reached good to excellent prediction or detection quality for real-world wafer data.

['Thomas Olschewski']

2021-11-05

['defect-detection']

['computer-vision']

[ 2.71459937e-01 -8.71672928e-02 -2.36163169e-01 -6.92029655e-01 -6.90078259e-01 -1.10867225e-01 1.82368517e-01 4.52750474e-01 -2.08071977e-01 8.32248926e-01 -5.74891210e-01 -2.58031845e-01 -1.02057207e+00 -1.03837955e+00 -1.28411084e-01 -8.41619849e-01 -8.42880756e-02 1.08907175e+00 3.71191442e-01 -1.92055851e-01 7.47090816e-01 9.32457805e-01 -2.01011038e+00 3.29872072e-01 5.58355689e-01 1.35178328e+00 4.58147913e-01 5.89479446e-01 -1.60435826e-01 2.51789570e-01 -9.19150293e-01 -2.79915094e-01 1.97542742e-01 -4.29136813e-01 -5.85747719e-01 3.60207930e-02 1.15372375e-01 5.85811973e-01 2.27123886e-01 9.20291305e-01 2.22325385e-01 -1.54921472e-01 1.28196204e+00 -1.02164674e+00 -7.34473914e-02 4.42141891e-01 -3.84830773e-01 1.75380856e-01 2.74912447e-01 -5.45613170e-01 8.80409718e-01 -7.87086248e-01 6.92160666e-01 1.02507186e+00 3.51798207e-01 1.49317741e-01 -1.26199210e+00 -6.05438352e-01 -4.60108906e-01 -4.88426834e-02 -1.62364745e+00 -9.34095979e-02 4.92838383e-01 -6.21781468e-01 1.04019690e+00 4.30796891e-01 5.97238421e-01 5.13548374e-01 9.86503363e-01 3.97344112e-01 9.64344800e-01 -5.87840199e-01 5.00784278e-01 4.72356737e-01 6.58177555e-01 5.82267523e-01 4.94586945e-01 2.10197657e-01 -2.88826048e-01 -2.57527590e-01 5.14669120e-01 -1.13469856e-02 4.11837101e-01 -6.11136317e-01 -9.19365287e-01 9.55965459e-01 2.68737793e-01 8.28033328e-01 -3.14706802e-01 -2.04630360e-01 4.60875481e-01 6.63690031e-01 3.63124609e-01 1.06763101e+00 -6.27033949e-01 -1.29847735e-01 -8.97177219e-01 2.91861892e-01 1.03879309e+00 9.66879725e-01 6.09781981e-01 -1.79220647e-01 -9.21863504e-03 7.12664962e-01 1.18650898e-01 3.24553728e-01 3.99027646e-01 -3.63262773e-01 3.23804736e-01 6.39528275e-01 1.05660573e-01 -1.11958241e+00 -7.26042628e-01 -5.66994309e-01 -6.85551822e-01 5.07390976e-01 3.21109354e-01 1.44321591e-01 -8.37969720e-01 9.47326541e-01 6.98022619e-02 -6.48150921e-01 1.94385678e-01 5.66991687e-01 3.91861022e-01 7.23247409e-01 -3.90328228e-01 -5.88382542e-01 1.26318896e+00 -4.17011738e-01 -7.85600543e-01 2.54212558e-01 6.37155473e-01 -8.66413832e-01 3.50339711e-01 1.16164374e+00 -8.07388425e-01 -7.52570033e-01 -1.37275231e+00 4.05260623e-01 -6.12590671e-01 5.89612901e-01 9.29889321e-01 1.00959086e+00 -6.29679084e-01 8.22479784e-01 -4.26817000e-01 -6.71084642e-01 1.87492192e-01 9.10102725e-01 -3.95865500e-01 -2.66296804e-01 -6.25226736e-01 7.97478914e-01 6.96093440e-01 -9.64419637e-03 -2.49024719e-01 -3.24071527e-01 -6.65811896e-01 -2.97756810e-02 1.92771643e-01 -2.40065791e-02 7.01776743e-01 -6.23073936e-01 -1.45827937e+00 4.12419677e-01 6.18928522e-02 -2.17982292e-01 1.35502905e-01 1.54645830e-01 -7.90483892e-01 -2.85885155e-01 -2.12519377e-01 1.28967583e-01 9.44335938e-01 -9.01811957e-01 -7.25086927e-01 -5.03692985e-01 -6.02305770e-01 -3.04905653e-01 -4.56832796e-01 -1.60833225e-02 5.12672625e-02 -3.62934709e-01 6.40890300e-01 -5.03889084e-01 -2.61504203e-01 -4.16729331e-01 -2.83128232e-01 -4.65247631e-01 7.30174541e-01 -2.37894773e-01 1.43533361e+00 -2.07793188e+00 1.22820586e-01 9.24998879e-01 -1.51400007e-02 -3.84742115e-03 -2.74422653e-02 5.70295215e-01 -2.61608601e-01 -2.02364311e-01 7.47041404e-03 3.07170451e-01 -1.62702531e-01 -8.20250735e-02 4.52834442e-02 5.52112579e-01 1.98142961e-01 2.75720805e-01 -6.74531341e-01 -2.79229224e-01 3.68878096e-01 -1.35171533e-01 -2.67146111e-01 -4.04319391e-02 1.45374358e-01 -2.22868681e-01 -5.70812225e-01 1.00137353e+00 7.40070224e-01 -2.88445093e-02 2.86656827e-01 -3.52168679e-01 -4.68409359e-02 6.19539358e-02 -1.33773184e+00 1.13645577e+00 -3.95598620e-01 4.94065642e-01 -3.13307345e-01 -1.20656669e+00 1.95969236e+00 4.34828214e-02 8.28363657e-01 -4.83070612e-01 4.03543383e-01 5.23005962e-01 4.61343825e-01 -1.76993385e-01 6.53171182e-01 -2.03944102e-01 -2.19609708e-01 9.01547894e-02 1.85210407e-01 -3.36687982e-01 4.03589219e-01 -2.22731292e-01 1.30489182e+00 -5.50279021e-01 5.75971723e-01 -8.99225116e-01 4.98036116e-01 -7.78157189e-02 4.82381165e-01 5.26704609e-01 9.88226235e-02 2.93163955e-01 6.48075104e-01 -4.99681652e-01 -9.92234945e-01 -9.56523657e-01 -5.22455871e-01 3.80172849e-01 1.39565483e-01 -5.42646289e-01 -2.41938621e-01 -6.66773319e-01 4.27167356e-01 5.64842641e-01 -5.23177087e-01 -7.07372278e-02 -1.64690137e-01 -9.87661123e-01 -1.28348976e-01 2.50318021e-01 -6.39353041e-03 -8.13860774e-01 -3.75974000e-01 3.95585865e-01 7.13578582e-01 -7.05536246e-01 3.91348153e-01 8.01634550e-01 -1.16501296e+00 -1.04904556e+00 -3.68279397e-01 -8.83052468e-01 8.19014847e-01 -1.31627977e-01 1.07824337e+00 -3.18561256e-01 -9.48254764e-01 -2.16378346e-01 -3.01349044e-01 -5.41107297e-01 -5.98426938e-01 3.98672044e-01 2.78988242e-01 -2.98386574e-01 5.68998516e-01 -1.22212805e-01 3.05139478e-02 7.65367568e-01 -4.47893679e-01 -6.75833941e-01 8.68437886e-01 9.96847987e-01 4.99590993e-01 9.99749303e-01 8.66896868e-01 -9.19157982e-01 5.09940624e-01 -5.38381696e-01 -1.10286152e+00 1.12808146e-01 -9.17973161e-01 1.54483989e-01 3.89900982e-01 -1.42216191e-01 -7.46164203e-01 5.69603622e-01 -2.16773838e-01 -6.14765659e-02 3.87284271e-02 2.58629262e-01 -2.01151818e-01 -1.79688022e-01 7.85862386e-01 -1.25201821e-01 4.34446111e-02 -6.35218859e-01 -1.60345823e-01 7.30870128e-01 -4.26830091e-02 -3.27474833e-01 7.58007765e-01 -1.49272352e-01 4.84570920e-01 -9.39069211e-01 -1.21339880e-01 -7.35289216e-01 -7.58106947e-01 -2.17124283e-01 5.87738872e-01 -2.72987664e-01 -7.34444022e-01 2.87392944e-01 -8.00044477e-01 3.85347426e-01 -2.68039823e-01 6.70553803e-01 -5.42332351e-01 -1.77153274e-02 -3.52700949e-01 -1.04178596e+00 6.04682341e-02 -1.06703925e+00 9.68466878e-01 7.86960721e-02 -3.59490454e-01 -7.67960012e-01 -2.05548421e-01 1.77652389e-02 1.09559000e-01 -6.89760670e-02 1.20988607e+00 -9.12906289e-01 -4.08117324e-01 -7.99965024e-01 -1.67163044e-01 2.48849571e-01 5.24774730e-01 1.04783699e-01 -6.29409075e-01 -6.32209718e-01 5.88316508e-02 -1.97078198e-01 9.41332221e-01 4.81702685e-01 1.23690414e+00 4.91092324e-01 -9.00264025e-01 9.94974002e-02 1.71015263e+00 8.39810729e-01 8.04482222e-01 1.14429398e-02 7.91343376e-02 6.59279227e-01 1.42236209e+00 4.37049925e-01 -6.32952571e-01 9.30675805e-01 3.08085978e-01 -1.32791162e-01 3.91272277e-01 4.01030779e-01 -2.18123868e-02 6.03323579e-01 1.70627728e-01 -3.83020520e-01 -6.64304018e-01 3.09536278e-01 -1.58389306e+00 -5.45526505e-01 -1.64979696e-01 2.31484461e+00 1.89937830e-01 5.89401662e-01 2.18108743e-01 5.98979354e-01 5.87668657e-01 -4.34565842e-01 -2.24564552e-01 -8.17221642e-01 -6.16886429e-02 3.18094641e-01 6.16221607e-01 2.91362703e-01 -1.17674470e+00 4.93817627e-01 7.45268726e+00 1.22435737e+00 -9.01698947e-01 -3.25405657e-01 7.25998104e-01 9.05268267e-02 1.79718941e-01 -5.63529670e-01 -1.20371807e+00 3.05578709e-01 9.35481906e-01 -1.35906190e-01 -2.12047547e-01 1.10023773e+00 -4.77456450e-01 -5.74546635e-01 -1.34793210e+00 1.07514560e+00 3.73472944e-02 -1.36263323e+00 -2.76403934e-01 4.01282102e-01 5.96112430e-01 -6.15523815e-01 1.17024809e-01 1.35677665e-01 2.35017017e-03 -1.02199602e+00 2.19982266e-01 3.75356972e-01 5.43448687e-01 -1.10304129e+00 1.20010388e+00 2.75115483e-02 -1.04385698e+00 -4.79605675e-01 -7.32256114e-01 9.45686102e-02 -2.84092247e-01 1.11062264e+00 -1.14548063e+00 7.93297172e-01 4.42799151e-01 6.96876407e-01 -3.36149901e-01 1.27796090e+00 5.85095525e-01 1.13265380e-01 -4.27918255e-01 -5.44040561e-01 -8.52367654e-02 -1.11911036e-01 3.34372342e-01 1.22362494e+00 8.15774918e-01 -2.40693271e-01 8.83076638e-02 5.08871019e-01 4.40086693e-01 3.58723640e-01 -8.56171966e-01 1.43681709e-02 3.29351544e-01 1.11688447e+00 -1.27362609e+00 -2.74385028e-02 -3.88549864e-02 6.01763189e-01 -1.36700869e-01 -2.78513670e-01 -3.89083833e-01 -7.96754718e-01 5.07421315e-01 1.55587077e-01 6.13602996e-01 -6.22517951e-02 -2.98823684e-01 -3.08096856e-01 -4.96713668e-02 -4.73437011e-01 4.92580682e-01 -1.54292911e-01 -1.57312727e+00 7.44969130e-01 8.11515003e-02 -1.49908519e+00 -4.58361983e-01 -1.34142745e+00 -5.97178861e-02 8.15142751e-01 -8.73073995e-01 -3.62220556e-01 5.36044352e-02 2.30503470e-01 6.44090414e-01 -9.49112177e-01 9.56758499e-01 2.77887940e-01 -3.09177816e-01 5.35130143e-01 4.67418432e-01 -3.84871840e-01 5.02967000e-01 -1.15817297e+00 -4.34154011e-02 1.75018877e-01 -1.94254577e-01 3.81464660e-01 8.14121127e-01 -6.38688505e-01 -1.51522338e+00 -7.77401924e-01 9.24932003e-01 -1.57196671e-01 5.96296668e-01 -5.40564001e-01 -6.60180807e-01 2.73913424e-03 -2.47525692e-01 -9.04984921e-02 5.98814726e-01 4.90826666e-01 1.50737658e-01 -6.76770747e-01 -1.55714750e+00 -7.83263221e-02 5.60859084e-01 6.22112118e-02 -5.08667491e-02 5.65794587e-01 8.47382173e-02 -2.55512327e-01 -1.24662232e+00 4.66830015e-01 5.48327684e-01 -8.68196905e-01 7.76564360e-01 -2.14964077e-01 2.84732878e-01 -1.74754724e-01 -2.66508996e-01 -1.12512374e+00 -5.21721900e-01 -3.25558662e-01 3.21700990e-01 1.07090974e+00 6.65117264e-01 -6.59269571e-01 1.08192217e+00 3.16041261e-02 7.59774726e-03 -1.07687390e+00 -1.12425280e+00 -1.30782628e+00 -2.77729958e-01 -1.20999649e-01 4.30213362e-01 6.32710516e-01 1.66858301e-01 2.66429573e-01 -1.29573226e-01 -3.57944593e-02 5.01007199e-01 2.36627907e-01 3.96375090e-01 -1.88882661e+00 -3.04327369e-01 -3.08606505e-01 -1.41783237e+00 -6.13222122e-01 -3.46424244e-02 -8.11187387e-01 2.03299969e-01 -1.16231084e+00 3.89755927e-02 -1.01727915e+00 -6.74375236e-01 -1.96235301e-03 3.70917797e-01 2.03094870e-01 -5.41637391e-02 -2.52709687e-02 -2.42786139e-01 6.38148412e-02 8.57304752e-01 -2.46742964e-01 -2.67992556e-01 4.55674350e-01 -2.78026670e-01 3.18934381e-01 7.51047611e-01 -5.81153631e-01 -5.04124224e-01 3.41230452e-01 1.81790203e-01 2.01031566e-01 -2.54725754e-01 -1.47080946e+00 -6.91792592e-02 1.10143483e-01 7.63255596e-01 -8.69211137e-01 4.88105536e-01 -1.07896042e+00 2.20511615e-01 7.69500196e-01 -1.16706863e-01 -7.88033009e-04 1.15186393e-01 4.40760285e-01 -2.37743586e-01 -7.29143858e-01 7.25900412e-01 2.17686906e-01 -7.25675464e-01 -1.53043091e-01 -4.65175241e-01 -9.72401857e-01 1.40391445e+00 -4.73016381e-01 -9.90959406e-02 2.16456875e-02 -8.58199418e-01 -1.99378550e-01 2.67369121e-01 3.74298573e-01 5.57515681e-01 -1.30493474e+00 -5.26105404e-01 4.44846153e-01 4.73770738e-01 -7.32477903e-01 -1.25301361e-01 3.69389236e-01 -4.93851602e-01 7.05837965e-01 -4.25526351e-01 -8.12875867e-01 -1.58061886e+00 6.54692709e-01 1.30536228e-01 -4.31809664e-01 -3.33221316e-01 8.41575623e-01 -2.85716444e-01 -1.95461169e-01 1.22530028e-01 -5.54872155e-01 -2.43942603e-01 2.06314564e-01 4.31986451e-01 4.54835176e-01 6.81176186e-01 -2.56232738e-01 -4.53028500e-01 7.18349159e-01 -2.66434520e-01 2.07614183e-01 1.22654092e+00 3.75113875e-01 -1.22360423e-01 7.24456012e-01 1.39800906e+00 -1.62078470e-01 -5.67858219e-01 1.16163410e-01 3.41397583e-01 -7.27145076e-01 9.79862064e-02 -8.47777128e-01 -9.61151123e-01 4.02754366e-01 1.03942394e+00 6.63076997e-01 1.35099030e+00 1.69121269e-02 7.88353384e-02 8.61217499e-01 8.01143467e-01 -1.21077418e+00 8.94739330e-02 3.89411569e-01 8.28688920e-01 -1.22961366e+00 5.34788609e-01 -1.03861952e+00 -2.24428445e-01 1.53844833e+00 5.14728844e-01 -1.48928359e-01 1.03837037e+00 5.54643273e-01 -2.72952557e-01 -2.71369874e-01 -7.60069549e-01 1.86079860e-01 1.45359993e-01 7.02415049e-01 4.77712810e-01 6.06836006e-02 -7.54797935e-01 3.71254712e-01 -1.29944250e-01 -1.12723194e-01 3.06328595e-01 1.07403362e+00 -5.95716774e-01 -1.42589033e+00 -4.68756497e-01 9.08610284e-01 -9.23634917e-02 3.13404799e-01 -6.82168230e-02 1.06676102e+00 1.84285820e-01 1.00205445e+00 4.73068953e-01 -6.36427939e-01 3.46581072e-01 -7.58647621e-02 7.27556348e-01 -6.19608283e-01 -4.65762317e-01 5.06464764e-02 2.05387622e-01 -6.52262926e-01 5.74387014e-02 -8.51279080e-01 -1.05760956e+00 -1.04720779e-02 -6.05770469e-01 3.75790149e-01 1.14047718e+00 6.52262330e-01 1.74314645e-03 7.25181997e-01 1.11327124e+00 -5.79438567e-01 -3.42103422e-01 -8.69530261e-01 -1.40602994e+00 -1.76894426e-01 3.45870815e-02 -8.15480053e-01 2.30249595e-02 -4.63786364e-01]

[8.240446090698242, 4.236546516418457]

21fd3f9c-f26b-4c72-a880-9583cd4c4098

neural-architecture-search-as-multiobjective

2208.04321

https://arxiv.org/abs/2208.04321v2

https://arxiv.org/pdf/2208.04321v2.pdf

Neural Architecture Search as Multiobjective Optimization Benchmarks: Problem Formulation and Performance Assessment

The ongoing advancements in network architecture design have led to remarkable achievements in deep learning across various challenging computer vision tasks. Meanwhile, the development of neural architecture search (NAS) has provided promising approaches to automating the design of network architectures for lower prediction error. Recently, the emerging application scenarios of deep learning have raised higher demands for network architectures considering multiple design criteria: number of parameters/floating-point operations, and inference latency, among others. From an optimization point of view, the NAS tasks involving multiple design criteria are intrinsically multiobjective optimization problems; hence, it is reasonable to adopt evolutionary multiobjective optimization (EMO) algorithms for tackling them. Nonetheless, there is still a clear gap confining the related research along this pathway: on the one hand, there is a lack of a general problem formulation of NAS tasks from an optimization point of view; on the other hand, there are challenges in conducting benchmark assessments of EMO algorithms on NAS tasks. To bridge the gap: (i) we formulate NAS tasks into general multi-objective optimization problems and analyze the complex characteristics from an optimization point of view; (ii) we present an end-to-end pipeline, dubbed $\texttt{EvoXBench}$, to generate benchmark test problems for EMO algorithms to run efficiently -- without the requirement of GPUs or Pytorch/Tensorflow; (iii) we instantiate two test suites comprehensively covering two datasets, seven search spaces, and three hardware devices, involving up to eight objectives. Based on the above, we validate the proposed test suites using six representative EMO algorithms and provide some empirical analyses. The code of $\texttt{EvoXBench}$ is available from $\href{https://github.com/EMI-Group/EvoXBench}{\rm{here}}$.

['Kalyanmoy Deb', 'Kay Chen Tan', 'Yaochu Jin', 'Ran Cheng', 'Zhichao Lu']

2022-08-08

['multiobjective-optimization']

['methodology']

[ 1.27306581e-01 -4.27804202e-01 2.09154829e-01 -3.15797418e-01 -3.93459409e-01 -2.53960073e-01 -6.77479059e-02 -7.05830976e-02 -4.80170071e-01 6.38718843e-01 -5.50370693e-01 -4.71893102e-01 -7.23391414e-01 -5.38503587e-01 -5.86831927e-01 -7.51125693e-01 2.52150986e-02 1.69742808e-01 -2.07841501e-01 -2.39325523e-01 3.20175439e-01 3.57003510e-01 -1.76382446e+00 5.25493957e-02 9.90403771e-01 1.54099834e+00 2.24650323e-01 4.32272136e-01 1.73659310e-01 4.18114990e-01 -5.72363019e-01 -7.82277167e-01 1.17253035e-01 -7.33256191e-02 -6.51355922e-01 -2.96441108e-01 1.93514824e-02 9.75635052e-02 -1.01247653e-01 1.17704713e+00 9.38982427e-01 2.61569262e-01 3.44965011e-01 -1.47428429e+00 -5.41013777e-01 3.25603962e-01 -4.40969229e-01 3.74544829e-01 -3.59518230e-01 2.09247887e-01 1.19115269e+00 -8.86723578e-01 1.88257009e-01 9.07027125e-01 5.45732260e-01 5.04490018e-01 -7.55895615e-01 -5.92972934e-01 3.41804057e-01 5.74353337e-01 -1.34328091e+00 -4.61995810e-01 9.94995415e-01 -1.22838661e-01 1.08713484e+00 5.51882267e-01 6.51157022e-01 1.27958655e+00 8.48178342e-02 8.01308990e-01 9.83431756e-01 -1.86587766e-01 4.48723614e-01 8.60732123e-02 3.44896108e-01 8.26880395e-01 2.42796555e-01 1.90338776e-01 -5.15101016e-01 7.99716935e-02 2.46701896e-01 -3.48951072e-01 -1.45109192e-01 -9.23292059e-03 -7.40833402e-01 6.90244615e-01 2.27573052e-01 1.91218764e-01 -4.21198934e-01 1.90786213e-01 7.35699832e-01 2.31006339e-01 2.60568500e-01 5.09609520e-01 -4.74107683e-01 -3.05272818e-01 -5.32774985e-01 1.38091013e-01 7.48020709e-01 7.64891207e-01 5.04633844e-01 6.35086119e-01 -5.97539730e-02 1.01523268e+00 2.14231804e-01 1.45057961e-01 4.13142294e-01 -8.08684170e-01 5.09692788e-01 5.54227233e-01 -2.86262631e-01 -1.16232693e+00 -5.52012324e-01 -1.03648794e+00 -9.85665917e-01 1.57167077e-01 -3.98994014e-02 -5.03608227e-01 -2.86807328e-01 1.82735288e+00 3.24981153e-01 -5.61522413e-03 -5.97645491e-02 9.18999135e-01 1.08220267e+00 4.38438296e-01 -3.42733860e-02 -2.35218052e-02 1.45939648e+00 -1.21261775e+00 -3.38456780e-01 -2.85854131e-01 4.37183022e-01 -5.67037046e-01 1.08020329e+00 4.65409577e-01 -9.86591399e-01 -6.06009245e-01 -1.28901374e+00 1.43577963e-01 -4.51361269e-01 5.55496633e-01 7.17514038e-01 1.01299644e+00 -1.04852796e+00 5.19463360e-01 -7.91299462e-01 1.04304990e-02 4.46671456e-01 6.04481936e-01 1.97323397e-01 3.55274260e-01 -1.06328034e+00 6.91523790e-01 2.97818840e-01 6.42134905e-01 -7.78648674e-01 -6.04293168e-01 -5.29855371e-01 4.03910249e-01 4.99559850e-01 -7.79952586e-01 8.15426052e-01 -1.08735943e+00 -1.68863642e+00 5.72308362e-01 8.38830769e-02 -1.25377789e-01 4.17627841e-01 -8.72137472e-02 -5.56845367e-01 -2.37424299e-01 -4.83075619e-01 4.87846255e-01 7.89960086e-01 -1.08049655e+00 -4.46575344e-01 -3.75648469e-01 1.91524923e-01 1.77320436e-01 -9.22945559e-01 2.81549871e-01 -5.26254117e-01 -6.49760067e-01 -4.59641218e-01 -9.90227938e-01 -8.42485577e-02 -1.85246870e-01 -6.37764573e-01 -5.79271652e-02 6.67563975e-01 -3.82773340e-01 1.40246892e+00 -2.02724290e+00 4.44492877e-01 2.32518554e-01 3.58965486e-01 5.34107447e-01 -2.27011666e-01 -1.27599075e-01 -2.18340874e-01 1.24867059e-01 -8.09778869e-02 -5.32266259e-01 3.63524824e-01 -1.12284847e-01 6.16332106e-02 1.40354499e-01 1.75462022e-01 7.81527400e-01 -4.39313859e-01 -3.15061986e-01 2.13495612e-01 6.77021801e-01 -5.85114360e-01 5.42799272e-02 -2.01208383e-01 1.34899735e-01 -7.87831247e-01 8.82561266e-01 4.44037288e-01 -4.46026355e-01 7.75644854e-02 -4.91169840e-01 -3.10576688e-02 -1.19416572e-01 -1.28068328e+00 1.34718895e+00 -4.67418253e-01 6.55294120e-01 2.22467408e-01 -1.37781954e+00 1.02674592e+00 2.06203520e-01 5.80536246e-01 -6.96138263e-01 7.62488127e-01 3.41882944e-01 3.05321276e-01 -5.10259867e-01 4.46030140e-01 5.20318627e-01 5.69242910e-02 2.27077946e-01 4.87044305e-02 4.74377424e-01 2.44199678e-01 -5.85431993e-01 1.13301957e+00 9.46791992e-02 -1.43449232e-01 -2.69653589e-01 6.02856576e-01 -4.48387302e-02 6.59902692e-01 4.23071235e-01 -3.24232280e-01 1.52609423e-01 3.42773765e-01 -5.57884634e-01 -7.82086432e-01 -8.30049932e-01 -1.54246986e-01 1.26473510e+00 8.39920044e-02 -2.91041553e-01 -8.88234138e-01 -3.31757933e-01 -4.03570056e-01 6.71432555e-01 -4.17174757e-01 -1.76677018e-01 -3.92087936e-01 -1.44063747e+00 7.72501290e-01 3.63352418e-01 7.37326026e-01 -1.04698062e+00 -1.17153573e+00 -3.74298752e-03 -2.20451504e-02 -1.13810980e+00 -1.32609621e-01 2.35811129e-01 -7.74612129e-01 -8.45024645e-01 -4.35363203e-01 -5.58008373e-01 4.84724134e-01 -5.18886484e-02 1.23673761e+00 4.04348701e-01 -5.38945198e-01 2.07794055e-01 -2.02292949e-01 -5.57222903e-01 -3.79525349e-02 3.05898279e-01 4.79572080e-02 4.28712666e-02 3.17099661e-01 -7.82062888e-01 -6.71498358e-01 5.70931673e-01 -6.39369845e-01 1.72606528e-01 5.70844352e-01 7.88388610e-01 6.12870157e-01 1.40229881e-01 6.12150252e-01 -3.13512385e-01 7.33559430e-01 -4.41295981e-01 -7.94336200e-01 5.74170291e-01 -6.11949801e-01 -1.86098814e-02 7.64147878e-01 -2.66092032e-01 -8.84217262e-01 -4.58595514e-01 -4.01792020e-01 -5.71369350e-01 2.54324004e-02 6.61612451e-01 -2.55358905e-01 -1.45741835e-01 4.36593026e-01 2.14233190e-01 -1.29603446e-01 -3.73208016e-01 -2.24637657e-01 5.72638869e-01 1.53601930e-01 -9.49992657e-01 2.92376012e-01 4.66900319e-02 1.50821775e-01 -7.96399951e-01 -5.46051741e-01 2.54067630e-01 -1.20724235e-02 -4.93574917e-01 7.87883162e-01 -5.85857868e-01 -1.28104496e+00 3.59365880e-01 -9.33290005e-01 -2.71749109e-01 2.93694377e-01 3.87323618e-01 -3.22175562e-01 8.78346860e-02 -3.40920657e-01 -7.80046642e-01 -9.26934779e-01 -1.84809709e+00 5.85499167e-01 5.07366598e-01 -1.46592081e-01 -9.66198623e-01 -4.22735661e-01 5.89403033e-01 6.10662878e-01 2.35969588e-01 1.20287395e+00 -5.24126053e-01 -3.81224811e-01 8.23787525e-02 -2.61478722e-01 4.36188728e-01 -2.85648704e-01 3.17004383e-01 -1.00654209e+00 -2.85470396e-01 7.27272183e-02 -4.13987428e-01 3.38542640e-01 4.43300754e-01 1.78770244e+00 1.50764182e-01 -1.21260226e-01 8.97864938e-01 1.43483126e+00 5.14533579e-01 4.93247509e-01 5.76803863e-01 5.28789163e-01 4.87025678e-01 2.97633439e-01 6.95756197e-01 4.12011772e-01 7.54634261e-01 6.55674696e-01 3.95745821e-02 2.21849471e-01 4.09176707e-01 2.16491818e-01 8.65651548e-01 -3.24995041e-01 -6.44130945e-01 -8.59604597e-01 2.01465443e-01 -1.55251026e+00 -6.10270321e-01 1.40584186e-01 1.92595947e+00 3.65106583e-01 1.27844065e-01 -6.84762448e-02 4.31025811e-02 7.62705505e-01 1.62628114e-01 -9.90658283e-01 -6.76128149e-01 1.55828688e-02 3.18664819e-01 5.74105084e-02 1.56701449e-02 -9.21595037e-01 4.64589268e-01 4.33648825e+00 1.12667477e+00 -1.50711226e+00 1.99451029e-01 1.09027803e+00 -4.43845719e-01 -1.11276940e-01 -3.23616564e-01 -9.44837153e-01 6.27258718e-01 8.02752852e-01 -3.63808572e-02 6.71826303e-01 9.25291002e-01 1.13032989e-01 1.46985680e-01 -9.89251971e-01 1.25323653e+00 -9.50182155e-02 -1.31276822e+00 -3.16018164e-01 -7.23341182e-02 4.72769558e-01 2.23735750e-01 4.47313100e-01 5.40074766e-01 7.30704842e-03 -9.89589810e-01 7.87781000e-01 1.17601790e-01 6.30805910e-01 -7.22855926e-01 5.45723021e-01 3.84169780e-02 -1.17732131e+00 -2.84858286e-01 -2.29803726e-01 1.96289808e-01 -8.07262808e-02 5.54573894e-01 -1.10397965e-01 8.78850460e-01 1.07294667e+00 3.11698198e-01 -5.82116365e-01 9.65910733e-01 1.67851910e-01 3.83453846e-01 -2.62271971e-01 -4.92672801e-01 1.95362613e-01 -3.20283949e-01 6.82145238e-01 7.98489034e-01 6.36199534e-01 -1.34487584e-01 -3.46806914e-01 9.91125822e-01 -2.16336489e-01 6.88752830e-02 3.76171097e-02 -2.55775750e-02 4.69647914e-01 1.51753581e+00 -7.04292417e-01 2.46771455e-01 -3.11474264e-01 6.46281064e-01 4.34418142e-01 2.45155454e-01 -1.43359315e+00 -4.41902429e-01 8.03837717e-01 -6.28815889e-01 1.45240605e-01 -5.08902222e-02 -6.13002539e-01 -1.05964863e+00 3.27067852e-01 -1.16344893e+00 2.92201579e-01 -6.41470551e-01 -9.95499969e-01 1.08852983e+00 -1.65097609e-01 -9.56311822e-01 1.78589836e-01 -8.55870187e-01 -6.74171090e-01 7.19064236e-01 -1.31907415e+00 -6.97883248e-01 -6.03654981e-01 2.67849356e-01 6.11691892e-01 -5.62118411e-01 8.24242175e-01 7.08341300e-01 -1.36503625e+00 9.56043720e-01 2.16459915e-01 -1.16214812e-01 1.01802319e-01 -6.57606244e-01 2.22946659e-01 6.00602031e-01 -1.17363378e-01 4.12259221e-01 6.38899446e-01 -8.53782743e-02 -1.76261389e+00 -7.67799735e-01 2.39331648e-01 -4.04482037e-02 5.47121465e-01 -2.94505417e-01 -7.05450654e-01 2.60660738e-01 1.61684826e-01 -5.25105000e-02 7.29794860e-01 2.28176728e-01 1.03776924e-01 -2.79533982e-01 -1.00806689e+00 7.30648279e-01 1.10716689e+00 -1.81390159e-02 1.44567251e-01 9.08989608e-02 3.52386385e-01 -4.81813997e-01 -9.96192515e-01 5.83596528e-01 5.96006513e-01 -1.11429477e+00 1.02007723e+00 -3.70457977e-01 6.32502019e-01 -3.83637510e-02 -3.25703025e-01 -1.16819119e+00 4.56192940e-02 -4.92304116e-01 -1.86097160e-01 1.17777181e+00 5.38886249e-01 -7.70594180e-01 8.68386865e-01 7.90458560e-01 -4.55020159e-01 -1.65679967e+00 -1.08201313e+00 -5.98755419e-01 -2.03350365e-01 -5.04883409e-01 6.58106267e-01 7.05614686e-01 -5.14606357e-01 2.44728163e-01 -2.27852702e-01 2.25597210e-02 4.48700517e-01 6.92567602e-02 3.56594443e-01 -9.62465823e-01 -6.89361572e-01 -8.97949278e-01 -1.17475763e-01 -6.80297792e-01 1.64095744e-01 -6.75291538e-01 -3.45805347e-01 -9.97858763e-01 -7.40559474e-02 -7.66495287e-01 -4.53410685e-01 4.04577464e-01 -6.98651597e-02 -1.17465919e-02 2.74697632e-01 -1.35610566e-01 -6.32744133e-01 9.94010508e-01 9.36745465e-01 -6.46134466e-02 -3.28235328e-02 4.30317111e-02 -6.80234373e-01 7.08727300e-01 7.95987248e-01 -3.80481780e-01 -4.95180756e-01 -1.03459251e+00 5.71424425e-01 -5.37453517e-02 4.56925154e-01 -1.26991522e+00 1.79031610e-01 -2.37443228e-03 6.78405613e-02 -2.53695995e-01 4.83797103e-01 -8.01540792e-01 3.44680071e-01 3.76801193e-01 -7.13970885e-02 6.65312111e-01 4.82397169e-01 2.34387234e-01 -9.85379964e-02 -4.85171765e-01 7.13014007e-01 2.57451206e-01 -9.55426097e-01 4.19705182e-01 -2.14533731e-01 1.37703270e-01 1.06944799e+00 -3.44430238e-01 -4.40711379e-01 1.99860588e-01 -6.23655438e-01 4.62891191e-01 7.67664425e-03 4.62896913e-01 5.02538383e-01 -9.55377102e-01 -4.55518425e-01 6.71995878e-02 -1.70553923e-01 -1.82796985e-01 6.46730423e-01 1.00135481e+00 -5.20764291e-01 2.82583445e-01 -4.65860754e-01 -4.34553444e-01 -1.32561886e+00 2.10340187e-01 7.59950697e-01 -3.34641904e-01 5.60807772e-02 1.09407830e+00 -1.33456782e-01 -4.28890735e-01 3.30522507e-01 1.52044937e-01 -1.42706349e-01 4.07608822e-02 1.71152666e-01 7.25223303e-01 4.27082330e-01 -3.32231134e-01 -4.88473654e-01 4.40129161e-01 8.03607702e-02 4.31530505e-01 1.62477076e+00 6.96131960e-02 -8.12167600e-02 -9.51015204e-02 1.29250276e+00 -5.30451715e-01 -9.66451466e-01 3.73429246e-02 -7.85807222e-02 -1.16652496e-01 2.91000217e-01 -7.73119569e-01 -1.56547606e+00 8.80226552e-01 8.31508279e-01 2.09981967e-02 1.50110102e+00 -4.66712534e-01 6.08537793e-01 5.61581194e-01 3.15695614e-01 -1.30127323e+00 1.18393518e-01 5.30347586e-01 7.25730479e-01 -1.12540066e+00 -5.69587946e-02 -2.65180677e-01 -5.32497585e-01 1.11896157e+00 9.39633191e-01 2.29433447e-01 6.22796416e-01 2.50582635e-01 -2.59023875e-01 -4.14722264e-01 -7.65518486e-01 1.22215472e-01 2.70430833e-01 1.53492600e-01 3.37259531e-01 -1.01308651e-01 -2.42516875e-01 9.89468932e-01 -1.58077314e-01 -3.25130597e-02 -9.03463215e-02 7.33309090e-01 -1.21808155e-02 -9.96652365e-01 -2.26359367e-01 5.34396231e-01 -4.82492596e-01 -2.86532864e-02 -3.10486406e-02 5.69019616e-01 1.69360131e-01 9.52420294e-01 -1.12265326e-01 -6.98354006e-01 1.72534153e-01 -1.22026272e-01 2.48565853e-01 8.83497857e-03 -8.40503335e-01 -3.82797480e-01 3.50443125e-01 -2.92910874e-01 -1.34942293e-01 -5.07225752e-01 -9.52727795e-01 -6.00352049e-01 -2.47319356e-01 8.43024403e-02 8.98399711e-01 1.00492191e+00 7.71745086e-01 1.06473780e+00 4.34618801e-01 -8.74243975e-01 -7.79331326e-01 -6.25368953e-01 -1.87758908e-01 -1.17863379e-01 -3.46379936e-01 -9.83547807e-01 -9.28703174e-02 -5.27031183e-01]

[8.270698547363281, 3.2147719860076904]

646f6bbf-ce99-441e-bb52-0436d69f83e8

norm-of-word-embedding-encodes-information

2212.09663

https://arxiv.org/abs/2212.09663v2

https://arxiv.org/pdf/2212.09663v2.pdf

Norm of Word Embedding Encodes Information Gain

Distributed representations of words encode lexical semantic information, but what type of information is encoded, and how? Focusing on the skip-gram with negative-sampling method, we found that the squared norm of static word embedding encodes the information gain conveyed by the word; the information gain is defined by the Kullback-Leibler divergence of the co-occurrence distribution of the word to the unigram distribution of the corpus. Our findings are explained by the theoretical framework of the exponential family of probability distributions and confirmed through precise experiments that remove spurious correlations arising from word frequency. We demonstrate that both the KL divergence and the squared norm of embedding provide a useful metric of a word's informativeness in tasks such as keyword extraction, part-of-speech discrimination, and hypernym classification.

['Hidetoshi Shimodaira', 'Sho Yokoi', 'Momose Oyama']

2022-12-19

['keyword-extraction']

['natural-language-processing']

[ 2.19863802e-01 3.26079458e-01 -4.55873996e-01 -4.41477507e-01 -7.35532820e-01 -6.97148204e-01 4.92524922e-01 6.54227257e-01 -8.06723297e-01 4.37296987e-01 6.74832404e-01 -2.91507900e-01 -1.52364343e-01 -8.19572747e-01 -3.28790635e-01 -7.51196027e-01 -3.10420662e-01 2.77514964e-01 9.31656137e-02 -1.84701949e-01 2.79722005e-01 1.79123908e-01 -1.61554539e+00 4.01723720e-02 3.41391265e-01 1.08357263e+00 1.24484457e-01 4.37756032e-01 -4.25065488e-01 3.45428646e-01 -7.40143299e-01 -3.64362180e-01 4.59308363e-02 -4.37873363e-01 -7.56398320e-01 -3.53372604e-01 -7.95520172e-02 -6.57070354e-02 -4.10690933e-01 1.31007409e+00 1.99306265e-01 2.77057827e-01 9.41016376e-01 -1.02740836e+00 -8.88031185e-01 5.89647830e-01 -7.14066029e-02 4.92655128e-01 4.15414244e-01 -3.49347949e-01 1.86892772e+00 -9.92636561e-01 6.63823783e-01 1.25031781e+00 6.21613264e-01 2.48497948e-01 -1.15529704e+00 -4.17127967e-01 -9.11696702e-02 7.95701668e-02 -1.35581088e+00 -1.62314251e-01 2.79548138e-01 -5.26082039e-01 1.07750273e+00 1.58749819e-01 5.72537601e-01 8.78106833e-01 3.73526692e-01 5.14478683e-01 5.42364419e-01 -7.46479392e-01 3.04530799e-01 3.06873500e-01 2.21860915e-01 6.62398756e-01 6.56987369e-01 5.52157871e-02 -6.79510772e-01 -4.49127495e-01 3.29266638e-01 -1.03259146e-01 -2.57986814e-01 -3.64845753e-01 -8.89001310e-01 1.23472524e+00 -3.72703224e-02 6.01916552e-01 -3.10532957e-01 4.21349406e-01 5.33546269e-01 1.92263409e-01 5.80194712e-01 4.77584153e-01 -5.84842801e-01 -2.96131164e-01 -3.86286765e-01 8.26010630e-02 9.32009280e-01 7.39785314e-01 7.52771616e-01 -1.49488226e-01 5.51599562e-02 8.13024402e-01 5.74312031e-01 6.02567136e-01 8.69404495e-01 -4.85791653e-01 1.37411848e-01 3.82148325e-01 -2.20749766e-01 -1.10008717e+00 3.80943865e-02 -8.82371441e-02 -9.39696133e-02 -2.92399377e-01 2.71135420e-01 -1.11523882e-01 -3.21169257e-01 2.13171887e+00 9.17986259e-02 -4.92624551e-01 -2.54424065e-02 5.37898481e-01 4.47912395e-01 5.47119915e-01 3.00316483e-01 -3.05636346e-01 1.41941214e+00 -2.22288489e-01 -8.98079574e-01 -3.56505483e-01 6.43605173e-01 -5.63664138e-01 9.61539567e-01 -1.84011117e-01 -8.43313277e-01 -8.92074257e-02 -1.10199118e+00 1.16113806e-02 -6.25192165e-01 -5.12079477e-01 6.78655386e-01 7.69905031e-01 -7.08156049e-01 5.55809498e-01 -4.47921604e-01 -2.93721110e-01 8.08998421e-02 -8.71646106e-02 -6.88890219e-01 1.72606081e-01 -1.44117522e+00 1.01591468e+00 5.84896207e-01 -6.13067985e-01 -3.45595986e-01 -6.71585083e-01 -1.05172372e+00 3.62412453e-01 2.79730987e-02 -2.10559562e-01 1.12974763e+00 -8.34441543e-01 -9.30085361e-01 7.73669302e-01 -2.32623622e-01 -2.90847123e-01 -4.02199268e-01 2.00890452e-01 -4.17961031e-01 2.63389885e-01 1.24289438e-01 1.82589903e-01 8.06114614e-01 -7.34971344e-01 -4.15919632e-01 -5.85864902e-01 -2.82213002e-01 8.14950466e-02 -7.01113641e-01 -1.54223725e-01 2.37451583e-01 -6.67646587e-01 2.57387340e-01 -4.76406038e-01 1.20707475e-01 1.49807662e-01 1.20352469e-01 -3.24102461e-01 2.98550487e-01 -7.52201915e-01 1.45833349e+00 -2.52116680e+00 -3.08657587e-01 4.21625406e-01 6.67484403e-02 1.82146933e-02 -3.01969647e-01 8.56512845e-01 -1.68914735e-01 1.73866346e-01 -2.02259138e-01 3.37919682e-01 3.47609580e-01 1.54918835e-01 -3.87045741e-01 6.73005402e-01 7.82719851e-02 6.80256605e-01 -1.17770386e+00 -2.29100063e-01 -1.52496845e-01 5.46629131e-01 -3.34396809e-01 -1.46718428e-01 -1.29966103e-02 -5.38438380e-01 -2.33659089e-01 7.03270584e-02 2.38669410e-01 -1.45324364e-01 3.92981410e-01 -1.17075302e-01 3.72347951e-01 8.75141978e-01 -9.15524840e-01 1.35795462e+00 -3.96922588e-01 8.63793850e-01 -2.68584967e-01 -8.55905771e-01 8.85539770e-01 3.41224432e-01 2.64050066e-01 -7.62306213e-01 1.70987807e-02 2.06637755e-01 1.36777654e-01 -3.23821515e-01 6.42901838e-01 -5.12209833e-01 -1.62879035e-01 5.95242918e-01 2.56206810e-01 1.92048848e-01 5.55380434e-02 1.67024612e-01 1.19842792e+00 -3.22674215e-01 8.19546878e-01 -4.84981865e-01 -7.43266940e-02 -3.52386653e-01 2.54478008e-01 5.43765962e-01 -1.53848410e-01 1.55712008e-01 7.46765792e-01 4.11617421e-02 -1.05864477e+00 -1.43799341e+00 -5.96415460e-01 1.26851308e+00 3.03958263e-03 -7.84523785e-01 -4.83293980e-01 -6.11741543e-01 4.40192103e-01 1.10979128e+00 -6.86528265e-01 -6.42260909e-01 8.80067050e-02 -6.52255416e-01 7.37147808e-01 5.65573692e-01 -2.73029000e-01 -7.64836669e-01 -5.22459805e-01 -6.89285621e-03 -1.31748945e-01 -8.25323343e-01 -8.56195569e-01 5.32946408e-01 -6.01613879e-01 -1.10476184e+00 -1.62700817e-01 -7.16433764e-01 4.96489018e-01 2.05386095e-02 1.12041700e+00 -3.76081467e-01 -6.07461214e-01 7.37207115e-01 -4.06250089e-01 -2.05873460e-01 -4.91744041e-01 -3.12911063e-01 -9.97000113e-02 -1.97419718e-01 1.04558253e+00 -4.58457083e-01 -5.59736669e-01 -3.81776430e-02 -9.95924652e-01 -8.13301742e-01 4.26861435e-01 9.13257182e-01 4.59964305e-01 1.77361071e-02 5.74906647e-01 -5.81786215e-01 9.84749317e-01 -5.82861602e-01 -1.54544443e-01 2.39568681e-01 -7.98410892e-01 4.94900078e-01 4.47414555e-02 -3.78633469e-01 -7.26378262e-01 -3.66133302e-01 -8.43849331e-02 1.74018517e-01 1.96977034e-01 5.93239486e-01 3.06044538e-02 3.55985820e-01 4.99707758e-01 2.56298482e-01 5.60084730e-02 -3.11282575e-01 6.49034798e-01 9.43726361e-01 1.80919230e-01 -3.92535657e-01 1.92216799e-01 3.36726278e-01 -3.12000439e-02 -1.18427622e+00 -7.53116071e-01 -8.21030259e-01 -3.80548596e-01 4.46427554e-01 9.40343559e-01 -6.43363595e-01 -6.34535968e-01 -2.79456079e-01 -1.14838290e+00 3.85430694e-01 -7.04064548e-01 6.77681804e-01 -3.70255113e-01 5.23910701e-01 -4.14140522e-01 -1.08158743e+00 -3.16286325e-01 -5.60502946e-01 1.04057872e+00 -3.19212526e-01 -7.80758142e-01 -1.35978818e+00 3.60306621e-01 -2.51954675e-01 3.42691004e-01 -2.02196285e-01 1.60251820e+00 -1.32252491e+00 1.22435682e-01 -8.00231636e-01 -2.01668724e-01 6.02389038e-01 3.03926796e-01 -2.87450820e-01 -8.77046347e-01 -5.76994494e-02 2.55848151e-02 -1.01493374e-01 9.77375865e-01 3.00621599e-01 8.75968575e-01 -5.17299294e-01 -1.00469626e-01 1.26051635e-01 1.39611351e+00 2.99687497e-02 5.55429399e-01 -1.32889987e-03 1.88325450e-01 6.46811545e-01 2.41107360e-01 5.24577975e-01 7.99214244e-02 3.37008864e-01 1.50460273e-01 8.76072347e-01 5.55235036e-02 -7.31307685e-01 3.82942051e-01 1.06928682e+00 6.11421704e-01 -2.09340081e-01 -7.36834347e-01 7.95762658e-01 -1.40645528e+00 -1.03877997e+00 2.05097780e-01 2.68620849e+00 9.96656775e-01 9.83454436e-02 -2.75683366e-02 2.06244439e-01 5.61970413e-01 4.23161954e-01 -7.36570507e-02 -4.97548729e-01 -6.11967556e-02 4.09869701e-01 7.15416908e-01 8.47066879e-01 -6.35036409e-01 6.18084252e-01 7.35919285e+00 1.11491740e+00 -6.64982259e-01 8.07430074e-02 9.15804431e-02 1.44944504e-01 -8.65739822e-01 -4.96150702e-02 -6.49932742e-01 5.29132128e-01 1.19079566e+00 -5.40831566e-01 2.14724958e-01 8.28002274e-01 -3.92711252e-01 -1.95793539e-01 -1.26276636e+00 7.92276800e-01 1.26832291e-01 -8.59812856e-01 1.16116874e-01 2.29551584e-01 1.44886434e-01 -3.25730741e-02 4.32339236e-02 8.45169574e-02 9.12721828e-02 -9.16197598e-01 5.31285226e-01 3.02028298e-01 8.86390805e-01 -6.89382017e-01 5.68905234e-01 1.80398524e-01 -1.07941186e+00 2.21200138e-02 -6.01234376e-01 5.51285371e-02 -1.47052035e-01 8.21185708e-01 -1.01213801e+00 -1.75019443e-01 1.95441335e-01 3.13889623e-01 -2.07273453e-01 6.21351719e-01 -3.12665850e-01 5.97340345e-01 -2.38029137e-01 -5.23622215e-01 -5.50074168e-02 -1.44156128e-01 6.70468569e-01 1.36915791e+00 2.79683411e-01 -8.43892768e-02 -2.19638705e-01 8.79872024e-01 -7.69111514e-02 2.74368852e-01 -8.44674885e-01 -8.63511682e-01 8.86236668e-01 9.51420665e-01 -5.36091268e-01 -1.96658805e-01 -3.88298839e-01 8.65620196e-01 1.59259081e-01 1.84873283e-01 -1.60350382e-01 -9.43238437e-01 1.11887980e+00 -5.69427479e-03 6.27702177e-01 -1.27459809e-01 -2.08403558e-01 -7.53330410e-01 1.80403963e-01 -4.27954853e-01 3.46523672e-01 -4.69842553e-01 -1.64665997e+00 3.49019587e-01 2.04911917e-01 -5.87128222e-01 -5.37575901e-01 -9.47928786e-01 -2.38679886e-01 1.06238878e+00 -9.44973350e-01 -2.29643732e-01 4.91802543e-01 7.45557845e-02 1.41375825e-01 -2.07558498e-01 1.32333457e+00 3.89338750e-03 7.92213902e-02 6.96255863e-01 3.34824860e-01 3.12368691e-01 3.72079790e-01 -1.35283935e+00 3.78880888e-01 1.66772500e-01 5.18377066e-01 9.18562889e-01 8.16203237e-01 -6.26026273e-01 -1.34127748e+00 -5.42034805e-01 1.35619020e+00 -3.69714916e-01 1.02935696e+00 -2.20638663e-01 -7.73114502e-01 5.39781213e-01 -1.37088716e-01 -3.03946878e-03 1.12605059e+00 3.59586775e-01 -1.02417421e+00 5.24727479e-02 -1.15528679e+00 3.67253453e-01 7.51632214e-01 -1.16449380e+00 -8.71823251e-01 3.14130217e-01 1.11444008e+00 3.02556187e-01 -7.86386311e-01 -7.76227983e-03 7.51084626e-01 -6.52590096e-01 8.56066823e-01 -9.45650041e-01 2.27950811e-01 2.74499208e-01 -9.87648308e-01 -1.39026630e+00 -3.03563386e-01 -1.47408232e-01 1.02362908e-01 8.05311799e-01 4.95897114e-01 -6.85024202e-01 3.46617758e-01 5.16341984e-01 3.31662804e-01 -6.83432519e-01 -1.13638663e+00 -8.17872643e-01 1.26594365e-01 -5.08765697e-01 3.35810155e-01 6.06662869e-01 6.93742156e-01 5.05025327e-01 2.76110590e-01 -2.26953536e-01 5.66196799e-01 -3.75653505e-01 -9.69469249e-02 -1.19877195e+00 -5.01809895e-01 -3.17895323e-01 -9.77828383e-01 -9.58975315e-01 2.96650887e-01 -1.14123750e+00 1.09211907e-01 -1.11099124e+00 3.10464799e-01 2.38846943e-01 -2.56912380e-01 1.80588081e-01 -1.51989283e-02 -1.56586051e-01 -6.65174574e-02 -1.33017421e-01 -3.37014169e-01 7.16721535e-01 6.62266791e-01 8.26622546e-03 2.29449525e-01 -2.76647747e-01 -7.06801295e-01 8.38004887e-01 4.22410995e-01 -6.14212096e-01 -5.04496992e-01 -2.63024598e-01 7.33883381e-01 -1.05921470e-01 2.13757262e-01 -2.67445683e-01 -1.72794580e-01 4.38854136e-02 4.28040102e-02 -2.15801716e-01 4.22863096e-01 -8.41665149e-01 -4.55406100e-01 4.20436174e-01 -9.01715219e-01 1.77843794e-01 -5.19205891e-02 9.82834220e-01 -2.73534566e-01 -7.85167694e-01 5.23827970e-01 -8.45623612e-02 -5.01131713e-01 -8.82299244e-02 -5.31514466e-01 5.61749995e-01 7.28136361e-01 -1.35871703e-02 -2.92851835e-01 -5.57137311e-01 -4.00244206e-01 -4.26809132e-01 2.12916031e-01 2.72642463e-01 6.59877598e-01 -1.45574486e+00 -5.41005492e-01 3.41131657e-01 3.34478468e-01 -9.48919475e-01 -2.95178175e-01 5.20334959e-01 -1.69789135e-01 6.18035734e-01 2.27351949e-01 4.03427295e-02 -1.05196440e+00 3.70849431e-01 1.19652919e-01 1.13363072e-01 -4.06492442e-01 7.88428485e-01 2.53305256e-01 -2.54882574e-01 3.04389775e-01 -2.44877443e-01 1.90092012e-01 4.24605459e-01 7.46114194e-01 3.89395982e-01 1.67414770e-01 -6.33693576e-01 -6.59490347e-01 1.84874371e-01 7.01780547e-04 -4.65370566e-01 1.04196465e+00 -3.50350738e-02 -3.63990098e-01 9.32658315e-01 1.88014495e+00 5.23271002e-02 -5.79196870e-01 -3.54888260e-01 3.13635319e-01 -6.01855516e-01 3.60795707e-02 -5.03742278e-01 -3.79269481e-01 8.35577071e-01 4.97832477e-01 5.97880304e-01 4.96664882e-01 3.84541184e-01 7.09263682e-01 3.74244720e-01 2.80989617e-01 -1.08912849e+00 1.45822940e-02 5.51582813e-01 5.47240078e-01 -8.75813544e-01 -3.99917401e-02 -2.79957503e-01 -5.31767666e-01 9.44820940e-01 -1.98446736e-01 -3.72808911e-02 1.09720612e+00 1.35802329e-01 -2.56602377e-01 -2.22627789e-01 -8.29965055e-01 -3.02867204e-01 4.66966569e-01 7.48040199e-01 6.71168387e-01 1.63439482e-01 -5.62891901e-01 7.53308356e-01 -4.66922581e-01 -5.39321542e-01 1.81695461e-01 6.46387935e-01 -9.11280751e-01 -8.42720211e-01 1.09655559e-01 5.36427855e-01 -5.38953602e-01 -4.36776668e-01 -4.29355234e-01 4.89888161e-01 -2.26947561e-01 8.20626199e-01 3.23257297e-01 -2.90958375e-01 -9.83298123e-02 5.89353740e-01 6.42178833e-01 -8.85229945e-01 -1.15014217e-03 -3.74000102e-01 1.37454018e-01 -3.08947951e-01 6.49618432e-02 -6.03348315e-01 -1.23004043e+00 -6.89966008e-02 -4.59199786e-01 4.08590436e-01 1.01713753e+00 9.71417964e-01 1.70332819e-01 1.59250051e-01 3.92493099e-01 -1.22250363e-01 -1.29525530e+00 -1.02729356e+00 -9.75124300e-01 5.61364830e-01 2.54463345e-01 -3.86357576e-01 -8.08284461e-01 -2.34455645e-01]

[10.434910774230957, 8.782041549682617]

801b43a3-1d8f-4f27-b1cb-81953d953a4c

realistic-data-enrichment-for-robust-image

2304.09534

https://arxiv.org/abs/2304.09534v1

https://arxiv.org/pdf/2304.09534v1.pdf

Realistic Data Enrichment for Robust Image Segmentation in Histopathology

Poor performance of quantitative analysis in histopathological Whole Slide Images (WSI) has been a significant obstacle in clinical practice. Annotating large-scale WSIs manually is a demanding and time-consuming task, unlikely to yield the expected results when used for fully supervised learning systems. Rarely observed disease patterns and large differences in object scales are difficult to model through conventional patient intake. Prior methods either fall back to direct disease classification, which only requires learning a few factors per image, or report on average image segmentation performance, which is highly biased towards majority observations. Geometric image augmentation is commonly used to improve robustness for average case predictions and to enrich limited datasets. So far no method provided sampling of a realistic posterior distribution to improve stability, e.g. for the segmentation of imbalanced objects within images. Therefore, we propose a new approach, based on diffusion models, which can enrich an imbalanced dataset with plausible examples from underrepresented groups by conditioning on segmentation maps. Our method can simply expand limited clinical datasets making them suitable to train machine learning pipelines, and provides an interpretable and human-controllable way of generating histopathology images that are indistinguishable from real ones to human experts. We validate our findings on two datasets, one from the public domain and one from a Kidney Transplant study.

['Bernhard Kainz', 'Candice Roufosse', 'Callum Arthurs', 'James Ball', 'Sarah Cechnicka']

2023-04-19

['whole-slide-images', 'image-augmentation']

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

[ 6.64247990e-01 5.97859740e-01 -1.55453429e-01 -3.82683605e-01 -1.22878492e+00 -6.10855103e-01 4.67175901e-01 7.58980095e-01 -5.09693682e-01 6.79787576e-01 6.49567395e-02 -6.46701813e-01 -1.57535583e-01 -7.22841442e-01 -7.79084921e-01 -9.31985915e-01 1.25197947e-01 9.88192141e-01 2.04846069e-01 2.30534151e-01 2.63946876e-02 5.73819399e-01 -1.01820242e+00 5.14083564e-01 9.25558150e-01 5.25580764e-01 2.22544909e-01 7.85681486e-01 -2.71644086e-01 5.80001593e-01 -5.17383039e-01 -5.07390022e-01 2.36894161e-01 -5.37114859e-01 -6.66903317e-01 2.17147037e-01 4.73168671e-01 -2.10835353e-01 4.43827510e-02 9.18996632e-01 6.83890164e-01 -5.68145812e-01 1.04400945e+00 -9.57421958e-01 -3.77810210e-01 7.25584507e-01 -7.36956179e-01 1.46914944e-01 -2.26218909e-01 6.73094332e-01 7.39421844e-01 -2.19911486e-01 1.10571718e+00 9.42763031e-01 8.17873836e-01 5.04901528e-01 -1.66615772e+00 -5.29258609e-01 -1.02232762e-01 -1.13441646e-01 -1.18271291e+00 -2.12148815e-01 4.01743680e-01 -7.74332523e-01 5.81267655e-01 5.97064197e-01 7.19786406e-01 9.10894990e-01 1.45032302e-01 7.30932713e-01 1.23170102e+00 -3.55625927e-01 2.23682716e-01 3.32566321e-01 -2.31237993e-01 3.05040240e-01 5.08773386e-01 -2.03137562e-01 -5.42541817e-02 -3.03162217e-01 7.25636184e-01 8.00680220e-02 -3.06367904e-01 -4.99511123e-01 -1.59836912e+00 7.71350920e-01 4.69941467e-01 2.50067264e-01 -3.00099373e-01 2.60583069e-02 4.08855438e-01 -3.45688537e-02 6.83123827e-01 5.40503025e-01 -3.39084595e-01 2.25911930e-01 -1.14010918e+00 2.42520854e-01 3.76350969e-01 5.10894179e-01 2.92388737e-01 -4.55279231e-01 -2.55900979e-01 6.36379123e-01 1.92134231e-02 5.39699376e-01 5.40966451e-01 -8.13219726e-01 4.26998846e-02 8.63667846e-01 -6.17547240e-03 -6.86096668e-01 -7.31496036e-01 -5.20800292e-01 -9.65182841e-01 3.21906328e-01 1.09458971e+00 1.08499296e-01 -1.19033885e+00 1.58010042e+00 3.75650644e-01 -1.34743452e-01 -2.25160196e-01 8.44269633e-01 4.62969333e-01 1.41784608e-01 3.84740710e-01 -1.29406720e-01 1.45751095e+00 -4.51538265e-01 -5.56490183e-01 -9.12931263e-02 1.20811212e+00 -6.68065012e-01 1.29785955e+00 3.89265954e-01 -9.98307526e-01 5.85059449e-02 -7.63187647e-01 6.76655769e-02 -4.52894986e-01 2.35585988e-01 7.21368492e-01 6.89800501e-01 -1.04936481e+00 5.50820827e-01 -1.22473943e+00 -5.48393011e-01 1.13118672e+00 4.05254811e-01 -4.56173718e-01 -1.54326707e-01 -6.22439325e-01 1.00207901e+00 2.55400538e-01 8.42916593e-02 -5.15105963e-01 -1.25154066e+00 -7.20448315e-01 -2.42478698e-01 1.91209033e-01 -7.56942272e-01 9.59647238e-01 -1.00223267e+00 -9.88191247e-01 1.29486418e+00 3.70760285e-03 -5.40397286e-01 8.94837260e-01 5.72857678e-01 5.34447953e-02 1.74163848e-01 3.98315862e-02 9.61188972e-01 4.30690914e-01 -1.32517576e+00 -2.15419501e-01 -5.52393556e-01 -4.60075140e-01 3.70567963e-02 -1.12272941e-01 -1.64523587e-01 -2.07934961e-01 -7.04377115e-01 1.57956183e-02 -9.78335381e-01 -8.70348632e-01 4.74967301e-01 -5.04038453e-01 2.36712858e-01 4.75456566e-01 -7.97409236e-01 7.98584461e-01 -1.98182607e+00 -2.62717098e-01 4.49941427e-01 3.89617294e-01 1.70563519e-01 1.62961986e-02 -1.40474960e-01 -1.19120039e-01 5.32151401e-01 -4.47659373e-01 -1.99661136e-01 -3.65517773e-02 3.33120711e-02 5.34766950e-02 8.82510483e-01 4.85015094e-01 9.86304939e-01 -9.68318522e-01 -8.00953925e-01 2.29632840e-01 3.29446048e-01 -7.52137959e-01 6.24585412e-02 -1.64347067e-01 6.87859654e-01 5.41860647e-02 4.54681873e-01 8.10284376e-01 -6.51222825e-01 4.46923643e-01 -2.55046517e-01 1.97776705e-01 3.62737402e-02 -9.97793257e-01 1.45734191e+00 -3.14399868e-01 4.79782313e-01 -4.81654033e-02 -8.49447668e-01 5.73656082e-01 1.85218364e-01 4.21375543e-01 -4.61246014e-01 1.74395088e-02 3.32287550e-01 4.25313026e-01 -5.13013184e-01 -8.98179337e-02 -6.81442142e-01 1.27236515e-01 4.47075963e-01 -2.04317495e-02 -6.23226047e-01 1.54145405e-01 2.21383423e-01 1.24252284e+00 -1.87541202e-01 2.03423753e-01 -4.12054628e-01 5.25602773e-02 4.24997896e-01 5.57189047e-01 7.95046687e-01 -3.09209079e-01 1.06459117e+00 8.72889519e-01 -3.19997668e-01 -1.44010198e+00 -1.00915539e+00 -5.01261830e-01 4.77693528e-01 -1.56939134e-01 9.40383896e-02 -6.20337367e-01 -9.57385421e-01 -3.26414146e-02 3.44592035e-01 -8.82957101e-01 -1.92417074e-02 -2.67010927e-01 -1.34415650e+00 6.34039164e-01 4.45304960e-01 -2.99579084e-01 -7.68071890e-01 -5.52981079e-01 9.28246081e-02 -5.74726425e-02 -8.66210580e-01 -1.53845608e-01 2.70234227e-01 -8.46899629e-01 -1.10162389e+00 -1.13490713e+00 -5.82257867e-01 1.17542076e+00 -1.67412907e-01 1.23075926e+00 2.75173783e-01 -9.00028348e-01 -2.36613974e-01 -1.28938712e-03 -9.37796116e-01 -7.09174573e-01 9.02885571e-02 -1.92000359e-01 -1.48730829e-01 3.91459078e-01 -2.00702384e-01 -1.04783678e+00 1.93563655e-01 -1.22983682e+00 2.30312005e-01 1.02459872e+00 1.03121865e+00 8.89499009e-01 -2.30747283e-01 5.85095704e-01 -1.40743792e+00 1.22074649e-01 -4.03173387e-01 -4.19557244e-01 3.21074903e-01 -4.93663520e-01 -2.39381157e-02 4.52289701e-01 -5.07302761e-01 -9.25389886e-01 2.24067226e-01 -1.52032152e-01 -1.74891904e-01 -4.31723326e-01 3.57030183e-01 1.59708321e-01 7.38479197e-02 9.92086172e-01 -1.23029612e-01 6.24812245e-01 -1.19072244e-01 8.57973099e-02 5.12305260e-01 3.56238931e-01 -2.75529772e-01 5.59032559e-01 9.71059501e-01 9.76884589e-02 -6.51394129e-01 -7.06230640e-01 -5.77234745e-01 -5.53068995e-01 -1.84845719e-02 7.46463478e-01 -8.26569676e-01 -4.23742503e-01 3.79684508e-01 -6.69335663e-01 -9.46323276e-01 -4.60193932e-01 4.90924776e-01 -4.07694846e-01 2.10634783e-01 -6.73738718e-01 -4.27626044e-01 -1.93445131e-01 -1.34429455e+00 1.26050818e+00 8.52535143e-02 -6.30594730e-01 -1.15156400e+00 1.51747018e-01 3.85064811e-01 5.42529762e-01 6.37223363e-01 1.08881807e+00 -8.80801141e-01 -4.26079959e-01 -3.62554103e-01 -3.94979358e-01 8.81082863e-02 1.46883592e-01 4.97168362e-01 -9.20986116e-01 -2.10122094e-01 -3.69167447e-01 -4.68574196e-01 8.81454051e-01 6.87957227e-01 1.49076545e+00 -1.06178403e-01 -5.65382540e-01 4.25289840e-01 1.40512657e+00 -2.56999046e-01 6.13781035e-01 8.41507688e-02 5.14417470e-01 9.01982546e-01 4.58935499e-01 1.40966162e-01 2.74883062e-01 5.05058408e-01 2.84542978e-01 -7.27101982e-01 -3.24160725e-01 -5.75514548e-02 -3.56014341e-01 2.02790469e-01 1.15531452e-01 -1.80657864e-01 -1.33189917e+00 9.75142419e-01 -1.36906791e+00 -6.89283729e-01 -3.90846997e-01 2.06055450e+00 1.10344553e+00 3.13635558e-01 7.39935413e-02 7.37585351e-02 6.31567359e-01 -3.32486600e-01 -4.90803152e-01 -1.96340084e-02 -1.68598607e-01 1.86306417e-01 7.89925992e-01 3.39926511e-01 -7.82368839e-01 4.98524338e-01 6.43222094e+00 8.77782941e-01 -1.17404091e+00 2.90123355e-02 1.37721717e+00 -2.23509997e-01 -6.35125816e-01 -8.58200341e-03 -4.49176103e-01 4.68475640e-01 8.59238923e-01 1.16159268e-01 -1.99066401e-01 4.91985172e-01 5.32969415e-01 -2.97143042e-01 -1.09313917e+00 7.34849334e-01 -2.50042200e-01 -1.58603871e+00 1.33845016e-01 4.44949508e-01 7.67454624e-01 -8.13999400e-02 3.32107782e-01 -2.71769110e-02 4.43509996e-01 -1.32912219e+00 2.78071791e-01 5.31441271e-01 1.05876100e+00 -3.83842856e-01 1.29969227e+00 2.63278216e-01 -2.88751364e-01 3.57956648e-01 -3.51098388e-01 3.92597318e-01 -3.67561472e-03 9.73363340e-01 -1.50639141e+00 2.07441181e-01 5.19150674e-01 3.47445577e-01 -8.63795280e-01 1.18860173e+00 3.16968530e-01 7.85250783e-01 -5.05331337e-01 1.47714972e-01 -9.62939207e-03 9.06972364e-02 1.06971882e-01 1.47348917e+00 2.74469793e-01 -6.14200197e-02 -2.55349606e-01 6.76334023e-01 -6.08804822e-03 3.33952546e-01 -5.35268605e-01 -7.67132044e-02 1.91102237e-01 1.59934485e+00 -1.27170122e+00 -3.21549296e-01 -2.75036395e-01 4.29027796e-01 2.40656048e-01 1.95004970e-01 -5.80030084e-01 -5.01920357e-02 2.66855270e-01 4.93447632e-01 1.53882499e-03 3.20769459e-01 -7.14901686e-01 -9.57130253e-01 -2.15535656e-01 -9.81500387e-01 5.70511103e-01 -5.16171217e-01 -1.53603303e+00 3.52432549e-01 -2.74529845e-01 -1.26847887e+00 -8.44516829e-02 -7.60863066e-01 -5.18069685e-01 9.20417011e-01 -1.51754975e+00 -1.21884155e+00 -4.72492129e-01 1.50743276e-01 6.40356541e-02 4.07792360e-01 8.08209538e-01 2.26478130e-01 -3.96952063e-01 7.40778625e-01 5.23415580e-02 4.49371859e-02 1.12571490e+00 -1.56791174e+00 -5.37983477e-02 4.25657451e-01 -2.81963110e-01 4.92739081e-01 9.45212662e-01 -5.25385141e-01 -8.66354048e-01 -1.02051282e+00 6.80374801e-01 -7.86634624e-01 6.00637078e-01 -2.59250671e-01 -1.17042065e+00 5.88106871e-01 -1.89866543e-01 3.91679585e-01 1.24816167e+00 -1.73506439e-01 1.62990913e-02 4.81937304e-02 -1.43054223e+00 6.81391776e-01 6.35917008e-01 -1.54701531e-01 -8.83933455e-02 5.07473350e-01 2.18017668e-01 -5.73140800e-01 -1.09873712e+00 5.23567080e-01 3.47470820e-01 -7.63519049e-01 7.69409299e-01 -7.59730101e-01 4.72141773e-01 -3.35604191e-01 2.17686281e-01 -1.31597233e+00 -1.36637334e-02 -9.16621312e-02 3.86665553e-01 1.07918870e+00 8.80363822e-01 -4.50030118e-01 1.16710186e+00 7.88250148e-01 -7.04536156e-04 -8.24163198e-01 -7.16637015e-01 -3.89498502e-01 4.63082463e-01 -2.84031630e-01 5.47196507e-01 9.79280472e-01 1.48847224e-02 -3.08046222e-01 2.54110903e-01 1.72947705e-01 6.60670996e-01 -5.65318167e-02 8.15041006e-01 -1.13038015e+00 -3.23432952e-01 -7.19745815e-01 -6.34983063e-01 -2.88213611e-01 -1.03221282e-01 -1.13408935e+00 6.08231276e-02 -1.56633615e+00 7.04524517e-01 -8.31858099e-01 -3.56924124e-02 3.19531560e-01 -4.50228781e-01 7.82202244e-01 -2.76127666e-01 1.88906923e-01 -3.62249225e-01 5.92540354e-02 1.51261497e+00 -3.04768115e-01 1.51362307e-02 -1.32180661e-01 -9.40007687e-01 8.42001855e-01 7.33722866e-01 -5.67896962e-01 -3.34578454e-01 -8.74857306e-02 3.28670651e-01 -7.31012970e-02 5.80149949e-01 -7.63097882e-01 1.34213185e-02 -1.83483124e-01 8.02991211e-01 -4.45552081e-01 -1.26611009e-01 -7.04811871e-01 3.65392238e-01 6.59691691e-01 -4.90051150e-01 -4.27062720e-01 2.28034779e-01 4.35120642e-01 -9.89575833e-02 -1.24761455e-01 8.82060468e-01 -3.66111606e-01 5.13982736e-02 3.26905549e-01 -3.17303717e-01 5.55876270e-02 1.01853871e+00 -1.92847490e-01 -4.95695382e-01 -2.61927396e-01 -8.40559304e-01 2.24830031e-01 7.97887623e-01 -1.97028667e-01 1.91126794e-01 -1.02009130e+00 -1.01447761e+00 -6.36725798e-02 3.03999603e-01 4.94673550e-01 6.31630003e-01 1.26707280e+00 -9.54418242e-01 2.04910591e-01 -3.68084535e-02 -1.03422058e+00 -1.07937717e+00 5.42730331e-01 4.09818411e-01 -7.95034647e-01 -4.46898013e-01 7.11519480e-01 5.19663572e-01 -6.72774732e-01 -1.44621506e-01 -4.79279727e-01 1.73386075e-02 2.64816694e-02 5.48459172e-01 -1.50769964e-01 3.21914852e-01 -3.27413410e-01 -2.74932504e-01 2.68173277e-01 -4.00472432e-01 3.29192281e-02 1.27378261e+00 -2.14300733e-02 -9.11652818e-02 4.35792089e-01 1.01479614e+00 -6.52830824e-02 -1.39041424e+00 -6.71083853e-02 6.22820621e-03 -5.35086274e-01 -7.59591116e-03 -8.58132362e-01 -8.74065459e-01 8.24701369e-01 6.62170827e-01 2.40485772e-01 8.61497521e-01 2.00825438e-01 3.55191380e-01 -1.90750927e-01 5.67726493e-02 -8.55978191e-01 -4.23755571e-02 -3.20351899e-01 6.22676790e-01 -1.51693308e+00 3.19077261e-02 -4.56444174e-01 -7.02880919e-01 8.73244584e-01 4.42660093e-01 2.09895089e-01 3.43038350e-01 5.30035019e-01 4.58182812e-01 -2.51982868e-01 -7.80606270e-01 -1.47968763e-02 1.50834963e-01 7.27708280e-01 4.96644557e-01 6.84963018e-02 -2.17039913e-01 3.81358534e-01 -1.12242095e-01 6.25822246e-02 7.83365488e-01 7.59891152e-01 -8.99643600e-02 -1.01769280e+00 -3.58082801e-01 1.02285683e+00 -8.83979499e-01 5.43285832e-02 -1.69988930e-01 9.17050362e-01 1.55952066e-01 4.05038685e-01 3.49122167e-01 1.93955392e-01 1.22551076e-01 -5.01201116e-02 4.53686535e-01 -6.40690386e-01 -3.37882936e-01 4.65685204e-02 -1.41952753e-01 -2.08239004e-01 -4.04087752e-01 -8.27944100e-01 -1.17757964e+00 -2.64589161e-01 -3.28088433e-01 -1.38205796e-01 6.03689075e-01 8.25925171e-01 1.86560929e-01 6.84072196e-01 1.83805093e-01 -7.05332041e-01 -3.36035639e-01 -8.86315763e-01 -5.21477520e-01 7.15698838e-01 2.49777883e-01 -3.02850515e-01 -4.73591417e-01 3.17406893e-01]

[14.997896194458008, -2.799880027770996]

55faf512-12c1-47f8-88bf-2c11a38b3982

robust-federated-learning-with-noisy

1911.00251

https://arxiv.org/abs/1911.00251v1

https://arxiv.org/pdf/1911.00251v1.pdf

Robust Federated Learning with Noisy Communication

Federated learning is a communication-efficient training process that alternates between local training at the edge devices and averaging the updated local model at the central server. Nevertheless, it is impractical to achieve a perfect acquisition of the local models in wireless communication due to noise, which also brings serious effects on federated learning. To tackle this challenge, we propose a robust design for federated learning to alleviate the effects of noise in this paper. Considering noise in the two aforementioned steps, we first formulate the training problem as a parallel optimization for each node under the expectation-based model and the worst-case model. Due to the non-convexity of the problem, a regularization for the loss function approximation method is proposed to make it tractable. Regarding the worst-case model, we develop a feasible training scheme which utilizes the sampling-based successive convex approximation algorithm to tackle the unavailable maxima or minima noise condition and the non-convex issue of the objective function. Furthermore, the convergence rates of both new designs are analyzed from a theoretical point of view. Finally, the improvement of prediction accuracy and the reduction of loss function are demonstrated via simulations for the proposed designs.

['Yunfei Chen', 'Weidong Wang', 'Fan Ang', 'Li Chen', 'Nan Zhao', 'F. Richard Yu']

2019-11-01

['robust-design']

['miscellaneous']

[ 7.32192583e-03 9.32375193e-02 -1.37245730e-01 -2.06606627e-01 -9.28042650e-01 -1.33409083e-01 -1.61015272e-01 -4.96582547e-03 -3.46189648e-01 9.09845352e-01 -1.02398917e-01 -3.93248677e-01 -5.50800383e-01 -7.68241405e-01 -8.15452635e-01 -1.09491467e+00 -7.55363563e-03 -2.00364426e-01 -4.10845339e-01 2.76579529e-01 -3.69759202e-01 2.45342746e-01 -1.01603019e+00 -2.75163472e-01 1.21212661e+00 1.54797626e+00 3.22407812e-01 2.13633910e-01 -1.03090271e-01 6.37895644e-01 -4.30161655e-01 -5.05131543e-01 3.99393231e-01 -4.01033580e-01 -3.07067156e-01 2.73026496e-01 -8.72640684e-03 -4.70508337e-01 -3.80399227e-01 1.39290798e+00 7.45085895e-01 9.23192054e-02 8.15260038e-02 -1.10039568e+00 1.55286223e-01 4.56677169e-01 -6.53266549e-01 -5.68910502e-02 -1.27089545e-01 -3.44509721e-01 8.23689163e-01 -8.08948994e-01 2.48624906e-01 6.29773080e-01 9.19691741e-01 3.35314482e-01 -7.53835380e-01 -7.59885013e-01 3.95526797e-01 1.68707073e-01 -1.46461761e+00 -5.43622196e-01 8.77868235e-01 5.24831638e-02 6.90687969e-02 3.54932249e-01 4.30322468e-01 6.01639450e-01 -1.25294194e-01 9.01785433e-01 9.33151662e-01 -4.83297020e-01 4.38925117e-01 3.83614063e-01 -1.20388411e-01 7.12264061e-01 4.05376792e-01 -7.19862282e-02 -3.90057236e-01 -4.63016689e-01 4.94352669e-01 3.40834439e-01 -4.47817326e-01 -3.60498846e-01 -6.71553969e-01 4.26200360e-01 4.52087224e-01 2.25348294e-01 -7.44018793e-01 -2.33718492e-02 3.44806433e-01 2.09452793e-01 7.43544579e-01 -4.07555223e-01 -6.32384300e-01 -1.62334647e-02 -1.22480178e+00 -2.44285613e-01 9.43467915e-01 9.51429427e-01 5.36191165e-01 1.92904755e-01 -2.07567081e-01 7.01921284e-01 3.66774023e-01 5.28681397e-01 9.49140042e-02 -6.25259221e-01 9.28126752e-01 2.13870108e-01 3.51352513e-01 -1.13798904e+00 -2.04980090e-01 -1.32799220e+00 -1.09447956e+00 -2.80374318e-01 5.02986193e-01 -7.30951846e-01 -1.41207457e-01 1.82019603e+00 7.26398587e-01 5.50524950e-01 6.43549040e-02 1.01534688e+00 3.30323994e-01 6.55353367e-01 -7.61572123e-02 -9.82887864e-01 8.71719778e-01 -9.31455374e-01 -8.88839781e-01 1.38554662e-01 8.00744712e-01 -4.13252205e-01 4.81093407e-01 3.36283147e-01 -1.09770739e+00 -1.06876820e-01 -1.16635132e+00 5.94534039e-01 1.66073889e-01 5.86513400e-01 3.31914425e-01 8.87025833e-01 -7.61693954e-01 4.54940319e-01 -7.64605820e-01 2.53919195e-02 5.12738407e-01 4.38813806e-01 3.00023879e-04 -1.10809959e-01 -9.54063475e-01 2.43335620e-01 1.80972051e-02 4.97468233e-01 -5.66600621e-01 -8.27378094e-01 -3.45780194e-01 3.42826039e-01 4.79520768e-01 -6.37515485e-01 1.12074041e+00 -1.03235269e+00 -1.53076410e+00 1.87815636e-01 -3.40212673e-01 -4.49860096e-01 9.93434608e-01 -6.49531484e-02 -3.44695717e-01 -7.31615582e-03 -2.14512408e-01 -6.33798480e-01 8.72426331e-01 -1.10529363e+00 -8.21690679e-01 -5.37055194e-01 3.57837975e-02 3.24531913e-01 -8.75012517e-01 -3.83752942e-01 -4.85736340e-01 -7.44398415e-01 1.74148694e-01 -6.24384820e-01 -5.36895514e-01 1.17032260e-01 -1.95024341e-01 2.24460736e-01 7.75246024e-01 -8.19287598e-01 1.41256249e+00 -2.32321405e+00 -3.25308770e-01 5.49624979e-01 2.32863829e-01 1.96130574e-01 2.60850221e-01 2.95272440e-01 3.06284040e-01 -1.79767400e-01 -2.07257885e-02 -6.58043265e-01 -1.79249540e-01 7.42541552e-02 -1.07871890e-01 7.18600988e-01 -5.18615663e-01 3.39710325e-01 -8.37508917e-01 -5.64755142e-01 9.53695327e-02 5.32895029e-01 -4.59291577e-01 2.56189913e-01 7.66025186e-02 5.23286343e-01 -1.00499916e+00 3.77934635e-01 1.05374897e+00 -2.56181151e-01 5.01900136e-01 -4.54509169e-01 -7.02719986e-02 -8.13311636e-02 -1.45751238e+00 1.66819727e+00 -1.10674906e+00 -4.18880954e-02 1.01053643e+00 -1.43400586e+00 7.24432647e-01 5.80591738e-01 9.34312344e-01 -5.01088321e-01 1.78009674e-01 4.46834087e-01 -5.61686635e-01 -3.18034351e-01 -9.22358781e-02 -1.97292730e-01 2.90663481e-01 3.88733834e-01 -1.06271029e-01 7.35434890e-01 -3.15361500e-01 -6.76337779e-02 9.81331229e-01 -1.27308175e-01 2.21683532e-01 -2.63133794e-01 7.27244556e-01 -4.38140213e-01 9.71341312e-01 6.39689445e-01 -1.99956596e-01 7.56627694e-02 -3.46585661e-02 -2.40959600e-01 -5.32175303e-01 -6.17264450e-01 -4.96489964e-02 8.42909992e-01 3.45858961e-01 -3.46524119e-01 -7.55471110e-01 -7.57107675e-01 -1.27703190e-01 5.41802466e-01 -1.80916101e-01 -1.21701650e-01 -2.14319855e-01 -1.06165981e+00 2.77904004e-01 1.65550932e-01 8.25314581e-01 -1.34088203e-01 -5.25009334e-01 2.81080812e-01 -2.66971737e-01 -1.04327035e+00 -5.08386910e-01 3.60961378e-01 -9.43415523e-01 -8.03627372e-01 -8.85548592e-01 -7.66198754e-01 9.05261278e-01 4.80882674e-01 5.25108099e-01 1.01986945e-01 9.96100828e-02 4.54521060e-01 -2.39133731e-01 -2.49428988e-01 -2.32268050e-02 -2.72775181e-02 7.77089298e-02 5.57742059e-01 -1.99240670e-01 -7.38990963e-01 -7.63364315e-01 3.52658272e-01 -5.16583920e-01 -8.10620412e-02 6.29157186e-01 9.77585256e-01 7.09740698e-01 6.53419137e-01 7.81158447e-01 -7.26226449e-01 4.98507679e-01 -6.99313462e-01 -7.09788561e-01 5.36551118e-01 -6.44837081e-01 -1.19651213e-01 1.13878417e+00 -2.49819070e-01 -1.41257334e+00 3.39045912e-01 3.63790430e-02 -4.62238222e-01 3.91844779e-01 5.40769935e-01 -4.93135571e-01 -4.25083727e-01 2.66664028e-01 4.20090139e-01 -9.17539839e-03 -4.67579216e-01 1.95961937e-01 8.85960340e-01 3.34157526e-01 -6.80732131e-01 1.06898880e+00 6.81057334e-01 2.31894925e-01 -8.54425848e-01 -9.63058233e-01 -4.32290703e-01 -1.23213299e-01 -3.23898345e-01 1.09929420e-01 -1.08820558e+00 -8.45137537e-01 1.57024190e-01 -9.72135782e-01 1.45864069e-01 -3.12498063e-01 6.47565603e-01 -4.01878297e-01 5.56368589e-01 -3.62358212e-01 -1.21293521e+00 -8.78792882e-01 -8.22223842e-01 4.78493452e-01 3.08216572e-01 4.76862699e-01 -1.11077893e+00 -3.03195387e-01 1.94643483e-01 5.55530608e-01 2.71861106e-01 4.68192488e-01 -3.92136276e-01 -4.52924103e-01 -4.09379691e-01 -1.36005312e-01 3.23556930e-01 1.76989764e-01 -5.82914531e-01 -8.28429937e-01 -6.35424256e-01 5.74829757e-01 -2.15427373e-02 3.19045395e-01 3.88340890e-01 1.41242671e+00 -6.86718822e-01 -4.94804204e-01 8.71278822e-01 1.65302598e+00 5.44026569e-02 -1.26113696e-02 7.10679591e-02 3.73352617e-01 2.25060746e-01 5.88094652e-01 1.16552591e+00 2.96410590e-01 6.04875922e-01 3.96043450e-01 -1.21159650e-01 2.80023694e-01 -2.30832562e-01 1.47051141e-01 1.01351213e+00 2.76175767e-01 -1.52129591e-01 -3.97241682e-01 2.78979123e-01 -2.04398417e+00 -8.05384457e-01 1.12294868e-01 2.54036355e+00 6.29756868e-01 -9.85274091e-02 -1.43956795e-01 2.88795948e-01 8.71493757e-01 -1.09657198e-02 -7.73794115e-01 6.78712176e-03 -1.82489567e-02 -2.25479424e-01 7.69384742e-01 1.67845890e-01 -8.92170012e-01 1.05110995e-01 5.15862751e+00 1.12157333e+00 -1.27766073e+00 4.91449237e-01 8.18763018e-01 -1.87430471e-01 -2.77475595e-01 5.52780963e-02 -5.07231593e-01 7.24717617e-01 7.09001899e-01 -3.76255363e-01 6.81534469e-01 9.61090207e-01 5.90561748e-01 8.02418217e-02 -6.24906361e-01 1.30164504e+00 -2.69030184e-01 -1.04411244e+00 -4.22137618e-01 1.91350877e-02 7.26722002e-01 -4.88780960e-02 -8.88951495e-02 7.28671029e-02 -1.55294135e-01 -4.51000094e-01 6.90228164e-01 5.50724864e-01 6.36779249e-01 -8.89649093e-01 6.26624584e-01 8.00310969e-01 -1.32315862e+00 -4.48393196e-01 -3.30569893e-01 -5.91547228e-02 1.08083270e-01 1.19023561e+00 -3.06862265e-01 1.01165605e+00 5.68934977e-01 4.60353017e-01 7.88250640e-02 1.32432938e+00 1.90878585e-01 8.08749020e-01 -6.85575008e-01 4.38215770e-02 5.92360180e-03 -4.30416673e-01 7.18253732e-01 7.76467800e-01 5.39665222e-01 3.20828683e-03 3.45946103e-01 5.62875628e-01 -2.93299198e-01 3.63230556e-01 -1.59732610e-01 4.62633759e-01 7.17449546e-01 1.41932571e+00 -2.26910695e-01 -4.83415509e-03 -7.01368093e-01 8.11740935e-01 3.01397651e-01 5.82579732e-01 -8.36503685e-01 -2.58411765e-01 2.31904298e-01 -3.57530154e-02 2.35967010e-01 -6.49791062e-02 -3.72492462e-01 -1.12336516e+00 6.84408545e-01 -5.74820876e-01 3.73224974e-01 5.47274649e-02 -1.12540412e+00 3.99215370e-01 -3.69191855e-01 -1.40111125e+00 3.48902009e-02 1.19030304e-01 -8.43106806e-01 7.15381265e-01 -1.59445095e+00 -8.79677296e-01 -4.15813684e-01 6.90847635e-01 2.14917317e-01 4.10161689e-02 6.38366044e-01 8.96152794e-01 -8.93417716e-01 1.14802885e+00 8.67805302e-01 -2.40015328e-01 2.73867518e-01 -6.00639760e-01 -3.93465132e-01 8.73304129e-01 -1.90838426e-01 3.63913119e-01 5.64174831e-01 -2.84131646e-01 -1.63718307e+00 -1.22657442e+00 6.01259172e-01 5.93578458e-01 3.92260909e-01 -2.04126760e-01 -5.01502275e-01 1.26826495e-01 -3.45426887e-01 6.21378243e-01 6.14341021e-01 -3.23370919e-02 1.51674360e-01 -6.73031211e-01 -1.39099860e+00 3.38935584e-01 9.75320220e-01 -3.94994557e-01 3.92524719e-01 5.54281235e-01 4.09829885e-01 -3.59777719e-01 -1.02089262e+00 4.17403668e-01 4.16540980e-01 -5.85160077e-01 5.92375636e-01 -2.82738239e-01 -2.10365921e-01 -2.61253029e-01 -2.59677827e-01 -1.23331225e+00 -1.15412073e-02 -1.08788979e+00 -6.05709851e-01 1.42035413e+00 3.78443122e-01 -7.58328140e-01 1.16345680e+00 6.13690913e-01 1.25530571e-01 -1.16129017e+00 -1.43733644e+00 -8.74985695e-01 -3.39726835e-01 -3.53801966e-01 5.27572334e-01 6.28972054e-01 2.18396723e-01 -3.16292867e-02 -6.73020840e-01 4.27470118e-01 9.02551651e-01 1.79130882e-01 5.94852149e-01 -9.41967964e-01 -5.74012518e-01 1.34846181e-01 -1.02996327e-01 -1.44175673e+00 3.38616073e-02 -6.64794683e-01 -1.43152094e-02 -1.28763187e+00 2.54251976e-02 -8.76845896e-01 -5.35609663e-01 1.63343847e-01 -1.07482627e-01 -4.35939431e-01 2.66801268e-01 3.04173470e-01 -8.62768531e-01 9.11040843e-01 1.03734386e+00 -1.26137719e-01 -2.32842237e-01 7.34640658e-01 -6.36530161e-01 4.65104133e-01 7.47868001e-01 -4.16945279e-01 -5.47906041e-01 -5.19284308e-01 1.85059488e-01 5.02133906e-01 9.73810032e-02 -1.07734752e+00 5.20188391e-01 1.25005528e-01 1.28037289e-01 -3.76909018e-01 2.35216156e-01 -1.60897195e+00 1.17806114e-01 4.28705752e-01 -9.34039727e-02 -4.11766887e-01 -1.93366647e-01 9.18239951e-01 -9.86020267e-02 -1.79407686e-01 6.91935778e-01 2.91313440e-01 -1.88911796e-01 5.78387141e-01 9.94399935e-03 -1.37704685e-01 1.19162178e+00 -1.29533798e-01 8.56059268e-02 -7.49508679e-01 -6.32596135e-01 6.14457488e-01 -1.56922229e-02 -1.09358624e-01 2.71642119e-01 -1.28361964e+00 -4.23181355e-01 2.67644431e-02 -3.76342088e-01 3.07946336e-02 5.98762155e-01 1.30714643e+00 4.32152152e-02 2.58849025e-01 4.32324082e-01 -2.93988645e-01 -9.41374123e-01 4.59735483e-01 6.18294358e-01 -4.39370543e-01 -3.84095639e-01 4.99494135e-01 -1.21327884e-01 -3.35437417e-01 6.79238081e-01 9.60183740e-02 2.03536928e-01 -2.56416380e-01 4.21640486e-01 7.16896236e-01 3.34870934e-01 -3.35822195e-01 -2.22303838e-01 3.05034548e-01 8.28014016e-02 3.41723353e-01 1.20049500e+00 -6.51422679e-01 5.24487495e-02 4.11086157e-02 1.38884521e+00 1.82324648e-01 -1.42898428e+00 -6.49731219e-01 -3.08686823e-01 -3.53675634e-01 3.79907697e-01 -4.52955574e-01 -1.55870342e+00 3.87075752e-01 7.56245136e-01 -1.58855394e-02 1.49645281e+00 -6.21687531e-01 1.15043437e+00 3.47232014e-01 6.98795617e-01 -1.24262655e+00 -5.63871562e-01 1.73355237e-01 2.40016341e-01 -9.92481709e-01 1.14640281e-01 -5.40937483e-01 -2.67339200e-02 1.03508377e+00 2.97744215e-01 1.73251942e-01 1.15099084e+00 2.35897213e-01 -1.23179808e-01 2.46567711e-01 -5.10676384e-01 1.32206663e-01 1.69508997e-02 1.91471949e-01 1.44488171e-01 1.34612992e-01 -7.56806016e-01 1.11576915e+00 1.54655769e-01 1.55787691e-01 -2.17771567e-02 8.79626989e-01 -1.74398139e-01 -7.75333703e-01 -3.91263992e-01 5.89797914e-01 -7.54120827e-01 1.34180471e-01 2.69525856e-01 2.34569341e-01 7.85932094e-02 1.32249200e+00 -4.14146543e-01 -3.53054076e-01 3.17392975e-01 -2.70812482e-01 8.02133679e-02 -8.86368677e-02 -3.71365249e-01 2.01530263e-01 -2.81066932e-02 -5.03873110e-01 -3.30621213e-01 -2.97458231e-01 -1.19615853e+00 -2.39206985e-01 -8.21668565e-01 6.11482143e-01 8.63019049e-01 1.09316170e+00 4.05097753e-01 3.89869481e-01 1.33605719e+00 -5.00490665e-01 -1.30707216e+00 -4.80441809e-01 -8.74928772e-01 1.65804237e-01 2.32144102e-01 -3.38170379e-01 -4.77047861e-01 -4.81804430e-01]

[5.966122150421143, 5.673704147338867]

8e8b6aca-d79c-404c-b456-8da05c14694b

deep-double-self-expressive-subspace

2306.11592

https://arxiv.org/abs/2306.11592v1

https://arxiv.org/pdf/2306.11592v1.pdf

Deep Double Self-Expressive Subspace Clustering

Deep subspace clustering based on auto-encoder has received wide attention. However, most subspace clustering based on auto-encoder does not utilize the structural information in the self-expressive coefficient matrix, which limits the clustering performance. In this paper, we propose a double self-expressive subspace clustering algorithm. The key idea of our solution is to view the self-expressive coefficient as a feature representation of the example to get another coefficient matrix. Then, we use the two coefficient matrices to construct the affinity matrix for spectral clustering. We find that it can reduce the subspace-preserving representation error and improve connectivity. To further enhance the clustering performance, we proposed a self-supervised module based on contrastive learning, which can further improve the performance of the trained network. Experiments on several benchmark datasets demonstrate that the proposed algorithm can achieve better clustering than state-of-the-art methods.

['Jun Zhou', 'Shanxiong Chen', 'Yunpeng Ma', 'Ling Zhao']

2023-06-20

['contrastive-learning', 'contrastive-learning', 'clustering']

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

[-3.67706746e-01 -4.81113195e-01 -5.00949435e-02 -2.79139698e-01 -2.26634428e-01 -4.33849633e-01 1.48930967e-01 -4.64567751e-01 -2.39007041e-01 1.48815051e-01 5.63373387e-01 2.31067523e-01 -2.35462174e-01 -6.43027306e-01 -4.69542325e-01 -1.09462237e+00 1.90464631e-01 7.86635578e-02 9.77649242e-02 6.71655163e-02 4.56757098e-02 1.25752851e-01 -1.28874183e+00 1.95039541e-01 1.18223739e+00 7.70941138e-01 3.01686347e-01 -1.27876222e-01 -1.93165675e-01 4.91466165e-01 -2.67595738e-01 2.06426039e-01 1.38125628e-01 -5.25543213e-01 -3.05234939e-01 2.61549741e-01 -1.57501444e-01 -2.26580054e-01 -7.78614998e-01 1.25907385e+00 4.80350077e-01 6.34452105e-02 7.08668411e-01 -1.10729682e+00 -6.62227333e-01 8.28092158e-01 -6.32871509e-01 -8.61957222e-02 7.04049990e-02 -6.28286079e-02 7.87184119e-01 -9.45313215e-01 5.09148717e-01 1.18594408e+00 5.30836642e-01 2.59685159e-01 -1.15677679e+00 -1.14217985e+00 2.70339698e-02 5.22770762e-01 -1.96598113e+00 -2.63181061e-01 1.27197254e+00 -4.10534590e-01 4.03257608e-01 -6.56866208e-02 8.20152760e-01 8.34306002e-01 -3.11138570e-01 9.73413587e-01 9.29270387e-01 -1.27512529e-01 2.20095351e-01 1.04135543e-01 -1.57842506e-02 8.43399823e-01 1.60426244e-01 -1.38807863e-01 -2.04033077e-01 7.02333301e-02 7.67232478e-01 4.54130501e-01 -4.99985009e-01 -8.64400029e-01 -1.36064458e+00 8.82632792e-01 8.04338038e-01 6.37005329e-01 -1.02428928e-01 -8.24031606e-02 3.63091618e-01 -7.39849210e-02 1.72866717e-01 1.86693862e-01 -4.09319401e-02 1.24095306e-01 -9.37251151e-01 -2.32707039e-01 4.69476581e-01 7.96655595e-01 1.00490499e+00 1.73122942e-01 1.20398207e-02 8.78742516e-01 5.25207877e-01 3.82894039e-01 8.37000668e-01 -1.02747238e+00 1.96616650e-01 1.13240111e+00 -4.54988718e-01 -1.22767043e+00 -2.93994099e-01 -7.76574254e-01 -1.34359396e+00 -2.03629419e-01 -1.10000461e-01 -1.41910210e-01 -6.17795348e-01 1.75334227e+00 2.81122446e-01 5.59908271e-01 1.16778567e-01 1.21488988e+00 6.61207795e-01 8.09173346e-01 -4.00752515e-01 -4.69647557e-01 8.40184689e-01 -9.99493659e-01 -9.03140843e-01 2.96765476e-01 6.79459214e-01 -5.42893469e-01 9.95706618e-01 2.36477032e-01 -5.63084900e-01 -6.25190437e-01 -1.23649514e+00 1.02568194e-01 -1.46483347e-01 3.82934093e-01 6.51687860e-01 3.17767739e-01 -6.69444084e-01 5.60308933e-01 -8.36230814e-01 -2.58894861e-01 4.04801548e-01 3.98889691e-01 -5.05094171e-01 -1.26147151e-01 -1.15248632e+00 1.57329485e-01 7.39236534e-01 -1.32802919e-01 -5.18112361e-01 -4.23205644e-01 -7.33651459e-01 2.84451157e-01 3.68122816e-01 -5.14741540e-01 4.79817599e-01 -8.45109224e-01 -1.34512258e+00 3.43785346e-01 -2.66957432e-01 -5.81642538e-02 1.21142052e-01 -6.52691871e-02 -5.24513960e-01 2.89314091e-01 3.09874713e-01 6.13434970e-01 7.94598758e-01 -1.41828370e+00 -2.65663445e-01 -5.00139236e-01 -4.93847132e-01 4.47252840e-01 -1.04970956e+00 -4.22100514e-01 -1.02193868e+00 -6.53405666e-01 6.11162126e-01 -9.07339931e-01 -2.41191193e-01 -3.69685501e-01 -3.86564583e-01 -1.52285025e-01 1.24135768e+00 -4.01979536e-01 1.60087729e+00 -2.67904091e+00 4.68384534e-01 4.80257839e-01 3.55635732e-01 1.16455927e-01 7.90196378e-03 3.30406994e-01 -2.40813822e-01 -1.79266986e-02 -3.81199092e-01 4.41337712e-02 -1.47769451e-01 9.28349271e-02 -1.76459715e-01 4.89306301e-01 -1.96930885e-01 7.06698418e-01 -8.23181510e-01 -8.09070587e-01 4.02778357e-01 6.79830909e-01 -5.93724430e-01 2.24579945e-01 3.52103025e-01 3.85415047e-01 -5.17253160e-01 1.43210039e-01 7.90436029e-01 -3.55262965e-01 1.42552897e-01 -8.56618106e-01 -8.88425037e-02 -3.05166483e-01 -1.48834217e+00 2.13164878e+00 -1.64810807e-01 3.72009367e-01 6.43304959e-02 -1.05899525e+00 9.32554960e-01 4.20406014e-02 9.06573117e-01 -2.62635291e-01 1.64442092e-01 1.73386097e-01 3.11408937e-01 -2.15908200e-01 -3.29007432e-02 6.67846724e-02 2.98335910e-01 3.43007237e-01 5.73178641e-02 4.54054832e-01 1.95849791e-01 5.67378223e-01 8.46951723e-01 -2.44042370e-02 -7.36099556e-02 -4.32853490e-01 9.00157988e-01 -1.69710830e-01 8.32026362e-01 -7.64521444e-03 -1.28583074e-01 5.31376839e-01 2.54974335e-01 -1.37399286e-01 -9.74092066e-01 -1.00124419e+00 -1.76459476e-01 4.81154859e-01 4.10577685e-01 -7.04518855e-01 -1.14157772e+00 -5.87183118e-01 -1.30285040e-01 4.35603887e-01 -4.75412339e-01 -6.90553844e-01 -3.92261386e-01 -7.11082399e-01 3.14271331e-01 6.50518954e-01 9.08436298e-01 -6.49173498e-01 -1.21613018e-01 1.59625292e-01 -2.75719374e-01 -8.22030962e-01 -8.44272137e-01 -2.92306934e-02 -9.26248968e-01 -9.32844937e-01 -7.33864725e-01 -9.95183825e-01 7.97824442e-01 6.82504475e-01 3.12584221e-01 1.60202593e-01 -7.61836767e-02 1.90130964e-01 -4.73197371e-01 1.57364845e-01 5.59781864e-02 5.38520701e-02 4.01135594e-01 3.74420434e-01 5.75548291e-01 -8.93029273e-01 -7.43673444e-01 3.33175123e-01 -9.21266377e-01 1.18403733e-01 8.42865527e-01 9.40888822e-01 6.93959594e-01 5.63289642e-01 5.23933709e-01 -6.60599291e-01 5.69296956e-01 -2.82499433e-01 -1.99757010e-01 -3.80257629e-02 -8.21454048e-01 2.40314290e-01 1.02048862e+00 -4.47562784e-01 -9.49230611e-01 5.57565391e-01 1.40317947e-01 -1.19691753e+00 -1.07646612e-02 5.05796492e-01 -6.16518855e-01 1.08314894e-01 2.70479292e-01 7.30576038e-01 1.47239268e-01 -5.43083489e-01 5.20468593e-01 7.95462668e-01 6.02442563e-01 -2.68504620e-01 9.99655783e-01 4.60365325e-01 -1.34949833e-01 -5.75778246e-01 -7.02753186e-01 -7.29086280e-01 -8.05846274e-01 -8.94077942e-02 9.30381715e-01 -1.25156903e+00 -8.07981670e-01 2.33722880e-01 -7.78159857e-01 1.68565020e-01 -1.79395396e-02 7.04349220e-01 -3.64201188e-01 7.32894301e-01 -6.47946179e-01 -5.87135732e-01 -1.95580319e-01 -1.19282508e+00 8.41306567e-01 1.76717579e-01 2.70610988e-01 -8.41335356e-01 -2.24965508e-03 2.41139695e-01 1.30175084e-01 -8.16090219e-03 6.93609834e-01 -4.06367272e-01 -4.59599525e-01 -3.92673500e-02 -2.91676372e-01 2.55756736e-01 1.79957181e-01 -9.11440179e-02 -7.43991792e-01 -4.15038973e-01 1.41029298e-01 -6.22194335e-02 1.20033348e+00 2.51580924e-01 1.54552460e+00 -3.58831376e-01 -5.66949129e-01 9.73078430e-01 1.41019595e+00 2.29769155e-01 4.48632538e-01 1.51601985e-01 1.38316023e+00 3.71641934e-01 3.55880469e-01 3.76050055e-01 2.95291305e-01 4.75277275e-01 1.75770462e-01 -4.50965762e-02 3.68379876e-02 -5.54920733e-01 3.60605150e-01 1.58542514e+00 6.61805496e-02 4.21161950e-01 -8.29979062e-01 4.88355875e-01 -2.15226102e+00 -1.02611673e+00 -1.88119531e-01 1.84383225e+00 6.92828178e-01 -1.77664071e-01 -4.45513278e-02 2.94097602e-01 8.97232115e-01 1.29971415e-01 -7.70943582e-01 3.92749548e-01 -2.59807020e-01 -4.32696670e-01 1.66838765e-01 1.29880384e-01 -1.22056699e+00 1.08809614e+00 5.63591671e+00 1.06123495e+00 -1.02092803e+00 -1.94270462e-02 3.81364882e-01 3.33207324e-02 -3.13903183e-01 5.06272689e-02 -4.21810299e-01 8.21005046e-01 5.19932389e-01 -1.85653508e-01 6.27363086e-01 1.06316698e+00 8.31182450e-02 5.07069349e-01 -9.70797658e-01 1.31215048e+00 2.52110362e-01 -1.23914695e+00 1.75558552e-01 1.05041012e-01 7.75309026e-01 -2.19882160e-01 5.19081615e-02 3.14499438e-01 1.46713614e-01 -8.26926768e-01 1.27072394e-01 4.58124816e-01 8.01748157e-01 -1.03537548e+00 7.95673490e-01 4.83410090e-01 -1.39749146e+00 -2.03131393e-01 -6.50717258e-01 2.06192806e-01 -1.53022334e-01 7.87702441e-01 -3.25074583e-01 7.07716823e-01 8.58985245e-01 1.43327689e+00 -6.15544796e-01 8.28112423e-01 1.54379038e-02 7.36745238e-01 -2.75438398e-01 1.21017955e-01 2.25824624e-01 -8.12903464e-01 4.17850912e-01 1.09295189e+00 2.99513370e-01 3.20278764e-01 4.46276903e-01 9.79044259e-01 -8.49979594e-02 3.95076632e-01 -6.42636657e-01 -4.88844886e-02 6.72359943e-01 1.44839036e+00 -5.98855793e-01 -3.81709427e-01 -4.42998409e-01 1.25349975e+00 3.30835521e-01 4.72601175e-01 -6.00011051e-01 -4.94335592e-01 3.48639518e-01 -6.59548044e-02 3.88141304e-01 -3.38571131e-01 -1.92729771e-01 -1.49894655e+00 -8.91591236e-02 -8.58237207e-01 3.10884237e-01 -7.27553129e-01 -1.27234173e+00 3.12666148e-01 -2.76986659e-01 -1.48780096e+00 1.30555600e-01 -4.10340577e-01 -7.36120701e-01 5.16709626e-01 -9.48916376e-01 -1.14196563e+00 -6.04327559e-01 9.55328465e-01 2.10506991e-01 -6.29492044e-01 6.78266108e-01 5.32929778e-01 -9.85635579e-01 5.83303034e-01 7.07427382e-01 4.25171912e-01 8.51165056e-01 -1.20362508e+00 -4.87780094e-01 7.65844464e-01 3.31058353e-01 9.98611033e-01 2.36591235e-01 -5.60149431e-01 -1.57983637e+00 -1.20878422e+00 -2.97248289e-02 5.22727109e-02 6.11678541e-01 -2.79933453e-01 -9.90973949e-01 3.95848393e-01 2.07592949e-01 -1.39413282e-01 8.93145561e-01 1.96202651e-01 -5.62219262e-01 -5.58306456e-01 -6.87157035e-01 6.18899345e-01 1.10110652e+00 -6.38832629e-01 -4.98665869e-01 2.76310980e-01 9.24862385e-01 7.92741850e-02 -1.00373888e+00 5.02768993e-01 3.99612069e-01 -1.06530619e+00 8.41964006e-01 -2.25053146e-01 4.39443737e-01 -7.78333902e-01 -1.77095145e-01 -1.51787317e+00 -8.81305456e-01 -1.47405833e-01 -2.70390421e-01 1.37659991e+00 5.29562384e-02 -3.44552606e-01 8.31113279e-01 5.43602258e-02 -2.46767968e-01 -8.55873406e-01 -7.09460557e-01 -7.08594203e-01 -1.54594902e-03 -9.03244689e-02 6.31474018e-01 1.38038468e+00 1.89506114e-01 8.58894229e-01 -2.43599892e-01 9.15863067e-02 7.62592077e-01 3.86434138e-01 7.42809117e-01 -1.25572526e+00 -9.25916508e-02 -4.58525389e-01 -4.14214432e-01 -1.03117967e+00 3.72385442e-01 -1.12778735e+00 -2.16889426e-01 -1.50363386e+00 6.90692842e-01 -2.41448715e-01 -5.92277229e-01 3.41420650e-01 -3.19797009e-01 4.67136018e-02 5.36799170e-02 6.30405009e-01 -8.56624901e-01 1.10658669e+00 1.27005053e+00 -2.12286174e-01 -2.16511935e-01 -6.52141929e-01 -7.06067443e-01 6.40937746e-01 7.19574153e-01 -2.21468687e-01 -6.05422556e-01 -2.56816059e-01 -1.95459262e-01 -2.20085904e-01 -1.92681700e-02 -1.36720765e+00 5.48517883e-01 1.54361576e-01 8.60386431e-01 -7.74738252e-01 2.38718107e-01 -1.16192579e+00 9.39184800e-02 6.03121877e-01 -1.69909433e-01 -2.35941142e-01 -1.61438733e-01 8.68628621e-01 -5.13834596e-01 2.92900819e-02 7.00511754e-01 2.27029826e-02 -4.87030476e-01 5.64658999e-01 -7.67415091e-02 -2.39640042e-01 1.01357687e+00 -1.69281438e-01 5.13683707e-02 -3.78934622e-01 -4.71792608e-01 5.39722025e-01 7.08204389e-01 3.20392549e-01 7.67655969e-01 -1.89021850e+00 -5.23572505e-01 4.06485975e-01 4.93249558e-02 6.96979836e-02 4.72173452e-01 8.37154210e-01 -2.86818504e-01 2.28597999e-01 -2.78682262e-01 -7.84922063e-01 -1.01972294e+00 9.17022884e-01 1.62501618e-01 4.57050800e-02 -6.72900379e-01 4.48125780e-01 4.43558782e-01 -4.44597095e-01 7.00210482e-02 3.48165095e-01 -4.79085773e-01 -2.72471886e-02 4.11335975e-01 3.68836790e-01 -5.65516770e-01 -8.06950688e-01 -4.43136692e-01 9.49868143e-01 -1.22839592e-01 -1.21111207e-01 1.24928606e+00 -1.40717253e-01 -2.89127886e-01 5.24633348e-01 1.63445807e+00 -9.94992163e-03 -1.12337661e+00 -2.87630618e-01 -2.95166671e-01 -4.51719105e-01 3.87791395e-01 -2.86270320e-01 -1.45143342e+00 1.13911808e+00 8.00118208e-01 -1.28434092e-01 1.28582799e+00 -1.98338673e-01 9.09884393e-01 6.25771165e-01 1.01680279e-01 -1.24542475e+00 3.22889000e-01 2.71242350e-01 5.47717631e-01 -1.25000429e+00 8.24272484e-02 -4.96707797e-01 -7.01541245e-01 8.81556630e-01 8.49279523e-01 -3.31594557e-01 8.74253809e-01 -1.42370192e-02 1.53181208e-02 -1.63008049e-01 -3.88459474e-01 -2.43035540e-01 4.37096804e-01 3.99818420e-01 4.18179154e-01 5.66923805e-02 -3.68394345e-01 8.66795301e-01 -1.77462637e-01 -3.63989443e-01 2.18094617e-01 2.06358746e-01 -4.53490853e-01 -1.06816018e+00 -2.95628309e-01 5.49813747e-01 -1.90589558e-02 -7.11788330e-03 -6.16674006e-01 3.45479071e-01 1.91303730e-01 9.29377794e-01 1.95737984e-02 -9.60833132e-01 1.70383267e-02 2.23827362e-03 7.13480264e-02 -4.07037973e-01 -7.37906396e-02 6.78799868e-01 -4.92727846e-01 -5.87908745e-01 -4.78138775e-01 -6.12853110e-01 -1.60317087e+00 -7.14570060e-02 -3.87854517e-01 4.65271682e-01 3.67506027e-01 5.70809007e-01 5.89770317e-01 5.34825683e-01 1.06139779e+00 -4.43362504e-01 -2.56180733e-01 -9.52154279e-01 -8.05116653e-01 7.77127683e-01 -1.04643360e-01 -6.93078518e-01 -4.11937505e-01 7.22382143e-02]

[8.659581184387207, 3.9919283390045166]

86907ee1-8538-4dae-a021-b8e7cbd11583

spatial-temporal-recurrent-graph-neural

2210.15177

https://arxiv.org/abs/2210.15177v1

https://arxiv.org/pdf/2210.15177v1.pdf

Spatial-Temporal Recurrent Graph Neural Networks for Fault Diagnostics in Power Distribution Systems

Fault diagnostics are extremely important to decide proper actions toward fault isolation and system restoration. The growing integration of inverter-based distributed energy resources imposes strong influences on fault detection using traditional overcurrent relays. This paper utilizes emerging graph learning techniques to build a new temporal recurrent graph neural network models for fault diagnostics. The temporal recurrent graph neural network structures can extract the spatial-temporal features from data of voltage measurement units installed at the critical buses. From these features, fault event detection, fault type/phase classification, and fault location are performed. Compared with previous works, the proposed temporal recurrent graph neural networks provide a better generalization for fault diagnostics. Moreover, the proposed scheme retrieves the voltage signals instead of current signals so that there is no need to install relays at all lines of the distribution system. Therefore, the proposed scheme is generalizable and not limited by the number of relays installed. The effectiveness of the proposed method is comprehensively evaluated on the Potsdam microgrid and IEEE 123-node system in comparison with other neural network structures.

['Rob Hovsapian', 'Mayank Panwar', 'Thai-Thanh Nguyen', 'Tuyen Vu', 'Bang Nguyen']

2022-10-27

['fault-detection']

['miscellaneous']

[-3.43582898e-01 -3.23172957e-01 7.43489265e-02 1.63933873e-01 -9.31805596e-02 -4.98178154e-01 1.81539521e-01 3.45476687e-01 3.81299406e-02 8.24863255e-01 -3.62337977e-01 -5.75143576e-01 -6.33201241e-01 -8.23470950e-01 -7.16641396e-02 -9.55994844e-01 -6.51326478e-01 2.30754554e-01 8.13412443e-02 -3.31746578e-01 2.17315644e-01 9.85885561e-01 -1.23449135e+00 -1.31796673e-01 9.95051384e-01 8.49727750e-01 1.00057878e-01 5.02338588e-01 3.62922013e-01 9.10878479e-01 -1.29724741e+00 5.34819067e-01 -1.10364571e-01 -5.27450144e-01 -2.71593362e-01 1.56985298e-01 -6.55788124e-01 -3.71819019e-01 -9.36592758e-01 1.14643586e+00 4.65242773e-01 4.10761297e-01 7.39457846e-01 -1.50511491e+00 -4.49697793e-01 7.42781758e-01 -4.35255706e-01 7.60481775e-01 2.31337473e-01 -2.13852957e-01 7.50949562e-01 -8.03611219e-01 2.32604131e-01 6.83378816e-01 5.81766605e-01 -2.52053380e-01 -5.48017561e-01 -4.09861356e-01 3.38651270e-01 1.14605737e+00 -1.49288630e+00 1.37288958e-01 1.25709009e+00 -1.24513730e-01 1.25455904e+00 2.72406757e-01 8.58637869e-01 4.72818613e-01 7.64999449e-01 3.76027465e-01 8.15496922e-01 -4.77407634e-01 3.67337763e-01 -4.74705786e-01 6.29917622e-01 6.36547208e-01 5.06899238e-01 -3.02450329e-01 -1.37072057e-01 6.71301736e-03 4.76388425e-01 2.89701760e-01 -9.40300226e-01 4.35478017e-02 -5.87981403e-01 7.45281577e-01 6.25497222e-01 9.33949947e-01 -5.14874578e-01 -8.82506818e-02 5.62907338e-01 4.74231064e-01 5.31336069e-01 3.46570574e-02 -7.51558319e-02 1.49953191e-03 -8.73669028e-01 -2.56542474e-01 7.70543218e-01 5.93066394e-01 3.20810288e-01 1.14690280e+00 3.78768563e-01 5.09872317e-01 1.88856035e-01 3.79932910e-01 7.23673522e-01 -1.68713078e-01 2.07768962e-01 8.98291647e-01 -1.04273029e-01 -1.17990386e+00 -9.18263733e-01 -6.63057268e-01 -9.89864647e-01 3.74106675e-01 2.60981675e-02 -3.86086196e-01 -8.71673048e-01 9.66910601e-01 6.32438902e-03 2.93289334e-01 9.90227386e-02 6.03472829e-01 3.97801578e-01 1.06024468e+00 -6.99594378e-01 -6.48259878e-01 9.91493702e-01 -4.12838668e-01 -1.33509123e+00 2.64371902e-01 6.80068672e-01 -3.62071157e-01 2.71172076e-01 6.02508128e-01 -6.79079950e-01 -2.45436072e-01 -1.65259433e+00 6.22870684e-01 -7.64276922e-01 3.51402164e-01 2.53736973e-01 3.89004469e-01 -1.15843272e+00 8.34482193e-01 -9.81350243e-01 -1.65958568e-01 -2.52859145e-01 3.82006317e-01 -2.35895574e-01 -2.28241435e-03 -1.54344690e+00 1.17080796e+00 6.34966850e-01 1.05033243e+00 -7.09004283e-01 -1.75810292e-01 -1.00299275e+00 6.39801025e-02 3.90299618e-01 -1.93746132e-03 7.95427799e-01 -5.05730212e-01 -1.25107086e+00 -3.98601443e-01 1.08059116e-01 -8.38398337e-01 2.45277703e-01 9.23968703e-02 -1.05042362e+00 5.15887141e-01 -8.42409208e-02 -6.95876122e-01 6.97648644e-01 -8.89124751e-01 -6.05687201e-01 -2.93835670e-01 -3.04056823e-01 2.29505643e-01 -3.41485947e-01 -5.35102904e-01 1.26465529e-01 -5.82289696e-01 3.55417818e-01 -4.96095747e-01 -2.08114952e-01 -6.74001873e-01 -5.64695001e-01 -4.66461182e-01 1.64640796e+00 -9.74824905e-01 1.40728784e+00 -1.87964702e+00 6.88299462e-02 6.18247151e-01 1.20763309e-01 2.49075413e-01 4.28625792e-01 1.12848926e+00 -5.28246343e-01 -3.41230929e-01 -2.51654774e-01 3.25967014e-01 -2.80893654e-01 3.82526726e-01 -1.35180010e-02 1.03314435e+00 1.02389187e-01 5.81045091e-01 -6.26545608e-01 6.95324615e-02 5.71905136e-01 5.72270334e-01 1.85186744e-01 -4.85398285e-02 2.75494277e-01 9.50099155e-02 -4.44836020e-01 3.86650711e-01 4.31666344e-01 -1.56230018e-01 2.28833437e-01 -2.79120117e-01 -2.02395871e-01 2.21253410e-02 -1.31259167e+00 1.24635410e+00 -3.10181141e-01 8.08859885e-01 -7.10383877e-02 -1.71612906e+00 1.10206926e+00 7.75200307e-01 6.98772907e-01 -7.26982534e-01 3.56063634e-01 1.12581134e-01 1.78914905e-01 -5.43069959e-01 3.62877771e-02 3.86354536e-01 1.18893065e-01 3.71605903e-01 1.98604822e-01 1.63618445e-01 6.47030592e-01 1.07012525e-01 1.01694095e+00 -2.78502464e-01 2.10687026e-01 -4.69362676e-01 6.08225524e-01 -6.07040934e-02 8.54632676e-01 2.19909295e-01 1.75527230e-01 3.94368917e-02 5.16208887e-01 -2.42111206e-01 -4.73899156e-01 -8.07310402e-01 1.66315436e-02 1.70391947e-01 1.94929987e-01 -1.88937619e-01 -5.29299796e-01 -5.28239310e-01 3.27614062e-02 8.14043999e-01 -4.97366518e-01 -4.96613890e-01 -6.31330729e-01 -9.22718167e-01 4.36522484e-01 4.19114679e-01 4.85742182e-01 -7.63745606e-01 -3.97052854e-01 4.54580158e-01 1.84465006e-01 -6.66591465e-01 -1.29214013e-02 5.47423124e-01 -7.55590439e-01 -1.41304755e+00 -8.77528667e-01 -1.03849280e+00 8.27520967e-01 2.23693952e-01 5.87035298e-01 2.29437083e-01 -5.36471069e-01 1.11862220e-01 -6.44235373e-01 -9.02708322e-02 -2.40494877e-01 -1.12722762e-01 1.86932027e-01 1.10306598e-01 -2.09744662e-01 -8.50194693e-01 -3.50432754e-01 1.74303114e-01 -8.37604761e-01 -5.29544950e-01 2.95843571e-01 1.13530231e+00 -3.01366206e-02 1.18277264e+00 1.31123173e+00 -6.25753641e-01 9.28553164e-01 -6.30385220e-01 -9.60743189e-01 4.21913117e-01 -1.03340232e+00 -2.11290315e-01 9.66464996e-01 -9.39157829e-02 -1.01913071e+00 -3.67469966e-01 8.49023908e-02 -3.82330269e-01 1.58422798e-01 1.05418372e+00 -1.47745118e-01 -1.60814255e-01 3.52286071e-01 2.58680284e-01 1.66109223e-02 -4.29044604e-01 1.45049572e-01 5.57066619e-01 5.51476181e-01 1.55618131e-01 6.81084156e-01 1.90721508e-02 1.82665870e-01 -1.02570224e+00 2.18295708e-01 -3.46678287e-01 -6.36373758e-01 -5.38574517e-01 4.31698561e-01 -6.77441359e-01 -1.02416444e+00 8.30449283e-01 -1.06880164e+00 7.15596825e-02 4.07127440e-02 6.46390736e-01 5.72589748e-02 6.87094629e-01 -1.19082534e+00 -1.08028448e+00 -3.43424946e-01 -1.04155600e+00 2.14378461e-01 2.86042005e-01 2.29377374e-01 -1.51375353e+00 -3.80969346e-01 -3.64886850e-01 2.05925465e-01 4.45125967e-01 1.10804427e+00 -7.54017651e-01 -2.72960156e-01 -7.03720987e-01 1.51259840e-01 4.60351586e-01 6.11937821e-01 1.66153029e-01 -4.00508493e-01 -7.10744619e-01 5.01144171e-01 3.40505153e-01 3.69523942e-01 4.66796398e-01 4.43053156e-01 -1.91976026e-01 -3.11434656e-01 1.55522704e-01 1.62971902e+00 1.13714278e+00 3.80054802e-01 1.46890596e-01 7.49880612e-01 3.68837506e-01 3.11557263e-01 4.89758700e-01 2.79922813e-01 1.76864967e-01 4.11960244e-01 -1.95504636e-01 2.43622601e-01 2.52622545e-01 4.67326701e-01 1.33040500e+00 1.12728395e-01 -7.32816935e-01 -9.74899232e-01 6.80429697e-01 -1.78617775e+00 -7.54394114e-01 -3.23772997e-01 1.84192598e+00 9.17099565e-02 1.86140046e-01 -1.50281399e-01 1.21857691e+00 1.00443447e+00 1.17142804e-01 -6.35614872e-01 -3.03161949e-01 -4.10469353e-01 1.09755471e-01 5.56400180e-01 5.17667949e-01 -7.12072313e-01 1.14139050e-01 5.25838804e+00 7.77559280e-01 -1.29183531e+00 -1.49991617e-01 2.62012213e-01 2.09550902e-01 5.31644747e-02 -1.68282881e-01 -1.42139360e-01 3.55470896e-01 1.15278244e+00 -4.87062544e-01 1.60482526e-01 3.90394598e-01 5.13395846e-01 -3.62129509e-01 -7.23701835e-01 7.57327080e-01 1.74798042e-01 -9.81994927e-01 4.84448411e-02 -3.63968253e-01 5.35267532e-01 -9.86398086e-02 -3.49857420e-01 -1.51373938e-01 1.52644709e-01 -7.72956014e-01 3.59242618e-01 5.03172100e-01 2.20230728e-01 -1.36989033e+00 1.11895537e+00 1.47706479e-01 -1.38026083e+00 -4.50400352e-01 -1.90151110e-03 -1.44808799e-01 7.12915242e-01 8.90605688e-01 -1.03577840e+00 1.57220280e+00 7.13659763e-01 1.18895137e+00 -4.31649983e-01 1.07464838e+00 -3.38954240e-01 7.09045827e-01 -2.18009949e-01 -4.40209396e-02 1.66012093e-01 -4.86719757e-01 4.67472523e-01 7.12292969e-01 4.12011921e-01 -5.34592457e-02 3.35180521e-01 3.62536520e-01 2.81103075e-01 -7.34247938e-02 -8.49550128e-01 3.74070145e-02 5.84299147e-01 1.05311882e+00 -1.18448806e+00 -2.04948276e-01 -3.90498102e-01 7.90329278e-01 -1.52548969e-01 8.88189316e-01 -5.16220927e-01 -1.02564836e+00 1.65998265e-01 -3.85360867e-01 -2.73440219e-02 -3.43721539e-01 -2.07059070e-01 -9.25567329e-01 1.11182615e-01 -5.11318862e-01 6.16135180e-01 -8.39718521e-01 -1.21658802e+00 8.23101044e-01 1.28899813e-01 -1.36122549e+00 -7.89840460e-01 -4.66484398e-01 -9.71392035e-01 9.47314918e-01 -1.28089392e+00 -7.57753968e-01 -2.81022415e-02 7.65199840e-01 5.53722918e-01 -8.09547380e-02 6.31904781e-01 4.47272629e-01 -1.32836211e+00 1.65393963e-01 3.39410692e-01 4.18007553e-01 -8.10894072e-02 -1.11629212e+00 1.66622892e-01 1.49873710e+00 3.50284688e-02 8.83351862e-02 6.93126440e-01 -8.37005317e-01 -1.50299489e+00 -6.34293437e-01 3.94282103e-01 6.99374497e-01 8.76993895e-01 -9.48780254e-02 -1.19870698e+00 6.44599974e-01 7.68150270e-01 -1.54471517e-01 1.37744263e-01 -3.73900235e-01 2.65496045e-01 -1.98187754e-01 -1.01681626e+00 4.40943718e-01 1.53124318e-01 -6.33313477e-01 -5.58056056e-01 3.27005863e-01 3.06318134e-01 -1.81130633e-01 -9.45935190e-01 5.58361948e-01 -7.48299137e-02 -6.06175840e-01 5.43990672e-01 1.17407426e-01 -5.63977420e-01 -5.81110954e-01 3.13903809e-01 -1.77097857e+00 -2.05653846e-01 -6.24001980e-01 -5.82631826e-02 9.74998832e-01 4.03305113e-01 -1.45068157e+00 4.51321959e-01 -9.94072333e-02 -5.48804820e-01 -7.32595086e-01 -1.22294152e+00 -6.69407129e-01 -3.44786495e-01 -1.63000934e-02 4.77215827e-01 1.09384644e+00 5.40138662e-01 2.71253318e-01 -1.58900648e-01 6.83344007e-01 5.22561789e-01 -5.63258566e-02 -2.28695199e-01 -1.23726785e+00 3.66648436e-02 -4.18876469e-01 -7.32072592e-01 -2.77568370e-01 2.18258664e-01 -6.77970529e-01 -9.28871334e-02 -1.91722023e+00 -9.11213458e-01 -1.55281544e-01 -6.18406415e-01 5.32869935e-01 7.15163723e-02 -1.34014204e-01 -2.90280253e-01 3.11455131e-02 1.56506211e-01 6.07329428e-01 7.82951474e-01 -2.91386843e-01 4.66721579e-02 8.08140934e-02 3.25351894e-01 3.73632729e-01 1.03614485e+00 -6.40992671e-02 -8.97294760e-01 -1.71519563e-01 -8.32336470e-02 7.91587353e-01 -1.10424168e-01 -1.30891693e+00 7.44224668e-01 4.52721715e-01 3.96174937e-01 -1.02457285e+00 1.20975509e-01 -1.00964653e+00 4.01151687e-01 8.68402719e-01 3.44108760e-01 1.04672480e+00 2.23377988e-01 6.04872465e-01 -5.49958408e-01 -2.28082925e-01 2.49736577e-01 2.14449540e-01 -8.10899794e-01 -1.97312217e-02 -1.12519741e+00 -6.52536929e-01 1.15217698e+00 -1.45257369e-01 -5.18515706e-01 -4.42418426e-01 -8.54908824e-01 4.55588669e-01 1.12618525e-02 4.52805340e-01 8.58142257e-01 -1.08547294e+00 -5.90073049e-01 2.87175119e-01 -3.06022912e-01 -1.92336515e-01 4.48400170e-01 7.55929351e-01 -7.21906781e-01 5.58826804e-01 -2.37306967e-01 -4.94212836e-01 -9.87009108e-01 4.38501298e-01 5.15798032e-01 -8.54532421e-02 -8.35148931e-01 2.73041427e-01 -6.40653551e-01 4.25598204e-01 1.45409912e-01 -4.50728625e-01 -7.54703224e-01 4.85996574e-01 2.09813878e-01 8.06175590e-01 6.52984917e-01 -3.12101245e-01 -2.52418011e-01 2.81795889e-01 3.58868986e-02 -1.33695593e-02 1.30169249e+00 -5.36543354e-02 -5.42588532e-01 9.37701821e-01 1.09345102e+00 -5.22162557e-01 -8.35761130e-01 2.57986069e-01 2.63403565e-01 3.59830022e-01 2.53533006e-01 -5.59848368e-01 -1.42835152e+00 6.27870023e-01 6.05651200e-01 8.79538119e-01 1.44962680e+00 -7.10820735e-01 3.86711895e-01 3.64644885e-01 6.56129122e-01 -9.54917729e-01 -4.69794601e-01 2.94004172e-01 7.09902823e-01 -5.26054323e-01 3.24695185e-02 -3.60164195e-01 7.33633200e-03 1.40500581e+00 1.56236321e-01 -5.09962142e-01 9.54600155e-01 4.44802552e-01 9.44811571e-03 -3.52906823e-01 -6.50547802e-01 5.94953634e-02 4.82299551e-02 5.64011991e-01 1.89602688e-01 -5.41153476e-02 -6.13282263e-01 1.14952855e-01 2.89845038e-02 -3.10547948e-01 1.05154240e+00 1.31717718e+00 -2.44620204e-01 -9.75736141e-01 -4.48190868e-01 5.55977404e-01 -3.98611218e-01 2.30345204e-01 1.24523371e-01 1.08283889e+00 -4.35069919e-01 1.34588742e+00 5.70337400e-02 -3.94140571e-01 4.93046463e-01 1.63317919e-01 1.18185751e-01 -3.36148322e-01 -4.41547573e-01 -3.82934473e-02 -8.92763585e-03 -1.30063742e-01 -1.24978848e-01 -5.34057736e-01 -1.66822910e+00 8.30446854e-02 -8.95631075e-01 1.00625503e+00 7.18964934e-01 1.10412848e+00 9.52332914e-02 1.12393606e+00 1.02804446e+00 -8.55582476e-01 -1.67171717e-01 -1.26132894e+00 -1.11643398e+00 -1.69201806e-01 3.83426994e-01 -8.26893330e-01 -6.65011704e-01 -3.79522890e-01]

[6.207886219024658, 2.491515636444092]

446d08c5-1151-495f-9cd9-16bc4c0149af

an-experiment-in-integrating-sentiment

https://aclanthology.org/W12-5503

https://aclanthology.org/W12-5503.pdf

An Experiment in Integrating Sentiment Features for Tech Stock Prediction in Twitter

['Tien Thanh Vu', 'Nigel Collier', 'Shu Chang', 'Quang Thuy Ha']

2012-12-01

an-experiment-in-integrating-sentiment-1

https://aclanthology.org/W12-5503

https://aclanthology.org/W12-5503.pdf

ws-2012-12

['twitter-sentiment-analysis', 'stock-market-prediction', 'stock-prediction']

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

[-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]

[-7.315954208374023, 3.6651344299316406]

5451e98b-f360-4e83-beb2-476a46c5359a

towards-exploiting-sticker-for-multimodal

https://aclanthology.org/2022.coling-1.591

https://aclanthology.org/2022.coling-1.591.pdf

Towards Exploiting Sticker for Multimodal Sentiment Analysis in Social Media: A New Dataset and Baseline

Sentiment analysis in social media is challenging since posts are short of context. As a popular way to express emotion on social media, stickers related to these posts can supplement missing sentiments and help identify sentiments precisely. However, research about stickers has not been investigated further. To this end, we present a Chinese sticker-based multimodal dataset for the sentiment analysis task (CSMSA). Compared with previous real-world photo-based multimodal datasets, the CSMSA dataset focuses on stickers, conveying more vivid and moving emotions. The sticker-based multimodal sentiment analysis task is challenging in three aspects: inherent multimodality of stickers, significant inter-series variations between stickers, and complex multimodal sentiment fusion. We propose SAMSAM to address the above three challenges. Our model introduces a flexible masked self-attention mechanism to allow the dynamic interaction between post texts and stickers. The experimental results indicate that our model performs best compared with other models. More researches need to be devoted to this field. The dataset is publicly available at https://github.com/Logos23333/CSMSA.

['Yi Cai', 'Haopeng Ren', 'Weizhao Li', 'Feng Ge']

coling-2022-10

['multimodal-sentiment-analysis', 'multimodal-sentiment-analysis']

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

[-5.08209355e-02 -4.19506490e-01 -1.06771417e-01 -6.31312847e-01 -6.46841884e-01 -5.27900100e-01 5.54603696e-01 3.21024477e-01 -2.25480929e-01 1.98554993e-01 6.02209032e-01 4.06270981e-01 4.21699971e-01 -3.97054672e-01 -4.90959615e-01 -7.52277255e-01 2.70489365e-01 -1.79577231e-01 -6.80639073e-02 -8.71872842e-01 3.26735318e-01 -1.52063087e-01 -1.52178907e+00 8.51157963e-01 6.71806216e-01 1.15778434e+00 3.67005080e-01 5.77453732e-01 -4.68640149e-01 8.52959871e-01 -6.54281497e-01 -5.85859776e-01 -2.96002895e-01 -5.02523899e-01 -5.55517673e-01 1.61372796e-01 2.66854227e-01 8.25039074e-02 -3.41522060e-02 1.02245498e+00 7.04041064e-01 1.14213258e-01 5.01215398e-01 -1.57641339e+00 -8.93970370e-01 7.95974195e-01 -9.23137724e-01 -5.01631834e-02 6.20423734e-01 -1.00404233e-01 1.16902792e+00 -9.25093889e-01 4.53971177e-01 1.41724634e+00 3.94709349e-01 4.11037713e-01 -6.42877102e-01 -6.80483997e-01 4.68147665e-01 2.84143388e-01 -1.10443592e+00 -3.23020309e-01 1.00081265e+00 -2.79167771e-01 4.77823645e-01 4.87807572e-01 7.88644612e-01 1.54494429e+00 6.36255927e-03 1.33203208e+00 1.24531448e+00 -2.81537473e-01 -8.30800757e-02 3.83879036e-01 2.96616286e-01 4.72770959e-01 -4.55706358e-01 -7.57199109e-01 -1.15863216e+00 9.67651904e-02 2.01160133e-01 3.57083619e-01 -2.13473827e-01 1.63825020e-01 -1.40199804e+00 5.91539562e-01 6.68985248e-01 5.57022631e-01 -2.89459914e-01 -4.95448522e-02 5.44954896e-01 2.23655224e-01 5.76580822e-01 4.76727001e-02 -3.38547796e-01 -4.22700822e-01 -6.16070151e-01 3.85159887e-02 4.74409223e-01 9.41694260e-01 7.42508471e-01 -2.29445070e-01 -1.13818251e-01 1.02837861e+00 4.34790105e-01 8.45060885e-01 7.14174867e-01 -5.07366657e-01 6.33250654e-01 1.01639760e+00 5.83644435e-02 -1.56705058e+00 -6.99470580e-01 2.20492482e-01 -8.49549770e-01 -1.96526736e-01 1.68291628e-01 -4.31782544e-01 -4.14693624e-01 1.58390343e+00 4.09682631e-01 -2.72387803e-01 7.30397254e-02 9.82526600e-01 1.52869701e+00 1.07234943e+00 1.93077251e-01 -1.10176317e-01 1.82964420e+00 -1.01939619e+00 -1.16193128e+00 -2.61878550e-01 4.49172974e-01 -1.16321719e+00 1.73685503e+00 2.26487279e-01 -9.36697483e-01 -4.81559843e-01 -9.68634963e-01 -2.66630411e-01 -7.74174571e-01 3.53222519e-01 3.23123217e-01 2.56848305e-01 -7.65327632e-01 2.02695414e-01 -6.89687371e-01 -6.14501297e-01 3.78375739e-01 -1.43884070e-04 -2.92593241e-01 1.74823791e-01 -1.13470185e+00 5.62469065e-01 -1.07572496e-01 4.56894368e-01 9.39661190e-02 -2.56659120e-01 -8.96394014e-01 -2.04110667e-01 3.55115920e-01 -3.76726359e-01 1.39893448e+00 -1.50953174e+00 -1.59230006e+00 7.40993261e-01 -6.05066359e-01 1.99268296e-01 9.04042125e-02 -3.42076123e-01 -5.87807834e-01 3.60591501e-01 8.19761604e-02 6.15119517e-01 7.95532465e-01 -1.45191801e+00 -4.68217850e-01 -4.57870275e-01 1.72756985e-01 5.63371837e-01 -8.94591749e-01 3.50306809e-01 -6.66585982e-01 -7.66246557e-01 1.75878238e-02 -1.06629086e+00 2.75426591e-03 -4.20537889e-01 -5.73123157e-01 -1.34581164e-01 1.05433834e+00 -2.66661823e-01 1.41516924e+00 -2.22420096e+00 2.13897392e-01 -1.13653429e-01 1.02674142e-02 -2.01458544e-01 -1.26488194e-01 1.12898540e+00 9.59001631e-02 1.48171261e-01 -1.46205738e-01 -8.63182425e-01 3.60112607e-01 -6.42506257e-02 -4.13884193e-01 2.51295805e-01 5.24027199e-02 1.06116664e+00 -7.68186748e-01 -6.75708890e-01 1.41612038e-01 6.00994349e-01 -1.29806185e-02 2.84028985e-02 -1.06077984e-01 6.05044007e-01 -5.95367193e-01 1.02487171e+00 6.14132166e-01 -4.31597412e-01 -1.17947266e-01 -6.88738942e-01 -3.29060674e-01 -3.67586166e-01 -8.68699431e-01 1.69322491e+00 -4.42865372e-01 8.51186693e-01 3.14840935e-02 -5.03946066e-01 8.30072880e-01 2.16793358e-01 4.61927325e-01 -8.01751733e-01 5.42270184e-01 -1.47370547e-01 -5.19852281e-01 -9.89980519e-01 1.07637644e+00 -2.51953483e-01 -4.96273309e-01 4.12444711e-01 -2.63284981e-01 -1.33726344e-01 3.29854041e-01 4.87888783e-01 3.76688659e-01 -2.39567719e-02 6.85856491e-02 1.10174969e-01 4.79610562e-01 -4.42529246e-02 2.12151095e-01 4.05330539e-01 -2.89063245e-01 6.93605244e-01 5.59897780e-01 -2.03619733e-01 -4.41695124e-01 -5.09314179e-01 1.39180213e-01 1.44672513e+00 6.45465851e-01 -9.45504904e-01 -6.63544714e-01 -3.73323977e-01 -3.63359779e-01 1.27730355e-01 -8.24301541e-01 2.49615490e-01 -9.58527625e-02 -9.99661088e-01 1.09906778e-01 4.32493776e-01 6.35143042e-01 -1.25548160e+00 -3.33239883e-01 -9.02194455e-02 -8.64163518e-01 -1.26553893e+00 -5.87195575e-01 -3.95463169e-01 -3.77040118e-01 -1.04440725e+00 -8.57216656e-01 -8.90431762e-01 7.04096198e-01 5.37523925e-01 9.26534176e-01 3.66400108e-02 -1.12090427e-02 6.82129145e-01 -9.88488019e-01 -7.37203658e-01 1.24245808e-01 4.08564880e-02 -5.45220859e-02 6.15302861e-01 5.92806041e-01 -1.58474579e-01 -8.55106115e-01 5.12510121e-01 -1.13676345e+00 2.59202868e-01 3.62783790e-01 6.02865934e-01 4.52475250e-01 -3.54548767e-02 5.48977256e-01 -7.95350492e-01 8.02038729e-01 -8.06468844e-01 -1.24052621e-01 2.36959569e-02 6.76603690e-02 -7.72674203e-01 5.66887140e-01 -6.28084421e-01 -1.15484035e+00 3.47850099e-02 2.37519983e-02 -5.77023849e-02 -1.16413981e-01 8.27259958e-01 6.38745949e-02 1.46963179e-01 3.01684469e-01 2.88047418e-02 -7.32607841e-02 -2.71430463e-01 2.76492804e-01 1.01874053e+00 1.47359401e-01 -4.05046523e-01 4.48218852e-01 8.60039294e-01 -2.67004728e-01 -1.27757311e+00 -1.14670825e+00 -6.24210596e-01 -3.35660964e-01 -6.74771428e-01 1.06509292e+00 -1.03395474e+00 -1.18870699e+00 8.66908669e-01 -8.77604008e-01 -3.48486960e-01 2.67185003e-01 1.63806424e-01 -3.09091330e-01 6.97605610e-01 -7.31262743e-01 -8.68053675e-01 -4.33641702e-01 -1.13724852e+00 1.36411428e+00 5.51083028e-01 -4.45634782e-01 -9.63062167e-01 4.37699631e-02 7.64893651e-01 2.90715426e-01 3.59454155e-01 1.78370997e-01 -1.76041424e-01 -1.24816887e-01 -4.15873766e-01 -1.85379133e-01 1.94701731e-01 2.23515689e-01 3.91097099e-01 -1.05698788e+00 -6.24643043e-02 7.57165179e-02 -6.87324643e-01 7.01856196e-01 2.17122555e-01 1.11630750e+00 -3.82234514e-01 -8.01082049e-03 1.15422957e-01 1.14929724e+00 1.95320621e-02 5.36961019e-01 5.78748226e-01 8.66073072e-01 8.95096719e-01 9.09523487e-01 8.17351401e-01 1.03865647e+00 4.56036985e-01 5.43774128e-01 -3.57805580e-01 3.87033105e-01 -7.05141202e-02 7.27922499e-01 1.33999455e+00 1.45155974e-02 -4.37829524e-01 -8.87325406e-01 4.46415216e-01 -2.18257904e+00 -1.08998501e+00 -4.10121202e-01 1.44161749e+00 5.51692009e-01 -2.80014813e-01 2.94177979e-01 1.04815289e-01 6.04344487e-01 4.63501185e-01 -2.18452722e-01 -2.59066939e-01 -6.00645542e-01 -5.18501997e-01 -8.92039835e-02 3.31398398e-01 -1.30638325e+00 8.91792774e-01 4.98527908e+00 6.75187469e-01 -1.30415428e+00 -3.66174430e-02 4.84528899e-01 -2.94827819e-01 -2.55978525e-01 -3.37389052e-01 -6.59928799e-01 8.20568264e-01 5.59063196e-01 1.00088485e-01 7.25141540e-02 4.27648544e-01 3.95097941e-01 -2.99932152e-01 -7.02476919e-01 1.36499321e+00 4.57943738e-01 -1.07813537e+00 -7.33900741e-02 -4.02339876e-01 7.48077273e-01 -5.54920686e-03 4.06126559e-01 1.84487373e-01 -1.90287918e-01 -7.49484956e-01 8.85436296e-01 5.83032966e-01 3.56452495e-01 -7.20580876e-01 8.10456455e-01 3.86679657e-02 -1.41332245e+00 -1.08253621e-01 1.99836381e-02 -2.28365585e-02 3.54130864e-01 2.85976440e-01 -2.32757479e-01 5.04779637e-01 1.11181152e+00 1.18862414e+00 -6.96593344e-01 5.77195764e-01 -1.27908170e-01 4.64683533e-01 -2.58474886e-01 -6.03751361e-01 4.10173088e-01 -4.09117192e-01 3.84623706e-01 1.65371573e+00 2.46348694e-01 2.02590451e-01 8.59071463e-02 2.84125894e-01 4.82736304e-02 6.38341427e-01 -4.62380052e-01 -3.33155304e-01 2.27822661e-01 1.74932313e+00 -9.06366885e-01 -3.88592482e-01 -6.23665750e-01 1.15906620e+00 2.89723389e-02 5.39863467e-01 -7.85336375e-01 -3.59914064e-01 3.15429032e-01 -2.06537053e-01 1.72326323e-02 -2.04340756e-01 -1.70865893e-01 -1.47390270e+00 1.31580189e-01 -9.37868714e-01 4.65327919e-01 -1.27644455e+00 -1.66193593e+00 6.70455456e-01 -2.57045895e-01 -1.43471706e+00 3.72991502e-01 -5.69341242e-01 -7.65087664e-01 5.13272643e-01 -1.56218672e+00 -1.44022083e+00 -7.06467807e-01 8.74896944e-01 6.52726412e-01 1.96968108e-01 7.25252509e-01 3.02943408e-01 -8.81062925e-01 3.97723764e-01 -1.02963135e-01 -6.23321086e-02 9.76541638e-01 -1.24033046e+00 -2.28124075e-02 5.26350021e-01 -9.38699991e-02 4.78744954e-01 7.92673349e-01 -4.34757352e-01 -1.77519441e+00 -6.61232233e-01 9.21027899e-01 -6.06941342e-01 1.16580808e+00 -2.65649974e-01 -7.36774921e-01 6.71753049e-01 7.59689987e-01 -5.46268463e-01 1.24423563e+00 1.70250624e-01 -9.54116434e-02 -1.77616864e-01 -6.79519057e-01 1.04122245e+00 4.52497602e-01 -4.68672186e-01 -3.78103375e-01 3.64532709e-01 4.80162948e-01 -4.26889420e-01 -8.02626252e-01 1.40021756e-01 5.44919193e-01 -9.25289392e-01 6.95216238e-01 -1.23625308e-01 8.98431182e-01 -3.92610490e-01 -2.84437448e-01 -1.35999608e+00 3.07982862e-01 -6.98659062e-01 1.45745277e-01 1.60281277e+00 2.53595650e-01 -3.51475030e-01 5.04499853e-01 4.53691304e-01 -4.33904752e-02 -6.75526023e-01 -5.11678874e-01 -7.98944756e-02 -3.90245438e-01 -5.95601857e-01 5.94345093e-01 1.21081412e+00 6.71018004e-01 6.83163762e-01 -7.11105704e-01 1.95367053e-01 2.79879451e-01 5.02643406e-01 1.02594602e+00 -7.00255930e-01 1.50157928e-01 -6.77882731e-01 -1.51616290e-01 -8.89239490e-01 -3.83828245e-02 -6.02455080e-01 -1.01717897e-01 -1.57927668e+00 2.86461204e-01 -1.34880021e-01 -1.49758726e-01 5.47569454e-01 -2.48775065e-01 6.23973310e-01 5.08336067e-01 7.31154829e-02 -1.04492736e+00 8.05829287e-01 1.55108881e+00 -1.82605654e-01 -2.00535014e-01 -2.66839594e-01 -9.57792819e-01 9.69704330e-01 1.07985556e+00 -5.80663569e-02 -2.90680796e-01 -3.38083535e-01 7.94627488e-01 -1.49739042e-01 1.51751250e-01 -4.86916959e-01 3.41074109e-01 -2.67709464e-01 2.37890109e-01 -1.12364316e+00 8.38226080e-01 -9.19171154e-01 -1.44973725e-01 -8.17516297e-02 -2.06960514e-01 3.96574795e-01 3.45699966e-01 3.73215497e-01 -5.81186056e-01 -2.69307941e-02 4.99927431e-01 -9.29976553e-02 -7.61535108e-01 2.22936973e-01 -5.80149174e-01 -1.18940003e-01 7.25520670e-01 2.85437293e-02 -4.75919425e-01 -7.48754680e-01 -6.85816646e-01 6.46595240e-01 4.32249099e-01 6.17481470e-01 6.99442446e-01 -1.45581889e+00 -5.05891204e-01 -2.83049166e-01 5.38885772e-01 -3.70372087e-02 9.02179062e-01 9.00144398e-01 -1.60486996e-01 -1.23619638e-01 -7.27131292e-02 -5.57117343e-01 -1.60887861e+00 2.62199402e-01 -1.71467632e-01 2.75748581e-01 -3.31539899e-01 7.34782338e-01 1.36394143e-01 -2.99834758e-01 3.06596577e-01 -1.97798729e-01 -6.35185122e-01 9.22147393e-01 7.74625719e-01 2.44720250e-01 -3.04327726e-01 -1.18028736e+00 -1.41864359e-01 9.01318312e-01 7.63767511e-02 -2.63429922e-03 1.26694739e+00 -8.89654577e-01 -3.01533610e-01 1.09430528e+00 1.21732557e+00 1.83778286e-01 -9.13515568e-01 -1.91584110e-01 -3.25511724e-01 -3.23079944e-01 -2.98858702e-01 -8.34007859e-01 -8.98530245e-01 9.69494760e-01 2.77037621e-01 5.87336063e-01 1.27619016e+00 9.08315256e-02 1.03564250e+00 3.60989809e-01 -2.20567197e-01 -1.38551939e+00 5.79543352e-01 5.57877779e-01 1.27595389e+00 -1.77647901e+00 -1.29689604e-01 -2.64001191e-01 -1.42538929e+00 1.11513531e+00 5.72759688e-01 2.89446861e-01 5.43489099e-01 -1.18475389e-02 6.59761488e-01 -4.72225308e-01 -6.13676667e-01 -2.85701483e-01 3.20349604e-01 1.74822375e-01 5.38696408e-01 1.77006200e-02 -9.79404449e-02 9.47874546e-01 -4.82909441e-01 -2.35815212e-01 6.12559736e-01 1.02579141e+00 -2.61068016e-01 -8.75276625e-01 -5.73433459e-01 -7.78042749e-02 -5.58808148e-01 -1.22767366e-01 -6.29815876e-01 5.98574519e-01 -1.40773907e-01 1.31669164e+00 -3.64769660e-02 -4.29656625e-01 4.51753438e-01 -1.21109270e-01 1.11607552e-01 -3.07792276e-01 -7.13466883e-01 2.72136360e-01 1.06522061e-01 -4.69686329e-01 -1.04144657e+00 -6.74697757e-01 -1.31529498e+00 -4.53151047e-01 -5.11600301e-02 -7.72047266e-02 1.04425454e+00 6.92422152e-01 3.11649352e-01 4.19258505e-01 9.20100033e-01 -1.27487576e+00 2.57529467e-01 -9.32929993e-01 -5.15829206e-01 6.51007891e-01 4.98940557e-01 -4.88732249e-01 -3.04283261e-01 2.11774498e-01]

[13.063501358032227, 5.246790885925293]

e6d91af0-2758-4a75-9dac-37ef5a70372e

target-conditioned-sampling-optimizing-data

1905.08212

https://arxiv.org/abs/1905.08212v1

https://arxiv.org/pdf/1905.08212v1.pdf

Target Conditioned Sampling: Optimizing Data Selection for Multilingual Neural Machine Translation

To improve low-resource Neural Machine Translation (NMT) with multilingual corpora, training on the most related high-resource language only is often more effective than using all data available (Neubig and Hu, 2018). However, it is possible that an intelligent data selection strategy can further improve low-resource NMT with data from other auxiliary languages. In this paper, we seek to construct a sampling distribution over all multilingual data, so that it minimizes the training loss of the low-resource language. Based on this formulation, we propose an efficient algorithm, Target Conditioned Sampling (TCS), which first samples a target sentence, and then conditionally samples its source sentence. Experiments show that TCS brings significant gains of up to 2 BLEU on three of four languages we test, with minimal training overhead.

['Xinyi Wang', 'Graham Neubig']

2019-05-20

target-conditioned-sampling-optimizing-data-1

https://aclanthology.org/P19-1583

https://aclanthology.org/P19-1583.pdf

acl-2019-7

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

['natural-language-processing']

[ 1.19914465e-01 -1.31187662e-01 -6.09809995e-01 -3.23676050e-01 -1.53187037e+00 -5.51123083e-01 5.74709237e-01 -1.61795706e-01 -8.71508181e-01 1.33966458e+00 2.08582968e-01 -8.39182198e-01 4.22450751e-01 -5.19014001e-01 -8.82782340e-01 -4.84351367e-01 3.12799096e-01 8.00551534e-01 -4.22522843e-01 -8.97463560e-02 -2.05203846e-01 2.30697885e-01 -8.64056468e-01 2.91447014e-01 1.35178363e+00 3.88140529e-01 8.38958561e-01 2.12091997e-01 -1.75069571e-01 5.40838778e-01 -5.46554983e-01 -6.50614679e-01 4.25374299e-01 -7.04809606e-01 -8.60107243e-01 -2.20364407e-01 4.14684772e-01 -1.08467326e-01 -2.68375911e-02 1.12713099e+00 6.64263368e-01 -5.22277057e-02 5.72995186e-01 -6.27640724e-01 -5.49516559e-01 1.31920815e+00 -5.39592564e-01 1.78956166e-01 -1.96547452e-02 7.05481023e-02 1.14732385e+00 -1.39940250e+00 8.76829922e-01 1.27530766e+00 3.52339149e-01 7.03492701e-01 -1.40515661e+00 -7.88424253e-01 1.30529374e-01 -1.34555370e-01 -1.45782149e+00 -7.21379995e-01 4.47793782e-01 7.51685426e-02 1.08802032e+00 1.24730445e-01 4.35371548e-01 1.34612656e+00 2.26892218e-01 1.30168247e+00 1.40477169e+00 -7.50216961e-01 -4.97606732e-02 2.86880910e-01 -3.42333138e-01 5.35407722e-01 1.50396645e-01 -9.23401117e-02 -6.47733867e-01 -1.34838475e-02 3.50483239e-01 -4.81598169e-01 -3.20238531e-01 -6.31795600e-02 -1.47402394e+00 8.19848418e-01 -1.16499318e-02 3.31720054e-01 -3.36162925e-01 -1.91183120e-01 3.21783245e-01 7.28520453e-01 9.78055894e-01 4.55169678e-01 -9.30314422e-01 -2.24744514e-01 -1.11084938e+00 -1.59536302e-02 7.61712611e-01 1.27930129e+00 9.42954063e-01 1.91812381e-01 -1.09615527e-01 1.19523335e+00 1.56388190e-02 1.01224554e+00 5.75072527e-01 -4.77758825e-01 1.07547164e+00 1.26625970e-01 -2.86816031e-01 8.71034432e-03 3.91706415e-02 -5.17579079e-01 -8.11721623e-01 -3.71841013e-01 4.27283973e-01 -6.84687734e-01 -7.43535280e-01 2.02655840e+00 1.22037701e-01 -4.53720599e-01 4.62548703e-01 7.85203040e-01 2.07401559e-01 8.24674487e-01 -2.84056254e-02 -5.34374237e-01 9.90635872e-01 -9.95780945e-01 -6.70066535e-01 -6.27754390e-01 1.08121812e+00 -1.08665490e+00 1.32593179e+00 2.06787154e-01 -1.05981767e+00 -1.89737871e-01 -9.37286198e-01 -1.87923729e-01 -1.29234642e-01 4.94370908e-01 7.09706962e-01 4.82467592e-01 -1.00612748e+00 4.94724095e-01 -8.07329118e-01 -3.39223176e-01 2.31331855e-01 3.30692023e-01 -4.18784738e-01 -4.29874986e-01 -1.33316529e+00 1.24107468e+00 4.39254165e-01 -5.30882999e-02 -9.84236956e-01 -5.87188721e-01 -7.17785537e-01 -2.74152517e-01 3.03836733e-01 -5.75302541e-01 1.23331642e+00 -1.29481518e+00 -1.74152434e+00 7.48405099e-01 -3.07816118e-01 -5.79751849e-01 6.48230553e-01 -3.37991059e-01 -1.41463101e-01 -3.04822326e-01 2.45557234e-01 7.43046641e-01 5.09607911e-01 -8.96425784e-01 -7.42155552e-01 -1.72499403e-01 -2.48490751e-01 5.36570132e-01 -6.04730666e-01 3.78422052e-01 -6.34306908e-01 -6.19473994e-01 -1.60082757e-01 -1.09918177e+00 -1.78329870e-01 -6.18899763e-01 -4.52844203e-01 -2.58611381e-01 2.40793094e-01 -9.95952129e-01 1.00612473e+00 -1.86668634e+00 4.74264562e-01 -2.59392373e-02 -2.92042822e-01 1.61824465e-01 -4.81721073e-01 3.37018400e-01 2.30877206e-01 1.91924974e-01 -1.36830732e-01 -4.85977829e-01 -2.69229650e-01 2.93560535e-01 -1.74005926e-01 3.31732780e-01 2.91312277e-01 8.86327505e-01 -8.61694157e-01 -5.56760371e-01 -1.76204264e-01 2.91601539e-01 -6.19581640e-01 -1.68587267e-02 -3.80545467e-01 5.94219387e-01 -2.58714110e-01 6.18121803e-01 4.51321959e-01 3.70745361e-02 5.92455029e-01 1.08628914e-01 -2.26473317e-01 8.66755784e-01 -6.68491483e-01 1.98203969e+00 -9.74284410e-01 7.72909045e-01 4.76432554e-02 -6.47822380e-01 8.27691793e-01 4.36675757e-01 2.45546922e-01 -7.34960616e-01 2.69713234e-02 7.56218672e-01 2.70470619e-01 -1.17009141e-01 4.46868420e-01 -1.23535387e-01 -8.48691836e-02 5.51277578e-01 2.61366397e-01 6.89899996e-02 3.46460909e-01 1.00181140e-01 7.13438570e-01 3.13021123e-01 2.34038889e-01 -6.06873989e-01 1.60064906e-01 1.86540648e-01 7.80095696e-01 5.61622322e-01 6.73826411e-02 2.34716699e-01 2.01055527e-01 -3.47458683e-02 -1.27582645e+00 -9.71110106e-01 -1.56408384e-01 1.13036919e+00 -5.61640501e-01 -3.17305714e-01 -8.91611218e-01 -8.73057604e-01 -3.11505526e-01 8.68911743e-01 -9.85009968e-02 1.60914510e-01 -9.24588263e-01 -8.43292594e-01 5.28311312e-01 5.36524653e-02 3.47759962e-01 -8.97126436e-01 -2.15298366e-02 1.09259382e-01 -6.68864131e-01 -9.65939403e-01 -8.41508985e-01 4.86816227e-01 -1.02863598e+00 -4.17076170e-01 -9.12457883e-01 -9.29549336e-01 7.65228152e-01 2.38951772e-01 1.28240013e+00 -3.62287670e-01 3.31289202e-01 -3.08114588e-01 -2.57235289e-01 -3.75220358e-01 -7.45205641e-01 7.58143067e-01 4.80021536e-01 -2.36444235e-01 4.15114999e-01 -2.52536923e-01 3.12760696e-02 -8.11668858e-02 -3.69705379e-01 4.04075801e-01 1.03343916e+00 9.81910706e-01 7.82599211e-01 -1.91092417e-01 6.60524130e-01 -1.04042804e+00 6.16455615e-01 -5.06014109e-01 -5.85689247e-01 5.29699206e-01 -6.77179456e-01 3.03524613e-01 9.77657974e-01 -6.27451658e-01 -1.06697965e+00 -4.63653766e-02 2.48251315e-02 -4.29954529e-01 3.10630709e-01 9.19504583e-01 -4.44162518e-01 2.01733053e-01 5.64016759e-01 4.63726103e-01 -2.05631956e-01 -7.59129524e-01 3.48864198e-01 9.13903475e-01 9.82020795e-02 -1.02960598e+00 6.19920015e-01 -1.66713744e-01 -3.56517971e-01 -8.14813673e-01 -9.36368704e-01 7.60916388e-03 -6.41884685e-01 3.67988162e-02 4.61727381e-01 -1.22099602e+00 2.49491408e-02 1.07657246e-01 -1.17944205e+00 -6.20679736e-01 -1.16358593e-01 9.75599706e-01 -4.89971399e-01 -9.26127564e-03 -7.38118470e-01 -6.81400001e-01 -4.86573577e-01 -1.37419891e+00 8.15924406e-01 -3.20433766e-01 -1.46060750e-01 -9.93991733e-01 -1.04482919e-01 3.26531202e-01 3.84805202e-01 -4.65393633e-01 9.50822771e-01 -7.74746656e-01 -7.00612843e-01 1.90401133e-02 4.37025875e-02 5.96469700e-01 2.52689600e-01 -2.48004586e-01 -7.35637307e-01 -7.36644447e-01 1.12065701e-02 -5.81593454e-01 6.23513401e-01 2.37701133e-01 7.35137820e-01 -4.96234208e-01 -1.22343950e-01 6.03509188e-01 1.44496548e+00 -1.35340437e-01 1.88484937e-01 2.00348929e-01 7.10289001e-01 6.26980305e-01 8.54832530e-01 -1.16029657e-01 3.02679986e-01 5.73034644e-01 -2.91436017e-01 -1.32500574e-01 -2.19125211e-01 -5.15240133e-01 8.90516400e-01 1.48705554e+00 3.84577550e-02 -1.40788466e-01 -8.63059461e-01 6.13583565e-01 -1.53877938e+00 -5.77970803e-01 2.31864274e-01 2.40441585e+00 1.59759855e+00 3.23814601e-02 -6.10915385e-02 -4.41713005e-01 5.73952198e-01 -2.00517744e-01 -5.06437600e-01 -2.72315711e-01 -4.39376920e-01 2.09291011e-01 8.50086093e-01 6.49676561e-01 -7.37078190e-01 1.50922883e+00 6.43176126e+00 1.23180854e+00 -1.23519623e+00 4.69038337e-01 7.33841717e-01 -4.66426760e-01 -5.94455242e-01 6.49481639e-02 -1.12405252e+00 4.43230659e-01 1.37793374e+00 -4.45260197e-01 7.76275992e-01 6.82241559e-01 3.36073518e-01 1.49112090e-01 -1.20035124e+00 5.80901802e-01 4.82774079e-02 -1.05154943e+00 3.58214021e-01 2.73500919e-01 9.62109625e-01 7.05123723e-01 1.38088055e-02 5.17693698e-01 7.04287171e-01 -6.70834184e-01 8.66690755e-01 3.46300378e-02 1.18012428e+00 -9.97495949e-01 4.96242851e-01 8.55608523e-01 -7.23903894e-01 3.62621039e-01 -5.85745454e-01 2.33571216e-01 -5.41861579e-02 5.70337117e-01 -9.50555921e-01 6.23203814e-01 3.34379554e-01 5.60299754e-01 -3.26793790e-01 5.62020302e-01 -4.47327316e-01 8.78950715e-01 -3.75045508e-01 -2.41210237e-01 2.41642416e-01 -3.17068636e-01 5.13098776e-01 1.43009293e+00 5.14237046e-01 -4.96690601e-01 3.67055625e-01 5.28692365e-01 -5.29978096e-01 7.29821146e-01 -9.00344312e-01 -1.09884672e-01 6.08184099e-01 1.03611565e+00 -2.81500548e-01 -3.97407532e-01 -5.97993255e-01 1.20552886e+00 8.23334694e-01 4.45084035e-01 -5.13437808e-01 -2.01283600e-02 6.06496632e-01 -2.60062367e-01 1.74033782e-03 -3.10895801e-01 -1.80103123e-01 -1.56046844e+00 1.66668087e-01 -1.38337350e+00 2.01296777e-01 -2.75301248e-01 -1.24451506e+00 7.17032671e-01 -1.49403214e-01 -1.30661285e+00 -4.73207325e-01 -4.37329590e-01 9.67755169e-02 1.44826174e+00 -1.50975132e+00 -1.31815910e+00 7.98935652e-01 5.28911293e-01 8.84880543e-01 -4.82761472e-01 9.69626904e-01 6.70311272e-01 -6.90910876e-01 1.00731468e+00 3.88010830e-01 2.09168717e-01 8.85860324e-01 -1.13306570e+00 3.80194515e-01 1.10355103e+00 4.80224550e-01 7.96568990e-01 4.96178240e-01 -7.04532266e-01 -1.60433102e+00 -1.17559278e+00 1.58412147e+00 -3.35364759e-01 6.66174948e-01 -4.60297942e-01 -5.69573581e-01 9.15423155e-01 5.20003200e-01 -4.94791776e-01 6.75778508e-01 3.65207016e-01 -2.81462222e-01 -8.99269432e-02 -9.15997386e-01 1.03891671e+00 9.24827933e-01 -6.53839707e-01 -2.29365468e-01 6.58035576e-01 8.00401747e-01 -3.82864058e-01 -8.65561128e-01 2.79060841e-01 1.08203374e-01 -2.14063272e-01 6.52575254e-01 -7.02514887e-01 4.53563958e-01 2.46958639e-02 -3.74500573e-01 -1.92480934e+00 -3.77034433e-02 -7.08353996e-01 1.99400499e-01 1.22051668e+00 1.12631893e+00 -5.90394676e-01 5.32419503e-01 1.86588526e-01 -2.77765095e-01 -7.02378035e-01 -1.13213801e+00 -9.95213032e-01 7.02687800e-01 -4.23317045e-01 3.70978206e-01 1.12837982e+00 -8.88807513e-03 8.15690935e-01 -7.04174876e-01 -1.70375202e-02 7.15263844e-01 -6.64847419e-02 6.93805993e-01 -7.61317551e-01 -3.60289961e-01 -2.93752462e-01 4.41054195e-01 -1.28476691e+00 5.90993106e-01 -1.49272692e+00 1.99833289e-01 -1.43856740e+00 3.86918664e-01 -6.12678647e-01 -1.89798012e-01 6.28043890e-01 -2.28891507e-01 2.08624676e-01 1.42683387e-01 3.83497685e-01 -2.86491960e-01 4.99831408e-01 1.50920808e+00 -1.31306887e-01 -5.40057272e-02 -1.07871965e-01 -6.33068442e-01 3.87665451e-01 8.29847097e-01 -5.77729046e-01 -2.97075748e-01 -1.06294930e+00 1.36087283e-01 9.71008837e-02 -4.11497802e-01 -6.33845747e-01 -1.82056166e-02 -3.73543590e-01 2.11115882e-01 -5.97934186e-01 1.72155231e-01 -6.85321450e-01 -9.51052383e-02 4.27063346e-01 -5.34598470e-01 1.22668333e-01 1.28305838e-01 4.22820300e-02 -2.05859885e-01 -3.28511417e-01 8.16282153e-01 -2.24880964e-01 -1.13191813e-01 3.48642766e-01 -5.36042690e-01 1.89746395e-01 4.65123802e-01 4.48789090e-01 -8.37765355e-03 -2.49841183e-01 -1.61832511e-01 2.56946743e-01 4.07468736e-01 4.68639016e-01 2.49778092e-01 -1.42066622e+00 -1.25398719e+00 2.46911541e-01 5.88121191e-02 -1.95279837e-01 -3.08080822e-01 8.19793522e-01 -2.55532563e-01 7.03746259e-01 -2.39988025e-02 -3.78194004e-01 -1.04784691e+00 2.33754605e-01 1.87154457e-01 -4.19840932e-01 -3.71405214e-01 7.52383351e-01 -2.04436749e-01 -8.28720629e-01 -8.69857073e-02 -1.42407930e-03 1.70178011e-01 5.35768643e-02 3.83480340e-01 7.52128363e-02 1.62841678e-01 -6.60857975e-01 -2.07233846e-01 3.19688097e-02 -4.85700101e-01 -6.87047839e-01 1.10396469e+00 -1.65183887e-01 -2.79136360e-01 6.39475107e-01 1.24988914e+00 4.13870782e-01 -1.00165915e+00 -7.86761522e-01 7.66153410e-02 -4.62145209e-01 -6.81140646e-03 -8.43537509e-01 -9.54756677e-01 8.83006036e-01 1.72490194e-01 -2.47681648e-01 1.07075572e+00 -2.15841547e-01 8.06036055e-01 6.05308115e-01 8.62640738e-01 -1.21550536e+00 -4.65174794e-01 9.21381712e-01 6.70503318e-01 -1.23076570e+00 -1.31367564e-01 -1.07096091e-01 -6.75949216e-01 7.87320197e-01 5.81007481e-01 2.95421839e-01 3.41133624e-01 3.69815260e-01 2.71276832e-01 4.35696959e-01 -1.06181264e+00 -1.31176069e-01 1.35590822e-01 2.64068961e-01 8.69608521e-01 3.98738176e-01 -6.49270177e-01 2.80394197e-01 -4.81351435e-01 -2.27400437e-01 2.28789255e-01 6.06848776e-01 -2.71540403e-01 -1.59463990e+00 -1.24583200e-01 5.35632551e-01 -8.75748813e-01 -8.05200696e-01 -3.51588517e-01 6.52740359e-01 -1.27989054e-01 8.55069041e-01 -4.11399417e-02 -3.62283885e-01 -2.10058447e-02 2.81588614e-01 5.12644529e-01 -8.03044200e-01 -4.68217790e-01 4.08492029e-01 5.11242867e-01 -2.07466170e-01 -2.58024633e-01 -8.71878088e-01 -8.91321719e-01 -4.06321734e-01 -3.61226708e-01 3.48839223e-01 9.14020240e-01 9.75467861e-01 1.87974304e-01 9.87742841e-02 7.28062212e-01 -5.08907378e-01 -8.14258456e-01 -1.40060830e+00 -2.29177758e-01 2.89017055e-02 1.60266623e-01 -1.24627605e-01 -2.77022719e-01 -9.12793130e-02]

[11.637834548950195, 10.279297828674316]

14cda3bf-61d4-49c2-a04c-a32bd0e9cf20

an-overview-on-the-evaluated-video-retrieval

2306.13118

https://arxiv.org/abs/2306.13118v1

https://arxiv.org/pdf/2306.13118v1.pdf

An overview on the evaluated video retrieval tasks at TRECVID 2022

The TREC Video Retrieval Evaluation (TRECVID) is a TREC-style video analysis and retrieval evaluation with the goal of promoting progress in research and development of content-based exploitation and retrieval of information from digital video via open, tasks-based evaluation supported by metrology. Over the last twenty-one years this effort has yielded a better understanding of how systems can effectively accomplish such processing and how one can reliably benchmark their performance. TRECVID has been funded by NIST (National Institute of Standards and Technology) and other US government agencies. In addition, many organizations and individuals worldwide contribute significant time and effort. TRECVID 2022 planned for the following six tasks: Ad-hoc video search, Video to text captioning, Disaster scene description and indexing, Activity in extended videos, deep video understanding, and movie summarization. In total, 35 teams from various research organizations worldwide signed up to join the evaluation campaign this year. This paper introduces the tasks, datasets used, evaluation frameworks and metrics, as well as a high-level results overview.

['Georges Quenot', 'Yvette Graham', 'Jeffrey Liu', 'Lukas Diduch', 'Eliot Godard', 'Andrew Delgado', 'Yooyoung Lee', 'Afzal Godil', 'Jonathan Fiscus', 'Asad Butt', 'Keith Curtis', 'George Awad']

2023-06-22

['video-retrieval', 'video-understanding', 'ad-hoc-video-search', 'retrieval']

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

[ 3.59854877e-01 -6.17948174e-01 -2.04009414e-02 -4.39307302e-01 -1.47969043e+00 -8.54912043e-01 8.62499177e-01 5.77403486e-01 -8.07411134e-01 4.24099475e-01 7.19062746e-01 3.13072592e-01 -5.62091172e-02 -1.49380490e-01 -2.97055215e-01 -3.73250932e-01 -2.88379669e-01 3.59600186e-01 3.74018699e-01 -7.16248378e-02 8.32732439e-01 2.18607605e-01 -1.90684152e+00 6.21372879e-01 3.49487156e-01 1.17046142e+00 1.61681980e-01 1.13073754e+00 1.14801675e-01 8.25932026e-01 -5.87406576e-01 -5.11139333e-01 -1.74246673e-02 -1.94169328e-01 -1.07508945e+00 -1.85395956e-01 7.16287673e-01 -6.13181591e-01 -4.06487912e-01 8.37836087e-01 8.18264425e-01 3.30350041e-01 7.56837666e-01 -9.73583162e-01 -4.05737966e-01 6.48994893e-02 -3.66623342e-01 5.35640121e-01 1.03237176e+00 -1.80143248e-02 9.70375657e-01 -6.91650987e-01 9.47398961e-01 9.90369737e-01 3.59832257e-01 2.84308374e-01 -3.01531345e-01 -3.37401241e-01 -4.08624768e-01 5.95886528e-01 -1.51597035e+00 -5.89407027e-01 2.52074897e-01 -6.68954611e-01 1.24759769e+00 4.91100073e-01 3.87478113e-01 9.27672267e-01 1.54186353e-01 9.40035641e-01 4.49873447e-01 -5.30987799e-01 1.86764508e-01 -9.79057848e-02 3.20457757e-01 2.95301944e-01 2.74757575e-02 -2.64096975e-01 -5.52807033e-01 -1.21624447e-01 2.28021443e-01 -2.74569482e-01 -2.39613548e-01 -1.50950775e-01 -1.29311121e+00 6.23987794e-01 -8.92462358e-02 4.37529743e-01 -5.26614487e-01 1.25819638e-01 1.37075210e+00 5.04311562e-01 4.36800480e-01 3.41154665e-01 -3.63847613e-02 -7.53525794e-01 -1.20267642e+00 6.69206440e-01 6.34400606e-01 9.88441527e-01 2.70141631e-01 -2.38655552e-01 -4.71536338e-01 8.60672712e-01 4.03824717e-01 6.32700682e-01 4.91553128e-01 -1.49715412e+00 6.91711009e-01 2.65184343e-01 2.98013002e-01 -1.27090108e+00 -1.78843200e-01 4.16027755e-01 -2.98317492e-01 -3.24649930e-01 -1.86203495e-01 7.33393207e-02 -9.40292060e-01 1.10465968e+00 -1.73345834e-01 -2.32615352e-01 4.65248637e-02 8.64291310e-01 1.15800643e+00 9.40912008e-01 7.65709952e-02 -2.96615392e-01 1.50614572e+00 -6.88028216e-01 -9.07544553e-01 2.04617038e-01 6.72490656e-01 -1.12450778e+00 5.16020417e-01 3.33628714e-01 -1.34627903e+00 -4.82657224e-01 -1.07653821e+00 -1.52501136e-01 -5.10942519e-01 -1.19325832e-01 1.46708518e-01 8.49582374e-01 -1.29710889e+00 3.60963315e-01 -5.50927162e-01 -9.27053928e-01 1.03407331e-01 -1.24020211e-01 -5.50666332e-01 -4.53238577e-01 -1.27650380e+00 8.43824089e-01 5.52949488e-01 -6.43684641e-02 -1.20587671e+00 -2.81420529e-01 -8.78481865e-01 -7.66968131e-02 2.52064645e-01 -1.92289278e-01 1.07504320e+00 -5.53479493e-01 -9.29506242e-01 1.26679778e+00 -7.59106725e-02 -6.36196256e-01 7.44332969e-02 -3.41940761e-01 -7.23971903e-01 6.80116892e-01 2.01127708e-01 8.36221516e-01 6.98279023e-01 -5.61861038e-01 -8.31599653e-01 -5.32986104e-01 5.34606129e-02 5.45721233e-01 -3.77163410e-01 1.11881757e+00 -1.17748141e+00 -5.78307688e-01 -2.06170872e-01 -8.61333251e-01 1.80376500e-01 -5.24148166e-01 7.32394978e-02 -2.73455560e-01 8.31559896e-01 -9.86988902e-01 1.54026818e+00 -2.04598951e+00 -1.80442980e-03 2.50356309e-02 2.07908690e-01 4.72108155e-01 -3.51345092e-01 7.93851614e-01 1.11244872e-01 3.09266776e-01 2.83310175e-01 -1.69184700e-01 -1.01585902e-01 -1.40631422e-01 -7.79193416e-02 3.22380453e-01 -1.86988309e-01 5.96983910e-01 -7.34847903e-01 -7.17480719e-01 2.72385508e-01 3.06446731e-01 -2.04981819e-01 -4.18277979e-02 4.42815945e-02 2.08949760e-01 -5.47196925e-01 7.32991159e-01 4.83108312e-01 -1.61066756e-01 2.89967842e-02 -1.69563398e-01 -3.61464500e-01 -3.76401804e-02 -1.03088474e+00 1.86371708e+00 1.49673179e-01 1.33124352e+00 -6.09035306e-02 -9.71986413e-01 5.06390035e-01 8.95329058e-01 7.17057705e-01 -8.27095985e-01 2.35545591e-01 2.11381346e-01 -8.11479211e-01 -8.85036647e-01 1.27157593e+00 5.18035829e-01 -2.58042455e-01 5.29813528e-01 1.51551932e-01 1.88594237e-02 8.79516006e-01 6.60445750e-01 1.15472555e+00 -4.24803309e-02 1.00047782e-01 1.04125492e-01 7.67024219e-01 2.53766358e-01 1.88756779e-01 7.82505631e-01 -5.87167323e-01 7.58453965e-01 6.89689592e-02 -6.26776695e-01 -1.37636268e+00 -5.56971550e-01 -8.02009776e-02 1.10354674e+00 8.70818049e-02 -9.80096638e-01 -1.07799375e+00 -3.30267459e-01 -5.38697779e-01 2.31063277e-01 -3.33512843e-01 1.75274074e-01 -5.27121186e-01 -8.76124874e-02 1.07089007e+00 3.26312155e-01 5.72993457e-01 -1.20734501e+00 -6.87156022e-01 4.15516049e-02 -8.11381817e-01 -1.38165426e+00 -6.16385281e-01 -5.82273066e-01 -6.00343227e-01 -1.22360826e+00 -1.32764328e+00 -8.35211158e-01 -4.95679006e-02 5.20536184e-01 9.89084601e-01 -9.88662019e-02 -5.27534366e-01 1.15827966e+00 -8.17223549e-01 -2.73398668e-01 -2.60254830e-01 -1.60169587e-01 -2.75040921e-02 -4.50129896e-01 7.29574442e-01 2.48299927e-01 -5.51804125e-01 4.48319048e-01 -1.15554488e+00 -4.16293561e-01 1.14298925e-01 2.53330529e-01 5.17496467e-01 -1.05960980e-01 2.90176600e-01 -2.44823277e-01 9.30790246e-01 -2.85731763e-01 -5.44694841e-01 6.91959262e-01 -5.27266026e-01 -2.13198513e-01 -3.89733970e-01 2.34360889e-01 -1.05503857e+00 -1.94044724e-01 1.25318943e-02 -3.25698286e-01 -2.58190453e-01 8.80915940e-01 2.19012603e-01 -1.20910697e-01 6.83960199e-01 1.71278119e-01 -1.09071694e-01 -3.36550176e-01 -8.28722678e-03 9.67816293e-01 5.84355414e-01 -4.35164183e-01 1.98290572e-01 1.71944410e-01 -2.67883778e-01 -1.21871138e+00 -6.69089854e-01 -1.12315035e+00 -3.30488414e-01 -9.28197563e-01 1.40871489e+00 -1.16825628e+00 -4.51485902e-01 6.91266060e-01 -1.29169345e+00 3.55214477e-01 1.58449367e-01 6.09721959e-01 -5.68694174e-01 8.21519256e-01 -7.16074288e-01 -6.16967440e-01 -5.68371058e-01 -1.30384588e+00 1.21757865e+00 1.50494903e-01 -2.54563242e-01 -7.87009835e-01 5.04439831e-01 9.65826988e-01 5.84348440e-01 2.96831548e-01 3.64844710e-01 -5.61554193e-01 -5.34155130e-01 -7.57925868e-01 -3.30147624e-01 5.93902647e-01 -4.16152060e-01 2.47934431e-01 -7.79668510e-01 -4.11695570e-01 -2.80579954e-01 -6.64793789e-01 8.44447553e-01 7.24862635e-01 9.14671421e-01 6.88681677e-02 -1.21939078e-01 1.13202788e-01 1.28853083e+00 4.51949149e-01 1.12456822e+00 4.36593890e-01 2.78095126e-01 7.51702011e-01 8.73761237e-01 4.77970511e-01 2.76657224e-01 8.95545363e-01 1.13904499e-01 6.13208950e-01 1.55528694e-01 2.18624398e-01 2.96030521e-01 9.01380181e-01 -6.84190154e-01 -5.78320205e-01 -9.75417852e-01 4.17979598e-01 -1.76071525e+00 -1.32031989e+00 -1.97177142e-01 2.40268970e+00 2.43262783e-01 -1.00589931e-01 -7.00044557e-02 9.86065343e-02 7.84891844e-01 3.58214378e-01 -1.22222140e-01 -3.37865531e-01 -2.80511647e-01 -3.28939706e-01 4.79997486e-01 9.18759480e-02 -1.33970118e+00 7.14286149e-01 7.04573870e+00 1.02123237e+00 -6.25101507e-01 1.09789163e-01 3.26020032e-01 -1.72773260e-04 1.03355153e-02 -2.58744389e-01 -6.55976772e-01 2.44422376e-01 1.40517473e+00 -5.67009211e-01 2.54788578e-01 4.87968266e-01 2.78126597e-01 -1.83328539e-01 -8.38880360e-01 1.29813814e+00 6.33779645e-01 -1.68123329e+00 5.11038676e-02 2.63443552e-02 6.88427806e-01 4.13537413e-01 -1.40734777e-01 4.35149282e-01 -3.13421667e-01 -6.13805056e-01 7.52935231e-01 5.89181960e-01 9.98886764e-01 -5.22264719e-01 9.94162560e-01 -1.54876038e-01 -1.20830333e+00 5.13203181e-02 -1.60837114e-01 2.36098453e-01 4.16543484e-01 -8.11123177e-02 -4.08743888e-01 6.16845548e-01 1.08534849e+00 7.64217436e-01 -6.37858450e-01 1.36305106e+00 3.36044729e-01 4.04130161e-01 4.99871522e-02 1.28880620e-03 3.39678675e-01 1.39835328e-01 8.40211272e-01 1.55312335e+00 1.20114908e-01 1.87897936e-01 1.54123247e-01 -2.33562142e-01 -7.68764243e-02 2.60092199e-01 -6.54847622e-01 -4.23365444e-01 3.74565870e-01 9.54945505e-01 -7.78751254e-01 -5.25975823e-01 -4.42765981e-01 9.73138988e-01 -4.50105101e-01 3.73761624e-01 -7.82752335e-01 -6.72468007e-01 4.34598088e-01 -9.87697691e-02 1.91317439e-01 -1.33567527e-01 3.31349581e-01 -9.46793377e-01 6.16837181e-02 -1.03213167e+00 7.81870484e-01 -1.08066928e+00 -7.63777256e-01 6.92657113e-01 4.86248612e-01 -1.43718171e+00 -6.32384837e-01 -4.01574790e-01 -2.70277947e-01 6.55939400e-01 -1.00053275e+00 -5.34217000e-01 -3.94007236e-01 6.47825658e-01 8.83428991e-01 -4.72757131e-01 8.84275615e-01 7.18741298e-01 -3.46836239e-01 4.26268250e-01 2.71339804e-01 6.51867464e-02 1.00038648e+00 -5.51904261e-01 1.77116394e-01 7.55848408e-01 -1.96494877e-01 5.04314899e-01 7.09130764e-01 -6.79848075e-01 -1.42083657e+00 -8.13237131e-01 1.03339744e+00 -4.21037406e-01 6.22490883e-01 2.23820120e-01 -4.48566049e-01 4.37836111e-01 4.12542731e-01 -5.31116247e-01 7.27273226e-01 -2.90599316e-01 -3.30620736e-01 -3.64532508e-02 -1.07460237e+00 2.21392378e-01 4.23455179e-01 -9.42306399e-01 -4.96407270e-01 6.71628773e-01 7.85703123e-01 -2.18380973e-01 -9.28888083e-01 3.23814064e-01 8.12881172e-01 -7.18034685e-01 8.75251532e-01 -6.05275810e-01 5.69197357e-01 -2.12768883e-01 -6.24249101e-01 -5.57787538e-01 -4.99573834e-02 -5.29117525e-01 7.90022388e-02 1.05868697e+00 2.38038659e-01 5.43804765e-02 5.91059744e-01 8.20419908e-01 -3.89654726e-01 4.08632681e-02 -7.13526905e-01 -6.68729365e-01 -4.61845368e-01 -9.00974691e-01 1.73463091e-01 5.99078596e-01 -1.68538913e-02 1.43362522e-01 -6.09268904e-01 -2.30744302e-01 6.81169689e-01 -5.78593373e-01 6.31994784e-01 -1.17489457e+00 3.23386610e-01 -3.80765110e-01 -1.09894073e+00 -7.66482890e-01 -2.58521914e-01 -3.80725205e-01 -3.77550498e-02 -1.65319562e+00 5.47739923e-01 3.06945235e-01 -2.68861860e-01 3.84901427e-02 3.86765033e-01 5.61503589e-01 2.99221724e-01 5.72502613e-01 -1.56007659e+00 6.87104510e-03 9.42425072e-01 -3.49559098e-01 1.84397683e-01 -2.00815722e-01 -4.61922556e-01 4.84642446e-01 6.07930124e-01 -2.91973948e-01 -1.25048071e-01 -6.66055262e-01 4.70278352e-01 3.36776912e-01 2.04546854e-01 -1.13093507e+00 4.82406676e-01 1.33712471e-01 1.46414295e-01 -1.09986281e+00 5.16104221e-01 -6.14812315e-01 1.36602893e-01 1.95297960e-03 -6.25994980e-01 2.96518177e-01 1.14958793e-01 4.58592564e-01 -8.08453798e-01 -6.41793072e-01 3.86109143e-01 -1.67670727e-01 -1.26796412e+00 2.06830323e-01 -6.64220989e-01 1.10951483e-01 1.10142863e+00 -3.96886945e-01 -5.41652739e-01 -8.88462842e-01 -5.09144962e-01 5.33665955e-01 4.08812404e-01 6.12780571e-01 7.56315053e-01 -1.19036198e+00 -9.87104237e-01 -1.91841096e-01 5.38795173e-01 -7.77716756e-01 7.44874418e-01 7.43819714e-01 -1.06243193e+00 1.07848787e+00 -2.92416573e-01 -5.89922547e-01 -1.79505897e+00 3.53560984e-01 -1.94619834e-01 -4.37388808e-01 -3.35448742e-01 4.57994252e-01 -3.04496288e-01 1.55965865e-01 6.43355906e-01 2.83298969e-01 -7.11557150e-01 2.41093054e-01 1.32824826e+00 6.60957932e-01 4.08121973e-01 -1.01060426e+00 -2.70888299e-01 6.17295861e-01 -3.61205071e-01 -3.11406791e-01 1.08756495e+00 -5.25999069e-01 2.36336254e-02 1.70343697e-01 1.51939976e+00 -5.90981126e-01 -6.06531322e-01 8.51476938e-02 2.28534549e-01 -3.28562051e-01 1.85014188e-01 -6.28417552e-01 -7.50258982e-01 7.13003695e-01 8.73558104e-01 4.95225877e-01 1.22354412e+00 -1.14189513e-01 7.68518746e-01 7.30190694e-01 3.78066599e-01 -1.49597108e+00 7.38022700e-02 8.71294558e-01 9.54338074e-01 -1.19160736e+00 1.39217302e-01 1.14309363e-01 -8.22032690e-01 9.84569728e-01 -5.26294075e-02 2.21979052e-01 4.53145236e-01 -1.76185697e-01 -4.16720994e-02 -4.85640734e-01 -7.60440946e-01 -3.94384973e-02 5.30737698e-01 6.41974032e-01 6.92383051e-01 -1.18378036e-01 -5.33373654e-01 -7.14970529e-02 3.11906725e-01 1.93194807e-01 3.97838414e-01 1.16603208e+00 -6.67439401e-01 -8.64033222e-01 -3.78817201e-01 4.00539368e-01 -9.34590280e-01 -3.86345126e-02 -2.11252153e-01 5.22679031e-01 -5.44440091e-01 1.11060357e+00 4.79517356e-02 -3.60404432e-01 2.82115608e-01 -1.06362812e-01 3.85615408e-01 -2.99720168e-01 -4.74833488e-01 3.76224443e-02 5.74989557e-01 -8.85578096e-01 -7.53850937e-01 -8.64877701e-01 -8.60314608e-01 -3.07271719e-01 -7.86972046e-02 5.31648934e-01 1.19695413e+00 7.39993572e-01 3.70952457e-01 9.85112786e-02 4.64658797e-01 -1.15298271e+00 -2.36430869e-01 -1.02607703e+00 -4.60246146e-01 4.61598903e-01 -3.26046981e-02 -3.01923960e-01 -2.09790215e-01 4.28874999e-01]

[10.463508605957031, 0.7297830581665039]

27733d3e-7e30-449d-8563-3c2cc6203e59

discriminative-representation-combinations

1808.08802

http://arxiv.org/abs/1808.08802v2

http://arxiv.org/pdf/1808.08802v2.pdf

Discriminative Representation Combinations for Accurate Face Spoofing Detection

Three discriminative representations for face presentation attack detection are introduced in this paper. Firstly we design a descriptor called spatial pyramid coding micro-texture (SPMT) feature to characterize local appearance information. Secondly we utilize the SSD, which is a deep learning framework for detection, to excavate context cues and conduct end-to-end face presentation attack detection. Finally we design a descriptor called template face matched binocular depth (TFBD) feature to characterize stereo structures of real and fake faces. For accurate presentation attack detection, we also design two kinds of representation combinations. Firstly, we propose a decision-level cascade strategy to combine SPMT with SSD. Secondly, we use a simple score fusion strategy to combine face structure cues (TFBD) with local micro-texture features (SPMT). To demonstrate the effectiveness of our design, we evaluate the representation combination of SPMT and SSD on three public datasets, which outperforms all other state-of-the-art methods. In addition, we evaluate the representation combination of SPMT and TFBD on our dataset and excellent performance is also achieved.

['Tianwei Lin', 'Xu Zhao', 'Xiao Song', 'Liangji Fang']

2018-08-27

['face-presentation-attack-detection']

['computer-vision']

[ 1.04253873e-01 -7.57527530e-01 3.91742811e-02 -2.42376864e-01 -8.29163432e-01 -4.13683534e-01 5.67815304e-01 -2.55097449e-01 2.06981927e-01 -8.78933445e-03 3.80585700e-01 -1.33358359e-01 8.03512782e-02 -6.95467591e-01 -3.95561874e-01 -8.29163790e-01 1.10292390e-01 -4.21737462e-01 2.72951216e-01 -1.66901469e-01 4.87484872e-01 9.23210382e-01 -1.60951245e+00 8.73628855e-01 2.66962051e-01 1.57032454e+00 -1.53051749e-01 2.35335052e-01 1.76034987e-01 5.02138376e-01 -6.57633841e-01 -7.13663340e-01 3.37540627e-01 -1.93322092e-01 -2.19803557e-01 1.19010165e-01 5.69634795e-01 -8.86506915e-01 -6.41064048e-01 1.07494342e+00 1.06643271e+00 -3.67301792e-01 8.51602077e-01 -1.24073350e+00 -6.65846884e-01 -1.20854311e-01 -1.20840442e+00 3.13139170e-01 7.73269594e-01 2.24866793e-01 4.16117072e-01 -1.19821310e+00 5.42264342e-01 1.72666788e+00 6.55599296e-01 4.37230289e-01 -7.27479219e-01 -1.09454131e+00 -2.87637375e-02 5.37569940e-01 -1.60018098e+00 -6.26670182e-01 1.16049254e+00 -3.76208395e-01 5.04823923e-01 3.45883131e-01 1.41756818e-01 1.26561272e+00 3.97739142e-01 8.46214473e-01 1.09362102e+00 -3.23013961e-01 -1.31751314e-01 7.08418638e-02 -9.14735422e-02 9.45378423e-01 1.66038036e-01 2.84037679e-01 -5.27572155e-01 -5.89545071e-01 7.30741262e-01 3.08245510e-01 -2.07010731e-01 1.49888635e-01 -4.49931949e-01 6.98914111e-01 2.58662641e-01 3.62691641e-01 -2.47169003e-01 -1.11685663e-01 5.27463853e-01 2.03224495e-01 2.58092314e-01 -3.07118148e-01 7.95173943e-02 3.24604809e-01 -5.98080099e-01 5.60181327e-02 2.85801232e-01 5.63741028e-01 4.31248486e-01 -5.46253771e-02 -7.94252634e-01 1.01510108e+00 5.54869235e-01 5.47200441e-01 4.23379838e-01 -6.80322587e-01 4.62351143e-01 4.54543710e-01 -2.41378799e-01 -1.54148066e+00 -1.10073417e-01 -1.72387362e-02 -8.39970291e-01 1.98535681e-01 -5.36974482e-02 4.71904278e-02 -7.00317383e-01 1.51152980e+00 3.57757628e-01 3.38398010e-01 -9.30794477e-02 9.73076046e-01 1.07871222e+00 4.62789625e-01 -5.27923927e-02 -3.28198463e-01 1.77347553e+00 -5.05477428e-01 -7.24939346e-01 1.78488433e-01 4.65907335e-01 -1.05479991e+00 8.15981865e-01 6.27108276e-01 -8.09896171e-01 -6.40302062e-01 -1.03222418e+00 7.83103630e-02 -2.97072619e-01 5.64067364e-01 3.73809874e-01 1.02990007e+00 -1.05821872e+00 3.61770988e-01 -2.56930918e-01 -4.91668060e-02 6.49404824e-01 4.12599772e-01 -6.11990750e-01 -3.24968517e-01 -1.17458844e+00 5.46997011e-01 -2.49221146e-01 1.98101133e-01 -1.00534534e+00 -2.36453474e-01 -7.40574956e-01 5.77030890e-02 5.49923480e-02 -2.45124340e-01 9.58773255e-01 -7.29770720e-01 -1.51081324e+00 9.10624266e-01 -2.92310268e-01 2.70911455e-01 1.19854510e-01 2.53292263e-01 -7.04227805e-01 5.45246065e-01 -1.51382223e-01 3.35142463e-01 1.29801571e+00 -1.20553160e+00 -3.15643042e-01 -6.80887043e-01 -7.11337328e-02 2.79088225e-02 -6.80011749e-01 7.80769110e-01 -4.50619876e-01 -8.30766678e-01 9.94366482e-02 -3.67598295e-01 2.93996304e-01 3.95815879e-01 -3.79348576e-01 -4.01297241e-01 1.38693976e+00 -9.39738989e-01 1.42620075e+00 -2.31648922e+00 -1.59053966e-01 3.30681682e-01 4.05802339e-01 6.43809855e-01 -2.83335060e-01 3.26060683e-01 -1.45572364e-01 -8.20525736e-02 1.01021126e-01 -5.80537200e-01 8.28224048e-02 -3.05854321e-01 -4.17614996e-01 6.32867098e-01 5.05771637e-01 6.99144483e-01 -3.53482008e-01 -5.01179397e-01 1.53593510e-01 6.27453446e-01 -5.10777295e-01 2.45686218e-01 3.46421301e-01 6.11728244e-02 -6.69248164e-01 1.05230200e+00 1.44674873e+00 2.24843249e-01 -6.02954580e-03 -4.39740956e-01 5.89414835e-02 -4.59959023e-02 -1.09664297e+00 1.17000675e+00 -2.16470793e-01 3.64505082e-01 7.67102763e-02 -6.86364532e-01 1.15553129e+00 3.66264105e-01 2.16108978e-01 -9.40863073e-01 4.68746454e-01 7.33878836e-02 -3.35643321e-01 -6.22748017e-01 -3.82293016e-02 6.89769909e-02 1.20006718e-01 4.00322407e-01 -7.71224424e-02 4.02529448e-01 -3.52909535e-01 7.87626132e-02 1.15375066e+00 -2.26802111e-01 -6.72356337e-02 -7.56518319e-02 1.02086246e+00 -8.18563342e-01 5.77167511e-01 1.89979583e-01 -5.08236229e-01 7.60519505e-01 8.76545131e-01 -4.87872541e-01 -5.96080065e-01 -1.03313565e+00 -1.98514849e-01 9.10030007e-01 3.43349069e-01 -3.79856259e-01 -8.53062570e-01 -9.57608163e-01 1.46176383e-01 -9.45458934e-02 -6.47742867e-01 -3.04745495e-01 -3.95044893e-01 -7.07323194e-01 7.51217067e-01 5.13333857e-01 7.74374008e-01 -9.81232405e-01 -1.51540384e-01 -2.09007084e-01 1.19762355e-02 -9.42191064e-01 -6.89455092e-01 -5.45698583e-01 -2.65134692e-01 -1.03768754e+00 -7.81648338e-01 -8.72000456e-01 5.52142859e-01 6.35187685e-01 4.77508485e-01 3.30929965e-01 -7.06506729e-01 2.54079401e-01 -3.59337181e-01 -1.71414748e-01 9.36350878e-03 -6.81883931e-01 2.31200039e-01 5.48924983e-01 4.32984591e-01 -4.37230498e-01 -1.03474796e+00 4.42366779e-01 -8.60356152e-01 -4.13890958e-01 6.55873179e-01 5.04916966e-01 2.67911613e-01 4.10040058e-02 3.44284207e-01 -2.92495936e-01 8.52486372e-01 -3.34739625e-01 -3.71980160e-01 3.34158033e-01 7.56669492e-02 -3.89407933e-01 3.07029456e-01 -6.06902361e-01 -1.07597351e+00 -1.40233383e-01 -2.79153645e-01 -8.85388315e-01 -1.44564435e-01 1.62611201e-01 -5.43660760e-01 -5.38448095e-01 5.31704843e-01 3.23555231e-01 2.05803514e-01 -6.20621502e-01 -6.41865060e-02 1.19875252e+00 3.17012012e-01 -4.29906547e-01 7.01183140e-01 5.23347259e-01 1.18855573e-01 -6.67304635e-01 -3.51801932e-01 -3.95852625e-01 -3.51868093e-01 -2.59868294e-01 6.60863340e-01 -9.04907346e-01 -1.20348561e+00 9.32475388e-01 -1.40010715e+00 3.45571429e-01 5.81754804e-01 1.96109086e-01 -2.34081820e-01 9.92158234e-01 -9.38725710e-01 -1.00439954e+00 -5.06157339e-01 -1.32829165e+00 1.66093302e+00 1.90014079e-01 4.71138358e-01 -5.42641759e-01 -2.47477040e-01 3.04616779e-01 3.91387552e-01 4.90997970e-01 5.35258591e-01 -2.92106986e-01 -3.22362989e-01 -4.11667407e-01 -8.02980304e-01 5.15591085e-01 2.05567136e-01 1.43641159e-01 -1.16845214e+00 -5.45078874e-01 2.86168814e-01 -2.20753327e-01 8.79200041e-01 1.37324065e-01 1.60371590e+00 -3.17752451e-01 -4.08987850e-01 5.61845183e-01 1.21137595e+00 3.04637164e-01 9.26357925e-01 6.80612447e-03 5.99575222e-01 7.07182705e-01 5.41963577e-01 6.20039821e-01 1.61987498e-01 1.04589999e+00 1.93617940e-01 -2.02596858e-01 -2.86049008e-01 -1.27009302e-01 6.71099365e-01 4.30513501e-01 1.32802516e-01 -1.11401446e-01 -6.00636542e-01 -3.27201262e-02 -1.47744906e+00 -9.62764084e-01 4.73335311e-02 2.03302455e+00 3.26200038e-01 -1.72328830e-01 1.61123514e-01 3.40208590e-01 9.80012596e-01 1.04940668e-01 -1.61712423e-01 -4.49112892e-01 -1.70607597e-01 2.26291031e-01 -1.59897264e-02 1.62685663e-01 -1.22861516e+00 7.39172697e-01 5.65465593e+00 1.49266112e+00 -1.42484748e+00 2.78081179e-01 8.29146147e-01 2.75125504e-01 -2.94445418e-02 -4.43013251e-01 -9.26308870e-01 7.68468499e-01 4.48145509e-01 2.72923380e-01 2.04925463e-01 6.23856664e-01 2.52727307e-02 1.50797889e-01 -6.84621871e-01 1.39608276e+00 5.16268134e-01 -1.07780802e+00 2.70258307e-01 2.60475188e-01 3.54401469e-01 -7.16205657e-01 6.03468537e-01 1.01431362e-01 -3.86878639e-01 -8.98125768e-01 4.33629632e-01 4.28555608e-01 1.04673409e+00 -8.83547843e-01 7.12211728e-01 -1.39801010e-01 -1.42364609e+00 -5.55595636e-01 -6.05145872e-01 3.32968026e-01 -2.26148158e-01 2.85729438e-01 -3.40401143e-01 6.90372825e-01 5.18172860e-01 5.15515625e-01 -6.35738432e-01 1.06576538e+00 -1.03064910e-01 1.09026939e-01 -2.23778307e-01 2.12732375e-01 5.30040599e-02 1.04861565e-01 4.03340876e-01 1.24917412e+00 3.99097949e-01 1.66417271e-01 2.50387471e-02 6.24343157e-01 -3.90923694e-02 1.90768555e-01 -5.62082410e-01 2.01749638e-01 6.25869393e-01 1.40600336e+00 -3.66785139e-01 -1.70206815e-01 -3.34492594e-01 9.80557501e-01 9.20876637e-02 8.27127248e-02 -8.07389379e-01 -6.24013543e-01 7.32251525e-01 2.72395134e-01 4.35866475e-01 9.48409885e-02 1.87747031e-01 -1.01540637e+00 3.17761272e-01 -9.97447371e-01 4.70640481e-01 -7.87845433e-01 -1.45522690e+00 7.99437344e-01 -7.74918795e-02 -1.53001893e+00 9.27510560e-02 -9.85228539e-01 -1.02990377e+00 9.61006880e-01 -1.47588801e+00 -1.36333144e+00 -5.12762487e-01 8.43810320e-01 2.69094557e-01 -5.07887006e-01 6.00201309e-01 6.23797715e-01 -9.47326362e-01 1.13823831e+00 -3.77685501e-04 3.04291815e-01 7.23308146e-01 -3.03228080e-01 3.29539955e-01 6.20287001e-01 -2.31432423e-01 6.20902121e-01 2.00150814e-02 -6.58461273e-01 -1.64237356e+00 -9.54975128e-01 4.22026575e-01 -2.32166916e-01 2.42819026e-01 -4.49195504e-01 -7.97366440e-01 -4.77247797e-02 -8.01652744e-02 1.53009221e-01 6.05107248e-01 -3.35058540e-01 -8.90235782e-01 -3.95869285e-01 -1.60852492e+00 4.94353920e-01 8.88143420e-01 -7.70039618e-01 -2.58692682e-01 2.46385902e-01 4.74094331e-01 -2.38123298e-01 -6.74869955e-01 7.17073083e-01 8.36132705e-01 -1.18750060e+00 1.07542372e+00 -1.34757265e-01 1.60271212e-01 -3.88638854e-01 -4.07259911e-01 -5.44597626e-01 -4.83431756e-01 -6.09768867e-01 -1.53006211e-01 1.34823287e+00 -1.59228593e-01 -5.85617542e-01 6.61549866e-01 5.31135239e-02 1.04839593e-01 -9.46476042e-01 -1.01061714e+00 -8.02425623e-01 -2.19556436e-01 3.25675458e-02 9.13357377e-01 7.76615083e-01 -2.99410801e-02 -8.35571811e-02 -3.19328249e-01 2.46015742e-01 6.78248882e-01 -4.58959714e-02 4.01815325e-01 -9.64427650e-01 -9.16549489e-02 -5.19668818e-01 -9.01666522e-01 -7.91073501e-01 3.99045981e-02 -3.52384746e-01 -3.22143286e-01 -9.26542997e-01 5.11776745e-01 -4.55924422e-02 -4.99544054e-01 3.88944924e-01 -1.11671083e-01 7.98942626e-01 1.60089955e-01 1.93720311e-01 -3.58521283e-01 6.27620041e-01 1.18285048e+00 -9.24657881e-02 1.32868454e-01 -2.40058407e-01 -7.61762202e-01 4.98069435e-01 6.26893699e-01 -1.41213581e-01 -1.26213118e-01 -3.36840093e-01 -5.55364251e-01 1.32219180e-01 5.99013984e-01 -1.07734501e+00 1.12039767e-01 7.50022233e-02 1.01385498e+00 -6.63739979e-01 6.43889844e-01 -4.03393745e-01 -2.73501456e-01 5.06758392e-01 1.00164682e-01 -4.48133573e-02 3.55461985e-01 4.21243697e-01 -1.82453737e-01 7.42931366e-02 8.93506825e-01 2.96172202e-01 -4.66038823e-01 6.71543479e-01 -2.31666699e-01 -6.82817578e-01 1.07792556e+00 -4.52986509e-01 -5.02940893e-01 -2.65301049e-01 -2.91329473e-01 -1.76391840e-01 1.79133549e-01 6.96968973e-01 1.16492641e+00 -1.76923907e+00 -7.08349288e-01 7.49333143e-01 5.71727566e-02 -8.14917564e-01 6.07674241e-01 7.72012413e-01 -4.71891701e-01 2.95655131e-01 -4.62879270e-01 -4.77918476e-01 -1.70668840e+00 7.44101822e-01 1.10319741e-01 1.03039451e-01 -3.06931943e-01 1.01847601e+00 6.17565155e-01 3.04440167e-02 3.01074684e-01 1.93641633e-01 -4.20440316e-01 -1.13484543e-02 1.11389971e+00 1.37075394e-01 9.50968117e-02 -8.65381896e-01 -5.79117298e-01 8.04007590e-01 -3.37436348e-01 1.08825602e-02 8.94725561e-01 -2.97618248e-02 -2.60559648e-01 -5.97977579e-01 1.66281188e+00 -2.93575861e-02 -1.06860948e+00 -1.74688995e-01 -2.70573527e-01 -1.02115130e+00 1.93776593e-01 -5.90569496e-01 -1.29145038e+00 1.30133712e+00 1.11073542e+00 -1.00986257e-01 1.52763581e+00 -1.80193216e-01 8.97174180e-01 -7.36060087e-03 3.95733476e-01 -6.04119122e-01 6.51845276e-01 1.15562454e-01 1.20870757e+00 -9.16040063e-01 -1.66877642e-01 -7.46282876e-01 -6.65890515e-01 1.14386296e+00 6.49974346e-01 -1.49374053e-01 8.00819218e-01 2.07492337e-01 -1.03289135e-01 -2.68992782e-01 -6.51494980e-01 -2.29042277e-01 4.95832682e-01 7.06055403e-01 1.86415210e-01 -1.96647629e-01 -1.56781405e-01 7.33285487e-01 1.29144475e-01 -1.07680544e-01 -4.56245355e-02 8.58025134e-01 -5.42494595e-01 -1.27374852e+00 -5.29828548e-01 3.42631578e-01 -5.77151179e-01 7.23645017e-02 -6.34456217e-01 2.17059374e-01 3.50244045e-01 1.02902961e+00 -4.06003334e-02 -1.00292552e+00 2.82525748e-01 -3.50340486e-01 6.98342383e-01 -3.46881628e-01 -5.19877315e-01 3.57545137e-01 -2.88952321e-01 -6.70450687e-01 1.79062605e-01 -5.47014415e-01 -4.77379382e-01 -5.77389121e-01 -3.82021129e-01 -1.09280907e-01 4.91565734e-01 6.38608098e-01 5.89382708e-01 9.97457132e-02 1.31682301e+00 -1.04648709e+00 -5.73769391e-01 -8.51551473e-01 -7.14537263e-01 2.43107900e-01 3.62168670e-01 -9.74030793e-01 -3.29093069e-01 -3.85407150e-01]

[13.023133277893066, 1.1014189720153809]

5a9f235b-99da-4a01-82c2-16df69d54bab

rethinking-cross-entropy-loss-for-stereo

2306.15612

https://arxiv.org/abs/2306.15612v1

https://arxiv.org/pdf/2306.15612v1.pdf

Rethinking Cross-Entropy Loss for Stereo Matching Networks

Despite the great success of deep learning in stereo matching, recovering accurate and clearly-contoured disparity map is still challenging. Currently, L1 loss and cross-entropy loss are the two most widely used loss functions for training the stereo matching networks. Comparing with the former, the latter can usually achieve better results thanks to its direct constraint to the the cost volume. However, how to generate reasonable ground-truth distribution for this loss function remains largely under exploited. Existing works assume uni-modal distributions around the ground-truth for all of the pixels, which ignores the fact that the edge pixels may have multi-modal distributions. In this paper, we first experimentally exhibit the importance of correct edge supervision to the overall disparity accuracy. Then a novel adaptive multi-modal cross-entropy loss which encourages the network to generate different distribution patterns for edge and non-edge pixels is proposed. We further optimize the disparity estimator in the inference stage to alleviate the bleeding and misalignment artifacts at the edge. Our method is generic and can help classic stereo matching models regain competitive performance. GANet trained by our loss ranks 1st on the KITTI 2015 and 2012 benchmarks and outperforms state-of-the-art methods by a large margin. Meanwhile, our method also exhibits superior cross-domain generalization ability and outperforms existing generalization-specialized methods on four popular real-world datasets.

['Xijun Zhao', 'Jingyun Fu', 'Chenyu Qiao', 'Zhiyu Xiang', 'Peng Xu']

2023-06-27

['stereo-matching-1', 'domain-generalization']

['computer-vision', 'methodology']

[ 9.98689383e-02 -1.87134504e-01 -4.26247418e-01 -5.08305430e-01 -9.56075191e-01 -1.30063266e-01 4.17975932e-01 -2.60188341e-01 -4.17846173e-01 9.17980909e-01 2.01911822e-01 -4.14292179e-02 8.64188373e-02 -9.52558756e-01 -8.93879712e-01 -8.09820712e-01 4.01775897e-01 2.76639014e-01 2.38555625e-01 -9.54610705e-02 1.78885087e-01 1.39943540e-01 -1.71873510e+00 2.69270223e-02 1.48886335e+00 1.31146872e+00 1.52789623e-01 6.36087805e-02 -8.19531158e-02 5.70838630e-01 -1.52047589e-01 -6.40275598e-01 7.00478017e-01 -3.49718124e-01 -2.99366474e-01 -1.07000142e-01 1.04611921e+00 -5.48830807e-01 -6.62468135e-01 1.51846063e+00 7.98858225e-01 3.91599312e-02 6.00735545e-01 -1.15419555e+00 -5.04925489e-01 2.14370400e-01 -8.91413808e-01 -2.83028483e-02 6.09889962e-02 3.30568850e-01 1.01488674e+00 -6.19497538e-01 4.95149940e-01 1.11571562e+00 7.94547737e-01 4.61724132e-01 -1.11863673e+00 -1.07434440e+00 4.94297780e-02 3.13879997e-01 -1.43579888e+00 -2.82562971e-01 8.39391649e-01 -4.45199728e-01 3.73716563e-01 -8.38205591e-02 5.16004324e-01 9.31931078e-01 1.64882496e-01 8.23028505e-01 1.15313053e+00 -1.05651394e-01 -4.71507944e-02 -3.42914388e-02 -2.60182440e-01 4.96347010e-01 3.46937299e-01 5.43518424e-01 -4.82312948e-01 1.94511324e-01 9.23470676e-01 -1.59409970e-01 -5.78279495e-01 -6.23109758e-01 -9.49277461e-01 7.27880299e-01 7.63410807e-01 -3.09613682e-02 -2.92621523e-01 6.74291030e-02 3.65897745e-01 -7.48070143e-03 5.93390346e-01 -9.04111541e-04 -1.93071842e-01 -2.79907007e-02 -8.99083734e-01 3.59419078e-01 2.17955947e-01 9.73757088e-01 1.04246426e+00 -9.76257175e-02 -4.17687893e-01 9.59254146e-01 1.84701428e-01 4.37708467e-01 3.35077375e-01 -9.38088655e-01 8.77115190e-01 6.40205860e-01 6.40499890e-02 -1.03442979e+00 -1.01066113e-01 -7.11887419e-01 -1.15738535e+00 3.74393255e-01 6.96720004e-01 -3.48916799e-02 -9.96193767e-01 1.94224250e+00 2.97827870e-01 3.50961685e-01 -8.74128118e-02 1.18628359e+00 8.09640527e-01 4.05301601e-01 -5.42152449e-02 2.20646396e-01 9.40448284e-01 -1.01660502e+00 -5.01206756e-01 -4.67812300e-01 1.75482899e-01 -6.78983390e-01 9.58158851e-01 1.06087528e-01 -1.02725089e+00 -6.51544034e-01 -1.11155009e+00 -3.42944860e-01 4.99318726e-02 -5.67086525e-02 7.11207688e-01 4.04144228e-01 -9.07252729e-01 6.28774166e-01 -4.91864890e-01 3.87957841e-02 7.44806886e-01 1.75429180e-01 -2.28179947e-01 -4.42199498e-01 -1.26316977e+00 7.14983642e-01 3.09119463e-01 1.19607970e-01 -4.03009087e-01 -1.03589153e+00 -1.02716684e+00 -1.09330446e-01 4.02505100e-02 -9.60369587e-01 8.59616280e-01 -1.02270591e+00 -1.38173282e+00 1.14627147e+00 -1.50537610e-01 -4.54056829e-01 1.04345417e+00 -5.19955084e-02 -1.02115810e-01 -1.11426055e-01 2.97287613e-01 1.04722393e+00 5.81691802e-01 -1.12603199e+00 -8.69757473e-01 -4.10239846e-01 5.62605215e-03 3.79901677e-01 -1.63806662e-01 -6.55979455e-01 -6.59300327e-01 -8.21233690e-01 3.18316221e-01 -6.76982760e-01 -1.27613721e-02 3.94987524e-01 -4.30611223e-01 -8.69954471e-03 2.61741757e-01 -5.89610100e-01 1.10812807e+00 -2.22886872e+00 -5.15392013e-02 7.62517676e-02 5.69974631e-02 1.06976002e-01 -2.24170107e-02 -1.33924216e-01 -3.25238565e-03 -3.01868647e-01 -4.71728861e-01 -4.51638758e-01 3.02401073e-02 8.13350528e-02 -3.22620720e-01 5.95977426e-01 -3.26024480e-02 8.06548357e-01 -9.02541339e-01 -5.90509236e-01 4.29067552e-01 6.62418485e-01 -7.26127148e-01 2.02435493e-01 1.20592536e-02 7.18231440e-01 -2.76728541e-01 5.65670252e-01 1.20785725e+00 -1.13901749e-01 -3.25768262e-01 -3.95570636e-01 7.99246356e-02 1.34258747e-01 -1.19049990e+00 1.93598700e+00 -5.41003406e-01 8.20368350e-01 -3.58254462e-02 -8.96693885e-01 1.04057753e+00 -8.92196149e-02 3.52256805e-01 -1.13667238e+00 1.05793782e-01 5.80017626e-01 -7.69812986e-02 -2.38241315e-01 3.54981333e-01 -3.56318533e-01 2.60804623e-01 -1.59960508e-01 -2.92711973e-01 -1.28248826e-01 -1.95770767e-02 -2.84329742e-01 4.02219683e-01 1.47773400e-01 5.35053089e-02 -3.60246360e-01 6.75324380e-01 -3.00068200e-01 1.07640898e+00 4.13929760e-01 -3.38642657e-01 1.15209746e+00 2.04524264e-01 -2.76944280e-01 -1.02916539e+00 -1.07140267e+00 -5.37376463e-01 5.49272180e-01 8.42838228e-01 2.00133026e-01 -6.27627790e-01 -4.47579533e-01 2.43906453e-01 5.40943027e-01 -5.56035280e-01 -1.72839269e-01 -4.74523276e-01 -6.67860150e-01 3.64969730e-01 6.27678394e-01 1.21692324e+00 -6.98360026e-01 -1.86358348e-01 -8.34631175e-03 -5.28084993e-01 -1.17441308e+00 -8.27927768e-01 -2.37567991e-01 -9.20666933e-01 -1.01917565e+00 -1.05000651e+00 -8.85598242e-01 7.63134122e-01 1.09528877e-01 1.13703012e+00 -8.69116262e-02 -1.41519651e-01 -2.87180394e-01 1.36313275e-01 -2.14140758e-01 -1.72276609e-02 9.86139178e-02 -4.59164053e-01 1.18599415e-01 4.79730874e-01 -6.83940709e-01 -1.11853373e+00 5.15742421e-01 -8.15202713e-01 3.91891301e-01 3.88582289e-01 9.91720140e-01 7.53892660e-01 -3.48379947e-02 2.93310910e-01 -6.68466806e-01 7.87632987e-02 -2.69353837e-01 -7.83895254e-01 5.25562689e-02 -6.55076325e-01 1.48040578e-01 5.14518559e-01 -2.32350543e-01 -1.07838428e+00 -6.59880638e-02 -2.21560851e-01 -4.94520873e-01 4.66781482e-02 1.63671281e-02 -3.60152245e-01 -1.82173833e-01 4.43454474e-01 2.68243223e-01 -4.72192951e-02 -2.70684242e-01 -3.53030860e-02 4.20098007e-01 7.33122468e-01 -6.65278137e-01 7.42236257e-01 7.90181100e-01 2.20971033e-01 -4.23046350e-01 -9.83624756e-01 -4.04098779e-01 -5.36609888e-02 -1.33014694e-01 6.78437591e-01 -1.26690078e+00 -5.51113605e-01 9.79159832e-01 -9.45676148e-01 -3.51818204e-01 -2.69915998e-01 4.84056234e-01 -4.55874562e-01 4.73256677e-01 -4.85213220e-01 -5.01563907e-01 -3.45472813e-01 -1.22062981e+00 1.23021019e+00 5.23322761e-01 3.13544601e-01 -1.01823425e+00 1.54267978e-02 4.86706465e-01 6.22688293e-01 3.33537459e-01 6.92211866e-01 1.24438792e-01 -8.40825081e-01 2.65363399e-02 -6.69610739e-01 5.60194790e-01 8.93952549e-02 -2.71176219e-01 -1.23693669e+00 -2.50216335e-01 -1.78235933e-01 -2.87812024e-01 1.24526548e+00 7.74672925e-01 1.31745183e+00 1.49673522e-01 -2.62605488e-01 1.30511725e+00 1.60496640e+00 4.35860902e-02 1.06403840e+00 6.11270845e-01 7.18127191e-01 8.21807861e-01 6.38345659e-01 2.58434743e-01 6.52438581e-01 7.82199502e-01 6.26021147e-01 -3.06867182e-01 -3.57546896e-01 -5.38457453e-01 1.09494897e-02 2.99889445e-01 6.55816346e-02 -1.14593014e-01 -6.81916654e-01 5.81795812e-01 -1.82657516e+00 -8.37380946e-01 -7.82632828e-02 2.50884151e+00 9.51300383e-01 2.97676414e-01 -3.29321995e-02 -9.16651264e-02 9.91664231e-01 3.16383332e-01 -8.78669322e-01 1.86341032e-01 -6.20689452e-01 1.15892217e-01 7.61978447e-01 7.35921800e-01 -1.15599072e+00 6.97013080e-01 5.06328487e+00 1.24647129e+00 -1.30141282e+00 -1.74615651e-01 1.10217977e+00 9.62338671e-02 -5.57633817e-01 -6.52329251e-02 -8.60994935e-01 8.18949044e-01 8.54315683e-02 -1.45984098e-01 2.77934939e-01 7.63308227e-01 1.48122748e-02 -2.25721627e-01 -1.06892276e+00 1.46963203e+00 -5.02648167e-02 -1.34751725e+00 -4.86362278e-02 7.05119148e-02 1.17532146e+00 1.43770844e-01 1.67798847e-01 2.05768675e-01 -4.37612794e-02 -1.12226057e+00 7.37020552e-01 4.60874408e-01 1.02189577e+00 -6.68737710e-01 8.58503103e-01 2.36369118e-01 -1.24289823e+00 1.06238037e-01 -5.13286173e-01 1.76873505e-02 2.23540828e-01 8.74713957e-01 -4.28386517e-02 6.22863412e-01 6.33862913e-01 9.55338180e-01 -3.35647851e-01 1.46747291e+00 -2.24369586e-01 7.77768120e-02 -3.75472128e-01 3.26478511e-01 1.59216255e-01 -6.02086604e-01 4.22921360e-01 9.88909841e-01 3.65991443e-01 -3.14110041e-01 -9.83088985e-02 9.17890489e-01 -2.65279770e-01 7.36917893e-04 -4.39378977e-01 6.79992914e-01 4.79164124e-01 7.97932267e-01 -2.32122600e-01 -2.19635755e-01 -4.21213925e-01 8.06927800e-01 2.91052133e-01 4.37712461e-01 -8.36342871e-01 -3.38303268e-01 9.96389329e-01 3.74122709e-01 2.44786650e-01 2.27000073e-01 -6.75962150e-01 -1.30155981e+00 4.93201256e-01 -7.09565997e-01 4.02196348e-01 -6.89731479e-01 -1.57420647e+00 5.24940908e-01 -1.67053297e-01 -1.60310721e+00 -4.80389819e-02 -4.96275574e-01 -7.01140463e-01 1.12115836e+00 -2.18131447e+00 -9.05775726e-01 -7.40751565e-01 5.88042080e-01 3.48361224e-01 -5.46068549e-02 1.70588717e-01 8.73312652e-01 -6.06942832e-01 8.87744963e-01 1.46264598e-01 2.12998912e-01 9.15158570e-01 -1.09806740e+00 2.88891971e-01 8.28007877e-01 -4.17297542e-01 2.45756000e-01 6.37489736e-01 -3.55940223e-01 -8.32496166e-01 -1.10400319e+00 7.46942580e-01 5.63098826e-02 3.65672708e-01 -1.27707019e-01 -9.61242259e-01 3.92974794e-01 6.84100613e-02 1.72107056e-01 -4.40688469e-02 -3.16458523e-01 -4.03774500e-01 -6.01698935e-01 -1.20529616e+00 6.65725589e-01 1.29760933e+00 -4.38755691e-01 -2.56409466e-01 3.86050791e-02 3.62602293e-01 -8.76912892e-01 -5.79136550e-01 8.32082570e-01 5.18318653e-01 -1.55786049e+00 1.05077183e+00 -8.35625306e-02 8.04918706e-01 -3.25653255e-01 -1.58184141e-01 -1.16553855e+00 2.21691281e-02 -4.81412232e-01 3.39086264e-01 1.29889500e+00 2.20710129e-01 -9.46015179e-01 1.09121299e+00 5.61481893e-01 -1.72920913e-01 -8.56026769e-01 -1.15246153e+00 -7.44305789e-01 3.59168530e-01 -4.71054792e-01 6.52713537e-01 8.54057491e-01 -4.48187798e-01 9.33096036e-02 -4.37506706e-01 -2.07200050e-02 1.12565088e+00 4.17800903e-01 8.68714809e-01 -1.10017943e+00 -2.73317665e-01 -8.15862298e-01 -6.69932544e-01 -1.69934392e+00 2.03971162e-01 -8.53344440e-01 9.81922671e-02 -1.41038513e+00 2.94273198e-01 -6.99814677e-01 -5.84464930e-02 4.92585748e-02 -4.12631005e-01 3.19691002e-01 -1.90864459e-01 1.24634765e-01 -9.18135941e-02 9.34386313e-01 1.62327635e+00 -2.20130846e-01 -3.84127488e-03 9.63466838e-02 -6.26008093e-01 6.77919626e-01 6.57131732e-01 -2.08069056e-01 -4.17585433e-01 -6.14369512e-01 1.66675046e-01 -1.37737066e-01 4.59601194e-01 -1.01996148e+00 2.85111696e-01 -9.75259542e-02 2.17967182e-01 -5.69847465e-01 2.24117145e-01 -7.89070368e-01 9.61431265e-02 1.98257297e-01 -1.23647943e-01 -3.24144483e-01 2.06019014e-01 4.53903496e-01 -6.25073493e-01 1.39844567e-01 1.24110961e+00 2.05796510e-01 -7.03511715e-01 7.12355554e-01 5.08092940e-01 6.81353450e-01 8.26875329e-01 -5.25150716e-01 -3.78346115e-01 -4.46495831e-01 -2.27833241e-02 4.69227910e-01 7.65697837e-01 3.59927654e-01 4.46012557e-01 -1.58146811e+00 -8.54083717e-01 3.39428842e-01 2.53447026e-01 5.31098545e-01 5.30804276e-01 8.12425971e-01 -7.00044155e-01 2.09267035e-01 -3.01304132e-01 -8.21875751e-01 -7.76320398e-01 1.75026134e-01 8.49028885e-01 -1.51587352e-01 -6.73805177e-01 9.24490392e-01 7.43911743e-01 -3.06366831e-01 5.46804011e-01 -2.71438897e-01 1.10769831e-01 -1.75932959e-01 2.58429676e-01 3.84942770e-01 -1.95035972e-02 -5.97350597e-01 -2.54522771e-01 1.04852831e+00 1.53299496e-01 2.41848350e-01 1.06776214e+00 -1.64605185e-01 1.27885178e-01 1.25808820e-01 1.41657639e+00 -1.42478853e-01 -1.78278685e+00 -4.14872706e-01 -4.25239742e-01 -9.46979165e-01 2.11729452e-01 -4.55758333e-01 -1.62962878e+00 1.01624513e+00 8.10000777e-01 -3.43229204e-01 1.26897013e+00 -2.97911316e-01 1.13818693e+00 -2.17975080e-01 3.54092181e-01 -1.13208604e+00 -2.02767372e-01 2.74467766e-01 6.80758297e-01 -1.72512054e+00 -8.72255415e-02 -6.03869021e-01 -5.67317784e-01 9.03133214e-01 9.34020102e-01 -2.54020512e-01 6.24220550e-01 8.77361670e-02 1.70887157e-01 9.36411917e-02 -1.41629994e-01 -1.66935489e-01 5.34698963e-01 4.89435911e-01 3.63828391e-01 -8.24812427e-02 -3.50058347e-01 1.05292276e-01 -2.20237479e-01 7.78928446e-03 -1.74160041e-02 3.48937184e-01 -1.69987783e-01 -9.66406167e-01 -1.68690354e-01 3.80249053e-01 -3.16912562e-01 -2.26823971e-01 1.59896433e-01 8.18033278e-01 2.97307789e-01 6.03360355e-01 2.80855834e-01 -2.10038275e-01 5.26991367e-01 -4.84189898e-01 4.57296371e-01 -1.60753012e-01 -1.83336139e-01 -7.93139338e-02 -1.65927842e-01 -6.59506798e-01 -4.71940309e-01 -5.45177519e-01 -1.01919746e+00 -5.83156705e-01 -2.18096867e-01 -2.79940069e-01 2.98875391e-01 6.63423002e-01 3.73418182e-01 1.77478626e-01 7.23019481e-01 -7.73124278e-01 -5.06290555e-01 -6.97660804e-01 -6.20840788e-01 7.74701655e-01 4.23416078e-01 -8.44316363e-01 -7.00238287e-01 -2.83505589e-01]

[8.879772186279297, -2.2902231216430664]

8205c95c-0339-4449-ab50-e3579afd58e0

back-to-reality-leveraging-pattern-driven

2110.08604

https://arxiv.org/abs/2110.08604v3

https://arxiv.org/pdf/2110.08604v3.pdf

Improving Implicit Sentiment Learning via Local Sentiment Aggregation

Aspect-based sentiment classification (ABSC) has revealed the potential dependency of sentiment polarities among different aspects. Our study further explores this phenomenon, positing that adjacent aspects often exhibit similar sentiments, a concept we term "aspect sentiment coherency." We argue that the current research landscape has not fully appreciated the significance of modeling aspect sentiment coherency. To address this gap, we introduce a local sentiment aggregation paradigm (LSA) that facilitates fine-grained sentiment coherency modeling. This approach enables the extraction of implicit sentiments for aspects lacking explicit sentiment descriptions. Leveraging gradient descent, we design a differential-weighted sentiment aggregation window that guides the modeling of aspect sentiment coherency. Experimental results affirm the efficacy of LSA in learning sentiment coherency, as it achieves state-of-the-art performance across three public datasets, thus significantly enhancing existing ABSC models. We have made our code available, providing a ready tool for existing methods to harness the potential of sentiment coherency information.

['Ke Li', 'Heng Yang']

2021-10-16

['sentiment-dependency-learning']

['natural-language-processing']

[-1.51161075e-01 -5.94823472e-02 -4.97588009e-01 -6.63262844e-01 -7.12627113e-01 -6.67105556e-01 8.45211327e-01 4.89930838e-01 -1.17148072e-01 3.30672085e-01 8.90338421e-01 -3.72267306e-01 8.44581202e-02 -8.40386331e-01 -2.35146105e-01 -5.18554211e-01 1.57260805e-01 -9.41932499e-02 -3.17062050e-01 -6.12077475e-01 5.86776972e-01 -1.72096401e-01 -1.63109052e+00 6.84131026e-01 5.96010745e-01 9.76833880e-01 -3.25487375e-01 2.75979251e-01 -4.42518771e-01 8.52308095e-01 -5.76053977e-01 -7.58255541e-01 5.58662042e-02 -2.40468845e-01 -6.65531695e-01 7.38043711e-02 1.76982880e-01 2.03556880e-01 5.72533131e-01 8.02417755e-01 6.33452386e-02 -1.97651997e-01 5.72355449e-01 -1.21141303e+00 -9.49410856e-01 7.71327198e-01 -7.02512562e-01 1.38206035e-01 5.06520510e-01 -5.28104976e-02 1.98323834e+00 -8.89373779e-01 5.48827767e-01 8.21961045e-01 1.03533077e+00 1.51507422e-01 -1.11286747e+00 -4.99490201e-01 7.25018322e-01 -1.61372647e-01 -8.99421930e-01 4.10245694e-02 1.01165807e+00 -4.85498279e-01 1.39763737e+00 3.89611214e-01 1.29886174e+00 8.00127864e-01 5.60887218e-01 1.14163744e+00 1.60877597e+00 -2.97499955e-01 2.57287860e-01 4.57309961e-01 6.83810592e-01 1.21746615e-01 4.69691187e-01 -3.88148487e-01 -9.50369716e-01 -1.72902644e-01 -2.31175601e-01 5.13352379e-02 1.70127526e-01 -2.60589957e-01 -8.86585712e-01 9.12174225e-01 3.30394894e-01 3.98871839e-01 -4.28943992e-01 -3.38122636e-01 4.31966871e-01 4.92678821e-01 1.24311781e+00 8.17965269e-01 -9.40554500e-01 -2.73726732e-01 -7.85071850e-01 3.95823568e-01 1.04215825e+00 6.60705268e-01 9.72798944e-01 -9.03734788e-02 2.29610782e-02 8.34115982e-01 5.32561779e-01 3.42724770e-01 5.10880589e-01 -4.21374917e-01 1.40258536e-01 1.31918180e+00 -6.51474223e-02 -1.11524534e+00 -4.76775080e-01 -5.03485620e-01 -3.53870660e-01 1.17050759e-01 -3.29503357e-01 -2.02396810e-01 -4.47123945e-01 1.47701085e+00 3.57940525e-01 -1.66545525e-01 1.06275082e-01 5.70652068e-01 8.98606896e-01 9.93614569e-02 2.64085174e-01 7.13538229e-02 1.73836672e+00 -8.97083759e-01 -5.99378228e-01 -4.34945077e-01 6.02424026e-01 -9.98432636e-01 1.45486593e+00 3.05712581e-01 -7.96267271e-01 -1.05778694e-01 -1.19990778e+00 5.03413901e-02 -9.64116991e-01 -3.16716433e-01 1.51727748e+00 9.18302059e-01 -1.13467932e+00 2.28282198e-01 -7.58234382e-01 -3.33413035e-01 4.58927065e-01 7.84339979e-02 -2.34832644e-01 3.80776763e-01 -9.83740568e-01 7.14535058e-01 -2.02436909e-01 -1.19670540e-01 -2.24003568e-01 -1.14734662e+00 -1.03795385e+00 -1.38770252e-01 1.18494578e-01 -8.61300409e-01 1.19725621e+00 -1.17286885e+00 -9.98997569e-01 8.46220553e-01 -3.43314439e-01 -2.26519257e-02 -1.61844179e-01 -4.94513541e-01 -6.10290885e-01 -3.09477329e-01 4.15244818e-01 3.04812193e-01 5.72354317e-01 -1.30144298e+00 -8.08462143e-01 -3.98271322e-01 4.52731818e-01 4.52702969e-01 -9.26479876e-01 2.27203593e-01 -2.02973709e-01 -7.55041182e-01 -9.88901928e-02 -8.69861305e-01 -2.62571245e-01 -6.17814481e-01 -3.31622809e-01 -2.62473792e-01 6.94037318e-01 1.58889458e-01 1.46465516e+00 -1.92148626e+00 -3.91468048e-01 1.28450304e-01 2.31986091e-01 -2.15647921e-01 -2.09901348e-01 6.84176922e-01 -1.45949861e-02 2.09603727e-01 -1.65725559e-01 -6.87989175e-01 1.38687462e-01 -3.30167934e-02 -5.89711487e-01 1.25831231e-01 4.82151598e-01 1.14129603e+00 -8.44438612e-01 -8.72887969e-02 -1.39843762e-01 5.91543674e-01 -9.12423730e-01 5.90404309e-02 -2.41280034e-01 1.15356423e-01 -6.99246824e-01 1.09124744e+00 6.72848284e-01 -5.29818177e-01 2.81167120e-01 -3.05755973e-01 -4.45942193e-01 9.03221190e-01 -8.32258463e-01 1.17530966e+00 -6.21837020e-01 5.50986886e-01 -2.82023668e-01 -6.02952480e-01 8.32104623e-01 3.05862855e-02 7.37005591e-01 -6.57893240e-01 9.90187675e-02 -4.75360192e-02 -4.12491649e-01 -9.39082131e-02 1.18431973e+00 -3.90149534e-01 -4.65108633e-01 1.02853465e+00 -1.43642515e-01 -6.60030544e-01 3.35287690e-01 2.67182052e-01 8.19734335e-01 1.91912353e-01 5.30653238e-01 -6.00481629e-01 4.19801682e-01 2.72669822e-01 2.62552917e-01 3.19526494e-01 -4.41248305e-02 5.27414620e-01 8.05120230e-01 -5.83842099e-01 -6.04538202e-01 -8.14376235e-01 -2.37894550e-01 1.33913767e+00 1.63745005e-02 -1.00775874e+00 -3.47838312e-01 -7.99635470e-01 2.08622858e-01 6.17406726e-01 -1.14461982e+00 1.09661713e-01 -1.24644667e-01 -1.38415265e+00 -5.72589273e-03 6.34487391e-01 7.14867041e-02 -1.10192168e+00 -2.96785325e-01 -9.04503912e-02 1.23644830e-03 -9.86009002e-01 -2.41438821e-01 3.79575580e-01 -7.64338255e-01 -1.00346863e+00 -8.61724839e-02 -4.10359263e-01 7.16771543e-01 7.49096811e-01 1.87044251e+00 8.53584260e-02 1.26581728e-01 6.34208739e-01 -5.74151993e-01 -5.82646132e-01 8.35599154e-02 4.30187315e-01 -7.07882568e-02 -1.49630815e-01 1.23850977e+00 -5.71203649e-01 -8.20517778e-01 5.40121496e-02 -1.04591572e+00 -1.66694596e-01 4.58776474e-01 4.68349725e-01 6.25556827e-01 -2.20000684e-01 6.58237815e-01 -1.47430098e+00 1.20344019e+00 -8.55734229e-01 -2.34307408e-01 -2.48251706e-01 -1.39629376e+00 -2.72501290e-01 3.41538072e-01 2.54507214e-02 -1.02621150e+00 -6.19730115e-01 -1.50919929e-01 5.66516936e-01 9.07968506e-02 1.00966680e+00 3.08149636e-01 8.14594105e-02 1.79775774e-01 5.63673079e-02 -2.96477467e-01 -2.25627705e-01 5.49198151e-01 6.09716177e-01 -2.06534088e-01 -5.06399214e-01 4.83856112e-01 7.87612379e-01 -4.22448814e-01 -4.75598156e-01 -1.52196598e+00 -8.14259410e-01 -3.27837795e-01 -7.03821406e-02 5.26502490e-01 -1.33321202e+00 -5.03006518e-01 4.45025742e-01 -5.71543515e-01 4.90389951e-02 -4.97151256e-01 1.73689455e-01 -2.77021229e-01 1.00536525e-01 -5.12848496e-01 -4.70377356e-01 -6.44110501e-01 -9.02924776e-01 1.26075411e+00 2.52043128e-01 -9.78802264e-01 -1.49170125e+00 6.79291368e-01 6.72649860e-01 6.70866489e-01 1.05524518e-01 7.65662968e-01 -6.69520617e-01 -4.00820494e-01 -3.73711467e-01 3.19710523e-02 2.91403174e-01 5.66659570e-01 4.63202670e-02 -1.14014781e+00 3.15736351e-03 5.22775464e-02 -3.50264281e-01 8.65513325e-01 3.72898817e-01 4.73710716e-01 -1.53179124e-01 -4.30022702e-02 4.25650090e-01 1.55214536e+00 -4.05283749e-01 2.66225904e-01 1.07058477e+00 6.53420568e-01 7.72058427e-01 8.56380165e-01 7.02995479e-01 1.15466809e+00 3.10320139e-01 2.14370772e-01 -2.34627575e-01 8.13803896e-02 -1.22590058e-01 4.41117018e-01 1.16522288e+00 1.26217246e-01 9.84116569e-02 -7.22340524e-01 8.31404269e-01 -1.57888484e+00 -5.68055749e-01 -3.42421502e-01 1.52595735e+00 8.89503777e-01 2.21503481e-01 -1.45414164e-02 -8.01773444e-02 -5.39353378e-02 8.58137608e-01 -3.48098040e-01 -9.01561558e-01 -3.57823074e-01 8.24156180e-02 -3.11054084e-02 5.54508150e-01 -1.25212693e+00 1.13809443e+00 6.93887377e+00 2.85571903e-01 -9.57289100e-01 1.53520666e-02 8.11529934e-01 -1.30663782e-01 -1.26849949e+00 1.75180286e-01 -8.93974841e-01 2.41440088e-01 5.92339635e-01 -1.97002530e-01 -1.20365784e-01 1.11672068e+00 2.77853727e-01 -2.98486471e-01 -7.32676387e-01 2.10017085e-01 3.46449256e-01 -1.13977873e+00 3.10404450e-01 -1.20575607e-01 1.29777348e+00 1.54098019e-01 4.05467123e-01 3.51499051e-01 6.37250185e-01 -7.05157340e-01 3.82104039e-01 3.76191795e-01 1.21728286e-01 -7.79952705e-01 9.43496644e-01 -4.45276409e-01 -1.44622684e+00 1.17652304e-01 -6.88266158e-02 -3.68716896e-01 7.88723752e-02 9.64783490e-01 -6.27333999e-01 5.02014160e-01 9.84335959e-01 1.30029631e+00 -6.95671678e-01 2.99991459e-01 -3.31361324e-01 5.39020717e-01 1.57752767e-01 -1.89180791e-01 4.22509372e-01 -4.00004268e-01 3.74459296e-01 1.30430508e+00 5.62796593e-02 -4.13120687e-01 -1.16837434e-01 4.83885497e-01 1.24902315e-01 4.77731615e-01 -6.62311077e-01 -2.52241015e-01 1.01379462e-01 1.68452084e+00 -8.39483738e-01 -2.37502590e-01 -1.01358354e+00 5.64154506e-01 1.90254495e-01 9.70779806e-02 -2.43658155e-01 -4.66828458e-02 1.31013429e+00 -1.72832608e-01 5.39132237e-01 -5.49982972e-02 -1.13826776e+00 -1.36858046e+00 1.53220683e-01 -1.09287846e+00 3.83733600e-01 -6.99069142e-01 -1.50895929e+00 8.03826213e-01 -2.54101872e-01 -1.39784145e+00 -2.13530540e-01 -6.72359228e-01 -7.74196386e-01 6.47506952e-01 -1.99337828e+00 -1.59334075e+00 -2.02224538e-01 4.11017448e-01 5.37442684e-01 -2.14545742e-01 9.47466731e-01 -1.20894261e-01 -2.21764103e-01 4.53187168e-01 -3.26529920e-01 -2.58095413e-01 7.64305294e-01 -1.51813543e+00 6.55672789e-01 7.20860839e-01 1.77243799e-01 1.19005954e+00 8.13680172e-01 -6.59701228e-01 -1.40472519e+00 -7.76341081e-01 1.43449605e+00 -1.17989826e+00 1.34979296e+00 -2.95591235e-01 -6.10636234e-01 6.63656592e-01 5.82230747e-01 -4.10427332e-01 1.78214657e+00 9.60457861e-01 -7.39658535e-01 -2.60829777e-01 -7.48674154e-01 8.54745507e-01 5.43107510e-01 -6.96335912e-01 -7.50345230e-01 -6.83320016e-02 7.41646886e-01 1.77681014e-01 -9.24665511e-01 4.68241811e-01 8.78621340e-01 -1.26038909e+00 7.28923082e-01 -4.38132435e-01 9.16186392e-01 -3.13973427e-01 -3.92388582e-01 -1.46747410e+00 -1.27604261e-01 -2.98055083e-01 -1.17979489e-01 1.43333650e+00 8.07196200e-01 -7.48831928e-01 8.67496669e-01 6.52368784e-01 -5.92933148e-02 -1.13374317e+00 -2.76159048e-01 -1.43671215e-01 2.21067145e-01 -8.77910078e-01 9.62359250e-01 1.09058082e+00 4.63206530e-01 4.06546474e-01 1.21162916e-02 -8.72652233e-02 7.45381713e-02 6.80014849e-01 6.91757798e-01 -9.54138517e-01 -2.13248178e-01 -7.96020508e-01 -1.79568335e-01 -5.97427428e-01 2.80253857e-01 -7.23593831e-01 -3.87803435e-01 -1.47349668e+00 5.49756765e-01 -3.87454301e-01 -7.44989038e-01 5.28532743e-01 -7.16616571e-01 7.99422324e-01 -1.81651078e-02 1.46931738e-01 -9.08487022e-01 5.12555480e-01 1.17888045e+00 -1.76723704e-01 -1.04829289e-01 1.24583267e-01 -1.88631392e+00 9.34232771e-01 8.18969190e-01 -3.95289034e-01 -5.65654576e-01 -2.88071871e-01 1.21336627e+00 -1.00069809e+00 -2.67845720e-01 -4.84469175e-01 5.60224056e-02 -1.70161858e-01 1.33646637e-01 -6.95448756e-01 2.39884794e-01 -8.02149355e-01 -2.69888133e-01 -6.91906959e-02 -4.05566514e-01 4.15769219e-01 2.31933370e-01 3.07582855e-01 -5.32110214e-01 1.89511731e-01 -3.81864458e-02 7.47558707e-03 -7.48973012e-01 1.05436437e-01 -5.13086915e-01 1.76984742e-01 6.51223421e-01 -7.45682716e-02 -3.79838705e-01 -3.86081815e-01 -1.10825174e-01 1.39230072e-01 5.99639595e-01 7.54711151e-01 4.22092348e-01 -1.19613695e+00 -5.32819211e-01 4.23886567e-01 6.57365561e-01 -5.05438745e-01 5.29207997e-02 8.10046554e-01 9.11232159e-02 3.43918085e-01 1.28989130e-01 -2.63792157e-01 -1.31924593e+00 2.04274744e-01 -1.10228494e-01 -5.63943148e-01 -3.47107381e-01 7.60143101e-01 1.63490281e-01 -7.66870558e-01 -5.64453065e-01 -2.30774745e-01 -7.37810075e-01 7.39831924e-01 5.61303020e-01 -1.64137722e-03 2.67062187e-01 -6.09883785e-01 -4.64530379e-01 7.87973940e-01 -4.56050724e-01 -4.87921238e-02 1.55493689e+00 -3.69481891e-01 -5.85878372e-01 7.11410999e-01 1.12097371e+00 6.10069871e-01 -9.77220476e-01 2.57542846e-03 1.96384847e-01 -3.65080327e-01 -8.30293372e-02 -9.08412755e-01 -9.55374897e-01 2.91822106e-01 9.79988603e-04 7.87267447e-01 1.17872882e+00 1.40347555e-01 5.72355449e-01 2.40833640e-01 2.04129294e-01 -1.11721539e+00 1.07746199e-01 9.55151200e-01 5.12760758e-01 -1.47759783e+00 4.64753807e-01 -2.50457883e-01 -1.18688357e+00 6.67039514e-01 5.54925859e-01 -2.66604424e-01 1.01757455e+00 7.12036729e-01 8.06842983e-01 -5.54669201e-01 -1.18221140e+00 -3.41937304e-01 4.69471365e-01 3.87286305e-01 1.12986302e+00 3.11608553e-01 -5.87921321e-01 8.59425664e-01 -8.51996779e-01 -4.10945505e-01 4.32481617e-01 1.04950345e+00 -9.33396518e-02 -1.32776046e+00 1.10613860e-01 5.50695717e-01 -9.52463508e-01 -6.90102220e-01 -7.58284152e-01 5.93990803e-01 -1.51757717e-01 1.15261579e+00 1.30822361e-01 -4.51453090e-01 2.58703709e-01 -4.60041910e-02 -2.43874818e-01 -7.50099003e-01 -1.25817275e+00 -2.31266301e-02 3.54170710e-01 -6.56620741e-01 -7.69854605e-01 -7.24666595e-01 -6.83708549e-01 -2.76257008e-01 -1.32040888e-01 2.57323831e-01 9.29184973e-01 1.09364617e+00 6.15514159e-01 6.57641768e-01 9.03872669e-01 -4.14359093e-01 5.24412692e-02 -9.16189909e-01 -5.10924518e-01 4.65883553e-01 4.47297543e-01 -2.92175949e-01 -5.74541390e-01 -6.99165091e-02]

[11.472517013549805, 6.725875377655029]

8afeedb6-48ae-43d2-83ac-5c7bcafb3492

ef-bv-a-unified-theory-of-error-feedback-and

2205.04180

https://arxiv.org/abs/2205.04180v4

https://arxiv.org/pdf/2205.04180v4.pdf

EF-BV: A Unified Theory of Error Feedback and Variance Reduction Mechanisms for Biased and Unbiased Compression in Distributed Optimization

In distributed or federated optimization and learning, communication between the different computing units is often the bottleneck and gradient compression is widely used to reduce the number of bits sent within each communication round of iterative methods. There are two classes of compression operators and separate algorithms making use of them. In the case of unbiased random compressors with bounded variance (e.g., rand-k), the DIANA algorithm of Mishchenko et al. (2019), which implements a variance reduction technique for handling the variance introduced by compression, is the current state of the art. In the case of biased and contractive compressors (e.g., top-k), the EF21 algorithm of Richt\'arik et al. (2021), which instead implements an error-feedback mechanism, is the current state of the art. These two classes of compression schemes and algorithms are distinct, with different analyses and proof techniques. In this paper, we unify them into a single framework and propose a new algorithm, recovering DIANA and EF21 as particular cases. Our general approach works with a new, larger class of compressors, which has two parameters, the bias and the variance, and includes unbiased and biased compressors as particular cases. This allows us to inherit the best of the two worlds: like EF21 and unlike DIANA, biased compressors, like top-k, whose good performance in practice is recognized, can be used. And like DIANA and unlike EF21, independent randomness at the compressors allows to mitigate the effects of compression, with the convergence rate improving when the number of parallel workers is large. This is the first time that an algorithm with all these features is proposed. We prove its linear convergence under certain conditions. Our approach takes a step towards better understanding of two so-far distinct worlds of communication-efficient distributed learning.

['Peter Richtárik', 'Kai Yi', 'Laurent Condat']

2022-05-09

['distributed-optimization']

['methodology']

[ 7.11763501e-02 3.12127713e-02 -2.96769500e-01 1.10682569e-01 -7.57625401e-01 -5.12932420e-01 4.99145627e-01 2.83590555e-01 -6.76832676e-01 9.60713923e-01 4.35374267e-02 -3.39399517e-01 -4.59881365e-01 -9.37147081e-01 -8.98630083e-01 -1.13569629e+00 -3.78538996e-01 6.43677831e-01 1.50857985e-01 -2.26297930e-01 2.98864156e-01 3.01565170e-01 -1.59116316e+00 2.70767161e-03 6.98793530e-01 1.02806187e+00 1.33767858e-01 8.61770451e-01 -1.29187360e-01 6.12646699e-01 -4.00696814e-01 -6.19197845e-01 6.55523658e-01 -5.58974445e-01 -1.15782118e+00 -2.63400018e-01 7.68008530e-02 -2.72047162e-01 -2.01884910e-01 1.08101487e+00 7.13960111e-01 1.35271717e-02 2.69434214e-01 -9.27402139e-01 -6.24107383e-02 1.29370785e+00 -4.53735501e-01 -8.02927930e-03 -2.37117428e-02 -8.49937499e-02 9.59081054e-01 -3.77413541e-01 6.32902682e-01 8.83579671e-01 7.68144846e-01 5.02076864e-01 -1.29449284e+00 -4.96491194e-01 -1.50875598e-01 2.83137530e-01 -1.41713119e+00 -3.21662873e-01 5.64453423e-01 -1.18876763e-01 7.62266636e-01 5.41611910e-01 6.52331889e-01 5.60117245e-01 1.38433799e-01 8.94919276e-01 1.25119328e+00 -6.71804726e-01 6.23699367e-01 1.48911864e-01 -1.25432566e-01 6.44392014e-01 6.22017384e-01 3.92377645e-01 -7.48916209e-01 -3.95421535e-01 2.21679494e-01 1.02708250e-01 -4.99290854e-01 -6.37810588e-01 -1.02790558e+00 9.30149019e-01 2.52736598e-01 6.27411664e-01 -1.91170022e-01 3.11161250e-01 6.98595166e-01 6.65309787e-01 4.56332117e-01 3.50089729e-01 -3.99014860e-01 -4.22703087e-01 -1.16968155e+00 4.68419760e-01 1.27429855e+00 7.93480754e-01 7.88041234e-01 -2.11014897e-01 7.14153517e-03 5.43118119e-01 -7.15725422e-02 3.00462544e-01 7.41540968e-01 -1.02438164e+00 4.08280581e-01 1.58788040e-01 -1.39029890e-01 -7.94367909e-01 -2.11641654e-01 -5.22173762e-01 -1.08065677e+00 2.50268400e-01 5.66056252e-01 -1.91323787e-01 -2.73004562e-01 1.76309657e+00 2.81215280e-01 -6.10680953e-02 1.93723664e-01 8.26807141e-01 5.90747036e-02 5.01290202e-01 -4.89546090e-01 -6.56276941e-01 1.05441105e+00 -8.99332702e-01 -5.49828529e-01 3.39607060e-01 8.90855193e-01 -7.07367957e-01 6.89178228e-01 7.39151776e-01 -1.50992835e+00 -3.01015265e-02 -1.02346730e+00 6.71906322e-02 -3.57171297e-01 -3.77603441e-01 8.02669823e-01 8.25851381e-01 -1.19243503e+00 1.08427918e+00 -9.10184920e-01 -1.04284771e-01 3.24217260e-01 4.54682767e-01 -5.84675558e-02 -1.54626504e-01 -9.15857375e-01 6.59141064e-01 3.48130554e-01 -1.78359970e-01 -8.41328800e-01 -4.76213038e-01 -1.91379771e-01 2.04094350e-01 4.93598551e-01 -7.73327827e-01 1.22895002e+00 -1.00268757e+00 -1.79893291e+00 5.78065574e-01 6.78150579e-02 -8.34986567e-01 9.68746781e-01 -5.98952025e-02 3.76453936e-01 1.46794021e-01 -4.17674422e-01 7.06891268e-02 7.09091485e-01 -1.12517178e+00 -5.43741882e-01 -4.76837009e-01 4.26475592e-02 1.55847147e-01 -5.56362033e-01 -1.50952831e-01 -1.42506525e-01 -5.35007179e-01 -5.29857762e-02 -9.72389698e-01 -2.99368382e-01 -1.72093287e-01 -1.15631707e-01 -1.88017517e-01 4.80916977e-01 -1.93030491e-01 1.30461669e+00 -1.87929106e+00 7.28849530e-01 3.58329982e-01 4.04466689e-01 2.39598006e-01 1.75667584e-01 6.93635702e-01 3.75220120e-01 1.62585288e-01 -5.62552035e-01 -5.81099510e-01 1.19998150e-01 4.46145922e-01 -2.08507866e-01 6.86469197e-01 -5.57279170e-01 5.92936993e-01 -9.01772320e-01 -4.36946273e-01 -2.73796469e-01 4.57002014e-01 -7.48370469e-01 1.91015095e-01 -1.70245558e-01 9.43701118e-02 -3.22899759e-01 2.39655435e-01 5.12431204e-01 -1.62404150e-01 3.97129893e-01 1.91079795e-01 -2.23525941e-01 1.50147080e-01 -1.57958853e+00 1.66269171e+00 -6.15338922e-01 3.68098050e-01 6.28639281e-01 -1.28574538e+00 6.90933049e-01 4.91024077e-01 5.22560239e-01 -2.65397251e-01 1.22470528e-01 7.79887497e-01 -1.13466479e-01 -2.67017841e-01 6.14355385e-01 -2.05495134e-01 1.23978913e-01 8.40763628e-01 -1.33714393e-01 -3.35772932e-01 3.08642179e-01 4.03944612e-01 1.29453671e+00 -1.50092155e-01 2.78200537e-01 -3.73925805e-01 4.87158597e-01 -2.09461629e-01 4.05518144e-01 1.04117131e+00 8.62625614e-02 4.82778519e-01 3.76127392e-01 -3.96000326e-01 -1.21113825e+00 -5.47782898e-01 -1.96236998e-01 1.17756701e+00 -6.68415353e-02 -8.48757029e-01 -7.28450418e-01 -5.27970254e-01 1.08927466e-01 3.10724735e-01 -3.87404025e-01 5.45253120e-02 -6.29915178e-01 -6.87633276e-01 4.81094629e-01 1.64870009e-01 5.15782058e-01 -7.89474547e-01 -9.60907698e-01 1.32466897e-01 -8.89145434e-02 -5.19181669e-01 -1.70329303e-01 6.26224399e-01 -1.22166789e+00 -9.86111045e-01 -7.09796548e-01 -4.46185589e-01 3.94703865e-01 7.48639628e-02 1.09476721e+00 2.82301843e-01 -9.13081765e-02 4.47708100e-01 -2.83342898e-01 -4.00871664e-01 -5.02377748e-01 4.44415390e-01 2.46396549e-02 -1.17543004e-01 -8.22122246e-02 -8.01564038e-01 -4.90126789e-01 2.68629398e-02 -1.23224711e+00 -1.74452409e-01 5.78316867e-01 1.09060788e+00 6.24697208e-01 6.88043758e-02 1.33049592e-01 -1.05052650e+00 5.97840726e-01 -4.67176288e-01 -7.14955270e-01 8.92108008e-02 -1.13452530e+00 4.47299957e-01 1.09280658e+00 -3.01196128e-01 -5.36851406e-01 -6.23804145e-02 1.96115058e-02 -4.62757260e-01 2.43930608e-01 4.06087726e-01 1.54276073e-01 -3.22199672e-01 5.62511325e-01 4.62454081e-01 2.45839804e-01 -6.51788056e-01 4.10170704e-01 8.47767651e-01 3.05989534e-01 -8.13322306e-01 6.31689250e-01 5.73043287e-01 2.18245745e-01 -4.60898250e-01 -4.66897488e-01 -3.41260612e-01 -1.72891989e-01 1.73349783e-01 3.08773428e-01 -5.01066446e-01 -8.35383713e-01 3.98204863e-01 -1.07056701e+00 -3.05784494e-01 -7.99432158e-01 5.61248004e-01 -8.91754925e-01 4.79208708e-01 -8.26741695e-01 -9.45310712e-01 -5.67326128e-01 -1.20658064e+00 7.69004524e-01 -7.87172914e-02 2.81074733e-01 -9.56945240e-01 2.61114955e-01 3.87252346e-02 8.32448006e-01 3.50861102e-02 7.82293499e-01 -8.33753288e-01 -6.40343189e-01 -1.82874307e-01 1.91302478e-01 4.46204305e-01 -3.48588765e-01 -3.69255722e-01 -5.79935849e-01 -7.70355880e-01 4.10751849e-01 -3.96865487e-01 9.18232739e-01 4.43906151e-02 1.34099507e+00 -7.19560444e-01 -1.57655865e-01 7.52606750e-01 1.57554090e+00 -1.02705680e-01 3.88326049e-01 2.89308965e-01 3.60710770e-01 4.24680620e-01 1.95398465e-01 6.01975739e-01 1.86560556e-01 6.53374672e-01 5.02545476e-01 2.80771852e-01 1.18885994e-01 5.11072427e-02 3.31077009e-01 1.45878303e+00 -4.70975101e-01 -1.20571844e-01 -6.07265115e-01 4.21357095e-01 -1.89164639e+00 -1.09765589e+00 9.99304503e-02 2.62332487e+00 1.18277383e+00 -6.59477338e-02 1.38881698e-01 4.94060874e-01 4.82206523e-01 1.64259225e-01 -3.61813813e-01 -7.26978838e-01 2.99238376e-02 4.78298068e-01 9.00789976e-01 5.21769762e-01 -6.74636424e-01 2.82005459e-01 5.98477030e+00 9.83465493e-01 -9.64524627e-01 3.41197163e-01 4.62296605e-01 -4.14320350e-01 -4.43026483e-01 1.65420786e-01 -6.63452625e-01 6.60161316e-01 1.13431013e+00 -2.94157535e-01 1.14523709e+00 8.18002164e-01 -3.29952598e-01 -8.53712857e-02 -1.11054504e+00 1.06954372e+00 1.31461501e-01 -1.31414640e+00 -3.69261891e-01 3.55006576e-01 8.82513762e-01 1.01450823e-01 -1.77940696e-01 1.00341938e-01 5.04259109e-01 -9.24733102e-01 6.24220669e-01 4.33387101e-01 5.03375888e-01 -9.67399299e-01 9.22842443e-01 5.47354579e-01 -8.57623100e-01 -3.34758699e-01 -4.16591287e-01 -2.65369713e-02 2.67507005e-02 8.71858478e-01 -1.89205125e-01 7.19856501e-01 7.07036436e-01 2.70409465e-01 -1.68896973e-01 1.03568375e+00 4.45805937e-02 7.94852376e-01 -6.76507235e-01 -2.06092387e-01 1.53144330e-01 -3.40471655e-01 7.57300675e-01 1.17107117e+00 3.85202408e-01 -1.15609899e-01 7.20510781e-02 3.34568888e-01 -2.57672876e-01 3.69080633e-01 -4.72343534e-01 2.50991106e-01 4.70347494e-01 1.13834262e+00 -4.28827405e-01 -5.21331131e-01 -2.56978750e-01 6.42520905e-01 4.80522245e-01 8.46147258e-03 -5.41353106e-01 -5.05136132e-01 3.09321046e-01 7.84842893e-02 6.05629265e-01 -2.41299272e-01 -2.42202118e-01 -1.09669709e+00 6.51480034e-02 -1.19350624e+00 5.57552695e-01 1.68137699e-01 -1.14309990e+00 4.65215743e-01 1.74344897e-01 -8.76219928e-01 -3.30043674e-01 -1.79490015e-01 -3.74718219e-01 5.77163815e-01 -1.59606731e+00 -6.91906571e-01 -7.19035864e-02 6.94411099e-01 1.80610880e-01 -1.44689515e-01 6.77126169e-01 2.89710015e-01 -3.11264396e-01 7.52874255e-01 8.57163608e-01 -4.23570246e-01 5.13687074e-01 -1.29165637e+00 -2.85187960e-01 6.92454278e-01 1.59939840e-01 6.10908508e-01 7.70672739e-01 -2.63154924e-01 -1.96119738e+00 -6.82654917e-01 9.57032204e-01 1.34232551e-01 5.08178651e-01 -2.95802087e-01 -7.96682417e-01 3.44056964e-01 1.96951479e-01 6.42407360e-03 5.38342357e-01 1.25270143e-01 -3.13036799e-01 -4.54145372e-01 -1.18249249e+00 2.44797960e-01 9.83688414e-01 -4.07840669e-01 -2.09560826e-01 7.11811841e-01 4.69807774e-01 -3.96619827e-01 -9.10228491e-01 5.55969849e-02 3.75386447e-01 -1.32043898e+00 6.37761652e-01 -3.31862271e-01 3.78135532e-01 -7.80286491e-02 -2.88314432e-01 -1.19494474e+00 8.21065828e-02 -1.18477130e+00 -6.19154871e-01 1.06790340e+00 1.01731911e-01 -9.52224910e-01 7.46627033e-01 2.38866717e-01 -3.88114788e-02 -1.01453996e+00 -1.24690402e+00 -1.02429783e+00 4.15727764e-01 -2.65010267e-01 7.35410690e-01 6.38898015e-01 1.60790488e-01 -2.65342686e-02 -4.09651399e-01 -4.56861764e-01 6.47632480e-01 4.11877155e-01 8.05230141e-01 -9.94910061e-01 -1.02970243e+00 -6.21798396e-01 -2.26934806e-01 -1.11076963e+00 -9.34545025e-02 -1.17270148e+00 1.71467774e-02 -9.22972679e-01 3.54923308e-01 -6.91177666e-01 -1.40086740e-01 2.43745282e-01 2.22155333e-01 1.61401883e-01 4.61339712e-01 5.17713428e-01 -5.18756211e-01 4.43622708e-01 9.74861681e-01 -1.90237835e-02 -3.32016259e-01 1.32633537e-01 -5.01527607e-01 4.81494814e-01 8.02438259e-01 -6.59350634e-01 -1.98293954e-01 -4.71710801e-01 4.84219730e-01 5.11023663e-02 1.51341200e-01 -1.03089797e+00 6.43193185e-01 4.21506912e-02 -2.95551002e-01 -8.79900679e-02 9.21239797e-03 -8.38141263e-01 3.23282957e-01 8.18417132e-01 -5.45024276e-01 2.89015830e-01 -3.44017923e-01 6.77829683e-01 -1.30441368e-01 -6.21669590e-01 7.85035372e-01 -4.27406520e-01 8.61368850e-02 1.99396282e-01 -2.18486920e-01 1.58378407e-01 8.89950573e-01 -3.45001779e-02 -2.02961877e-01 -4.90084112e-01 -4.65657979e-01 6.26070648e-02 5.28041720e-01 -2.82805443e-01 2.67651409e-01 -9.46859837e-01 -8.18207502e-01 8.45125169e-02 -4.10098433e-01 3.02125305e-01 -1.05001643e-01 1.28258705e+00 -4.24618661e-01 2.32911274e-01 1.98698148e-01 -4.01577234e-01 -1.12283766e+00 7.37608492e-01 2.46225268e-01 -6.57585740e-01 -5.81620216e-01 6.92882478e-01 -4.20206189e-01 -1.31344065e-01 4.46805745e-01 -3.45190242e-02 3.82862955e-01 -1.07546546e-01 4.59929019e-01 6.95144475e-01 2.83998251e-01 -2.23619357e-01 -7.19463602e-02 4.61814344e-01 -1.50578655e-02 -9.69460085e-02 1.47809267e+00 -1.68321177e-01 -4.67535704e-01 3.92514676e-01 1.39603841e+00 3.09924066e-01 -8.44029248e-01 -3.12652558e-01 -8.80323574e-02 -4.58115220e-01 1.83118939e-01 -5.40132225e-01 -1.39692247e+00 8.49576771e-01 5.61710536e-01 4.31927770e-01 1.28701770e+00 -1.10714763e-01 9.71569955e-01 4.92792964e-01 7.93870687e-01 -1.09549308e+00 -3.51081014e-01 4.35674727e-01 5.60513973e-01 -7.84965217e-01 2.93128431e-01 1.03533007e-01 -3.52151632e-01 1.23260021e+00 7.41369724e-02 -9.19202864e-02 4.19444084e-01 4.58293915e-01 -3.67187530e-01 -5.48232161e-02 -8.44943762e-01 -7.66702294e-02 -2.44644120e-01 2.53039330e-01 2.51969755e-01 1.00077689e-01 -1.03530097e+00 3.62388432e-01 -3.99881303e-01 3.00344285e-02 4.72234011e-01 1.17169642e+00 -3.54743928e-01 -1.38451529e+00 -3.95074636e-01 4.31778103e-01 -6.76173329e-01 -3.76874879e-02 -1.07389554e-01 5.98865509e-01 2.10524291e-01 7.76027143e-01 -1.01058081e-01 -2.55395979e-01 -1.17302820e-01 1.46776021e-01 4.81020868e-01 -1.68309525e-01 -1.15196013e+00 -2.15603486e-01 -2.09670424e-01 -6.26112401e-01 -5.68392694e-01 -5.37212789e-01 -1.00369883e+00 -8.30067575e-01 -4.66674089e-01 6.59696102e-01 8.82971883e-01 7.58646965e-01 3.68149251e-01 2.38233656e-02 7.87562251e-01 -8.61234725e-01 -1.47852337e+00 -5.36172807e-01 -7.55584002e-01 3.14198375e-01 2.25355446e-01 -1.91117197e-01 -6.89960122e-01 -2.97581047e-01]

[6.310446739196777, 4.84208345413208]

6d9cee5e-66ce-464e-be2e-bfaa03372fa4

non-rigid-medical-image-registration-using

2302.10343

https://arxiv.org/abs/2302.10343v1

https://arxiv.org/pdf/2302.10343v1.pdf

Non-rigid Medical Image Registration using Physics-informed Neural Networks

Biomechanical modelling of soft tissue provides a non-data-driven method for constraining medical image registration, such that the estimated spatial transformation is considered biophysically plausible. This has not only been adopted in real-world clinical applications, such as the MR-to-ultrasound registration for prostate intervention of interest in this work, but also provides an explainable means of understanding the organ motion and spatial correspondence establishment. This work instantiates the recently-proposed physics-informed neural networks (PINNs) to a 3D linear elastic model for modelling prostate motion commonly encountered during transrectal ultrasound guided procedures. To overcome a widely-recognised challenge in generalising PINNs to different subjects, we propose to use PointNet as the nodal-permutation-invariant feature extractor, together with a registration algorithm that aligns point sets and simultaneously takes into account the PINN-imposed biomechanics. The proposed method has been both developed and validated in both patient-specific and multi-patient manner.

['Yipeng Hu', 'Zeike A. Taylor', 'Dean C. Barratt', 'Mark Emberton', 'Shaheer U. Saeed', 'Zachary M. C. Baum', 'Zhe Min']

2023-02-20

['medical-image-registration']

['medical']

[ 5.57132304e-01 4.19316322e-01 -1.79214016e-01 -2.59046972e-01 -4.98275667e-01 -5.62604308e-01 8.19626451e-01 3.50316793e-01 -5.79750896e-01 4.44706023e-01 3.06476951e-01 -2.30906606e-01 -8.73821795e-01 -4.84996825e-01 -5.31362712e-01 -6.60534561e-01 -5.65437794e-01 8.27147067e-01 1.11877054e-01 -2.91998386e-01 3.72265577e-01 9.57913637e-01 -8.32449019e-01 -1.50646091e-01 7.74822891e-01 5.98247230e-01 2.51976907e-01 7.45416403e-01 2.13839471e-01 1.83989391e-01 7.54662156e-02 -1.22060806e-01 3.23624015e-01 -2.32518911e-01 -1.06425667e+00 -3.04957747e-01 4.00441587e-01 7.27853030e-02 -2.36687195e-02 7.66350389e-01 6.83576584e-01 2.79477298e-01 7.16881931e-01 -5.50456703e-01 -3.00048202e-01 4.32267636e-01 -4.52194840e-01 2.56244808e-01 2.34392703e-01 6.73983246e-02 2.99923092e-01 -4.90531415e-01 1.02590013e+00 8.28949928e-01 9.11913276e-01 5.82095087e-01 -1.26262999e+00 -6.21260442e-02 -5.18876851e-01 -1.29247963e-01 -1.01854312e+00 -2.93255597e-01 9.96492505e-01 -5.76996565e-01 6.32419944e-01 8.11769485e-01 8.40478599e-01 9.53944147e-01 1.13918614e+00 1.70570359e-01 1.31152666e+00 -3.98144186e-01 2.85059154e-01 -2.37225294e-01 -6.94568679e-02 2.42481619e-01 3.97621393e-01 4.53152269e-01 -1.65532902e-01 -3.68014425e-01 1.24166465e+00 -1.90243170e-01 -3.52426529e-01 -6.55008018e-01 -1.37316358e+00 5.11035979e-01 7.08555996e-01 9.08640265e-01 -6.85943127e-01 3.60974729e-01 4.70527381e-01 -2.92226613e-01 3.16742897e-01 6.31496668e-01 -2.08113998e-01 -2.05073029e-01 -1.02982199e+00 2.26711527e-01 7.31983006e-01 2.93797404e-01 1.99331269e-01 -2.33950868e-01 -1.87379017e-01 3.84356052e-01 6.64697170e-01 2.14491054e-01 6.77243233e-01 -7.82650113e-01 9.74090397e-02 4.60346758e-01 -1.46573007e-01 -1.31278408e+00 -8.88863742e-01 -5.50826311e-01 -1.22343767e+00 1.53917789e-01 3.36143136e-01 2.64678776e-01 -1.10380101e+00 1.46929896e+00 6.67539179e-01 6.21729791e-01 -6.51165992e-02 1.17327547e+00 7.53523290e-01 -1.33344729e-03 2.66738564e-01 -2.30256528e-01 1.42575192e+00 -3.20893794e-01 -5.16539454e-01 -2.05418393e-01 6.14735603e-01 -6.92717195e-01 4.90163118e-01 1.66012011e-02 -1.32508099e+00 -3.93734574e-01 -8.19489300e-01 -7.84317497e-03 -1.36691153e-01 -3.12802911e-01 7.61744082e-01 5.63712955e-01 -1.06442142e+00 1.20999050e+00 -1.52209628e+00 -5.33974707e-01 2.64902174e-01 8.69532526e-01 -6.72339916e-01 1.85770243e-01 -1.04934883e+00 1.46546960e+00 5.54749131e-01 9.90408421e-01 -3.74477863e-01 -8.89161170e-01 -8.56886029e-01 -9.48718861e-02 3.08602080e-02 -9.98100400e-01 8.08917999e-01 -2.95823902e-01 -1.63049972e+00 9.21879470e-01 9.06651467e-02 -2.81355113e-01 7.81477928e-01 1.73342928e-01 -1.54268339e-01 -4.54628654e-02 -2.74355769e-01 3.23987752e-01 2.81084687e-01 -1.35694551e+00 4.69070613e-01 -4.51333612e-01 -3.12615573e-01 3.47787470e-01 -1.08436100e-01 -2.07350954e-01 -3.65166217e-01 -4.66408253e-01 6.23511195e-01 -1.35275471e+00 -8.06665421e-01 2.51992028e-02 -5.56619465e-01 3.50413352e-01 3.17776322e-01 -9.20199811e-01 6.57448530e-01 -1.80640304e+00 6.44743741e-01 7.28781939e-01 2.43252426e-01 2.66508162e-01 7.40024671e-02 4.10584956e-01 -2.01115057e-01 -6.38296306e-02 -5.85166395e-01 -2.01904446e-01 -2.60315210e-01 3.18932205e-01 3.45370531e-01 8.94815981e-01 -1.84415071e-03 1.18674910e+00 -8.65589261e-01 -5.92620850e-01 5.74217618e-01 7.28509188e-01 -2.19307452e-01 -2.44994964e-02 2.88354516e-01 1.23170304e+00 -7.21971512e-01 2.87670583e-01 9.91682470e-01 1.10259175e-01 3.56353462e-01 -5.87734103e-01 -1.41182646e-01 -1.81676939e-01 -1.11206114e+00 2.24316287e+00 -4.43044633e-01 -9.52739082e-03 1.28329962e-01 -8.26251209e-01 7.80987144e-01 3.80986243e-01 1.05523562e+00 -5.31678200e-01 2.60396421e-01 5.27738273e-01 4.12059635e-01 -5.79471946e-01 3.94632488e-01 -3.31028789e-01 -6.70031309e-02 -2.50631664e-02 3.34927849e-02 -3.44671458e-01 -2.34531865e-01 -2.87478298e-01 1.00634909e+00 6.78790331e-01 4.28017825e-01 -9.95195389e-01 8.30222487e-01 6.94977641e-02 3.35490823e-01 5.66957533e-01 -1.89977944e-01 6.55322671e-01 1.30241700e-02 -5.58646202e-01 -8.63275349e-01 -9.69861031e-01 -5.27682245e-01 4.01986428e-02 3.82464260e-01 2.79101729e-01 -4.88606334e-01 -1.56290233e-01 2.13196263e-01 3.79967541e-01 -7.76420236e-01 -1.27294868e-01 -1.10525322e+00 -8.13729048e-01 3.81148905e-01 3.16252828e-01 -1.84669346e-01 -1.18926585e+00 -7.01749861e-01 5.47301292e-01 7.84397423e-02 -9.30123031e-01 -1.73607886e-01 3.75244558e-01 -1.24319828e+00 -9.67291355e-01 -9.25007164e-01 -4.02642339e-01 7.13605762e-01 -4.27454501e-01 9.32490528e-01 1.14907540e-01 -4.99112666e-01 3.22134525e-01 6.81546554e-02 -1.07309751e-01 -7.71380901e-01 1.13933228e-01 1.80024989e-02 -3.05461198e-01 -1.39395416e-01 -8.04193020e-01 -8.38112593e-01 1.57136872e-01 -9.92561758e-01 -5.29435575e-02 8.11899424e-01 5.63226044e-01 7.02438593e-01 -4.22386140e-01 2.78741091e-01 -8.24163139e-01 7.17886269e-01 -2.32854277e-01 -2.71989256e-01 8.86728913e-02 -2.87578791e-01 -8.75639636e-03 4.25129920e-01 -3.87435794e-01 -8.42220545e-01 2.85084307e-01 -4.31501657e-01 -1.70836538e-01 -3.38169396e-01 9.17908788e-01 2.94756591e-01 -8.09120536e-01 7.23921597e-01 2.20422462e-01 2.70595282e-01 -3.02635610e-01 2.32304081e-01 1.17476523e-01 8.33342731e-01 -6.28425419e-01 9.81045425e-01 5.48199713e-01 9.03051257e-01 -6.91312790e-01 1.46049559e-01 -5.17178535e-01 -1.11406064e+00 -2.59612918e-01 9.49724555e-01 -2.60315627e-01 -1.04902577e+00 3.01763147e-01 -1.07147205e+00 -1.86620682e-01 -3.39014351e-01 8.62062871e-01 -7.70533741e-01 6.56391740e-01 -6.19379103e-01 -6.79556429e-01 -5.34640610e-01 -1.41093576e+00 1.04984617e+00 2.52191842e-01 -4.18236732e-01 -1.48705399e+00 4.10787016e-01 8.15794393e-02 8.05500031e-01 1.02282429e+00 6.61197543e-01 -7.03574240e-01 -5.55863380e-01 -3.84696066e-01 1.07314736e-01 -2.66463429e-01 2.72219688e-01 -2.21070841e-01 -8.59723330e-01 -3.86668086e-01 3.11965704e-01 2.10126221e-01 3.61041397e-01 9.20617759e-01 9.41659272e-01 9.95351300e-02 -6.20328128e-01 5.22620738e-01 1.81602132e+00 1.52598023e-01 7.49564648e-01 2.96955466e-01 7.54431009e-01 5.73443592e-01 1.81167856e-01 -3.46438549e-02 7.17093199e-02 9.10997391e-01 6.89898133e-01 -3.73449504e-01 1.71610951e-01 2.95416445e-01 -3.91722351e-01 1.00425184e+00 -7.79962659e-01 8.78984705e-02 -1.08646989e+00 4.28605914e-01 -1.84035337e+00 -6.11544669e-01 -4.16852564e-01 2.44820809e+00 7.23232687e-01 1.19248867e-01 -4.44083095e-01 -3.33358347e-01 4.05513674e-01 -8.43562782e-02 -2.62800872e-01 -3.87587458e-01 3.87196422e-01 5.30793190e-01 8.44989419e-01 6.86715424e-01 -9.08565164e-01 1.73527434e-01 5.51716089e+00 4.71915066e-01 -1.35222864e+00 1.18837714e-01 3.62076551e-01 4.35998410e-01 -2.36302957e-01 3.72648016e-02 -1.24150559e-01 2.86262751e-01 1.05271018e+00 -2.74065859e-03 1.65616110e-01 4.34572160e-01 6.44346118e-01 -2.08734199e-01 -1.26662767e+00 4.40692693e-01 -2.00730875e-01 -1.59816146e+00 -4.72233623e-01 4.59057420e-01 2.57077247e-01 -4.46362086e-02 -1.89528525e-01 -2.27156505e-01 -5.07008135e-01 -1.16702497e+00 2.51871169e-01 1.09917808e+00 5.06682515e-01 -3.08463961e-01 9.54318345e-01 3.79450679e-01 -1.07032049e+00 4.79310155e-01 -3.20502520e-02 3.50833863e-01 4.38623369e-01 4.94399935e-01 -9.41732049e-01 1.22573066e+00 3.72079670e-01 4.64592993e-01 -3.60832393e-01 1.28251266e+00 5.06681979e-01 1.47285655e-01 -4.82882261e-01 2.41420373e-01 1.45715341e-01 -2.92880893e-01 6.97474599e-01 1.07844174e+00 4.20209467e-01 -2.47811675e-01 5.73407970e-02 9.95875657e-01 5.18760264e-01 1.62268221e-01 -4.93526548e-01 2.11791351e-01 5.46216220e-03 1.59634864e+00 -1.29734170e+00 1.81224510e-01 2.11104140e-01 9.32269394e-01 -1.37385666e-01 -7.85409007e-03 -4.76477116e-01 1.18704319e-01 1.92509413e-01 2.30912164e-01 -2.14468792e-01 -4.62158620e-01 -2.50440955e-01 -8.09072137e-01 3.36373560e-02 -2.06790715e-01 -9.42305326e-02 -5.43245554e-01 -1.27466810e+00 6.12647891e-01 4.50921148e-01 -1.13490069e+00 -4.87932712e-01 -7.01471567e-01 -7.20058084e-01 1.49909163e+00 -1.56254494e+00 -1.41475928e+00 -3.14486265e-01 1.61949620e-01 -1.47547096e-01 5.46133876e-01 1.17177689e+00 2.81400412e-01 -2.76692063e-02 6.53878972e-02 -4.64219823e-02 -8.64956304e-02 6.27118826e-01 -1.36945939e+00 3.41089547e-01 5.40346920e-01 -4.89389181e-01 1.16734707e+00 9.90037084e-01 -9.33444262e-01 -1.88255501e+00 -7.02153087e-01 7.84867406e-01 -4.73453879e-01 5.72821915e-01 -1.09549291e-01 -1.06428897e+00 4.14083540e-01 1.60348400e-01 5.64007998e-01 5.60170770e-01 -2.36586019e-01 8.07498693e-01 2.27960765e-01 -1.40674257e+00 4.87499475e-01 8.44226599e-01 -1.61441758e-01 -5.02771318e-01 4.56088066e-01 1.06481649e-01 -1.26985240e+00 -1.64637637e+00 6.57530487e-01 5.57092369e-01 -6.78425610e-01 1.29983783e+00 -2.71799296e-01 3.36447328e-01 -1.84911817e-01 3.43142748e-01 -1.21754801e+00 -4.10078406e-01 -7.22837031e-01 -2.58457456e-02 7.89195895e-01 -1.06415123e-01 -6.52073562e-01 1.06801522e+00 8.45383406e-01 -4.23832625e-01 -8.85786116e-01 -1.68703091e+00 -5.88849723e-01 2.75144786e-01 -2.26579577e-01 1.50672063e-01 1.14404690e+00 -1.67581350e-01 -6.45213068e-01 -1.24907508e-01 3.35116625e-01 5.98025918e-01 -2.22397581e-01 2.41970077e-01 -1.51211715e+00 -2.78691739e-01 -3.88155520e-01 -8.11465740e-01 -3.96339595e-01 -8.47379416e-02 -1.19191277e+00 1.38152033e-01 -1.58217049e+00 5.65619506e-02 -7.71378696e-01 -3.83256257e-01 6.84019402e-02 8.67252331e-03 2.61776298e-01 1.37305213e-02 4.13207531e-01 3.23997885e-01 2.14349121e-01 1.76050019e+00 2.14962378e-01 -4.31737363e-01 4.92024794e-02 -3.63675356e-01 5.90810001e-01 4.14232671e-01 -4.05827701e-01 -1.84368387e-01 8.79444405e-02 1.22262597e-01 3.54433388e-01 6.69761240e-01 -1.02909935e+00 3.44980747e-01 -2.90964514e-01 4.21418190e-01 -4.43312138e-01 2.50339478e-01 -1.35457861e+00 7.21112072e-01 9.28062022e-01 -1.44738287e-01 -1.67660359e-02 4.33500409e-01 4.59479332e-01 4.14073095e-02 -4.76710081e-01 5.23098350e-01 -2.40720570e-01 -4.14145678e-01 2.80288905e-01 6.91966247e-03 -4.66995001e-01 8.80952179e-01 -6.14206016e-01 9.51792579e-03 1.43294305e-01 -1.08329809e+00 -1.55279934e-01 2.53533810e-01 1.06701940e-01 4.25582111e-01 -1.31938708e+00 -6.34873152e-01 1.22261189e-01 -6.84849471e-02 2.55703688e-01 7.15939224e-01 1.44524407e+00 -8.99928510e-01 4.80483532e-01 -3.64585370e-01 -9.92429316e-01 -8.74606371e-01 1.88673422e-01 7.37590790e-01 -6.54725075e-01 -7.46515632e-01 5.30114293e-01 -3.27770710e-01 -9.03652310e-01 -3.88371497e-01 -6.79471076e-01 -3.61486107e-01 -5.05757093e-01 -2.78348088e-01 2.82661282e-02 4.22409147e-01 -9.88880754e-01 -7.17874289e-01 9.78975892e-01 2.31998056e-01 7.21186697e-02 1.51676083e+00 7.85854831e-03 -2.46559516e-01 1.94443479e-01 9.29619193e-01 9.26446021e-02 -1.00954950e+00 -7.91029539e-03 1.23937562e-01 -2.03379244e-01 1.62260488e-01 -8.33845675e-01 -8.23255062e-01 4.75187123e-01 8.40678871e-01 2.00504482e-01 9.21199918e-01 -1.61776513e-01 4.85130787e-01 -9.10396799e-02 3.21085364e-01 -6.17590249e-01 -5.38331032e-01 3.61958146e-02 1.05443418e+00 -1.04178703e+00 4.34069335e-01 -7.90804863e-01 -2.57221758e-01 1.33151126e+00 2.56519556e-01 -3.52387816e-01 7.23097265e-01 4.01912361e-01 -3.35048065e-02 -2.78214574e-01 1.11319408e-01 3.85587662e-01 8.01686406e-01 6.77840948e-01 7.80050218e-01 -8.18109140e-02 -7.95989811e-01 1.13751948e-01 7.65160378e-03 5.23474634e-01 4.02828455e-01 1.27054107e+00 1.75808132e-01 -1.31258953e+00 -4.13520962e-01 2.41890803e-01 -3.89114380e-01 5.33958077e-02 8.14949498e-02 1.02928305e+00 2.09722146e-01 5.26967570e-02 -5.80575615e-02 -6.36099279e-03 3.56522232e-01 -1.98121816e-01 7.05620110e-01 -4.12954181e-01 -1.07778871e+00 2.03015074e-01 -1.56438127e-01 -5.66674948e-01 -9.03818190e-01 -7.25374699e-01 -1.34140897e+00 1.43983752e-01 -2.97980011e-01 -1.23059161e-01 1.23461437e+00 1.04303908e+00 3.56134534e-01 8.70642543e-01 2.32342273e-01 -1.70834720e+00 -4.08732384e-01 -9.20616567e-01 -3.92386258e-01 6.23072088e-01 2.17358157e-01 -6.98615849e-01 -4.94550951e-02 -2.16431290e-01]

[14.023699760437012, -2.6408796310424805]

f80b3898-5fc6-4375-96c9-a2d1299fa591

distilling-vision-language-pre-training-to

2212.09335

https://arxiv.org/abs/2212.09335v1

https://arxiv.org/pdf/2212.09335v1.pdf

Distilling Vision-Language Pre-training to Collaborate with Weakly-Supervised Temporal Action Localization

Weakly-supervised temporal action localization (WTAL) learns to detect and classify action instances with only category labels. Most methods widely adopt the off-the-shelf Classification-Based Pre-training (CBP) to generate video features for action localization. However, the different optimization objectives between classification and localization, make temporally localized results suffer from the serious incomplete issue. To tackle this issue without additional annotations, this paper considers to distill free action knowledge from Vision-Language Pre-training (VLP), since we surprisingly observe that the localization results of vanilla VLP have an over-complete issue, which is just complementary to the CBP results. To fuse such complementarity, we propose a novel distillation-collaboration framework with two branches acting as CBP and VLP respectively. The framework is optimized through a dual-branch alternate training strategy. Specifically, during the B step, we distill the confident background pseudo-labels from the CBP branch; while during the F step, the confident foreground pseudo-labels are distilled from the VLP branch. And as a result, the dual-branch complementarity is effectively fused to promote a strong alliance. Extensive experiments and ablation studies on THUMOS14 and ActivityNet1.2 reveal that our method significantly outperforms state-of-the-art methods.

['Qi Tian', 'Yanfeng Wang', 'Jianlong Chang', 'Ya zhang', 'Peisen Zhao', 'Jinxiang Liu', 'Kunhao Zheng', 'Chen Ju']

2022-12-19

http://openaccess.thecvf.com//content/CVPR2023/html/Ju_Distilling_Vision-Language_Pre-Training_To_Collaborate_With_Weakly-Supervised_Temporal_Action_Localization_CVPR_2023_paper.html

http://openaccess.thecvf.com//content/CVPR2023/papers/Ju_Distilling_Vision-Language_Pre-Training_To_Collaborate_With_Weakly-Supervised_Temporal_Action_Localization_CVPR_2023_paper.pdf

cvpr-2023-1

['weakly-supervised-temporal-action', 'action-localization']

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

[ 3.06526959e-01 -1.21735111e-02 -6.67196274e-01 -1.49552315e-01 -8.54440272e-01 -4.70769048e-01 6.96306229e-01 -3.18835676e-01 -4.44344074e-01 7.86265075e-01 2.47135729e-01 1.74460579e-02 2.83138275e-01 -4.30651993e-01 -6.96241975e-01 -1.06736434e+00 8.20913613e-02 4.31906953e-02 7.89071620e-01 6.46759644e-02 7.70923682e-03 1.87266394e-02 -1.34496105e+00 8.38024318e-01 8.24295402e-01 1.15912735e+00 2.37741426e-01 2.30779111e-01 -2.34538123e-01 1.39825022e+00 -3.07587266e-01 -3.95805389e-02 3.12026799e-01 -5.46593249e-01 -7.78767049e-01 2.12337732e-01 2.25060001e-01 -4.47411597e-01 -4.14888561e-01 1.00236714e+00 2.33855978e-01 1.22592799e-01 4.09308851e-01 -1.57396889e+00 -4.88382012e-01 4.71231997e-01 -7.90459633e-01 2.01086313e-01 3.25953603e-01 4.66575414e-01 1.00843298e+00 -8.70118082e-01 6.56155825e-01 1.07983303e+00 5.15318513e-01 5.87057650e-01 -9.14400160e-01 -6.15724146e-01 6.78924501e-01 4.93337214e-01 -1.29146826e+00 -3.36640596e-01 8.21378946e-01 -4.50860381e-01 5.66375911e-01 -3.84679474e-02 8.45467627e-01 1.53867459e+00 -2.80074403e-02 1.36550999e+00 1.20239282e+00 -2.58728623e-01 3.01520020e-01 -1.29190751e-03 -4.41564135e-02 8.62149537e-01 -1.82233647e-01 1.36305213e-01 -6.10574841e-01 2.23852947e-01 7.41136968e-01 1.69894964e-01 -3.70604128e-01 -4.80954677e-01 -1.39896727e+00 4.29767847e-01 4.95363295e-01 5.62904060e-01 -3.54581982e-01 3.70822906e-01 3.49975765e-01 -1.17222123e-01 4.41835463e-01 -9.78067145e-03 -4.95000750e-01 -2.63050884e-01 -9.24325526e-01 -7.54037201e-02 4.07040685e-01 8.88609886e-01 7.99892366e-01 -3.27252001e-02 -6.92377269e-01 6.70633674e-01 2.53053635e-01 1.61171332e-01 5.91059983e-01 -1.08334351e+00 6.51454926e-01 6.90813839e-01 1.27221972e-01 -7.06021011e-01 -1.23494834e-01 -4.69663411e-01 -6.17000461e-01 8.94365162e-02 5.88228643e-01 -6.66407049e-02 -9.85572457e-01 1.71852553e+00 4.48152363e-01 7.67652929e-01 1.66882072e-02 1.03619850e+00 5.90801418e-01 6.05166674e-01 2.05730110e-01 -2.30731726e-01 1.15214932e+00 -1.63559401e+00 -7.01571226e-01 -3.91645223e-01 7.35016942e-01 -2.98734188e-01 1.08875942e+00 3.60691875e-01 -8.01775873e-01 -7.37221301e-01 -9.79819357e-01 1.46413303e-03 -1.71948001e-01 4.79532808e-01 6.71503961e-01 2.41913959e-01 -7.74936378e-01 4.64550555e-01 -1.01704550e+00 -2.95075983e-01 7.81676888e-01 -9.70982686e-02 -4.98453617e-01 -1.47975817e-01 -9.99120653e-01 6.84921086e-01 6.85008049e-01 2.55225092e-01 -1.32450473e+00 -4.92624760e-01 -7.90828526e-01 -2.11583108e-01 8.84147644e-01 -2.08268270e-01 1.17496049e+00 -1.29461467e+00 -1.48282683e+00 6.71349883e-01 -1.90476298e-01 -3.84942561e-01 7.16326535e-01 -2.57905155e-01 -1.99213102e-01 4.11285758e-01 3.92813861e-01 7.59668350e-01 7.92899489e-01 -1.21644330e+00 -1.09339166e+00 -7.01621622e-02 5.08943379e-01 3.21772635e-01 -2.94186354e-01 -2.30286002e-01 -8.08063209e-01 -6.29998326e-01 1.76199242e-01 -8.61277938e-01 -1.93732068e-01 1.97768971e-01 -3.53123963e-01 -3.03850740e-01 7.94604242e-01 -6.66674674e-01 1.13188779e+00 -2.30657411e+00 5.73445819e-02 -1.26780540e-01 1.23370998e-01 1.84156016e-01 -2.86875784e-01 8.94813985e-02 -1.65919796e-01 -1.47755176e-01 -9.98132974e-02 -5.06605983e-01 -1.23712532e-01 4.90764916e-01 -3.32920223e-01 5.80832601e-01 3.31426919e-01 9.81392801e-01 -1.34174371e+00 -8.11455846e-01 3.87831032e-01 1.87481135e-01 -5.45440316e-01 8.14308971e-02 -4.18931514e-01 8.98742676e-01 -6.76842213e-01 1.02391315e+00 4.42036539e-01 -3.02617192e-01 1.71776772e-01 -3.26602757e-01 -2.36003250e-01 9.68878046e-02 -1.15451980e+00 2.09493208e+00 -2.52101600e-01 3.23710710e-01 5.81288431e-03 -1.42573583e+00 4.78687465e-01 3.04653138e-01 8.59520376e-01 -8.15917730e-01 8.37517828e-02 1.21904455e-01 -1.66768894e-01 -6.22139990e-01 5.69837354e-02 5.43474741e-02 4.06792611e-02 1.66338101e-01 2.35921592e-01 3.91877890e-01 1.95657164e-01 1.86139256e-01 1.19914675e+00 8.93973470e-01 1.76862568e-01 -1.41450774e-03 7.92732477e-01 -6.60175830e-02 1.04475188e+00 6.89578891e-01 -6.69330657e-01 5.54596663e-01 7.44596779e-01 -2.01270834e-01 -4.69147354e-01 -9.43545818e-01 1.91224709e-01 1.10586584e+00 4.42520499e-01 -3.68089467e-01 -7.87085295e-01 -1.30623126e+00 -3.18718344e-01 5.16864836e-01 -7.28739202e-01 -2.22902119e-01 -6.95515394e-01 -5.11871934e-01 5.44516623e-01 7.83617914e-01 1.04316473e+00 -1.17892158e+00 -5.11433303e-01 1.40620768e-01 -5.27706265e-01 -1.50805497e+00 -4.41677630e-01 2.76349455e-01 -5.85405111e-01 -1.05933130e+00 -7.51903355e-01 -7.01939046e-01 5.84159434e-01 3.79408985e-01 7.66988873e-01 6.54021502e-02 1.33444751e-02 3.06932986e-01 -7.59357631e-01 -1.78001896e-02 7.13621825e-03 -2.38156080e-01 -1.15632370e-01 4.80108052e-01 3.30293357e-01 -6.22653484e-01 -6.88280463e-01 4.27824795e-01 -6.32686496e-01 2.61831135e-01 7.50269651e-01 8.12303603e-01 8.38704586e-01 1.22518465e-02 3.19013596e-01 -5.24063587e-01 -1.10202372e-01 -3.81273389e-01 -3.65901440e-01 3.88760597e-01 -2.41480723e-01 7.27959583e-03 4.86176223e-01 -7.37511098e-01 -1.02658880e+00 3.57192069e-01 3.19277793e-02 -7.27954447e-01 -2.32786089e-01 3.46259415e-01 -4.35072929e-01 1.00268628e-02 3.51546943e-01 5.29089868e-01 -3.42030197e-01 -3.89167070e-01 4.42424268e-01 3.24468821e-01 6.53237939e-01 -5.14648855e-01 6.34996951e-01 7.04461575e-01 -2.96870798e-01 -4.12992895e-01 -1.20119846e+00 -5.20371377e-01 -7.90863454e-01 -4.94278520e-01 1.22357130e+00 -1.00331807e+00 -4.17533517e-01 5.59184432e-01 -1.18205452e+00 -5.35248280e-01 -4.29593533e-01 6.24395430e-01 -5.10172129e-01 4.63454306e-01 -3.19518417e-01 -7.39791572e-01 1.92080915e-01 -1.19781101e+00 1.16443121e+00 2.35284790e-01 1.68833792e-01 -6.79962158e-01 -5.79863563e-02 5.13156712e-01 -9.68942605e-03 2.11583212e-01 4.09928352e-01 -6.11717582e-01 -7.51848638e-01 -3.88093255e-02 -4.48348135e-01 5.66204965e-01 1.16117664e-01 -2.70175457e-01 -1.09796786e+00 -9.43331420e-02 -6.81167319e-02 -4.79118019e-01 1.05217266e+00 2.52860904e-01 1.38631153e+00 -9.93728340e-02 -4.37735915e-01 6.35327280e-01 1.07042885e+00 1.64034173e-01 6.41319394e-01 3.35066557e-01 7.47206092e-01 4.47883815e-01 1.04097283e+00 4.15965229e-01 3.32584381e-01 8.66741836e-01 5.13671696e-01 -1.28008023e-01 -3.03993493e-01 -4.24966723e-01 7.63631642e-01 3.94267350e-01 -3.88646245e-01 -1.37600943e-01 -6.25458777e-01 4.12726343e-01 -2.20915937e+00 -1.10656786e+00 6.76147491e-02 2.01323891e+00 8.92717540e-01 1.80657491e-01 3.11886705e-02 4.90551069e-02 5.34465134e-01 5.58263958e-01 -5.04235268e-01 5.56673586e-01 -9.88419503e-02 -3.51487584e-02 3.98508281e-01 2.98520356e-01 -1.38836956e+00 1.09413445e+00 5.00863600e+00 1.15322566e+00 -9.84536469e-01 5.86931467e-01 4.64775980e-01 -1.44663304e-01 1.42519042e-01 3.16356778e-01 -7.10914731e-01 6.64605916e-01 3.49714696e-01 3.36880177e-01 2.40953609e-01 8.54306340e-01 2.75783747e-01 -3.34476918e-01 -1.27333665e+00 1.11404729e+00 3.04305088e-02 -1.13307440e+00 -2.14851126e-01 2.63540685e-04 6.77680969e-01 -3.20401527e-02 -1.71051204e-01 7.68906057e-01 1.49630979e-01 -5.70698261e-01 1.09182978e+00 4.71197546e-01 5.35212696e-01 -3.39563966e-01 6.26103163e-01 5.48709869e-01 -1.62165940e+00 -3.03980350e-01 -1.64635614e-01 -6.40327185e-02 3.64466459e-01 4.97948289e-01 -3.31916958e-01 7.19256520e-01 8.02213371e-01 1.16322064e+00 -5.59476912e-01 8.18495154e-01 -6.68280005e-01 6.74577594e-01 -6.40990660e-02 2.48722374e-01 6.22029901e-01 -3.41314524e-01 4.54330385e-01 1.01585221e+00 7.57297575e-02 1.41388044e-01 7.13784277e-01 8.51694703e-01 1.24405861e-01 -2.03233808e-01 -3.36226255e-01 -1.40363306e-01 2.25757867e-01 1.19880283e+00 -7.93397903e-01 -4.77410764e-01 -5.85918069e-01 1.25440562e+00 3.37834418e-01 5.30418098e-01 -1.36794806e+00 6.52225316e-02 4.17293936e-01 2.53617531e-03 3.25538367e-01 -2.06500202e-01 -8.67910217e-03 -1.44542611e+00 2.60573238e-01 -7.06999660e-01 3.66049588e-01 -7.53938615e-01 -9.91353989e-01 2.97323614e-01 1.01361588e-01 -1.63984764e+00 1.44584840e-02 -4.15062547e-01 -6.02667093e-01 5.07468402e-01 -1.69231796e+00 -1.38910306e+00 -4.71808642e-01 8.30068529e-01 8.41781080e-01 4.28484613e-03 3.99084330e-01 5.06797910e-01 -8.37467194e-01 4.31513995e-01 -2.52086550e-01 3.19532573e-01 6.20680571e-01 -1.12676251e+00 -3.75451058e-01 1.04726613e+00 2.91245818e-01 2.14669943e-01 2.73834646e-01 -6.18843853e-01 -1.00062883e+00 -1.28428900e+00 5.43522894e-01 -4.62829113e-01 7.14361548e-01 -3.00742090e-01 -6.82696700e-01 6.61073804e-01 5.50787374e-02 3.99070144e-01 3.48570645e-01 -1.60531640e-01 -2.64974564e-01 -2.01075226e-01 -8.13588381e-01 6.26266301e-01 1.41381669e+00 -5.05032539e-01 -5.63365877e-01 6.79523289e-01 7.13533103e-01 -2.20756009e-01 -5.01214564e-01 4.82106388e-01 5.14225483e-01 -1.03860462e+00 9.75036383e-01 -3.58372003e-01 5.46263158e-01 -6.28120482e-01 -2.47690707e-01 -8.25966775e-01 -1.28409624e-01 -3.91972840e-01 -3.36331040e-01 1.38513589e+00 1.87988788e-01 -4.90844101e-01 7.58651197e-01 2.35701546e-01 -2.53488034e-01 -9.11637068e-01 -9.54092443e-01 -9.14718568e-01 -2.18694612e-01 -5.13584495e-01 3.93776931e-02 8.70268583e-01 -6.56358227e-02 1.38490573e-01 -5.20245492e-01 1.27095520e-01 3.94069403e-01 6.60901964e-02 6.78177297e-01 -8.34968388e-01 -5.07286787e-01 -3.00886631e-01 -2.42383078e-01 -1.50577796e+00 2.94455767e-01 -7.56714344e-01 3.34899545e-01 -1.35659790e+00 2.15946704e-01 -4.39052850e-01 -5.77135623e-01 9.94470239e-01 -1.62033200e-01 3.37548435e-01 2.62696534e-01 3.69361281e-01 -1.01639915e+00 7.53574312e-01 1.42862356e+00 -2.27311268e-01 -1.77515626e-01 -5.00888005e-02 -3.97357851e-01 9.89460588e-01 4.79720980e-01 -4.91718441e-01 -6.47441983e-01 -3.23458046e-01 -2.97079444e-01 -4.17266972e-02 6.41702890e-01 -1.11287999e+00 2.50994861e-01 -4.18319613e-01 3.07357907e-01 -6.26529455e-01 4.28876430e-01 -7.89770305e-01 -3.63297194e-01 4.10532236e-01 -2.74437368e-01 -4.49576676e-01 -1.54554546e-01 8.58665287e-01 -3.53959918e-01 -1.43147543e-01 7.49819875e-01 -2.97562242e-01 -1.25317025e+00 4.80261058e-01 -2.11469993e-01 2.73060426e-02 1.35119510e+00 -3.11546862e-01 -2.84544796e-01 -1.79114401e-01 -7.46145368e-01 4.45438027e-01 2.42817715e-01 4.04324740e-01 4.08733368e-01 -1.41181171e+00 -4.36223119e-01 8.39679390e-02 1.25695899e-01 1.31001413e-01 2.88861096e-01 1.50022829e+00 -1.93733573e-01 2.50143051e-01 -3.79371047e-02 -8.04645956e-01 -9.78452742e-01 5.77990949e-01 2.77737767e-01 -5.67202985e-01 -7.68977225e-01 9.84118998e-01 4.56373394e-01 -1.96863189e-01 4.61487114e-01 -4.54687327e-01 -2.82464027e-01 1.72378853e-01 5.49242139e-01 1.96243539e-01 -2.71928698e-01 -6.22293949e-01 -5.62738478e-01 4.76255089e-01 1.18395664e-01 -3.08084071e-01 1.01425374e+00 -1.06754988e-01 5.19144088e-02 4.31919843e-01 9.79345262e-01 -1.17703371e-01 -1.71117926e+00 -3.30140054e-01 -5.23504205e-02 -5.21563172e-01 6.01616269e-03 -7.93176830e-01 -1.28305221e+00 8.97931397e-01 5.70258677e-01 -1.74869269e-01 1.25360000e+00 7.14589581e-02 6.76416457e-01 2.68132627e-01 5.44995189e-01 -1.23167312e+00 6.07147455e-01 4.85369861e-01 6.61003590e-01 -1.19112062e+00 -1.05954029e-01 -4.97623712e-01 -9.39366579e-01 8.40678990e-01 9.58894789e-01 -3.89689244e-02 5.09784818e-01 8.99847783e-03 -1.55506194e-01 9.69286636e-02 -6.65124774e-01 -4.24112558e-01 1.59351140e-01 5.97654641e-01 3.42613645e-02 -3.11116755e-01 -3.84947628e-01 9.08236921e-01 5.37265897e-01 2.82609433e-01 1.36915013e-01 1.21489966e+00 -3.24409515e-01 -1.05930507e+00 -7.74644837e-02 8.00634250e-02 -2.87576050e-01 -6.20732009e-02 -2.06627071e-01 7.19362199e-01 7.45689332e-01 9.22193289e-01 -1.84860528e-01 -5.77453375e-01 6.01864457e-02 4.76446413e-02 3.63455772e-01 -5.41867018e-01 -2.59151787e-01 1.63060129e-01 3.44208814e-02 -1.13690484e+00 -8.94635141e-01 -6.12702250e-01 -1.38319206e+00 3.45250070e-01 -4.05616581e-01 4.40839194e-02 2.06592396e-01 1.26203847e+00 1.56703413e-01 7.20245957e-01 5.18800914e-01 -9.83492613e-01 -4.35607195e-01 -9.81495321e-01 -5.02033651e-01 3.21623802e-01 2.34588712e-01 -1.05944455e+00 -3.95740241e-01 2.44996205e-01]

[8.466673851013184, 0.633354663848877]

97568580-b168-47ac-899a-b65fec7d8cc8

dynamic-survival-prediction-in-intensive-care

1909.07214

https://arxiv.org/abs/1909.07214v2

https://arxiv.org/pdf/1909.07214v2.pdf

Dynamic survival prediction in intensive care units from heterogeneous time series without the need for variable selection or pre-processing

We present a machine learning pipeline and model that uses the entire uncurated EHR for prediction of in-hospital mortality at arbitrary time intervals, using all available chart, lab and output events, without the need for pre-processing or feature engineering. Data for more than 45,000 American ICU patients from the MIMIC-III database were used to develop an ICU mortality prediction model. All chart, lab and output events were treated by the model in the same manner inspired by Natural Language Processing (NLP). Patient events were discretized by percentile and mapped to learnt embeddings before being passed to a Recurrent Neural Network (RNN) to provide early prediction of in-patient mortality risk. We compared mortality predictions with the Simplified Acute Physiology Score II (SAPS II) and the Oxford Acute Severity of Illness Score (OASIS). Data were split into an independent test set (10%) and a ten-fold cross-validation was carried out during training to avoid overfitting. 13,233 distinct variables with heterogeneous data types were included without manual selection or pre-processing. Recordings in the first few hours of a patient's stay were found to be strongly predictive of mortality, outperforming models using SAPS II and OASIS scores within just 2 hours and achieving a state of the art Area Under the Receiver Operating Characteristic (AUROC) value of 0.80 (95% CI 0.79-0.80) at 12 hours vs 0.70 and 0.66 for SAPS II and OASIS at 24 hours respectively. Our model achieves a very strong performance of AUROC 0.86 (95% CI 0.85-0.86) for in-patient mortality prediction after 48 hours on the MIMIC-III dataset. Predictive performance increases over the first 48 hours of the ICU stay, but suffers from diminishing returns, providing rationale for time-limited trials of critical care and suggesting that the timing of decision making can be optimised and individualised.

['Pietro Liò', 'Jacob Deasy', 'Ari Ercole']

2019-09-13

['icu-mortality']

['medical']

[ 3.04278079e-02 -7.71968290e-02 2.82289907e-02 -3.10031086e-01 -8.21961045e-01 -5.17196715e-01 9.03877318e-02 1.30159712e+00 -6.56517446e-01 5.91206968e-01 6.15238488e-01 -9.82114673e-01 -5.55823207e-01 -6.42157555e-01 -6.06561825e-02 -3.88623983e-01 -5.30667067e-01 6.43128335e-01 -2.69024074e-01 3.32180679e-01 -1.24743246e-01 6.26155674e-01 -8.93852293e-01 5.25938630e-01 4.42020535e-01 8.93554688e-01 -3.46884787e-01 1.17258537e+00 4.06019129e-02 1.11285341e+00 -4.87680644e-01 1.35646716e-01 3.73331040e-01 -5.90297639e-01 -4.45720434e-01 -6.24818087e-01 -1.58186674e-01 -3.10901761e-01 -3.06301028e-01 -6.94727823e-02 1.00666809e+00 -1.00505568e-01 7.27867126e-01 -8.36671650e-01 3.09530622e-03 3.74974996e-01 2.45341033e-01 1.80731311e-01 2.43960455e-01 3.77785444e-01 6.60485625e-01 -6.56584978e-01 2.04065502e-01 6.70509517e-01 1.12532806e+00 2.84717411e-01 -1.40248358e+00 -2.83572733e-01 -4.52016026e-01 -2.35403150e-01 -1.32410514e+00 1.74264815e-02 -5.75531200e-02 -7.71062315e-01 1.22141349e+00 3.14981550e-01 8.09807181e-01 8.17225218e-01 7.60650277e-01 -9.73425582e-02 8.08005512e-01 -3.06022614e-01 2.31432602e-01 1.80384532e-01 3.16161543e-01 4.94327813e-01 3.57230455e-01 2.13094085e-01 -1.29377395e-01 -7.10934877e-01 7.68657923e-01 9.91883814e-01 -2.83383757e-01 -5.18413354e-03 -1.58344507e+00 6.95985138e-01 1.81025967e-01 -8.29444826e-02 -7.31496990e-01 -4.09069091e-01 8.25700879e-01 5.95556080e-01 7.63579365e-03 6.34526134e-01 -1.15692115e+00 -4.02147233e-01 -9.02117312e-01 -3.23241621e-01 1.07610750e+00 4.57776487e-01 4.64152806e-02 -2.03797355e-01 -4.06602234e-01 8.48385990e-01 4.42784689e-02 1.57523572e-01 6.84477508e-01 -7.80320227e-01 2.48413742e-01 7.75300920e-01 1.52994514e-01 -4.03556973e-01 -1.10599911e+00 -5.27464449e-01 -1.05197346e+00 1.06265083e-01 5.01277864e-01 -5.39684355e-01 -6.71528876e-01 1.32532418e+00 -2.03326777e-01 -6.29337132e-02 4.29240793e-01 3.96167636e-01 5.67161441e-01 4.94012028e-01 5.15866756e-01 -5.16766310e-01 1.46051443e+00 -3.92691851e-01 -8.85003954e-02 2.78636724e-01 1.21612442e+00 -4.08470750e-01 7.08373845e-01 5.67203939e-01 -7.80547917e-01 -2.49394357e-01 -6.33971930e-01 2.70011276e-01 -1.87389329e-01 -6.12585768e-02 1.01781689e-01 3.82018864e-01 -9.11801755e-01 9.61890876e-01 -9.38272119e-01 -6.13825142e-01 4.06750262e-01 4.36755329e-01 -4.98593867e-01 -4.93569970e-02 -1.03648186e+00 1.00357258e+00 3.12283427e-01 -3.47754091e-01 -7.01301932e-01 -1.20049644e+00 -7.76865125e-01 1.61961988e-01 -3.67914587e-01 -1.35585046e+00 7.35488951e-01 -4.18972164e-01 -1.11891174e+00 6.87687814e-01 -9.89912376e-02 -7.25241840e-01 4.24408257e-01 -4.89349067e-01 -4.12709296e-01 2.65118927e-01 -1.67295739e-01 2.59667654e-02 3.22112411e-01 -6.16061449e-01 -6.03390753e-01 -4.02428418e-01 -4.35278356e-01 7.09198564e-02 -5.09266704e-02 2.63622552e-01 3.47217500e-01 -6.07277274e-01 -1.47441626e-01 -8.46765578e-01 -5.71958661e-01 -3.77795190e-01 -5.01740649e-02 7.82428682e-02 3.60435456e-01 -9.49345469e-01 1.55577159e+00 -2.10156131e+00 -2.29299635e-01 2.19356939e-02 4.70077634e-01 1.69494048e-01 1.55967489e-01 7.19428957e-01 -6.02306068e-01 2.72988588e-01 -1.75482273e-01 -1.26219705e-01 -4.62460369e-01 -3.85303013e-02 -2.39899635e-01 4.36868429e-01 3.85083944e-01 8.84845972e-01 -8.50642025e-01 -2.79275477e-01 6.51459694e-01 5.34016132e-01 -5.66368878e-01 8.41931283e-01 5.53858399e-01 6.35574937e-01 -9.63459089e-02 4.30300295e-01 -1.33977726e-01 -3.44811469e-01 2.01974347e-01 2.27005512e-01 -1.35263190e-01 3.06906164e-01 -5.38987458e-01 1.15042222e+00 -4.55252498e-01 3.88598710e-01 -4.16946769e-01 -8.46482992e-01 1.10843480e+00 7.32495725e-01 9.63906765e-01 -2.92140722e-01 2.29224980e-01 1.59976512e-01 2.30667546e-01 -5.33368051e-01 -5.64557552e-01 -6.52138531e-01 -1.58028677e-01 4.00767893e-01 -5.19383103e-02 6.80609867e-02 -2.85119981e-01 -6.34789281e-03 1.55277574e+00 -1.68854758e-01 7.34244585e-01 -4.18607831e-01 2.37425372e-01 1.50473654e-01 6.94945991e-01 9.01734650e-01 -3.03188831e-01 1.16609228e+00 7.49905646e-01 -9.28071678e-01 -9.92932618e-01 -1.24386120e+00 -6.48511410e-01 6.84132278e-01 -7.72990644e-01 -5.29986560e-01 -2.81844288e-01 -5.09762287e-01 1.55069858e-01 9.06503439e-01 -7.04279900e-01 -3.24182898e-01 -4.70243245e-01 -1.01957178e+00 6.76046431e-01 6.33995950e-01 -2.67079800e-01 -1.00516641e+00 -1.21083570e+00 8.21378708e-01 2.46914670e-01 -4.68270898e-01 1.12568527e-01 7.37056911e-01 -1.15256572e+00 -1.25194454e+00 -1.02235746e+00 -3.79437804e-01 2.86417902e-01 -4.22498494e-01 1.16422200e+00 -1.14881665e-01 -4.84138399e-01 4.38357055e-01 -1.68681875e-01 -6.97710514e-01 -3.64522338e-01 -1.26203954e-01 4.91290003e-01 -3.59092534e-01 5.46321630e-01 -6.27757668e-01 -1.06612027e+00 1.36743262e-01 -6.52730644e-01 -2.08272472e-01 5.64408481e-01 1.05112362e+00 4.24314290e-01 -4.27045614e-01 9.47140515e-01 -6.17681205e-01 3.75584900e-01 -9.61888790e-01 -1.46767423e-01 -9.06879082e-04 -1.21467733e+00 -2.52179772e-01 1.10808492e+00 -2.86070764e-01 -2.75107622e-01 -1.02016069e-01 1.74088687e-01 -4.12239432e-01 -5.27690649e-01 5.32015860e-01 4.06901717e-01 7.63519645e-01 5.90466022e-01 -1.80606693e-01 2.31302574e-01 -5.39652467e-01 -4.65974569e-01 6.99245811e-01 2.14923605e-01 -2.90441245e-01 2.84330279e-01 -1.46458689e-02 2.47645825e-01 -5.30953825e-01 -5.29451489e-01 -7.96016693e-01 -7.61870027e-01 3.38142961e-01 9.92645741e-01 -1.02763867e+00 -8.92367542e-01 8.84644687e-02 -7.45662093e-01 -5.37066936e-01 -6.28656447e-01 9.60001588e-01 -5.76446593e-01 4.38691974e-02 -6.89669430e-01 -6.77033842e-01 -7.43424118e-01 -8.36128771e-01 6.17826462e-01 -4.66574654e-02 -9.56250370e-01 -1.12678063e+00 4.97524977e-01 -1.30617037e-01 4.08133447e-01 8.47156048e-01 1.33532965e+00 -1.29938686e+00 1.45344287e-01 -6.44523382e-01 -2.42252827e-01 3.80429924e-01 5.42230844e-01 2.73816250e-02 -5.90849876e-01 -2.24044621e-01 -4.44480665e-02 3.40570137e-02 6.02636695e-01 5.55367529e-01 1.02316594e+00 -2.77930498e-01 -1.27947375e-01 8.16150844e-01 1.55341113e+00 5.14370680e-01 5.78072071e-01 4.35805649e-01 5.78328013e-01 5.32543957e-01 7.24597946e-02 7.23150253e-01 3.00288618e-01 5.35357222e-02 -5.08128814e-02 -1.45959988e-01 3.78515542e-01 -4.47795875e-02 2.32496426e-01 5.17071843e-01 3.46155576e-02 4.76901755e-02 -1.48942280e+00 5.66980302e-01 -1.52819753e+00 -6.21018648e-01 -4.13944721e-01 2.78363180e+00 6.79142654e-01 8.75571594e-02 2.11256191e-01 1.45790121e-03 2.87725180e-01 -3.17658156e-01 -4.27563936e-01 -9.79118884e-01 1.80234704e-02 4.62176442e-01 4.99843657e-01 2.93202519e-01 -8.24122787e-01 2.15319946e-01 6.42900085e+00 -4.24672097e-01 -1.04092467e+00 -2.67910630e-01 1.04030752e+00 -3.55233192e-01 5.57375252e-01 8.32426026e-02 -2.31315270e-01 4.40057188e-01 1.93803585e+00 -2.85887718e-01 2.29683489e-01 6.28444612e-01 6.59730732e-01 1.16877608e-01 -1.30946243e+00 8.67412329e-01 -3.20403397e-01 -1.12080467e+00 -2.92382956e-01 -2.30007470e-01 4.11027431e-01 2.87003338e-01 -3.47142190e-01 4.48802948e-01 2.53561795e-01 -1.46482515e+00 -1.95831247e-02 8.67573261e-01 1.34227359e+00 -6.70190632e-01 1.20812786e+00 9.33100060e-02 -7.20115006e-01 -4.44766134e-01 -5.55427223e-02 -3.26834381e-01 2.75549352e-01 7.08410740e-01 -1.15838277e+00 3.36484879e-01 8.37997735e-01 6.79062307e-01 -4.92730558e-01 9.85954523e-01 3.28344911e-01 9.30888951e-01 -4.71713096e-01 4.23984379e-01 9.09814164e-02 1.15426794e-01 3.47887695e-01 1.40543616e+00 3.92027259e-01 5.11931658e-01 -1.15444519e-01 3.18447322e-01 1.81810930e-01 2.78602630e-01 -7.88737535e-01 3.81320305e-02 1.35755733e-01 9.44218636e-01 -2.85503566e-01 -5.89469790e-01 -4.45337892e-01 7.01475739e-01 -1.09420195e-02 4.46714044e-01 -4.60702091e-01 -6.51582897e-01 4.39920038e-01 5.62739432e-01 -2.32499719e-01 9.91369933e-02 -8.38471949e-01 -1.08582067e+00 -4.52084154e-01 -7.35632300e-01 9.68858838e-01 -4.75963116e-01 -1.25049710e+00 5.73483527e-01 -2.17946365e-01 -1.39921153e+00 -4.81103837e-01 -6.41924381e-01 -7.49684691e-01 1.39532769e+00 -9.78919327e-01 -2.55885810e-01 -2.58220345e-01 1.85385570e-01 2.19976112e-01 -2.51029193e-01 1.61089897e+00 -7.31820567e-03 -6.46564543e-01 3.70571315e-01 2.89440483e-01 3.29736978e-01 1.05366039e+00 -1.21464753e+00 -1.56977892e-01 1.30499080e-01 -8.11552823e-01 1.06009686e+00 3.91409576e-01 -5.87999880e-01 -9.66279924e-01 -1.49108672e+00 1.05730700e+00 -1.05093062e+00 5.91632545e-01 3.12203228e-01 -9.89467800e-01 6.87664747e-01 -7.17170388e-02 -2.19469573e-02 1.39476824e+00 2.75103822e-02 -1.11421086e-01 3.04533504e-02 -1.19187331e+00 2.71579534e-01 4.77872819e-01 -3.21307033e-01 -9.24785078e-01 9.26728845e-02 4.86656785e-01 -2.25778315e-02 -1.67713249e+00 8.05969834e-01 8.83060992e-01 -8.33053231e-01 9.38422799e-01 -1.17538559e+00 4.96220648e-01 8.47627223e-03 -9.32555795e-02 -1.15952730e+00 -5.80378532e-01 -7.15822756e-01 -5.58630899e-02 7.04147398e-01 4.93949711e-01 -1.17464519e+00 2.59344608e-01 9.77811873e-01 -1.14351511e-01 -1.13554573e+00 -7.32924759e-01 -2.82625616e-01 3.26644510e-01 -1.78295538e-01 4.24749285e-01 1.02151394e+00 2.51171261e-01 2.32793912e-01 -1.46097979e-02 1.15280844e-01 3.38931203e-01 -1.76646948e-01 4.94548082e-01 -1.52339113e+00 -2.18275592e-01 -4.42494363e-01 -3.41958880e-01 -3.03329714e-02 -3.75371933e-01 -8.02187145e-01 -4.58114535e-01 -1.99115467e+00 3.02577466e-01 -5.42195201e-01 -1.19938791e+00 5.95262229e-01 -2.99586147e-01 -2.05399096e-01 -4.81840931e-02 3.95832449e-01 -1.11114765e-02 1.04986437e-01 2.97640055e-01 3.71762544e-01 -7.30409265e-01 -4.16096933e-02 -6.07449949e-01 6.26389802e-01 1.12980020e+00 -7.74778247e-01 -1.74225315e-01 -1.92224365e-02 -2.17884153e-01 6.76162541e-01 4.02872562e-01 -9.76306558e-01 -2.13032410e-01 -2.82699496e-01 8.51151347e-01 -3.38326305e-01 -1.32313013e-01 -9.07003760e-01 3.47267777e-01 8.37956667e-01 -3.85729641e-01 5.82934499e-01 4.88779187e-01 3.36201966e-01 -2.34463364e-02 2.48439774e-01 7.64026344e-01 2.18487671e-03 1.11803032e-01 3.04729998e-01 -5.93996823e-01 9.22236368e-02 9.43977833e-01 -1.80540115e-01 -1.09347478e-01 -1.18067712e-01 -1.01187217e+00 1.02965519e-01 4.83676791e-01 3.74532878e-01 4.50544536e-01 -9.09428298e-01 -1.00210500e+00 2.22873390e-01 3.07435513e-01 4.82814340e-03 1.77817777e-01 1.23482275e+00 -8.27161252e-01 5.68368852e-01 5.10521047e-03 -5.10430098e-01 -8.97537410e-01 6.23117566e-01 4.68244612e-01 -3.21921974e-01 -1.09348285e+00 3.40418339e-01 2.59123743e-01 -2.26303756e-01 1.09017864e-01 -3.90195042e-01 -4.20802049e-02 6.52699769e-02 6.30830586e-01 4.93679911e-01 -1.95141695e-02 -2.61912256e-01 -5.83943129e-01 2.66642213e-01 1.40464902e-01 1.75642297e-01 1.48257172e+00 8.58830288e-02 -2.04314157e-01 9.10119176e-01 1.55291641e+00 -2.52655536e-01 -8.29696953e-01 2.54568428e-01 1.99782290e-02 1.39246449e-01 -2.58240819e-01 -1.06193292e+00 -3.39377522e-01 9.70678329e-01 8.63187432e-01 4.58541662e-02 1.11962795e+00 -2.20537513e-01 6.21583581e-01 2.53225088e-01 4.50130068e-02 -6.05617166e-01 -4.41945672e-01 5.43767750e-01 6.03008091e-01 -8.92886400e-01 -2.05645591e-01 4.63035166e-01 -8.32664728e-01 1.35990310e+00 -6.55821189e-02 -2.15655938e-01 8.98204923e-01 6.72649816e-02 3.46249849e-01 -6.37618303e-02 -1.04752636e+00 5.25499821e-01 2.21909825e-02 3.35884035e-01 7.60593593e-01 4.07540709e-01 -3.09890896e-01 9.13013160e-01 -6.82736039e-02 3.15042049e-01 4.33854282e-01 8.93406212e-01 -1.88676924e-01 -8.67848933e-01 -2.65500218e-01 1.15486693e+00 -9.56006527e-01 -3.16777706e-01 -5.99433109e-02 5.56410909e-01 -7.83283636e-02 9.41368401e-01 1.91626027e-02 -1.80875674e-01 5.68499267e-01 9.20551002e-01 -1.47787958e-01 -7.55045831e-01 -7.71368504e-01 -1.63283139e-01 1.38754830e-01 -4.09218162e-01 2.17863336e-01 -6.81171656e-01 -1.43824637e+00 2.68395338e-02 2.91894257e-01 2.81900883e-01 2.19090268e-01 5.61913967e-01 8.23201239e-01 8.57450306e-01 6.86374962e-01 -3.74497235e-01 -5.29068708e-01 -1.00888264e+00 -4.93619591e-01 3.11640978e-01 7.99975753e-01 -2.32081354e-01 -7.24339068e-01 1.44851342e-01]

[8.014100074768066, 6.169409275054932]

ef772e52-4b54-444c-ac11-05cf06a46002

prores-exploring-degradation-aware-visual

2306.13653

https://arxiv.org/abs/2306.13653v1

https://arxiv.org/pdf/2306.13653v1.pdf

ProRes: Exploring Degradation-aware Visual Prompt for Universal Image Restoration

Image restoration aims to reconstruct degraded images, e.g., denoising or deblurring. Existing works focus on designing task-specific methods and there are inadequate attempts at universal methods. However, simply unifying multiple tasks into one universal architecture suffers from uncontrollable and undesired predictions. To address those issues, we explore prompt learning in universal architectures for image restoration tasks. In this paper, we present Degradation-aware Visual Prompts, which encode various types of image degradation, e.g., noise and blur, into unified visual prompts. These degradation-aware prompts provide control over image processing and allow weighted combinations for customized image restoration. We then leverage degradation-aware visual prompts to establish a controllable and universal model for image restoration, called ProRes, which is applicable to an extensive range of image restoration tasks. ProRes leverages the vanilla Vision Transformer (ViT) without any task-specific designs. Furthermore, the pre-trained ProRes can easily adapt to new tasks through efficient prompt tuning with only a few images. Without bells and whistles, ProRes achieves competitive performance compared to task-specific methods and experiments can demonstrate its ability for controllable restoration and adaptation for new tasks. The code and models will be released in \url{https://github.com/leonmakise/ProRes}.

['Lefei Zhang', 'Xinggang Wang', 'Qian Zhang', 'Guoli Wang', 'Tianheng Cheng', 'Jiaqi Ma']

2023-06-23

['deblurring', 'image-restoration']

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

[ 2.38105908e-01 -3.48325133e-01 1.45708367e-01 -2.92991042e-01 -7.52300680e-01 -4.53764081e-01 5.34158647e-01 -6.92100942e-01 -5.28340079e-02 4.49857712e-01 4.89753097e-01 -4.00411963e-01 6.61954805e-02 -1.57587767e-01 -8.11595380e-01 -7.93731034e-01 3.12470287e-01 -3.86477053e-01 1.30788192e-01 -2.61593282e-01 1.99362531e-01 3.58165681e-01 -1.40429735e+00 3.87930095e-01 1.05701411e+00 8.46135378e-01 8.34661365e-01 1.03190422e+00 3.17295372e-01 9.56679940e-01 -6.00385249e-01 -5.07544577e-02 3.62353027e-01 -2.76140630e-01 -5.61468005e-01 3.42581391e-01 7.45429456e-01 -8.11884046e-01 -6.31425440e-01 1.06195271e+00 7.42561460e-01 1.95621312e-01 4.62593883e-01 -1.12341106e+00 -1.58945072e+00 3.10520351e-01 -5.06932199e-01 4.41372454e-01 1.40782267e-01 8.88951242e-01 6.64394557e-01 -9.83808637e-01 3.75607878e-01 1.28135192e+00 5.28693497e-01 9.41411078e-01 -1.42028332e+00 -5.83116055e-01 4.53038454e-01 5.12320101e-01 -8.23144078e-01 -9.35996890e-01 7.07070708e-01 -4.73485380e-01 1.00847316e+00 3.10716271e-01 2.91852891e-01 1.50192928e+00 3.89797866e-01 8.97806048e-01 1.26987541e+00 -1.68857202e-01 1.41516834e-01 -2.38868654e-01 1.33201897e-01 1.80235118e-01 -2.08855271e-02 4.66173768e-01 -5.08175552e-01 4.55370665e-01 1.01003003e+00 6.93771988e-02 -7.95540333e-01 -8.13520327e-02 -1.24967110e+00 3.06401074e-01 6.19191468e-01 -7.93638602e-02 -3.59877437e-01 3.34394187e-01 3.27449560e-01 5.07717133e-01 4.53715026e-01 4.10346508e-01 -6.06662571e-01 1.97687238e-01 -6.21831119e-01 1.12551227e-01 1.48613468e-01 9.53089774e-01 6.42920494e-01 6.24054193e-01 -6.47349179e-01 1.09109175e+00 1.27449155e-01 4.38563734e-01 5.31303287e-01 -1.31680536e+00 2.16527417e-01 -9.09404829e-03 3.40391994e-01 -6.74555421e-01 -4.06099498e-01 -7.11291850e-01 -1.00293815e+00 5.29140234e-01 1.49301931e-01 -1.75268680e-01 -1.40384734e+00 1.87118065e+00 4.04553488e-02 3.52091521e-01 -3.38496566e-02 1.21258569e+00 8.42340767e-01 7.26892591e-01 9.69543904e-02 -1.69370592e-01 1.33824432e+00 -1.34358180e+00 -8.14897418e-01 -7.05249369e-01 -2.04409555e-01 -8.49692106e-01 1.67820418e+00 5.81941664e-01 -1.04514861e+00 -8.90406311e-01 -1.08356428e+00 -3.82816851e-01 1.57381266e-01 2.37275347e-01 5.41151166e-01 3.18278730e-01 -1.50998342e+00 4.75708544e-01 -9.31075573e-01 -2.82148123e-01 4.09505486e-01 1.27762288e-01 -1.08336091e-01 -3.82820040e-01 -9.62047935e-01 9.57634151e-01 -1.07995316e-01 2.81230867e-01 -1.47937453e+00 -7.69499481e-01 -8.01731110e-01 1.10995639e-02 2.77012765e-01 -1.08966529e+00 1.54916072e+00 -1.13575709e+00 -1.60364616e+00 5.71450830e-01 -1.42741084e-01 -3.04902107e-01 3.77738684e-01 -5.06870687e-01 -5.89154422e-01 6.56386688e-02 9.70644597e-03 7.23779738e-01 1.46523750e+00 -1.60985172e+00 -3.85757834e-01 -5.24313562e-03 1.82377920e-01 2.94419616e-01 -4.00188893e-01 1.07961170e-01 -4.94954318e-01 -9.46499705e-01 -8.85963738e-02 -5.72790921e-01 -2.71523923e-01 2.44943276e-01 -3.37681800e-01 1.51568785e-01 7.92082071e-01 -8.47154677e-01 9.69664097e-01 -2.24832463e+00 1.52817756e-01 -6.21343374e-01 3.64869267e-01 4.13571179e-01 -6.85514629e-01 1.62699848e-01 -2.23660246e-01 1.35953752e-02 -1.53166115e-01 -4.91335034e-01 4.41476963e-02 1.69504777e-01 -5.66716850e-01 3.74540538e-01 2.75316089e-01 9.44103837e-01 -8.73073757e-01 -3.27660479e-02 4.53360051e-01 7.08090961e-01 -5.30858219e-01 4.16023612e-01 -2.66965508e-01 6.26209378e-01 -1.47536442e-01 7.31584787e-01 7.87087142e-01 -3.02496463e-01 -4.61498685e-02 -6.01235807e-01 -1.77402925e-02 2.47295260e-01 -7.54340470e-01 1.60314310e+00 -6.40134394e-01 7.80711293e-01 2.91162580e-01 -9.08600211e-01 7.55104065e-01 2.81757981e-01 6.57545626e-02 -9.28610504e-01 -1.48535267e-01 4.44071107e-02 -1.96690276e-01 -6.70934081e-01 5.56825280e-01 5.37018441e-02 3.47236127e-01 2.76887804e-01 1.96318284e-01 -1.56716570e-01 -1.96372792e-02 9.89551544e-02 1.28861403e+00 3.21780175e-01 1.32544518e-01 -1.57750681e-01 1.80882379e-01 -1.92111596e-01 6.31500661e-01 8.65849793e-01 -4.94615287e-01 9.95888770e-01 8.24101642e-02 -3.57293189e-01 -9.17213976e-01 -1.30111742e+00 -2.58848052e-02 1.33856833e+00 3.26642334e-01 -1.61781698e-01 -6.12538576e-01 -3.15579385e-01 -1.81679681e-01 7.41434395e-01 -4.24005210e-01 -3.30248713e-01 -4.91173953e-01 -7.85872698e-01 8.32865760e-02 4.10751164e-01 5.73155522e-01 -1.16736341e+00 -4.05564576e-01 5.02049662e-02 -3.55063319e-01 -1.05875099e+00 -9.64337528e-01 8.33088905e-02 -9.01200891e-01 -8.73247266e-01 -9.00548577e-01 -7.79600620e-01 5.92738211e-01 9.59722281e-01 1.17358303e+00 -8.77812132e-03 -6.58691078e-02 6.58837259e-01 -2.61167347e-01 1.31092280e-01 -3.39395434e-01 -3.25605810e-01 1.25540808e-01 2.82317307e-02 -8.00215155e-02 -8.81384730e-01 -1.19709706e+00 4.13823187e-01 -1.08696508e+00 2.08592057e-01 6.47739768e-01 1.05108202e+00 4.96969134e-01 -2.67585069e-01 7.03592300e-01 -5.51211357e-01 9.21967089e-01 -3.31252158e-01 -4.07002717e-01 1.59363374e-01 -8.01572323e-01 -9.48966369e-02 7.68348753e-01 -7.72223353e-01 -1.32560420e+00 -1.26023069e-01 -1.10145435e-02 -6.57048941e-01 -1.80557609e-01 3.18260074e-01 -1.98223159e-01 -4.17401120e-02 9.86449301e-01 3.55941236e-01 -1.91673547e-01 -7.27112055e-01 6.23759866e-01 6.45352483e-01 1.09294665e+00 -5.21216929e-01 7.72108138e-01 2.07781747e-01 -5.34180462e-01 -5.88224590e-01 -7.89922655e-01 -1.27237335e-01 -1.10677607e-01 -2.94118643e-01 6.15741551e-01 -1.30196536e+00 -5.84196448e-01 8.74397516e-01 -1.12625277e+00 -9.98163044e-01 -1.43942967e-01 1.15995117e-01 -6.45026684e-01 4.81268704e-01 -9.67533469e-01 -4.60434020e-01 -4.53101307e-01 -1.31124628e+00 8.50028396e-01 4.25865710e-01 -4.23029698e-02 -8.38945270e-01 -2.62580842e-01 5.83240509e-01 8.63569081e-01 -2.05161542e-01 6.27403915e-01 6.02862053e-02 -8.02282214e-01 1.19312726e-01 -5.00566006e-01 8.65784466e-01 2.36450091e-01 -1.91966772e-01 -1.14932096e+00 -5.48807204e-01 2.66900539e-01 -4.34236258e-01 1.17386663e+00 5.91854513e-01 1.39757621e+00 -5.18503487e-01 -1.43535552e-03 9.95567381e-01 1.22330630e+00 7.04285130e-02 9.11507487e-01 5.51218688e-01 6.91167057e-01 1.23574726e-01 3.68420839e-01 2.61844546e-01 6.03927374e-01 7.82186508e-01 6.27971232e-01 -2.46811405e-01 -8.69513214e-01 -1.51267439e-01 8.37457120e-01 6.07904375e-01 -3.43383215e-02 -3.74903530e-01 -6.89728796e-01 5.26680291e-01 -1.68116641e+00 -7.42003739e-01 1.68251228e-02 1.88671267e+00 1.15088236e+00 -1.34660378e-01 -5.04345834e-01 -3.09206575e-01 6.44973993e-01 4.03570265e-01 -9.93953824e-01 -9.47612822e-02 -2.67981470e-01 1.24169877e-02 3.48327905e-01 7.46623337e-01 -1.01469791e+00 1.08347952e+00 6.53426218e+00 7.24240065e-01 -1.16178083e+00 3.03981364e-01 5.98281562e-01 -1.81441307e-01 -4.98871356e-01 6.01464063e-02 -4.56280738e-01 4.18405443e-01 5.22263467e-01 -2.63717417e-02 8.86549473e-01 5.83283484e-01 6.50278091e-01 -5.41135622e-03 -9.91643429e-01 1.07214558e+00 6.11025095e-02 -1.08902848e+00 5.60123548e-02 -3.39683712e-01 8.14689100e-01 2.19020605e-01 5.39234579e-01 2.68850178e-01 6.40419900e-01 -9.76471961e-01 8.72907937e-01 6.49343789e-01 9.83520389e-01 -6.84874579e-02 2.25318819e-01 9.19288769e-03 -6.08674526e-01 -3.31495106e-01 -3.93709093e-01 -6.48000538e-02 3.01434159e-01 6.36398673e-01 -2.55920231e-01 2.08625346e-01 9.26142454e-01 9.33962822e-01 -6.40419841e-01 1.00682473e+00 -6.46585345e-01 7.81931043e-01 2.03414768e-01 7.06249714e-01 -1.46569073e-01 3.50434855e-02 6.52464807e-01 1.12421656e+00 3.82090956e-01 3.29331495e-02 2.24001426e-02 7.60340095e-01 1.56854205e-02 -4.90956128e-01 -1.70660451e-01 3.70164931e-01 3.58972311e-01 1.23182309e+00 -1.22826360e-01 -1.93836555e-01 -3.37850928e-01 1.44590437e+00 2.48000607e-01 1.06208587e+00 -7.55746901e-01 -3.27254310e-02 1.07857978e+00 -5.88568412e-02 1.53289363e-01 -3.32984298e-01 -5.14565408e-01 -1.52692235e+00 3.57892774e-02 -1.22443938e+00 5.82468733e-02 -1.46175361e+00 -1.49891210e+00 7.72223294e-01 -2.39739284e-01 -1.12635481e+00 3.31747532e-02 -7.04725027e-01 -7.23267794e-01 9.66073871e-01 -1.73115802e+00 -1.16426599e+00 -5.75161397e-01 7.89089978e-01 9.01930630e-01 -7.24372193e-02 4.58562642e-01 3.24737281e-01 -8.02400649e-01 5.91716707e-01 5.39023764e-02 -2.04387650e-01 1.24722612e+00 -1.20493567e+00 6.98654771e-01 1.32496023e+00 -2.83137709e-01 6.28801703e-01 1.09176445e+00 -4.63419199e-01 -1.46658885e+00 -1.19565034e+00 4.35841262e-01 -3.77193421e-01 6.44223094e-01 -2.44432613e-01 -1.01178217e+00 8.06989312e-01 4.71070856e-01 1.77204892e-01 1.69129997e-01 -9.62611809e-02 -6.93174422e-01 -5.66944242e-01 -8.86610150e-01 1.03795850e+00 1.30897939e+00 -5.49549222e-01 -2.44277820e-01 5.02212584e-01 1.03553331e+00 -5.93284130e-01 -5.26063681e-01 2.89193481e-01 3.12766850e-01 -1.10694277e+00 1.36996317e+00 -4.48464751e-01 6.45883441e-01 -4.89558607e-01 -2.40381166e-01 -1.67685235e+00 -7.78953910e-01 -7.55237401e-01 -2.87753642e-01 1.13576448e+00 1.46431088e-01 -7.71883845e-01 5.66489138e-02 6.81289971e-01 -6.74980640e-01 -3.41874480e-01 -5.71664393e-01 -8.53311479e-01 -5.68007641e-02 -4.39162672e-01 4.15985316e-01 7.68108428e-01 -4.02039081e-01 3.69908422e-01 -9.10726011e-01 3.32583725e-01 7.39815116e-01 -9.60735306e-02 7.41776764e-01 -5.09748042e-01 -5.44413090e-01 -4.22311038e-01 7.50097483e-02 -1.52588952e+00 -8.31565410e-02 -6.09420478e-01 3.82784814e-01 -1.75618076e+00 2.44769961e-01 -2.96198964e-01 -4.54004705e-01 8.03375721e-01 -5.71956277e-01 3.56611669e-01 3.42454523e-01 4.70351458e-01 -5.94273806e-01 7.38097131e-01 1.64286184e+00 -3.85613650e-01 -1.81700632e-01 -4.84321229e-02 -1.24136770e+00 3.50381553e-01 1.00278986e+00 -4.79025245e-02 -6.39459431e-01 -1.14736974e+00 -8.61030221e-02 9.30641890e-02 8.12289119e-01 -8.17551315e-01 2.18196452e-01 -3.26545507e-01 4.09527898e-01 -2.26487592e-02 2.85710037e-01 -4.58350569e-01 1.95341974e-01 1.68187767e-01 -3.18885118e-01 -1.07662611e-01 2.79998899e-01 6.41609430e-01 -1.11278035e-01 8.13088343e-02 8.85069907e-01 -3.56976725e-02 -9.40318882e-01 3.97082865e-01 -4.73147810e-01 -1.77887917e-01 5.60626209e-01 -1.35104656e-01 -8.69159579e-01 -6.74644291e-01 -8.07271540e-01 3.02666783e-01 5.63857079e-01 6.49837017e-01 8.66403103e-01 -1.13693643e+00 -8.81234348e-01 -4.73765917e-02 -4.78474982e-03 -1.94561616e-01 8.05862367e-01 6.65754318e-01 -2.07453609e-01 -5.71545437e-02 -3.65393609e-01 -5.76109111e-01 -1.03541124e+00 8.37063253e-01 5.80436528e-01 6.20886981e-02 -8.58521163e-01 7.43946135e-01 6.10417962e-01 -2.03398451e-01 2.14404136e-01 -2.65461981e-01 -4.07328792e-02 -4.49095100e-01 7.96894133e-01 1.66339964e-01 -1.47013301e-02 -1.59132928e-01 -5.93108386e-02 3.19324493e-01 -2.27571398e-01 4.63979989e-02 1.35998666e+00 -6.63028598e-01 -1.01551272e-01 1.43240899e-01 7.76305437e-01 -3.57418418e-01 -2.20132971e+00 -3.28008503e-01 -4.04031783e-01 -6.26638114e-01 3.47523659e-01 -1.33714175e+00 -1.40619159e+00 8.85746360e-01 8.36559057e-01 -1.27604857e-01 1.81657803e+00 -1.34710312e-01 6.26133323e-01 4.61570658e-02 1.21052817e-01 -6.41039312e-01 4.46535230e-01 4.81309742e-01 1.46676779e+00 -1.24492967e+00 -1.24549404e-01 -1.70466602e-01 -7.77752757e-01 9.15371656e-01 8.77724707e-01 -1.93851050e-02 2.85612047e-01 2.30668202e-01 4.84491557e-01 5.83426692e-02 -1.02194166e+00 -1.41733497e-01 2.04914555e-01 1.09889889e+00 2.35571638e-01 1.36714913e-02 -1.77408233e-02 6.26626313e-01 4.86926660e-02 1.41234875e-01 6.74485207e-01 3.99853110e-01 -4.43972200e-01 -9.72472131e-01 -5.65467000e-01 3.41496825e-01 -2.34613076e-01 -4.72403139e-01 1.29793063e-01 1.75872445e-01 3.04012224e-02 1.21922171e+00 -2.53997624e-01 -4.63367850e-01 2.73895770e-01 -4.19632763e-01 4.90788847e-01 -4.28228110e-01 -4.54243571e-01 5.62251583e-02 -1.15446001e-01 -6.29774928e-01 -1.52694404e-01 -4.27707911e-01 -6.84650242e-01 -2.93231279e-01 6.33490756e-02 -5.28546274e-01 2.94686645e-01 7.01291859e-01 5.60315669e-01 8.15891147e-01 5.81796408e-01 -1.21940231e+00 -7.22905517e-01 -1.02994239e+00 -2.66847074e-01 4.07673985e-01 8.49442959e-01 -4.76426482e-01 -4.09936458e-01 4.07265007e-01]

[11.21603775024414, -2.267251491546631]

fb1e42af-38fb-469b-aee6-bbaddc08abc8

modeling-irregular-time-series-with

2111.11344

https://arxiv.org/abs/2111.11344v3

https://arxiv.org/pdf/2111.11344v3.pdf

Modeling Irregular Time Series with Continuous Recurrent Units

Recurrent neural networks (RNNs) are a popular choice for modeling sequential data. Modern RNN architectures assume constant time-intervals between observations. However, in many datasets (e.g. medical records) observation times are irregular and can carry important information. To address this challenge, we propose continuous recurrent units (CRUs) -- a neural architecture that can naturally handle irregular intervals between observations. The CRU assumes a hidden state, which evolves according to a linear stochastic differential equation and is integrated into an encoder-decoder framework. The recursive computations of the CRU can be derived using the continuous-discrete Kalman filter and are in closed form. The resulting recurrent architecture has temporal continuity between hidden states and a gating mechanism that can optimally integrate noisy observations. We derive an efficient parameterization scheme for the CRU that leads to a fast implementation f-CRU. We empirically study the CRU on a number of challenging datasets and find that it can interpolate irregular time series better than methods based on neural ordinary differential equations.

['Maja Rudolph', 'Stefan Lessmann', 'Mazin Eltayeb', 'Mona Schirmer']

2021-11-22

['irregular-time-series']

['time-series']

[ 1.25253901e-01 -1.53388269e-02 -2.94768397e-04 -3.83143872e-01 -6.18789017e-01 -1.22369312e-01 3.99198234e-01 -1.01870336e-01 -4.33173686e-01 6.13801658e-01 2.73196578e-01 -5.79935133e-01 8.93785283e-02 -5.47051072e-01 -7.28915155e-01 -6.85618401e-01 -2.42090523e-01 2.33843446e-01 -7.43333325e-02 -9.18387249e-02 -2.80079424e-01 3.41610730e-01 -1.21465051e+00 1.02129407e-01 6.56720042e-01 1.00084603e+00 3.16580743e-01 8.54381800e-01 7.16428459e-02 1.25103307e+00 -4.78895932e-01 1.06195718e-01 1.77226692e-01 -6.80680156e-01 -4.95119125e-01 -1.42503351e-01 -4.15209174e-01 -3.33600879e-01 -6.88363314e-01 7.54528522e-01 4.52412695e-01 4.43904221e-01 4.75294918e-01 -8.13661337e-01 -6.34855866e-01 6.06349230e-01 4.39098477e-02 3.62650633e-01 -6.53132284e-03 1.75526515e-02 7.70863831e-01 -4.78950262e-01 6.32241309e-01 1.11382210e+00 1.21333802e+00 7.09055305e-01 -1.52269661e+00 -2.71556467e-01 3.44594181e-01 1.71234943e-02 -1.25745785e+00 -4.73560214e-01 5.74572563e-01 -2.00301588e-01 1.27951932e+00 1.10372819e-01 7.04881966e-01 1.41427231e+00 7.26361334e-01 1.00038910e+00 5.96523225e-01 -7.82043412e-02 4.82042193e-01 -3.38951766e-01 2.85849243e-01 3.56346101e-01 -1.89248994e-01 1.24734975e-01 -3.93896788e-01 -1.41553476e-01 1.11094844e+00 7.77626634e-01 -2.26016909e-01 2.72344381e-01 -1.13082790e+00 7.35681951e-01 4.38329577e-01 1.27520293e-01 -8.49181056e-01 5.29072762e-01 6.62549078e-01 6.61411941e-01 4.53026623e-01 1.90135881e-01 -5.47597229e-01 -3.57334018e-01 -1.07036901e+00 1.85448468e-01 9.67556655e-01 7.81586587e-01 4.14845258e-01 3.66269618e-01 -1.33080065e-01 5.37988484e-01 1.73570856e-01 3.33192736e-01 9.81294334e-01 -8.99510026e-01 4.12306422e-03 2.56422043e-01 2.75172312e-02 -6.71313822e-01 -6.44129813e-01 -6.75598681e-01 -1.53650606e+00 -2.08504260e-01 1.80776224e-01 -5.03291488e-01 -1.06469858e+00 1.86201680e+00 1.17186353e-01 8.75894129e-01 3.28302264e-01 5.63235223e-01 5.16189098e-01 9.59432662e-01 -3.00419152e-01 -6.52553439e-01 1.03229916e+00 -7.19716191e-01 -1.27744079e+00 -7.41246566e-02 6.58530593e-01 -2.14247778e-01 5.31046450e-01 2.67463028e-01 -1.23629689e+00 -5.70349693e-01 -7.04963326e-01 -2.90102988e-01 -1.53745219e-01 1.95218399e-01 4.42923009e-01 -7.31053874e-02 -1.17569458e+00 9.94325936e-01 -1.77189040e+00 5.56074493e-02 -1.53219894e-01 3.73598158e-01 2.98726056e-02 5.62575936e-01 -1.47685027e+00 7.45915532e-01 2.24319205e-01 7.26676464e-01 -7.62223661e-01 -6.57792568e-01 -1.03803194e+00 2.72410493e-02 1.75467268e-01 -7.62247443e-01 1.77294552e+00 -6.63500428e-01 -1.88193405e+00 1.40196443e-01 -7.37345517e-01 -1.25066698e+00 3.43094379e-01 -3.25547248e-01 -5.83508849e-01 -1.70007259e-01 -1.30618230e-01 1.31661907e-01 8.67350042e-01 -4.57616389e-01 -3.27153981e-01 -8.81474407e-04 -4.48005438e-01 -9.38379541e-02 1.21035591e-01 -1.49832517e-01 -4.26257610e-01 -9.10464942e-01 3.20326388e-01 -1.15625250e+00 -8.80149364e-01 -4.08525497e-01 -4.12268490e-01 -1.86982989e-01 6.83646202e-01 -7.36045420e-01 1.70299017e+00 -2.04914165e+00 2.02764332e-01 -2.35015620e-02 8.91591907e-02 1.98690772e-01 3.18723232e-01 4.75560129e-01 -9.90381688e-02 -3.06986749e-01 -3.28333229e-01 -9.34258163e-01 -1.28811821e-01 7.40177453e-01 -7.70728350e-01 3.82333010e-01 2.41564438e-01 1.28983104e+00 -7.45729744e-01 7.32766017e-02 -6.19307102e-05 8.81352603e-01 -3.17040980e-01 3.68392795e-01 -4.18426335e-01 4.33695704e-01 -2.91476071e-01 1.57440290e-01 2.14721143e-01 -6.42062366e-01 1.06966719e-01 1.78941980e-01 -1.94846854e-01 7.36019969e-01 -1.11922371e+00 1.69571328e+00 -6.28418505e-01 8.57355237e-01 -1.69144779e-01 -9.29859519e-01 9.08932209e-01 6.88387871e-01 3.68453890e-01 -5.12981296e-01 2.11882859e-01 1.63596123e-01 -1.51739433e-01 -2.90308863e-01 6.12469554e-01 -1.88001528e-01 -7.45090097e-02 6.24614000e-01 -6.51023304e-03 2.65028656e-01 -6.81092963e-02 -6.71735480e-02 1.32253957e+00 6.48259884e-03 1.69176117e-01 1.28365085e-01 2.00910613e-01 -5.43715835e-01 1.09051466e+00 8.73148680e-01 1.13639064e-01 7.56687284e-01 5.62649846e-01 -8.15454900e-01 -8.93093646e-01 -7.24422753e-01 3.24995965e-02 6.79218948e-01 -3.03971678e-01 -4.29366052e-01 -4.86525506e-01 -3.08301765e-02 -2.20922664e-01 6.07587218e-01 -8.02202463e-01 -2.66595155e-01 -6.71492040e-01 -6.58088923e-01 6.24030173e-01 9.44852173e-01 1.69260204e-01 -1.28326654e+00 -1.01018977e+00 9.26679075e-01 -8.85598660e-02 -1.02036583e+00 -5.97576022e-01 8.11040640e-01 -1.09761834e+00 -3.53907317e-01 -7.62199461e-01 -5.91757119e-01 4.62044477e-01 -1.75674513e-01 9.50159311e-01 -1.84053048e-01 -8.72934330e-03 6.36879727e-03 -5.99284470e-02 -3.93381596e-01 -5.55829108e-01 2.49476612e-01 2.96476543e-01 1.01195455e-01 2.39778683e-01 -8.40282083e-01 -4.47453022e-01 1.72124803e-02 -1.14554679e+00 1.78494528e-01 3.87159556e-01 1.03917027e+00 8.33852768e-01 -4.88043912e-02 5.88761568e-01 -8.71967554e-01 7.33990133e-01 -5.29129207e-01 -7.17477441e-01 9.07857046e-02 -3.57625276e-01 5.60152829e-01 8.43661726e-01 -6.86507165e-01 -9.42772567e-01 1.21940255e-01 -3.99392575e-01 -7.89691985e-01 7.12821111e-02 8.15494657e-01 4.87175912e-01 5.66873014e-01 3.92979473e-01 5.05597889e-01 1.24576494e-01 -6.35330021e-01 1.01679303e-01 4.52309489e-01 8.09804380e-01 -2.74468541e-01 5.79330400e-02 4.49683338e-01 -5.72784878e-02 -7.98648059e-01 -8.03023756e-01 -3.74235690e-01 -3.50766361e-01 2.24780664e-01 6.30810618e-01 -1.06557846e+00 -9.52547848e-01 5.14991522e-01 -1.36188400e+00 -7.74370313e-01 -6.10264778e-01 7.23697126e-01 -6.93808019e-01 2.96769682e-02 -1.36116171e+00 -1.11354828e+00 -4.03939456e-01 -1.04580665e+00 1.00616944e+00 2.49805287e-01 -4.42253679e-01 -1.37133515e+00 3.38445783e-01 -7.23442852e-01 5.17135441e-01 3.54567051e-01 3.78368288e-01 -4.37416166e-01 -3.78134817e-01 -3.25865567e-01 2.52771884e-01 4.33999270e-01 2.29697991e-02 7.25283697e-02 -8.38875830e-01 -1.56385362e-01 5.66184759e-01 1.76287182e-02 1.15385234e+00 8.26521397e-01 1.08417785e+00 -3.58239770e-01 -2.35573605e-01 8.25985253e-01 1.19753492e+00 4.14745212e-01 6.10660613e-01 -1.09462261e-01 4.76342589e-01 5.10418154e-02 -7.48524815e-02 5.42222261e-01 5.32713175e-01 3.08308572e-01 1.65305927e-01 8.22381228e-02 3.28909725e-01 -2.43589759e-01 7.21242487e-01 1.47727942e+00 -1.47774220e-01 -6.38635308e-02 -8.57946157e-01 6.76870465e-01 -2.28603816e+00 -1.05997217e+00 -1.78085595e-01 2.01993108e+00 1.03138089e+00 2.28960603e-01 -1.68046862e-01 7.78097063e-02 3.94167393e-01 1.37127668e-01 -9.64130759e-01 -5.36172807e-01 -1.40622377e-01 2.14188233e-01 4.64419186e-01 4.82096404e-01 -9.46622908e-01 6.69963241e-01 6.96809053e+00 2.52722949e-01 -1.37336850e+00 4.90412787e-02 4.59131867e-01 -1.04241781e-01 -7.07060099e-02 -2.10407540e-01 -1.00267565e+00 6.50251865e-01 1.86111856e+00 -1.02544241e-01 2.78558075e-01 6.21827781e-01 5.60895741e-01 2.23864108e-01 -1.23598909e+00 1.03341210e+00 -4.17338818e-01 -1.52063382e+00 -3.04035872e-01 -7.44648790e-03 8.12553525e-01 4.31856692e-01 1.55962884e-01 3.77048105e-01 6.44674897e-01 -1.11947644e+00 5.13433933e-01 1.00856614e+00 4.73411381e-01 -5.92959642e-01 6.44458830e-01 6.42802238e-01 -1.22812068e+00 -1.21444777e-01 -5.00239789e-01 -3.91027719e-01 6.22465432e-01 8.24767292e-01 -4.75617528e-01 4.58098233e-01 5.49230754e-01 1.25831652e+00 4.24984619e-02 7.81875312e-01 -1.40521586e-01 8.03532779e-01 -5.98581791e-01 1.66216418e-01 3.54241222e-01 -2.42959052e-01 3.60300779e-01 1.19548666e+00 4.23580259e-01 -2.48054918e-02 -1.11557215e-01 8.32594991e-01 1.02787301e-01 -4.94855553e-01 -5.48265934e-01 -1.02600440e-01 1.94232166e-01 7.20537782e-01 -5.57579696e-01 -4.37424451e-01 -4.18906748e-01 1.15761542e+00 3.64473403e-01 7.41161466e-01 -7.87674785e-01 -3.15902591e-01 8.89951646e-01 -2.92706966e-01 5.71137547e-01 -5.32014966e-01 -1.44938856e-01 -1.51192343e+00 1.79757610e-01 -6.80843234e-01 2.97777474e-01 -6.03250861e-01 -1.14986944e+00 9.04223859e-01 -4.89465326e-01 -1.25912344e+00 -9.79144633e-01 -3.00020963e-01 -6.22263908e-01 9.96766508e-01 -1.23692286e+00 -6.48282051e-01 2.24210024e-01 5.91915965e-01 6.01775527e-01 2.00012162e-01 1.00322413e+00 1.73942909e-01 -8.44440877e-01 3.79444182e-01 5.44228733e-01 1.76128551e-01 1.53911918e-01 -1.18320084e+00 1.01262116e+00 9.03618217e-01 -7.94052333e-03 1.01259255e+00 7.73047447e-01 -7.21242368e-01 -1.40019989e+00 -1.37624168e+00 1.09698284e+00 -2.61735499e-01 7.74918139e-01 -3.55721325e-01 -1.50501561e+00 1.22858906e+00 8.22927654e-02 2.15514630e-01 5.32274783e-01 -9.10878181e-02 -9.03341398e-02 1.65127546e-01 -5.23990870e-01 5.41399240e-01 8.19920421e-01 -9.50729847e-01 -5.88996172e-01 8.12328979e-02 9.59871054e-01 -8.50293219e-01 -8.36773932e-01 1.17420547e-01 5.79904974e-01 -8.52123380e-01 6.89948559e-01 -7.55550265e-01 3.40433985e-01 -2.10963041e-01 1.42742038e-01 -1.51908493e+00 -1.03238024e-01 -1.32650149e+00 -9.29969728e-01 5.39007604e-01 3.33319515e-01 -8.30880821e-01 5.42127311e-01 8.47975314e-01 -2.27123529e-01 -9.56811011e-01 -8.65860224e-01 -8.80164385e-01 -2.20473155e-01 -7.79320419e-01 5.50121784e-01 6.48358405e-01 -1.89010128e-01 1.29863933e-01 -6.75823748e-01 2.94408798e-01 1.95959806e-01 -1.67134535e-02 3.05760443e-01 -1.16510677e+00 -4.23233122e-01 -1.81085914e-01 -3.08464170e-01 -1.72866011e+00 3.13206673e-01 -5.41465819e-01 2.83201903e-01 -1.39203179e+00 -3.61665100e-01 -1.27663195e-01 -4.99659061e-01 4.18083698e-01 -1.40209481e-01 -1.67905927e-01 -1.80551156e-01 2.83395976e-01 -5.67176402e-01 7.26075590e-01 8.63294005e-01 3.70350689e-01 -6.86250746e-01 3.87531757e-01 -2.13996097e-01 6.70582056e-01 7.38992691e-01 -4.19458568e-01 -3.55210871e-01 -4.15671051e-01 4.38795924e-01 6.36906028e-01 3.53219420e-01 -9.51892614e-01 6.73505664e-01 1.13157779e-01 1.54672772e-01 -8.82604659e-01 4.71523643e-01 -6.29103541e-01 6.00474477e-01 5.76966584e-01 -5.71152508e-01 4.43199605e-01 6.80033863e-02 8.13538373e-01 -4.47028965e-01 1.95665538e-01 4.83228773e-01 -6.87589720e-02 -3.13041568e-01 3.85268450e-01 -7.54049480e-01 -8.95246714e-02 4.94564921e-01 8.17163959e-02 3.29096615e-01 -7.85432756e-01 -1.13846207e+00 2.18432218e-01 1.02046728e-01 2.54482597e-01 4.97525275e-01 -1.48763871e+00 -4.36574042e-01 5.41298330e-01 -4.79829252e-01 4.51944888e-01 3.45503837e-01 1.09370446e+00 -2.61399746e-01 5.69443524e-01 2.89874643e-01 -6.13614440e-01 -6.78369343e-01 4.40715343e-01 5.90527654e-01 -6.27594292e-01 -1.01874971e+00 7.11004019e-01 -9.96151287e-03 -2.88016528e-01 3.48146588e-01 -1.17401803e+00 7.91872144e-02 9.29374993e-02 8.32421422e-01 2.64551014e-01 1.18466459e-01 -4.95438457e-01 7.88854342e-03 1.27517059e-01 -1.25299618e-01 -2.12448135e-01 1.60964012e+00 -2.31057510e-01 -8.53227228e-02 1.34155345e+00 1.38718724e+00 -7.37973809e-01 -1.62464929e+00 -7.00611711e-01 1.09293655e-01 3.33161384e-01 6.55149445e-02 -1.00922912e-01 -1.04020166e+00 9.74267840e-01 2.82036334e-01 6.66733861e-01 1.06882763e+00 -3.73359412e-01 1.31342185e+00 5.43990672e-01 -1.96773168e-02 -1.04440463e+00 -4.39986765e-01 1.16863465e+00 6.50322199e-01 -9.44714904e-01 -4.62716311e-01 2.90820509e-01 -4.85302478e-01 1.19836557e+00 -1.16661467e-01 -3.58019888e-01 1.01356804e+00 6.40329361e-01 1.78104281e-01 7.31064454e-02 -1.50109243e+00 -1.85279958e-02 1.71532661e-01 1.40827268e-01 5.50781846e-01 5.64831495e-02 1.43333107e-01 7.44756460e-01 -1.85917363e-01 3.47536296e-01 5.25093555e-01 1.00404942e+00 1.09244637e-01 -7.50004888e-01 -1.10823020e-01 4.33179587e-01 -5.89226484e-01 -8.76929238e-02 3.04782569e-01 2.17622727e-01 -4.63871926e-01 6.16331339e-01 4.88778502e-01 -1.71694055e-01 1.50323480e-01 2.99463004e-01 -6.67065978e-02 -4.39539373e-01 -6.87917352e-01 3.64175737e-01 -3.85893792e-01 -7.52292156e-01 -2.49219567e-01 -8.13124180e-01 -1.51402652e+00 -2.37250775e-01 -2.05488563e-01 1.52241111e-01 5.58610201e-01 9.64017630e-01 6.25478208e-01 1.00206411e+00 4.48782384e-01 -9.25050676e-01 -7.75301218e-01 -8.68505836e-01 -6.85392141e-01 1.50317162e-01 1.17043257e+00 -2.39974186e-01 -3.28103662e-01 2.29786769e-01]

[7.168252468109131, 3.352914571762085]

08802fa6-ffe0-47b2-9702-467a7702ef4c

study-on-unsupervised-statistical-machine

https://aclanthology.org/R19-1068

https://aclanthology.org/R19-1068.pdf

Study on Unsupervised Statistical Machine Translation for Backtranslation

Machine Translation systems have drastically improved over the years for several language pairs. Monolingual data is often used to generate synthetic sentences to augment the training data which has shown to improve the performance of machine translation models. In our paper, we make use of an Unsupervised Statistical Machine Translation (USMT) to generate synthetic sentences. Our study compares the performance improvements in Neural Machine Translation model when using synthetic sentences from supervised and unsupervised Machine Translation models. Our approach of using USMT for backtranslation shows promise in low resource conditions and achieves an improvement of 3.2 BLEU score over the Neural Machine Translation model.

['Mydhili K. Nair', 'Ch', 'Anush Kumar', 'Nihal V. Nayak', 'Aditya ra']

2019-09-01

ranlp-2019-9

['unsupervised-machine-translation']

['natural-language-processing']

[ 6.01551473e-01 2.95545101e-01 -2.86725789e-01 -4.43405420e-01 -1.38461804e+00 -5.18893719e-01 1.17292166e+00 -4.32895720e-01 -4.35454696e-01 1.69275248e+00 4.42391813e-01 -8.18511963e-01 8.29994261e-01 -5.56122243e-01 -9.07375395e-01 -2.13373721e-01 9.14881706e-01 1.10205019e+00 -3.57373595e-01 -6.62555873e-01 9.38479304e-02 1.83957621e-01 -7.52709687e-01 6.80127978e-01 1.18270922e+00 5.96275888e-02 2.54235119e-01 7.43236840e-01 -3.74420285e-01 2.67845899e-01 -7.84464896e-01 -7.53210723e-01 6.48175955e-01 -1.07956946e+00 -9.12642300e-01 -1.38004243e-01 4.05105680e-01 -1.33663446e-01 -8.76625255e-02 8.21625531e-01 7.72225976e-01 -3.23110163e-01 7.11213648e-01 -8.91778111e-01 -9.27325904e-01 9.82431650e-01 -3.49972963e-01 2.05839723e-01 4.17771667e-01 2.58847147e-01 7.28563428e-01 -1.24710858e+00 9.80517268e-01 1.24031401e+00 5.67334354e-01 7.47611403e-01 -1.44693530e+00 -5.24313927e-01 -9.47049499e-01 -2.60507852e-01 -1.11789513e+00 -6.64884865e-01 4.66794848e-01 -6.29712939e-02 1.49052620e+00 1.12592041e-01 3.24095368e-01 1.47650492e+00 5.92028737e-01 7.06467450e-01 1.74079025e+00 -1.14075291e+00 -1.68584093e-01 5.27791739e-01 -3.79828990e-01 3.04349124e-01 1.51799336e-01 3.31739396e-01 -4.17575538e-01 -5.92947789e-02 7.31279194e-01 -6.15706325e-01 4.39152807e-01 4.43246692e-01 -1.65432930e+00 8.94617200e-01 -1.72653235e-02 5.38727582e-01 -4.39071059e-01 4.78974879e-02 4.20638323e-01 9.39918995e-01 8.44754815e-01 9.60609794e-01 -5.33481061e-01 -3.75861198e-01 -1.07010913e+00 1.62349045e-01 7.51874447e-01 1.17701018e+00 5.93339086e-01 3.97138149e-01 -3.57955426e-01 1.28120720e+00 -2.74087459e-01 9.98777151e-01 9.23995972e-01 -6.47061229e-01 1.02483273e+00 3.54850560e-01 1.36728719e-01 -3.88458937e-01 2.58658469e-01 -2.80742675e-01 -6.17223799e-01 -2.42663905e-01 2.39271507e-01 -5.91108739e-01 -1.06098986e+00 1.47089946e+00 -1.09594516e-01 -4.66547281e-01 6.25041604e-01 5.34029603e-01 3.45429897e-01 9.40288365e-01 -1.02794930e-01 -4.05736566e-01 6.55429125e-01 -1.25146234e+00 -7.17996061e-01 -2.28671119e-01 1.04142964e+00 -1.44155920e+00 1.26498258e+00 -5.56652881e-02 -1.37834513e+00 -5.86448491e-01 -8.97000432e-01 1.71951577e-02 -3.61606777e-01 2.52338767e-01 4.03254300e-01 6.36962056e-01 -1.24179077e+00 7.37595320e-01 -6.02175832e-01 -7.04480410e-01 9.25035328e-02 5.78112125e-01 -4.95313257e-01 6.80684252e-03 -1.35834062e+00 1.40913427e+00 4.26594973e-01 -2.13605657e-01 -5.27603507e-01 -2.19195820e-02 -5.98711431e-01 -4.10936773e-01 -3.23403239e-01 -8.86296272e-01 1.67833781e+00 -1.45960820e+00 -1.95353472e+00 9.16389108e-01 -4.56028491e-01 -6.92118168e-01 7.18556464e-01 -1.33361638e-01 -3.41849476e-01 -1.95438460e-01 3.56221169e-01 9.03972387e-01 4.85119313e-01 -7.52370059e-01 -3.64145577e-01 3.08671650e-02 -4.36466157e-01 4.32436764e-01 -3.99812013e-02 6.48152113e-01 1.42269984e-01 -8.49065602e-01 -2.65400618e-01 -1.19095314e+00 -4.09566462e-01 -7.98256159e-01 -2.67090291e-01 -2.10911948e-02 4.01277691e-01 -9.17277694e-01 7.62963295e-01 -1.36712837e+00 2.03867301e-01 -3.02473962e-01 -5.85547507e-01 3.41773331e-01 -4.97450322e-01 9.26150858e-01 3.19813890e-03 2.05280334e-01 -4.76909518e-01 -3.66149783e-01 -2.28720218e-01 4.41516578e-01 -4.24066037e-01 -1.62812129e-01 5.18262804e-01 1.49454463e+00 -8.22777927e-01 -5.46112120e-01 -2.68935084e-01 6.95632100e-02 -6.32802099e-02 1.66763991e-01 -1.98769614e-01 6.26901746e-01 -3.03577274e-01 4.20368224e-01 3.82408500e-01 1.91321358e-01 4.00756635e-02 5.88714838e-01 3.60175818e-02 8.19646358e-01 -1.62779719e-01 1.87102997e+00 -5.62818050e-01 1.01856983e+00 -6.63648903e-01 -6.35856450e-01 1.20638084e+00 6.76503658e-01 -9.23118442e-02 -8.37758958e-01 1.64122224e-01 9.29571629e-01 3.40786725e-01 -3.40138048e-01 5.55205047e-01 -4.66717005e-01 6.71504885e-02 9.82821882e-01 1.17371611e-01 -6.57798767e-01 2.63316393e-01 -5.03983125e-02 6.92058086e-01 4.58389670e-01 2.12195352e-01 -4.12799627e-01 4.72849280e-01 6.61799848e-01 2.25270823e-01 5.12232661e-01 6.91786706e-02 6.83415174e-01 1.70307960e-02 -5.30635357e-01 -1.89442968e+00 -9.89066362e-01 2.42840052e-01 7.14940906e-01 -6.12374425e-01 3.48244868e-02 -1.12532592e+00 -6.94453061e-01 -5.27452648e-01 1.05297208e+00 -2.16110885e-01 -7.43861198e-02 -1.08086562e+00 -1.07704639e+00 1.07339358e+00 3.62209141e-01 4.63481665e-01 -1.24480796e+00 1.68420047e-01 4.14130479e-01 -5.51478148e-01 -1.10198307e+00 -5.59968829e-01 -1.29834577e-01 -1.31454504e+00 -2.24856168e-01 -9.86526430e-01 -1.03009069e+00 6.86552346e-01 1.46424860e-01 1.31957912e+00 -2.94884413e-01 1.94305167e-01 -3.98523748e-01 -2.65868396e-01 -7.14785039e-01 -1.34753597e+00 5.65423310e-01 5.74298799e-01 -3.52301180e-01 7.36897767e-01 -6.18189991e-01 -7.34049305e-02 3.15131336e-01 -7.63113618e-01 5.80025256e-01 9.43568766e-01 1.03050256e+00 2.66544044e-01 -8.65535557e-01 9.17887449e-01 -8.40927601e-01 1.23982680e+00 -3.00069034e-01 -1.35042682e-01 2.29525045e-01 -9.54159856e-01 4.36884493e-01 7.68119812e-01 -5.68552792e-01 -1.02557945e+00 -7.24400356e-02 1.97740085e-02 5.37175871e-02 -7.42240995e-02 3.19844782e-01 3.21514517e-01 8.51793587e-02 1.09664190e+00 6.53411925e-01 1.86834559e-01 -3.75283450e-01 2.15767443e-01 1.32999754e+00 2.46273234e-01 -6.48899376e-01 9.52940822e-01 -2.77016193e-01 -8.13718513e-02 -3.99360478e-01 -3.52237523e-01 6.95724338e-02 -1.01419997e+00 3.02985534e-02 4.40390378e-01 -8.66319239e-01 6.69680357e-01 3.31266135e-01 -1.41365325e+00 -4.80514020e-01 -9.32243690e-02 7.52421498e-01 -8.04826677e-01 -2.45017745e-02 -9.13549840e-01 -3.85006487e-01 -8.07536542e-01 -1.32807922e+00 1.08057213e+00 -7.35388175e-02 -6.20550990e-01 -9.12690103e-01 5.05909979e-01 5.16492426e-01 5.57209253e-01 1.04594104e-01 6.27400875e-01 -8.49967659e-01 -2.83732444e-01 -2.15013474e-01 4.38693278e-02 4.86940920e-01 4.50419486e-01 -2.06502065e-01 -4.93678898e-01 -1.42279238e-01 -8.33082050e-02 -3.99633288e-01 2.55447090e-01 -6.46275282e-02 1.63894556e-02 -4.45267469e-01 -6.83443025e-02 2.23129615e-01 1.06089222e+00 1.86776772e-01 7.06028283e-01 4.19772238e-01 4.09265339e-01 5.39392769e-01 5.63815236e-01 -4.86635864e-01 -3.99344563e-02 6.85022056e-01 -6.58676147e-01 -1.72310457e-01 -2.43474901e-01 -5.23912191e-01 8.10549736e-01 1.43938613e+00 -2.35281333e-01 3.44195440e-02 -1.03666234e+00 4.82334077e-01 -1.67111981e+00 -8.31461787e-01 -3.24916959e-01 1.99305284e+00 1.38256323e+00 7.96385407e-02 -2.11728550e-02 -2.34454885e-01 8.24742734e-01 -2.40069255e-01 -8.49707946e-02 -1.37329423e+00 -3.55522007e-01 6.19210660e-01 5.50721407e-01 7.38013625e-01 -4.84791726e-01 1.49583948e+00 7.31394243e+00 7.31360197e-01 -1.23558021e+00 3.16797644e-01 7.93670654e-01 5.28026894e-02 -2.40353063e-01 -2.72874366e-02 -6.18363917e-01 5.11783361e-01 1.64049888e+00 -4.25357878e-01 6.21948779e-01 3.77669722e-01 6.44989908e-01 2.19217837e-01 -1.23205340e+00 6.19644284e-01 2.14853376e-01 -1.23040199e+00 5.67253232e-01 5.16858436e-02 1.45642674e+00 3.81257176e-01 -6.73597604e-02 4.60890919e-01 4.89919513e-01 -1.07421017e+00 2.94991165e-01 1.40409097e-01 1.12456489e+00 -6.90336108e-01 8.94478023e-01 6.54361665e-01 -3.11327517e-01 5.81971347e-01 -5.81519783e-01 -3.38112324e-01 1.25385091e-01 1.16357878e-01 -1.73691320e+00 4.82756287e-01 -6.47081509e-02 3.08175832e-01 -4.98723269e-01 4.58055019e-01 -3.54914814e-01 9.01845574e-01 -2.81443477e-01 -2.45298803e-01 6.07240915e-01 -5.48713803e-01 5.05043626e-01 1.30278814e+00 5.64927638e-01 -3.11753720e-01 -2.33824149e-01 6.55715466e-01 -3.70165020e-01 5.83204150e-01 -1.04482770e+00 -5.56031227e-01 1.00053288e-01 6.87984765e-01 -3.59408468e-01 -6.83175147e-01 -2.33591512e-01 1.57916963e+00 2.17766494e-01 3.22722346e-01 -5.44270575e-01 -3.10190946e-01 1.76595688e-01 -1.22896461e-02 -3.60677004e-01 -3.84469956e-01 -6.90111399e-01 -1.25734603e+00 1.87029481e-01 -1.31714463e+00 -3.89618397e-01 -1.02359438e+00 -1.00107896e+00 1.15404022e+00 -7.37069920e-02 -1.46790838e+00 -9.50415432e-01 -3.79392207e-01 -5.67804992e-01 1.50646138e+00 -8.01816344e-01 -1.35317945e+00 5.75288117e-01 -6.09690649e-03 1.26493156e+00 -7.64745235e-01 1.00901103e+00 -4.99298610e-03 -2.49024287e-01 8.27455223e-01 5.08137703e-01 1.88669026e-01 1.14776301e+00 -1.06407368e+00 1.32013977e+00 9.54483211e-01 4.05169845e-01 7.04314590e-01 7.87155747e-01 -8.98696542e-01 -1.14033699e+00 -1.08767629e+00 1.88430548e+00 -9.67941344e-01 5.44454992e-01 -2.99065381e-01 -3.52029324e-01 7.48108149e-01 9.31679368e-01 -7.52559364e-01 6.34930968e-01 -3.89865607e-01 -2.24693820e-01 1.61447644e-01 -1.17215085e+00 9.83107924e-01 8.43169570e-01 -6.62957847e-01 -9.98492777e-01 6.16486073e-01 8.89585018e-01 -1.13784418e-01 -6.95957363e-01 3.77843022e-01 5.54935157e-01 -2.53599733e-01 5.16836286e-01 -1.06873035e+00 9.98267174e-01 -3.64413038e-02 -1.15822054e-01 -1.79585016e+00 8.00963193e-02 -1.07811689e+00 3.02411497e-01 9.30228889e-01 1.35225928e+00 -7.57157147e-01 7.09644020e-01 3.98781598e-01 4.07519117e-02 -6.53065026e-01 -8.18525374e-01 -8.43187451e-01 6.93479359e-01 5.36803342e-02 6.46718681e-01 9.29996550e-01 1.81369290e-01 1.00023365e+00 -5.40880024e-01 -5.44378161e-01 1.85772598e-01 -4.71680388e-02 1.01660109e+00 -6.73144579e-01 -4.57242310e-01 -3.33666503e-01 -1.24940567e-03 -8.75236392e-01 6.01062812e-02 -1.22971487e+00 1.74648706e-02 -1.54971147e+00 3.31115335e-01 1.52001698e-02 1.98850390e-02 1.76948294e-01 -3.19031060e-01 7.08556235e-01 7.03781620e-02 5.13091326e-01 2.49356896e-01 3.54354262e-01 1.60207009e+00 -2.36549780e-01 -2.29463950e-01 6.19135834e-02 -5.70299745e-01 1.12883456e-01 1.30744350e+00 -6.74473763e-01 -2.83672690e-01 -9.03919041e-01 1.65625215e-01 6.92294240e-02 -5.03105462e-01 -6.74378633e-01 -2.83240616e-01 -2.14761227e-01 3.00401837e-01 -1.44588217e-01 4.65136617e-02 -3.98830593e-01 2.09202722e-01 5.56918025e-01 -6.71010733e-01 7.81003475e-01 2.39540666e-01 7.45348856e-02 -3.01175028e-01 -1.31512880e-01 5.08048296e-01 -4.13115710e-01 1.58219472e-01 -9.98185277e-02 -6.41341805e-01 -2.69818842e-01 5.95278382e-01 -2.98853368e-01 -7.69141167e-02 -4.05948550e-01 -4.64447916e-01 -1.83614865e-01 6.97702646e-01 6.18340313e-01 2.73303360e-01 -1.52253377e+00 -1.39436841e+00 1.15667500e-01 9.58100520e-03 -5.99709928e-01 -7.98662007e-01 7.83320665e-01 -9.39721882e-01 8.04009795e-01 -5.82719684e-01 -3.67252797e-01 -1.01237273e+00 2.10759848e-01 2.00234488e-01 -5.07079363e-01 -2.56482363e-01 4.12124008e-01 -4.42710370e-01 -8.90525699e-01 -5.77267826e-01 8.04530177e-03 2.20014274e-01 -6.24016762e-01 2.67157078e-01 2.15765342e-01 9.98769030e-02 -8.47327054e-01 3.31166357e-01 2.18155548e-01 -3.21332157e-01 -9.14396584e-01 1.04408658e+00 4.89359125e-02 -3.41178864e-01 5.65758348e-01 1.03676033e+00 9.07911658e-02 -4.10432965e-01 -2.99090475e-01 1.44081488e-01 -1.33506998e-01 -5.80248713e-01 -1.14871061e+00 -2.37562045e-01 7.63556540e-01 1.85584709e-01 -3.43258768e-01 8.54769588e-01 -5.29102206e-01 1.19737923e+00 8.66459429e-01 8.17894340e-01 -1.12662435e+00 -4.74572241e-01 7.92465210e-01 8.75628293e-01 -1.42338204e+00 -3.98826897e-01 -1.27249718e-01 -6.12474620e-01 1.09878862e+00 4.17454451e-01 -2.06404194e-01 -1.83774129e-01 2.57479906e-01 5.23661077e-01 5.71920693e-01 -8.26056659e-01 3.22266281e-01 4.17162739e-02 5.17784536e-01 8.22635829e-01 3.63898546e-01 -1.18775594e+00 -1.46245614e-01 -6.86290622e-01 9.33630094e-02 5.47682464e-01 7.35908985e-01 -2.09338918e-01 -1.98712730e+00 -3.85218471e-01 3.73259574e-01 -7.64460921e-01 -6.96776509e-01 -1.29649401e+00 5.71957409e-01 -2.92579561e-01 1.07367921e+00 -1.99147686e-01 -4.82440650e-01 -3.77383381e-02 6.59537435e-01 6.26471460e-01 -8.15031230e-01 -9.07252491e-01 -1.23724990e-01 7.06535041e-01 -3.02509796e-02 -3.03494871e-01 -5.61961353e-01 -7.79322505e-01 -2.31489301e-01 -2.24800318e-01 6.99670374e-01 1.05322969e+00 1.05323219e+00 3.80856574e-01 -6.46289289e-02 8.88348818e-01 -6.39055967e-01 -7.71786451e-01 -1.78129852e+00 3.10717195e-01 2.35516474e-01 -1.45090505e-01 1.04854003e-01 -2.03071465e-03 3.87240648e-01]

[11.599553108215332, 10.398356437683105]

6b0c30a0-daf8-48c2-9cd9-6e3d89405010

expanded-parts-model-for-human-attribute-and

http://openaccess.thecvf.com/content_cvpr_2013/html/Sharma_Expanded_Parts_Model_2013_CVPR_paper.html

http://openaccess.thecvf.com/content_cvpr_2013/papers/Sharma_Expanded_Parts_Model_2013_CVPR_paper.pdf

Expanded Parts Model for Human Attribute and Action Recognition in Still Images

We propose a new model for recognizing human attributes (e.g. wearing a suit, sitting, short hair) and actions (e.g. running, riding a horse) in still images. The proposed model relies on a collection of part templates which are learnt discriminatively to explain specific scale-space locations in the images (in human centric coordinates). It avoids the limitations of highly structured models, which consist of a few (i.e. a mixture of) 'average' templates. To learn our model, we propose an algorithm which automatically mines out parts and learns corresponding discriminative templates with their respective locations from a large number of candidate parts. We validate the method on recent challenging datasets: (i) Willow 7 actions [7], (ii) 27 Human Attributes (HAT) [25], and (iii) Stanford 40 actions [37]. We obtain convincing qualitative and state-of-the-art quantitative results on the three datasets.

['Gaurav Sharma', 'Frederic Jurie', 'Cordelia Schmid']

2013-06-01

cvpr-2013-6

['action-recognition-in-still-images']

['computer-vision']

[ 3.44261557e-01 2.34875068e-01 -2.10322529e-01 -7.41514146e-01 -8.73582840e-01 -3.01939875e-01 7.24083126e-01 -2.87746161e-01 -2.72093773e-01 5.48808873e-01 3.74496460e-01 3.62630039e-01 -8.80158618e-02 -1.62187755e-01 -8.13354969e-01 -6.40197754e-01 -1.40873984e-01 8.64102483e-01 4.81180876e-01 -7.18474761e-02 3.02557349e-01 6.38737202e-01 -1.76581812e+00 4.41042721e-01 4.34558839e-01 9.38050926e-01 1.16141126e-01 5.92925131e-01 3.58544499e-01 8.47168744e-01 -4.83395159e-01 -6.84569240e-01 2.47782096e-01 -4.12428737e-01 -8.05658638e-01 7.49178290e-01 9.06080544e-01 -2.41773069e-01 -3.16899955e-01 7.82883823e-01 1.59658626e-01 1.60984948e-01 1.01765311e+00 -1.46060002e+00 -4.08177793e-01 1.67112499e-01 -7.78859735e-01 1.02332234e-03 6.16623580e-01 -3.51330568e-03 1.05438399e+00 -9.30222452e-01 9.42373991e-01 1.39879382e+00 7.61517167e-01 6.50829494e-01 -1.11780453e+00 -3.62603635e-01 1.86057493e-01 3.66152942e-01 -1.40295994e+00 -6.11532032e-01 6.90114617e-01 -4.40888584e-01 7.35325456e-01 3.45398873e-01 7.70456612e-01 1.40939879e+00 4.33733582e-01 1.29242432e+00 1.20695388e+00 -3.68178248e-01 3.23341966e-01 2.64073163e-02 -4.47872877e-02 7.24898398e-01 4.22315411e-02 -1.45220891e-01 -5.60159385e-01 -3.15142035e-01 8.51125240e-01 2.77139902e-01 2.19974592e-01 -9.76675868e-01 -1.38153362e+00 6.32225633e-01 2.83124000e-01 6.32153675e-02 -5.01648128e-01 1.44949630e-01 1.61158368e-01 -2.81237513e-01 2.75163859e-01 1.59582511e-01 -5.99443316e-01 -1.75500453e-01 -8.09083223e-01 4.58050758e-01 6.30904675e-01 1.26908469e+00 9.67754602e-01 -3.30123782e-01 -9.30766314e-02 7.92252123e-01 8.21916759e-02 5.45131803e-01 1.85349956e-01 -1.20231450e+00 2.76065320e-01 6.51851833e-01 3.89450997e-01 -7.19897747e-01 -4.37487721e-01 -1.22206189e-01 -7.05504596e-01 1.06453799e-01 5.74217319e-01 1.88630700e-01 -1.08480418e+00 1.52952242e+00 4.94744003e-01 1.27514154e-01 -2.94807732e-01 8.15709352e-01 4.08823043e-01 1.06500618e-01 2.69556046e-01 1.36284947e-01 1.53023672e+00 -1.09947801e+00 -6.32410944e-01 -4.45870817e-01 4.06042904e-01 -7.17893302e-01 9.60396290e-01 3.94438237e-01 -9.83066559e-01 -8.27632427e-01 -6.42532349e-01 -1.22387387e-01 -4.62018996e-01 5.59786737e-01 7.47981846e-01 4.21804875e-01 -8.60239208e-01 6.58731639e-01 -8.35646272e-01 -9.10108626e-01 5.96112728e-01 2.37641320e-01 -7.62216091e-01 -3.12078372e-02 -6.23462379e-01 9.51026201e-01 4.46082987e-02 -1.15886115e-01 -9.67863142e-01 -2.76264042e-01 -9.80899215e-01 -3.58832628e-01 4.50329900e-01 -7.31600523e-01 1.12697363e+00 -1.10279739e+00 -1.25623262e+00 1.37786067e+00 -3.58960301e-01 -4.05522168e-01 7.78264165e-01 -7.37253249e-01 -3.24777365e-01 4.63381588e-01 3.49182725e-01 1.01462567e+00 9.71699059e-01 -1.34693301e+00 -8.13988805e-01 -6.28655910e-01 6.00208491e-02 1.17082633e-01 2.68030837e-02 3.15165490e-01 -4.73729312e-01 -7.35340416e-01 2.74160743e-01 -1.35723805e+00 -2.65817940e-01 1.81914344e-01 -6.69145942e-01 -2.15202749e-01 8.37463379e-01 -7.98549831e-01 8.46930563e-01 -1.94293475e+00 1.80141538e-01 1.41323460e-02 1.24742359e-01 -6.10940084e-02 3.38682979e-02 2.91253775e-01 5.13621718e-02 -2.12192312e-01 -2.08296239e-01 -6.34054363e-01 1.26904607e-01 5.46399534e-01 1.15661427e-01 7.96425760e-01 2.15223104e-01 1.02423501e+00 -7.29683101e-01 -9.06434476e-01 3.52824986e-01 3.94556731e-01 -3.58179480e-01 2.77473480e-01 7.64455050e-02 5.04260302e-01 -6.28106356e-01 1.02574718e+00 4.47327763e-01 -2.22207144e-01 -2.18143705e-02 -2.13991925e-01 7.21322522e-02 -6.20967969e-02 -1.17277002e+00 1.90258694e+00 1.65888434e-03 2.45848879e-01 -1.97991863e-01 -8.02425563e-01 8.10294330e-01 2.69262344e-01 5.34494758e-01 -2.55499274e-01 1.33540735e-01 -1.12024136e-02 -1.65545806e-01 -7.54049301e-01 3.24765444e-01 -6.07372485e-02 -3.31119984e-01 3.93966973e-01 2.96610802e-01 2.27336481e-01 1.95482180e-01 1.63361058e-01 1.06511283e+00 6.68434978e-01 7.24327862e-01 -3.85646284e-01 3.24761033e-01 -1.03813060e-01 5.72960913e-01 7.25525081e-01 -6.61549628e-01 9.58896160e-01 3.88439715e-01 -8.54502022e-01 -1.29138660e+00 -1.02932501e+00 5.18109184e-03 1.25248575e+00 1.62850589e-01 -3.45185101e-01 -8.51043522e-01 -9.45415556e-01 9.30806547e-02 4.78520811e-01 -1.26293325e+00 5.74800409e-02 -7.25929558e-01 -3.07211995e-01 3.42702717e-01 1.04577935e+00 4.00943488e-01 -1.23409963e+00 -1.15995026e+00 -7.91415200e-02 -3.43488783e-01 -1.37205946e+00 -5.17455995e-01 1.75336495e-01 -7.77262568e-01 -1.16462457e+00 -6.84988916e-01 -5.74017942e-01 8.89867842e-01 2.20175326e-01 1.31762624e+00 -2.10460667e-02 -3.78770649e-01 6.20072246e-01 -4.77910697e-01 -5.20619094e-01 -1.79965869e-02 -1.36569679e-01 2.16957584e-01 3.17560315e-01 4.26799119e-01 -4.66794133e-01 -5.82911670e-01 7.64088154e-01 -4.28995103e-01 1.45987868e-01 1.14813948e+00 7.36704886e-01 8.91484976e-01 -2.42151037e-01 1.87797800e-01 -9.97541726e-01 -2.18812615e-01 -2.14708135e-01 -1.59858376e-01 3.68205488e-01 -8.41700584e-02 -1.64301306e-01 2.92439252e-01 -6.55099154e-01 -1.07835245e+00 7.47004628e-01 2.48307988e-01 -5.80279529e-01 -8.70408416e-01 -2.27466971e-01 -3.50099564e-01 3.41535211e-02 6.64430976e-01 2.05233917e-01 -6.97190985e-02 -7.08656311e-01 4.57568884e-01 4.50368971e-01 6.99599862e-01 -7.13020921e-01 8.65334094e-01 7.87682354e-01 8.19990411e-02 -6.44086838e-01 -1.05270243e+00 -6.83273613e-01 -1.48223555e+00 -4.77245569e-01 1.15360582e+00 -8.76854718e-01 -5.70532322e-01 3.70018244e-01 -8.73565614e-01 -2.28327602e-01 -3.12244445e-01 3.61086339e-01 -1.12224400e+00 2.86132812e-01 -5.42938054e-01 -7.67323375e-01 8.31875056e-02 -7.95297384e-01 1.74867678e+00 2.92985350e-01 -5.80063045e-01 -6.81465745e-01 -6.03131168e-02 5.53415775e-01 -3.53101268e-02 6.87300265e-01 5.40184498e-01 -6.58779740e-01 -4.99316633e-01 -4.97110218e-01 -5.52674662e-03 1.73110768e-01 3.01696472e-02 5.71064949e-02 -9.87276971e-01 -1.87567636e-01 -3.03974301e-01 -4.99440759e-01 5.34091234e-01 3.86103392e-01 1.16172957e+00 -2.63684988e-01 -6.38084888e-01 4.25255388e-01 1.02067125e+00 2.49741703e-01 8.63607645e-01 3.37535560e-01 6.81090534e-01 7.33867943e-01 1.05258119e+00 4.47098702e-01 3.02409828e-01 9.19511735e-01 2.13260993e-01 -1.08902223e-01 -1.21504039e-01 -3.93331140e-01 2.11573958e-01 2.67603427e-01 -7.01217055e-01 1.79456577e-01 -7.09178448e-01 6.03380561e-01 -2.03331470e+00 -1.11584306e+00 -9.04429406e-02 1.90987599e+00 3.76094222e-01 1.10387675e-01 5.25650442e-01 -8.39178935e-02 6.42897367e-01 6.13543279e-02 -5.02140343e-01 -1.31880650e-02 -7.98745379e-02 1.53356388e-01 2.42833793e-01 -4.08631563e-02 -1.65968752e+00 8.80164742e-01 6.76442385e+00 6.30008399e-01 -4.02457267e-01 -3.68876494e-02 7.40134001e-01 1.81179747e-01 2.39288330e-01 1.00581475e-01 -7.85906553e-01 2.49063075e-01 5.93697309e-01 2.72324026e-01 4.93789464e-02 1.30400479e+00 -1.28060624e-01 -2.25710571e-01 -1.29300892e+00 8.62858891e-01 3.04588854e-01 -1.00398278e+00 -1.06995560e-01 1.29561350e-01 7.22064495e-01 -4.87262726e-01 -7.25187808e-02 2.75756449e-01 1.60523117e-01 -1.00520539e+00 1.04158413e+00 8.62139106e-01 5.70216298e-01 -5.26098907e-01 7.17724919e-01 4.19976115e-01 -1.23804736e+00 -3.57659273e-02 -5.31346500e-01 9.71904248e-02 3.31883468e-02 8.82413611e-02 -4.80796307e-01 4.61018533e-01 1.01346242e+00 9.63034034e-01 -8.98464561e-01 7.91105270e-01 -3.37962955e-01 3.09578896e-01 -1.73880115e-01 1.11010723e-01 1.19399197e-01 -2.43734434e-01 2.79703140e-01 1.23660052e+00 2.18021691e-01 4.22366947e-01 2.23403469e-01 6.25601530e-01 1.99109122e-01 -2.54316349e-02 -5.50945997e-01 3.55788499e-01 1.43568799e-01 1.45569110e+00 -8.69256616e-01 -4.07322317e-01 -5.63992083e-01 1.09042048e+00 3.05363595e-01 3.17394674e-01 -7.86797166e-01 -1.24666549e-01 5.70858419e-01 4.02342796e-01 5.45790076e-01 -1.13011628e-01 -5.47486208e-02 -1.04127324e+00 1.02802604e-01 -9.22588825e-01 5.19422710e-01 -9.73427415e-01 -1.38172460e+00 5.33590257e-01 4.38036054e-01 -1.54623377e+00 -4.03390795e-01 -5.81236422e-01 -4.72594500e-01 4.50613141e-01 -9.17291343e-01 -1.56817997e+00 -4.98715907e-01 7.90054619e-01 6.85476840e-01 -1.04233615e-01 8.22352290e-01 1.91827163e-01 -4.16313857e-01 4.18511122e-01 -3.99567634e-01 1.95690110e-01 7.48752594e-01 -1.39588130e+00 3.70965600e-01 4.83699650e-01 3.88897866e-01 6.27157271e-01 9.14197147e-01 -5.55182219e-01 -1.16949117e+00 -8.96604002e-01 1.00647426e+00 -1.10273015e+00 5.42717040e-01 -5.72325230e-01 -7.58481920e-01 1.10339427e+00 -8.40600580e-02 2.78833151e-01 4.47612584e-01 1.00979328e-01 -2.37027436e-01 2.09153313e-02 -1.22859931e+00 3.77499938e-01 1.41187215e+00 -2.18169212e-01 -7.35541403e-01 4.86264408e-01 1.32278159e-01 -5.01696587e-01 -8.15717161e-01 5.07531404e-01 9.70416784e-01 -1.34092009e+00 1.17250073e+00 -8.99510145e-01 2.99699545e-01 -2.14570671e-01 -2.79557407e-01 -9.27291632e-01 -4.34215575e-01 -4.62784588e-01 -1.97162434e-01 8.63623798e-01 6.02737330e-02 -1.89449862e-01 9.71589863e-01 6.35393620e-01 -1.25619024e-01 -1.05353761e+00 -9.22005832e-01 -8.66513848e-01 -4.15432662e-01 3.63231264e-02 3.89542729e-01 6.35690928e-01 -2.81083375e-01 1.54002547e-01 -6.60906136e-01 -6.74775690e-02 8.28476250e-01 1.25581399e-01 1.14969671e+00 -1.33657110e+00 -1.35357574e-01 -6.84541017e-02 -8.17577660e-01 -9.86648262e-01 2.22485468e-01 -1.78563341e-01 1.04511432e-01 -1.44918299e+00 6.81221783e-01 -2.55614698e-01 -1.29203245e-01 7.29000688e-01 -2.44906276e-01 2.61141002e-01 1.98280923e-02 3.69321734e-01 -9.95231926e-01 5.24021089e-01 1.21634912e+00 1.27328396e-01 4.05592054e-01 2.85737574e-01 -5.04037321e-01 1.04995298e+00 5.67139685e-01 -6.23454928e-01 -1.57282576e-01 1.09379977e-01 -3.23710442e-01 -1.20071892e-03 6.35321975e-01 -1.16577423e+00 -2.24477984e-02 -3.68662030e-01 8.40603173e-01 -6.31453693e-01 6.76986992e-01 -8.12889993e-01 2.55726725e-01 2.74602532e-01 -3.66782457e-01 2.01901242e-01 -1.60537332e-01 7.34137475e-01 -1.74809411e-01 1.27260685e-02 9.63438570e-01 -3.49716008e-01 -9.64277506e-01 2.16773763e-01 -9.35030729e-02 -1.18831746e-01 1.20585907e+00 -5.01693130e-01 -2.05493391e-01 -5.13568640e-01 -9.13914561e-01 -1.68647338e-02 6.68115854e-01 4.62594301e-01 5.64536691e-01 -1.56798816e+00 -6.00595415e-01 -2.85233930e-02 4.33089942e-01 -2.34408930e-01 3.39688271e-01 9.93354619e-01 -2.13360712e-01 3.80046606e-01 -6.16991341e-01 -7.11168349e-01 -1.29829895e+00 6.43997014e-01 9.24872309e-02 -1.51491076e-01 -9.44090784e-01 7.09239960e-01 4.34356779e-01 -2.47008041e-01 1.71644390e-01 -1.22531720e-01 -1.16108224e-01 -1.46754012e-01 3.15851241e-01 5.36722958e-01 -1.52001262e-01 -1.20791042e+00 -6.19582534e-01 8.73995185e-01 -3.81326638e-02 5.20505868e-02 1.22957706e+00 7.98290689e-03 1.81741655e-01 5.57346046e-01 1.08456075e+00 -1.73218429e-01 -1.56669796e+00 -2.72377074e-01 4.52041738e-02 -7.37001061e-01 -7.08095849e-01 -6.10851347e-01 -8.59748483e-01 7.77683437e-01 5.23904443e-01 -1.24718726e-01 1.02046049e+00 4.88517076e-01 7.01838553e-01 3.10365915e-01 7.14723885e-01 -1.29285586e+00 5.13512135e-01 8.59326720e-02 9.90034044e-01 -1.19466662e+00 3.03046882e-01 -4.23998326e-01 -1.10348368e+00 1.00881946e+00 6.62256956e-01 -1.75874904e-01 4.46827650e-01 1.36945873e-01 2.06285611e-01 -4.75859940e-01 -6.46662593e-01 -3.66962194e-01 6.22067213e-01 6.46214366e-01 2.22797781e-01 4.44027185e-01 -6.85478523e-02 4.60123986e-01 -1.61906317e-01 -3.68955255e-01 2.65949845e-01 1.10462284e+00 -4.86944318e-01 -8.50625515e-01 -4.12944436e-01 4.10612613e-01 -2.42821500e-01 5.61146677e-01 -6.22584522e-01 1.28973699e+00 4.64395642e-01 7.14075029e-01 -1.02922790e-01 -4.55492824e-01 6.60058737e-01 2.84107208e-01 6.40575707e-01 -5.00215054e-01 -2.31614068e-01 2.27809712e-01 8.39302614e-02 -9.92152870e-01 -8.45623851e-01 -1.25554490e+00 -9.30395067e-01 1.20430700e-01 1.02214716e-01 -3.25621456e-01 3.80884737e-01 1.02709341e+00 1.47694141e-01 1.21952191e-01 2.87522405e-01 -1.18130875e+00 -5.25518179e-01 -1.19846654e+00 -8.97971094e-01 1.04119182e+00 -5.87442555e-02 -1.17322123e+00 -2.70050645e-01 4.31717306e-01]

[7.991342067718506, 0.2295798510313034]

4a337c14-c595-4269-a157-4b7566f7fa5c

a-harmonic-mean-linear-discriminant-analysis

1610.04631

http://arxiv.org/abs/1610.04631v2

http://arxiv.org/pdf/1610.04631v2.pdf

A Harmonic Mean Linear Discriminant Analysis for Robust Image Classification

Linear Discriminant Analysis (LDA) is a widely-used supervised dimensionality reduction method in computer vision and pattern recognition. In null space based LDA (NLDA), a well-known LDA extension, between-class distance is maximized in the null space of the within-class scatter matrix. However, there are some limitations in NLDA. Firstly, for many data sets, null space of within-class scatter matrix does not exist, thus NLDA is not applicable to those datasets. Secondly, NLDA uses arithmetic mean of between-class distances and gives equal consideration to all between-class distances, which makes larger between-class distances can dominate the result and thus limits the performance of NLDA. In this paper, we propose a harmonic mean based Linear Discriminant Analysis, Multi-Class Discriminant Analysis (MCDA), for image classification, which minimizes the reciprocal of weighted harmonic mean of pairwise between-class distance. More importantly, MCDA gives higher priority to maximize small between-class distances. MCDA can be extended to multi-label dimension reduction. Results on 7 single-label data sets and 4 multi-label data sets show that MCDA has consistently better performance than 10 other single-label approaches and 4 other multi-label approaches in terms of classification accuracy, macro and micro average F1 score.

['Shuai Zheng', 'Heng Huang', 'Chris Ding', 'Feiping Nie']

2016-10-14

['supervised-dimensionality-reduction']

['computer-vision']

[-2.24074826e-01 -5.00396132e-01 -3.77381235e-01 -4.52756017e-01 -6.79874420e-01 -5.53552926e-01 3.18655163e-01 1.84430435e-01 -1.45097882e-01 5.19220650e-01 8.90281722e-02 -1.12039275e-01 -5.23056388e-01 -6.67127609e-01 1.67872787e-01 -1.22866344e+00 -6.26220256e-02 4.04573590e-01 -4.45840582e-02 1.15984187e-01 1.48266137e-01 6.56749010e-01 -1.69803870e+00 2.00680688e-01 7.81866014e-01 7.68247664e-01 -8.81109238e-02 1.32809013e-01 -2.12401196e-01 1.81569785e-01 -3.66542041e-01 7.68948123e-02 1.99330688e-01 -6.80967927e-01 -5.49864411e-01 1.28720433e-01 4.46438789e-01 -1.87836133e-03 1.65996596e-01 1.12470269e+00 6.38756692e-01 9.15324241e-02 1.25206017e+00 -1.60688806e+00 -8.07097495e-01 -7.89228156e-02 -9.55637455e-01 -2.37447750e-02 2.60893285e-01 -3.52965653e-01 1.05011535e+00 -1.12055957e+00 4.37208682e-01 1.46172678e+00 6.34732246e-01 3.50647837e-01 -1.44196749e+00 -7.52732158e-01 -2.67128289e-01 7.66799152e-02 -1.54156840e+00 -1.61333419e-02 9.33587193e-01 -7.14027226e-01 5.77760994e-01 5.16374111e-01 5.85320532e-01 2.49736220e-01 4.73970026e-01 6.25137329e-01 1.55467069e+00 -6.25298500e-01 2.98652321e-01 1.34755149e-01 5.44103324e-01 4.96814102e-01 2.82074660e-01 -1.58880144e-01 -5.36035299e-02 -4.84843463e-01 1.42393500e-01 8.19666311e-02 1.23372734e-01 -1.02022564e+00 -1.07622087e+00 1.16572368e+00 2.07116231e-01 4.60228711e-01 -1.54612616e-01 -4.75226343e-01 4.36714321e-01 1.14161782e-01 6.30350232e-01 1.12769030e-01 -3.58904861e-02 2.40334257e-01 -8.51718426e-01 3.15050900e-01 4.38388735e-01 4.56328303e-01 8.27868998e-01 -3.46302211e-01 3.93055268e-02 1.26880598e+00 5.74019492e-01 7.80719042e-01 7.93101490e-01 -8.49387586e-01 3.39180902e-02 8.11609089e-01 -2.93440908e-01 -1.49934697e+00 -8.36409628e-01 -3.65318894e-01 -8.19930971e-01 4.63015258e-01 5.38619220e-01 2.66934454e-01 -4.67135340e-01 1.58134758e+00 4.66001958e-01 -3.32547486e-01 1.29714102e-01 1.02245903e+00 7.60898709e-01 5.69625258e-01 1.56358153e-01 -8.14029574e-01 1.31384122e+00 -6.67570531e-01 -9.02735353e-01 -3.26477289e-02 1.10420370e+00 -9.47642982e-01 1.19700253e+00 5.16800821e-01 -3.52758884e-01 -2.95354187e-01 -8.98668647e-01 -5.10298973e-03 -4.50560927e-01 3.72736990e-01 6.34316802e-01 8.48214626e-01 -6.10589623e-01 1.51945636e-01 -4.74756807e-01 -5.29261947e-01 3.07711184e-01 4.20181543e-01 -6.92858100e-01 -3.93668920e-01 -9.99065578e-01 7.49253392e-01 2.74582326e-01 -2.24827170e-01 -9.59625095e-02 -3.26133639e-01 -7.35646069e-01 -1.85687214e-01 3.67749035e-02 3.09272407e-04 5.61899722e-01 -4.41075146e-01 -1.03210795e+00 1.13609982e+00 -1.92246184e-01 1.72308713e-01 -1.15710065e-01 2.81861126e-01 -6.02344811e-01 -1.75405413e-01 3.50582689e-01 6.91849291e-01 4.76807535e-01 -1.24283040e+00 -3.84378970e-01 -7.31473684e-01 -2.98569679e-01 2.56601781e-01 -5.02297342e-01 -8.97891670e-02 2.05393076e-01 -4.23880756e-01 7.84921288e-01 -1.40473855e+00 1.87801585e-01 -1.13475487e-01 -2.57677436e-01 -7.51122177e-01 1.23655391e+00 -3.31505537e-01 1.61716175e+00 -2.52166986e+00 -5.48625141e-02 3.51612389e-01 2.18451083e-01 2.65484422e-01 -1.01188160e-01 2.96202749e-01 -3.84945363e-01 -2.27481246e-01 -1.54474154e-01 1.79449007e-01 -4.68659624e-02 1.09732069e-01 6.07295446e-02 8.29944372e-01 -1.50214016e-01 2.72279322e-01 -9.25289273e-01 -6.38739705e-01 3.16628695e-01 2.57094860e-01 -2.62339652e-01 -3.16239864e-01 3.87612104e-01 1.99422427e-02 -2.61279404e-01 6.43766105e-01 1.07486355e+00 -1.60248086e-01 2.79164374e-01 -4.53385383e-01 -1.11674801e-01 -2.16365308e-01 -1.34416425e+00 1.39650142e+00 -1.49013966e-01 6.95080578e-01 -4.73656744e-01 -9.76572037e-01 1.16833711e+00 6.18778355e-02 8.17634284e-01 -4.65316236e-01 1.68827698e-01 4.00727779e-01 1.80829763e-01 -4.40390825e-01 5.66119850e-02 -3.69415104e-01 -5.35624400e-02 5.70826232e-01 -1.31190985e-01 1.01148099e-01 3.68120223e-01 2.00995639e-01 5.45935154e-01 -3.09726000e-01 8.41592908e-01 -7.71306396e-01 8.59774709e-01 -7.93179125e-02 5.32255948e-01 -8.11879113e-02 -3.52703482e-01 6.37038529e-01 7.19472110e-01 -5.36859453e-01 -7.89263010e-01 -9.74265218e-01 -6.45601749e-01 1.06324589e+00 2.32829407e-01 -4.81951147e-01 -6.45324767e-01 -1.03579628e+00 2.42987767e-01 6.09408796e-01 -4.95324016e-01 -1.54947311e-01 -2.19967812e-01 -1.25836921e+00 2.14300409e-01 1.26589224e-01 3.04704726e-01 -5.59248745e-01 -7.33785778e-02 -1.75893888e-01 -1.18451469e-01 -4.36882049e-01 -5.20516455e-01 -6.20864183e-02 -8.46262574e-01 -1.32528484e+00 -8.14226389e-01 -9.76969779e-01 7.66340017e-01 7.45887339e-01 4.61572647e-01 -3.08613747e-01 -3.56032640e-01 1.97873890e-01 -2.17806131e-01 -1.60544738e-01 -3.30230087e-01 -1.05283320e-01 4.59515154e-01 3.56517620e-02 1.17885458e+00 -4.13674235e-01 -3.94514352e-01 7.16848195e-01 -6.19065344e-01 -3.12952787e-01 5.38589776e-01 8.64476740e-01 7.73177505e-01 3.87859970e-01 7.53030300e-01 -5.94784141e-01 7.48151243e-01 -5.50262094e-01 -3.32104087e-01 2.84062386e-01 -9.97294843e-01 -1.61835790e-01 2.21609339e-01 -5.29311478e-01 -5.49858749e-01 8.09525996e-02 3.07708561e-01 -1.50518775e-01 -1.48590505e-01 3.99221361e-01 -3.55307221e-01 -5.01200631e-02 9.81034636e-01 1.42084777e-01 3.87640566e-01 -5.09310305e-01 1.15446836e-01 1.23310113e+00 -9.15656090e-02 -1.88022569e-01 2.85507560e-01 4.75501746e-01 6.45497501e-01 -8.15938115e-01 -8.34238946e-01 -8.46356034e-01 -9.34063137e-01 -1.86407834e-01 8.61100614e-01 -7.93214500e-01 -6.38918579e-01 5.17831624e-01 -6.20244145e-01 2.93442667e-01 -3.21143456e-02 1.02287102e+00 -3.58923316e-01 6.76881850e-01 -8.95127729e-02 -6.66209102e-01 -1.79424942e-01 -1.29320204e+00 8.28753829e-01 4.18838225e-02 -4.05086905e-01 -1.04811978e+00 2.38923743e-01 3.01894844e-01 1.08862799e-02 2.62340218e-01 1.24959230e+00 -9.10105467e-01 2.93562204e-01 -3.31068844e-01 -3.21448863e-01 6.52859211e-01 6.36895895e-01 -1.66861504e-01 -7.47494876e-01 -5.12823999e-01 -9.29445848e-02 -1.91899195e-01 4.50343013e-01 6.08831406e-01 7.56851315e-01 -8.96550864e-02 -5.79527080e-01 6.64965734e-02 1.24016309e+00 4.66368407e-01 3.74272346e-01 3.17934573e-01 7.12072432e-01 8.22474718e-01 9.87626135e-01 2.94588894e-01 3.98048550e-01 8.25197637e-01 -1.05954204e-02 1.01956122e-01 -1.61842152e-01 2.93886989e-01 1.61352709e-01 9.86965179e-01 3.06596607e-01 -6.40337542e-02 -9.51972187e-01 2.89783657e-01 -1.60923088e+00 -7.73956478e-01 -7.28570998e-01 2.22072935e+00 6.03294015e-01 -3.49329889e-01 4.94137317e-01 4.96292293e-01 9.01639640e-01 4.74650897e-02 -3.94841015e-01 -3.01816195e-01 -2.88340926e-01 -4.72016096e-01 2.29334429e-01 4.28563327e-01 -1.35940599e+00 4.03966367e-01 6.13678455e+00 1.24101412e+00 -1.12640405e+00 9.10534896e-03 3.31059337e-01 -8.68436024e-02 -1.10043278e-02 -1.26540720e-01 -8.47103477e-01 6.75502777e-01 5.92962503e-01 -1.28694192e-01 1.04634576e-01 1.13383520e+00 1.27893433e-01 -3.79702866e-01 -9.71036375e-01 1.27448928e+00 2.94240057e-01 -5.98896444e-01 -1.36644274e-01 6.51754975e-01 8.62799346e-01 -4.52737242e-01 2.41309494e-01 1.30297899e-01 -9.67509076e-02 -5.56066573e-01 1.05332635e-01 1.14024334e-01 1.03648913e+00 -1.07671583e+00 7.99609423e-01 3.57632428e-01 -8.79812360e-01 -4.16387012e-03 -5.95270455e-01 3.98137234e-02 -1.76884964e-01 1.19148254e+00 -5.99242806e-01 2.92681664e-01 5.06313682e-01 7.99928367e-01 -6.26964211e-01 8.96163046e-01 2.19136253e-01 3.26615512e-01 -2.31672853e-01 7.54118338e-02 2.43952125e-01 -5.97002506e-01 5.14597714e-01 9.49265063e-01 5.32639444e-01 1.29230812e-01 2.79375792e-01 1.83765516e-01 3.66682678e-01 6.66196764e-01 -6.96924508e-01 1.40497997e-01 4.92249817e-01 1.28102410e+00 -1.04861343e+00 -8.04397985e-02 -4.75994378e-01 5.34586728e-01 -1.45957887e-01 2.10670922e-02 -3.70835394e-01 -4.09361720e-01 6.15474939e-01 1.50133684e-01 -1.73003882e-01 -1.72906935e-01 -3.78481895e-01 -9.69179571e-01 -2.11291581e-01 -7.90921211e-01 7.87850976e-01 -5.67552269e-01 -1.50356066e+00 3.40710372e-01 1.73089057e-01 -1.67497540e+00 -1.77119300e-01 -5.95688522e-01 -3.89640629e-02 7.38720477e-01 -1.01600397e+00 -1.09521735e+00 -9.28879529e-02 4.44392353e-01 3.23994160e-01 -4.67016876e-01 1.11228883e+00 5.62280059e-01 -4.48760271e-01 5.33775866e-01 6.22043252e-01 -1.16206907e-01 1.12518406e+00 -1.06569004e+00 -4.86805081e-01 3.60487908e-01 -4.13023606e-02 5.37015617e-01 6.00877285e-01 -5.82934916e-01 -8.74683499e-01 -8.27964604e-01 1.12084663e+00 -1.93527132e-01 3.84016961e-01 -2.09313601e-01 -9.03469145e-01 3.35962027e-01 -3.36579710e-01 1.76533401e-01 1.44367015e+00 2.01095730e-01 -4.30207461e-01 -2.70817012e-01 -1.50654244e+00 2.37573937e-01 5.55182278e-01 -4.04818177e-01 -3.09595108e-01 6.18525743e-01 1.27337843e-01 -4.77910452e-02 -1.09141791e+00 3.21566522e-01 8.96976829e-01 -9.91378248e-01 9.00969088e-01 -3.64846826e-01 1.26981452e-01 -6.54649794e-01 -4.45168436e-01 -1.30638278e+00 -6.12280548e-01 2.57873893e-01 1.41820848e-01 1.33031833e+00 3.54419760e-02 -6.95257843e-01 6.45715594e-01 3.88301820e-01 1.09887242e-01 -7.73084283e-01 -9.80048478e-01 -1.04745245e+00 2.46180251e-01 -9.89108533e-02 5.59633613e-01 1.37541914e+00 1.67302698e-01 3.10027927e-01 -3.42845947e-01 -9.71300751e-02 8.51010799e-01 2.90042371e-01 3.85993928e-01 -1.71776581e+00 2.80006438e-01 -2.35382408e-01 -8.14247370e-01 -3.58153939e-01 2.82983750e-01 -1.09586358e+00 -2.88190275e-01 -1.42083085e+00 3.46148759e-01 -7.81801045e-01 -2.64731914e-01 4.53656167e-01 8.97285193e-02 6.08704746e-01 8.72695446e-02 7.25291967e-01 -5.32106936e-01 3.92027110e-01 1.22802150e+00 2.76626386e-02 -4.40727174e-01 -7.50909895e-02 -5.14648318e-01 5.74351728e-01 7.44859219e-01 -7.96686769e-01 -5.33951879e-01 2.40581870e-01 4.35408950e-02 -2.74797946e-01 -6.83543682e-02 -9.75094855e-01 -6.97060227e-02 -2.81045943e-01 3.27164799e-01 -7.89776802e-01 1.46579608e-01 -8.30449164e-01 2.58176088e-01 4.32967037e-01 -2.65950829e-01 -2.98126847e-01 -9.33674425e-02 4.02436882e-01 -2.77385056e-01 -3.93247604e-01 9.40305293e-01 2.40327150e-01 -5.47200859e-01 -6.11712597e-02 -4.88208145e-01 -3.28097224e-01 1.40085876e+00 -4.08619165e-01 -2.19754621e-01 3.62732150e-02 -8.18731487e-01 -5.44961430e-02 3.99876535e-01 2.83919871e-01 2.85393149e-01 -1.97802925e+00 -5.14292955e-01 2.53972054e-01 3.39731604e-01 -3.44586790e-01 5.16537070e-01 1.08094263e+00 -2.89368510e-01 6.67087197e-01 -3.05562258e-01 -8.00090611e-01 -1.79391766e+00 7.43885756e-01 -3.91423851e-02 -1.46478504e-01 -2.51853049e-01 4.75107580e-01 4.34160143e-01 -7.83558786e-01 -3.04999739e-01 1.60407171e-01 -5.21205664e-01 7.00050950e-01 5.38816810e-01 1.02793181e+00 1.65067613e-01 -1.12741458e+00 -6.10949457e-01 1.05504477e+00 -1.39911659e-02 1.70310125e-01 1.09909320e+00 -3.90339583e-01 -6.44039631e-01 5.88014245e-01 1.73578465e+00 6.48996746e-03 -6.64897203e-01 -1.99391410e-01 -4.64486703e-02 -5.85397959e-01 2.93942660e-01 -5.77515900e-01 -8.09103847e-01 9.33407426e-01 1.08204496e+00 3.68893713e-01 1.21282446e+00 -3.36529240e-02 3.89819920e-01 2.22703442e-01 1.88339964e-01 -1.08761513e+00 6.35091364e-02 1.80870786e-01 7.29565382e-01 -1.36812365e+00 4.47292507e-01 -6.59467876e-01 -7.42438972e-01 1.25936735e+00 4.30625468e-01 -9.53904912e-02 9.81522679e-01 -7.23400190e-02 1.93677172e-01 -2.75442392e-01 -2.51647741e-01 1.50057673e-01 5.57435274e-01 6.03944242e-01 5.74924946e-01 1.65093377e-01 -1.08683372e+00 4.52182829e-01 -5.35237826e-02 -2.02190921e-01 1.90967634e-01 6.90682352e-01 -6.06434405e-01 -1.23032582e+00 -7.61931121e-01 5.12581050e-01 -3.02403629e-01 3.44425738e-01 -8.15923586e-02 7.70617127e-01 3.61456066e-01 8.48339319e-01 7.16326907e-02 -3.49807501e-01 5.87818101e-02 3.24128330e-01 1.79381400e-01 -4.78190124e-01 8.12680721e-02 2.43135110e-01 -2.08993196e-01 -2.65790105e-01 -6.52204633e-01 -9.69269753e-01 -1.29380071e+00 -3.79579216e-01 -6.66295290e-01 2.81626910e-01 8.52414966e-01 7.59498179e-01 3.01170498e-01 -8.65550041e-02 9.26622570e-01 -4.42803383e-01 -4.27884310e-01 -9.87835646e-01 -1.19539583e+00 2.97744870e-01 1.03660136e-01 -1.01153314e+00 -7.04274356e-01 1.35016833e-02]

[7.8944621086120605, 4.257002353668213]

b8c5a30e-1a20-4e19-8610-b1d770412841

dualhgnn-a-dual-hypergraph-neural-network-for

2306.04214

https://arxiv.org/abs/2306.04214v1

https://arxiv.org/pdf/2306.04214v1.pdf

DualHGNN: A Dual Hypergraph Neural Network for Semi-Supervised Node Classification based on Multi-View Learning and Density Awareness

Graph-based semi-supervised node classification has been shown to become a state-of-the-art approach in many applications with high research value and significance. Most existing methods are only based on the original intrinsic or artificially established graph structure which may not accurately reflect the "true" correlation among data and are not optimal for semi-supervised node classification in the downstream graph neural networks. Besides, while existing graph-based methods mostly utilize the explicit graph structure, some implicit information, for example, the density information, can also provide latent information that can be further exploited. To address these limitations, this paper proposes the Dual Hypergraph Neural Network (DualHGNN), a new dual connection model integrating both hypergraph structure learning and hypergraph representation learning simultaneously in a unified architecture. The DualHGNN first leverages a multi-view hypergraph learning network to explore the optimal hypergraph structure from multiple views, constrained by a consistency loss proposed to improve its generalization. Then, DualHGNN employs a density-aware hypergraph attention network to explore the high-order semantic correlation among data points based on the density-aware attention mechanism. Extensive experiments are conducted in various benchmark datasets, and the results demonstrate the effectiveness of the proposed approach.

['Qian Tao', 'Jun Yan', 'Jianpeng Liao']

2023-06-07

['multi-view-learning']

['computer-vision']

[-2.24402696e-01 3.64219725e-01 -6.59532666e-01 -3.78932118e-01 -2.99427807e-01 -1.56302452e-01 2.83255070e-01 2.80947592e-02 1.08387634e-01 5.77252328e-01 1.20756581e-01 -1.33912086e-01 -4.30938482e-01 -1.06220126e+00 -6.04375660e-01 -8.64014030e-01 -8.78586918e-02 5.41613936e-01 2.28806853e-01 8.08525160e-02 -8.35953727e-02 1.24510974e-01 -1.08952677e+00 -1.49611011e-01 1.10383451e+00 9.59229350e-01 1.67996839e-01 1.45162269e-01 -3.51620913e-01 7.87143588e-01 -1.47081047e-01 -3.90566677e-01 9.30695087e-02 -3.21289241e-01 -6.58163369e-01 7.01773986e-02 2.09345609e-01 -1.64340690e-01 -7.87134290e-01 1.43979895e+00 3.36176246e-01 1.73454285e-02 5.79787731e-01 -1.37076128e+00 -1.23575962e+00 9.13499057e-01 -9.88035798e-01 2.67109185e-01 -6.47417679e-02 -1.47369560e-02 1.44529390e+00 -7.71411896e-01 4.72789943e-01 1.33022070e+00 4.95724887e-01 2.54952192e-01 -1.17517066e+00 -7.44561493e-01 5.61205506e-01 4.01552320e-01 -1.61552691e+00 1.01289377e-01 1.28418875e+00 -1.86136305e-01 6.80645227e-01 -2.79033631e-01 9.28374946e-01 9.27698553e-01 6.91280067e-02 8.78645360e-01 8.57590079e-01 1.45258978e-02 -4.20259945e-02 1.96733847e-01 4.50595587e-01 1.01021349e+00 5.63116729e-01 -4.36471701e-02 -2.15696603e-01 6.00369535e-02 8.42102528e-01 2.28780538e-01 -5.24333417e-01 -7.33298540e-01 -7.86818087e-01 9.49553013e-01 1.13439214e+00 2.54075676e-01 -2.31877476e-01 -5.24008647e-03 5.80872595e-01 8.14963356e-02 6.88513994e-01 1.78390164e-02 -1.98008642e-01 2.54448831e-01 -4.70487535e-01 -2.90279180e-01 6.26245856e-01 9.51074779e-01 1.05203927e+00 1.56061083e-01 1.31116495e-01 8.76522422e-01 7.02200711e-01 2.05201909e-01 5.51707387e-01 -3.72421354e-01 7.02148557e-01 1.40283084e+00 -6.86723292e-01 -1.39043498e+00 -4.06544328e-01 -8.33240390e-01 -1.19652140e+00 -2.56729752e-01 1.33297024e-02 1.99667718e-02 -9.07880783e-01 1.87754107e+00 4.52933282e-01 4.73067492e-01 4.20267181e-03 9.01754081e-01 1.19393015e+00 6.44342601e-01 -8.89017899e-03 -1.36417449e-01 1.02573371e+00 -1.05641544e+00 -7.62581885e-01 -1.64803520e-01 6.64411366e-01 2.46328227e-02 1.04335511e+00 1.96061973e-02 -7.28313625e-01 -3.89508456e-01 -1.26678097e+00 3.89174335e-02 -4.32425439e-01 -2.18923420e-01 7.61972785e-01 4.73576814e-01 -9.54008579e-01 4.55514491e-01 -7.18161166e-01 -2.71168381e-01 7.21287489e-01 3.36234868e-01 -4.76709485e-01 -3.23917240e-01 -1.33923388e+00 2.39650205e-01 8.25208127e-01 3.26243192e-01 -6.33611977e-01 -5.65892994e-01 -1.11419547e+00 4.11125988e-01 7.14241385e-01 -5.77863395e-01 6.02176607e-01 -7.66547263e-01 -1.07622349e+00 6.34808302e-01 1.14347629e-01 -6.55274242e-02 1.96146965e-01 2.16013849e-01 -4.24801975e-01 2.92497784e-01 8.63441899e-02 3.18957329e-01 4.93907899e-01 -1.45395029e+00 -3.03525031e-01 -7.51440227e-01 9.64717790e-02 5.98396838e-01 -6.56515062e-01 -7.34132886e-01 -7.41275430e-01 -5.30473173e-01 4.95754391e-01 -7.70937026e-01 -1.39178708e-02 -3.51211607e-01 -7.00433731e-01 -2.91025996e-01 9.69831049e-01 -4.25504357e-01 1.45131779e+00 -1.83660769e+00 4.02544677e-01 6.06633008e-01 6.70881450e-01 1.58916533e-01 -1.43736869e-01 2.74310291e-01 -8.01573247e-02 1.06924459e-01 -1.07664935e-01 -4.37328070e-02 -3.26419294e-01 2.99652815e-01 8.83639827e-02 5.59192300e-01 -1.20645307e-01 1.12684309e+00 -1.07605994e+00 -6.38686657e-01 6.36519268e-02 5.00164151e-01 -2.97768801e-01 1.42875180e-01 5.50260674e-03 2.72443801e-01 -7.76597798e-01 7.01271951e-01 8.14179301e-01 -9.72066224e-01 4.93799776e-01 -3.49231482e-01 5.95597148e-01 -1.24983005e-01 -9.98037755e-01 1.58387983e+00 -2.67174661e-01 2.31271446e-01 2.11425703e-02 -1.33735311e+00 1.05346453e+00 6.26650751e-02 6.04560971e-01 -4.62610126e-01 1.01696379e-01 -9.70237050e-03 2.12025627e-01 -3.90269965e-01 1.08140647e-01 2.72954926e-02 2.05176994e-01 3.89102221e-01 2.29120061e-01 5.35184741e-01 9.56853181e-02 6.31278157e-01 9.42964137e-01 -6.70504943e-02 4.07150626e-01 -3.06917489e-01 6.82602406e-01 -3.50210130e-01 7.54375100e-01 3.36129099e-01 -1.26886651e-01 4.53203619e-01 8.29174280e-01 -3.17608029e-01 -6.79217458e-01 -9.58078861e-01 3.99582945e-02 6.88066781e-01 5.74383497e-01 -4.83254492e-01 -5.75659931e-01 -1.27354443e+00 7.25918636e-02 2.55540043e-01 -8.30794871e-01 -6.28859460e-01 -2.96020657e-01 -9.90341783e-01 1.12486079e-01 7.78432667e-01 6.83214486e-01 -8.91416728e-01 2.34694809e-01 -5.31032123e-02 1.31175201e-02 -1.12643504e+00 -5.70353985e-01 -1.95872933e-02 -1.03145301e+00 -1.28099799e+00 -5.99204540e-01 -7.92574704e-01 1.00837433e+00 6.62585735e-01 1.10285830e+00 4.16260689e-01 -1.51130464e-03 3.60877305e-01 -4.59098876e-01 1.89066604e-02 -9.24651697e-02 5.60280323e-01 -1.21228606e-01 2.02273026e-01 5.75320959e-01 -8.61478269e-01 -5.55753648e-01 2.54646420e-01 -7.37550318e-01 6.05149493e-02 8.05688858e-01 1.11066926e+00 6.46345675e-01 2.54158765e-01 9.25521433e-01 -1.45640957e+00 6.25831902e-01 -8.92593443e-01 -5.07254183e-01 3.72019053e-01 -1.02731609e+00 7.12340623e-02 8.11508060e-01 -1.74228743e-01 -9.91507113e-01 -2.98306733e-01 5.28435968e-02 -9.32581365e-01 1.89306349e-01 1.13380754e+00 -7.50261009e-01 -6.48130402e-02 9.11097974e-02 3.60554487e-01 1.42601818e-01 -2.76097089e-01 3.12502086e-01 2.69822389e-01 2.24927872e-01 -3.21266472e-01 8.17889690e-01 4.00844872e-01 2.45633006e-01 -3.94520909e-01 -9.71401870e-01 -5.34489393e-01 -6.80498898e-01 -1.72244653e-01 7.07940280e-01 -8.63306701e-01 -6.88752353e-01 4.07351196e-01 -7.23472476e-01 -3.74150351e-02 1.58749387e-01 4.87677753e-01 -1.90991029e-01 5.92059374e-01 -6.20860338e-01 -6.70689523e-01 -2.70196974e-01 -1.07720065e+00 8.08219850e-01 3.96748424e-01 5.50181091e-01 -1.54107642e+00 -4.63927500e-02 2.63693184e-01 2.66801771e-02 4.03040126e-02 1.21223724e+00 -8.12344730e-01 -8.17465603e-01 -2.43420452e-01 -7.01591730e-01 1.70319051e-01 3.31911713e-01 -2.74583161e-01 -7.31829464e-01 -3.92072499e-01 -2.99861342e-01 -3.76665056e-01 1.03642368e+00 3.88934195e-01 1.42231333e+00 -2.82124907e-01 -6.54601932e-01 8.44632208e-01 1.61005390e+00 -7.21913353e-02 3.95767540e-01 2.01640446e-02 1.47290850e+00 5.91608822e-01 2.33649984e-01 2.17016503e-01 7.90414333e-01 3.34044516e-01 6.87695801e-01 -8.48634019e-02 -1.73660982e-02 -6.99478030e-01 -5.80317900e-02 1.23246276e+00 7.37261772e-02 -3.99850100e-01 -8.08827043e-01 3.26939285e-01 -2.01693821e+00 -7.86216497e-01 -9.37551185e-02 1.94130778e+00 2.62570947e-01 1.46455646e-01 1.19378850e-01 -4.44022939e-02 1.05349588e+00 4.85786140e-01 -9.00261521e-01 1.20752819e-01 -6.51979223e-02 -5.05951107e-01 4.82127458e-01 2.71350771e-01 -9.72794712e-01 7.64509320e-01 5.06975317e+00 8.03627253e-01 -8.93723369e-01 -5.35819978e-02 8.76573503e-01 3.65471244e-01 -7.11386859e-01 -4.89934348e-02 -6.48452342e-01 5.37220359e-01 5.19940138e-01 -2.42100805e-01 4.30558592e-01 1.08057010e+00 -1.98288038e-01 3.17516267e-01 -8.14164579e-01 9.74818408e-01 2.13534147e-01 -1.23240268e+00 3.59408200e-01 4.17451859e-01 6.94403768e-01 5.48441038e-02 7.07812607e-02 5.05138218e-01 3.84702414e-01 -1.02272534e+00 1.17511367e-02 2.93275028e-01 6.09704435e-01 -9.03150260e-01 9.51516211e-01 3.78205299e-01 -1.76993167e+00 -9.83324721e-02 -5.09434283e-01 3.94370019e-01 4.30970117e-02 5.22075951e-01 -5.79008996e-01 1.06326270e+00 6.29204035e-01 1.28554952e+00 -7.07386851e-01 7.44242132e-01 -2.17901811e-01 6.28292322e-01 -1.11302190e-01 -1.52202398e-01 4.30133820e-01 -6.41830802e-01 5.61898351e-01 7.32376695e-01 1.87485516e-01 4.29161824e-02 5.17624259e-01 9.40958440e-01 -3.43534648e-01 2.93301672e-01 -7.72169173e-01 -1.61055654e-01 4.84302014e-01 1.55735362e+00 -9.24654543e-01 -1.76244706e-01 -8.40831757e-01 5.33599138e-01 1.01952362e+00 4.64643300e-01 -6.70951962e-01 -3.05547595e-01 1.37381703e-01 1.28224015e-01 1.55937016e-01 5.73982997e-03 -9.13140252e-02 -1.37908864e+00 8.91439840e-02 -5.41144192e-01 8.44911754e-01 -5.64949393e-01 -1.73619771e+00 5.65256000e-01 7.76343048e-02 -1.12736213e+00 1.60062797e-02 -4.58527476e-01 -8.00070226e-01 7.32158184e-01 -1.58836663e+00 -1.32929432e+00 -6.08382761e-01 3.95184010e-01 2.65324175e-01 -3.89781952e-01 4.56070215e-01 2.87280798e-01 -9.25776541e-01 7.15692222e-01 9.48847309e-02 2.91476578e-01 2.22390160e-01 -1.50060761e+00 1.21461585e-01 5.45485258e-01 1.23873174e-01 6.28192902e-01 5.45830205e-02 -9.09660280e-01 -1.61619151e+00 -1.34528112e+00 8.59308094e-02 3.00904047e-02 6.74501598e-01 -3.57531965e-01 -1.47569501e+00 7.68007100e-01 -1.65299941e-02 5.05169213e-01 6.92937672e-01 4.14697289e-01 -4.97878790e-01 -2.58241415e-01 -8.56304407e-01 3.56910706e-01 1.21945095e+00 -4.66405272e-01 -2.22102776e-01 2.15766132e-01 9.52426136e-01 -2.89666235e-01 -9.47173715e-01 5.50748229e-01 3.43854785e-01 -8.86570156e-01 8.21051538e-01 -5.61819196e-01 4.80883926e-01 -1.80797860e-01 2.86994632e-02 -1.48910797e+00 -5.60595989e-01 -1.20153978e-01 -4.86139566e-01 1.41941738e+00 3.54283750e-01 -8.66914749e-01 1.23835731e+00 3.47797036e-01 -1.41477659e-01 -1.26240885e+00 -7.58228481e-01 -7.05591857e-01 -9.39295068e-02 -5.08585721e-02 7.58959234e-01 1.14291704e+00 1.00275971e-01 8.67087305e-01 -3.22722763e-01 2.86961347e-01 7.69043863e-01 2.02696547e-01 5.82262695e-01 -1.46323562e+00 -3.07228625e-01 -5.71432590e-01 -7.15149283e-01 -1.04472315e+00 6.24071300e-01 -1.45869327e+00 -3.42870474e-01 -1.83579695e+00 6.19470060e-01 -4.99471456e-01 -6.05303168e-01 1.50366277e-01 -6.14554286e-01 -6.38420731e-02 -2.05023661e-01 2.04121947e-01 -7.45238185e-01 1.05509269e+00 1.41661692e+00 -2.42557570e-01 -1.73207685e-01 -2.22486347e-01 -8.77497852e-01 6.93961322e-01 6.35128319e-01 -2.27311715e-01 -1.02147818e+00 -3.91187906e-01 1.55743510e-01 1.36498198e-01 2.81990111e-01 -8.12889814e-01 2.61553079e-01 -2.02476531e-02 3.59120458e-01 -5.91340005e-01 1.36088103e-01 -9.19262528e-01 -1.29998296e-01 2.67801583e-01 -2.79012382e-01 2.74896566e-02 -2.23111495e-01 1.32361388e+00 -2.95707673e-01 -6.13438003e-02 5.93898356e-01 -2.03468859e-01 -7.53111124e-01 1.02199817e+00 4.13124651e-01 2.20535353e-01 9.69474435e-01 -2.42000952e-01 -3.49032670e-01 -4.36258376e-01 -4.71042186e-01 6.15069747e-01 3.85850549e-01 4.09754187e-01 7.28953302e-01 -1.65025377e+00 -4.97423559e-01 1.88648164e-01 3.15464020e-01 2.62337655e-01 5.76569736e-01 6.70183718e-01 -1.28300861e-01 1.82720736e-01 9.50459018e-02 -7.73807526e-01 -9.42462325e-01 8.86882544e-01 3.26049775e-01 -4.79182482e-01 -8.81639421e-01 7.00182557e-01 6.78387821e-01 -4.59979951e-01 9.75852013e-02 1.61450699e-01 -5.87993503e-01 -1.65357769e-01 2.16936529e-01 1.96733937e-01 -2.59436011e-01 -7.85870373e-01 -1.86135679e-01 6.29274487e-01 -3.70681822e-01 2.69749194e-01 1.17199993e+00 -2.79472977e-01 -7.52864778e-02 5.29507697e-01 1.50209844e+00 -3.65256101e-01 -1.11029613e+00 -5.87030351e-01 -1.69005424e-01 -5.63076377e-01 2.48673916e-01 -3.41679871e-01 -1.68983757e+00 9.88899589e-01 2.69915581e-01 4.28219914e-01 1.04932320e+00 1.54586822e-01 7.65696645e-01 4.17105764e-01 2.57578641e-01 -8.85435164e-01 2.62852937e-01 1.75880149e-01 5.92738867e-01 -1.49424028e+00 1.38676643e-01 -8.40776503e-01 -6.16294861e-01 9.78599250e-01 1.20002913e+00 -1.82801917e-01 1.05662560e+00 -1.89774454e-01 -2.77615279e-01 -4.95890290e-01 -6.34818196e-01 -2.33251080e-02 4.36788619e-01 6.28776789e-01 3.34297746e-01 -2.89674923e-02 7.22182766e-02 7.82609522e-01 2.73373038e-01 -3.20014626e-01 2.26283878e-01 4.54957634e-01 -2.07660407e-01 -8.39736938e-01 1.84770599e-01 9.63112354e-01 -2.26034939e-01 -2.76180655e-02 -3.38103503e-01 8.94169331e-01 -3.20827246e-01 5.11354327e-01 5.17712859e-03 -5.75123250e-01 4.85641286e-02 -3.75436366e-01 2.19947830e-01 -6.75903320e-01 -8.38675499e-02 -1.95938852e-02 -2.43905604e-01 -3.69260997e-01 -3.79664332e-01 -3.51688802e-01 -1.36566973e+00 -2.01054782e-01 -5.22978425e-01 1.97078958e-01 1.76711082e-01 8.03507626e-01 4.69093651e-01 6.88695014e-01 7.89132178e-01 -4.80559200e-01 -3.70286167e-01 -8.61558139e-01 -8.84025693e-01 4.22045469e-01 -1.06010037e-02 -9.57930446e-01 -3.81860405e-01 -4.50007290e-01]

[7.372117519378662, 6.193046569824219]

d00b7803-33a6-4671-9f15-a3917459586c

understanding-hindsight-goal-relabeling

2209.13046

https://arxiv.org/abs/2209.13046v2

https://arxiv.org/pdf/2209.13046v2.pdf

Understanding Hindsight Goal Relabeling from a Divergence Minimization Perspective

Hindsight goal relabeling has become a foundational technique in multi-goal reinforcement learning (RL). The essential idea is that any trajectory can be seen as a sub-optimal demonstration for reaching its final state. Intuitively, learning from those arbitrary demonstrations can be seen as a form of imitation learning (IL). However, the connection between hindsight goal relabeling and imitation learning is not well understood. In this paper, we propose a novel framework to understand hindsight goal relabeling from a divergence minimization perspective. Recasting the goal reaching problem in the IL framework not only allows us to derive several existing methods from first principles, but also provides us with the tools from IL to improve goal reaching algorithms. Experimentally, we find that under hindsight relabeling, Q-learning outperforms behavioral cloning (BC). Yet, a vanilla combination of both hurts performance. Concretely, we see that the BC loss only helps when selectively applied to actions that get the agent closer to the goal according to the Q-function. Our framework also explains the puzzling phenomenon wherein a reward of (-1, 0) results in significantly better performance than a (0, 1) reward for goal reaching.

['Bradly C. Stadie', 'Lunjun Zhang']

2022-09-26

['multi-goal-reinforcement-learning']

['methodology']

[-1.17965210e-02 1.28512517e-01 -4.38304037e-01 -1.61254965e-02 -6.26973033e-01 -6.36981130e-01 4.81806666e-01 -9.72292125e-02 -6.18578851e-01 9.04340863e-01 2.28311330e-01 -3.37908447e-01 -4.24288541e-01 -5.04186332e-01 -7.44761944e-01 -9.80494082e-01 -2.16095060e-01 1.48118839e-01 -5.11936890e-03 -5.89839220e-01 4.37566638e-01 4.50353742e-01 -1.24163568e+00 -2.53789723e-01 9.23097432e-01 4.21976596e-01 3.63610119e-01 7.21911550e-01 3.05253923e-01 9.03605402e-01 -3.94201934e-01 -5.28742187e-02 5.12120366e-01 -9.95918691e-01 -9.23638523e-01 -2.90375799e-01 2.67227292e-01 -5.24129212e-01 -3.11010569e-01 1.17366230e+00 3.40899050e-01 4.66342747e-01 6.07373357e-01 -1.61817789e+00 -4.64276850e-01 6.89971566e-01 -5.23165643e-01 -1.24521822e-01 3.60986352e-01 3.49660724e-01 1.30342567e+00 -8.07812288e-02 5.37396491e-01 1.32478929e+00 4.61285532e-01 8.42782915e-01 -1.27044797e+00 -3.98666739e-01 2.31823027e-01 2.36484051e-01 -8.06957662e-01 -1.66972950e-02 7.02283323e-01 -3.67475450e-01 6.70812368e-01 -3.38911451e-02 7.43964016e-01 1.01719046e+00 4.96068507e-01 9.60754454e-01 1.18806219e+00 -3.98542762e-01 3.80273134e-01 -3.71196568e-01 2.18931530e-02 7.54310429e-01 1.49832413e-01 6.71270072e-01 -4.06650811e-01 7.01752603e-02 9.04047787e-01 8.80539194e-02 -2.14122921e-01 -1.06727982e+00 -1.03496075e+00 1.07461715e+00 5.24894357e-01 2.92361170e-01 -5.03518343e-01 7.89547443e-01 1.73585773e-01 1.01702583e+00 -3.82821858e-01 8.78904700e-01 -3.35533440e-01 -3.59255940e-01 -4.39814031e-01 5.32641053e-01 5.94550252e-01 6.71434522e-01 8.33603382e-01 1.74364850e-01 -1.92102268e-01 4.23826277e-01 2.71313339e-01 3.63938987e-01 4.85652119e-01 -1.55078351e+00 2.56867260e-01 3.72452080e-01 3.75891626e-01 -4.89024878e-01 -5.65190494e-01 -3.03560883e-01 -2.39465803e-01 9.64116156e-01 8.98876548e-01 -4.07273471e-01 -5.27769029e-01 2.33095860e+00 4.93104309e-02 -6.29401058e-02 1.92927212e-01 1.04188001e+00 1.33038074e-01 3.95381749e-01 3.41330701e-03 -3.28570276e-01 6.87545121e-01 -1.18064988e+00 -3.80956382e-01 -2.54760653e-01 1.04295349e+00 -1.67916149e-01 1.35239434e+00 4.93675113e-01 -7.68804491e-01 -2.67912626e-01 -1.06118059e+00 1.50685742e-01 4.80295941e-02 -4.96388257e-01 5.86025000e-01 5.20448983e-01 -1.09823525e+00 1.09017360e+00 -6.76483393e-01 -6.56763554e-01 1.81594700e-01 2.02626258e-01 -2.93839186e-01 3.59590910e-02 -9.05022740e-01 1.26320732e+00 2.20949098e-01 -4.66430366e-01 -1.42103553e+00 -2.92289853e-01 -7.34213233e-01 -6.60515651e-02 6.74081445e-01 -6.03768945e-01 1.42599392e+00 -9.85659182e-01 -1.73533034e+00 4.90064114e-01 1.02430880e-01 -5.82610190e-01 6.20362639e-01 -2.31844723e-01 2.12326676e-01 1.15110837e-01 1.87604412e-01 6.20223701e-01 1.11163676e+00 -1.14513767e+00 -6.40009642e-01 -5.04454672e-01 5.11862218e-01 5.42195618e-01 -5.72649650e-02 -1.56481862e-01 2.98575222e-01 -3.81401122e-01 -1.32941619e-01 -1.05004585e+00 -4.84503180e-01 5.90703115e-02 -1.67241648e-01 -4.81038928e-01 3.79542887e-01 -2.98087031e-01 7.62312114e-01 -2.04320335e+00 6.74111664e-01 -4.56007607e-02 2.91644692e-01 -1.10069104e-01 -3.59886467e-01 7.33120978e-01 -7.86647201e-03 -1.90253004e-01 -2.23082863e-02 -2.15490907e-02 3.40408474e-01 4.04284418e-01 -3.28161061e-01 7.06373632e-01 -4.00617689e-01 1.16136205e+00 -1.29404819e+00 -2.66636252e-01 2.09274352e-01 -5.33859394e-02 -8.83268058e-01 1.95703536e-01 -2.11511940e-01 7.44723141e-01 -5.46932817e-01 2.61655927e-01 1.00195304e-01 -1.53974611e-02 2.07846090e-01 5.23664296e-01 -2.23397598e-01 -1.23736290e-02 -9.78450060e-01 1.78682709e+00 -2.71505624e-01 5.24039090e-01 1.20512456e-01 -1.32425892e+00 5.77441335e-01 9.01081879e-03 4.18152481e-01 -7.29231954e-01 1.59966260e-01 1.38786569e-01 3.45083237e-01 -4.52993900e-01 4.00682598e-01 -4.41162914e-01 -2.53336787e-01 6.76231384e-01 7.38903433e-02 -3.62306356e-01 1.60608381e-01 -2.96741445e-03 1.05385613e+00 5.54500461e-01 5.98009288e-01 -1.53747573e-01 2.16004029e-01 1.94631383e-01 4.41441745e-01 1.25046110e+00 -6.66228175e-01 1.28482327e-01 7.84669220e-01 -4.07802612e-02 -9.68687892e-01 -1.25566006e+00 2.80747294e-01 1.49967265e+00 3.33482623e-01 -2.31680498e-01 -5.46370685e-01 -8.19428205e-01 2.75197625e-01 8.33641648e-01 -7.36853182e-01 -4.88648027e-01 -7.14942932e-01 -1.36089504e-01 4.69736487e-01 4.18770880e-01 2.08799332e-01 -1.27002299e+00 -1.04938579e+00 1.40478089e-01 -2.29295902e-02 -4.65821683e-01 -5.08378983e-01 3.33305538e-01 -1.01992059e+00 -1.10181212e+00 -9.62997496e-01 -6.92531884e-01 3.09755534e-01 3.95481110e-01 6.75497830e-01 5.49418889e-02 4.44317386e-02 7.61493325e-01 -4.86856759e-01 -1.50644392e-01 -5.48155904e-01 -2.80129939e-01 4.01632071e-01 -4.82395709e-01 1.26002878e-01 -7.39097238e-01 -4.12724704e-01 2.21189350e-01 -5.39123952e-01 -1.25790268e-01 4.61168885e-01 9.62253809e-01 2.23102570e-01 -2.46832788e-01 7.33886182e-01 -1.74904481e-01 9.29373920e-01 -2.85567880e-01 -6.96240723e-01 1.69148788e-01 -7.33554840e-01 5.86090326e-01 9.65235054e-01 -5.59700131e-01 -6.32508159e-01 -5.98653890e-02 -8.12869072e-02 -2.26006180e-01 -1.47122055e-01 2.43310601e-01 9.04104933e-02 -1.47834048e-01 5.99829733e-01 4.29489404e-01 2.76568502e-01 -4.62519586e-01 8.43395472e-01 3.35789949e-01 2.71179765e-01 -7.39531934e-01 6.02788448e-01 4.53891337e-01 2.38963589e-01 -6.53203547e-01 -6.85292959e-01 -3.27718705e-01 -4.84067917e-01 -4.10573959e-01 8.04306626e-01 -5.74550927e-01 -1.19295299e+00 2.22539365e-01 -6.43162251e-01 -9.13591921e-01 -5.89837372e-01 7.01330841e-01 -1.46871054e+00 4.88767505e-01 -3.53923172e-01 -8.24085474e-01 1.23274386e-01 -1.13932061e+00 5.40407717e-01 3.50263178e-01 -2.29064718e-01 -7.92145073e-01 2.49175683e-01 2.71447445e-03 2.75023967e-01 6.89505562e-02 9.61891115e-01 -5.35929799e-01 -3.16086620e-01 2.43472576e-01 1.80509374e-01 1.58004954e-01 -3.71199511e-02 -5.43421984e-01 -4.01017964e-01 -6.24504685e-01 1.55660018e-01 -7.20118344e-01 6.97078586e-01 5.07721961e-01 6.32558763e-01 -2.64210612e-01 -1.34687811e-01 4.61683244e-01 1.51419449e+00 3.99701446e-01 3.56469005e-01 5.82217574e-01 4.35334176e-01 5.27739286e-01 8.00900757e-01 6.40477121e-01 2.57412493e-01 6.24452651e-01 7.85036087e-01 4.31826919e-01 7.71979615e-02 -6.29357159e-01 6.48971736e-01 1.46852419e-01 -5.03214188e-02 1.80397213e-01 -5.40146291e-01 4.54221964e-01 -2.12914467e+00 -1.27165365e+00 6.73164949e-02 2.36394024e+00 8.51067305e-01 -6.46295100e-02 6.10973895e-01 -1.52048945e-01 4.25932467e-01 6.86483681e-02 -9.02683854e-01 -3.81738663e-01 1.35319814e-01 -4.89443615e-02 4.22951192e-01 7.29421139e-01 -6.24458551e-01 1.10299659e+00 6.89975357e+00 7.06385195e-01 -1.01880765e+00 -1.46091972e-02 -5.80820907e-03 -1.41744301e-01 -5.26509099e-02 1.97510406e-01 -6.14642441e-01 2.20027253e-01 4.45769608e-01 -5.89841723e-01 9.68262434e-01 1.19165885e+00 4.38991845e-01 -3.10344070e-01 -1.41893375e+00 8.91615510e-01 -7.42224604e-02 -8.71628582e-01 -1.73710331e-01 1.56072631e-01 6.24567389e-01 7.37721380e-03 1.78403780e-02 6.74042583e-01 9.91568923e-01 -1.13327527e+00 7.50643849e-01 1.82359055e-01 3.41324270e-01 -7.76479542e-01 1.72716767e-01 8.63191068e-01 -7.73974001e-01 -5.00687957e-01 -3.92668575e-01 -6.27980173e-01 1.56129792e-01 -2.15383232e-01 -6.10095739e-01 4.33363497e-01 2.56411135e-01 6.24789715e-01 -1.66079774e-01 1.03430307e+00 -6.39370739e-01 2.43605465e-01 -1.21980205e-01 -4.30176020e-01 6.20775223e-01 -6.21885836e-01 7.18800664e-01 5.49244046e-01 1.73838988e-01 -1.32999390e-01 2.40206674e-01 7.90298939e-01 2.94982553e-01 7.05030188e-02 -8.91042173e-01 1.41521525e-02 -3.16241477e-03 8.56204391e-01 -3.85925710e-01 -1.97821274e-01 -1.60896048e-01 1.00851762e+00 7.49532521e-01 4.51365858e-01 -8.75151992e-01 -2.72840321e-01 9.67616320e-01 -2.83190131e-01 1.97177008e-01 -3.99908751e-01 -7.49499202e-02 -1.02920163e+00 -2.36368760e-01 -8.35495353e-01 3.20646912e-01 -7.01315165e-01 -8.85205388e-01 4.12183180e-02 -1.98175348e-02 -1.21359050e+00 -5.98529041e-01 -4.83042121e-01 -5.56639791e-01 6.11028373e-01 -1.37408829e+00 -6.81699574e-01 -1.62941702e-02 6.08640492e-01 4.59066451e-01 1.00457408e-02 6.31617367e-01 -1.84062347e-01 -2.74172753e-01 4.71896797e-01 4.02582109e-01 -1.13035388e-01 5.80702841e-01 -1.48624825e+00 -9.05507877e-02 6.85307801e-01 1.74145848e-01 6.24417365e-01 1.07076311e+00 -3.70133132e-01 -1.42990923e+00 -6.19656622e-01 5.62043309e-01 -3.72980684e-01 8.68177116e-01 1.12245470e-01 -4.59627271e-01 8.32035303e-01 1.11052930e-01 -5.20627677e-01 2.25938603e-01 1.83856711e-01 -2.40921378e-01 1.64606929e-01 -9.85226572e-01 1.30627811e+00 1.23562765e+00 -2.80425191e-01 -1.05756748e+00 2.81720936e-01 6.28855109e-01 -1.62805952e-02 -5.51055551e-01 2.44836453e-02 7.45488584e-01 -1.08686924e+00 9.38915730e-01 -1.17253458e+00 4.48530138e-01 -1.42552182e-01 -1.25457034e-01 -1.86805904e+00 -4.57202941e-01 -1.01708066e+00 4.91419714e-03 4.57780451e-01 9.26754549e-02 -6.82347715e-01 7.28431940e-01 5.28006675e-03 -1.21148832e-01 -5.89792550e-01 -1.07966304e+00 -1.44328463e+00 8.19367528e-01 -3.69569361e-01 2.16596007e-01 6.89516723e-01 7.30205774e-01 2.88192391e-01 -6.89400077e-01 -3.10358703e-01 8.65607500e-01 4.34722304e-01 9.70918953e-01 -9.75456119e-01 -5.47677457e-01 -7.62240648e-01 -7.92010128e-02 -1.68854225e+00 3.00248742e-01 -9.57202494e-01 2.38882989e-01 -1.59893477e+00 1.50697440e-01 -3.98682475e-01 -2.73733526e-01 4.50380981e-01 -7.21973274e-03 -1.66549504e-01 5.88855386e-01 2.00301662e-01 -6.66333556e-01 6.26846313e-01 1.65830982e+00 1.99757472e-01 -4.33347404e-01 6.33052811e-02 -8.22224379e-01 7.81627655e-01 1.16576970e+00 -7.44901836e-01 -3.37399691e-01 -1.53919190e-01 2.49876544e-01 4.62538511e-01 4.27107543e-01 -7.45621204e-01 1.12935051e-01 -6.27665997e-01 -2.55490661e-01 -2.06300750e-01 3.45517278e-01 -5.29538512e-01 -2.47722343e-01 1.11107731e+00 -5.50486207e-01 -2.99004093e-02 -2.52434462e-01 6.83652222e-01 2.45122120e-01 -6.00584507e-01 9.63628829e-01 -3.05549741e-01 -8.94253492e-01 -5.99679500e-02 -6.62838459e-01 4.22532320e-01 1.13792622e+00 -2.64619559e-01 -1.39566839e-01 -7.74112165e-01 -7.34701097e-01 5.05542040e-01 6.94526911e-01 3.99937332e-01 3.88912618e-01 -1.26625896e+00 -6.29970372e-01 1.99678037e-02 6.28302470e-02 -6.60176933e-01 7.89524540e-02 9.82400060e-01 -1.71919435e-01 4.97858465e-01 -5.58681548e-01 -4.02120352e-01 -1.07524228e+00 7.57025123e-01 4.35267031e-01 -3.56151432e-01 -7.70787418e-01 6.84409499e-01 3.05195063e-01 -5.00044525e-01 3.93018156e-01 -3.26577932e-01 -1.91079453e-01 -2.10049450e-01 3.59518558e-01 5.11453688e-01 -6.76480651e-01 -3.21580231e-01 -1.12384364e-01 5.81916332e-01 1.62065506e-01 -5.25825381e-01 1.20217407e+00 -1.55660540e-01 2.24937811e-01 5.07602274e-01 9.68729258e-01 -2.03841880e-01 -1.60755253e+00 1.69292074e-02 9.45594534e-02 -4.06194270e-01 -3.87572557e-01 -6.25677764e-01 -4.54795271e-01 8.47850561e-01 2.83401102e-01 2.07495108e-01 6.86547577e-01 7.97896981e-02 4.11312610e-01 7.61919558e-01 7.91371644e-01 -1.21215451e+00 5.93448579e-01 6.99599862e-01 9.96117353e-01 -1.22457862e+00 -2.05488324e-01 2.69841313e-01 -8.69321764e-01 8.76015544e-01 5.42330265e-01 -4.57120389e-01 2.25153431e-01 -4.12355252e-02 -9.91127118e-02 -1.80639494e-02 -5.58018088e-01 -5.38746893e-01 -2.65264779e-01 8.37099254e-01 -3.87317315e-02 1.28020972e-01 -5.12786031e-01 3.58585000e-01 -2.51976907e-01 3.52205783e-02 6.91562772e-01 9.75212336e-01 -9.96121287e-01 -1.17482615e+00 -3.12291801e-01 2.57212043e-01 -2.38672853e-01 1.31994739e-01 -2.92312175e-01 8.82767379e-01 -3.77910256e-01 9.78008747e-01 -2.04498604e-01 -2.37748116e-01 1.83027327e-01 -7.82204047e-03 9.52467859e-01 -6.19074762e-01 -3.49846274e-01 -1.56108707e-01 -2.73491204e-01 -7.58917391e-01 -1.82847276e-01 -6.56210840e-01 -1.47943664e+00 -4.02018547e-01 -4.01726514e-02 2.44214326e-01 4.25063848e-01 9.93136227e-01 -4.82602492e-02 3.63468528e-01 6.00413024e-01 -6.07395411e-01 -1.27168548e+00 -7.38701463e-01 -7.95926988e-01 3.72392684e-01 5.83876848e-01 -8.00378323e-01 -6.90250158e-01 -3.73813391e-01]

[4.064385890960693, 1.7886152267456055]

43248067-ae3d-4ad0-b72f-451997173ef5

cvpr19-tracking-and-detection-challenge-how

1906.04567

https://arxiv.org/abs/1906.04567v1

https://arxiv.org/pdf/1906.04567v1.pdf

CVPR19 Tracking and Detection Challenge: How crowded can it get?

Standardized benchmarks are crucial for the majority of computer vision applications. Although leaderboards and ranking tables should not be over-claimed, benchmarks often provide the most objective measure of performance and are therefore important guides for research. The benchmark for Multiple Object Tracking, MOTChallenge, was launched with the goal to establish a standardized evaluation of multiple object tracking methods. The challenge focuses on multiple people tracking, since pedestrians are well studied in the tracking community, and precise tracking and detection has high practical relevance. Since the first release, MOT15, MOT16 and MOT17 have tremendously contributed to the community by introducing a clean dataset and precise framework to benchmark multi-object trackers. In this paper, we present our CVPR19 benchmark, consisting of 8 new sequences depicting very crowded challenging scenes. The benchmark will be presented at the 4th BMTT MOT Challenge Workshop at the Computer Vision and Pattern Recognition Conference (CVPR) 2019, and will evaluate the state-of-the-art in multiple object tracking whend handling extremely crowded scenarios.

['Laura Leal-Taixe', 'Anton Milan', 'Javen Shi', 'Stefan Roth', 'Konrad Schindler', 'Daniel Cremers', 'Patrick Dendorfer', 'Ian Reid', 'Hamid Rezatofighi']

2019-06-10

['multiple-people-tracking']

['computer-vision']

[-1.75987363e-01 -5.33019423e-01 -3.13825086e-02 -5.47250314e-03 -5.89579225e-01 -4.79529202e-01 7.39685893e-01 1.53063118e-01 -7.39525259e-01 8.69229138e-01 -1.62496741e-04 2.42563695e-01 9.89674926e-02 -1.22748390e-01 -7.04504430e-01 -5.47965169e-01 -1.51615977e-01 7.81122446e-01 9.22202229e-01 1.31501462e-02 -1.80378631e-02 3.60756636e-01 -2.04093456e+00 1.53674722e-01 2.83457428e-01 7.00390995e-01 1.72167495e-01 1.01743555e+00 2.25593328e-01 7.45634139e-01 -9.32884932e-01 -6.81728184e-01 4.95239943e-01 -1.42209217e-01 -4.50946331e-01 -1.33373186e-01 1.41432035e+00 -1.15998387e-01 -1.06494635e-01 1.05019808e+00 7.37350047e-01 1.15656093e-01 3.65570426e-01 -1.79662168e+00 -1.01073578e-01 3.38709950e-01 -7.46201158e-01 8.34690034e-01 2.84502804e-01 4.31030124e-01 9.90964234e-01 -6.18309379e-01 8.10076118e-01 1.37647879e+00 9.84453976e-01 6.68863058e-01 -1.04943097e+00 -7.25976586e-01 2.63374686e-01 5.37242293e-01 -1.21585310e+00 -6.24385953e-01 5.26130795e-02 -9.07240510e-01 6.07731342e-01 6.44345939e-01 8.35654855e-01 1.17696488e+00 1.29986340e-02 1.12159145e+00 9.04108942e-01 -2.07917795e-01 -1.61158860e-01 2.70043015e-01 3.74619126e-01 5.41357994e-01 9.35658693e-01 4.04083461e-01 -4.11140859e-01 -1.45338207e-01 2.83926845e-01 -1.20838165e-01 -1.63080588e-01 -6.94555402e-01 -1.53352654e+00 7.01619506e-01 2.84634471e-01 2.70753980e-01 -1.70593917e-01 3.45852256e-01 6.39528096e-01 5.83685152e-02 1.04772843e-01 1.55934706e-01 -6.94698691e-02 -1.79371640e-01 -8.66287827e-01 8.96929502e-01 5.38331628e-01 9.79626596e-01 1.97642297e-02 -9.21241194e-02 -7.21709073e-01 6.13611102e-01 5.04584193e-01 8.90704572e-01 6.97909892e-02 -9.15924191e-01 3.98553878e-01 2.74569958e-01 6.32543147e-01 -7.91721404e-01 -4.36827451e-01 -2.81316161e-01 -5.11193752e-01 6.85149431e-01 8.51718247e-01 -2.93405801e-01 -5.44878066e-01 1.49739885e+00 4.51128393e-01 4.33396786e-01 -1.98331356e-01 9.02946234e-01 1.18614101e+00 2.38441005e-01 2.51010597e-01 -2.51377225e-01 1.45828676e+00 -1.14544439e+00 -5.11499822e-01 -2.32813999e-01 2.96776116e-01 -8.93259108e-01 4.47758108e-01 1.96182325e-01 -1.07455981e+00 -7.49223769e-01 -9.13344741e-01 4.45605278e-01 -2.11941794e-01 -1.31138518e-01 1.54187545e-01 7.97830582e-01 -1.06105804e+00 2.97819078e-01 -6.64985299e-01 -8.27911794e-01 4.51198608e-01 1.65517837e-01 -1.83991939e-01 -4.43411060e-02 -7.66381621e-01 1.35041511e+00 2.65890986e-01 -1.01608723e-01 -1.04457402e+00 -7.49680400e-01 -4.67403799e-01 -4.28807437e-01 4.44685191e-01 -6.72299802e-01 1.32721663e+00 -3.36800277e-01 -8.36367011e-01 1.10710573e+00 -6.70862431e-03 -8.58566225e-01 1.10534501e+00 -3.55138630e-01 -6.69593036e-01 -2.32965350e-01 4.74232048e-01 7.54909217e-01 7.27954745e-01 -1.29859757e+00 -1.25295341e+00 -8.04572254e-02 -1.37482435e-01 -5.38927242e-02 1.35985836e-01 6.14326239e-01 -5.75788856e-01 -3.49681675e-01 -8.48324358e-01 -9.98791277e-01 -3.45167577e-01 7.32014030e-02 -2.70741224e-01 -2.47486308e-01 1.07924426e+00 -2.47326151e-01 8.87879550e-01 -1.95460355e+00 5.61734661e-02 -4.05952305e-01 5.26912749e-01 6.91540241e-01 -7.37714097e-02 -1.46839113e-04 2.81255722e-01 -2.87906051e-01 2.96667606e-01 -6.64348066e-01 1.69966817e-01 7.81798586e-02 -8.54482353e-02 8.18251789e-01 -7.59398043e-02 8.53731871e-01 -1.12517488e+00 -8.50891590e-01 6.31376445e-01 2.69935429e-01 -9.53287855e-02 2.79049259e-02 -2.52052248e-01 5.30744910e-01 -2.12089717e-01 4.62809771e-01 6.01335049e-01 -3.31288159e-01 -2.74778247e-01 -1.20026186e-01 -4.71636683e-01 -5.44731915e-01 -1.56423521e+00 1.08013368e+00 2.34059453e-01 1.12795889e+00 9.46201980e-02 -4.18723226e-01 5.53955138e-01 2.69831479e-01 7.58426011e-01 -5.10057628e-01 1.57650962e-01 -5.07078134e-02 1.48624435e-01 -1.79041103e-01 8.19408953e-01 2.80803025e-01 1.21690795e-01 -9.62112620e-02 -1.51484653e-01 2.06532240e-01 9.81099546e-01 1.23982884e-01 1.09369922e+00 1.01051569e-01 4.55311358e-01 -3.29120785e-01 6.90729380e-01 2.86048889e-01 6.10785007e-01 1.08720708e+00 -9.47594881e-01 5.85714459e-01 -7.07376972e-02 -7.80877471e-01 -8.95867467e-01 -1.10014498e+00 -2.09599718e-01 1.24533343e+00 3.45838398e-01 -5.53106010e-01 -4.81466085e-01 -5.27366877e-01 1.39866337e-01 2.34774545e-01 -7.52164900e-01 4.51957673e-01 -7.19182312e-01 -7.73682356e-01 6.73886001e-01 4.27075118e-01 4.19433028e-01 -1.10506868e+00 -1.22597682e+00 1.64979383e-01 -1.74021214e-01 -1.44198167e+00 -5.28250992e-01 -2.52592295e-01 -3.65426093e-01 -1.61861646e+00 -1.10161185e+00 -5.31271160e-01 1.21727347e-01 6.35312259e-01 1.35992956e+00 3.06685179e-01 -6.49702132e-01 6.66544259e-01 -1.50041878e-01 -9.31099296e-01 -2.81292796e-01 -1.78693429e-01 2.63059884e-01 1.26881674e-02 3.27018678e-01 1.41606137e-01 -4.39942747e-01 7.36558855e-01 -3.05557549e-01 -2.44073391e-01 3.55475992e-01 4.48578089e-01 4.89705920e-01 -4.80194211e-01 -4.48761834e-03 -5.09757042e-01 2.94294685e-01 -1.24899419e-02 -1.21679389e+00 4.29658681e-01 -3.34627032e-02 -2.15606973e-01 1.52725145e-01 -4.00705844e-01 -7.02391088e-01 1.54982150e-01 1.32655993e-01 -4.46378797e-01 -2.97902197e-01 -5.00531197e-01 1.00005209e-01 -3.86059970e-01 8.38055253e-01 -9.05076340e-02 -4.61404115e-01 -4.03614700e-01 2.44303226e-01 1.87784746e-01 7.09937274e-01 -3.77036065e-01 9.98238981e-01 6.40791535e-01 2.40058914e-01 -1.10868764e+00 -8.33266973e-01 -1.05552125e+00 -5.62654555e-01 -7.08671927e-01 1.11922538e+00 -1.06076121e+00 -1.06198096e+00 4.77566004e-01 -1.36345220e+00 -9.86163095e-02 -2.97952026e-01 3.96123588e-01 -2.36080781e-01 3.31125885e-01 -1.05457895e-01 -9.24903631e-01 -1.87986016e-01 -1.18081725e+00 1.15771186e+00 3.88658792e-01 -2.13945299e-01 -9.37145889e-01 5.27199447e-01 3.28650087e-01 4.26910549e-01 5.97073555e-01 -1.77035317e-01 -6.16447508e-01 -8.35839331e-01 -2.26655632e-01 -3.18159431e-01 4.64245342e-02 -3.20090175e-01 1.78787485e-01 -1.02142966e+00 -5.35421014e-01 -6.74839556e-01 -1.39819354e-01 9.98306155e-01 6.80812240e-01 5.57348967e-01 3.37485045e-01 -8.26730788e-01 2.18158275e-01 1.31544185e+00 -5.83501980e-02 3.28771502e-01 6.42504692e-01 7.28688896e-01 3.38938236e-01 5.19304216e-01 4.32240427e-01 4.88309741e-01 1.18639064e+00 4.83682722e-01 2.23053858e-01 -6.07532918e-01 3.56062144e-01 4.64521885e-01 3.73297483e-01 -4.57764030e-01 -3.63462061e-01 -1.03141630e+00 4.81256902e-01 -1.94168067e+00 -1.41936862e+00 -7.70285249e-01 2.30265260e+00 8.89100879e-02 3.78253222e-01 8.77797782e-01 -1.28723457e-01 9.76457238e-01 2.52985388e-01 -1.95339054e-01 4.01815116e-01 -1.74786821e-01 -3.59997153e-01 5.84924698e-01 4.59637016e-01 -1.50964487e+00 6.73714340e-01 6.87411261e+00 6.18659377e-01 -5.62900782e-01 3.81994039e-01 -1.10968739e-01 -2.43253648e-01 6.80172980e-01 -4.29217130e-01 -1.70555067e+00 5.96175849e-01 9.79735851e-01 -2.82944739e-01 -7.39762112e-02 7.89603353e-01 -1.37423277e-01 -1.94754884e-01 -1.07639205e+00 1.16638744e+00 4.27523032e-02 -1.30079842e+00 -3.04805398e-01 1.85910787e-03 6.94477975e-01 6.10520244e-01 8.89984220e-02 4.09781545e-01 6.54584587e-01 -7.96741784e-01 1.05724752e+00 5.37008882e-01 2.61705101e-01 -2.71854669e-01 8.33542347e-01 1.32409260e-01 -1.68815541e+00 8.02802891e-02 -5.13883054e-01 1.36308879e-01 4.10213590e-01 1.64542213e-01 -5.76514065e-01 4.67861921e-01 1.13236606e+00 8.17082345e-01 -9.03481901e-01 2.01738048e+00 2.55335987e-01 1.45188496e-01 -2.48729885e-01 -2.48261839e-01 6.85057864e-02 2.28393823e-01 1.00636089e+00 1.42813504e+00 -9.14133564e-02 -3.79261553e-01 5.68818867e-01 1.62574127e-01 1.13791019e-01 -1.56615466e-01 -4.92931515e-01 2.69473135e-01 1.59106687e-01 1.34158444e+00 -8.64271641e-01 -4.36794400e-01 -5.75897098e-01 4.10432667e-01 1.84451088e-01 -5.23427874e-02 -1.13221586e+00 1.40652403e-01 1.01347792e+00 -4.58433554e-02 5.43145418e-01 -7.98700750e-02 1.11203484e-01 -9.41217959e-01 -9.52435955e-02 -7.02458024e-01 6.01913035e-01 -4.89596784e-01 -1.24235046e+00 6.78229094e-01 3.42292577e-01 -1.48542571e+00 -5.80485649e-02 -7.47530162e-01 -4.68122244e-01 4.60569441e-01 -1.32757378e+00 -1.16674793e+00 -6.99083686e-01 3.97860050e-01 5.44180453e-01 -3.72359723e-01 4.43084836e-01 5.88926375e-01 -5.93012273e-01 5.21467328e-01 1.88293740e-01 1.60805359e-01 7.98058748e-01 -1.24823022e+00 5.81939220e-01 1.02435887e+00 4.32692885e-01 1.28850609e-01 1.17215014e+00 -6.15662932e-01 -1.10316336e+00 -1.33340573e+00 5.82744896e-01 -1.39329875e+00 6.52003050e-01 -3.20868015e-01 -4.67009366e-01 7.67927110e-01 3.14005047e-01 4.12489682e-01 3.47622514e-01 -1.47908747e-01 -1.45573348e-01 -7.91448429e-02 -8.69287431e-01 4.49093789e-01 1.12662840e+00 1.97736725e-01 -5.90201378e-01 6.34769022e-01 4.63657618e-01 -7.16524184e-01 -6.40934110e-01 2.76488096e-01 6.10835016e-01 -1.12403119e+00 1.33083451e+00 -4.93587822e-01 -2.91627109e-01 -7.21898973e-01 -2.60598868e-01 -1.09928179e+00 -3.66023064e-01 -4.67370689e-01 -2.91228205e-01 1.10629594e+00 1.59402415e-01 -2.30599552e-01 7.93072879e-01 1.27357349e-01 -1.41917160e-02 -6.15600459e-02 -1.06978106e+00 -1.33085442e+00 -2.55020142e-01 -2.68917322e-01 3.79379094e-01 3.37408453e-01 -7.61758864e-01 2.99497575e-01 -7.84546196e-01 1.11633338e-01 1.47094584e+00 -1.28876075e-01 1.35959280e+00 -1.70224392e+00 -2.63795942e-01 -8.12770545e-01 -9.59679604e-01 -6.96180820e-01 -3.73254210e-01 -5.73860705e-01 2.26927385e-01 -1.59119022e+00 4.80804026e-01 -7.62251094e-02 -2.11568743e-01 7.01015443e-02 -3.95573318e-01 5.39976835e-01 7.83617735e-01 1.80729881e-01 -1.59707642e+00 1.48949280e-01 1.16704142e+00 -3.71418387e-01 2.87584126e-01 4.06449705e-01 -3.06561649e-01 5.30632913e-01 3.82396460e-01 -5.66135108e-01 1.28983274e-01 -4.05022502e-01 -2.23659024e-01 -3.96448731e-01 8.54608953e-01 -1.68668246e+00 5.52294672e-01 2.95005785e-03 4.18643057e-01 -8.66416693e-01 6.49065733e-01 -8.25572014e-01 3.81675512e-01 9.35300171e-01 7.61642307e-02 3.23444545e-01 4.45965022e-01 6.18580043e-01 -2.29402985e-02 6.33587390e-02 1.07600093e+00 -1.71576828e-01 -1.22556973e+00 5.53177297e-01 -4.56139483e-02 4.81646389e-01 1.53291667e+00 -2.22778603e-01 -6.48129642e-01 9.56523567e-02 -6.38445437e-01 5.58490336e-01 3.24230880e-01 7.64893711e-01 2.71069199e-01 -1.37936699e+00 -1.15411878e+00 -2.40325645e-01 4.19823557e-01 -5.32246947e-01 4.63944450e-02 9.22959089e-01 -3.52781564e-01 6.22692347e-01 -5.13695240e-01 -1.08787191e+00 -1.97279787e+00 5.25685668e-01 2.71022290e-01 -3.88110518e-01 -7.97995925e-01 8.44848514e-01 -7.23614916e-02 -1.95575375e-02 5.83248973e-01 -1.31819353e-01 -3.58714819e-01 -2.31797881e-02 1.05606306e+00 8.19703877e-01 -1.32521600e-01 -1.05303967e+00 -7.62446046e-01 6.01041555e-01 -4.50199172e-02 4.08178940e-02 1.14229727e+00 8.57309029e-02 2.65713215e-01 5.09818792e-01 6.65974915e-01 -2.28907689e-01 -1.31400871e+00 -1.79578796e-01 4.40508962e-01 -7.10955441e-01 -3.33318383e-01 -6.05372965e-01 -8.06973457e-01 6.30393028e-01 1.02678728e+00 3.58160853e-01 3.20158958e-01 6.28985763e-02 5.22761583e-01 3.29743356e-01 5.88422358e-01 -6.99653268e-01 4.47151251e-02 5.72992980e-01 8.44132125e-01 -1.56089342e+00 1.30507141e-01 -2.88523864e-02 -6.24187052e-01 8.66889179e-01 9.78843212e-01 9.35534984e-02 3.69257390e-01 4.08000708e-01 1.46192312e-01 -2.23522216e-01 -5.02079606e-01 -6.93662584e-01 5.65075099e-01 9.36262608e-01 2.75646985e-01 3.89572233e-02 1.62150208e-02 3.60550508e-02 5.49042877e-03 -1.42679363e-01 1.36641309e-01 8.16447318e-01 -7.62689888e-01 -1.03079605e+00 -9.38871980e-01 4.52129364e-01 -3.61494392e-01 3.99855226e-01 -3.76127005e-01 1.05722618e+00 3.94141376e-01 1.05168498e+00 -1.33409157e-01 -2.21102640e-01 6.65275156e-01 -4.15773839e-01 7.91707277e-01 -3.17974836e-01 -7.49395072e-01 -4.40290123e-01 2.13170767e-01 -6.25776768e-01 -7.43243933e-01 -1.14536834e+00 -6.03197157e-01 -5.38442492e-01 -2.56047010e-01 2.34767631e-01 5.62094331e-01 9.84929442e-01 7.92593583e-02 6.20469987e-01 -2.09491372e-01 -1.07815695e+00 -2.45052785e-01 -8.93651247e-01 -3.01165015e-01 5.34474075e-01 5.38376510e-01 -1.19965541e+00 6.70245662e-02 -5.89771047e-02]

[6.384876728057861, -2.0085182189941406]

756966ee-cd20-44a3-afe5-8e9ab8b4d89a

mask-shadownet-towards-shadow-removal-via

https://ieeexplore.ieee.org/document/9408351

https://ieeexplore.ieee.org/document/9408351

Mask-ShadowNet: Towards Shadow Removal via Masked Adaptive Instance Normalization

Shadow removal is an important yet challenging task in image processing and computer vision. Existing methods are limited in extracting good global features due to the interference of shadow. And also, most of them ignore a fact that features inside and outside the shaded area should be treated disparately because of different semantics or materials. In this letter, we propose a novel deep neural network Mask-ShadowNet for shadow removal. The core of our approach is a well-designed masked adaptive instance normalization (MAdaIN) mechanism with embedded aligners that serves two goals: 1) producing hidden features that considering an illumination consistency of different regions. 2) treating the feature statistics of shadow and non-shadow areas discriminately based on the shadow mask. Experimental results demonstrate that the proposed model outperforms the state-of-the-art on the ISTD benchmark.

['Yong Du', 'Junyu Dong', 'Bing Peng', 'Shengfeng He']

2021-04-19

['shadow-removal', 'image-shadow-removal']

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

[ 5.88666320e-01 -9.72131938e-02 3.03585202e-01 -6.46879792e-01 -1.33827049e-02 -2.07514353e-02 5.69162369e-01 -4.55857605e-01 -2.60926455e-01 6.29722893e-01 1.99179426e-01 -1.76894248e-01 2.12612525e-01 -4.34356242e-01 -3.61164868e-01 -1.15774608e+00 3.07225525e-01 6.05951212e-02 6.34925067e-01 -1.43167675e-01 1.29564315e-01 6.74193859e-01 -1.70602596e+00 2.17749223e-01 1.01955235e+00 1.11372030e+00 4.45261955e-01 1.98672488e-01 -1.57619655e-01 4.43051577e-01 -8.75711441e-01 9.53384861e-02 6.00516737e-01 -5.64026713e-01 -1.04693949e-01 1.80770293e-01 6.53811991e-01 -5.22199452e-01 -2.44408116e-01 9.64661956e-01 5.78906357e-01 6.61720484e-02 6.46491528e-01 -1.33105254e+00 -6.93872094e-01 4.78767641e-02 -7.48176098e-01 1.04103938e-01 -2.74009734e-01 -2.05741413e-02 6.33837402e-01 -9.54212308e-01 3.12035471e-01 1.21418810e+00 6.03874743e-01 2.29816988e-01 -9.02928829e-01 -5.53079844e-01 3.73165458e-01 3.99560034e-01 -1.34528971e+00 -3.43521297e-01 8.89686286e-01 -1.63191885e-01 4.16516721e-01 5.50235271e-01 5.35971284e-01 9.02609050e-01 6.70162082e-01 7.35286415e-01 1.71038711e+00 -4.05072868e-01 2.32978731e-01 1.81784302e-01 2.29424998e-01 7.58613408e-01 3.02324772e-01 1.96132828e-02 -7.10746229e-01 4.30394448e-02 4.24977690e-01 2.34019130e-01 -5.43523371e-01 -5.67755580e-01 -1.05443871e+00 6.51626706e-01 4.56594914e-01 1.04392752e-01 -2.13646144e-01 -1.41670063e-01 1.06414288e-01 -2.74047311e-02 6.54851913e-01 1.72781590e-02 -5.28364360e-01 6.98170602e-01 -9.45888758e-01 1.16334751e-01 6.64202332e-01 7.35601604e-01 7.97819018e-01 2.19587758e-01 -3.77128243e-01 5.18240333e-01 1.44114062e-01 8.38392496e-01 3.79546344e-01 -4.74645168e-01 -1.32663567e-02 5.58432877e-01 1.10343397e-01 -1.10470700e+00 -6.65597618e-01 -6.00273192e-01 -9.03977752e-01 8.08764338e-01 1.28242671e-01 8.95803869e-02 -1.42675483e+00 1.51975954e+00 4.76924568e-01 2.95796186e-01 9.07671899e-02 1.09131479e+00 1.28996289e+00 3.40733796e-01 -3.28542829e-01 -2.13510394e-01 1.34833145e+00 -1.18336987e+00 -1.05098987e+00 -5.16993701e-01 -9.59376916e-02 -1.00167227e+00 1.03323352e+00 2.10165054e-01 -5.89559972e-01 -6.12620533e-01 -1.38020933e+00 -1.85764804e-01 -7.73150682e-01 3.34746301e-01 7.25323439e-01 6.17205679e-01 -7.74998963e-01 4.02227551e-01 -3.93967956e-01 -3.95371944e-01 4.16939050e-01 2.44512305e-01 -1.08650967e-01 2.20610008e-01 -7.70407796e-01 9.38704729e-01 2.26912379e-01 5.13630569e-01 -7.58930504e-01 -4.57771450e-01 -6.31314278e-01 6.77958354e-02 5.68183482e-01 -4.58071381e-01 8.79435360e-01 -1.29636812e+00 -1.47062325e+00 7.99099445e-01 -3.29983920e-01 -3.46628845e-01 5.95444739e-01 -5.87079167e-01 -3.55842352e-01 -3.08863252e-01 -1.81638282e-02 1.57134503e-01 1.18430209e+00 -1.81231999e+00 -5.95756888e-01 -4.95784670e-01 -2.33547300e-01 5.19202709e-01 -2.79788762e-01 -4.52359915e-02 -4.47443664e-01 -8.05692375e-01 3.43018353e-01 -8.37287724e-01 -5.02268188e-02 1.35154352e-01 -7.48435259e-01 2.31704295e-01 1.52432227e+00 -7.71724761e-01 1.01230121e+00 -2.03406024e+00 -2.84337014e-01 2.27547616e-01 3.39128137e-01 5.05223215e-01 -1.65864155e-02 -9.56741907e-03 8.78030062e-02 -4.01741117e-01 -6.30925357e-01 -4.21095878e-01 8.46223757e-02 3.11743677e-01 -6.81578040e-01 8.39411557e-01 4.34759557e-02 6.25367820e-01 -4.88426059e-01 -5.44594228e-01 5.21232724e-01 5.53845108e-01 1.36835426e-01 3.28123748e-01 -6.13492727e-02 4.22989577e-01 -3.18620801e-01 8.62332165e-01 1.23012888e+00 1.03583105e-01 8.64583775e-02 -4.72904265e-01 -2.80585915e-01 4.85675186e-02 -1.30971611e+00 1.29270923e+00 -1.71142638e-01 7.83277512e-01 1.97163209e-01 -5.27722895e-01 8.30237508e-01 -1.90342650e-01 4.17986155e-01 -8.70951295e-01 4.28222835e-01 1.22107245e-01 6.72231913e-02 -2.79467672e-01 4.96325761e-01 1.00453667e-01 1.95039883e-01 3.44885737e-01 -3.60490054e-01 -1.68717027e-01 -4.07626063e-01 -2.14928389e-01 9.44339633e-01 2.77937382e-01 4.44774300e-01 -4.63137835e-01 5.82368970e-01 -3.78525347e-01 9.64454353e-01 7.70459652e-01 -4.14796948e-01 1.01812971e+00 1.65215865e-01 -4.65600997e-01 -5.28026164e-01 -1.17884016e+00 -2.42278159e-01 1.12258255e+00 6.55829847e-01 6.80965856e-02 -6.98678792e-01 -8.08295369e-01 1.59541756e-01 6.22381389e-01 -9.12407100e-01 5.96639328e-02 -6.44953728e-01 -1.03519845e+00 1.89526841e-01 3.28877717e-01 8.08098137e-01 -1.13018215e+00 -1.08399177e+00 -1.72435284e-01 9.99400467e-02 -1.22885501e+00 -2.98564553e-01 4.79866594e-01 -3.11787337e-01 -1.04235613e+00 -4.80743051e-01 -6.76873505e-01 6.45676792e-01 1.04999483e+00 1.04449332e+00 2.33118489e-01 -5.46302915e-01 -1.07730515e-02 -3.36165100e-01 -9.80579972e-01 -1.01541497e-01 -2.20467374e-01 -5.45821749e-02 4.55333710e-01 1.94056541e-01 -5.24418533e-01 -7.70523369e-01 3.93696487e-01 -1.01813555e+00 3.30242187e-01 9.16423857e-01 6.74207449e-01 5.10190189e-01 -5.85107692e-02 1.04770809e-01 -1.07697070e+00 4.49278414e-01 -9.07409787e-02 -5.66459179e-01 4.53365475e-01 -8.30913126e-01 -6.45941794e-02 5.16947746e-01 -5.91312759e-02 -1.51360297e+00 4.53050621e-02 5.28403878e-01 -2.53447473e-01 -3.38104606e-01 -2.72138566e-01 -6.47058606e-01 -2.72899270e-01 3.39968532e-01 2.34740987e-01 -2.59315342e-01 -4.11826134e-01 3.61020297e-01 5.59384704e-01 4.96574223e-01 -1.58273473e-01 1.13565242e+00 1.15686643e+00 2.69224465e-01 -8.40980232e-01 -1.00381815e+00 -4.90642965e-01 -8.64233196e-01 -2.54757881e-01 8.61192107e-01 -6.26695752e-01 -3.35544735e-01 8.83901656e-01 -1.10546637e+00 -3.15194666e-01 -1.32365808e-01 5.50945103e-03 -1.63717702e-01 6.17551804e-01 5.50853424e-02 -9.53810334e-01 -4.28123921e-01 -1.02460301e+00 1.29192865e+00 6.38382614e-01 2.42170036e-01 -6.02473378e-01 -6.55288100e-02 1.65442020e-01 5.79357028e-01 5.14958918e-01 7.33434558e-01 -4.38510299e-01 -7.47214139e-01 1.86296642e-01 -6.63707614e-01 5.16829133e-01 4.42488611e-01 -3.50957364e-02 -1.45969403e+00 -4.80628945e-02 2.06075087e-01 1.51078459e-02 1.24339926e+00 4.76199865e-01 1.15273142e+00 -4.70387749e-02 -3.50401938e-01 9.07738924e-01 1.46459532e+00 1.17814228e-01 6.31763995e-01 3.95592839e-01 8.36335242e-01 6.51550353e-01 7.93683887e-01 4.94934589e-01 1.19338676e-01 6.86860263e-01 6.76715374e-01 -8.22159529e-01 -5.61645865e-01 3.96109968e-01 3.05301905e-01 5.03445566e-01 -7.85583556e-02 -3.11239034e-01 -6.04444563e-01 4.01314974e-01 -2.02822733e+00 -6.78808749e-01 -3.46181840e-01 1.99797833e+00 5.16654849e-01 1.12628065e-01 -5.15722871e-01 -1.58319548e-01 6.50010943e-01 7.27327347e-01 -6.51280165e-01 -2.86966681e-01 -8.14413786e-01 4.75911379e-01 7.72288501e-01 2.99867839e-01 -1.33708799e+00 1.06741583e+00 5.60097551e+00 8.66114616e-01 -1.07628882e+00 1.32296458e-01 4.83144730e-01 1.02686375e-01 2.54396722e-03 5.30810794e-03 -6.85025513e-01 4.22777742e-01 1.46786749e-01 3.80025387e-01 3.83688450e-01 6.97577715e-01 2.58735836e-01 -6.42508566e-01 -6.63749993e-01 8.46662343e-01 5.73677182e-01 -7.79358923e-01 -1.23153597e-01 1.64924916e-02 9.12775159e-01 -2.58828755e-02 2.86542833e-01 2.67029293e-02 6.73326552e-02 -9.83144641e-01 8.25395703e-01 7.45653272e-01 4.15518433e-01 -3.76687974e-01 8.69212210e-01 -2.90025119e-02 -1.27397096e+00 -2.70398501e-02 -3.72865170e-01 1.70434383e-03 -9.35449600e-02 9.45544422e-01 -8.47451270e-01 5.86788118e-01 8.87304187e-01 2.31093585e-01 -7.30590641e-01 8.50552201e-01 -3.76719892e-01 3.73127848e-01 -1.03149585e-01 1.35895699e-01 2.52805389e-02 -5.02679706e-01 4.48886603e-01 1.31514454e+00 -1.90610941e-02 -1.41601995e-01 2.72897780e-01 8.61021101e-01 7.62725025e-02 1.44080515e-03 -5.44072270e-01 5.25775790e-01 2.41936535e-01 1.53556895e+00 -1.08906829e+00 -2.23286361e-01 -5.45343995e-01 1.25906670e+00 -2.75084823e-02 4.59866613e-01 -1.20431828e+00 -2.48550326e-01 7.25503683e-01 -2.42349416e-01 3.49478662e-01 -2.02383116e-01 -8.08838785e-01 -7.85260677e-01 2.64682680e-01 -7.41444409e-01 -1.78124476e-02 -8.92991245e-01 -1.13871694e+00 6.25631571e-01 -2.37507835e-01 -1.28478265e+00 5.35675406e-01 -7.44043648e-01 -7.44004726e-01 8.21023047e-01 -2.05167556e+00 -1.49873292e+00 -1.06240737e+00 4.99129504e-01 6.39293909e-01 -1.48743272e-01 6.74063623e-01 3.52301374e-02 -5.74015617e-01 2.93921858e-01 3.86321574e-01 5.24461009e-02 1.10503972e+00 -1.29525661e+00 2.57463813e-01 1.13524473e+00 -4.76217903e-02 3.99234772e-01 9.00059342e-01 -6.53473198e-01 -1.09623742e+00 -1.13907027e+00 5.79504371e-01 -1.95777863e-01 3.02920640e-01 -4.92821693e-01 -8.53536785e-01 3.16524178e-01 4.68092710e-01 6.22461140e-02 5.40122509e-01 -1.01190589e-01 -3.46421599e-01 -5.05658925e-01 -9.89725113e-01 6.52537584e-01 1.10994744e+00 -2.68692791e-01 -5.18658400e-01 4.32005227e-01 4.87265974e-01 -5.48768818e-01 -1.55357584e-01 7.79874861e-01 5.79947770e-01 -1.56560707e+00 8.87613297e-01 -2.12005749e-01 7.63818249e-02 -5.91550231e-01 -4.11934912e-01 -1.11731303e+00 -1.86935991e-01 -4.64835107e-01 -9.13808793e-02 1.25127065e+00 -9.79130790e-02 -8.11426580e-01 7.39947140e-01 3.32670063e-01 -3.76233101e-01 -8.75265300e-01 -9.68367219e-01 -6.95010245e-01 -5.44305801e-01 -4.96970303e-02 6.54981673e-01 7.53976643e-01 -9.77913916e-01 2.00879499e-01 -3.94937128e-01 4.99752313e-01 6.50709093e-01 7.03899801e-01 1.07273078e+00 -1.40634525e+00 -9.32672247e-02 -3.58665675e-01 -7.40084574e-02 -6.65636063e-01 2.78454244e-01 -3.97077441e-01 4.30111885e-01 -1.58036458e+00 4.80110526e-01 -3.69084984e-01 -5.78872025e-01 4.73294973e-01 -3.70061994e-01 3.71758342e-01 1.63686275e-01 8.74872208e-02 -5.86917579e-01 9.26631868e-01 1.20180893e+00 2.37556472e-02 -1.04377583e-01 -2.63455696e-02 -4.62932140e-01 1.03774965e+00 8.59940112e-01 -3.40416700e-01 -2.54996240e-01 -3.81768823e-01 -4.32551980e-01 -7.90486097e-01 7.13374972e-01 -1.16384041e+00 -1.29578710e-01 -3.58404219e-01 6.31719172e-01 -9.09258485e-01 6.42762125e-01 -1.11958861e+00 -2.52221674e-01 3.72252643e-01 1.95046306e-01 -1.21381596e-01 2.15043407e-02 5.78661442e-01 9.66703445e-02 1.60031438e-01 7.94649601e-01 1.59960333e-03 -8.82201612e-01 6.61230460e-03 -2.12147146e-01 -2.30787158e-01 1.06510746e+00 -3.53552192e-01 -6.37868822e-01 -2.31245518e-01 -1.13235384e-01 -1.56017557e-01 4.49943274e-01 4.84360695e-01 7.39534199e-01 -1.00468910e+00 -5.41817427e-01 3.87400240e-01 -1.09471586e-02 -3.44390608e-03 2.08162144e-01 8.69844198e-01 -4.10386056e-01 3.30987215e-01 -3.62282574e-01 -5.11724472e-01 -1.68248451e+00 3.06714654e-01 4.92848307e-01 -1.78155556e-01 -7.23711610e-01 7.83261776e-01 9.61869597e-01 -3.63578349e-01 3.99406672e-01 -4.30048466e-01 -9.16683972e-02 -9.17841867e-02 2.86475062e-01 3.47281545e-01 2.64498234e-01 -8.18280399e-01 -3.84286493e-01 5.44478059e-01 8.01427588e-02 1.71777189e-01 1.28914213e+00 -7.91653916e-02 -3.76975447e-01 3.44196349e-01 8.03283513e-01 2.49723524e-01 -1.42795300e+00 -3.37537408e-01 -2.84991771e-01 -7.23625660e-01 1.66502550e-01 -9.22894001e-01 -1.16698110e+00 5.89293361e-01 1.18102050e+00 9.12540704e-02 1.48965538e+00 -3.58631849e-01 8.34507465e-01 2.35779315e-01 1.44842952e-01 -1.32668889e+00 -7.36596584e-02 5.23498178e-01 1.06124902e+00 -1.22699440e+00 4.84529376e-01 -6.14643514e-01 -5.69523811e-01 9.92532372e-01 7.89768636e-01 -1.06732033e-01 6.71532989e-01 4.45908427e-01 5.05398273e-01 -2.21606702e-01 -1.79598778e-01 -6.14206493e-01 5.46914637e-01 7.34809399e-01 5.67420311e-02 1.18681297e-01 -4.20683116e-01 5.78482449e-01 1.66337844e-02 -5.57541788e-01 4.19328898e-01 1.06853306e+00 -6.78070188e-01 -8.02609563e-01 -7.80450284e-01 2.82980561e-01 -1.50730506e-01 -2.70094782e-01 -8.35421801e-01 9.75556433e-01 6.90203607e-01 9.27690685e-01 -1.57057956e-01 -3.56114388e-01 1.86820433e-01 -1.49736404e-01 3.22014123e-01 -4.31184918e-01 -4.77787256e-01 1.15445472e-01 -1.85859159e-01 -6.97184384e-01 -5.40010154e-01 -5.65144062e-01 -1.26594162e+00 -3.75246033e-02 -4.37909424e-01 -3.21432114e-01 9.14461076e-01 1.04748392e+00 2.94037133e-01 8.76768649e-01 6.86307907e-01 -1.05491173e+00 -2.91870981e-01 -8.64970744e-01 -7.80270875e-01 4.29483801e-01 5.17823994e-01 -9.10338819e-01 -2.91246265e-01 3.25209908e-02]

[10.845865249633789, -4.08622932434082]

a26b6b43-6b73-4655-8ac3-81c9eee9a8b7

high-dimensional-smoothed-entropy-estimation

2305.04712

https://arxiv.org/abs/2305.04712v2

https://arxiv.org/pdf/2305.04712v2.pdf

High-Dimensional Smoothed Entropy Estimation via Dimensionality Reduction

We study the problem of overcoming exponential sample complexity in differential entropy estimation under Gaussian convolutions. Specifically, we consider the estimation of the differential entropy $h(X+Z)$ via $n$ independently and identically distributed samples of $X$, where $X$ and $Z$ are independent $D$-dimensional random variables with $X$ sub-Gaussian with bounded second moment and $Z\sim\mathcal{N}(0,\sigma^2I_D)$. Under the absolute-error loss, the above problem has a parametric estimation rate of $\frac{c^D}{\sqrt{n}}$, which is exponential in data dimension $D$ and often problematic for applications. We overcome this exponential sample complexity by projecting $X$ to a low-dimensional space via principal component analysis (PCA) before the entropy estimation, and show that the asymptotic error overhead vanishes as the unexplained variance of the PCA vanishes. This implies near-optimal performance for inherently low-dimensional structures embedded in high-dimensional spaces, including hidden-layer outputs of deep neural networks (DNN), which can be used to estimate mutual information (MI) in DNNs. We provide numerical results verifying the performance of our PCA approach on Gaussian and spiral data. We also apply our method to analysis of information flow through neural network layers (c.f. information bottleneck), with results measuring mutual information in a noisy fully connected network and a noisy convolutional neural network (CNN) for MNIST classification.

['Yuancheng Yu', 'Brian Kingsbury', 'Kristjan Greenewald']

2023-05-08

['dimensionality-reduction']

['methodology']

[ 8.27239081e-03 3.54232937e-01 4.05681163e-01 -3.08096170e-01 -6.59293771e-01 -4.05713469e-01 1.13815993e-01 -1.38416037e-01 -1.03796899e+00 9.55538452e-01 -3.51075172e-01 -3.38853776e-01 -5.61625957e-01 -6.86568558e-01 -7.45118141e-01 -1.17526853e+00 -7.01009929e-01 3.01840246e-01 -3.56045574e-01 3.66364777e-01 7.39997923e-02 5.30702055e-01 -1.43842995e+00 -4.38065022e-01 6.22290969e-01 1.63107634e+00 1.24538448e-02 8.46749663e-01 5.61850779e-02 5.43206930e-01 -5.57461381e-01 -8.57105613e-01 3.70525718e-01 -4.67258871e-01 -7.86677897e-01 -4.47761804e-01 9.61497352e-02 -3.08174402e-01 -7.06160665e-01 1.65522134e+00 5.70760608e-01 1.86273500e-01 9.47173119e-01 -9.41075325e-01 -3.67537916e-01 8.16281557e-01 -2.82606244e-01 4.05589461e-01 -5.86585641e-01 5.23514785e-02 1.01921725e+00 -6.94285750e-01 4.73456442e-01 9.51000392e-01 8.38445723e-01 4.41527843e-01 -1.50642002e+00 -9.38029706e-01 -3.13912839e-01 -2.29369700e-01 -1.68880105e+00 -3.20484519e-01 6.64049685e-01 -5.11954606e-01 1.02633369e+00 -2.76955254e-02 6.45791352e-01 6.62863433e-01 3.32364053e-01 5.71683526e-01 8.50203753e-01 -1.00194983e-01 5.29422998e-01 2.20039099e-01 2.43369788e-01 7.51655340e-01 4.62495178e-01 1.57591313e-01 -2.16954336e-01 1.00851180e-02 9.24952865e-01 -1.64606899e-01 -3.05076331e-01 -1.05626553e-01 -8.34848225e-01 7.91126370e-01 2.99270838e-01 3.41854066e-01 -2.42256999e-01 4.43660647e-01 4.39493686e-01 4.37595755e-01 4.10241604e-01 3.49010736e-01 -7.64249980e-01 -3.68945718e-01 -8.53671014e-01 1.92568764e-01 1.07884407e+00 1.18329847e+00 7.83803344e-01 1.81790993e-01 2.94880986e-01 5.34621060e-01 3.08348656e-01 7.48918831e-01 3.32816511e-01 -1.35723054e+00 4.67451870e-01 2.53381431e-01 -3.34743798e-01 -8.33070040e-01 -4.33236659e-01 -8.37151349e-01 -1.53843963e+00 3.75354111e-01 6.38471246e-01 -5.92459023e-01 -3.45081925e-01 2.08812833e+00 -4.09435958e-01 -5.92689574e-01 2.48916432e-01 5.31754851e-01 3.10909539e-01 1.37513697e-01 -2.04842314e-01 -4.66192186e-01 1.02498651e+00 -1.51687056e-01 -6.39573336e-01 -2.26393431e-01 6.55287743e-01 -3.10720682e-01 6.10856593e-01 2.98546404e-01 -1.45745242e+00 -9.20778215e-02 -1.06424391e+00 -9.44926068e-02 -2.28377223e-01 3.61992195e-02 4.96098191e-01 8.36204946e-01 -1.19851768e+00 1.12862110e+00 -9.47126567e-01 3.98235470e-01 1.00670886e+00 7.83894002e-01 -4.41163182e-01 3.36629570e-01 -1.07291317e+00 4.63623255e-01 4.19926554e-01 1.94386691e-01 -7.12956607e-01 -8.33971322e-01 -6.63008749e-01 2.77752072e-01 -2.90524453e-01 -3.72573644e-01 9.28894222e-01 -6.73995316e-01 -1.37218988e+00 6.45827830e-01 3.50234173e-02 -7.49837399e-01 4.00157750e-01 -4.93069291e-02 8.55983347e-02 2.33334795e-01 -2.99553096e-01 4.92779702e-01 5.67805409e-01 -7.21059859e-01 -1.02736123e-01 -8.84732842e-01 -3.66179943e-01 -9.98981372e-02 -4.14716959e-01 -1.39190704e-01 1.74703225e-02 -3.68857235e-01 5.39768577e-01 -8.21099341e-01 -2.39352539e-01 4.74633425e-02 -4.88674045e-01 4.22417104e-01 5.10328710e-01 -8.19183528e-01 1.18725741e+00 -2.16559124e+00 1.65761143e-01 4.65644330e-01 6.40713930e-01 1.32904172e-01 3.24458897e-01 -1.48348495e-01 -2.74223149e-01 3.96551102e-01 -5.80863297e-01 -5.34684062e-01 1.81404680e-01 1.19930347e-02 4.16236594e-02 8.64964485e-01 1.71406582e-01 7.56532192e-01 -4.94557053e-01 -2.46905103e-01 -1.78323880e-01 8.10328543e-01 -6.59964502e-01 -5.02496213e-02 1.34374246e-01 7.92533457e-02 -2.15353921e-01 1.40914991e-01 9.53568041e-01 -4.44108456e-01 7.18117878e-02 -1.70421764e-01 5.12285568e-02 1.70884967e-01 -1.26922274e+00 1.37543738e+00 -3.02931219e-01 1.18184614e+00 4.24687147e-01 -1.30031836e+00 8.20348084e-01 1.22699484e-01 4.28423017e-01 -4.17702079e-01 6.95901275e-01 2.09456518e-01 4.20114428e-01 -1.24993227e-01 4.43457440e-02 -3.96224856e-01 1.66398481e-01 4.13737595e-01 5.66906571e-01 -7.35549107e-02 -5.08555099e-02 6.33368641e-02 1.37579918e+00 -8.10136497e-01 1.42869771e-01 -7.27421761e-01 2.09389925e-01 -7.83822954e-01 3.00880939e-01 7.08046079e-01 -5.06559551e-01 3.68851304e-01 1.28791511e+00 1.53486669e-01 -1.38582647e+00 -1.17804730e+00 -6.47593737e-01 4.10766840e-01 -2.62690425e-01 -1.26656964e-01 -1.17541158e+00 -3.67136717e-01 -1.48214951e-01 6.64126575e-01 -6.29066765e-01 -2.27402315e-01 -1.94160998e-01 -1.13436854e+00 8.55375707e-01 3.96781117e-01 8.88568223e-01 -7.78151989e-01 -5.49697936e-01 -4.49970290e-02 1.97587743e-01 -8.90056670e-01 -1.88224167e-01 9.51068401e-01 -1.00804543e+00 -7.45009243e-01 -6.65301979e-01 -4.84608710e-01 7.31880009e-01 -5.04035830e-01 8.73987138e-01 -6.17867827e-01 -3.64294410e-01 3.43101025e-01 3.99174273e-01 -2.81998217e-01 -1.77187100e-01 6.10987619e-02 2.20851034e-01 -3.99532586e-01 4.74182606e-01 -1.05694771e+00 -7.66719520e-01 2.58641213e-01 -8.20082188e-01 -4.97987002e-01 6.71817303e-01 9.13026631e-01 6.62678659e-01 4.46212620e-01 2.07975760e-01 -3.32679540e-01 6.10672832e-01 -3.37174594e-01 -1.01677299e+00 -3.24479282e-01 -6.32734597e-01 5.91059327e-01 7.70436347e-01 -3.84703577e-01 -6.25084400e-01 -2.19826207e-01 -3.11973482e-01 -6.26598179e-01 5.72834983e-02 3.73652011e-01 -2.11173549e-01 -2.26332650e-01 6.23029351e-01 3.12361151e-01 3.84690493e-01 -4.40412134e-01 6.24249876e-02 4.35666502e-01 5.06345570e-01 -2.83440232e-01 4.21674490e-01 3.66583973e-01 4.31988120e-01 -8.14757705e-01 -5.08482337e-01 9.15482640e-02 -6.02817714e-01 7.94379935e-02 8.96553576e-01 -7.65488982e-01 -1.58213758e+00 5.37621081e-01 -1.08786356e+00 -4.05715346e-01 -5.51688075e-01 8.89007986e-01 -8.50740969e-01 1.02596991e-01 -7.50301063e-01 -1.24941933e+00 -4.74922359e-01 -1.15144289e+00 5.65866828e-01 2.36264706e-01 1.12613961e-01 -1.19581163e+00 -3.48224878e-01 -2.45314628e-01 5.74367523e-01 -1.03076383e-01 1.00091505e+00 -8.08674932e-01 -4.51816559e-01 -4.53157902e-01 -4.34382766e-01 1.05214357e+00 -3.74013454e-01 -1.48124576e-01 -1.06131208e+00 -5.27586341e-02 5.87833107e-01 -2.10952565e-01 9.98198569e-01 8.37701738e-01 1.49304497e+00 -7.18431830e-01 -6.32283539e-02 9.46119189e-01 1.39640164e+00 2.99341008e-02 6.38318896e-01 -2.04688862e-01 2.25474447e-01 3.76958370e-01 -1.81301713e-01 6.78337395e-01 -1.20439716e-01 9.78084560e-03 4.95563060e-01 3.44918430e-01 2.06601322e-01 2.26622730e-01 2.59585291e-01 9.26200628e-01 -4.91363220e-02 -5.38373888e-02 -8.87384892e-01 2.30170161e-01 -1.22216129e+00 -1.14486933e+00 7.37835467e-02 2.41910386e+00 1.08791924e+00 4.02532876e-01 -1.72965437e-01 2.17927784e-01 5.11657536e-01 -1.22954957e-01 -7.83345520e-01 -1.77633196e-01 -3.45000833e-01 5.81275403e-01 1.01876163e+00 5.50132155e-01 -1.15873492e+00 3.58323276e-01 6.05147839e+00 1.02245545e+00 -8.73167396e-01 1.43945897e-02 1.04064417e+00 -2.24248841e-01 -1.07570931e-01 -3.85486871e-01 -9.36996698e-01 6.69640005e-01 1.30320704e+00 -1.36641502e-01 6.19270742e-01 1.09363604e+00 -2.89851099e-01 -1.65555984e-01 -1.05595922e+00 1.32880390e+00 -4.24659014e-01 -1.31341851e+00 -6.00082755e-01 4.86587375e-01 5.83635211e-01 4.39229220e-01 6.40189528e-01 9.86786485e-02 4.56791699e-01 -1.13027716e+00 3.82875234e-01 4.49807942e-01 9.76108730e-01 -9.10898626e-01 7.98065305e-01 4.33305979e-01 -8.83197129e-01 -1.92017972e-01 -7.09372818e-01 8.94587673e-03 -1.22691989e-01 1.14688182e+00 -5.42916775e-01 -1.64254293e-01 7.66477525e-01 4.57149625e-01 -4.46764193e-02 7.41001308e-01 3.19558114e-01 3.30695897e-01 -7.44093060e-01 -2.70696163e-01 -2.30139098e-03 -5.25267839e-01 5.42210221e-01 1.10566056e+00 5.98712802e-01 2.68568665e-01 -8.56035054e-01 1.32279289e+00 -3.94384354e-01 -1.66570187e-01 -5.71620643e-01 -3.90886903e-01 3.80066842e-01 9.85796928e-01 -7.20193207e-01 -2.54563894e-02 -5.20102642e-02 8.89361858e-01 3.02683055e-01 4.71782953e-01 -5.58706760e-01 -8.89897645e-01 1.14792466e+00 -1.48926154e-01 5.25000811e-01 -4.83228058e-01 -6.42701864e-01 -8.55196118e-01 1.39871314e-01 -2.43938699e-01 -5.54846115e-02 -1.27324119e-01 -1.20240664e+00 6.08293056e-01 -1.90445378e-01 -9.36827838e-01 -1.60899103e-01 -1.01835835e+00 -1.20017521e-01 1.21771109e+00 -1.08278513e+00 8.03155545e-03 2.76981980e-01 5.06957412e-01 -3.38927388e-01 -3.49910617e-01 8.42683911e-01 4.61272597e-01 -6.81138098e-01 1.00688016e+00 8.71349275e-01 5.99997401e-01 -9.95389223e-02 -1.31000996e+00 3.18680137e-01 4.77078766e-01 -3.20665985e-01 7.78412819e-01 7.99101830e-01 -3.40667069e-01 -1.32464564e+00 -9.73687410e-01 6.21117711e-01 -1.14037380e-01 7.63367474e-01 -6.40836954e-01 -8.58612955e-01 6.63119256e-01 -1.74913123e-01 3.28258991e-01 9.21300948e-01 -1.86180174e-01 -2.88242906e-01 -4.60881479e-02 -1.47738326e+00 3.58183950e-01 1.10233557e+00 -7.44021297e-01 4.31153290e-02 1.04009926e-01 6.73630059e-01 -1.43189043e-01 -1.30327618e+00 2.15106130e-01 5.47286153e-01 -1.27257419e+00 8.77915442e-01 -5.27170300e-01 3.94813716e-01 4.03904080e-01 -5.53036988e-01 -1.08745539e+00 1.67048827e-01 -7.97224283e-01 -1.42466933e-01 7.28292525e-01 5.75531423e-01 -7.14644194e-01 9.60423112e-01 9.80721295e-01 1.07737407e-01 -7.20742345e-01 -1.48156559e+00 -7.36159623e-01 5.53076088e-01 -7.42824674e-01 -7.38293957e-03 4.93610322e-01 1.18820786e-01 2.99146008e-02 2.70072877e-01 -8.41269176e-03 8.55782330e-01 -6.71766400e-01 2.27172211e-01 -1.24320507e+00 -3.74278784e-01 -9.60423172e-01 -8.36091161e-01 -1.03265512e+00 3.11703920e-01 -9.15644050e-01 1.39180431e-02 -6.95475101e-01 2.08982885e-01 -4.01461035e-01 -1.55286118e-01 -1.22865133e-01 3.60083520e-01 -2.12109596e-01 -1.19810112e-01 -1.22820556e-01 -3.24500471e-01 7.97423482e-01 9.64770019e-01 1.40131041e-01 2.13253424e-02 1.07393295e-01 -4.99640137e-01 8.78856301e-01 6.66374445e-01 -3.93183172e-01 -3.32995236e-01 -2.33957261e-01 4.12405282e-01 2.15378001e-01 3.95933598e-01 -1.32855725e+00 2.61067837e-01 3.75261605e-01 4.73842502e-01 -3.62080634e-01 5.46985507e-01 -7.28502631e-01 -1.07646406e-01 3.73052180e-01 -4.85653222e-01 -6.81048408e-02 1.60327822e-01 3.81186366e-01 -1.16551533e-01 -3.10131967e-01 1.14403415e+00 -1.07082136e-01 3.41282010e-01 5.74835122e-01 -2.26854861e-01 2.98508167e-01 7.50739813e-01 2.08992496e-01 -3.84969145e-01 -5.32667756e-01 -7.36557305e-01 -2.55897045e-01 1.66545719e-01 -6.12034023e-01 5.99896431e-01 -1.17371655e+00 -3.69080454e-01 5.39383054e-01 -4.94306415e-01 2.94201195e-01 4.52809185e-01 1.01975000e+00 -4.98135805e-01 4.32888180e-01 -2.43511517e-03 -5.28266132e-01 -6.30653322e-01 3.62860203e-01 7.71052718e-01 -2.43479043e-01 -2.02737540e-01 1.43281138e+00 -4.01547700e-02 -1.49863333e-01 4.24577564e-01 -4.66163516e-01 1.85676515e-01 5.34923784e-02 5.49301028e-01 4.20715034e-01 1.09365985e-01 -3.93409282e-01 -1.41464770e-01 1.96033776e-01 -1.20476454e-01 -4.62819457e-01 1.25117552e+00 -4.34693024e-02 -3.74765277e-01 4.47587222e-01 2.34302974e+00 -5.66448748e-01 -1.54431009e+00 -3.52858990e-01 -2.04638928e-01 -3.91769707e-02 2.70400405e-01 -3.26375306e-01 -1.39099586e+00 1.35727525e+00 8.50330114e-01 3.43778282e-01 8.77223015e-01 2.09621102e-01 3.21190625e-01 7.04275966e-01 5.26854470e-02 -1.15056336e+00 -1.32874072e-01 7.57813573e-01 5.07958293e-01 -9.34602857e-01 -3.69587928e-01 1.52941927e-01 -3.55435848e-01 1.13919735e+00 2.00728357e-01 -1.29440784e-01 1.43897605e+00 6.20454073e-01 -4.03844386e-01 -1.95278943e-01 -5.79545438e-01 1.20611906e-01 1.84960831e-02 4.16930825e-01 3.57178003e-01 -1.29222609e-02 -8.72327164e-02 9.47964191e-01 -5.74456573e-01 -2.02728465e-01 2.33912706e-01 6.14592969e-01 -4.11965132e-01 -4.63553667e-01 1.75813362e-01 8.16193223e-01 -6.51639998e-01 -4.31173593e-01 7.59741291e-02 6.83427811e-01 -1.85345605e-01 3.33055675e-01 5.21105766e-01 -4.11821157e-01 -1.90052822e-01 3.38560224e-01 4.27000493e-01 1.91806093e-01 -7.97942188e-03 -1.56014219e-01 -4.49970663e-01 -4.90077585e-01 -7.50546111e-03 -9.74121213e-01 -1.35466421e+00 -5.77567816e-01 -1.58027321e-01 -2.46293768e-02 1.05815721e+00 9.12537694e-01 3.17118436e-01 3.75529796e-01 5.07915437e-01 -7.37333655e-01 -8.90167356e-01 -9.63088334e-01 -1.11126089e+00 -7.87918083e-03 5.57702601e-01 -5.04508197e-01 -1.01068711e+00 -2.53538042e-01]

[7.852762222290039, 3.618507146835327]

3c89c8b7-f413-4669-9919-f64380a32655

deep-learning-for-punctuation-restoration-in

https://aclanthology.org/W17-2319

https://aclanthology.org/W17-2319.pdf

Deep Learning for Punctuation Restoration in Medical Reports

In clinical dictation, speakers try to be as concise as possible to save time, often resulting in utterances without explicit punctuation commands. Since the end product of a dictated report, e.g. an out-patient letter, does require correct orthography, including exact punctuation, the latter need to be restored, preferably by automated means. This paper describes a method for punctuation restoration based on a state-of-the-art stack of NLP and machine learning techniques including B-RNNs with an attention mechanism and late fusion, as well as a feature extraction technique tailored to the processing of medical terminology using a novel vocabulary reduction model. To the best of our knowledge, the resulting performance is superior to that reported in prior art on similar tasks.

['David Suendermann-Oeft', 'Wael Salloum', 'Greg Finley', 'Mark Miller', 'Erik Edwards']

2017-08-01

ws-2017-8

['punctuation-restoration']

['natural-language-processing']

[ 5.69456160e-01 5.06347835e-01 3.07035983e-01 -5.22538662e-01 -1.17210484e+00 -5.10343015e-01 2.01183945e-01 7.33426988e-01 -7.04515219e-01 7.35443234e-01 7.22576261e-01 -7.71885216e-01 -2.52940834e-01 -7.28089958e-02 -4.14884210e-01 -4.61303502e-01 2.28274137e-01 6.63935304e-01 -4.69597787e-01 -3.66666466e-01 2.18253329e-01 4.56701368e-01 -1.09383273e+00 4.81781006e-01 4.86017644e-01 6.47582710e-01 2.34656468e-01 8.19811046e-01 -5.43652058e-01 1.03509510e+00 -8.57720971e-01 -5.54924726e-01 8.94645974e-03 -4.04143989e-01 -9.56461251e-01 1.41752169e-01 3.17461342e-01 -1.82895213e-01 -2.25090846e-01 8.61400962e-01 5.89407742e-01 7.94194713e-02 4.54764277e-01 7.30816871e-02 -4.89709377e-01 8.96325469e-01 6.92707896e-02 3.22621942e-01 7.08739519e-01 -3.35913748e-01 7.04433143e-01 -7.22504258e-01 7.49352038e-01 8.27269733e-01 7.64916897e-01 7.12684155e-01 -1.21348286e+00 -3.32529217e-01 -1.39927134e-01 -5.66543043e-02 -1.38903594e+00 -8.33271027e-01 5.51692426e-01 -1.65844113e-01 1.43342006e+00 3.65503579e-01 4.65865880e-01 1.05066037e+00 6.43253446e-01 5.27502775e-01 8.55453014e-01 -1.06857598e+00 1.66466281e-01 3.25328767e-01 1.57645583e-01 3.83502781e-01 1.67640802e-02 -2.35441297e-01 -6.82216823e-01 6.20750785e-02 2.27272004e-01 5.20013971e-04 -4.77676779e-01 3.64537209e-01 -1.06234896e+00 4.87567723e-01 -7.00468272e-02 5.33215940e-01 -6.55242980e-01 -2.13149086e-01 6.00338280e-01 3.12661856e-01 3.54046643e-01 6.85704112e-01 -5.86643219e-01 -3.34512621e-01 -1.44275725e+00 3.97933796e-02 1.11905801e+00 1.03234041e+00 1.88416004e-01 -1.21720515e-01 -1.67331368e-01 8.75375569e-01 2.77752161e-01 8.10362995e-02 8.44408810e-01 -5.66382051e-01 6.34368479e-01 2.96255380e-01 -7.34320581e-02 -4.60378528e-01 -6.65825009e-01 -5.20345151e-01 -7.31046617e-01 -7.86021352e-02 3.51685345e-01 -2.27981344e-01 -1.21855044e+00 1.52360702e+00 8.37097019e-02 -1.72345296e-01 4.27700549e-01 5.39485991e-01 9.69507039e-01 4.59501892e-01 1.25178531e-01 -5.77292979e-01 1.69781435e+00 -7.69517362e-01 -1.44892180e+00 -2.06848517e-01 6.96764648e-01 -1.41049826e+00 6.30502224e-01 7.32626081e-01 -1.27441764e+00 -2.21298248e-01 -8.34018171e-01 -4.00081515e-01 -3.02270710e-01 -2.19614934e-02 2.42231190e-01 7.50732839e-01 -1.23531163e+00 8.34606171e-01 -8.04370522e-01 -4.95692551e-01 3.27508636e-02 5.48879623e-01 -6.37369633e-01 1.60743788e-01 -1.10725319e+00 1.15553308e+00 4.04615045e-01 3.28223586e-01 -1.43631950e-01 -6.73082232e-01 -1.19419229e+00 7.54083171e-02 3.43804985e-01 -6.84315681e-01 1.84019780e+00 -6.46436155e-01 -1.91704202e+00 8.51237357e-01 -2.53492892e-01 -6.61838353e-01 2.95684516e-01 -4.68610853e-01 -4.77392972e-01 2.25913063e-01 -4.76348214e-02 3.97088528e-01 6.52195811e-01 -7.39179969e-01 -5.62846601e-01 -2.52030551e-01 -3.69610190e-01 3.66206825e-01 -3.48771252e-02 3.04424644e-01 -5.31538427e-01 -9.57419932e-01 2.28060424e-01 -7.11616158e-01 -3.74156237e-01 -5.93375802e-01 -5.32100737e-01 -1.27841845e-01 5.21629453e-02 -1.33691740e+00 1.69572830e+00 -2.14070725e+00 5.13555147e-02 1.35355756e-01 2.33961493e-01 3.07864040e-01 2.31671393e-01 6.34265244e-01 -4.65356112e-01 1.41821653e-01 -3.38462800e-01 -1.01560140e+00 -1.81684986e-01 2.89844066e-01 -2.93557584e-01 4.78669345e-01 1.30577624e-01 6.16118312e-01 -6.19537592e-01 -5.50167263e-01 2.44870976e-01 4.91394550e-01 -4.96892720e-01 1.24640517e-01 1.78239346e-01 2.22346544e-01 -1.64323404e-01 5.81881404e-01 1.29562184e-01 -9.20791179e-02 3.34645957e-01 7.83782825e-02 -2.97265768e-01 1.11654425e+00 -9.90833879e-01 2.02379727e+00 -5.53986967e-01 4.13982362e-01 1.49775997e-01 -6.25455081e-01 6.70336723e-01 8.17567289e-01 3.56586307e-01 -4.91504401e-01 3.93384784e-01 3.91681463e-01 4.53670733e-02 -4.86191720e-01 8.38238597e-01 -4.18559551e-01 -7.42077753e-02 1.34933576e-01 1.04406446e-01 -2.39422411e-01 2.03164712e-01 1.87496141e-01 1.19801533e+00 9.69571546e-02 9.29993451e-01 -1.29845485e-01 4.70047921e-01 1.82099803e-03 3.89579028e-01 7.91885376e-01 -1.30169373e-02 9.85968590e-01 4.30099100e-01 -2.86306798e-01 -1.08732247e+00 -5.30768633e-01 -3.35560709e-01 5.57671010e-01 -5.49302816e-01 -6.88399792e-01 -1.02009130e+00 -4.62060839e-01 -5.22958219e-01 1.12783694e+00 -5.17591178e-01 5.60540371e-02 -8.56027603e-01 -3.60262513e-01 5.52641332e-01 3.35524261e-01 -2.74536192e-01 -1.27600849e+00 -5.54190695e-01 8.46110106e-01 -4.71242182e-02 -1.17852223e+00 -7.41807938e-01 9.08578992e-01 -9.49403048e-01 -7.71522164e-01 -6.26695335e-01 -9.42662179e-01 4.80253607e-01 -2.54617065e-01 9.55558836e-01 -6.04545735e-02 -1.67015567e-01 2.06708580e-01 -6.26157224e-01 -5.65416396e-01 -7.11310744e-01 1.19710434e-03 2.08437201e-02 -2.04191029e-01 3.74496311e-01 -5.32426834e-01 -3.24746937e-01 -6.01225615e-01 -1.04468834e+00 1.63633900e-03 9.56610620e-01 8.81640017e-01 7.43378878e-01 -4.15326089e-01 1.78765863e-01 -9.55538630e-01 7.44336247e-01 -1.86573058e-01 -9.53932777e-02 1.86544254e-01 -5.12332141e-01 1.00469910e-01 8.47130477e-01 -1.92191347e-01 -8.49800467e-01 2.10549638e-01 -5.60424268e-01 -2.06296504e-01 -4.71968234e-01 6.94676578e-01 1.46774068e-01 1.87671974e-01 6.90660655e-01 4.12993938e-01 9.39017534e-02 -6.12210453e-01 1.71104327e-01 1.06138420e+00 6.29608572e-01 7.18618110e-02 2.80034900e-01 1.54729456e-01 -2.77452379e-01 -8.35582018e-01 -7.70650446e-01 -7.92333186e-01 -5.19917130e-01 1.34589896e-01 8.14142346e-01 -6.39337957e-01 -5.04702389e-01 2.02986225e-01 -1.54293168e+00 2.62251087e-02 -4.64509130e-01 7.72365630e-01 -5.51034808e-01 6.46049142e-01 -8.11210215e-01 -6.84722483e-01 -6.94909215e-01 -1.06511152e+00 1.04958546e+00 1.17482804e-01 -8.92027020e-01 -7.20207632e-01 1.82752669e-01 4.20518696e-01 1.41128168e-01 -1.42839551e-01 6.48001969e-01 -1.14241374e+00 -5.47668599e-02 -3.42316210e-01 3.55083108e-01 4.05968487e-01 3.63571316e-01 -3.49057198e-01 -7.83269882e-01 1.01892151e-01 4.59684610e-01 6.89566657e-02 7.26644754e-01 3.66449535e-01 9.38072324e-01 -5.32196462e-01 -1.70616955e-01 5.37559927e-01 1.06562328e+00 3.86009216e-01 7.51489878e-01 3.33146036e-01 4.00777578e-01 5.08873880e-01 1.91208035e-01 4.54192489e-01 4.39501375e-01 4.72750008e-01 -5.48831522e-02 8.39539468e-02 -3.53639126e-01 1.19036227e-01 1.50495321e-01 1.19776845e+00 3.70733976e-01 -2.86956728e-01 -1.09112251e+00 7.12356627e-01 -1.52438688e+00 -6.22826040e-01 1.47539089e-02 2.07466555e+00 1.22202110e+00 1.17565058e-01 -4.77378309e-01 2.44404718e-01 5.01733184e-01 -6.04286306e-02 -5.78714013e-02 -1.11040807e+00 1.01997994e-01 7.16387808e-01 6.36414826e-01 7.60747731e-01 -9.62310374e-01 7.80582488e-01 6.54780674e+00 7.07046330e-01 -1.02311361e+00 7.95197487e-02 4.25355971e-01 -3.21170747e-01 -1.87864378e-01 -2.99357057e-01 -8.82794499e-01 4.06379968e-01 1.44222140e+00 2.38541186e-01 3.81371945e-01 5.93773007e-01 4.81246740e-01 -1.12785794e-01 -1.10193741e+00 1.09075141e+00 4.96956319e-01 -1.34853470e+00 -1.01820841e-01 -1.34000108e-01 2.87565500e-01 -1.03270598e-01 -2.75388867e-01 1.70077160e-01 -1.38520539e-01 -1.00201476e+00 8.07738841e-01 5.81473887e-01 6.20988727e-01 -5.95252931e-01 9.56846356e-01 1.80365145e-01 -6.19294286e-01 1.75094754e-01 1.15407884e-01 5.00631705e-02 4.48126704e-01 4.17436749e-01 -1.19221711e+00 5.87847292e-01 4.81940478e-01 3.46334249e-01 -5.41834906e-02 1.01582873e+00 -4.44454759e-01 4.91798341e-01 -4.94898558e-01 -1.68996990e-01 4.60226595e-01 2.65010953e-01 6.84017241e-01 1.60693741e+00 3.57276708e-01 4.30882603e-01 -1.07253768e-01 3.01170051e-01 -1.53823331e-01 7.40278959e-01 -2.72267967e-01 -1.01717450e-01 2.28701591e-01 1.16061044e+00 -6.24630034e-01 -3.72067034e-01 -4.27041024e-01 1.01098454e+00 3.15956742e-01 8.34836885e-02 -3.41581494e-01 -5.93514025e-01 4.48545396e-01 5.83786815e-02 5.76913476e-01 8.18481017e-03 -5.49979746e-01 -8.41499388e-01 3.70839477e-01 -1.13196373e+00 2.84916252e-01 -5.29860497e-01 -9.47390854e-01 1.06604016e+00 -4.16973323e-01 -1.11691988e+00 -5.92286170e-01 -5.10496736e-01 -1.42460093e-01 9.81389225e-01 -1.68181431e+00 -7.11474836e-01 3.62436444e-01 3.26133519e-01 7.32866466e-01 -8.52195397e-02 1.25233209e+00 4.25428420e-01 -3.58749241e-01 5.44536889e-01 1.53425723e-01 -7.93277323e-02 9.32515383e-01 -1.35983658e+00 2.78615534e-01 8.04625273e-01 2.40814656e-01 1.08981109e+00 9.84361231e-01 -7.57708549e-01 -1.02873313e+00 -8.88762414e-01 1.76994348e+00 -2.64421612e-01 5.71929157e-01 -1.62528574e-01 -9.43247080e-01 6.47491097e-01 4.71368641e-01 -4.46224004e-01 1.04936314e+00 3.05138435e-02 7.04893619e-02 4.90312278e-03 -8.66024971e-01 7.28643715e-01 4.64448214e-01 -5.00684083e-01 -1.21749139e+00 5.38110912e-01 6.89220667e-01 -1.07608378e+00 -7.11174726e-01 2.11071193e-01 4.32634026e-01 -5.08280516e-01 4.45348442e-01 -4.47529346e-01 3.11760008e-01 -2.23333940e-01 -4.80074296e-03 -1.06811833e+00 -1.81352451e-01 -1.29131210e+00 -6.57550171e-02 9.45633233e-01 6.64038122e-01 -2.76727110e-01 6.09686673e-01 7.89447308e-01 -7.54284382e-01 -6.99256241e-01 -1.33937192e+00 -2.78370380e-01 -1.42893970e-01 -7.01051772e-01 3.43712717e-01 6.09274864e-01 4.34426486e-01 2.24620700e-01 -4.72989768e-01 2.09718905e-02 1.34417266e-01 -4.29206789e-01 1.85113207e-01 -7.77817130e-01 -1.89113662e-01 -6.01208627e-01 -3.69328707e-01 -7.70054162e-01 3.79690863e-02 -7.99236178e-01 3.23936284e-01 -1.63249350e+00 -1.96883887e-01 9.06905159e-03 -1.95125431e-01 5.82050383e-01 -6.01374656e-02 1.40702546e-01 8.05063322e-02 -1.26719639e-01 -3.74608845e-01 1.70766801e-01 7.71371603e-01 1.09821223e-01 -5.75793207e-01 -3.69588435e-02 -1.03820384e+00 6.32161796e-01 8.50666046e-01 -7.79091716e-01 6.60686195e-02 -1.96945727e-01 2.05104455e-01 3.04688841e-01 -1.01854868e-01 -6.79834902e-01 5.98291218e-01 2.78953403e-01 2.26726457e-01 -6.76911592e-01 2.45245188e-01 -8.27303112e-01 8.99022445e-02 3.98881286e-01 -3.88156801e-01 1.59877673e-01 5.59395194e-01 1.97507545e-01 -2.47960612e-01 -5.52034736e-01 4.23492730e-01 -2.86249965e-01 -2.24632114e-01 -3.55526984e-01 -9.64738429e-01 -3.40203315e-01 6.21465564e-01 -3.42459559e-01 4.79283109e-02 -3.05371821e-01 -1.17346776e+00 -1.12309933e-01 -2.17692554e-02 2.67449915e-01 6.40102565e-01 -7.69047022e-01 -7.09384382e-01 1.10820502e-01 -1.69954553e-01 5.58298901e-02 4.20848787e-01 1.05309582e+00 -7.66951025e-01 8.22484255e-01 2.55874932e-01 -1.59063369e-01 -1.39598823e+00 5.66340089e-01 3.57480288e-01 -7.34927297e-01 -1.06435525e+00 6.13785684e-01 -1.49495170e-01 -3.45227748e-01 4.31680977e-01 -7.67989635e-01 -4.90214437e-01 2.14646265e-01 9.38847005e-01 -1.25159919e-01 8.99428070e-01 -4.94706273e-01 -6.04050696e-01 3.39604318e-01 -5.73599637e-01 -2.06062436e-01 1.33827639e+00 -3.66829425e-01 -4.03488651e-02 2.53139853e-01 8.13492417e-01 1.70753896e-01 -5.00956953e-01 -2.27534428e-01 9.64469537e-02 1.84281021e-01 2.21745208e-01 -1.19772983e+00 -4.94269192e-01 5.45997262e-01 2.13685289e-01 7.02105612e-02 1.13676023e+00 -9.90739018e-02 8.45062673e-01 7.16991723e-01 3.47772730e-03 -1.18530810e+00 -4.22441930e-01 5.80873430e-01 1.08605480e+00 -9.27579999e-01 1.40283350e-02 -2.54240751e-01 -5.39864540e-01 1.33686340e+00 -1.59939647e-01 7.26944059e-02 5.47239184e-01 5.44392586e-01 5.39863467e-01 -4.66294289e-02 -5.92755139e-01 -5.98038211e-02 4.69437629e-01 2.33135536e-01 7.19666779e-01 -3.40660848e-02 -6.31128430e-01 7.14281559e-01 -5.91677070e-01 -9.59745497e-02 7.81967700e-01 1.01271093e+00 -1.29347816e-01 -1.11175871e+00 -6.37882948e-01 4.62284803e-01 -1.26711178e+00 -8.17810357e-01 -2.08746627e-01 6.62949026e-01 -9.11429673e-02 1.09988260e+00 -4.32335632e-03 7.42281750e-02 5.56610107e-01 4.21971768e-01 3.46396267e-01 -1.03887236e+00 -1.21030056e+00 5.45832992e-01 3.13049525e-01 -5.39757967e-01 -3.17479879e-01 -1.08551931e+00 -1.39456511e+00 1.23751484e-01 -1.94557145e-01 1.44279912e-01 8.62349093e-01 1.14601219e+00 3.06763649e-01 1.01510596e+00 4.02065292e-02 -5.66372395e-01 -5.07795572e-01 -1.28025138e+00 -5.53864598e-01 1.40383095e-01 7.17279077e-01 1.25687364e-02 -1.69253334e-01 2.84034371e-01]

[11.257030487060547, 9.725858688354492]

8dbeac49-5092-4a3d-ab28-1bb9412fa0a0

automated-essay-scoring-with-discourse-aware

https://aclanthology.org/W19-4450

https://aclanthology.org/W19-4450.pdf

Automated Essay Scoring with Discourse-Aware Neural Models

Automated essay scoring systems typically rely on hand-crafted features to predict essay quality, but such systems are limited by the cost of feature engineering. Neural networks offer an alternative to feature engineering, but they typically require more annotated data. This paper explores network structures, contextualized embeddings and pre-training strategies aimed at capturing discourse characteristics of essays. Experiments on three essay scoring tasks show benefits from all three strategies in different combinations, with simpler architectures being more effective when less training data is available.

['Mari Ostendorf', 'Farah Nadeem', 'Huy Nguyen', 'Yang Liu']

2019-08-01

ws-2019-8

['automated-essay-scoring']

['natural-language-processing']

[-1.27656505e-01 7.86587074e-02 -4.24154192e-01 -6.95657909e-01 -5.42877674e-01 -6.14825249e-01 7.28448987e-01 6.42366111e-01 -7.56507874e-01 7.55455256e-01 6.62145734e-01 -2.71721333e-01 -3.60912800e-01 -7.66655505e-01 1.03525501e-02 -8.10680017e-02 2.90445030e-01 4.00631815e-01 -5.26285022e-02 -3.82370174e-01 8.60668123e-01 1.79834530e-01 -1.40993690e+00 6.28951490e-02 8.84063482e-01 7.10235298e-01 -2.18884442e-02 9.36175883e-01 -5.30982018e-01 8.46904337e-01 -8.59401226e-01 -5.84165752e-01 4.18060757e-02 -4.28819150e-01 -8.41585457e-01 -5.69544658e-02 3.16972435e-01 -2.81619996e-01 -4.28784132e-01 6.80037498e-01 4.65073615e-01 3.15345436e-01 8.27492654e-01 -7.56570101e-01 -1.08607519e+00 4.07526791e-01 5.28788194e-02 3.70820135e-01 7.29129434e-01 7.69394487e-02 1.58972836e+00 -1.02597725e+00 3.91424209e-01 7.28980243e-01 1.02695251e+00 5.84757626e-01 -1.22366536e+00 -1.27624929e-01 -2.83132464e-01 4.40296590e-01 -6.86818063e-01 -4.15797800e-01 8.86873245e-01 -5.37848592e-01 1.11901879e+00 1.75072208e-01 7.33125448e-01 9.19460952e-01 5.16115986e-02 7.38583088e-01 1.23283160e+00 -7.29427218e-01 1.03324339e-01 4.29744542e-01 8.31156194e-01 6.86014593e-01 3.90805185e-01 3.57307382e-02 -7.82389283e-01 -3.11262369e-01 4.39508766e-01 -4.22635600e-02 -3.48507345e-01 -1.07228711e-01 -6.66777670e-01 1.34096837e+00 8.26068446e-02 5.42451143e-01 -2.95227766e-01 -1.68576062e-01 5.19507706e-01 7.93711722e-01 5.62560558e-01 1.07743478e+00 -3.88015747e-01 -5.43128014e-01 -1.16914368e+00 4.34649318e-01 1.06969190e+00 3.88248712e-01 6.95239067e-01 7.54520670e-03 -3.78490925e-01 1.07471621e+00 1.41608909e-01 -2.66921669e-01 9.39916015e-01 -7.84060955e-01 4.81729120e-01 1.07760882e+00 1.17095836e-01 -1.07322955e+00 -5.61181724e-01 -1.19670190e-01 -2.22790599e-01 1.65251106e-01 5.59541106e-01 -2.91519582e-01 -6.11467719e-01 1.26767206e+00 -1.37242824e-01 -5.14552832e-01 8.16212371e-02 8.21340799e-01 1.02135599e+00 4.70062792e-01 -1.36789694e-01 7.56227449e-02 1.09030473e+00 -1.08456135e+00 -7.57175863e-01 -1.31646410e-01 6.94330633e-01 -5.58023870e-01 1.39568806e+00 3.67026538e-01 -1.21091497e+00 -3.71436924e-01 -1.16078496e+00 -2.63500065e-01 -5.33710837e-01 1.16849253e-02 5.93714476e-01 1.02554274e+00 -9.59149718e-01 9.41990316e-01 -3.45522285e-01 -5.31335175e-02 5.11605084e-01 4.39556539e-01 -3.16267252e-01 6.37216344e-02 -1.16993415e+00 1.34775984e+00 1.45857915e-01 -1.60691082e-01 -6.88548908e-02 -6.81374907e-01 -9.17578995e-01 4.54104900e-01 -7.20715597e-02 -4.25811231e-01 1.33618200e+00 -8.65178049e-01 -1.95247304e+00 6.33661985e-01 1.01680301e-01 -3.07237506e-01 2.67288834e-01 -2.70551860e-01 -1.16364881e-01 7.81321228e-02 -3.59377086e-01 2.19130561e-01 5.63249171e-01 -3.78625989e-01 -3.54755133e-01 -8.39901567e-02 2.27892607e-01 2.27668539e-01 -1.14124238e+00 1.93507314e-01 2.32496664e-01 -5.00125647e-01 -4.52542514e-01 -5.30332446e-01 -1.02811880e-01 -1.78926632e-01 1.15356088e-01 -8.34855258e-01 7.15171576e-01 -5.37625909e-01 1.50320053e+00 -1.44001651e+00 -9.40665603e-02 1.52258679e-01 3.80287856e-01 6.72162354e-01 -1.81520477e-01 5.30328751e-01 2.35722169e-01 1.00130916e-01 1.46080106e-01 -5.74411750e-02 3.66225541e-01 1.35490866e-02 3.92056465e-01 1.88020691e-01 3.45665365e-01 8.82783771e-01 -7.96495199e-01 -4.46296036e-01 1.46282077e-01 9.28793326e-02 -6.32632554e-01 2.98999786e-01 -1.74764574e-01 -2.26166874e-01 -3.11453074e-01 4.32485521e-01 -7.75260553e-02 -2.13559210e-01 2.40608588e-01 2.91179538e-01 -9.09806266e-02 9.15544331e-01 -9.21208620e-01 1.23089957e+00 -5.61479747e-01 1.17338181e+00 -1.68950275e-01 -9.90076065e-01 1.22744596e+00 4.52818006e-01 3.82277519e-02 -5.27389526e-01 3.36653143e-01 1.38433442e-01 2.30010584e-01 -7.40277588e-01 9.11133707e-01 -1.01557650e-01 8.53401609e-04 9.35242593e-01 2.10756496e-01 -1.27947420e-01 4.73934889e-01 8.52043405e-02 1.38622403e+00 -2.69981474e-01 5.51413059e-01 -2.04749167e-01 3.55635285e-01 1.37918115e-01 4.30965960e-01 6.66548014e-01 -3.68854642e-01 4.20806348e-01 6.59005642e-01 -6.98889315e-01 -1.20025110e+00 -3.44379306e-01 -1.83206186e-01 1.18885541e+00 -4.45574015e-01 -6.34675264e-01 -5.68476140e-01 -8.54027033e-01 9.83568728e-02 5.96468270e-01 -7.67687559e-01 -8.39441493e-02 -6.33286059e-01 -5.05525410e-01 4.77414638e-01 7.20026314e-01 1.07728712e-01 -1.14347744e+00 -1.00255716e+00 4.66399014e-01 3.37196201e-01 -5.55636406e-01 -3.19047421e-01 5.17098427e-01 -9.14799690e-01 -1.15931296e+00 -6.06872916e-01 -6.21858001e-01 4.60814387e-01 1.10232249e-01 1.26330304e+00 6.46110296e-01 -2.49442905e-01 3.69815081e-01 -3.96615356e-01 -4.74527061e-01 -3.22128981e-01 3.15967321e-01 -8.09514821e-02 -5.75392008e-01 8.19355726e-01 -2.49278799e-01 -3.23311150e-01 -6.10267418e-03 -6.09293818e-01 -4.57543552e-01 4.33267653e-01 1.46181309e+00 -2.23107591e-01 -3.42290133e-01 8.48121405e-01 -1.05063164e+00 1.68489587e+00 -2.80779183e-01 -1.64039388e-01 2.55879581e-01 -1.07009482e+00 1.68437406e-01 7.06077874e-01 -3.47512811e-01 -9.81928527e-01 -3.95229816e-01 -8.83936733e-02 1.38524696e-01 -5.38685098e-02 9.50692177e-01 2.83040524e-01 -1.94610447e-01 9.75719810e-01 -2.75742441e-01 2.48808190e-01 -5.17622471e-01 7.57811368e-02 7.53751397e-01 4.18908820e-02 -5.85150719e-01 3.42814296e-01 -2.93727070e-01 -3.50829691e-01 -7.24012017e-01 -8.92090440e-01 -3.87955099e-01 -6.50850654e-01 -3.05831552e-01 3.90475422e-01 -2.40852982e-01 -4.24382508e-01 1.54292852e-01 -9.86338258e-01 -1.05962083e-01 -3.93710107e-01 5.93876064e-01 -3.39411587e-01 1.87377289e-01 -7.35152900e-01 -5.41823626e-01 -4.56191182e-01 -7.22961903e-01 1.72700554e-01 7.72085190e-01 -8.15706909e-01 -1.31728339e+00 4.45535302e-01 4.19246644e-01 6.72722220e-01 -1.05892010e-01 1.13393271e+00 -1.25337493e+00 5.68635948e-02 -6.28908336e-01 -1.99288025e-01 4.42075044e-01 1.08224690e-01 1.77558497e-01 -8.99503708e-01 1.71197280e-01 -9.74027067e-02 -6.24145985e-01 8.65670204e-01 2.16076881e-01 1.01985002e+00 -4.38992709e-01 3.07049960e-01 1.07202873e-01 1.15701282e+00 -6.36257008e-02 2.48503804e-01 7.08256721e-01 3.38472962e-01 6.74179971e-01 4.84305203e-01 5.07127762e-01 4.17743713e-01 4.31700617e-01 1.06653273e-01 2.49045193e-01 -5.23952916e-02 1.17872031e-02 3.13520372e-01 1.01329088e+00 -2.27978677e-01 -6.31944165e-02 -1.03448617e+00 7.18987584e-01 -1.69266927e+00 -1.07727849e+00 -2.94649333e-01 1.65921748e+00 1.15034926e+00 1.91014901e-01 2.82926679e-01 5.33505380e-01 4.54821497e-01 3.23181152e-01 -2.83430398e-01 -1.04978716e+00 -5.39032789e-03 7.22793281e-01 6.86149746e-02 5.05408823e-01 -8.51140857e-01 5.66881299e-01 7.51010513e+00 3.53379816e-01 -7.76301444e-01 6.66416287e-02 1.90231666e-01 -4.16776329e-01 -3.49842727e-01 -1.94675967e-01 -6.37735963e-01 2.88384140e-01 1.27689672e+00 -3.12319189e-01 -4.57955636e-02 9.14442182e-01 -1.98246036e-02 -1.69549853e-01 -1.09601605e+00 5.44693470e-01 2.17244953e-01 -1.63748658e+00 -3.58234227e-01 -1.43409118e-01 6.35273397e-01 -2.30564877e-01 -1.31844759e-01 4.29969519e-01 3.49604815e-01 -1.19377041e+00 4.17156398e-01 5.01968980e-01 2.73727953e-01 -7.88580954e-01 1.09305954e+00 2.01361850e-01 -4.95427608e-01 -3.67440581e-01 -4.48236674e-01 -8.73485625e-01 -1.26077414e-01 4.88474548e-01 -8.17471802e-01 -6.86574914e-03 3.00076604e-01 6.28419161e-01 -7.83611178e-01 1.05316293e+00 -4.65283841e-01 7.45523036e-01 -2.45866552e-03 -9.16216731e-01 4.41429734e-01 -1.17399476e-01 1.87684387e-01 1.32709122e+00 2.52703484e-02 9.80091318e-02 -6.90785004e-03 6.20515049e-01 -1.44878119e-01 8.65488350e-02 -5.76108336e-01 -2.11534292e-01 5.53073943e-01 1.45906079e+00 -3.71488780e-01 -2.94875592e-01 -5.30203164e-01 5.45501769e-01 6.92520618e-01 -1.83620498e-01 -3.92096251e-01 -7.14850843e-01 4.80292976e-01 1.00018485e-02 1.32947326e-01 -2.38680780e-01 -7.38408625e-01 -9.03043985e-01 -1.92148417e-01 -9.82351243e-01 3.53970140e-01 -4.18304563e-01 -1.33083391e+00 3.22070211e-01 -3.82852882e-01 -6.77558839e-01 -5.31706810e-01 -1.00746572e+00 -9.90016043e-01 7.98757255e-01 -1.43900204e+00 -7.32485533e-01 -2.14589566e-01 2.09559888e-01 5.87472796e-01 -5.91980457e-01 1.19531882e+00 2.05819383e-01 -4.74946946e-01 6.73381627e-01 2.16381788e-01 1.99306443e-01 8.28273356e-01 -1.57669520e+00 4.77165636e-03 2.79722720e-01 3.62197042e-01 5.05501032e-01 6.07913375e-01 -2.17298746e-01 -1.04959011e+00 -4.65298355e-01 1.45044184e+00 -6.07983053e-01 8.12591732e-01 -2.40233503e-02 -1.14315104e+00 2.92880327e-01 5.34744918e-01 -3.84186119e-01 1.23618042e+00 7.95877874e-01 -2.88366497e-01 3.02674025e-01 -1.07453358e+00 4.87100422e-01 3.53232384e-01 -6.48456216e-01 -1.12074840e+00 1.64108202e-01 4.09666486e-02 -1.34548068e-01 -1.06044102e+00 -1.14476345e-01 6.62907064e-01 -1.03029513e+00 6.69805944e-01 -8.64581287e-01 9.20750856e-01 2.75439769e-01 1.77581102e-01 -1.59951293e+00 -4.88905579e-01 -3.33602220e-01 -3.56512010e-01 1.15219045e+00 5.22674084e-01 -3.53869915e-01 1.15358484e+00 1.05680835e+00 -1.23347312e-01 -1.04712200e+00 -6.43195570e-01 -6.44251287e-01 3.05230975e-01 3.74034047e-02 5.77616096e-01 1.19240522e+00 6.66279256e-01 6.59243166e-01 -1.54652536e-01 -7.03572810e-01 1.12415448e-01 -5.78089133e-02 5.51408350e-01 -1.58792853e+00 -6.45105988e-02 -9.70446885e-01 -4.30419773e-01 -5.29197097e-01 3.46443623e-01 -7.19966650e-01 -8.71071219e-02 -1.52308273e+00 1.72602385e-01 -3.79861802e-01 -2.15326145e-01 4.54523981e-01 -4.70486492e-01 2.21401438e-01 5.99284396e-02 4.25637700e-02 -3.30043793e-01 4.35680658e-01 7.95227289e-01 -9.60564762e-02 -2.78313935e-01 -1.19075611e-01 -8.16450655e-01 9.42762017e-01 1.23366058e+00 -5.41739047e-01 -2.93285668e-01 -5.37657738e-01 2.88692772e-01 -1.28013968e-01 2.09282726e-01 -7.66356230e-01 4.53708172e-01 -2.95097768e-01 5.76872706e-01 -3.75292122e-01 2.61706084e-01 -3.95946622e-01 -4.52049792e-01 1.82447761e-01 -7.74865627e-01 4.28702861e-01 -6.32137507e-02 2.47321218e-01 -3.69162142e-01 -1.20075035e+00 7.24760473e-01 -2.04767585e-01 -4.95373100e-01 -1.52349070e-01 -5.62183380e-01 1.53650343e-01 6.76862061e-01 -5.80249310e-01 -2.93498278e-01 -4.49525088e-01 -5.25974095e-01 3.68226394e-02 3.60143572e-01 4.01853502e-01 6.23109519e-01 -1.24069309e+00 -6.49451435e-01 1.21328987e-01 -4.05850559e-02 -4.99093413e-01 -1.89669162e-01 5.74002504e-01 -6.09362662e-01 6.68097019e-01 -4.01670754e-01 -7.86826387e-02 -1.53460658e+00 -1.21771164e-01 5.80306873e-02 -7.10955620e-01 -5.33133566e-01 9.59397137e-01 -7.79461563e-01 -5.51811397e-01 2.17210144e-01 6.32300153e-02 -5.76438725e-01 4.22013611e-01 6.65833116e-01 5.09312928e-01 3.50546807e-01 -1.40455082e-01 -5.58501109e-02 1.88599244e-01 -3.66302490e-01 -8.13536495e-02 1.70111668e+00 3.22384477e-01 2.47225851e-01 3.73559058e-01 8.39597404e-01 -8.02277103e-02 -9.53536510e-01 -3.22278470e-01 6.32685900e-01 -6.59242928e-01 3.64796847e-01 -8.71660411e-01 -6.93756759e-01 1.01876104e+00 5.22063337e-02 4.52177346e-01 5.92174947e-01 -4.79627818e-01 6.27416849e-01 6.09361827e-01 -1.00316793e-01 -1.74861443e+00 4.05882657e-01 8.26053739e-01 6.84321404e-01 -1.07318270e+00 3.28993380e-01 3.81952465e-01 -6.19033933e-01 1.67994404e+00 6.34960353e-01 -5.51136136e-01 6.27339303e-01 1.16293296e-01 6.25208626e-03 -3.92948270e-01 -9.03368354e-01 1.55457452e-01 4.64860231e-01 4.81076300e-01 9.88432348e-01 -2.24444978e-02 -7.49172390e-01 1.02683377e+00 -4.26465333e-01 -3.07013490e-03 8.44378471e-01 1.08261263e+00 -6.96277738e-01 -1.40260506e+00 -2.97083676e-01 1.05502343e+00 -6.74336374e-01 -1.45813465e-01 -8.10795426e-01 8.10558915e-01 -2.26488099e-01 9.07391250e-01 -1.61059201e-02 -1.86239660e-01 3.34865838e-01 4.89314944e-01 6.31233454e-01 -8.51824164e-01 -1.28323269e+00 -5.30166686e-01 5.98684669e-01 -1.77203521e-01 -4.64919627e-01 -8.03635538e-01 -7.75876164e-01 -6.73017979e-01 -8.02268744e-01 3.19958448e-01 6.93974733e-01 9.43116426e-01 2.27456570e-01 5.48138916e-01 4.69323486e-01 -6.95358455e-01 -1.04437268e+00 -1.21053350e+00 -5.11751890e-01 1.95231065e-01 2.53123552e-01 -6.41906559e-01 -2.82569617e-01 -1.73940584e-01]

[11.305092811584473, 9.327591896057129]

665e7eeb-397a-463d-8d59-15f02bec61c6

audio-barlow-twins-self-supervised-audio

2209.14345

https://arxiv.org/abs/2209.14345v1

https://arxiv.org/pdf/2209.14345v1.pdf

Audio Barlow Twins: Self-Supervised Audio Representation Learning

The Barlow Twins self-supervised learning objective requires neither negative samples or asymmetric learning updates, achieving results on a par with the current state-of-the-art within Computer Vision. As such, we present Audio Barlow Twins, a novel self-supervised audio representation learning approach, adapting Barlow Twins to the audio domain. We pre-train on the large-scale audio dataset AudioSet, and evaluate the quality of the learnt representations on 18 tasks from the HEAR 2021 Challenge, achieving results which outperform, or otherwise are on a par with, the current state-of-the-art for instance discrimination self-supervised learning approaches to audio representation learning. Code at https://github.com/jonahanton/SSL_audio.

['Bjorn W. Schuller', 'Pancham Shukla', 'Harry Coppock', 'Jonah Anton']

2022-09-28

['environmental-sound-classification']

['audio']

[ 6.3489944e-01 2.1561986e-01 -6.5963879e-02 -4.8996505e-01 -1.6303165e+00 -5.0842887e-01 6.6291636e-01 3.4469047e-01 -3.0959782e-01 3.9912844e-01 6.1080468e-01 2.9051554e-01 -4.1381276e-01 -4.0473774e-01 -7.0630640e-01 -3.9345676e-01 -4.9303821e-01 6.2825894e-01 1.2062046e-01 -1.5046081e-01 -1.5449743e-01 -2.9926878e-01 -2.2116563e+00 7.6230043e-01 2.0100521e-01 1.4520819e+00 -4.4982278e-01 1.0364136e+00 2.7444512e-01 8.3762217e-01 -1.7837113e-01 -2.1205574e-01 2.6945418e-01 -1.0248015e-01 -8.7741107e-01 -4.0576187e-01 1.1798915e+00 -3.2782357e-02 -4.8394349e-01 7.3859286e-01 8.7570918e-01 1.1509267e-01 5.6306314e-01 -1.3634977e+00 -5.5845284e-01 1.1107079e+00 -5.5098861e-01 4.2168829e-01 5.2242678e-01 -7.5056940e-02 1.5261754e+00 -6.4915836e-01 1.8308157e-01 1.2845393e+00 1.0778520e+00 6.1149704e-01 -1.4851182e+00 -1.1723567e+00 -3.2981534e-02 4.5974624e-01 -1.4624734e+00 -1.1431655e+00 7.4382126e-01 -6.2202895e-01 8.1157410e-01 2.2826011e-01 5.4014313e-01 1.4546926e+00 -5.6381369e-01 9.7550237e-01 9.0514940e-01 -6.0362363e-01 3.2718173e-01 -1.2183820e-01 1.9614592e-01 5.1669770e-01 -3.2102916e-01 3.9617467e-01 -1.1093371e+00 -4.5567459e-01 2.0327465e-01 -5.4616445e-01 -7.9251811e-02 -3.4331557e-01 -1.0837671e+00 7.3345542e-01 2.4497990e-01 3.1311753e-01 -8.4465586e-02 7.0798993e-01 8.7363076e-01 4.9972063e-01 6.2541455e-01 2.8993106e-01 -3.9706546e-01 -6.1816233e-01 -1.0977460e+00 2.6532710e-01 4.8022258e-01 5.7304937e-01 5.0735211e-01 5.1276976e-01 -5.9269354e-02 1.1636412e+00 1.6551946e-01 2.4162374e-01 7.0836258e-01 -1.1895357e+00 2.7663800e-01 -8.0259502e-02 -3.7735721e-01 -3.7092495e-01 -2.7398556e-01 -6.3625222e-01 -6.7398345e-01 1.6807097e-01 3.3663777e-01 -9.3728215e-02 -7.1670860e-01 1.8319937e+00 6.0128357e-02 7.3393089e-01 3.8603105e-02 6.6524947e-01 1.0078723e+00 6.5648532e-01 -6.6765279e-02 1.0294471e-01 1.2074895e+00 -8.7259084e-01 -3.0843386e-01 -2.4370329e-01 3.3122689e-01 -8.1926286e-01 9.5336819e-01 9.3116897e-01 -8.7562156e-01 -9.2612284e-01 -1.2170751e+00 4.5528419e-02 -3.3171780e-02 1.1774406e-01 8.7401664e-01 7.1645272e-01 -9.4630551e-01 6.9661230e-01 -6.8936837e-01 -2.7010801e-01 4.7014993e-01 1.9093227e-01 -5.3302079e-01 7.8689486e-02 -1.1974148e+00 3.2561308e-01 4.4734940e-01 -4.1639894e-01 -1.3999695e+00 -1.3283767e+00 -8.6885965e-01 -1.9665654e-01 2.7121097e-01 -4.3427438e-01 1.8638756e+00 -1.0627143e+00 -1.3115073e+00 9.8149294e-01 1.8622106e-01 -9.8868626e-01 4.9068466e-01 -5.8106768e-01 -7.5952023e-01 2.4101938e-01 1.3190949e-01 1.0210787e+00 1.1811268e+00 -1.0577140e+00 -7.1036428e-01 -2.4959680e-01 -2.1279383e-01 -2.8676167e-02 -3.8681126e-01 -4.0423404e-02 -7.3677689e-02 -7.2230172e-01 -3.6229506e-01 -8.8412905e-01 1.2526616e-01 1.5424606e-01 -2.4008390e-01 -3.7482715e-01 6.4486337e-01 -5.4235333e-01 9.7112513e-01 -2.6069922e+00 -1.4836644e-01 -8.5348718e-02 2.9365698e-01 7.2479665e-02 -5.3633398e-01 3.9867809e-01 -6.7572421e-01 -3.0351257e-01 -2.8033245e-01 -7.2194451e-01 2.8400481e-01 -2.4879070e-03 -5.4824752e-01 3.8846320e-01 1.3286112e-01 2.2998057e-01 -1.1014364e+00 -2.3496936e-01 4.0899452e-02 5.5562162e-01 -7.1098763e-01 1.9341512e-01 -7.2628036e-02 4.9126238e-02 -7.1611002e-02 6.6218615e-01 2.1966635e-01 2.5864971e-01 -1.5983109e-01 -2.7381521e-01 2.0990033e-02 5.3790796e-01 -1.2606344e+00 2.1016853e+00 -4.9970609e-01 8.6587834e-01 5.6991383e-02 -1.1857250e+00 8.4647793e-01 4.8239762e-01 6.5372276e-01 -5.4989928e-01 -1.6071962e-01 1.7545719e-01 -5.8741976e-02 -2.7765679e-01 2.6463521e-01 -2.6762614e-01 -6.0005140e-02 4.5570219e-01 7.7157062e-01 -1.5120897e-01 4.8014071e-02 2.6473939e-01 1.1856902e+00 1.1547305e-01 2.3332626e-01 -4.8191834e-02 1.7983091e-01 -2.7727810e-01 4.2200524e-01 8.1370109e-01 -1.4799704e-01 1.0200654e+00 8.4885180e-02 -1.5267599e-01 -8.1423116e-01 -1.5108790e+00 -2.7386948e-01 1.6219090e+00 -7.0647705e-01 -9.3621534e-01 -5.8856004e-01 -3.3285832e-01 3.2223165e-01 5.1284695e-01 -7.5737542e-01 -3.1638360e-01 2.7216133e-03 -2.5948894e-01 1.2637683e+00 6.4175874e-01 2.2431922e-01 -8.6413670e-01 -4.1384712e-01 3.0433592e-01 4.3464556e-02 -9.9915475e-01 -2.6027849e-01 4.1103670e-01 -5.1860696e-01 -1.0219254e+00 -6.4689958e-01 -4.6648359e-01 -2.5298566e-01 -1.0781129e-01 1.2476107e+00 -4.8238859e-01 -7.8315103e-01 8.1473452e-01 -3.4440589e-01 -8.6160707e-01 -3.5803559e-01 2.2668608e-01 5.7229400e-01 1.4511675e-01 1.9267832e-01 -9.6778005e-01 -4.0811533e-01 -1.5031560e-01 -6.2643749e-01 -3.4073687e-01 2.4507311e-01 7.6161134e-01 5.5672282e-01 -1.2118195e-01 9.6614462e-01 -6.4890176e-01 4.1426465e-01 -5.2082902e-01 -2.2636168e-01 -2.7648073e-01 -5.6166339e-01 -1.0533282e-01 3.1645617e-01 -4.4886309e-01 -6.2945461e-01 2.8310558e-01 -2.9840079e-01 -7.9440403e-01 -4.6204606e-01 1.4318235e-01 2.0956054e-01 2.2443683e-01 1.0200917e+00 3.1789247e-02 1.9514054e-01 -7.1938133e-01 5.2604616e-01 9.3802387e-01 8.6595851e-01 -6.9690073e-01 7.2906786e-01 4.1050720e-01 -3.3922890e-01 -7.9936028e-01 -1.1073411e+00 -7.7980185e-01 -4.0240639e-01 -1.3656804e-01 6.2315750e-01 -1.4387978e+00 -6.3946861e-01 3.9011404e-01 -7.0605171e-01 -2.5406703e-01 -7.8033447e-01 5.3455204e-01 -1.0204909e+00 1.7087454e-01 -4.1989914e-01 -1.1006830e+00 -4.5166284e-01 -4.7647411e-01 1.2870708e+00 -1.9706807e-01 -6.7207807e-01 -6.7309713e-01 6.5280509e-01 6.3498557e-01 4.0035617e-01 -3.4749165e-02 6.4693964e-01 -9.0016454e-01 6.7020074e-02 -1.5750585e-01 -1.0029064e-01 6.5083319e-01 4.5183752e-02 -2.3073759e-02 -1.8992729e+00 -3.4589332e-01 -5.7396752e-01 -9.1730398e-01 1.3642851e+00 1.7579757e-01 1.2130852e+00 -1.6101295e-01 1.5526788e-01 5.1232427e-01 1.0103672e+00 -2.5942066e-01 2.3188990e-01 2.8556964e-01 3.9608824e-01 6.1162275e-01 5.0055265e-01 7.9285419e-01 3.4412774e-01 7.4243361e-01 4.5606431e-01 -1.6984522e-02 -3.8224614e-01 -5.5673307e-01 5.1832753e-01 7.1609813e-01 -5.4908089e-02 3.0876353e-01 -9.0313041e-01 9.1460723e-01 -1.8714331e+00 -1.2150562e+00 3.7484923e-01 2.0389767e+00 9.7501099e-01 1.3234642e-01 6.1027247e-01 8.7150115e-01 5.2803487e-01 4.3392080e-01 -4.0972814e-01 -2.7366352e-01 1.4458723e-01 7.1286291e-01 1.8528779e-01 3.1722164e-01 -1.5338334e+00 6.1330855e-01 6.7997885e+00 1.0363035e+00 -1.0394418e+00 3.5393575e-01 2.5073761e-01 -5.7380217e-01 1.8174147e-02 -1.0804134e-01 -4.6215025e-01 2.7357358e-01 1.3000575e+00 -3.6523128e-01 6.1441648e-01 8.8866836e-01 -8.5176937e-02 3.3084616e-01 -1.4044911e+00 1.5250502e+00 2.0660833e-01 -1.2757348e+00 -1.7931864e-01 -3.2594475e-01 3.8334259e-01 6.0904932e-01 2.5690174e-01 5.6324935e-01 4.3252456e-01 -1.0435684e+00 9.9039942e-01 3.4538278e-01 1.0305306e+00 -6.2256581e-01 4.2398927e-01 9.6943360e-03 -1.3344456e+00 -3.5362729e-01 -1.3558666e-01 -1.1626635e-01 2.7115250e-02 7.0009631e-01 -8.9554220e-01 4.9713358e-01 1.2339303e+00 9.8228562e-01 -7.3720407e-01 1.1693304e+00 6.1228029e-02 1.1807288e+00 -5.3151369e-01 3.7920320e-01 1.3815817e-01 3.6819175e-01 6.1683142e-01 1.5292672e+00 2.8765321e-01 -4.2056105e-01 1.7139544e-01 3.6817729e-01 -1.7131434e-01 4.3189716e-02 -6.9403750e-01 -3.2387733e-01 4.4882852e-01 1.0207745e+00 2.6350815e-02 -2.9811314e-01 -1.5951445e-02 5.5595750e-01 2.1631958e-01 -7.7295210e-03 -5.2369422e-01 -3.9330712e-01 1.0231614e+00 1.6027068e-01 2.5463885e-01 1.6813058e-01 7.7649005e-02 -7.6995927e-01 -1.5147196e-01 -9.8760337e-01 8.5110903e-01 -7.6570082e-01 -1.6031791e+00 3.4671265e-01 1.4227631e-02 -1.5717657e+00 -6.0991555e-01 -3.7031046e-01 -5.6565160e-01 2.2802234e-01 -1.4792044e+00 -1.1566261e+00 -1.6622046e-01 6.6791111e-01 4.7048125e-01 -6.8515849e-01 1.2116567e+00 5.1465487e-01 -3.6016408e-02 9.1664970e-01 8.0025487e-02 1.2777536e-01 1.1366379e+00 -1.2775350e+00 4.6618360e-01 1.2144039e-01 8.4274709e-01 2.4415842e-01 6.5446621e-01 -1.1924491e-02 -1.2976661e+00 -9.2718059e-01 3.9242417e-01 -3.7797415e-01 1.0928179e+00 -5.7270378e-01 -6.9811118e-01 5.8991694e-01 2.7400756e-01 3.3809417e-01 1.1632026e+00 6.8480295e-01 -1.0634217e+00 -6.5027553e-01 -9.4603771e-01 9.0368330e-02 1.1784128e+00 -1.1323246e+00 -8.1764680e-01 2.7468395e-01 7.1543068e-01 -1.6952444e-02 -7.9370409e-01 4.7024760e-01 1.0189375e+00 -9.4712412e-01 1.1750212e+00 -7.5826585e-01 2.8792882e-01 -1.5320742e-01 -5.6513250e-01 -1.4062177e+00 -2.4336335e-01 -9.5815188e-01 -1.1346362e-01 1.6024604e+00 3.5702184e-01 -3.9556402e-01 7.9320270e-01 -3.3581603e-01 -1.8434356e-01 -4.1133964e-01 -1.5043534e+00 -9.7820628e-01 -1.0489296e-01 -1.1177610e+00 3.5938534e-01 1.1225852e+00 1.2298443e-01 5.3863466e-01 -4.1682139e-01 -9.2497736e-02 1.0066731e+00 -2.3828937e-01 7.7578330e-01 -1.6209986e+00 -8.7456226e-01 -3.2530746e-01 -9.9511486e-01 -4.6951270e-01 3.8895223e-01 -1.1137714e+00 -9.9648252e-02 -1.0646311e+00 1.0715357e-02 -3.6897832e-01 -7.9358137e-01 7.9047734e-01 4.6341389e-01 6.7946428e-01 4.6077159e-01 -2.8559666e-02 -7.3222768e-01 6.8520474e-01 3.3963355e-01 -6.3215172e-01 7.1785250e-04 8.3461024e-02 -9.7346282e-01 7.4107456e-01 8.1001985e-01 -4.8678452e-01 -5.4394937e-01 -3.7590325e-01 -4.8736468e-02 -1.0205542e-01 5.5707133e-01 -1.6611713e+00 1.3209291e-01 4.6557099e-01 1.4470634e-01 -2.5076351e-01 8.4446222e-01 -4.4876924e-01 8.5016675e-02 2.1105283e-01 -8.9038414e-01 -2.8127965e-01 5.0499904e-01 6.3565058e-01 -4.1951203e-01 -1.4754362e-01 6.2065232e-01 5.6523986e-02 -6.4765418e-01 1.5905242e-01 -3.4361601e-01 2.8433663e-01 4.3804872e-01 -1.7087441e-02 -2.9409444e-01 -6.3782614e-01 -1.0257137e+00 6.2535793e-02 -9.8842859e-02 8.5718417e-01 4.3955106e-01 -1.4387705e+00 -9.9350148e-01 1.8095572e-01 5.8010638e-01 -2.6277772e-01 3.7889561e-01 4.9925587e-01 9.2583276e-02 2.4578881e-01 -2.5911972e-01 -7.5656193e-01 -1.4285421e+00 1.4452493e-01 2.0442921e-01 8.7623946e-02 -4.5988780e-01 1.0124987e+00 3.4257475e-02 -6.4937395e-01 5.7458961e-01 -8.8566490e-02 -5.5776159e-03 3.8207501e-01 7.7800417e-01 4.4411668e-01 2.7360210e-01 -2.4012256e-01 -2.8288069e-01 5.1197857e-01 -7.9499744e-02 -3.8944343e-01 1.5146496e+00 3.4862959e-01 3.6615682e-01 1.1172874e+00 1.1884420e+00 -7.8224212e-02 -1.0969107e+00 -3.5919812e-01 -5.5499811e-02 -3.4618807e-01 3.0138016e-01 -7.4420226e-01 -8.2470316e-01 1.0263516e+00 1.2634290e+00 1.3905318e-01 8.8245565e-01 2.0798060e-01 5.8851612e-01 5.2711397e-01 3.4804481e-01 -1.1983409e+00 2.5502852e-01 3.8615477e-01 1.0248114e+00 -1.1765951e+00 2.1910576e-02 -2.5084823e-01 -6.8745220e-01 9.3632126e-01 3.5412446e-01 -3.7309337e-01 8.6019385e-01 5.2728063e-01 2.5123698e-01 1.1325279e-01 -1.2019935e+00 -3.6339971e-01 2.9005137e-01 1.0902992e+00 7.0383894e-01 9.0331174e-02 4.5269480e-01 8.8218331e-01 -5.8793497e-01 1.6677478e-01 6.3235596e-02 7.1073443e-01 -4.3479821e-01 -9.1933739e-01 -2.6981130e-01 4.6692988e-01 -3.5445371e-01 -7.4834906e-02 -5.5171853e-01 3.8469896e-01 9.9510618e-02 1.0471441e+00 2.8114030e-01 -5.1474136e-01 4.4159994e-01 2.9595098e-01 5.4107541e-01 -6.9029617e-01 -6.4651835e-01 6.5751038e-02 5.5964673e-01 -5.3923208e-01 -4.5778412e-01 -7.6900393e-01 -1.1268452e+00 1.3086382e-01 6.7339174e-02 2.1455768e-01 4.5199537e-01 5.5092877e-01 3.2941583e-01 4.7181624e-01 5.6884325e-01 -1.1514640e+00 -7.4680424e-01 -1.1125162e+00 -7.0975506e-01 3.2894352e-01 6.8377471e-01 -7.1065652e-01 -4.1534522e-01 5.3851523e-02]

[15.270251274108887, 5.142383575439453]

116879c1-1704-4e1e-9d2c-9ac3939b08ac

twina-at-semeval-2017-task-4-twitter

https://aclanthology.org/S17-2109

https://aclanthology.org/S17-2109.pdf

TWINA at SemEval-2017 Task 4: Twitter Sentiment Analysis with Ensemble Gradient Boost Tree Classifier

This paper describes the TWINA system, with which we participated in SemEval-2017 Task 4B (Topic Based Message Polarity Classification {--} Two point scale) and 4D (two-point scale Tweet quantification). We implemented ensemble based Gradient Boost Trees classification method for both the tasks. Our system could perform well for the task 4D and ranked 13th among 15 teams, for the task 4B our model ranked 23rd position.

['Suresh Kumar Sanampudi', 'Naveen Kumar Laskari']

2017-08-01

semeval-2017-8

['twitter-sentiment-analysis']

['natural-language-processing']

[-1.84112042e-01 1.67462543e-01 -2.72682130e-01 -4.11252290e-01 -8.52974176e-01 -4.47982430e-01 1.24702799e+00 7.27419972e-01 -6.31848931e-01 8.97987783e-01 3.18840116e-01 -3.88025582e-01 7.00317547e-02 -6.44330740e-01 -1.96910694e-01 -2.82211602e-01 -4.15838927e-01 7.53316998e-01 3.73548478e-01 -8.82528841e-01 8.35903108e-01 -4.04129922e-01 -1.13009429e+00 8.84967387e-01 4.38405126e-01 1.48241079e+00 -5.92311144e-01 1.00332534e+00 -1.34621114e-01 8.78291905e-01 -7.81471908e-01 -9.54424262e-01 -1.82671677e-02 9.43103209e-02 -1.13743782e+00 -7.16337800e-01 6.39631808e-01 2.86989123e-01 2.88931698e-01 7.22483218e-01 2.20935866e-01 -1.13035366e-01 1.10115564e+00 -1.69680917e+00 -1.93924636e-01 9.04174566e-01 -6.99555635e-01 5.75049162e-01 4.52280939e-01 -5.77737629e-01 1.37387323e+00 -1.16412568e+00 6.20718539e-01 1.13251913e+00 8.98914099e-01 1.73416644e-01 -1.04980731e+00 -9.08533394e-01 -3.41146179e-02 4.73861784e-01 -7.22476721e-01 -9.23328940e-03 9.24480677e-01 -7.24430799e-01 1.18931758e+00 3.80196422e-01 6.30383670e-01 1.42377484e+00 5.01292527e-01 6.36545420e-01 2.15460038e+00 6.07402287e-02 7.90424049e-02 3.23560417e-01 9.82937813e-01 2.06129283e-01 8.93312618e-02 -4.16574508e-01 -9.49391603e-01 -5.07923245e-01 -4.14382517e-01 -5.19234896e-01 4.27878112e-01 4.07515675e-01 -1.39790440e+00 9.56738710e-01 2.66724944e-01 4.81289148e-01 -3.62416536e-01 -5.69101907e-02 8.48503590e-01 9.25013304e-01 1.27103412e+00 6.56666100e-01 -9.23449755e-01 -4.67327267e-01 -9.70921874e-01 7.42912889e-01 1.05210733e+00 4.19596165e-01 4.98430550e-01 -2.50906080e-01 -4.79428887e-01 1.00423634e+00 2.61362314e-01 3.66122276e-01 6.73749328e-01 -6.55123949e-01 6.58681810e-01 6.93510473e-01 1.49937794e-01 -1.38935983e+00 -6.01052284e-01 -2.93643683e-01 -6.26681387e-01 1.70845821e-01 1.39424220e-01 -2.93252915e-01 -7.89787889e-01 1.21890831e+00 7.40551651e-02 2.42736991e-02 -4.93232235e-02 5.82959831e-01 1.42539203e+00 8.24395061e-01 4.37586248e-01 -3.07325840e-01 1.65097952e+00 -7.28073180e-01 -5.94055831e-01 -1.62075058e-01 8.31514239e-01 -8.34998846e-01 1.08538854e+00 8.34896266e-01 -9.25394237e-01 -2.26125315e-01 -1.32501578e+00 -1.39327375e-02 -9.84212935e-01 -3.13753545e-01 4.53804880e-01 6.37076914e-01 -1.44844151e+00 5.46818197e-01 -4.87101115e-02 -1.32750839e-01 4.45809960e-01 3.60140502e-01 -4.78232861e-01 5.14548540e-01 -1.75164962e+00 1.11671972e+00 9.35702398e-02 -3.65265042e-01 -6.82591558e-01 -6.36194527e-01 -5.05750239e-01 -2.51111746e-01 -3.08777720e-01 -2.76694536e-01 1.47769845e+00 -3.12781036e-01 -1.63657820e+00 1.52905893e+00 -1.39700249e-01 -7.26748347e-01 4.89957035e-01 -1.24928802e-01 -4.67653394e-01 -5.88180244e-01 4.12650585e-01 5.10930240e-01 7.24468410e-01 -7.63066709e-01 -1.00892735e+00 -4.01983619e-01 -9.96436030e-02 2.29821637e-01 -4.19979244e-01 2.11365536e-01 5.75292408e-01 -4.38972324e-01 -8.49652886e-02 -8.82162869e-01 5.73621094e-02 -1.14185572e+00 -3.67451012e-01 -9.33869541e-01 1.20220530e+00 -6.53811455e-01 1.22035408e+00 -1.50284159e+00 5.16768210e-02 1.82100639e-01 4.92853373e-01 -2.54398704e-01 1.97963759e-01 4.01979566e-01 -1.37839168e-01 4.36306089e-01 2.95921974e-02 -8.01780701e-01 4.05804999e-02 -3.05065930e-01 -4.05963689e-01 3.76502842e-01 -5.16358390e-02 6.21399224e-01 -8.98315251e-01 -6.86111450e-01 -3.53062153e-01 -7.99850374e-02 -1.39376715e-01 -1.95530161e-01 -3.27240005e-02 -6.72717616e-02 -3.68524909e-01 6.96101487e-01 5.70914507e-01 -2.32672110e-01 6.68701380e-02 -1.81616321e-02 -3.58753443e-01 7.97319770e-01 -3.73865336e-01 9.43475008e-01 -4.22249258e-01 1.18511796e+00 6.24986514e-02 -1.00945377e+00 1.02534628e+00 5.07275343e-01 3.04416239e-01 -8.17400992e-01 4.82750237e-01 3.56680602e-01 9.23945010e-02 -1.42257646e-01 8.04515243e-01 -2.72706956e-01 -9.10375416e-01 3.44319850e-01 -6.48334548e-02 -4.36228484e-01 2.29581445e-01 6.90193176e-01 1.15862560e+00 -5.20899773e-01 2.57853806e-01 -7.76145935e-01 5.97426653e-01 2.24366456e-01 1.65328026e-01 5.75952411e-01 -5.77343166e-01 2.69418955e-01 1.11806858e+00 -6.23293102e-01 -1.05578756e+00 -3.80974829e-01 -3.02853525e-01 1.83324075e+00 6.10441789e-02 -7.33713031e-01 -2.70127714e-01 -1.11445844e+00 1.32849291e-02 8.51032674e-01 -1.17589200e+00 4.42848206e-01 -3.81101936e-01 -1.06324482e+00 4.66443270e-01 -1.01411971e-03 3.76939446e-01 -1.02404439e+00 7.73690874e-03 1.11917071e-01 -5.77637315e-01 -9.02071118e-01 7.57601634e-02 3.46415430e-01 -6.84829891e-01 -8.90890241e-01 -5.61237395e-01 -8.10830474e-01 -1.81256961e-02 -1.52330697e-01 1.41598344e+00 -3.55857760e-01 3.48166525e-01 -6.99822828e-02 -3.87781620e-01 -9.06136751e-01 -2.31858864e-01 3.86503905e-01 5.47515787e-02 -8.14163238e-02 4.92513210e-01 -6.03945196e-01 -4.66730088e-01 2.12678120e-01 1.21164910e-01 8.55362322e-03 1.55507132e-01 5.77384651e-01 5.13204001e-02 -5.48205733e-01 1.03207731e+00 -1.37546885e+00 1.52786195e+00 -8.50057900e-01 -7.07056969e-02 -3.19863290e-01 -9.80515182e-01 -4.96347457e-01 3.26826066e-01 -5.45355491e-03 -8.38741064e-01 -7.36067116e-01 -2.29195744e-01 4.69830215e-01 2.23490894e-01 6.26788616e-01 6.43388867e-01 2.13201925e-01 9.90672171e-01 -2.83987343e-01 -4.63734210e-01 -2.12868020e-01 -1.35696391e-02 1.14832222e+00 -5.01349010e-03 -4.20796245e-01 3.07136953e-01 3.24160099e-01 3.23932320e-02 -5.01811802e-01 -1.06960285e+00 -6.51810884e-01 -1.69546619e-01 -5.49591601e-01 9.50054467e-01 -1.15106905e+00 -8.64450157e-01 7.44274378e-01 -1.21695900e+00 -2.03659162e-01 4.63342816e-01 9.52860638e-02 -2.72213846e-01 -9.32931378e-02 -8.13880622e-01 -8.59810054e-01 -1.04120278e+00 -6.86043680e-01 1.11928749e+00 -2.15749159e-01 -8.67350340e-01 -7.78101146e-01 7.38524556e-01 4.83881295e-01 6.25493288e-01 4.91518110e-01 8.27364683e-01 -1.09429395e+00 6.23091877e-01 -4.22890693e-01 -1.99106738e-01 1.60141766e-01 -4.74948645e-01 -6.88047931e-02 -1.09583187e+00 1.23892389e-01 -8.01861435e-02 -8.24959934e-01 1.05812371e+00 1.85777277e-01 1.01609278e+00 -3.61797124e-01 -5.31899571e-01 -3.34187984e-01 6.27543986e-01 -2.48333544e-01 4.50661182e-01 6.44250154e-01 3.64634097e-01 6.83017790e-01 6.62163913e-01 2.58183569e-01 1.02977514e+00 7.89812922e-01 3.44870597e-01 3.21794838e-01 4.73199300e-02 3.58455605e-03 5.14761150e-01 1.35667050e+00 -1.38860732e-01 -2.47229755e-01 -1.15987563e+00 6.35939538e-01 -1.92882192e+00 -8.27496290e-01 -1.16110098e+00 1.49400163e+00 9.14455414e-01 6.71672761e-01 4.38784242e-01 5.65072715e-01 5.01760125e-01 3.33383083e-01 1.96225241e-01 -1.24794579e+00 -1.58656791e-01 -8.02436322e-02 1.96840733e-01 5.93930662e-01 -1.44629729e+00 7.15959966e-01 6.91750717e+00 9.95400190e-01 -1.16106093e+00 6.74002647e-01 1.01241648e+00 4.19762880e-02 -6.68065175e-02 -3.24730463e-02 -7.17954278e-01 7.66700506e-01 1.34254134e+00 -1.27784684e-01 -2.35686094e-01 1.03730536e+00 -1.94009408e-01 -6.53738156e-02 -9.43635404e-01 7.73303926e-01 -3.19203734e-02 -1.14078653e+00 -2.93683559e-01 9.10690501e-02 1.11267507e+00 6.37036502e-01 1.07724294e-01 1.08035970e+00 2.51306862e-01 -8.18239272e-01 7.94727802e-01 1.66268870e-01 7.68058300e-01 -3.72226626e-01 1.03606439e+00 4.69959676e-01 -6.81621850e-01 -4.29895297e-02 5.25321141e-02 -5.26280999e-01 -4.69580516e-02 9.23898101e-01 -1.12795615e+00 1.40426174e-01 1.09501994e+00 8.58180106e-01 -5.48877239e-01 5.59135139e-01 -8.59160796e-02 1.28133023e+00 -3.88898402e-01 -8.66247773e-01 2.45805621e-01 1.30955115e-01 9.25265372e-01 1.81956351e+00 2.26960406e-01 -1.23280853e-01 -1.45222902e-01 -1.94347829e-01 -7.62505412e-01 2.35380948e-01 -3.57229471e-01 2.33493239e-01 1.76663771e-01 1.84778285e+00 -7.26691484e-01 -7.36808062e-01 6.82564676e-02 4.62808490e-01 5.11858165e-01 -1.51809737e-01 -6.72863185e-01 -5.00010490e-01 1.05989859e-01 2.96177328e-01 -2.43199736e-01 1.13314889e-01 -1.04085910e+00 -9.87794518e-01 -4.28017706e-01 -7.14563906e-01 4.62835550e-01 -5.61000407e-01 -1.73605883e+00 7.33850598e-01 1.48357525e-01 -8.99176240e-01 -4.35623750e-02 -7.64420390e-01 -1.08096933e+00 5.18663585e-01 -1.38838720e+00 -1.12336195e+00 -1.63706213e-01 3.80603194e-01 4.88000721e-01 -2.69782215e-01 7.95263648e-01 4.57067728e-01 -5.18044718e-02 3.06042641e-01 -1.69298694e-01 -3.29694867e-01 1.02550924e+00 -1.81953788e+00 3.28837126e-01 -2.57170826e-01 -1.63404033e-01 2.24628851e-01 9.75372553e-01 -6.19328439e-01 -5.88057995e-01 -8.15325320e-01 1.67995489e+00 -9.83465493e-01 1.06466794e+00 -6.83227837e-01 -4.49518412e-01 2.92180210e-01 5.57412565e-01 -6.61985099e-01 7.46694982e-01 7.95040846e-01 -4.69608754e-01 -2.41989344e-01 -1.18618774e+00 3.13820899e-01 4.50339139e-01 -3.88789535e-01 -7.29893327e-01 7.36599028e-01 3.75266999e-01 -3.06007355e-01 -9.43310678e-01 6.13605201e-01 7.38135159e-01 -8.03491771e-01 3.37775856e-01 -4.81609434e-01 9.90680933e-01 -5.77107072e-02 -8.46145675e-02 -1.48112261e+00 -2.89704502e-01 -3.37261260e-01 -2.72565156e-01 9.55287874e-01 1.21033823e+00 -5.73956966e-01 6.71865165e-01 1.96638212e-01 -8.84486064e-02 -9.48078096e-01 -1.17963755e+00 -1.10797718e-01 4.55559224e-01 -7.01462984e-01 8.16793069e-02 1.18450868e+00 7.35854149e-01 1.21735394e+00 -4.57552731e-01 -6.51296437e-01 4.05004531e-01 -2.34973547e-03 6.21536255e-01 -1.65712750e+00 2.54318267e-01 -8.98104906e-01 -4.41700637e-01 -4.92636740e-01 5.26679270e-02 -9.33996260e-01 -1.93885624e-01 -1.50148594e+00 4.74621177e-01 -1.56919077e-01 -5.60994267e-01 3.38599682e-01 -2.60027945e-01 6.22487485e-01 3.13876718e-02 2.40529895e-01 -1.10250354e+00 2.49060482e-01 8.13875556e-01 -4.02027160e-01 3.09087634e-01 9.52899009e-02 -9.48081970e-01 7.05181777e-01 9.34748292e-01 -7.73139536e-01 -9.59656946e-03 1.06703006e-01 9.75492775e-01 -2.92255849e-01 1.12822898e-01 -6.40353560e-01 2.18714923e-01 2.50277787e-01 3.55237097e-01 -9.04685676e-01 5.41972876e-01 -3.09684962e-01 -4.30337131e-01 2.61174887e-01 -5.57363868e-01 4.86797914e-02 6.92951903e-02 4.97435957e-01 -3.42293084e-01 4.03068475e-02 5.67279816e-01 2.04669952e-01 -1.92839399e-01 2.42006523e-03 -5.95260561e-01 1.76396757e-01 9.07514393e-01 1.48383081e-01 -9.12021458e-01 -5.95175147e-01 -7.38978386e-01 5.02140105e-01 -3.02479595e-01 5.74694633e-01 3.08319390e-01 -1.25147617e+00 -1.30051363e+00 -6.02373719e-01 1.93283856e-01 -6.55268073e-01 7.17087016e-02 1.07300246e+00 -2.15286374e-01 4.58767593e-01 -1.78257316e-01 -4.45226789e-01 -1.34371996e+00 -2.23172531e-01 2.30780710e-02 -9.51501429e-01 1.69399247e-01 1.21163774e+00 -1.89125627e-01 -6.36194289e-01 -1.50428608e-01 -5.11648394e-02 -1.03644478e+00 6.25529945e-01 4.83533740e-01 8.73126924e-01 4.48309839e-01 -6.39576435e-01 -5.85684657e-01 4.01028991e-01 -3.75616759e-01 -4.48840111e-01 1.40934205e+00 1.24366835e-01 -4.76757467e-01 8.51527989e-01 1.44507420e+00 -7.38174235e-03 -1.79087207e-01 1.46676078e-01 3.03642660e-01 8.99346545e-02 4.70209152e-01 -1.21436954e+00 -1.16914183e-01 8.65400672e-01 3.77797872e-01 1.20904970e+00 4.80978131e-01 -1.24937862e-01 7.50974119e-01 4.58852530e-01 2.81323224e-01 -1.32559013e+00 1.09482855e-01 1.09332275e+00 1.25480771e+00 -1.68319309e+00 1.40365496e-01 -5.40630259e-02 -9.43002582e-01 7.77844667e-01 6.08210325e-01 -2.66631156e-01 1.05954468e+00 1.07598836e-02 5.73962927e-03 -7.44551003e-01 -1.32926440e+00 4.69550997e-01 6.22341096e-01 3.13485891e-01 8.67391288e-01 4.21570480e-01 -1.12942326e+00 8.76614034e-01 -6.49250090e-01 -2.50292629e-01 4.18508232e-01 5.50830245e-01 -5.16320169e-01 -8.51105392e-01 -1.14652878e-02 9.18506026e-01 -1.11108863e+00 -1.25696003e-01 -9.90968585e-01 5.83043039e-01 1.93672985e-01 1.33992422e+00 -1.63047627e-01 -9.97282803e-01 2.90036291e-01 1.35233417e-01 -2.85955966e-01 -4.18228298e-01 -1.20018017e+00 -2.90968060e-01 9.83781397e-01 -2.37590939e-01 -4.64682937e-01 -6.63248599e-01 -8.29010189e-01 -5.33697605e-01 -2.25520849e-01 5.57982683e-01 1.14498198e+00 6.60866857e-01 2.35125601e-01 2.64987797e-01 9.30753350e-01 -8.92311692e-01 -2.88269043e-01 -1.71222699e+00 -5.01293659e-01 3.38351607e-01 2.94858575e-01 -6.40424013e-01 -6.53459609e-01 -6.55068904e-02]

[11.162430763244629, 7.042892932891846]

8bb749e9-d1cc-4735-987b-539a417e9a14

on-the-relationships-between-graph-neural

2304.00146

https://arxiv.org/abs/2304.00146v1

https://arxiv.org/pdf/2304.00146v1.pdf

On the Relationships between Graph Neural Networks for the Simulation of Physical Systems and Classical Numerical Methods

Recent developments in Machine Learning approaches for modelling physical systems have begun to mirror the past development of numerical methods in the computational sciences. In this survey, we begin by providing an example of this with the parallels between the development trajectories of graph neural network acceleration for physical simulations and particle-based approaches. We then give an overview of simulation approaches, which have not yet found their way into state-of-the-art Machine Learning methods and hold the potential to make Machine Learning approaches more accurate and more efficient. We conclude by presenting an outlook on the potential of these approaches for making Machine Learning models for science more efficient.

['Nikolaus A. Adams', 'Andrea Panizza', 'Ludger Paehler', 'Artur P. Toshev']

2023-03-31

['physical-simulations']

['miscellaneous']

[-1.14860147e-01 -2.28815660e-01 -3.53215188e-01 2.11035274e-02 -1.19637810e-02 9.02133062e-02 1.20603681e+00 3.99793744e-01 -3.51736277e-01 9.35536385e-01 -2.65722573e-01 -1.01345587e+00 -1.57089084e-01 -1.23825049e+00 -4.20677006e-01 -9.61201549e-01 -6.67131841e-01 7.62189984e-01 2.37987265e-01 -5.26150882e-01 3.80590737e-01 1.01041782e+00 -1.46283686e+00 1.81393862e-01 3.33983332e-01 6.14868104e-01 -1.91629514e-01 1.05678809e+00 -1.15214651e-02 8.36786866e-01 -2.77337521e-01 1.99983761e-01 -8.57730433e-02 -5.54520488e-01 -7.04834461e-01 -2.77060360e-01 -1.78282514e-01 4.61474478e-01 -9.89812493e-01 6.95551634e-01 3.85076612e-01 5.46429276e-01 1.00935638e+00 -1.26854980e+00 -4.27603692e-01 2.12954566e-01 -2.23829776e-01 3.99662256e-01 2.06688091e-01 5.25336325e-01 5.80539644e-01 -5.22925198e-01 4.44989622e-01 1.30780447e+00 1.00264692e+00 4.83195394e-01 -9.63279247e-01 -3.41030926e-01 -1.37763634e-01 3.36675316e-01 -9.83137369e-01 -1.44469664e-01 7.64422894e-01 -4.19510752e-01 1.75174296e+00 2.92111397e-01 1.25902808e+00 5.00529587e-01 9.42380130e-01 3.27322930e-01 1.34632719e+00 -1.03745568e+00 6.78592443e-01 4.96408232e-02 2.30142370e-01 7.49882221e-01 4.91485834e-01 8.68263304e-01 -1.50464237e-01 -4.41617429e-01 8.53343010e-01 -3.49926323e-01 3.02502692e-01 -4.63306695e-01 -9.73599911e-01 1.34705007e+00 3.21698546e-01 3.81782502e-01 -3.66364628e-01 4.76285875e-01 5.51294208e-01 5.61647639e-02 4.57970589e-01 7.78204560e-01 -5.68122089e-01 -1.28065199e-01 -1.00537264e+00 8.70890021e-01 1.21122277e+00 3.45508277e-01 2.83266604e-01 7.09871709e-01 4.49887335e-01 2.93730140e-01 4.52691585e-01 4.12551284e-01 1.65192023e-01 -1.05581033e+00 -2.87839055e-01 2.38368183e-01 2.12680757e-01 -6.17313266e-01 -7.75880873e-01 -5.36040403e-02 -1.11154687e+00 9.95239139e-01 3.17022175e-01 -4.71402496e-01 -5.39013445e-01 8.18767846e-01 4.11959052e-01 3.35785717e-01 2.65318304e-01 3.44777822e-01 9.04570460e-01 1.12406278e+00 1.20822147e-01 -3.79832834e-01 8.70845675e-01 -1.14813197e+00 -5.24937987e-01 2.85315782e-01 9.92370367e-01 -6.75460994e-01 4.91949171e-01 3.73156667e-01 -1.34728217e+00 -6.71681821e-01 -1.10290742e+00 3.59200031e-01 -8.03623319e-01 -5.88351965e-01 1.38581276e+00 7.69064069e-01 -1.06288266e+00 1.46485937e+00 -1.39444315e+00 -3.85797083e-01 7.93614611e-02 5.40599525e-01 3.62219006e-01 5.20259917e-01 -1.29899335e+00 1.40425301e+00 1.45344079e-01 -3.40729237e-01 -4.88744825e-01 -8.70105982e-01 -5.49753249e-01 -3.29401970e-01 -3.89567427e-02 -9.42809939e-01 1.50219643e+00 -4.14674461e-01 -1.86929393e+00 4.03588861e-01 -3.43570858e-02 -6.96667850e-01 1.63560420e-01 3.32636118e-01 -6.63758636e-01 -9.86653566e-02 -7.35457063e-01 3.80702585e-01 1.66240737e-01 -1.26760352e+00 -5.96165061e-01 2.16313556e-01 -1.87067330e-01 -3.98740843e-02 1.55758247e-01 2.10595459e-01 2.50991106e-01 -2.29087055e-01 -3.99022758e-01 -1.08842945e+00 -1.10748363e+00 -3.23027037e-02 7.25299194e-02 -6.85924292e-01 7.00864792e-01 -1.69252872e-01 1.01737285e+00 -1.13573098e+00 1.64705142e-01 1.34978116e-01 3.16770136e-01 7.14653730e-01 2.96143264e-01 1.31243968e+00 -1.85867503e-01 -3.65728214e-02 -3.68536562e-02 -2.81232595e-01 -2.27146238e-01 3.94009829e-01 -1.16058350e-01 6.04248643e-01 1.97622459e-02 1.10346639e+00 -1.10949075e+00 -2.81132728e-01 8.83155942e-01 6.64290428e-01 -1.16558343e-01 -6.49299994e-02 -3.01518798e-01 5.91485083e-01 -5.80636442e-01 9.64148939e-02 4.02371168e-01 -3.98690850e-01 1.20640256e-01 1.73849702e-01 -5.99642038e-01 5.72586834e-01 -9.86223936e-01 7.12562144e-01 -4.27129149e-01 5.17855465e-01 -4.20455914e-03 -1.68971872e+00 5.84671795e-01 1.97346866e-01 6.00196064e-01 -6.41346335e-01 2.02423424e-01 1.31013379e-01 3.35009217e-01 -3.52388084e-01 3.09395820e-01 -3.24126124e-01 2.23068297e-01 3.96629661e-01 -2.41934419e-01 -7.87305892e-01 2.46726349e-02 3.91383003e-03 8.40953887e-01 8.54552090e-02 8.28095078e-01 -6.60759449e-01 6.99374974e-01 4.45202142e-01 -2.24997684e-01 1.16830492e+00 -7.30933249e-02 -1.08962201e-01 -4.29468043e-02 -1.10312891e+00 -1.67470264e+00 -5.67306042e-01 -1.07656084e-01 7.52291024e-01 -2.20367029e-01 -4.10731286e-01 -8.15635026e-01 -1.45408884e-01 3.22246045e-01 6.72226906e-01 -5.27963459e-01 2.96865683e-02 -8.86701286e-01 -1.14165509e+00 3.94668698e-01 5.46690166e-01 4.27319258e-02 -1.41018438e+00 -4.53100324e-01 2.82299072e-01 8.63038301e-01 -7.79391587e-01 6.90954268e-01 5.58081381e-02 -1.33204186e+00 -8.68294656e-01 -3.35263461e-01 -7.16241300e-01 1.59674749e-01 1.98357791e-01 1.30370724e+00 6.20825708e-01 -4.38123643e-01 2.04206079e-01 -4.78535183e-02 -4.34247851e-01 -1.07156003e+00 8.02458823e-02 5.44728339e-01 -1.01462495e+00 1.82549566e-01 -8.64041984e-01 -5.17202258e-01 -2.89956152e-01 -6.52290046e-01 2.94404268e-01 -3.78381386e-02 5.88889122e-01 1.75237879e-01 1.94413990e-01 5.31361699e-01 -7.12246120e-01 6.55117095e-01 -5.85428476e-01 -7.87869036e-01 -2.41511278e-02 -1.11204338e+00 8.32472071e-02 9.87959504e-01 -2.89024353e-01 -7.39239872e-01 -6.25839755e-02 -4.24151599e-01 4.07546647e-02 -1.48471445e-01 5.31840086e-01 6.76058054e-01 -9.11476851e-01 6.79111779e-01 2.99000502e-01 2.15556115e-01 -2.93854505e-01 4.45365161e-01 5.11576772e-01 1.00549631e-01 -4.60873991e-01 6.44219756e-01 4.53057498e-01 8.68959129e-01 -1.39073408e+00 -2.60575861e-01 -4.05630916e-01 -5.93432784e-01 -3.38405818e-01 6.30749166e-01 -3.72183710e-01 -9.22489107e-01 7.82077193e-01 -9.58725512e-01 -7.79407203e-01 -4.35363472e-01 6.89851642e-01 -7.74779081e-01 4.77281243e-01 -8.92550468e-01 -1.13627398e+00 -2.02052400e-01 -1.00869513e+00 5.00786364e-01 4.59899873e-01 -2.16345891e-01 -2.04022050e+00 7.60266185e-01 -1.40064925e-01 6.88440919e-01 2.22672179e-01 1.06239772e+00 -5.67794561e-01 -9.50783491e-02 -3.71479481e-01 1.37580395e-01 -1.46843031e-01 -1.97285160e-01 5.52894413e-01 -6.39029086e-01 -4.53636020e-01 5.45327589e-02 -1.00107938e-01 8.41716051e-01 6.50616348e-01 9.74360049e-01 -2.65657008e-01 -1.03811371e+00 4.69379932e-01 1.52019262e+00 4.30928349e-01 2.74901122e-01 2.33361661e-01 6.67651832e-01 3.29314440e-01 5.26075624e-02 2.94159204e-01 1.13479488e-01 5.18354177e-01 9.34664607e-02 -4.35199589e-01 -2.40652665e-01 7.61102289e-02 -9.37023982e-02 1.38103151e+00 -5.22477269e-01 -2.79600501e-01 -1.23438549e+00 2.28088498e-01 -1.99893010e+00 -9.28407848e-01 -6.53423727e-01 1.93811059e+00 5.66599965e-01 2.27171808e-01 3.51097345e-01 2.13188350e-01 6.62676990e-01 4.89913709e-02 -4.47352439e-01 -1.02023530e+00 3.39806497e-01 6.37403548e-01 6.06971502e-01 8.40094745e-01 -9.31510627e-01 1.10823607e+00 8.54746246e+00 9.22096670e-01 -1.24325514e+00 1.75427660e-01 4.66609746e-01 1.58250421e-01 8.04392844e-02 1.92375183e-01 -9.38796639e-01 2.96644032e-01 1.75176859e+00 -1.90617457e-01 7.37704337e-01 7.57286489e-01 6.30790293e-01 -3.03760618e-01 -8.64769816e-01 6.00543261e-01 -3.36566240e-01 -2.02994347e+00 -4.25546430e-02 1.55650824e-02 9.62935805e-01 5.60962141e-01 -3.18501025e-01 4.82754678e-01 7.86661386e-01 -1.22801864e+00 1.66514084e-01 5.34977317e-01 3.00195098e-01 -6.25311971e-01 4.82610166e-01 5.49185157e-01 -1.12660670e+00 3.12661111e-01 -7.08505690e-01 -1.23842764e+00 4.57513779e-01 6.27055168e-01 -5.39705336e-01 5.03236353e-01 4.14369762e-01 6.29425347e-01 -2.64770687e-01 1.34925568e+00 4.12192404e-01 8.34326804e-01 -4.22937661e-01 -1.01696026e+00 6.07031226e-01 -2.99241573e-01 5.67478299e-01 1.44225240e+00 -1.55971885e-01 2.49958828e-01 4.54442471e-01 4.83107775e-01 4.61274177e-01 -1.56793326e-01 -7.58732319e-01 -1.46887511e-01 2.27480099e-01 1.19692564e+00 -1.14277840e+00 -6.48657084e-01 -6.49369001e-01 6.83157369e-02 2.76244104e-01 1.46526292e-01 -8.47018063e-01 -8.87160227e-02 5.64807713e-01 2.08910838e-01 2.32057646e-01 -6.88767552e-01 -5.06190121e-01 -8.36921394e-01 -8.80526364e-01 -7.12626517e-01 -5.48821455e-03 -6.92539454e-01 -1.47801542e+00 2.93020874e-01 2.90477306e-01 -4.74149048e-01 -3.84892851e-01 -1.19707692e+00 -9.76859868e-01 6.39624774e-01 -1.23960030e+00 -8.55543017e-01 1.95959732e-01 7.81330839e-03 3.44190568e-01 -4.30568129e-01 1.17996144e+00 -2.08876982e-01 -3.53578120e-01 -8.80657509e-02 9.36353028e-01 -2.65028149e-01 1.30901542e-02 -1.09127593e+00 1.16341186e+00 2.02465862e-01 2.28030626e-02 2.30666712e-01 1.24570918e+00 -8.78006101e-01 -1.56549048e+00 -7.70438373e-01 6.81873679e-01 -4.52798784e-01 1.08316875e+00 -1.92114741e-01 -9.21180844e-01 3.40701401e-01 4.30781662e-01 1.01509549e-01 2.18299747e-01 1.75996542e-01 2.55909532e-01 3.34088743e-01 -7.33075917e-01 6.70205951e-01 8.22398186e-01 -3.03538471e-01 -4.78559375e-01 9.35580134e-01 2.25332364e-01 -3.01776856e-01 -8.35153699e-01 4.05333728e-01 5.26539922e-01 -7.44331539e-01 1.10832107e+00 -1.21648908e+00 3.41552764e-01 2.97972374e-02 4.49197471e-01 -1.29650509e+00 -5.56245744e-01 -9.62157547e-01 -5.99565208e-01 4.81722653e-01 4.11725283e-01 -8.15560818e-01 9.78375018e-01 2.65794873e-01 -1.39824986e-01 -1.32924342e+00 -1.18418992e+00 -7.75661528e-01 1.18674958e+00 -4.72024530e-01 3.62237126e-01 9.22220707e-01 5.35560310e-01 2.08799407e-01 -6.35350645e-02 -2.03058541e-01 4.82410669e-01 8.39438736e-02 4.52453941e-01 -1.23728573e+00 -2.98502058e-01 -9.20724094e-01 -5.62940180e-01 -5.65104306e-01 2.73751765e-01 -6.92605853e-01 -2.26632461e-01 -1.67393088e+00 1.08853579e-01 -4.40186769e-01 -9.98311490e-02 7.03134388e-02 1.16928734e-01 2.48765394e-01 -2.25030273e-01 2.69372076e-01 -5.52602053e-01 4.01455432e-01 1.22958577e+00 2.07074702e-01 -2.75139540e-01 2.12103337e-01 -8.46281126e-02 9.93181646e-01 1.19544017e+00 -3.84887457e-01 -1.25663787e-01 2.29524896e-01 4.13596123e-01 2.58411437e-01 5.51379263e-01 -1.45997608e+00 5.40451527e-01 -4.01861906e-01 4.89279777e-01 -3.95919114e-01 2.33605891e-01 -2.45920956e-01 -1.09327540e-01 9.54222679e-01 -2.14864805e-01 5.22910178e-01 5.67758083e-01 2.62692809e-01 2.72406518e-01 -2.70238936e-01 1.12946260e+00 -5.40592730e-01 -3.76681417e-01 1.31644517e-01 -1.18523860e+00 -6.50769472e-02 1.00131917e+00 -3.56716700e-02 -1.99249238e-01 -4.82673138e-01 -7.95304179e-01 -4.45734113e-02 5.63019633e-01 9.53359753e-02 1.50017828e-01 -8.88525128e-01 -5.86277962e-01 -2.19098464e-01 -5.28120935e-01 -7.38723457e-01 4.10302319e-02 5.22723138e-01 -9.26675439e-01 9.35886562e-01 -2.68717378e-01 -3.73884469e-01 -1.04492521e+00 1.08087838e+00 7.64564514e-01 -5.95189333e-01 -6.56205297e-01 3.45938534e-01 -1.59729689e-01 -5.60713708e-01 -2.94309258e-01 -2.37619117e-01 -2.88476139e-01 -8.60442400e-01 3.64057243e-01 8.72109294e-01 1.31931126e-01 -4.51435924e-01 -2.30309844e-01 5.08942127e-01 2.04833299e-01 -1.09498091e-01 1.51775789e+00 3.46539676e-01 -1.53667033e-01 7.51685798e-01 7.99376965e-01 -2.33830869e-01 -9.88361776e-01 7.76552558e-02 -2.18320623e-01 4.13848966e-01 5.15589595e-01 -5.68001628e-01 -8.20200205e-01 1.06912768e+00 5.00625074e-01 7.36005127e-01 3.23401242e-01 -6.94618821e-02 8.70673478e-01 6.34081900e-01 3.19318444e-01 -1.02334332e+00 -3.26813072e-01 7.79983580e-01 4.41391557e-01 -9.99398947e-01 6.01656199e-01 -5.05288363e-01 7.43762180e-02 1.48002732e+00 3.51976693e-01 -7.71570623e-01 1.28203046e+00 6.62464619e-01 -3.85651916e-01 -3.79588276e-01 -9.48214114e-01 6.29521385e-02 8.63413513e-02 8.20314348e-01 6.18016660e-01 2.11108476e-01 -3.21054846e-01 7.16041075e-03 -2.86293000e-01 9.55804735e-02 4.80915666e-01 1.07600570e+00 -7.01095760e-01 -1.49397457e+00 -2.32511759e-01 3.71800423e-01 -1.97488561e-01 -2.03767091e-01 -2.33374968e-01 1.20865524e+00 -7.60146081e-02 8.59601676e-01 1.06382556e-01 -5.28846197e-02 7.16992170e-02 1.17334537e-01 8.87240171e-01 -4.41578031e-01 -8.21317017e-01 -5.04094124e-01 2.15620801e-01 -1.86890781e-01 -4.46131825e-01 -7.43891239e-01 -1.30893219e+00 -1.08487439e+00 -4.05330569e-01 5.78918695e-01 5.97567976e-01 1.23825681e+00 2.04892918e-01 6.94818258e-01 2.73339003e-01 -1.61234951e+00 -4.05461311e-01 -7.63929248e-01 -4.18841898e-01 -4.27144885e-01 1.74622983e-01 -8.19722712e-01 -4.22023505e-01 -1.48917645e-01]

[6.363089084625244, 3.603257417678833]

75733754-f1ca-4b98-939d-19fb2c9961c0

convolutional-neural-networks-for-detecting

https://spie.org/Publications/Proceedings/Paper/10.1117/12.2571111

https://www.spiedigitallibrary.org/conference-proceedings-of-spie/11524/115240K/Convolutional-neural-networks-for-detecting-challenging-cases-in-cloud-masking/10.1117/12.2571111.full?SSO=1

Convolutional neural networks for detecting challenging cases in cloud masking using Sentinel-2 imagery

Cloud contamination represents a large obstacle for mapping the earth’s surface using remotely sensed data. Therefore, cloudy pixels should be identified and eliminated before any further data processing can be achieved. Although several threshold, multi-temporal and machine learning applications have been developed to tackle this issue, it still remains a challenge. The main challenges are imposed by the difficulty to detect thin clouds and to separate bright clouds from bright non-cloud objects. Convolutional neural networks (CNNs) have proven to be one of the most promising methods for image classification tasks and their use is rapidly increasing in remote sensing problems. CNNs present interesting properties for image processing since they directly exploit not only the spectral information but also the spatial covariance of the data. In this work, we study the applicability of CNNs in cloud detection of Sentinel-2 imagery, a complex remote sensing problem with crucial spatial context. A patch-to-pixel CNN architecture consisting of three convolutional layers and two fully connected layers is trained on a recently available manually created public dataset. The results were evaluated both qualitatively and quantitatively through comparison with ground truth cloud masks and state-of-the art pixel-based algorithms (Fmask, Sen2Cor). It was shown that the proposed architecture even though simpler than the deep learning architectures proposed by recent literature, performs very favorably, especially in the challenging cases. Besides the evaluation of the results, feature maps where observed as an initial effort to extract the weights of the useful kernels for cloud masking applications.

['Vassilia Karathanassi', 'Viktoria Kristollari']

2020-08-26

['cloud-detection']

['computer-vision']

[ 4.29148734e-01 -5.14386415e-01 4.06837285e-01 -3.70852560e-01 -6.70191348e-01 -6.72375500e-01 6.36815786e-01 1.90396413e-01 -6.21362388e-01 6.51280761e-01 -4.67345685e-01 -4.76255298e-01 -2.09241614e-01 -9.14521575e-01 -4.34988767e-01 -1.15822172e+00 -3.99951220e-01 1.08684830e-01 1.55189648e-01 -2.29788259e-01 4.41893972e-02 9.52498138e-01 -1.94196236e+00 3.28466624e-01 1.08129930e+00 1.12598002e+00 4.58645105e-01 4.63310421e-01 -6.06526770e-02 4.27919865e-01 -3.30268234e-01 1.06521584e-01 5.78025401e-01 -2.59155482e-01 -3.68531644e-01 7.94383734e-02 8.28951716e-01 -2.14851014e-02 3.60939652e-01 1.33512998e+00 4.71769303e-01 8.86970665e-03 5.32855272e-01 -6.36235237e-01 -1.39120549e-01 -1.64761115e-03 -4.49612558e-01 4.57211912e-01 -4.63828593e-01 2.67688513e-01 6.91747069e-01 -1.12444115e+00 1.86936766e-01 6.38526380e-01 6.68780565e-01 3.60322371e-02 -1.00099909e+00 -7.85361111e-01 5.84184639e-02 8.83258432e-02 -1.40812182e+00 -1.80783436e-01 4.93249536e-01 -8.83192778e-01 8.85209024e-01 4.84537989e-01 7.32132375e-01 3.27722728e-01 -5.98083176e-02 2.52947599e-01 1.70679724e+00 -4.15734619e-01 1.80516452e-01 3.12402576e-01 1.13815054e-01 1.94765732e-01 4.54830676e-01 3.28215122e-01 8.19717795e-02 -6.18184432e-02 4.76272732e-01 1.02903284e-01 -4.47246015e-01 -1.37893781e-01 -8.25548768e-01 9.71849918e-01 9.24774170e-01 6.23712242e-01 -7.85209954e-01 -8.85336567e-03 2.10010596e-02 1.09127060e-01 9.69094038e-01 4.53247517e-01 -3.79529119e-01 5.36622524e-01 -1.27048683e+00 5.06502271e-01 1.20463125e-01 3.30506176e-01 8.88867140e-01 3.42188597e-01 -3.44620012e-02 5.61608076e-01 6.54234141e-02 8.95622611e-01 -9.61375609e-02 -3.59036833e-01 3.21881443e-01 5.71543753e-01 4.18621510e-01 -1.02377498e+00 -3.62440646e-01 -7.98954189e-01 -1.10461807e+00 7.99569607e-01 1.27896771e-01 -9.93104875e-02 -1.02819049e+00 1.01855266e+00 1.83865771e-01 2.39967722e-02 1.18047602e-01 1.16044879e+00 6.11650825e-01 7.33748019e-01 7.24985451e-02 -1.11105330e-01 1.28749156e+00 -4.39103037e-01 -5.32088995e-01 -3.98842961e-01 1.69671819e-01 -6.84365809e-01 6.16835356e-01 2.96826839e-01 -5.73989153e-01 -6.35448277e-01 -9.40968812e-01 4.91987288e-01 -9.21486855e-01 3.77096415e-01 7.22598791e-01 6.80835307e-01 -1.12518489e+00 5.76821089e-01 -6.38764977e-01 -1.80552199e-01 5.36518812e-01 1.57354861e-01 -1.40490383e-01 7.13213952e-03 -1.12954676e+00 9.28234100e-01 4.46144849e-01 9.09430563e-01 -8.45352411e-01 -4.67108577e-01 -5.29082775e-01 2.55528569e-01 9.67207029e-02 -5.55274747e-02 5.93958378e-01 -1.42836177e+00 -7.86741316e-01 9.75692809e-01 2.07988266e-02 -5.95369041e-01 4.89473999e-01 -1.06007501e-01 -4.70566064e-01 2.92975187e-01 8.08939990e-03 4.90950882e-01 1.04679096e+00 -1.44013834e+00 -7.93138385e-01 -3.60786349e-01 -4.56486382e-02 5.90789057e-02 -7.57899210e-02 2.69247413e-01 4.43438143e-02 -5.26876569e-01 1.25875846e-01 -9.40809071e-01 -3.07318598e-01 -2.27138922e-02 -2.40061227e-02 2.09935606e-01 8.86398971e-01 -7.47958302e-01 7.34343529e-01 -2.19454885e+00 -2.92730212e-01 3.00978899e-01 8.44732299e-02 9.21209812e-01 -1.68679938e-01 3.93302023e-01 -3.40423137e-01 1.85105011e-01 -7.86802053e-01 3.61019969e-02 -4.80554432e-01 -2.11213976e-02 -2.46226788e-01 7.55160868e-01 7.60220945e-01 5.70420325e-01 -7.06511438e-01 -2.41721757e-02 5.52397549e-01 7.71903992e-01 -8.82591084e-02 2.19843328e-01 -2.93026835e-01 6.39048159e-01 -1.23260140e-01 5.35710037e-01 1.47676599e+00 6.63337037e-02 5.34065925e-02 6.15873113e-02 -6.40490174e-01 -3.55343185e-02 -1.13849771e+00 7.79251575e-01 -3.63899142e-01 9.15034056e-01 3.69616449e-01 -9.43169534e-01 9.54752326e-01 4.22042906e-01 1.72299385e-01 -8.60257447e-01 -1.02320209e-01 4.14837956e-01 1.51567087e-01 -7.73573995e-01 5.37859738e-01 -3.68228555e-01 6.37683868e-01 -6.01375401e-02 -3.98488343e-01 -2.61063606e-01 -2.32281432e-01 -3.05448651e-01 3.59475553e-01 -1.00529931e-01 4.36557271e-02 -5.26640773e-01 5.29089451e-01 2.92248309e-01 3.08163762e-01 6.52715743e-01 -9.81675610e-02 8.68502736e-01 4.88183722e-02 -7.72287250e-01 -8.01650226e-01 -6.72618330e-01 -4.22339827e-01 5.47326565e-01 -1.58499837e-01 3.23623747e-01 -5.83503008e-01 -3.38461846e-01 -6.28317222e-02 3.73708725e-01 -7.36628592e-01 3.85774076e-01 -4.50915575e-01 -1.27003014e+00 3.22375804e-01 3.26043010e-01 7.26317823e-01 -1.19641626e+00 -8.99677813e-01 1.32547721e-01 -1.07572041e-01 -1.30318797e+00 4.15446550e-01 3.19636762e-01 -9.96140718e-01 -1.24515641e+00 -5.76867461e-01 -3.75184089e-01 5.43162286e-01 9.04614925e-01 1.03077877e+00 3.68217021e-01 -5.24962485e-01 -2.39345491e-01 -4.40378040e-01 -7.49513030e-01 -1.99290603e-01 -4.52706851e-02 -4.28064227e-01 3.59544426e-01 3.71497989e-01 -4.27684188e-01 -7.30157733e-01 9.40938741e-02 -1.25092125e+00 -1.27573282e-01 7.46121526e-01 7.00443625e-01 4.83991593e-01 3.27229619e-01 2.03008622e-01 -8.37792993e-01 1.45914957e-01 -3.08392286e-01 -1.19237792e+00 -1.02211619e-02 -4.51989621e-01 -4.20150489e-01 3.66160780e-01 1.87318325e-01 -9.39945579e-01 3.46232831e-01 -7.95779377e-02 -3.13176274e-01 -5.67015052e-01 8.07841897e-01 6.13634065e-02 -4.69489664e-01 7.11159647e-01 3.00269455e-01 -2.02273309e-01 -4.01942074e-01 -4.86200601e-02 7.25097716e-01 2.96622157e-01 -2.75669675e-02 1.18205547e+00 9.89105821e-01 3.45319218e-04 -1.31228495e+00 -7.49214828e-01 -9.79329884e-01 -8.32087934e-01 -3.88744086e-01 1.18624604e+00 -1.22092545e+00 -2.44857326e-01 6.27114892e-01 -1.13363397e+00 -2.43564770e-01 1.13431640e-01 3.98710430e-01 2.09677637e-01 1.40301228e-01 7.89161697e-02 -1.42460036e+00 -5.62741518e-01 -9.15877402e-01 9.18088555e-01 1.72623962e-01 5.77987134e-01 -8.13606203e-01 -6.18622676e-02 3.11362743e-01 9.10013556e-01 6.42474830e-01 6.91621244e-01 -2.13130012e-01 -9.20457125e-01 -2.79760540e-01 -6.88224256e-01 7.87632465e-01 2.20807999e-01 2.35461488e-01 -1.62809277e+00 -4.29630339e-01 9.27110240e-02 1.52280964e-02 1.25607836e+00 4.58561778e-01 9.54879224e-01 -1.29494041e-01 4.02760543e-02 7.51713037e-01 2.03262615e+00 3.81824165e-03 9.05192614e-01 4.26282853e-01 4.83326584e-01 7.44943559e-01 5.06484747e-01 1.89317673e-01 -2.12362841e-01 4.33160901e-01 1.20717144e+00 -6.59955800e-01 1.21523164e-01 4.79455441e-01 -1.42767206e-01 2.45177031e-01 -6.63673460e-01 5.60335517e-02 -1.05581617e+00 8.43447208e-01 -1.53044653e+00 -1.13353205e+00 -7.18084753e-01 2.36844993e+00 2.61455655e-01 -5.80294877e-02 -2.68452644e-01 2.03165680e-01 6.08162761e-01 3.54082555e-01 -6.01994433e-02 -4.23026308e-02 -6.11768961e-01 7.32007742e-01 8.68211091e-01 4.21743989e-01 -1.64610386e+00 8.30344439e-01 5.38367128e+00 4.90790993e-01 -1.61803651e+00 2.32567549e-01 4.33940113e-01 2.87200928e-01 -1.23889362e-02 5.92251588e-03 -5.38272858e-01 2.62592852e-01 6.79960370e-01 6.70809388e-01 2.01072633e-01 5.55131733e-01 6.67981207e-01 -5.34167528e-01 -2.27380574e-01 8.01637590e-01 -2.04799071e-01 -1.39638245e+00 -2.36973688e-02 1.64460542e-03 9.57921684e-01 5.20523429e-01 1.26864269e-01 -9.11196694e-02 -1.39112845e-01 -1.17045760e+00 6.29086673e-01 3.87282103e-01 6.79775596e-01 -5.18850148e-01 1.10861456e+00 2.01355129e-01 -1.19891322e+00 -1.22221649e-01 -5.90031207e-01 -4.36484843e-01 -3.74823540e-01 9.87287939e-01 -6.82136893e-01 7.54637241e-01 9.56081212e-01 4.10495073e-01 -7.32797503e-01 1.37695205e+00 -2.39748493e-01 6.29584730e-01 -2.73766339e-01 1.66099042e-01 7.12742627e-01 -3.51150960e-01 3.16909611e-01 1.52619731e+00 3.06021541e-01 1.62178606e-01 -7.64093548e-02 1.01500309e+00 3.18726391e-01 6.21015765e-02 -6.61385119e-01 -8.48002639e-03 1.24637105e-01 1.54971445e+00 -7.90608943e-01 -1.83725268e-01 -3.44063252e-01 6.74308181e-01 -9.24329758e-02 5.06412148e-01 -5.59413135e-01 -3.14086139e-01 9.11874235e-01 2.96542466e-01 5.75546443e-01 -3.93600017e-01 -2.45425597e-01 -9.57570434e-01 2.67457366e-01 -8.42379987e-01 -1.33997034e-02 -8.37703109e-01 -9.61169720e-01 8.28870535e-01 -8.89914855e-02 -1.47638547e+00 1.60926610e-01 -8.46500278e-01 -8.28384399e-01 1.40168417e+00 -2.36890006e+00 -1.14131820e+00 -9.95262742e-01 3.32896352e-01 2.09121823e-01 1.69620469e-01 9.57696795e-01 3.16559792e-01 -1.81500211e-01 -1.90183297e-01 4.10274863e-01 2.51836509e-01 2.02821732e-01 -1.07845855e+00 4.04903620e-01 1.23051238e+00 -5.87787293e-02 2.67360419e-01 7.25143790e-01 -4.11080092e-01 -1.02755535e+00 -1.46240544e+00 7.54076958e-01 2.61434969e-02 3.16305727e-01 -4.37622994e-01 -1.27189529e+00 2.02060178e-01 2.35049650e-01 3.62757355e-01 3.21629077e-01 -3.42566997e-01 -2.61338890e-01 -3.30678582e-01 -9.77469027e-01 3.42560490e-03 3.21871907e-01 -5.82501590e-01 -2.03976974e-01 5.02879620e-01 1.80793479e-01 -1.42560363e-01 -2.59603977e-01 6.14713728e-01 2.73427963e-01 -1.49022210e+00 7.79362977e-01 -3.41297150e-01 3.74486715e-01 -5.07666945e-01 -2.22876042e-01 -1.30226791e+00 -3.10108215e-01 -4.03378569e-02 6.82231486e-01 8.40791643e-01 4.81591254e-01 -5.60274243e-01 6.56764805e-01 2.58418149e-03 -1.40466258e-01 -3.06905955e-01 -8.81101787e-01 -8.52134287e-01 1.56725317e-01 -3.60451341e-01 4.12738591e-01 1.14706719e+00 -9.64946806e-01 -1.78168342e-01 -2.01303795e-01 9.88874853e-01 6.32293880e-01 5.11785984e-01 6.51456237e-01 -1.58258605e+00 8.59783515e-02 -3.50903183e-01 -3.50408643e-01 -1.31056800e-01 -1.98923826e-01 -6.04780614e-01 1.34342358e-01 -1.46054268e+00 1.25363553e-02 -6.57955647e-01 -2.87990749e-01 3.80172133e-01 -3.74403477e-01 4.91041005e-01 2.16522217e-01 2.41150662e-01 1.61279500e-01 2.81236380e-01 8.73224676e-01 -3.95506889e-01 -7.90330023e-03 2.22474247e-01 -4.82324660e-02 4.95096415e-01 1.05800068e+00 -4.94466156e-01 1.20516494e-01 -7.02568412e-01 3.58038247e-01 -2.73878634e-01 8.53609860e-01 -1.32267189e+00 -5.94462827e-02 -1.65846929e-01 4.51571763e-01 -7.74694204e-01 3.38951796e-01 -9.96858239e-01 2.92098552e-01 5.51659822e-01 1.80710986e-01 -9.36047733e-02 4.13047880e-01 2.63214409e-01 -5.88658571e-01 -1.93791926e-01 1.08189964e+00 -5.05843759e-01 -8.07920754e-01 3.49650085e-01 -5.13936162e-01 -4.19692069e-01 8.24060321e-01 -1.60161838e-01 -1.37712389e-01 -1.87952518e-01 -5.91618001e-01 -5.63678928e-02 2.47610450e-01 1.41623735e-01 6.24713361e-01 -6.68212354e-01 -9.98923004e-01 3.55947137e-01 3.57520401e-01 1.08803205e-01 4.33010995e-01 8.55653882e-01 -8.78516793e-01 5.06385684e-01 -2.38309234e-01 -8.06851447e-01 -1.30590689e+00 5.41854143e-01 8.08682740e-01 -4.17184271e-02 -4.71359611e-01 6.92515373e-01 9.90219787e-02 -3.42576295e-01 -1.34493575e-01 -5.47211409e-01 -3.24540347e-01 2.55768478e-01 5.71277499e-01 -4.94899862e-02 7.41263390e-01 -6.42018616e-01 -3.02461207e-01 6.26972616e-01 4.46004748e-01 2.17437431e-01 1.62390697e+00 1.84212163e-01 -3.68546963e-01 2.10273921e-01 8.44955802e-01 -7.04188272e-02 -1.06428945e+00 -2.84275651e-01 -2.53107715e-02 -7.47356296e-01 4.05164272e-01 -7.64309347e-01 -1.43578207e+00 1.35623145e+00 1.11555147e+00 4.87978101e-01 1.26219118e+00 -4.70014095e-01 6.95522875e-02 2.74955720e-01 1.24411300e-01 -7.67716467e-01 -4.47516054e-01 4.32934821e-01 8.13206673e-01 -1.69877720e+00 1.47824258e-01 -5.13965905e-01 -2.20298618e-01 1.12599051e+00 2.49938354e-01 -2.02012300e-01 7.36394882e-01 3.37089561e-02 5.05498171e-01 -5.23078084e-01 -2.46538907e-01 -8.28962624e-01 1.33475944e-01 6.72150552e-01 4.61747080e-01 3.66828740e-01 -8.45236629e-02 -8.57858062e-02 2.67015010e-01 -4.85166013e-02 4.25029784e-01 7.89073348e-01 -6.43612206e-01 -6.06092870e-01 -7.83522964e-01 4.07254487e-01 -5.74901879e-01 -4.63968277e-01 -2.86188424e-01 8.97853255e-01 4.79676634e-01 9.13104713e-01 5.25284819e-02 -9.15552452e-02 7.43124261e-02 -1.23874068e-01 8.07949528e-02 -6.51855052e-01 -1.05257177e+00 5.11598326e-02 -1.26469597e-01 -1.98825285e-01 -9.57344592e-01 -6.20127618e-01 -6.60756469e-01 2.00880244e-02 -6.20337188e-01 2.14661494e-01 1.16631627e+00 8.12237263e-01 4.14602347e-02 4.13215667e-01 8.01858246e-01 -1.25475609e+00 -2.63330072e-01 -1.07830298e+00 -7.57725358e-01 8.86938199e-02 7.38202274e-01 -5.46185434e-01 -4.85991091e-01 -7.95058087e-02]

[9.748889923095703, -1.7264128923416138]

7f5e3dd3-6618-44c9-8782-46edb16b7f8b

clip2protect-protecting-facial-privacy-using-1

2306.10008

https://arxiv.org/abs/2306.10008v2

https://arxiv.org/pdf/2306.10008v2.pdf

CLIP2Protect: Protecting Facial Privacy using Text-Guided Makeup via Adversarial Latent Search

The success of deep learning based face recognition systems has given rise to serious privacy concerns due to their ability to enable unauthorized tracking of users in the digital world. Existing methods for enhancing privacy fail to generate naturalistic images that can protect facial privacy without compromising user experience. We propose a novel two-step approach for facial privacy protection that relies on finding adversarial latent codes in the low-dimensional manifold of a pretrained generative model. The first step inverts the given face image into the latent space and finetunes the generative model to achieve an accurate reconstruction of the given image from its latent code. This step produces a good initialization, aiding the generation of high-quality faces that resemble the given identity. Subsequently, user-defined makeup text prompts and identity-preserving regularization are used to guide the search for adversarial codes in the latent space. Extensive experiments demonstrate that faces generated by our approach have stronger black-box transferability with an absolute gain of 12.06% over the state-of-the-art facial privacy protection approach under the face verification task. Finally, we demonstrate the effectiveness of the proposed approach for commercial face recognition systems. Our code is available at https://github.com/fahadshamshad/Clip2Protect.

['Karthik Nandakumar', 'Muzammal Naseer', 'Fahad Shamshad']

2023-06-16

clip2protect-protecting-facial-privacy-using

http://openaccess.thecvf.com//content/CVPR2023/html/Shamshad_CLIP2Protect_Protecting_Facial_Privacy_Using_Text-Guided_Makeup_via_Adversarial_Latent_CVPR_2023_paper.html

http://openaccess.thecvf.com//content/CVPR2023/papers/Shamshad_CLIP2Protect_Protecting_Facial_Privacy_Using_Text-Guided_Makeup_via_Adversarial_Latent_CVPR_2023_paper.pdf

cvpr-2023-1

['face-recognition', 'face-verification']

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

[ 3.49001974e-01 1.90530926e-01 9.51459780e-02 -5.99103391e-01 -7.32500613e-01 -8.57331753e-01 5.64011991e-01 -6.31730497e-01 -4.96347137e-02 4.33610290e-01 9.30048078e-02 -2.71910548e-01 1.66949064e-01 -6.70301020e-01 -8.41568768e-01 -1.01986861e+00 1.00584909e-01 -6.58720434e-02 -5.96213102e-01 1.16324805e-01 -1.30683754e-03 8.25808346e-01 -1.36740339e+00 4.07670647e-01 6.25978351e-01 1.07784390e+00 -4.93907899e-01 4.15584445e-01 3.00367355e-01 2.54886508e-01 -4.36582416e-01 -1.00685883e+00 7.84700930e-01 -5.09923577e-01 -5.25035977e-01 1.24256976e-01 8.21185231e-01 -6.35848820e-01 -6.44442976e-01 1.38534248e+00 4.53683227e-01 -2.51127809e-01 4.80915248e-01 -1.67211223e+00 -1.06602383e+00 -2.53680032e-02 -5.41000903e-01 -3.82544398e-01 2.88896173e-01 1.87906012e-01 5.37829101e-01 -1.02572238e+00 4.57934439e-01 1.30670166e+00 4.65029091e-01 1.11274469e+00 -1.38258994e+00 -1.35152280e+00 -2.27081746e-01 -1.37211785e-01 -1.74042284e+00 -1.12074196e+00 6.96555912e-01 -2.78992504e-01 2.42634207e-01 4.20346737e-01 1.96305066e-01 1.25268865e+00 5.52993640e-02 4.18348044e-01 1.04856360e+00 -2.38012314e-01 -5.22016883e-02 4.38776493e-01 -4.40463483e-01 9.64291275e-01 3.14203203e-01 4.73088682e-01 -6.03827655e-01 -5.93453705e-01 7.64098763e-01 7.32431188e-02 -3.37552160e-01 -5.17991364e-01 -3.94301534e-01 8.15830648e-01 2.74720579e-01 4.85020354e-02 -2.13918731e-01 1.00365795e-01 6.60717636e-02 2.16533139e-01 2.69668698e-01 -2.39224150e-03 3.82523537e-02 3.95373046e-01 -9.40376997e-01 1.69102296e-01 8.17188323e-01 9.20753002e-01 6.46961629e-01 8.08097571e-02 -8.43338519e-02 6.03816390e-01 5.29652715e-01 6.30702198e-01 1.64306134e-01 -9.56863344e-01 1.84985414e-01 3.03258777e-01 7.65282139e-02 -1.33516979e+00 4.78315741e-01 -1.45377055e-01 -9.02456045e-01 5.29370129e-01 4.23483729e-01 -2.34999076e-01 -9.50405061e-01 1.97447789e+00 3.87729198e-01 2.35225752e-01 1.71244502e-01 6.37184203e-01 3.98009181e-01 6.30002379e-01 8.18688199e-02 -7.11037293e-02 1.36491787e+00 -4.81384367e-01 -6.79382563e-01 -1.24951929e-01 1.48258328e-01 -7.68561304e-01 7.74451196e-01 8.86602178e-02 -8.33994567e-01 -3.50706786e-01 -1.01035237e+00 7.25956410e-02 -1.43151209e-01 2.71420270e-01 5.60388863e-01 1.38067222e+00 -1.20467210e+00 3.44277859e-01 -6.24480605e-01 -1.44878328e-01 1.05829358e+00 7.39698589e-01 -8.85212719e-01 -2.03684494e-01 -1.02656651e+00 3.15798610e-01 -1.51098818e-02 1.71502635e-01 -1.16507280e+00 -6.70776069e-01 -9.12182331e-01 1.10584579e-01 1.77230567e-01 -4.51317161e-01 8.64662409e-01 -1.36312425e+00 -1.43108845e+00 1.35079074e+00 -2.91719019e-01 -3.17432940e-01 7.23097205e-01 -2.28004962e-01 -6.01777911e-01 8.89880881e-02 4.66886759e-02 7.55414903e-01 1.46247399e+00 -1.39013040e+00 -2.13498235e-01 -5.78224421e-01 -2.88518727e-01 -1.04276516e-01 -6.54536247e-01 2.63030887e-01 -5.39661765e-01 -6.82605088e-01 -8.55418295e-02 -1.15551949e+00 1.49876744e-01 5.80947459e-01 -4.49803293e-01 3.25614661e-01 1.14676511e+00 -7.64821649e-01 9.65640664e-01 -2.47193432e+00 -1.44438833e-01 4.56617981e-01 1.57599464e-01 4.97296304e-01 -2.11532593e-01 1.88873619e-01 -3.33937287e-01 2.95932502e-01 -2.61264205e-01 -6.30441964e-01 5.07588359e-03 -3.51062864e-02 -6.39504194e-01 8.46583545e-01 1.67475328e-01 7.64200032e-01 -3.74493003e-01 -2.12415963e-01 -1.31567419e-01 9.05625641e-01 -6.04662180e-01 2.86621392e-01 1.41549721e-01 4.45563167e-01 -3.12521279e-01 7.42254198e-01 1.28451955e+00 4.48614806e-02 2.09455848e-01 -2.45180004e-03 2.83570170e-01 -1.90256357e-01 -1.02938128e+00 1.31427097e+00 1.95553377e-02 4.59744185e-01 4.34768587e-01 -3.72567266e-01 8.93958330e-01 4.71258938e-01 2.49124005e-01 -2.99862683e-01 3.73163253e-01 2.09242795e-02 -2.79726326e-01 -2.38332108e-01 4.92564477e-02 -1.61529407e-01 9.33627039e-02 4.41575676e-01 4.85703796e-02 3.84928197e-01 -5.28200984e-01 7.51949921e-02 7.02469289e-01 8.52273591e-03 -1.30188078e-01 -1.34441689e-01 7.79780567e-01 -7.94562876e-01 5.49813807e-01 4.92348194e-01 -4.22494590e-01 5.87010086e-01 4.11739618e-01 -3.72922391e-01 -9.49560165e-01 -9.76210892e-01 -8.71604681e-02 8.07325006e-01 -1.29553825e-01 -1.92367733e-01 -1.15801978e+00 -1.01578057e+00 2.68888265e-01 4.38406289e-01 -8.58626306e-01 -3.95606190e-01 -3.57182443e-01 -4.14227754e-01 1.02431881e+00 6.77767694e-02 8.04473877e-01 -8.64094734e-01 -1.04276851e-01 -3.74186963e-01 8.17726087e-03 -9.26910400e-01 -8.72246802e-01 -5.41270614e-01 -5.39107919e-01 -1.03239715e+00 -6.45908892e-01 -8.23601902e-01 1.22939336e+00 4.71465401e-02 4.33888257e-01 3.54152203e-01 -3.70713562e-01 3.46042603e-01 1.47243872e-01 -2.47570381e-01 -6.77560449e-01 -2.78417349e-01 3.47161323e-01 8.82001400e-01 5.94245613e-01 -5.45027196e-01 -6.35353923e-01 3.28224659e-01 -1.03422892e+00 -1.87744886e-01 3.85600954e-01 7.11705387e-01 3.15966934e-01 7.36209899e-02 3.08217287e-01 -8.84672821e-01 4.86235738e-01 -3.28577995e-01 -8.49161446e-01 2.22889721e-01 -5.90845764e-01 5.80191165e-02 4.93088752e-01 -5.07216752e-01 -1.07204342e+00 4.82949048e-01 -3.19888681e-01 -6.72436416e-01 -2.35758334e-01 -1.68752894e-01 -7.35921800e-01 -6.54978752e-01 5.23640752e-01 4.56399143e-01 3.84792596e-01 -2.97263414e-01 4.13176537e-01 6.40704155e-01 6.55536234e-01 -5.05122423e-01 1.32275689e+00 6.72813296e-01 6.22395286e-03 -6.24299765e-01 -4.45499748e-01 1.96735561e-02 -5.46182573e-01 -2.39243835e-01 7.40926981e-01 -8.61615062e-01 -1.00857997e+00 7.90789425e-01 -9.73494112e-01 2.14185357e-01 1.51570022e-01 -4.89013717e-02 -4.85337943e-01 3.74211907e-01 -4.37353134e-01 -9.31432903e-01 -4.74023491e-01 -1.01394570e+00 1.06479871e+00 2.72179306e-01 5.76980691e-03 -6.06773138e-01 -1.77865729e-01 3.88541132e-01 3.85556847e-01 4.65081155e-01 5.78689396e-01 -4.38231498e-01 -8.16594243e-01 -5.29478073e-01 -7.20668659e-02 5.58101535e-01 4.17929500e-01 -7.06156567e-02 -1.25895727e+00 -6.48637056e-01 2.34012768e-01 -2.64486790e-01 5.73624134e-01 1.14250623e-01 1.17983007e+00 -8.87121737e-01 -3.43600720e-01 1.17022991e+00 1.37925816e+00 2.17995048e-01 9.38675880e-01 -1.96195580e-02 6.06128216e-01 6.63521409e-01 1.39847070e-01 5.32382846e-01 -9.65821370e-02 4.68513757e-01 4.36641604e-01 -1.53073505e-01 1.55595705e-01 -5.76004326e-01 4.25498158e-01 -6.24956489e-02 2.91009665e-01 -7.24187270e-02 -6.34378195e-01 3.83385867e-01 -1.56043494e+00 -1.11273849e+00 4.71722335e-01 2.32233715e+00 8.92404675e-01 -2.21783638e-01 -5.55409454e-02 -2.22514540e-01 7.93857992e-01 8.35732669e-02 -6.17762804e-01 -2.65104383e-01 -1.41762765e-02 1.46714941e-01 5.83183587e-01 5.82682431e-01 -1.24451923e+00 1.19323087e+00 5.86287117e+00 6.34131253e-01 -1.21711516e+00 1.03521705e-01 9.72488940e-01 -1.55816242e-01 -1.68351799e-01 -1.88843668e-01 -9.79782760e-01 4.30702716e-01 9.45796728e-01 -3.10259670e-01 5.91573954e-01 9.14017618e-01 -2.48329584e-02 4.62635547e-01 -1.03959918e+00 1.08394575e+00 3.21375430e-01 -1.37117338e+00 2.55268127e-01 5.42682350e-01 6.67512298e-01 -5.77389657e-01 7.16591239e-01 -2.11249545e-01 1.86712533e-01 -1.27964652e+00 6.63616180e-01 5.10894179e-01 1.26928008e+00 -1.05947113e+00 2.54830837e-01 1.09212294e-01 -8.96893799e-01 -4.65845466e-02 -4.04995143e-01 4.35311347e-01 -1.74644172e-01 8.63676965e-02 -8.29364598e-01 1.36130691e-01 5.58952093e-01 2.62724012e-01 -3.89121264e-01 4.95619029e-01 -3.55675876e-01 5.37073791e-01 -2.48540044e-01 4.61980671e-01 -6.51841462e-02 -1.53355330e-01 6.05482399e-01 8.89123619e-01 4.26557511e-01 2.65221387e-01 -2.41467044e-01 1.18477726e+00 -7.16869354e-01 4.04751785e-02 -1.05201042e+00 -2.60741651e-01 5.23530781e-01 1.18558311e+00 -2.45611757e-01 1.23744681e-01 -1.92684859e-01 1.50697529e+00 1.64441586e-01 4.09279734e-01 -8.55155706e-01 -2.50277668e-01 1.30556738e+00 1.94928408e-01 2.63197541e-01 -5.41800596e-02 -1.37487143e-01 -1.08175933e+00 1.21065274e-01 -1.21866024e+00 3.05999398e-01 -4.55829114e-01 -1.10172427e+00 7.97804177e-01 -5.11009872e-01 -1.04909611e+00 -2.00222582e-01 -4.19986695e-01 -3.46642196e-01 1.21451628e+00 -1.21859276e+00 -1.44202566e+00 -1.95591405e-01 9.49287415e-01 -1.65361911e-02 -5.77342451e-01 1.19503117e+00 2.48333499e-01 -5.12108147e-01 1.32795751e+00 1.19909659e-01 5.22461057e-01 7.35722721e-01 -5.92879891e-01 6.50613964e-01 1.18269920e+00 1.45481557e-01 9.75400925e-01 3.90597671e-01 -6.72735929e-01 -1.45605516e+00 -1.24489677e+00 8.91273558e-01 -5.34488201e-01 1.44899815e-01 -8.42985511e-01 -9.35599506e-01 8.44423473e-01 1.57586262e-01 1.75691143e-01 9.94673193e-01 -3.90320629e-01 -8.41240466e-01 -3.31984878e-01 -1.68964922e+00 5.97954571e-01 8.63157988e-01 -9.59774733e-01 -3.77013371e-03 1.47245899e-01 4.07573491e-01 -1.92176059e-01 -4.84051079e-01 1.18010521e-01 8.81268263e-01 -8.04282367e-01 9.66086030e-01 -7.31042683e-01 2.26605013e-01 -2.94648796e-01 -1.66085497e-01 -6.52937233e-01 -2.52873331e-01 -1.23913145e+00 -1.57156177e-02 1.52588320e+00 2.89063364e-01 -6.74708247e-01 1.25209737e+00 1.09335494e+00 5.98165154e-01 -3.75716299e-01 -9.89521265e-01 -6.76326931e-01 -3.76324877e-02 -7.03826547e-02 8.20682824e-01 1.01393747e+00 -4.67081308e-01 -3.16031486e-01 -8.38708103e-01 7.31851220e-01 1.01648009e+00 -1.47705019e-01 8.52786183e-01 -8.13364029e-01 -8.81487795e-04 -5.81162497e-02 -4.68658805e-01 -7.35053420e-01 4.96307254e-01 -9.20552433e-01 -1.67239681e-01 -6.61770880e-01 1.63478255e-01 -2.77075619e-01 -2.44137570e-01 7.98555493e-01 6.42786697e-02 5.83004951e-01 1.89896107e-01 1.22994944e-01 -4.78072528e-04 5.21952569e-01 8.54827046e-01 3.94673873e-04 4.35291044e-02 1.02441482e-01 -1.11134696e+00 4.11611259e-01 8.60323608e-01 -6.87133908e-01 -4.25320119e-01 -2.41970226e-01 -3.11112285e-01 -1.28824666e-01 6.09715343e-01 -9.79795218e-01 1.28992677e-01 -4.91849147e-02 6.40528500e-01 3.93442810e-02 4.04450238e-01 -1.13164461e+00 4.96631742e-01 4.59584594e-01 -3.27414423e-01 -1.86767831e-01 4.69553709e-01 4.20720607e-01 -3.88858467e-02 2.73142904e-02 1.21964955e+00 1.48915529e-01 -2.33173653e-01 7.69618213e-01 -3.67267355e-02 -3.03514361e-01 1.13966036e+00 -3.22587132e-01 -1.60706669e-01 -5.81016541e-01 -7.07990050e-01 -1.27259001e-01 7.12318659e-01 6.46418750e-01 7.59064853e-01 -1.48720431e+00 -8.46265137e-01 1.00577033e+00 1.70310810e-02 -7.76450634e-01 1.66913688e-01 1.54703066e-01 -3.74680758e-01 2.53157675e-01 -5.01518190e-01 -2.88676620e-01 -1.83867204e+00 6.57721937e-01 5.56408107e-01 3.65105152e-01 -3.35872114e-01 1.19089007e+00 5.49924970e-01 -1.44428194e-01 3.60687912e-01 3.75508845e-01 2.99468160e-01 -2.54662842e-01 7.70086169e-01 -4.89551313e-02 -1.42146066e-01 -1.14882600e+00 -4.47070569e-01 3.45309407e-01 -1.75326526e-01 -1.96334079e-01 1.17107785e+00 -9.40279663e-02 -2.31249630e-01 -5.44189990e-01 1.65496850e+00 3.00088704e-01 -1.52755415e+00 -8.22514892e-02 -3.10276091e-01 -1.09888804e+00 -1.70065910e-01 -6.66670799e-01 -1.32238042e+00 6.35930300e-01 9.11445916e-01 -1.22485667e-01 1.32293653e+00 -1.57318994e-01 7.06532776e-01 -5.60035370e-02 3.37726414e-01 -5.28300345e-01 -1.12504847e-01 -7.64448661e-03 1.06839752e+00 -1.10311377e+00 -1.86417416e-01 -5.49069762e-01 -5.73656261e-01 8.79069090e-01 4.17883039e-01 7.49620870e-02 7.76731610e-01 2.65870154e-01 1.87261209e-01 -8.04572273e-03 -4.04175282e-01 3.97586823e-01 3.88410836e-01 7.17106402e-01 5.78974076e-02 -6.39243843e-03 1.29923254e-01 5.34174025e-01 -2.73048967e-01 8.05166364e-03 1.06225252e-01 8.56168747e-01 -3.78454290e-02 -1.41493630e+00 -5.65234840e-01 2.67793555e-02 -7.63279378e-01 2.03438736e-02 -6.46533847e-01 3.36437255e-01 1.46923035e-01 7.55223095e-01 -2.01801106e-01 -4.62550044e-01 1.72487378e-01 3.11210275e-01 3.94990206e-01 -3.49121898e-01 -5.08384168e-01 -3.32763270e-02 -2.11511880e-01 -8.11252892e-01 6.83461949e-02 -8.43576193e-01 -8.53331804e-01 -6.81980312e-01 -1.62644740e-02 5.51836379e-02 6.15896821e-01 3.98555100e-01 7.76009440e-01 -1.91358417e-01 9.79344308e-01 -5.47120214e-01 -4.71181750e-01 -3.62542838e-01 -5.12075484e-01 5.26571453e-01 6.18514717e-01 -2.81459033e-01 -3.89133364e-01 4.55167115e-01]

[12.78046703338623, 0.7730129957199097]

2c835e44-aba1-4c41-bc29-19fc173ccb24

machine-reading-fast-and-slow-when-do-models

2209.07430

https://arxiv.org/abs/2209.07430v1

https://arxiv.org/pdf/2209.07430v1.pdf

Machine Reading, Fast and Slow: When Do Models "Understand" Language?

Two of the most fundamental challenges in Natural Language Understanding (NLU) at present are: (a) how to establish whether deep learning-based models score highly on NLU benchmarks for the 'right' reasons; and (b) to understand what those reasons would even be. We investigate the behavior of reading comprehension models with respect to two linguistic 'skills': coreference resolution and comparison. We propose a definition for the reasoning steps expected from a system that would be 'reading slowly', and compare that with the behavior of five models of the BERT family of various sizes, observed through saliency scores and counterfactual explanations. We find that for comparison (but not coreference) the systems based on larger encoders are more likely to rely on the 'right' information, but even they struggle with generalization, suggesting that they still learn specific lexical patterns rather than the general principles of comparison.

['Isabelle Augenstein', 'Anna Rogers', 'Sagnik Ray Choudhury']

2022-09-15

['coreference-resolution']

['natural-language-processing']

[ 4.41535741e-01 9.45090473e-01 -1.63556352e-01 -4.31703359e-01 -6.59346044e-01 -5.22382021e-01 8.68620455e-01 6.62283421e-01 -5.01918197e-01 6.06804669e-01 7.40206122e-01 -7.22687721e-01 -4.11063612e-01 -8.40419292e-01 -9.93384421e-01 -1.72687113e-01 1.04085788e-01 9.57064271e-01 4.59601372e-01 -6.70391619e-01 5.88500738e-01 1.04988374e-01 -1.76370084e+00 4.25471514e-01 1.07220125e+00 5.69290519e-01 3.34770083e-01 6.36362672e-01 -2.38968775e-01 1.12925589e+00 -3.12505811e-01 -5.17416179e-01 -1.17668845e-01 -7.92145431e-01 -1.56690097e+00 -6.02815211e-01 6.89304709e-01 -2.64252454e-01 -7.14516491e-02 1.13391638e+00 1.97521582e-01 2.26933897e-01 8.16606700e-01 -8.81956756e-01 -8.26280355e-01 1.15305948e+00 -2.10210025e-01 7.94680595e-01 7.18056679e-01 2.42909342e-01 1.54254246e+00 -3.03464234e-01 8.18344057e-01 1.79784751e+00 4.58594382e-01 7.47478187e-01 -1.38256836e+00 -2.68141359e-01 3.17634076e-01 4.51275200e-01 -6.35961115e-01 -4.13482249e-01 2.34154060e-01 -4.12380368e-01 1.25663185e+00 2.13615879e-01 3.47973645e-01 1.09273505e+00 3.44085217e-01 7.14810193e-01 1.33303165e+00 -5.62744141e-01 1.00138567e-01 3.59467752e-02 6.31099999e-01 5.18810570e-01 5.60801864e-01 4.52005595e-01 -7.82940447e-01 1.85395002e-01 5.77344000e-01 -4.99892086e-01 -6.51019752e-01 -4.22202438e-01 -1.42585218e+00 9.95931625e-01 5.61553597e-01 4.41952854e-01 -3.63167852e-01 -2.98608597e-02 2.44667038e-01 5.82606912e-01 -1.07178412e-01 1.10569370e+00 -8.61351907e-01 -4.68652733e-02 -6.44521236e-01 5.57045400e-01 8.76500905e-01 8.11868489e-01 5.97527683e-01 -3.22633773e-01 -1.64471358e-01 3.56289864e-01 2.71858126e-01 2.91718960e-01 6.26124024e-01 -1.14572358e+00 6.19400084e-01 5.04249215e-01 2.48611301e-01 -7.46575117e-01 -6.51944339e-01 -3.98173749e-01 -2.34532729e-01 1.54487982e-01 9.67367947e-01 -5.57141788e-02 -5.24506211e-01 2.37427807e+00 -2.64889032e-01 -5.03516197e-01 4.96486336e-01 8.29953790e-01 5.91077685e-01 4.38391805e-01 2.86796659e-01 -2.82158077e-01 1.37730515e+00 -6.97250009e-01 -4.66005117e-01 -6.79919660e-01 8.62611234e-01 -4.10188973e-01 1.45684910e+00 7.00837225e-02 -1.50648999e+00 -7.56242931e-01 -1.05475056e+00 -5.03098190e-01 -1.47074342e-01 -4.84251797e-01 6.88157380e-01 1.87725663e-01 -1.19381928e+00 9.37654436e-01 -4.64953393e-01 -7.56189525e-01 2.66575456e-01 7.14070946e-02 4.23943950e-03 1.92662001e-01 -1.47112668e+00 1.72412896e+00 6.42834723e-01 -4.52313513e-01 -8.47229362e-01 -7.32618570e-01 -7.40838408e-01 6.78175330e-01 3.49049956e-01 -7.66275167e-01 1.63901532e+00 -1.16663051e+00 -9.52899933e-01 1.11411738e+00 -2.18841642e-01 -7.12412298e-01 4.32564586e-01 -4.13978368e-01 -1.25166625e-01 7.98237696e-02 3.76897156e-01 8.30329835e-01 3.29820931e-01 -1.25801027e+00 -7.54746914e-01 -3.44235718e-01 4.16777551e-01 5.82544804e-01 3.99704486e-01 7.33494610e-02 5.79644918e-01 -2.67639369e-01 1.89675227e-01 -6.19562209e-01 1.57366827e-01 -1.69550315e-01 -2.62487650e-01 -5.75951278e-01 1.06744796e-01 -6.73277438e-01 8.96496832e-01 -1.87403691e+00 2.84486473e-01 -3.16412635e-02 2.21975803e-01 7.13846227e-03 -1.56687692e-01 3.83401573e-01 -6.17211461e-01 2.47656673e-01 -2.48143747e-02 4.87042159e-01 2.58225024e-01 9.68290120e-02 -6.66682959e-01 3.95506918e-02 1.94229141e-01 9.85002220e-01 -1.17389154e+00 -1.24020286e-01 -1.28242835e-01 -3.46712470e-01 -6.83076382e-01 1.38131425e-01 -5.63728094e-01 2.11723328e-01 -2.43176162e-01 -1.42272850e-02 3.24880153e-01 -3.76438797e-01 4.01971936e-01 -6.31837100e-02 -1.78906307e-01 1.12545216e+00 -8.40838373e-01 1.46758902e+00 -1.13400340e-01 8.41039896e-01 -1.79606169e-01 -9.65820968e-01 5.23256779e-01 1.61130905e-01 -5.55336237e-01 -8.90648127e-01 2.71444291e-01 3.90959948e-01 1.07765961e+00 -5.32449782e-01 6.95436820e-02 -4.65219289e-01 1.31754711e-01 8.86636794e-01 9.10986811e-02 9.80780497e-02 1.66837916e-01 4.78948444e-01 8.39823782e-01 1.35662779e-01 5.09423018e-01 -9.25857306e-01 4.26646531e-01 2.46059611e-01 2.40557760e-01 1.32437289e+00 -2.65796751e-01 2.88491189e-01 8.38617325e-01 -5.48830748e-01 -1.06893039e+00 -1.28979206e+00 -1.06302232e-01 1.29445148e+00 3.12392294e-01 -9.73772258e-02 -8.60700250e-01 -4.33324605e-01 -1.50195584e-01 1.76531041e+00 -8.07490885e-01 -4.56576824e-01 -4.59435046e-01 -4.49266464e-01 1.91590413e-01 6.98892474e-01 3.95057350e-01 -1.23406506e+00 -1.30294573e+00 1.63439736e-01 -3.43286872e-01 -5.87679505e-01 -1.93852499e-01 4.13350195e-01 -8.98888171e-01 -1.29598010e+00 -1.57883763e-01 -5.63440323e-01 4.37449247e-01 1.87900901e-01 1.52653110e+00 5.10736227e-01 3.54413360e-01 3.43794435e-01 -3.20504814e-01 -7.95170724e-01 -7.14628100e-01 8.67218897e-02 1.18199997e-01 -7.81819880e-01 7.99830973e-01 -4.74156588e-01 -4.77693170e-01 9.52726305e-02 -6.12001956e-01 2.66849548e-01 7.00465858e-01 8.60141158e-01 -1.07990108e-01 -2.71073908e-01 5.75481713e-01 -1.06943655e+00 9.71386254e-01 -4.40426916e-01 -2.17906550e-01 5.92319965e-01 -6.42305255e-01 5.81155241e-01 4.47203189e-01 -2.13969588e-01 -1.41101551e+00 -7.44806945e-01 1.13182487e-02 3.26034307e-01 -4.14134145e-01 2.91456074e-01 -2.52094597e-01 4.10761535e-01 9.77484524e-01 1.67129248e-01 9.88222808e-02 -3.85989189e-01 3.43510002e-01 7.52580911e-02 5.29186964e-01 -1.16160953e+00 7.27158546e-01 9.20450166e-02 -2.68748611e-01 -5.93080997e-01 -1.56864047e+00 -1.09174609e-01 -5.35005271e-01 2.68134445e-01 1.04048514e+00 -7.04482079e-01 -7.70409942e-01 -2.22167913e-02 -1.41455638e+00 -3.88610959e-01 -5.15182137e-01 4.77944434e-01 -9.44045603e-01 2.34434098e-01 -5.87122679e-01 -4.04133081e-01 -2.75745708e-02 -1.10423207e+00 6.18502319e-01 4.76595730e-01 -8.82904232e-01 -1.07957470e+00 -1.42535530e-02 2.95948833e-01 5.06309748e-01 -2.47577935e-01 1.75178862e+00 -1.30568874e+00 -4.86544907e-01 4.61466849e-01 -5.56854844e-01 -3.78008303e-03 -2.35663697e-01 -4.34839159e-01 -9.21595812e-01 -9.04212147e-02 4.24601436e-01 -5.46428025e-01 9.21739936e-01 3.41006190e-01 7.01367259e-01 -4.45829362e-01 -2.91072875e-01 2.03082278e-01 1.34518790e+00 2.14130998e-01 7.37031877e-01 4.27451879e-01 2.26938188e-01 1.08178508e+00 5.26864707e-01 -2.60423303e-01 5.65774083e-01 3.05550694e-01 2.93940455e-01 3.19608688e-01 -5.39387017e-03 -6.21482611e-01 2.88884908e-01 5.32888770e-01 -1.11584246e-01 1.56807341e-02 -1.03879547e+00 7.66070545e-01 -1.92642570e+00 -1.23383844e+00 -2.26564184e-01 2.06835055e+00 9.97164488e-01 5.29024303e-01 -1.35240793e-01 -1.91888630e-01 5.97140670e-01 1.08390383e-01 -6.21189237e-01 -8.79309356e-01 -1.59122929e-01 2.89098531e-01 1.63957596e-01 8.15720677e-01 -5.51204920e-01 1.07021821e+00 7.11267138e+00 4.13425773e-01 -6.15633905e-01 -1.82285216e-02 6.56155229e-01 3.13152552e-01 -6.98309124e-01 4.19650674e-01 -6.57562852e-01 1.76951867e-02 1.21380711e+00 -3.37182343e-01 3.79299849e-01 5.32052755e-01 1.05230436e-02 -4.53426987e-01 -1.78523278e+00 2.68585891e-01 1.03678741e-01 -1.34305418e+00 5.11706889e-01 -2.94428170e-01 5.42369008e-01 -1.21660717e-02 -2.61704892e-01 3.69700342e-01 6.08465612e-01 -1.11327910e+00 9.41326857e-01 4.17933226e-01 2.37989694e-01 -3.46197993e-01 6.92929029e-01 6.75828993e-01 -3.03107530e-01 -1.09355956e-01 -7.08539248e-01 -6.19674861e-01 -3.26885539e-03 -1.23288296e-01 -4.26971614e-01 5.62359653e-02 4.47763652e-01 2.09040895e-01 -6.62088454e-01 6.50397599e-01 -8.05524945e-01 3.99217188e-01 -6.89840987e-02 -2.38680303e-01 1.63815901e-01 1.70529276e-01 4.64378327e-01 8.36310089e-01 9.28058028e-02 2.62147605e-01 -3.53790045e-01 1.11103523e+00 7.68350884e-02 -3.33672166e-02 -4.56938386e-01 2.12535739e-01 4.83050317e-01 4.66404527e-01 -6.26679957e-01 -4.32513952e-01 -4.21320379e-01 6.82794571e-01 5.37609398e-01 1.18500642e-01 -6.93970263e-01 8.99386257e-02 4.58253145e-01 9.34505388e-02 5.52815720e-02 1.51121065e-01 -2.54483432e-01 -1.34547353e+00 -2.82473117e-01 -1.10209513e+00 4.79882687e-01 -1.20266712e+00 -1.04713321e+00 2.90871352e-01 1.47078544e-01 -3.66421193e-01 -2.97025532e-01 -9.36069906e-01 -8.54935288e-01 8.99655223e-01 -1.58470750e+00 -7.13994980e-01 1.28555551e-01 3.13245624e-01 6.69445992e-01 1.81985378e-01 9.09023225e-01 -5.71173549e-01 -1.16739266e-01 2.21305683e-01 -1.12154849e-01 -1.13865897e-01 4.05074865e-01 -1.57054234e+00 5.19911349e-01 7.46791959e-01 3.06056440e-01 1.00116706e+00 1.21714234e+00 -4.10083622e-01 -8.66324902e-01 -2.43539587e-01 1.40604341e+00 -9.23712432e-01 6.92228734e-01 -4.76270318e-02 -1.11856902e+00 1.06819797e+00 6.28500760e-01 -8.04451942e-01 4.45818305e-01 4.45488960e-01 -5.07464886e-01 3.12789559e-01 -1.01802874e+00 6.61324322e-01 1.18147099e+00 -4.54288453e-01 -1.83080578e+00 4.78037447e-01 7.56439030e-01 -1.77669078e-01 -4.33174938e-01 1.38313666e-01 3.43377739e-01 -1.47258508e+00 8.43838692e-01 -1.33091617e+00 8.07925642e-01 -1.45031855e-01 -8.73214751e-02 -1.53162670e+00 -5.23589194e-01 -2.91675538e-01 2.93278396e-01 1.01398051e+00 8.20889473e-01 -6.24260247e-01 2.27655500e-01 6.83224857e-01 9.89181325e-02 -5.21452606e-01 -8.35360765e-01 -3.98246646e-01 5.39312065e-01 -1.81746319e-01 4.96113300e-01 9.36350107e-01 2.53068417e-01 1.05880368e+00 2.32284173e-01 5.80793917e-02 5.47228813e-01 3.21771175e-01 2.66846627e-01 -1.56500149e+00 -2.76027352e-01 -6.97884798e-01 9.75393429e-02 -9.76333439e-01 2.82100916e-01 -8.20922494e-01 6.89887488e-03 -1.54801476e+00 4.68873680e-01 -2.23298687e-02 -1.70436010e-01 2.55032271e-01 -3.24442416e-01 -6.44181669e-01 2.53251940e-01 3.56225550e-01 -5.79348385e-01 1.13438830e-01 1.08806396e+00 7.47219995e-02 2.10694462e-01 -2.83073395e-01 -1.43719614e+00 1.10291457e+00 7.51655459e-01 -9.29665342e-02 -6.67192400e-01 -8.13155890e-01 7.30541348e-01 1.55179486e-01 3.42314839e-01 -8.18560719e-01 7.48436153e-02 -3.05205762e-01 3.79925013e-01 -1.45037189e-01 -3.88780415e-01 -4.87350076e-01 -3.87524605e-01 8.10069799e-01 -1.04359615e+00 4.12200093e-01 2.84405231e-01 3.21840227e-01 4.55560461e-02 -5.21458685e-01 9.21833158e-01 -5.86639047e-01 -8.65081966e-01 -4.41653013e-01 -4.23655570e-01 5.54173946e-01 6.48235559e-01 -1.75647348e-01 -8.83222401e-01 -6.09035194e-01 -8.55175316e-01 3.16379279e-01 5.73225200e-01 4.15673107e-01 3.43397737e-01 -7.69934058e-01 -7.05448568e-01 -7.30708763e-02 1.17990755e-01 -1.46316171e-01 5.43053336e-02 7.87400961e-01 -5.66699743e-01 8.10220182e-01 -2.78442591e-01 -1.26726344e-01 -5.34571707e-01 7.67311215e-01 5.96392334e-01 -2.96199232e-01 -4.05329198e-01 1.08366656e+00 7.45357633e-01 -3.82075787e-01 -5.02790995e-02 -4.17341530e-01 -4.61560398e-01 1.90832108e-01 6.23253405e-01 -1.96664105e-03 -2.98841327e-01 -5.52568376e-01 -2.82367080e-01 3.50319922e-01 -3.95802706e-01 1.85675725e-01 1.21880496e+00 -2.50760287e-01 -6.28201589e-02 5.49528241e-01 6.40156925e-01 -2.04877570e-01 -1.05443013e+00 -1.01647392e-01 5.12148201e-01 -2.21066877e-01 -3.50389510e-01 -1.20440447e+00 -2.95383781e-01 1.23334241e+00 2.30389521e-01 2.54861742e-01 5.70179582e-01 3.61080915e-01 4.14721459e-01 5.49543858e-01 3.27526689e-01 -9.36918795e-01 -2.33363405e-01 6.46842062e-01 1.01608729e+00 -1.00790799e+00 -9.79428813e-02 -2.19772384e-01 -5.02680123e-01 1.03440249e+00 8.90297592e-01 -1.66508570e-01 1.15045957e-01 -3.14915478e-01 3.00812963e-02 -4.03426141e-01 -1.24108768e+00 -3.28061402e-01 -6.32137235e-04 6.83187842e-01 6.76643431e-01 -8.85478780e-02 -5.87851763e-01 4.46272165e-01 -6.87134147e-01 -4.65568572e-01 8.03046465e-01 4.00864482e-01 -7.95057237e-01 -6.75225973e-01 -1.62005886e-01 7.17225313e-01 -3.01390916e-01 -3.27311546e-01 -6.50740027e-01 1.02939570e+00 2.28158105e-02 6.33152127e-01 2.75451958e-01 1.82455897e-01 1.97922051e-01 3.03766787e-01 7.71192491e-01 -9.34783280e-01 -3.87214780e-01 -3.42038572e-01 2.59342581e-01 -5.65261900e-01 -4.45196867e-01 -9.80171442e-01 -1.25791025e+00 -3.21203232e-01 -3.75666320e-01 3.41474593e-01 -3.35837789e-02 1.39927042e+00 4.24285121e-02 4.59913850e-01 -9.34764817e-02 -2.75152981e-01 -1.08869934e+00 -9.13796008e-01 -3.12174916e-01 8.68709147e-01 3.70560199e-01 -5.81743002e-01 -5.84235787e-01 -3.11349094e-01]

[9.912005424499512, 7.79224157333374]

df8d18f6-09d4-435d-a8fe-059b04a07c0f

multi-label-restricted-boltzmann-machine-for

1910.08149

https://arxiv.org/abs/1910.08149v1

https://arxiv.org/pdf/1910.08149v1.pdf

Multi Label Restricted Boltzmann Machine for Non-Intrusive Load Monitoring

Increasing population indicates that energy demands need to be managed in the residential sector. Prior studies have reflected that the customers tend to reduce a significant amount of energy consumption if they are provided with appliance-level feedback. This observation has increased the relevance of load monitoring in today's tech-savvy world. Most of the previously proposed solutions claim to perform load monitoring without intrusion, but they are not completely non-intrusive. These methods require historical appliance-level data for training the model for each of the devices. This data is gathered by putting a sensor on each of the appliances present in the home which causes intrusion in the building. Some recent studies have proposed that if we frame Non-Intrusive Load Monitoring (NILM) as a multi-label classification problem, the need for appliance-level data can be avoided. In this paper, we propose Multi-label Restricted Boltzmann Machine(ML-RBM) for NILM and report an experimental evaluation of proposed and state-of-the-art techniques.

['Sagar Verma', 'Shikha Singh', 'Angshul Majumdar']

2019-10-17

['non-intrusive-load-monitoring', 'non-intrusive-load-monitoring', 'non-intrusive-load-monitoring']

['knowledge-base', 'miscellaneous', 'time-series']

[ 1.97082847e-01 1.72661468e-02 -3.99287611e-01 -5.90496063e-01 -6.42266452e-01 -1.27360955e-01 4.72585350e-01 1.71747193e-01 -4.04542357e-01 8.67862701e-01 4.48835408e-03 -4.81167100e-02 -1.60105050e-01 -7.87657559e-01 -3.34038913e-01 -1.04076815e+00 2.39691362e-01 7.87291765e-01 -1.75445154e-02 4.30398732e-02 1.09382614e-01 5.29039383e-01 -1.58809638e+00 1.47294283e-01 4.75191951e-01 9.92757738e-01 1.83419377e-01 5.50958395e-01 2.41320148e-01 7.55565703e-01 -6.68640852e-01 2.72167742e-01 -4.83231246e-02 -2.30981916e-01 -1.01068902e+00 -7.33336732e-02 -9.56094936e-02 -4.07903761e-01 2.92163879e-01 1.00340223e+00 5.63949645e-01 3.71287763e-01 8.52881968e-01 -1.42257822e+00 -2.64329731e-01 5.86045206e-01 -9.09972310e-01 4.04560119e-01 2.46446282e-01 -4.01743501e-01 6.14329159e-01 1.50951579e-01 -2.16308594e-01 8.97633433e-01 3.64354044e-01 8.19383681e-01 -1.40675020e+00 -8.06959391e-01 1.71124965e-01 7.03495920e-01 -1.28978622e+00 -4.79374498e-01 1.16239917e+00 -8.71731117e-02 1.45541275e+00 1.03532887e+00 4.36409473e-01 1.25267768e+00 4.86648753e-02 5.40529728e-01 1.32322371e+00 -7.43677199e-01 6.14140213e-01 5.48976123e-01 3.46252143e-01 7.45253488e-02 4.35498476e-01 -2.03708053e-01 -3.17392275e-02 -6.93335354e-01 -4.14648280e-02 1.84861258e-01 -7.41194412e-02 1.31682783e-01 -3.50804061e-01 6.86895370e-01 -3.41087729e-02 8.14316988e-01 -4.70398635e-01 2.21876577e-02 5.37281215e-01 -8.23081732e-02 6.28211141e-01 -1.75946400e-01 -6.10866845e-01 -8.58051181e-02 -1.14393079e+00 -3.35167557e-01 9.30832505e-01 6.51349366e-01 5.42183161e-01 -1.98311448e-01 3.10732484e-01 8.62480104e-01 6.99696958e-01 1.94074735e-01 1.12332153e+00 -4.99903649e-01 6.06671199e-02 5.77378154e-01 -1.57565549e-02 -5.77799916e-01 -6.92040801e-01 1.56646892e-01 -1.17367876e+00 2.42898807e-01 -2.64115393e-01 2.27602571e-01 -5.92660964e-01 1.78762388e+00 5.32515407e-01 4.51521635e-01 -1.08883694e-01 2.53638506e-01 2.54589617e-01 7.02367604e-01 5.23984969e-01 -7.75181711e-01 1.27741838e+00 -7.39774227e-01 -9.14911091e-01 -1.16330229e-01 5.89454532e-01 -4.37433720e-01 8.63894463e-01 6.41645730e-01 -6.37944639e-01 -4.87859100e-01 -1.22787499e+00 3.24066758e-01 -6.49080873e-01 -4.38305229e-01 2.28744209e-01 1.21854389e+00 -8.40005457e-01 7.49551654e-01 -1.15886462e+00 -8.46399665e-01 3.15313071e-01 7.41512656e-01 -9.79829207e-03 3.46649021e-01 -8.11098099e-01 1.13752556e+00 5.63103378e-01 -9.44465585e-03 -7.45853126e-01 -2.25627288e-01 -3.44423413e-01 -2.20196530e-01 4.06543389e-02 -2.90255547e-01 1.26651752e+00 -5.03842175e-01 -1.42016554e+00 7.51267850e-01 6.10299781e-02 -1.98771417e-01 4.28777367e-01 -1.33933336e-01 -8.76821339e-01 -9.03117210e-02 -2.60059804e-01 1.96710393e-01 8.01867187e-01 -1.49459589e+00 -9.64151740e-01 -6.89396083e-01 -3.80781472e-01 -1.58579141e-01 -5.45032322e-01 -2.13516235e-01 4.55150008e-01 -1.57188624e-01 -6.20826744e-02 -9.14816320e-01 7.67332315e-02 -9.37024713e-01 -7.08483100e-01 -6.35224402e-01 1.48467875e+00 -6.44900262e-01 1.42842436e+00 -1.93730772e+00 -4.26168829e-01 2.21539482e-01 -3.54741633e-01 2.30253100e-01 3.75943244e-01 3.52356285e-01 -4.50601220e-01 2.48301044e-01 -9.40452814e-02 -7.95793295e-01 1.58703163e-01 6.93163455e-01 -1.38161391e-01 8.54374528e-01 -6.68451548e-01 5.94932139e-01 -5.28010666e-01 -6.67687237e-01 8.33733737e-01 4.91774380e-01 -1.35936290e-01 3.65036309e-01 -1.20402232e-01 5.11902988e-01 -4.31556374e-01 5.01042008e-01 4.55106348e-01 -9.54791531e-02 3.41379493e-01 -3.17788631e-01 2.26667508e-01 1.95884451e-01 -1.02319598e+00 1.40103877e+00 -6.48408949e-01 -8.84412415e-03 -2.38208562e-01 -1.37029731e+00 5.74799776e-01 8.00930917e-01 9.51086164e-01 -7.53198624e-01 3.39731276e-01 -6.35973215e-02 -3.37800682e-01 -7.25130022e-01 1.25812665e-01 -2.63347328e-01 -1.39380127e-01 6.79782152e-01 -1.85274616e-01 1.87481642e-01 -2.48553440e-01 -5.19533813e-01 1.20685780e+00 8.49641487e-03 5.89180768e-01 -3.57427835e-01 5.23496687e-01 -5.74810982e-01 3.97309601e-01 5.17643332e-01 -5.15214860e-01 -1.24840207e-01 -5.24400890e-01 -2.96558022e-01 -8.45000327e-01 -7.20007241e-01 -2.86648363e-01 1.33795404e+00 -2.65347719e-01 4.22721654e-02 -9.55112338e-01 -5.49583852e-01 -2.49101758e-01 1.40957510e+00 -6.19464517e-01 -4.18798178e-01 -6.75736725e-01 -1.46464145e+00 2.47917399e-01 2.74013907e-01 6.95526659e-01 -1.32361817e+00 -1.23287284e+00 5.41338742e-01 -2.79316336e-01 -7.96813965e-01 -4.09002155e-02 8.14137578e-01 -1.02316201e+00 -1.00705230e+00 7.09382677e-03 -5.40054560e-01 3.68368030e-01 -8.90254676e-02 1.11016011e+00 -1.23039104e-01 -4.50135976e-01 5.42049229e-01 -3.59967351e-01 -6.04999244e-01 -5.99117279e-01 5.04805744e-01 3.32372844e-01 3.60583104e-02 1.01643646e+00 -1.48457015e+00 -5.83761811e-01 2.37922594e-01 -7.48027921e-01 -2.56621987e-01 2.47990742e-01 2.24226400e-01 4.50561732e-01 6.77224100e-01 9.44933593e-01 -8.55664551e-01 5.06752729e-01 -6.73181713e-01 -3.64211291e-01 4.38405275e-01 -1.12234271e+00 -1.66964501e-01 4.46264029e-01 -7.42241919e-01 -1.04660511e+00 -1.01331212e-02 -2.06722900e-01 7.23150522e-02 -8.69702041e-01 -2.38676682e-01 -5.45839310e-01 3.91773254e-01 2.74179220e-01 1.06839135e-01 -7.98032284e-01 -8.29382539e-01 1.65644735e-01 1.18362474e+00 3.15301061e-01 -5.84441721e-01 4.06270921e-01 6.11351609e-01 2.00068429e-02 -8.24955285e-01 -8.10606122e-01 -6.89173400e-01 -6.65186763e-01 -3.14973176e-01 1.01206648e+00 -2.55543202e-01 -1.49657416e+00 3.57749015e-01 -7.10496306e-01 -3.87160003e-01 -4.79454994e-01 1.71760783e-01 -7.71698415e-01 2.98398614e-01 -3.08335006e-01 -1.65744030e+00 -7.50390828e-01 -7.30731606e-01 7.05459833e-01 2.28986397e-01 -6.21457100e-01 -9.22798514e-01 3.50331932e-01 6.35505736e-01 3.73729616e-01 3.27843457e-01 1.21052885e+00 -9.83611405e-01 7.32925069e-03 -4.00813818e-01 2.95344353e-01 5.14858723e-01 7.57868648e-01 -5.99083066e-01 -1.75452924e+00 -5.93930066e-01 8.87404203e-01 -2.26399943e-01 3.80670160e-01 3.75879824e-01 1.26461303e+00 -5.20296097e-01 -4.78823811e-01 5.90626150e-02 1.74386835e+00 6.85547292e-01 4.79751080e-01 4.49608326e-01 4.79097486e-01 3.38887841e-01 1.28853358e-02 4.77727026e-01 4.15094852e-01 5.51616311e-01 6.88190699e-01 1.03110127e-01 3.27818483e-01 -1.59134611e-03 2.09004790e-01 1.00183845e+00 5.31974882e-02 -5.32676697e-01 -3.59270215e-01 4.36371505e-01 -2.19892383e+00 -1.05659699e+00 1.80656597e-01 2.03957033e+00 8.51271093e-01 1.65158689e-01 3.38366300e-01 8.17030728e-01 6.79845333e-01 1.47642970e-01 -9.06735778e-01 -5.33257663e-01 4.70970154e-01 1.34154543e-01 6.90793991e-01 2.72153288e-01 -1.10422802e+00 1.51907936e-01 6.30730486e+00 8.04174185e-01 -7.10198760e-01 8.46386731e-01 6.94224834e-01 -3.33375543e-01 3.19825500e-01 -5.53552091e-01 -7.41665900e-01 8.11861515e-01 1.71620774e+00 1.45779774e-01 5.30456960e-01 1.15371430e+00 3.44148785e-01 -7.57464230e-01 -1.43883467e+00 1.19495988e+00 1.54445216e-01 -3.43104720e-01 -5.23585975e-01 3.85823786e-01 4.64292973e-01 4.72161695e-02 -3.15661430e-01 2.03653827e-01 4.60377306e-01 -8.57796490e-01 2.95168102e-01 6.73485041e-01 3.03635210e-01 -9.87133145e-01 5.71429312e-01 8.48702133e-01 -1.19201040e+00 -2.47840494e-01 -2.35288113e-01 3.71758528e-02 3.95137101e-01 6.21675491e-01 -5.66993594e-01 4.98973906e-01 9.39040303e-01 -2.47372203e-02 -3.75553459e-01 5.34030735e-01 2.10624397e-01 9.22263443e-01 -9.13493872e-01 1.40624985e-01 -1.82034910e-01 -3.45101297e-01 -1.29845291e-01 1.01960683e+00 1.39619902e-01 1.60746530e-01 3.80702168e-01 3.10243845e-01 1.43309787e-01 2.47480553e-02 -4.82904524e-01 4.95910168e-01 2.76899278e-01 1.28066349e+00 -8.52807999e-01 -4.85996038e-01 -4.28691834e-01 1.38147593e+00 1.86175704e-01 -1.78209245e-01 -8.31588209e-01 2.66861379e-01 5.68276882e-01 -1.57214496e-02 -1.02505192e-01 1.94264397e-01 -1.52503356e-01 -5.56218028e-01 -5.91983274e-02 -3.07459623e-01 7.08521962e-01 -4.38372612e-01 -1.45789814e+00 2.26226360e-01 3.32518160e-01 -4.88813847e-01 -3.04164231e-01 -6.58254772e-02 -4.25152838e-01 4.30201203e-01 -1.35541642e+00 -1.26867855e+00 -5.98319136e-02 7.15197742e-01 6.46433353e-01 1.89709112e-01 1.52733958e+00 7.66413093e-01 -8.03301871e-01 3.26512218e-01 1.64337888e-01 -2.43690565e-01 4.43946511e-01 -1.16947091e+00 -1.11638293e-01 2.59827703e-01 1.38866857e-01 -1.49548620e-01 9.19457495e-01 -6.75073504e-01 -8.35867345e-01 -1.16986513e+00 5.72353065e-01 -6.24332249e-01 1.83944881e-01 -3.44700307e-01 -7.81202555e-01 9.21687305e-01 3.56876552e-01 -2.16467395e-01 1.30697739e+00 1.85855087e-02 7.08337054e-02 -3.99005204e-01 -1.87382150e+00 -8.17175955e-02 6.41895533e-01 -5.45223534e-01 -7.21870661e-01 6.35811925e-01 1.17829658e-01 4.85794127e-01 -8.14866960e-01 3.40084821e-01 4.15670276e-01 -9.50044632e-01 8.24965298e-01 -2.93085575e-01 -7.12010920e-01 -9.62898508e-02 -7.55315959e-01 -1.10229731e+00 -3.13333213e-01 -1.73716962e-01 -7.50532508e-01 1.73978651e+00 -1.03036545e-01 -6.82250261e-01 6.49104774e-01 1.04927909e+00 2.68678904e-01 -4.43108231e-01 -1.37617445e+00 -7.73083389e-01 -2.19296694e-01 -6.99149609e-01 8.13148856e-01 9.51434612e-01 1.60008147e-01 3.48119467e-01 -6.59869850e-01 4.01867926e-01 7.89492965e-01 -2.77576208e-01 3.34981009e-02 -1.57887185e+00 -1.56035289e-01 -1.16867065e-01 -2.66485989e-01 -2.38420084e-01 1.83021992e-01 -6.71223164e-01 2.51456611e-02 -1.59632802e+00 3.95138741e-01 -2.76609570e-01 -8.87943804e-01 6.41454995e-01 4.49195385e-01 4.23590243e-01 -1.91107780e-01 2.59141833e-01 -6.04631782e-01 2.88790733e-01 2.46340021e-01 -2.30562001e-01 -2.98768491e-01 2.89416432e-01 -3.72496575e-01 1.00926232e+00 1.37202609e+00 -7.72782087e-01 -4.91801858e-01 1.59992918e-01 -2.80717500e-02 -3.52837592e-01 5.75187504e-02 -1.25039411e+00 1.16825595e-01 -2.43721604e-01 5.85648060e-01 -9.71254945e-01 6.56634808e-01 -1.65939701e+00 7.24058867e-01 4.96873319e-01 6.73469678e-02 8.25810432e-02 -1.43766263e-02 4.44969863e-01 6.66823506e-01 -4.49843466e-01 1.08530354e+00 -3.80668014e-01 -3.63282382e-01 3.26519762e-03 -5.56881309e-01 -6.38354659e-01 1.44656920e+00 -4.84827831e-02 -7.89979175e-02 -5.50552309e-02 -7.84257531e-01 -7.73754194e-02 6.72707930e-02 4.23296154e-01 -6.73517436e-02 -1.28480422e+00 -8.21400732e-02 1.34587973e-01 -1.88011646e-01 -2.06351191e-01 2.55189985e-01 4.75251675e-01 2.16487348e-01 4.07682776e-01 -9.86671820e-03 -3.71975690e-01 -1.37836742e+00 1.11223567e+00 3.80777061e-01 -5.67300916e-01 -6.05116308e-01 3.59516948e-01 -3.79665315e-01 -3.98705564e-02 4.14388597e-01 -4.17595863e-01 -2.54651934e-01 4.87030089e-01 5.75959802e-01 8.91377628e-01 3.87486339e-01 -6.67210400e-01 -4.90205616e-01 5.37147999e-01 -5.13820797e-02 -2.33407095e-02 1.45063245e+00 -5.48109412e-01 -2.02201590e-01 1.21220577e+00 1.18780315e+00 -5.13040841e-01 -4.01942760e-01 1.21855691e-01 3.67356151e-01 -3.88242826e-02 2.37374678e-01 -1.01736152e+00 -9.05483007e-01 5.39043605e-01 1.49616516e+00 8.27462614e-01 1.40410173e+00 3.32375616e-02 8.55621636e-01 3.89594197e-01 8.29477429e-01 -1.64999092e+00 -4.55423951e-01 -4.62061942e-01 4.15360570e-01 -1.36741090e+00 1.25750512e-01 2.11005688e-01 2.52428442e-01 6.66471303e-01 4.08775598e-01 1.61441967e-01 1.32803977e+00 4.08216387e-01 -9.54419747e-02 -2.26530265e-02 -4.83517349e-01 8.91951025e-02 -2.28925735e-01 7.26488352e-01 1.24492802e-01 4.35728073e-01 -3.50348167e-02 4.02648479e-01 -6.22145273e-02 7.74031207e-02 1.12334028e-01 1.21240306e+00 -6.34309769e-01 -1.32588363e+00 -6.03612542e-01 7.12562799e-01 -5.04454076e-01 4.67420042e-01 1.83236212e-01 3.99426669e-01 3.93164665e-01 1.54311764e+00 -4.17272970e-02 -2.69173175e-01 3.99694383e-01 7.65606880e-01 4.58401740e-01 -5.15152693e-01 -4.07802612e-01 3.40163745e-02 -1.02802709e-01 -3.84065598e-01 -8.12658489e-01 -6.42731369e-01 -1.09085834e+00 -4.49596763e-01 -4.33113903e-01 9.06529576e-02 9.20368135e-01 1.12484443e+00 -1.24063998e-01 7.56786942e-01 6.77707076e-01 -1.07956040e+00 -4.88042414e-01 -1.41527677e+00 -9.37475204e-01 5.01253784e-01 4.02845830e-01 -4.96854484e-01 -3.92688036e-01 3.40426862e-02]

[6.000402927398682, 2.5758440494537354]

4e796be2-5e6f-40c6-bb1e-b0ee3f949146

cross-lingual-fine-tuning-for-abstractive

https://aclanthology.org/2021.ranlp-main.74

https://aclanthology.org/2021.ranlp-main.74.pdf

Cross-lingual Fine-tuning for Abstractive Arabic Text Summarization

While abstractive summarization in certain languages, like English, has already reached fairly good results due to the availability of trend-setting resources, like the CNN/Daily Mail dataset, and considerable progress in generative neural models, progress in abstractive summarization for Arabic, the fifth most-spoken language globally, is still in baby shoes. While some resources for extractive summarization have been available for some time, in this paper, we present the first corpus of human-written abstractive news summaries in Arabic, hoping to lay the foundation of this line of research for this important language. The dataset consists of more than 21 thousand items. We used this dataset to train a set of neural abstractive summarization systems for Arabic by fine-tuning pre-trained language models such as multilingual BERT, AraBERT, and multilingual BART-50. As the Arabic dataset is much smaller than e.g. the CNN/Daily Mail dataset, we also applied cross-lingual knowledge transfer to significantly improve the performance of our baseline systems. The setups included two M-BERT-based summarization models originally trained for Hungarian/English and a similar system based on M-BART-50 originally trained for Russian that were further fine-tuned for Arabic. Evaluation of the models was performed in terms of ROUGE, and a manual evaluation of fluency and adequacy of the models was also performed.

['Attila Novák', 'Zijian Győző Yang', 'Mram Kahla']

https://aclanthology.org/2021.ranlp-1.74

https://aclanthology.org/2021.ranlp-1.74.pdf

ranlp-2021-9

['extractive-summarization']

['natural-language-processing']

[-1.19530752e-01 4.29220855e-01 9.34647918e-02 -3.91216576e-01 -1.22103167e+00 -5.98437071e-01 7.75737226e-01 3.64195287e-01 -7.06728339e-01 9.39335167e-01 8.93741786e-01 -3.01417768e-01 2.53160089e-01 -5.25953412e-01 -6.33261144e-01 -2.66368866e-01 1.11784615e-01 9.95662272e-01 2.63397321e-02 -7.79680490e-01 3.45885336e-01 2.49664038e-01 -9.66444194e-01 5.41195512e-01 1.27586412e+00 3.68764579e-01 6.84883147e-02 9.36447799e-01 2.21711352e-01 8.05276096e-01 -1.04804575e+00 -7.11524844e-01 -2.87521631e-01 -9.19439197e-01 -9.88718450e-01 2.34195534e-02 4.12206769e-01 -5.65422773e-01 -2.21053839e-01 5.55077195e-01 8.50504994e-01 1.44092202e-01 8.27277780e-01 -5.99980772e-01 -9.68182623e-01 1.27972662e+00 -2.74311215e-01 2.82629102e-01 3.51920068e-01 -9.30368155e-03 9.40741479e-01 -8.24524462e-01 7.76460707e-01 1.30451322e+00 6.61651552e-01 5.25682151e-01 -7.84425616e-01 -1.81715980e-01 -1.50504678e-01 -5.46139218e-02 -9.96633351e-01 -7.92569101e-01 4.19487745e-01 -5.79081848e-03 1.53464139e+00 4.77265716e-02 7.29252160e-01 1.03109980e+00 2.29913592e-01 1.07724977e+00 6.16013587e-01 -6.73742175e-01 -4.65999823e-03 9.54719037e-02 1.08401045e-01 7.22604334e-01 3.05833846e-01 -6.79362297e-01 -4.54683304e-01 1.64829090e-01 4.80488449e-01 -7.74994373e-01 -2.14851350e-01 4.17719930e-01 -1.32345188e+00 1.11931598e+00 2.95163006e-01 4.20811415e-01 -3.82469714e-01 -1.60547525e-01 8.42029572e-01 3.80913883e-01 7.97616780e-01 7.24538684e-01 -4.55052793e-01 -3.22468549e-01 -1.32586908e+00 3.05309802e-01 1.32149863e+00 9.31676745e-01 2.66358435e-01 5.16996562e-01 -1.91310525e-01 1.15338612e+00 9.58857313e-02 5.25983989e-01 7.95784295e-01 -6.40374720e-01 8.43161821e-01 4.49704528e-01 -4.81781922e-02 -8.18894207e-01 -5.51561832e-01 -2.29320660e-01 -7.82836795e-01 -4.61196452e-01 4.06624794e-01 -7.08726883e-01 -1.03023171e+00 1.35187745e+00 -3.16025704e-01 -7.28636682e-01 5.12364805e-01 6.61096394e-01 1.05095708e+00 1.20874476e+00 -1.89410567e-01 -3.03966373e-01 1.25395453e+00 -1.15277255e+00 -6.99129939e-01 -2.74124235e-01 6.96453869e-01 -8.98759425e-01 8.73543799e-01 3.14231902e-01 -1.56841910e+00 -2.03840032e-01 -1.10931218e+00 -3.46454769e-01 -3.85746688e-01 3.68593246e-01 2.86086231e-01 5.52595735e-01 -1.32274663e+00 3.38633239e-01 -8.92535329e-01 -1.01648891e+00 1.50496051e-01 1.71117932e-01 -4.67576057e-01 -8.50197375e-02 -1.40175307e+00 1.48099637e+00 8.41205657e-01 1.75296724e-01 -7.41403580e-01 -8.24183226e-02 -1.00863814e+00 1.94879249e-02 1.08258769e-01 -6.12507343e-01 1.49538541e+00 -1.06585848e+00 -1.79659975e+00 7.78783083e-01 1.84281066e-01 -7.50734687e-01 2.89947122e-01 -2.07425922e-01 -5.39053023e-01 2.83907205e-01 5.00310063e-02 8.60791385e-01 5.77418685e-01 -9.67012346e-01 -6.12695396e-01 -2.83736378e-01 -7.08417520e-02 5.70241511e-01 -3.06622982e-01 3.96990955e-01 -4.90363479e-01 -7.57085204e-01 -2.69929916e-01 -8.36336255e-01 -3.88744101e-02 -1.08424580e+00 -5.69877803e-01 -2.31469423e-01 4.17278141e-01 -1.45341384e+00 1.33457971e+00 -1.51809096e+00 3.66774797e-01 -1.33901566e-01 -3.77167732e-01 4.87271905e-01 -3.50324303e-01 1.02373672e+00 2.48412549e-01 3.36577632e-02 -4.34292525e-01 -3.65462184e-01 -7.66068622e-02 2.25098014e-01 -2.00379118e-01 2.22942278e-01 5.71364403e-01 1.07538295e+00 -9.15189028e-01 -4.43588763e-01 -2.58280396e-01 3.94301921e-01 -5.36365569e-01 -6.07841741e-03 -3.37023616e-01 1.91228002e-01 -1.33749485e-01 5.07497489e-01 8.22407305e-02 2.72845417e-01 2.99444441e-02 7.88078923e-03 -2.06450105e-01 5.36472082e-01 -5.64380527e-01 1.71197128e+00 -4.78595704e-01 7.05712855e-01 -2.20580734e-02 -1.01690221e+00 7.82915711e-01 5.38843811e-01 5.24021722e-02 -5.76805592e-01 3.47664326e-01 4.88306254e-01 1.22572064e-01 -3.67256165e-01 1.30978215e+00 5.82707264e-02 -4.34884548e-01 6.78635955e-01 4.46878850e-01 -6.78622186e-01 8.69400561e-01 4.08613533e-01 8.11367154e-01 1.31204978e-01 4.43982422e-01 -2.24517927e-01 3.82055104e-01 4.31427211e-01 1.11532748e-01 5.76528907e-01 1.77870870e-01 8.81353080e-01 4.85088587e-01 -8.74952599e-02 -1.19268715e+00 -6.87092245e-01 1.28773540e-01 1.19088554e+00 -5.22212446e-01 -6.00471318e-01 -1.27460063e+00 -5.88730752e-01 -4.52181548e-01 1.23674428e+00 -2.81774104e-01 -1.43105648e-02 -8.64097357e-01 -9.69700396e-01 1.07332170e+00 4.06475753e-01 5.63437581e-01 -1.63721478e+00 -4.77722496e-01 4.96512830e-01 -3.92986715e-01 -8.89705479e-01 -3.86171639e-01 1.75384730e-01 -8.21115971e-01 -5.88414967e-01 -1.08590055e+00 -9.39099789e-01 2.69710451e-01 -3.04604679e-01 1.36310923e+00 -3.04767758e-01 2.32639074e-01 3.02371651e-01 -5.78342915e-01 -7.81056523e-01 -1.05182099e+00 8.25883090e-01 -1.05615072e-01 -4.54234451e-01 1.47411421e-01 -2.66881678e-02 -1.05286166e-01 -3.00726980e-01 -1.20413578e+00 7.30395466e-02 8.26581478e-01 8.42834890e-01 1.12361230e-01 -3.39246809e-01 1.08478892e+00 -9.72597837e-01 1.16494381e+00 -4.22279328e-01 8.30479246e-03 1.78023890e-01 -2.33200341e-01 -3.07929423e-02 5.70873857e-01 -1.04722776e-01 -1.18716240e+00 -3.36495370e-01 -5.70709527e-01 4.36520308e-01 3.27119441e-03 1.21401215e+00 -4.60334942e-02 6.47728682e-01 8.98372173e-01 1.86668694e-01 3.32680997e-04 -3.77113789e-01 6.76262140e-01 1.14720070e+00 7.83251166e-01 -2.71534592e-01 2.40493417e-01 -2.11616442e-01 -7.54850805e-01 -1.39313459e+00 -8.07793796e-01 -8.81382823e-02 -6.16321385e-01 3.66823263e-02 7.88657725e-01 -9.89824235e-01 4.52987663e-02 7.39291549e-01 -1.25444472e+00 -5.24328649e-01 -3.31900477e-01 3.10221881e-01 -5.88459015e-01 3.20056051e-01 -1.08236444e+00 -3.42602700e-01 -8.31331789e-01 -9.97668743e-01 9.55295444e-01 5.45210615e-02 -5.69359839e-01 -1.16811585e+00 2.73416579e-01 2.42005482e-01 5.19147694e-01 9.70718339e-02 1.08474064e+00 -1.12631047e+00 1.92108229e-01 -4.62704271e-01 -2.96939593e-02 7.21678615e-01 4.12329957e-02 1.76250383e-01 -5.10020614e-01 -3.28316420e-01 -1.68649763e-01 -7.05750704e-01 1.02627754e+00 4.17023569e-01 3.81431818e-01 -5.53447545e-01 1.57544434e-01 -8.18122998e-02 8.81022751e-01 -3.80945094e-02 7.82423258e-01 4.23446983e-01 5.58046043e-01 5.46812534e-01 4.26826209e-01 2.29717821e-01 7.98790276e-01 3.36569965e-01 4.60094437e-02 6.33693859e-02 -3.04851294e-01 -8.86444151e-02 7.89101243e-01 1.69796908e+00 -1.89927995e-01 -6.90934837e-01 -9.36144650e-01 7.72089064e-01 -1.69591010e+00 -8.67062271e-01 5.08423448e-02 1.85484171e+00 1.23750603e+00 4.82525639e-02 2.58721799e-01 -1.55097440e-01 4.71052021e-01 2.17609927e-01 -1.65751487e-01 -9.98148263e-01 -3.70253980e-01 1.21579543e-01 2.09162131e-01 5.63391030e-01 -9.89199996e-01 1.33450449e+00 6.30380297e+00 6.71430409e-01 -1.05688190e+00 -1.43395923e-02 6.43259048e-01 -5.23520000e-02 -1.13900013e-01 -2.52020985e-01 -8.09295952e-01 2.57271796e-01 1.52662396e+00 -1.04058623e-01 2.77850598e-01 4.51353133e-01 2.04804391e-01 -3.30123574e-01 -9.55600142e-01 4.46886420e-01 7.16385961e-01 -1.22777891e+00 3.39936942e-01 -2.71153927e-01 9.10974801e-01 4.26262528e-01 -2.68058777e-01 6.26549542e-01 5.12245119e-01 -9.84358668e-01 8.30644727e-01 3.78060699e-01 7.60741055e-01 -1.02230656e+00 9.82285500e-01 5.66638947e-01 -3.66214931e-01 3.07172298e-01 -4.88047004e-01 1.13459684e-01 2.87367195e-01 3.38047594e-01 -1.02342868e+00 7.40459085e-01 3.09719384e-01 7.75188565e-01 -7.67644763e-01 7.22431242e-01 -4.57283080e-01 8.86673748e-01 -2.28307471e-01 -2.31028408e-01 6.33930862e-01 -1.60992458e-01 6.49328113e-01 1.80951333e+00 2.80585498e-01 -7.93303624e-02 -7.81325623e-03 1.33454040e-01 -3.38573843e-01 5.30255556e-01 -4.75695014e-01 -4.64039594e-01 1.00520246e-01 1.27664530e+00 -6.97183549e-01 -4.94356751e-01 -3.06581765e-01 1.12329316e+00 4.02938575e-01 2.78321624e-01 -6.19010091e-01 -7.36513853e-01 -1.99939460e-01 -1.27364352e-01 2.07993612e-01 -3.85001332e-01 -8.20483416e-02 -1.33291626e+00 -3.63547236e-01 -1.30923104e+00 3.19753617e-01 -7.93145716e-01 -1.05615318e+00 1.01314294e+00 7.43289739e-02 -6.01945579e-01 -8.61973643e-01 -4.21176463e-01 -5.56071341e-01 9.04216766e-01 -1.15582836e+00 -1.43897271e+00 1.37970254e-01 2.55986869e-01 1.13028765e+00 -5.57703853e-01 1.06649327e+00 3.34946648e-03 -7.01551318e-01 4.86055195e-01 3.34617466e-01 2.83518851e-01 1.15935791e+00 -1.25172246e+00 5.89402080e-01 8.41299176e-01 1.61855802e-01 4.82428640e-01 8.02407205e-01 -7.46519387e-01 -1.16649902e+00 -1.05811632e+00 1.19794476e+00 -4.50332075e-01 7.56808102e-01 -2.05928266e-01 -8.65258396e-01 9.68050063e-01 1.06547499e+00 -9.24841046e-01 4.99632180e-01 -5.76292612e-02 4.69242670e-02 2.66936809e-01 -8.25343966e-01 6.94991887e-01 4.11919385e-01 -1.91335022e-01 -1.02177048e+00 4.84003067e-01 6.74067736e-01 -5.63965619e-01 -8.72068822e-01 1.21213518e-01 3.99989992e-01 -6.35747254e-01 4.71770257e-01 -6.22856081e-01 8.80296648e-01 1.86304748e-01 -4.15096097e-02 -2.22101712e+00 -2.75145359e-02 -6.16212547e-01 2.43925098e-02 1.55601859e+00 7.25303948e-01 -4.69966143e-01 3.75777543e-01 2.51401146e-03 -6.30174696e-01 -5.36961436e-01 -5.46436429e-01 -3.89178783e-01 5.69582582e-01 -5.26929423e-02 2.47202173e-01 6.96253538e-01 2.12871477e-01 1.04334378e+00 -1.77876130e-01 -3.45340908e-01 -2.01831236e-02 -2.50530422e-01 6.75439835e-01 -9.69346523e-01 1.22695602e-02 -6.27887547e-01 -4.07154001e-02 -6.35182679e-01 3.42785776e-01 -1.14208901e+00 2.90726483e-01 -2.18975878e+00 7.09778890e-02 7.57887736e-02 3.14423412e-01 6.85304224e-01 -1.13515705e-01 4.87453908e-01 1.29279166e-01 1.52517915e-01 -5.04477799e-01 5.56890249e-01 9.76158738e-01 -8.64057243e-02 -4.15257871e-01 -1.16298646e-01 -8.82693529e-01 6.09392762e-01 9.73576307e-01 -1.33988991e-01 -2.06277668e-01 -6.96287990e-01 2.79132873e-01 9.02779102e-02 -3.08087528e-01 -8.72866392e-01 1.84905410e-01 4.41061795e-01 4.35583770e-01 -7.60104120e-01 1.53464332e-01 -6.61266893e-02 -3.81385297e-01 2.63904482e-01 -4.72674429e-01 3.20743442e-01 4.68799293e-01 2.89894436e-02 -4.58915770e-01 -4.39419061e-01 6.86736405e-01 -2.57672787e-01 -5.17547905e-01 -1.65359080e-01 -9.43059623e-01 3.59758466e-01 5.06685257e-01 -4.62970585e-02 -5.47171652e-01 -8.68982315e-01 -5.69099784e-01 2.63592809e-01 3.04554731e-01 3.45879376e-01 5.05790412e-01 -8.81270170e-01 -1.52652729e+00 -1.25861257e-01 -4.02550325e-02 -2.34571472e-02 -1.59974441e-01 9.00143564e-01 -9.49570775e-01 7.72818625e-01 -5.30282915e-01 -2.12130964e-01 -8.25098693e-01 1.38625354e-01 2.36971509e-02 -5.25870204e-01 -5.36546409e-01 4.42125171e-01 -3.71801555e-01 -6.52599454e-01 -2.80104224e-02 -2.37507969e-01 -5.42202353e-01 6.03252411e-01 4.59525585e-01 6.02701724e-01 5.01224816e-01 -9.65374231e-01 3.13491002e-02 -4.84859943e-02 -3.32157582e-01 -6.53044522e-01 1.54595304e+00 -8.53772908e-02 -4.19346869e-01 5.72411716e-01 9.08885121e-01 2.10360393e-01 -7.22713947e-01 3.28627303e-02 1.14965886e-01 4.98140693e-01 -7.92725459e-02 -1.13602853e+00 -6.55309260e-01 7.67131507e-01 -1.76977947e-01 2.43585333e-01 9.80604649e-01 -1.70119703e-01 8.18970978e-01 7.10604548e-01 -2.54010521e-02 -1.23958766e+00 1.14647206e-03 1.30335820e+00 1.32276917e+00 -1.05000532e+00 4.16544117e-02 1.79451391e-01 -1.06783783e+00 1.22803295e+00 2.68703222e-01 -2.47977838e-01 4.75540608e-02 -1.01640765e-02 1.09590977e-01 -1.43399596e-01 -4.48476553e-01 -5.95806874e-02 3.16208392e-01 4.60329652e-01 8.91924322e-01 4.84267287e-02 -5.23372054e-01 6.36985481e-01 -7.26370215e-01 -1.70639917e-01 1.01287925e+00 8.30143154e-01 -6.42208338e-01 -8.38979363e-01 -2.68399596e-01 5.67921877e-01 -7.82629550e-01 -3.55089456e-01 -6.70991123e-01 1.16250813e+00 -5.36652267e-01 1.02274990e+00 1.33032933e-01 1.52986050e-01 5.23037732e-01 3.48261118e-01 6.35266602e-01 -9.45467889e-01 -8.85308325e-01 2.06975266e-01 6.74738050e-01 2.18041763e-02 -2.57998586e-01 -8.14291656e-01 -1.18215704e+00 -2.81719893e-01 -2.37540007e-01 3.19216579e-01 7.90447056e-01 9.75144207e-01 3.08097955e-02 5.29457450e-01 4.33197431e-02 -1.23322678e+00 -5.59469819e-01 -1.70427966e+00 -5.49710035e-01 -8.91442001e-02 -3.88064943e-02 1.72038659e-01 5.83784096e-02 1.84011668e-01]

[12.023859977722168, 9.543588638305664]

b89e7411-a83c-4ffb-aa75-665d4416f7c2

tdbscan-spatiotemporal-density-clustering

https://online-journals.org/index.php/i-joe/article/view/3881/0

https://online-journals.org/index.php/i-joe/article/view/3881/3315

TDBSCAN: Spatiotemporal Density Clustering

Trajectory data generated from personal or vehicle use of GPS devices can be utilized for travel analysis and traffic information service, whereas trip segmentation is a key step toward the semantic labelling of the trajectories. Two issues are difficult to deal with by the traditional density-based algorithms, i. e. multiple stops at the same spatial location with different visit times and non-consecutive point sequence for stop definition due to signal drifting. This article aims to develop a modified density-based clustering algorithm, named T-DBSCAN, by considering the time-sequential characteristics of the GPS points along a trajectory. Two new premises (i.e. state continuity within a single stop and temporal disjuncture among stops) were proposed as a theoretical basis for regulating the trajectory point selection in clustering. An empirical test was performed using a GPS-based personal travel dataset collected in the city of Shanghai to compare T-DBSCAN against DBSCAN. The results indicated that T-DBSCAN effectively improved both accuracy and computational speed in trajectory segmentation.

['Ji M', 'Chen W', 'Wang J']

2014-01-01

international-journal-of-online-and

['unsupervised-spatial-clustering']

['time-series']

[-3.43276799e-01 -4.71911490e-01 -4.03127611e-01 -4.61407840e-01 -6.57618225e-01 -5.61713576e-01 4.21890944e-01 4.74010438e-01 -3.59796911e-01 6.84681833e-01 2.29912847e-01 -5.49980164e-01 -7.27123618e-01 -1.20869553e+00 -4.84136343e-01 -8.65805089e-01 -2.15327278e-01 8.02582562e-01 6.01531327e-01 3.39931957e-02 3.19513768e-01 7.33120084e-01 -1.66999543e+00 1.34927168e-01 1.43224192e+00 5.12233078e-01 1.88344106e-01 1.93649441e-01 -6.39427423e-01 -6.45609433e-03 -3.60523105e-01 -2.95729302e-02 2.20566168e-02 -4.11672831e-01 -5.19606352e-01 2.29352750e-02 -5.44019282e-01 -4.85291295e-02 -2.89527267e-01 8.69286656e-01 2.04495162e-01 6.69037402e-01 7.25253701e-01 -1.65313840e+00 -1.82392120e-01 7.22727239e-01 -4.19651777e-01 4.95983988e-01 2.18710721e-01 -1.88862979e-01 4.76849556e-01 -3.89090657e-01 1.69582531e-01 1.04220021e+00 7.66007841e-01 5.62085807e-02 -1.00811327e+00 -4.38464642e-01 2.57559121e-01 4.79173690e-01 -2.14199758e+00 -2.36247648e-02 5.76870978e-01 -3.53083760e-01 8.03108871e-01 5.88415325e-01 8.02594006e-01 4.57276672e-01 -5.74170575e-02 6.48183942e-01 5.48806727e-01 -7.04978406e-02 4.25242215e-01 4.75666970e-02 2.41375610e-01 -2.32390150e-01 3.62253517e-01 -5.67654185e-02 1.26182720e-01 -3.93498272e-01 3.59162480e-01 4.00661588e-01 1.35700583e-01 -1.37976989e-01 -9.49577570e-01 6.97095692e-01 3.73100847e-01 5.23007691e-01 -5.33902109e-01 -2.11107865e-01 3.71760219e-01 -4.92048636e-02 3.44652236e-01 -4.17839736e-01 -2.32125878e-01 -4.57872421e-01 -1.27130163e+00 4.34835374e-01 4.96640116e-01 1.25194252e+00 7.85551310e-01 -1.47221014e-01 -7.66607821e-02 4.70421076e-01 3.12152594e-01 7.57105052e-01 3.03354800e-01 -5.89179158e-01 4.80377316e-01 9.24359918e-01 3.47010463e-01 -1.20091176e+00 -5.76255441e-01 6.80460798e-05 -7.26173043e-01 -4.59560663e-01 6.22547150e-01 -2.19293386e-01 -8.56909394e-01 1.22207201e+00 5.89967430e-01 5.25925279e-01 -2.64982164e-01 7.48913527e-01 4.53022182e-01 1.00515819e+00 2.67961770e-01 -3.35874349e-01 1.03995144e+00 -2.03494221e-01 -7.16400802e-01 4.93997395e-01 1.09041584e+00 -2.51776218e-01 6.07336342e-01 8.76404569e-02 -7.30395555e-01 -3.33989769e-01 -3.84472877e-01 4.64206666e-01 -6.00455105e-01 -1.50045797e-01 3.79292935e-01 1.01954913e+00 -7.38963664e-01 2.43040815e-01 -1.11352170e+00 -4.66504157e-01 2.04427168e-01 4.00446534e-01 2.03167330e-02 -5.62133491e-02 -1.19012356e+00 4.59816784e-01 4.34128225e-01 2.06662491e-02 -3.18727106e-01 -5.75111985e-01 -6.20371819e-01 -2.05067415e-02 1.22612871e-01 -1.81828663e-01 7.07870483e-01 -5.04082501e-01 -8.43054295e-01 5.07937431e-01 -5.55266142e-01 -4.34632301e-01 4.87274498e-01 2.56828725e-01 -1.17726457e+00 -1.64468065e-01 4.52954143e-01 4.94740270e-02 1.77856565e-01 -1.15668428e+00 -1.42908061e+00 -4.88932967e-01 -3.89984667e-01 1.40970707e-01 2.87267357e-01 -1.46560192e-01 -6.88205659e-01 -2.86522388e-01 3.35583329e-01 -1.03611064e+00 -3.37651610e-01 -1.11211729e+00 -4.10811216e-01 -6.60299480e-01 7.94754982e-01 -4.23681438e-01 2.04161024e+00 -2.04246926e+00 -4.64687765e-01 9.21816170e-01 -3.68741751e-01 2.00406089e-01 5.07156432e-01 8.19340765e-01 2.04876438e-01 -3.21162194e-02 -3.44990253e-01 1.71249166e-01 -2.65075527e-02 3.02198321e-01 -2.08896711e-01 7.04212785e-01 -3.47182244e-01 6.21890604e-01 -1.18797839e+00 -5.42610407e-01 6.03415489e-01 3.16811055e-01 -2.99844831e-01 -3.53781551e-01 1.30099356e-01 6.35108888e-01 -6.45646870e-01 6.83180869e-01 9.54830825e-01 4.40575629e-01 1.76283680e-02 2.06513554e-01 -7.84960449e-01 2.28991479e-01 -1.41742909e+00 1.14072788e+00 -3.46726067e-02 4.26859766e-01 -2.66469181e-01 -9.55699265e-01 1.10415030e+00 2.71772206e-01 9.68079984e-01 -8.56434643e-01 -1.50928590e-02 2.91354686e-01 -5.73096238e-02 -5.59168041e-01 8.54874194e-01 1.80423781e-01 -4.20892388e-01 2.52672672e-01 -6.03345990e-01 2.33528063e-01 3.72788757e-01 -1.43562987e-01 7.62736857e-01 -1.27014086e-01 -9.01730210e-02 -5.59570253e-01 4.33380514e-01 6.69839084e-01 7.66718507e-01 5.21486342e-01 -2.53463745e-01 4.31189626e-01 2.73161590e-01 -2.99523950e-01 -9.40440893e-01 -1.20225871e+00 -4.64361876e-01 6.65920436e-01 5.80527484e-01 -3.02021384e-01 -9.00421977e-01 -3.01137447e-01 1.15730613e-01 1.24452567e+00 -2.37248719e-01 2.36922801e-02 -6.68108821e-01 -9.63190675e-01 5.37185013e-01 2.19289720e-01 3.85456800e-01 -5.16011775e-01 -3.34269643e-01 5.31424403e-01 -2.51496077e-01 -6.58397317e-01 -4.31646556e-01 -2.02966928e-01 -8.91367078e-01 -1.01775730e+00 -4.65328246e-01 -5.47026694e-01 8.12872350e-01 7.04917908e-01 5.75727344e-01 -6.23624995e-02 3.93194526e-01 3.65510553e-01 -4.96677548e-01 -1.86296955e-01 -5.85586540e-02 4.08919841e-01 1.43338129e-01 2.81684697e-01 1.02523458e+00 -5.72741508e-01 -7.93518722e-01 1.06076741e+00 -8.58508706e-01 -3.68271261e-01 -5.45706227e-02 9.27921161e-02 7.87907660e-01 9.56528187e-01 8.74784470e-01 -5.74067056e-01 7.14437783e-01 -1.18852079e+00 -5.81939280e-01 1.24724492e-01 -5.18048406e-01 -5.15196443e-01 3.29679668e-01 -9.24300775e-02 -1.07828641e+00 1.08067825e-01 -4.02783096e-01 1.52145326e-02 -5.39495587e-01 4.74263400e-01 -5.46014488e-01 3.45827430e-01 2.51306504e-01 4.08544242e-01 -3.85587573e-01 -3.89942318e-01 3.70468527e-01 9.81264412e-01 3.18817228e-01 -4.08563435e-01 3.81032139e-01 6.47718012e-01 -2.19383314e-01 -9.76167262e-01 1.03763759e-01 -1.02951717e+00 -9.41334188e-01 -3.88495028e-01 7.30460048e-01 -7.84839034e-01 -9.07564580e-01 3.14257890e-01 -6.72516048e-01 -7.49543533e-02 -1.17431760e-01 7.27917969e-01 -4.45926189e-01 2.80687988e-01 -6.21211939e-02 -1.14257336e+00 2.56746769e-01 -9.92730856e-01 6.88334703e-01 4.31967407e-01 -3.13669801e-01 -1.29628432e+00 1.18343823e-01 -7.00944811e-02 2.98060570e-02 2.75515229e-01 7.45539963e-01 -7.56141126e-01 -5.31594753e-01 -5.19763172e-01 -5.83920144e-02 -3.21156293e-01 4.66036320e-01 2.05072492e-01 -4.75719780e-01 6.07832102e-03 -1.99924201e-01 7.92122245e-01 4.06971365e-01 8.14229846e-01 9.10831928e-01 -2.68596083e-01 -1.00130618e+00 4.63876486e-01 1.35393441e+00 1.10674787e+00 9.82135892e-01 4.84072834e-01 8.02634835e-01 7.87823379e-01 9.78481472e-01 4.12770480e-01 7.37491548e-01 7.78570354e-01 1.49761915e-01 1.99526235e-01 5.04800439e-01 -5.14358461e-01 -6.28582314e-02 6.41415834e-01 -9.14854482e-02 -3.79504830e-01 -1.23515952e+00 1.25834072e+00 -2.05860591e+00 -1.28272092e+00 -1.05389130e+00 2.56923676e+00 2.25667596e-01 3.74954380e-02 7.62456179e-01 3.94436896e-01 1.15833628e+00 -3.69592279e-01 -4.87751752e-01 -3.46265972e-01 6.71800077e-02 -5.02068162e-01 1.07753634e+00 6.16903663e-01 -8.71803641e-01 7.90316284e-01 6.14862823e+00 1.18637395e+00 -7.85565078e-01 2.33496930e-02 4.75844771e-01 6.23582974e-02 -5.98081887e-01 -4.38561663e-02 -1.01917422e+00 1.11669362e+00 1.24198413e+00 -3.68332803e-01 1.76165208e-01 7.07387507e-01 1.02033031e+00 -5.06470740e-01 -5.26587009e-01 1.01050556e+00 -6.10831916e-01 -1.22845805e+00 -6.06300905e-02 2.67824233e-01 7.98905909e-01 1.70431677e-02 -1.77230000e-01 7.64106512e-02 3.40227723e-01 -6.92899644e-01 4.96101111e-01 5.97591341e-01 4.61319208e-01 -1.26478863e+00 4.74115431e-01 5.96933365e-01 -1.51349449e+00 -1.59673039e-02 -5.15390821e-02 8.65964442e-02 8.13773155e-01 7.35176921e-01 -7.83940375e-01 8.96880925e-01 7.41173446e-01 5.56140780e-01 -4.96362932e-02 1.50656509e+00 2.04986647e-01 8.37890923e-01 -6.96252525e-01 2.32688133e-02 5.73934674e-01 -8.49029839e-01 5.79004765e-01 1.34122944e+00 8.02747011e-01 1.03579357e-01 -7.60356039e-02 6.81608915e-01 6.39291823e-01 -1.62616611e-01 -7.09455490e-01 4.34775621e-01 1.04670155e+00 7.74338365e-01 -1.26224434e+00 -1.76600799e-01 -2.16035798e-01 4.64410752e-01 -4.08994496e-01 3.96164954e-01 -1.22673166e+00 -4.47694331e-01 6.54657483e-01 7.45459199e-01 1.79951325e-01 -6.29137337e-01 -4.61671919e-01 -5.95178783e-01 -2.42780134e-01 -7.46447891e-02 3.98537964e-01 -2.22835109e-01 -9.38706577e-01 2.38920048e-01 5.39562821e-01 -1.42661405e+00 -2.38829017e-01 9.83680710e-02 -8.22554469e-01 8.89396191e-01 -1.08625317e+00 -7.48349965e-01 -3.82316634e-02 9.58713472e-01 2.28152737e-01 2.28197962e-01 3.76902580e-01 8.25061321e-01 -6.77754998e-01 3.09428304e-01 8.69116604e-01 -1.85731594e-02 1.47511270e-02 -9.88959730e-01 6.25159442e-01 7.44006753e-01 -2.45411277e-01 6.35269523e-01 6.23686314e-01 -1.10332286e+00 -1.11119425e+00 -1.26365900e+00 1.16086280e+00 -3.66869330e-01 5.95704496e-01 -8.67675096e-02 -9.22403038e-01 5.69749236e-01 -3.92561585e-01 -7.41558969e-01 7.92542040e-01 5.40309623e-02 6.82992160e-01 -1.21502228e-01 -1.43343055e+00 6.61573827e-01 1.05645633e+00 -3.25278401e-01 -4.40132052e-01 6.68976307e-02 4.42734599e-01 -1.33237794e-01 -8.06810558e-01 6.59120306e-02 2.94449091e-01 -8.96760464e-01 8.58405232e-01 -8.72361765e-04 -5.59769988e-01 -8.01113486e-01 -2.45405845e-02 -1.25441408e+00 -4.55989182e-01 -5.18908381e-01 3.75868618e-01 1.63306963e+00 2.05388248e-01 -7.67957151e-01 8.27913046e-01 9.98867214e-01 -3.76662254e-01 -3.46556753e-01 -1.46229100e+00 -9.42219794e-01 -1.30403966e-01 -9.11453426e-01 1.35076213e+00 7.57045567e-01 1.61852464e-01 -3.00333202e-01 -1.03672810e-01 4.17125016e-01 5.85276484e-01 -1.27727211e-01 7.29442179e-01 -1.09698951e+00 8.05951118e-01 -4.91904378e-01 -4.94597733e-01 -1.07789421e+00 -1.32510170e-01 -6.89777851e-01 -6.47686496e-02 -1.78657997e+00 -5.95893025e-01 -1.05916584e+00 1.07973844e-01 -1.96951836e-01 1.73543125e-01 -2.78232902e-01 -4.48245108e-01 5.15715122e-01 -3.68809491e-01 4.44845587e-01 7.53152668e-01 1.96384728e-01 -1.03013265e+00 8.19129586e-01 -3.35969567e-01 4.26632494e-01 9.11739826e-01 -5.98268628e-01 -7.26815999e-01 -1.37948781e-01 2.12086245e-01 1.16340503e-01 -2.32188590e-03 -9.11569595e-01 5.55206299e-01 -5.17023265e-01 -2.51529813e-01 -1.43778205e+00 -2.60386020e-02 -1.29557049e+00 8.77138138e-01 2.27814898e-01 1.89292118e-01 1.18766330e-01 2.17964888e-01 7.63328731e-01 -2.29789689e-01 -6.20882474e-02 1.77138627e-01 2.15914249e-01 -8.39334369e-01 3.25370669e-01 -8.64903092e-01 -3.12707990e-01 1.54212916e+00 -9.55615938e-01 -5.13869636e-02 -1.25345603e-01 -7.67594039e-01 5.94354391e-01 4.41940397e-01 2.42545635e-01 3.58997613e-01 -1.71637344e+00 -2.96769500e-01 9.57940668e-02 2.89173406e-02 1.39086828e-01 7.82826066e-01 1.13718545e+00 -6.06215715e-01 7.83283591e-01 -1.46873564e-01 -7.77428031e-01 -9.38988090e-01 6.57910287e-01 2.64358908e-01 2.00939968e-01 -7.28587985e-01 1.99603617e-01 -1.77674606e-01 -3.93365532e-01 3.48610207e-02 -5.79186320e-01 -3.28643262e-01 4.27752316e-01 5.23325920e-01 1.14218831e+00 1.06546901e-01 -1.15181184e+00 -5.39637029e-01 5.52020431e-01 4.42852557e-01 -2.27684245e-01 8.67321372e-01 -8.29735935e-01 1.20871328e-01 8.09924006e-01 7.85143256e-01 -1.55129448e-01 -1.17672741e+00 -1.67807192e-02 4.72989142e-01 -6.30867898e-01 -1.12420112e-01 -2.72512615e-01 -8.10833037e-01 2.27258280e-01 4.39682454e-01 9.31739986e-01 1.03221655e+00 -1.66871741e-01 1.11219096e+00 -1.18299343e-01 6.34062588e-01 -1.44444346e+00 -1.23927236e+00 3.17846835e-01 1.47500724e-01 -9.62053657e-01 -3.83742690e-01 -3.91912431e-01 -7.23112345e-01 7.59044528e-01 1.14621028e-01 -9.82721522e-02 1.04738951e+00 -1.55639514e-01 -4.52018142e-01 -7.56073594e-02 -5.79067096e-02 -5.66187978e-01 2.06359416e-01 8.84248018e-01 9.77481902e-02 5.46604991e-01 -6.31509840e-01 5.83710134e-01 -1.44810930e-01 1.66065559e-01 2.26271898e-01 8.83997381e-01 -6.11786067e-01 -9.32171226e-01 -6.56985164e-01 5.03217280e-01 -1.95234925e-01 3.34508151e-01 2.65540555e-02 8.27936471e-01 4.82887596e-01 1.41311264e+00 6.39495850e-01 -5.07359207e-01 5.45968592e-01 -1.85601637e-02 -8.45765024e-02 -1.83680862e-01 -4.61414903e-01 -3.81524302e-02 8.24455768e-02 -2.72537142e-01 -4.49982673e-01 -1.13282716e+00 -1.75288713e+00 -9.08721745e-01 -1.72193453e-01 6.69396281e-01 7.47354686e-01 8.80967677e-01 5.19622624e-01 1.70599848e-01 7.80078530e-01 -6.16555572e-01 3.85075718e-01 -4.90222901e-01 -1.06828082e+00 2.38921642e-01 2.09353492e-01 -5.62091887e-01 -2.94133723e-01 -4.84685935e-02]

[6.260828971862793, 1.783690094947815]

19c04f84-af54-434c-af07-f7546ff03fb1

crossspeech-speaker-independent-acoustic

2302.14370

https://arxiv.org/abs/2302.14370v2

https://arxiv.org/pdf/2302.14370v2.pdf

CrossSpeech: Speaker-independent Acoustic Representation for Cross-lingual Speech Synthesis

While recent text-to-speech (TTS) systems have made remarkable strides toward human-level quality, the performance of cross-lingual TTS lags behind that of intra-lingual TTS. This gap is mainly rooted from the speaker-language entanglement problem in cross-lingual TTS. In this paper, we propose CrossSpeech which improves the quality of cross-lingual speech by effectively disentangling speaker and language information in the level of acoustic feature space. Specifically, CrossSpeech decomposes the speech generation pipeline into the speaker-independent generator (SIG) and speaker-dependent generator (SDG). The SIG produces the speaker-independent acoustic representation which is not biased to specific speaker distributions. On the other hand, the SDG models speaker-dependent speech variation that characterizes speaker attributes. By handling each information separately, CrossSpeech can obtain disentangled speaker and language representations. From the experiments, we verify that CrossSpeech achieves significant improvements in cross-lingual TTS, especially in terms of speaker similarity to the target speaker.

['Byeong-Yeol Kim', 'Il-Hwan Kim', 'Yoon-Cheol Ju', 'Hong-Sun Yang', 'Ji-Hoon Kim']

2023-02-28

['speech-synthesis']

['speech']

[-1.98617175e-01 1.65144041e-01 -6.94105551e-02 -6.13299131e-01 -1.52298880e+00 -6.87239110e-01 8.20578754e-01 -2.41918027e-01 1.32281810e-01 2.67220825e-01 5.92606127e-01 -4.23765332e-01 3.04048091e-01 -2.11063281e-01 -4.74290729e-01 -8.93708408e-01 1.63411900e-01 5.58217883e-01 -2.81850636e-01 -1.80050477e-01 -5.15458941e-01 1.18115142e-01 -1.38122475e+00 1.48815051e-01 7.59981394e-01 6.19323075e-01 3.01216274e-01 7.43617117e-01 -2.25543842e-01 4.94148314e-01 -7.47479320e-01 -3.45037580e-01 -2.30494048e-02 -8.18046093e-01 -3.25548202e-01 -9.75927934e-02 3.77443820e-01 1.14508666e-01 -2.97543198e-01 1.02799892e+00 5.83587468e-01 -1.66640565e-01 6.50459826e-01 -1.35443902e+00 -6.36431932e-01 1.17164743e+00 -2.76440173e-01 -9.33359787e-02 6.29914552e-02 -4.26691864e-03 1.35354519e+00 -1.04454851e+00 1.27830952e-01 1.84056938e+00 1.66191459e-01 6.98229909e-01 -1.46149802e+00 -9.94719386e-01 1.98240411e-02 -1.59054369e-01 -1.42628074e+00 -1.13821614e+00 8.75756741e-01 -5.23425221e-01 7.08923221e-01 3.00778776e-01 1.92380041e-01 1.59399700e+00 9.81965065e-02 9.71005261e-01 9.75263536e-01 -4.72304791e-01 8.93292427e-02 2.05238506e-01 2.71104481e-02 3.68193477e-01 -9.78207290e-02 5.34053564e-01 -1.00774109e+00 1.14209935e-01 2.81649590e-01 -7.69075632e-01 -2.98941076e-01 -1.87844262e-01 -1.30667901e+00 7.69118011e-01 -8.85977373e-02 4.68674302e-01 -7.81165436e-03 1.08899660e-02 4.69086587e-01 2.36684442e-01 3.81771356e-01 1.29765451e-01 -4.03136522e-01 -1.95317104e-01 -1.02513814e+00 -4.91581298e-02 7.70476639e-01 1.07787752e+00 5.56578577e-01 7.14653850e-01 -1.07701249e-01 1.04755104e+00 7.90086389e-01 1.15356374e+00 6.28065467e-01 -4.98535782e-01 6.46565735e-01 -6.49121702e-02 -4.96712297e-01 -5.11899054e-01 4.70568277e-02 -7.09189951e-01 -8.55556488e-01 -1.46536911e-02 2.27354348e-01 -1.18920006e-01 -9.03019965e-01 2.35763335e+00 1.87937096e-01 2.17266213e-02 4.73681629e-01 6.27513111e-01 8.53238285e-01 8.40069115e-01 2.04775315e-02 -2.66112119e-01 1.52539897e+00 -1.02221453e+00 -1.03998899e+00 -4.70673978e-01 6.04140699e-01 -9.78616655e-01 1.05039990e+00 1.09133542e-01 -1.22214079e+00 -8.11114967e-01 -1.19104779e+00 -7.29463398e-02 -1.16406485e-01 1.36952817e-01 -1.12204991e-01 8.65973949e-01 -8.50801826e-01 -1.34226054e-01 -8.66335273e-01 1.95032969e-01 -1.75132260e-01 8.96028206e-02 -4.74110782e-01 1.54034317e-01 -1.54809630e+00 8.70244503e-01 2.24398419e-01 -1.19389102e-01 -1.04792404e+00 -7.60705054e-01 -1.16390407e+00 1.25826582e-01 9.99977216e-02 -5.51193535e-01 1.47760713e+00 -4.65625584e-01 -1.82703638e+00 5.67780614e-01 -7.18861103e-01 -4.28425223e-01 7.56691098e-02 -1.18892258e-02 -9.83816922e-01 -1.88125908e-01 2.13794053e-01 4.00620610e-01 1.03972638e+00 -1.37856209e+00 -4.91515905e-01 -4.52347040e-01 -8.12442064e-01 2.82342464e-01 -1.55293331e-01 1.12975053e-01 -3.72089595e-01 -6.89429700e-01 3.02535176e-01 -9.64128435e-01 3.38220328e-01 -7.75698364e-01 -7.33640969e-01 -3.30045938e-01 8.07952166e-01 -6.28214538e-01 1.14046478e+00 -2.42864656e+00 3.83021951e-01 -8.85816291e-02 4.54327203e-02 8.64385590e-02 -2.25390688e-01 6.46924734e-01 -2.27040425e-01 1.27607286e-01 -5.66134416e-03 -1.15008950e+00 2.93468535e-01 3.22650313e-01 -6.22362256e-01 2.74134845e-01 2.02697724e-01 9.12724614e-01 -7.69448340e-01 -6.14954889e-01 1.28540888e-01 5.78819871e-01 -2.75348991e-01 2.78173655e-01 -3.63936089e-02 7.92744398e-01 -2.38571912e-01 2.77819782e-01 5.73014557e-01 3.11847717e-01 2.42340297e-01 -2.40327194e-01 -1.09789103e-01 1.07993174e+00 -7.50007272e-01 1.72667313e+00 -6.61722124e-01 6.76510274e-01 3.83582473e-01 -6.34517133e-01 9.86787319e-01 7.56278455e-01 1.34003028e-01 -6.55516446e-01 -2.95539126e-02 3.33435684e-01 4.38808948e-01 -1.54468462e-01 4.76711690e-01 -6.39512539e-01 -3.59749764e-01 4.90547150e-01 4.12095398e-01 -4.75982398e-01 -1.99134529e-01 2.24396914e-01 5.17247498e-01 -1.11725740e-01 1.33296013e-01 -2.35736981e-01 4.09077257e-01 -6.25204742e-01 4.72689480e-01 4.90726709e-01 -8.62059668e-02 6.11280441e-01 1.61660790e-01 4.81419563e-01 -7.64927208e-01 -1.70760262e+00 -1.17013663e-01 1.07875919e+00 -1.74278766e-01 -5.96301317e-01 -8.56724083e-01 -4.23818797e-01 -1.36585861e-01 1.30959535e+00 -2.47302875e-01 -3.07351649e-01 -5.42648315e-01 -3.06151807e-01 1.08199334e+00 2.23872632e-01 5.02931625e-02 -7.51475573e-01 3.87382239e-01 1.10267937e-01 -3.65795314e-01 -1.32584751e+00 -8.99100244e-01 2.03491479e-01 -3.91512692e-01 -3.32149148e-01 -4.74383384e-01 -6.27023995e-01 1.33386448e-01 1.56016663e-01 9.48281348e-01 -8.58004808e-01 2.76950091e-01 5.93906306e-02 -2.18558714e-01 -4.13753510e-01 -1.21576321e+00 1.77739605e-01 5.66110075e-01 2.53699332e-01 2.65061110e-01 -6.58240974e-01 -1.54869910e-02 3.40949804e-01 -4.93228495e-01 1.37882113e-01 5.06770015e-01 7.07828641e-01 3.66605550e-01 1.16122231e-01 7.46885061e-01 -4.10372525e-01 5.52760422e-01 -4.28933710e-01 -4.84920651e-01 1.80783138e-01 -3.81635815e-01 4.68891263e-01 6.47170901e-01 -3.45466554e-01 -1.20545876e+00 -2.03256786e-01 -1.73704401e-01 -3.77904832e-01 -1.74854100e-01 5.42033434e-01 -7.03123152e-01 6.76602960e-01 4.53246713e-01 4.99658525e-01 9.37715098e-02 -6.08392477e-01 5.55168986e-01 9.85631049e-01 7.46938229e-01 -6.02670252e-01 9.09980834e-01 -6.33314997e-02 -3.13314855e-01 -1.11068785e+00 -9.31033254e-01 -3.16268951e-01 -4.49061990e-01 -7.13154150e-04 9.17953372e-01 -1.25107825e+00 -5.61138928e-01 4.74200517e-01 -1.22469699e+00 -9.04030912e-03 -3.27233434e-01 8.53124082e-01 -4.31618631e-01 1.34526551e-01 -3.80327642e-01 -1.06481719e+00 -5.18555641e-02 -1.33204794e+00 1.41394138e+00 -2.16088563e-01 -3.00860375e-01 -1.03287745e+00 2.99599111e-01 5.50181746e-01 2.45762035e-01 -3.16231549e-01 8.68932426e-01 -8.76425862e-01 -4.65059608e-01 -4.59622890e-02 2.84121424e-01 7.50988543e-01 3.67247581e-01 -1.87381774e-01 -1.41075146e+00 -3.15280706e-01 3.39124501e-01 -1.20927446e-01 5.57035685e-01 2.68186808e-01 4.74587947e-01 -1.92136660e-01 -5.25055900e-02 4.48288500e-01 6.06726706e-01 1.47245154e-01 3.44039798e-01 -5.19716322e-01 9.37368929e-01 8.15858006e-01 1.29930109e-01 -8.59209374e-02 6.11271143e-01 9.36929464e-01 -3.62591036e-02 2.04472765e-02 -6.88013732e-01 -5.99677622e-01 1.05121636e+00 1.87325490e+00 4.95760381e-01 -3.88375998e-01 -7.95543134e-01 4.63371038e-01 -1.45793521e+00 -9.25346792e-01 -2.60028869e-01 2.36765289e+00 9.93175924e-01 -4.97668311e-02 4.22220565e-02 8.74390751e-02 7.55806684e-01 2.50147074e-01 -4.48779732e-01 -3.28982741e-01 -4.11711514e-01 -2.19516978e-02 1.80443153e-01 8.81698370e-01 -6.37476206e-01 1.10203660e+00 6.11065769e+00 1.10540295e+00 -1.03921235e+00 2.02528700e-01 1.51384801e-01 -1.44927487e-01 -6.88899577e-01 -9.01248157e-02 -1.06746364e+00 3.56885195e-01 1.46563125e+00 -4.39617991e-01 4.43253696e-01 5.16278684e-01 2.57159173e-01 3.45283359e-01 -1.51972640e+00 1.01257551e+00 1.68434814e-01 -8.28420222e-01 4.83484305e-02 3.43914300e-01 4.44268465e-01 3.76212090e-01 3.63720089e-01 4.23919171e-01 4.84645307e-01 -1.09543657e+00 1.19763160e+00 5.58090843e-02 8.44607174e-01 -7.37416148e-01 2.46695489e-01 5.36747277e-01 -1.24920607e+00 3.46370012e-01 2.48587057e-01 4.47228491e-01 4.79292989e-01 5.06118119e-01 -1.24899971e+00 8.29168081e-01 2.55333334e-01 4.44010884e-01 -1.77552760e-01 1.55388907e-01 -3.95568699e-01 1.08652651e+00 -1.18442200e-01 2.84871191e-01 5.41846938e-02 -2.32409850e-01 9.86518741e-01 1.29012156e+00 5.02972245e-01 -3.72823298e-01 2.18579277e-01 1.04700613e+00 -4.98621613e-02 2.15566903e-02 -5.89812338e-01 -5.65609515e-01 8.51751447e-01 9.93751884e-01 1.47358375e-02 -3.41662019e-01 -1.75061226e-01 8.27545881e-01 1.35711357e-01 4.69854087e-01 -6.28754199e-01 -9.52468291e-02 1.03763032e+00 -2.26919353e-01 1.51390776e-01 -5.27646959e-01 -2.31616125e-01 -1.24280918e+00 -7.28138629e-03 -9.85710144e-01 -1.46302700e-01 -5.34481466e-01 -1.48337865e+00 8.76517951e-01 -6.75298274e-02 -1.10235310e+00 -8.88006747e-01 -3.83098036e-01 -2.81961560e-01 1.49073923e+00 -1.20149565e+00 -1.48110116e+00 4.44394350e-01 4.62724179e-01 6.90483212e-01 -4.21096861e-01 1.17077327e+00 2.89135039e-01 -6.39492869e-01 9.84075189e-01 8.41896608e-02 1.38686493e-01 8.35013807e-01 -1.13080645e+00 7.74528742e-01 8.87221575e-01 5.39570510e-01 7.87215471e-01 8.26931536e-01 -4.43146735e-01 -1.32112491e+00 -1.09133363e+00 1.30149853e+00 -5.71007907e-01 8.13315094e-01 -9.67826068e-01 -8.53982687e-01 6.34852111e-01 3.45988810e-01 -1.60689324e-01 9.20480251e-01 3.99905264e-01 -9.02100444e-01 -9.82611999e-02 -6.52201474e-01 7.33278692e-01 7.83437967e-01 -1.25361526e+00 -6.75555348e-01 -2.27698907e-02 1.17528665e+00 -1.69208601e-01 -6.38932049e-01 -6.18766360e-02 5.94545007e-01 -8.79850090e-01 9.04057264e-01 -4.33990151e-01 1.86821260e-02 -4.55175459e-01 -7.07415462e-01 -1.71241331e+00 -1.31651402e-01 -8.89553964e-01 1.08349852e-01 1.67193282e+00 6.17755473e-01 -7.43466079e-01 1.30227163e-01 2.08691254e-01 -6.37922168e-01 -1.26376569e-01 -1.20303631e+00 -1.22051132e+00 3.57663184e-01 -8.72153580e-01 8.43082130e-01 8.57614756e-01 1.45898253e-01 9.56494510e-01 -5.00711501e-01 5.53470194e-01 7.61531711e-01 9.55307484e-02 6.63917840e-01 -9.58136499e-01 -6.18477583e-01 -4.08973187e-01 -9.48228836e-02 -1.07627547e+00 5.51364660e-01 -1.26607764e+00 4.53137606e-01 -1.01231921e+00 -1.19352922e-01 -3.10569137e-01 -2.52605498e-01 2.21268237e-01 -1.18344603e-02 -1.56105116e-01 3.37825507e-01 -2.58303564e-02 -1.18325330e-01 9.00503814e-01 1.21941781e+00 3.27413641e-02 -1.22301243e-01 -2.99366154e-02 -6.97546363e-01 6.45383894e-01 7.21864343e-01 -6.00612760e-01 -5.03934264e-01 -4.59590018e-01 -3.74612600e-01 3.21075708e-01 7.56694674e-02 -7.58410692e-01 2.16201007e-01 2.44279169e-02 -2.68644154e-01 -6.19741678e-01 7.63704240e-01 -3.39132875e-01 6.37030005e-02 1.45425633e-01 -5.51071703e-01 -3.49735886e-01 2.13604458e-02 4.24421370e-01 -5.02424717e-01 1.48309350e-01 7.61107028e-01 2.92413592e-01 1.24165021e-01 1.66176915e-01 -4.89918470e-01 3.54126900e-01 4.78852153e-01 1.78594127e-01 -2.77761906e-01 -5.50782740e-01 -5.00547647e-01 -3.19069251e-02 -1.41950836e-02 8.25975418e-01 2.72063822e-01 -1.58368838e+00 -1.04990530e+00 6.67022288e-01 3.28439474e-01 -2.51372129e-01 3.19856435e-01 6.96321368e-01 5.82340777e-01 7.96364129e-01 4.20163214e-01 -7.32352197e-01 -1.30647182e+00 2.99253106e-01 2.45965764e-01 1.12674661e-01 -1.17689177e-01 1.03161716e+00 8.03262830e-01 -7.42794812e-01 1.86590657e-01 -2.60774106e-01 2.48876035e-01 1.21090643e-01 4.50057358e-01 1.73669338e-01 2.22808290e-02 -1.45413721e+00 -5.55071115e-01 3.47328633e-01 -4.06953692e-03 -9.71728981e-01 8.71716797e-01 -4.43652272e-01 3.03006619e-02 1.14504409e+00 1.57926202e+00 6.93278253e-01 -1.03467762e+00 -2.93119937e-01 -1.29720449e-01 -4.50407416e-02 2.50296205e-01 -8.72617662e-01 -7.71128714e-01 1.20016563e+00 1.94464192e-01 4.01175082e-01 6.34246528e-01 3.48853558e-01 9.54990745e-01 -1.07583240e-01 2.56457657e-01 -9.03436005e-01 -1.44487202e-01 6.57519400e-01 1.03635406e+00 -1.20584404e+00 -6.79596961e-01 -4.17864978e-01 -8.99823308e-01 8.07646692e-01 2.02247307e-01 3.39241534e-01 6.79361641e-01 4.96377110e-01 3.48044157e-01 -4.42418791e-02 -8.63561749e-01 -1.03942573e-01 7.33880937e-01 4.84694600e-01 6.65277243e-01 6.20900452e-01 4.71496940e-01 7.34906137e-01 -1.00919580e+00 -8.08441579e-01 1.54316232e-01 1.66406557e-01 -2.90764034e-01 -1.51399255e+00 -5.23029387e-01 -2.89489418e-01 -2.75154531e-01 -5.06773293e-01 -3.41547549e-01 4.08986151e-01 -8.31514001e-02 1.41221642e+00 3.05021014e-02 -5.27691603e-01 2.25441679e-01 5.09072423e-01 4.34158325e-01 -7.58906305e-01 -3.09505522e-01 4.82848763e-01 3.70464772e-01 -3.79965186e-01 -2.20417269e-02 -8.69540155e-01 -1.33170748e+00 -1.15040593e-01 -4.59042162e-01 5.11908770e-01 1.18148935e+00 1.08595335e+00 4.25648153e-01 7.46420264e-01 8.31511617e-01 -6.44808054e-01 -8.36617708e-01 -1.00400925e+00 -7.37277865e-01 1.41238526e-01 8.27837825e-01 -3.95065308e-01 -7.21972167e-01 1.16464145e-01]

[14.816173553466797, 6.651287078857422]

247a6194-3228-443a-b9a5-21af4ce01900

infrared-image-super-resolution-systematic

2212.12322

https://arxiv.org/abs/2212.12322v1

https://arxiv.org/pdf/2212.12322v1.pdf

Infrared Image Super-Resolution: Systematic Review, and Future Trends

Image Super-Resolution (SR) is essential for a wide range of computer vision and image processing tasks. Investigating infrared (IR) image (or thermal images) super-resolution is a continuing concern within the development of deep learning. This survey aims to provide a comprehensive perspective of IR image super-resolution, including its applications, hardware imaging system dilemmas, and taxonomy of image processing methodologies. In addition, the datasets and evaluation metrics in IR image super-resolution tasks are also discussed. Furthermore, the deficiencies in current technologies and possible promising directions for the community to explore are highlighted. To cope with the rapid development in this field, we intend to regularly update the relevant excellent work at \url{https://github.com/yongsongH/Infrared_Image_SR_Survey

['Shinichiro Omachi', 'Xiaofeng Liu', 'Tomo Miyazaki', 'Yongsong Huang']

2022-12-22

['infrared-image-super-resolution']

['computer-vision']

[ 8.95087838e-01 -3.30598921e-01 -1.75317839e-01 -1.44723356e-01 -9.59324300e-01 -7.61261582e-02 1.95220247e-01 -6.80460989e-01 -2.71257609e-01 6.79177105e-01 1.40555620e-01 1.64117306e-01 -1.36912152e-01 -4.97154504e-01 -3.50665182e-01 -1.03622210e+00 1.47604495e-01 -4.14954573e-01 -1.30710155e-01 -4.04301196e-01 5.32184660e-01 5.78814864e-01 -1.90158105e+00 6.03322208e-01 8.54319990e-01 9.46972907e-01 6.55817509e-01 7.03744650e-01 2.67614257e-02 6.76431239e-01 -4.10717398e-01 1.75610051e-01 2.57628113e-01 -3.44350994e-01 -1.05705416e+00 -1.43717349e-01 1.13782954e+00 -6.43027008e-01 -3.39612007e-01 1.40944600e+00 8.48751664e-01 2.81089723e-01 1.14079736e-01 -4.09329087e-01 -1.12335455e+00 4.92684603e-01 -9.81720924e-01 1.12039793e+00 1.34116426e-01 -1.68301046e-01 5.03629327e-01 -1.07885730e+00 4.35328066e-01 1.11732221e+00 4.81986165e-01 7.20492184e-01 -9.83907998e-01 -8.72964442e-01 -2.49541432e-01 3.72048646e-01 -1.29296255e+00 -4.92014766e-01 7.02608109e-01 -9.69655216e-02 7.18809366e-01 3.84527922e-01 -3.06657795e-02 9.93231118e-01 3.14312458e-01 4.07394081e-01 1.37836456e+00 -5.43771923e-01 -1.35220084e-02 1.00171804e-01 2.59853542e-01 2.12007448e-01 2.52346575e-01 4.07896161e-01 -6.67496085e-01 2.01057792e-01 1.47797763e+00 -2.10960224e-01 -1.84174806e-01 1.92607418e-01 -1.07747304e+00 5.02125502e-01 5.31209052e-01 7.26984560e-01 -1.37502491e-01 6.91048428e-02 1.45400643e-01 2.80779719e-01 7.25064099e-01 4.65948790e-01 -3.13749939e-01 3.68191570e-01 -7.77317166e-01 9.88810286e-02 -4.12585407e-01 7.56707370e-01 5.41599214e-01 5.38199246e-01 1.22280873e-01 1.23287213e+00 -1.99185848e-01 2.66563535e-01 3.05448562e-01 -1.49824452e+00 6.60613999e-02 3.52262706e-03 1.71827361e-01 -5.39323866e-01 -4.30840492e-01 -4.32161450e-01 -1.36625671e+00 7.77718604e-01 8.57013389e-02 1.23478305e-02 -7.47298539e-01 1.13437212e+00 1.58110917e-01 1.42269254e-01 6.10719919e-02 1.19588470e+00 1.36131740e+00 5.20709455e-01 6.31797165e-02 -5.15539408e-01 1.72456980e+00 -5.33822119e-01 -7.44430006e-01 -3.85072768e-01 -1.22378431e-01 -8.63779068e-01 7.30866730e-01 5.48104584e-01 -1.33277011e+00 -1.11032951e+00 -9.94513512e-01 -5.13345301e-01 -2.34383941e-02 2.97348976e-01 6.23701274e-01 3.28755885e-01 -1.37075901e+00 5.93223572e-01 -5.66659033e-01 -5.36228657e-01 5.68010867e-01 2.45158061e-01 -1.38395607e-01 -3.55153441e-01 -1.39335966e+00 8.92458200e-01 4.67272907e-01 1.17939830e-01 -2.62169898e-01 -9.14794147e-01 -5.26814282e-01 -4.45706397e-01 1.45221815e-01 -7.35635817e-01 1.24799240e+00 -9.78360295e-01 -1.38106418e+00 9.98062134e-01 -2.39430279e-01 -8.47896039e-02 -2.87384931e-02 -4.30700928e-01 -7.38884091e-01 4.62640822e-01 -1.25650361e-01 5.49149334e-01 1.05359328e+00 -1.49668038e+00 -9.27371204e-01 -6.76901817e-01 -1.69433221e-01 6.15241647e-01 -8.79286081e-02 8.11750174e-01 -1.03879645e-01 -4.64007944e-01 4.08948511e-02 -6.46729112e-01 -3.16666663e-01 -1.44923016e-01 -6.30623251e-02 -1.34723052e-01 8.40546608e-01 -7.48573601e-01 8.46737921e-01 -2.10212922e+00 -5.79275787e-02 -6.09144032e-01 3.04795772e-01 4.33193535e-01 -1.01370350e-01 1.92977078e-02 -5.76360047e-01 1.14614159e-01 7.00342730e-02 -2.25441717e-02 -8.05627406e-01 -5.53169787e-01 -2.50545651e-01 6.18097603e-01 3.01645752e-02 6.51260853e-01 -7.03431308e-01 -5.09353876e-01 6.42762423e-01 1.16517949e+00 3.21030557e-01 -1.69297591e-01 2.60883778e-01 9.99342382e-01 -4.79612976e-01 6.96066678e-01 1.07390022e+00 -3.89820337e-01 1.15677062e-02 -7.50039160e-01 -6.79336071e-01 -2.98829675e-01 -1.08149385e+00 1.63705897e+00 -2.84443527e-01 6.93506002e-01 2.12711245e-01 -7.31901884e-01 9.19719160e-01 2.54170507e-01 7.51702011e-01 -1.14525282e+00 2.90356558e-02 -2.57693212e-02 -4.52835172e-01 -3.48563850e-01 7.48798788e-01 -6.14947639e-02 3.77614111e-01 3.05967927e-01 -3.12459022e-01 1.38531089e-01 -1.44107521e-01 -7.46353567e-02 4.46196675e-01 1.65981650e-01 3.64365697e-01 -2.80225813e-01 6.57758057e-01 2.02216823e-02 9.36886147e-02 6.05671346e-01 -2.87026197e-01 7.84008086e-01 -3.79983306e-01 -8.43722761e-01 -1.37116003e+00 -1.00841355e+00 -6.66682363e-01 1.32425737e+00 3.77437413e-01 2.94283926e-01 -5.70094526e-01 3.03208411e-01 -6.93911850e-01 3.85341257e-01 -6.21907532e-01 1.70311019e-01 -7.13515401e-01 -1.14017463e+00 1.37897238e-01 4.98424947e-01 9.30702209e-01 -1.23981750e+00 -7.03347266e-01 -2.47992128e-01 -5.50367415e-01 -1.31417966e+00 -1.44458890e-01 -2.29692668e-01 -1.15388036e+00 -8.22965622e-01 -8.67052257e-01 -9.63107646e-01 2.80989558e-01 1.14973164e+00 1.00792801e+00 3.41853090e-02 -8.51269007e-01 2.50599325e-01 -3.99318755e-01 -2.47716457e-01 -1.93774417e-01 -2.19253376e-01 1.22952789e-01 -3.52890968e-01 4.38973427e-01 -4.05194223e-01 -1.06039226e+00 1.42415926e-01 -1.00267684e+00 1.63827047e-01 6.95007205e-01 5.41324198e-01 8.53511453e-01 5.90411603e-01 3.44681263e-01 -5.33296525e-01 3.92961860e-01 -7.63808265e-02 -6.94283962e-01 6.52828664e-02 -6.52639568e-01 -1.92721680e-01 1.27424270e-01 -1.90038785e-01 -1.74850619e+00 -1.85880605e-02 1.36257991e-01 -2.12615162e-01 -4.88903463e-01 5.69588244e-02 1.41339496e-01 -4.13708717e-01 1.15960670e+00 4.30045813e-01 1.25927731e-01 -6.27218723e-01 3.96625578e-01 7.62972474e-01 8.02241564e-01 -1.45399705e-01 9.08054411e-01 9.34998035e-01 -4.37820591e-02 -1.43741989e+00 -1.20378673e+00 -6.01311028e-01 -8.37266028e-01 -2.57728636e-01 1.13989818e+00 -1.21199000e+00 -3.06735337e-01 4.79267061e-01 -9.42454934e-01 -1.38093859e-01 1.61332726e-01 4.46704358e-01 -5.23800492e-01 6.17253065e-01 -9.82117653e-01 -7.93184638e-01 -9.49973762e-01 -9.67642486e-01 1.18446481e+00 6.81306362e-01 1.90573707e-01 -7.43067920e-01 -9.08649564e-02 9.81857479e-01 8.55243564e-01 2.91708052e-01 4.42932785e-01 3.61289233e-01 -6.76614940e-01 1.83542371e-01 -8.29767287e-01 3.98709625e-01 2.75552213e-01 -1.50724888e-01 -1.26392281e+00 -4.08688307e-01 1.40784130e-01 -2.50504494e-01 9.25916433e-01 8.40317905e-01 1.39929354e+00 1.19174898e-01 -3.97155657e-02 8.73429298e-01 1.90942609e+00 2.15207741e-01 1.16146338e+00 7.03467369e-01 6.01187646e-01 6.67077839e-01 8.48263204e-01 2.66955048e-01 -9.42965820e-02 8.77880812e-01 4.12602544e-01 -4.43963051e-01 -4.99505699e-01 4.44869578e-01 2.09008917e-01 4.36488204e-02 -9.36271548e-01 2.82634467e-01 -4.55322832e-01 -9.24885422e-02 -1.28008401e+00 -1.54727364e+00 -4.96693462e-01 2.15155888e+00 7.24798620e-01 -5.66410959e-01 7.53709301e-02 -3.84227410e-02 1.04963100e+00 3.29218209e-01 -7.39470243e-01 2.82110553e-02 -3.51745516e-01 3.08954537e-01 5.73077679e-01 4.71450567e-01 -1.29039288e+00 1.00460708e+00 6.72666359e+00 7.71803379e-01 -1.15247810e+00 1.16553321e-01 7.12406278e-01 -5.12311682e-02 1.77779347e-01 -5.15822828e-01 -9.90518570e-01 6.21844977e-02 8.32519710e-01 -1.60672203e-01 7.74375141e-01 5.34684002e-01 5.48397958e-01 -1.50746614e-01 -5.09378910e-01 1.28090346e+00 1.75225034e-01 -1.43122709e+00 -1.00678235e-01 -4.93907817e-02 7.85406768e-01 5.89049816e-01 6.04901552e-01 -1.55541629e-01 5.25693316e-03 -1.21745813e+00 2.40442865e-02 3.62079442e-01 1.70875192e+00 -7.43730545e-01 6.66144967e-01 -1.99411839e-01 -1.26359236e+00 -1.90818444e-01 -7.99498558e-01 1.84870452e-01 -4.50718291e-02 4.36752498e-01 -1.95115432e-02 6.81543827e-01 1.49764812e+00 5.43089211e-01 -5.48514366e-01 6.97381616e-01 2.15452120e-01 9.75562558e-02 2.50545919e-01 7.98591435e-01 -4.16226059e-01 -2.51693100e-01 5.21749139e-01 1.06537020e+00 2.42647812e-01 5.22730589e-01 -2.07236767e-01 8.56035411e-01 3.05444658e-01 -2.04624310e-01 -4.16857719e-01 1.50337636e-01 3.48391175e-01 1.61987329e+00 -4.89136726e-01 -8.30443390e-03 -7.13322937e-01 1.11920679e+00 -1.60071060e-01 6.70059562e-01 -6.20900869e-01 -2.66918749e-01 7.99193859e-01 1.00493535e-01 -2.29096264e-01 -3.20678651e-01 -4.91437078e-01 -9.36461091e-01 -2.08937839e-01 -9.17999804e-01 6.08534634e-01 -1.38781714e+00 -1.09407520e+00 7.89721429e-01 1.69073686e-01 -1.37407172e+00 1.34847999e-01 -4.97482568e-01 -2.28625447e-01 1.17207158e+00 -1.89388025e+00 -1.11051452e+00 -9.71392214e-01 6.67229235e-01 9.66986954e-01 6.92451373e-03 5.62680483e-01 1.58572063e-01 -4.71377939e-01 2.65172064e-01 2.65346944e-01 -2.44632307e-02 7.05195665e-01 -8.25806499e-01 3.71774852e-01 8.34975243e-01 -2.70102084e-01 5.43249428e-01 9.58744764e-01 -4.48573560e-01 -1.36962831e+00 -1.06661856e+00 2.82423615e-01 -6.18623018e-01 3.92896086e-01 5.24463728e-02 -1.15741777e+00 5.05324066e-01 2.82034993e-01 3.89983621e-03 4.91290241e-01 -1.49236187e-01 -2.47099623e-01 -1.65707335e-01 -1.32195032e+00 4.76696163e-01 1.02380300e+00 -3.94174546e-01 -3.04038525e-01 3.14558148e-01 8.18035007e-01 -3.80769491e-01 -1.17052972e+00 3.32394421e-01 6.33083940e-01 -1.24020612e+00 1.62070870e+00 -2.60176007e-02 5.52245021e-01 -3.24290365e-01 3.27660963e-02 -8.37730229e-01 -9.47032988e-01 -2.36157000e-01 8.35099593e-02 8.79671454e-01 -1.30610779e-01 -5.14810801e-01 7.42431998e-01 5.09460926e-01 3.76950316e-02 -2.88186967e-01 -5.97516894e-01 -6.27980947e-01 1.01469465e-01 -9.12681073e-02 1.58055812e-01 9.28797901e-01 -2.61128366e-01 2.23132491e-01 -5.50503969e-01 5.03935993e-01 1.49783337e+00 7.83273578e-02 3.04513723e-01 -1.15197277e+00 2.09126219e-01 -4.16583866e-01 -1.05243661e-02 -4.63578671e-01 -2.88130432e-01 -4.97165531e-01 -1.87792525e-01 -1.70475388e+00 5.54050446e-01 3.13035795e-03 -4.99458969e-01 5.17269783e-02 -1.63506925e-01 7.94113517e-01 -6.53377175e-02 5.50922811e-01 -5.35146415e-01 2.29549799e-02 1.56678081e+00 1.30925521e-01 -2.80292720e-01 -3.45057901e-03 -1.23724747e+00 4.47903067e-01 1.45490777e+00 8.08651447e-02 -2.65197694e-01 -5.86942673e-01 1.00491241e-01 1.70219108e-01 4.40842152e-01 -8.41259897e-01 1.74052343e-01 -4.25243527e-01 7.50151336e-01 -6.54291034e-01 3.97118419e-01 -5.11695564e-01 2.39270434e-01 1.58419013e-01 -4.95252997e-01 -7.65145868e-02 2.26832867e-01 3.12478840e-01 -9.86758247e-02 1.27509594e-01 1.53692079e+00 -6.05035543e-01 -1.24348962e+00 5.26344180e-01 -1.82278737e-01 -4.01149124e-01 7.18129277e-01 -6.01410925e-01 -7.68688858e-01 -1.04641870e-01 -4.27496880e-01 -8.84910300e-02 5.50934255e-01 6.01719499e-01 9.75910902e-01 -1.02970791e+00 -1.03609836e+00 -2.28274185e-02 1.20148741e-01 4.70035858e-02 1.16538143e+00 7.57721424e-01 -4.33852404e-01 4.77121711e-01 -5.29191732e-01 -3.54689807e-01 -1.69963145e+00 7.60162592e-01 4.86423582e-01 1.73369616e-01 -1.09196842e+00 8.23951542e-01 4.28851992e-01 2.02165917e-01 -1.45625335e-03 1.49180025e-01 -7.37736166e-01 -4.51659411e-01 1.42816913e+00 6.57787263e-01 -5.26110381e-02 -5.80488682e-01 -1.45964801e-01 9.86076057e-01 -2.63191611e-01 1.69814691e-01 1.39769340e+00 -8.79032910e-01 -2.58581668e-01 2.65409708e-01 8.40298891e-01 -5.12724638e-01 -1.20650828e+00 -5.23390293e-01 -2.05575868e-01 -6.05499387e-01 5.67408800e-01 -8.08118165e-01 -1.14262927e+00 7.24322975e-01 1.28900731e+00 -1.45306185e-01 1.57952309e+00 2.99162030e-01 5.23070931e-01 -1.41343504e-01 5.43398142e-01 -1.06646454e+00 3.11070800e-01 1.79988101e-01 1.05940938e+00 -1.54524326e+00 4.72844481e-01 -5.68930805e-01 -7.09042132e-01 1.23371661e+00 6.44011915e-01 -2.50439961e-02 2.05858350e-01 3.23880404e-01 3.79202068e-01 -2.02326685e-01 -5.47898114e-01 -3.22459340e-01 1.04827024e-01 1.14297521e+00 9.15946603e-01 -3.73906255e-01 -1.72450662e-01 -1.73709631e-01 6.82840496e-02 3.32437456e-01 7.60186791e-01 4.18801248e-01 -8.45703304e-01 -7.61466861e-01 -6.58773065e-01 3.13994914e-01 -8.10341001e-01 -3.93586427e-01 -2.75128279e-02 3.50240111e-01 -2.03867406e-01 9.85602081e-01 6.29904717e-02 -2.82522500e-01 -1.41277775e-01 -4.01191622e-01 5.97559214e-01 -4.20971423e-01 2.29496565e-02 9.55391452e-02 -1.28590286e-01 -7.74454534e-01 -7.85234094e-01 -5.59031188e-01 -1.05471289e+00 -2.45235816e-01 2.51012314e-02 -3.32147419e-01 9.23062742e-01 4.10008192e-01 2.10724682e-01 4.69411731e-01 5.53371012e-01 -1.15309894e+00 -3.65275383e-01 -8.94192994e-01 -9.74674463e-01 -8.75834376e-03 4.48654115e-01 -4.14925665e-01 -5.34446537e-01 2.50370234e-01]

[10.571118354797363, -2.2070910930633545]

9127abb3-54ed-41ea-b1f8-cd2c8acfb8c9

amodal-3d-reconstruction-for-robotic

2009.13146

https://arxiv.org/abs/2009.13146v1

https://arxiv.org/pdf/2009.13146v1.pdf

Amodal 3D Reconstruction for Robotic Manipulation via Stability and Connectivity

Learning-based 3D object reconstruction enables single- or few-shot estimation of 3D object models. For robotics, this holds the potential to allow model-based methods to rapidly adapt to novel objects and scenes. Existing 3D reconstruction techniques optimize for visual reconstruction fidelity, typically measured by chamfer distance or voxel IOU. We find that when applied to realistic, cluttered robotics environments, these systems produce reconstructions with low physical realism, resulting in poor task performance when used for model-based control. We propose ARM, an amodal 3D reconstruction system that introduces (1) a stability prior over object shapes, (2) a connectivity prior, and (3) a multi-channel input representation that allows for reasoning over relationships between groups of objects. By using these priors over the physical properties of objects, our system improves reconstruction quality not just by standard visual metrics, but also performance of model-based control on a variety of robotics manipulation tasks in challenging, cluttered environments. Code is available at github.com/wagnew3/ARM.

['Siddhartha Srinivasa', 'Caelen Wang', 'Octavian Murad', 'Aaron Walsman', 'Pedro Domingos', 'Christopher Xie', 'William Agnew']

2020-09-28

['3d-object-reconstruction']

['computer-vision']

[-9.90456417e-02 -5.98656163e-02 4.87072505e-02 1.80068996e-03 -4.52962399e-01 -5.79163611e-01 5.81155121e-01 1.50582746e-01 -1.73765853e-01 3.50674719e-01 -2.30673589e-02 -1.17002688e-01 -4.26328070e-02 -5.73312104e-01 -9.69588757e-01 -2.57964015e-01 -1.64602622e-01 8.75789642e-01 6.09851241e-01 4.78061411e-04 3.72354537e-01 1.06508207e+00 -1.55070233e+00 5.90601750e-03 3.70922595e-01 9.97307360e-01 7.24747181e-01 6.44849896e-01 1.55149952e-01 6.25365615e-01 -1.19113140e-01 2.39358008e-01 5.38393974e-01 -3.42711881e-02 -5.79990923e-01 2.71679342e-01 3.62390429e-01 -5.29891849e-01 -4.24942583e-01 7.59831190e-01 3.93147469e-01 3.78224820e-01 8.00003827e-01 -1.03822756e+00 -3.70875031e-01 1.58309370e-01 -5.27095973e-01 -4.94734049e-02 5.26048124e-01 8.17098498e-01 5.77612936e-01 -1.07967496e+00 8.93493235e-01 1.58965838e+00 6.16774082e-01 5.31524360e-01 -1.60027099e+00 -3.44607979e-01 1.00643568e-01 8.22366402e-02 -1.29582572e+00 -5.58671832e-01 4.83641297e-01 -7.41022289e-01 1.14405584e+00 -2.43970603e-02 1.01888716e+00 7.60930836e-01 6.50139093e-01 5.22435725e-01 8.14057827e-01 -2.10204825e-01 4.12344635e-01 -1.85336858e-01 -2.10785538e-01 8.45791817e-01 4.25955415e-01 3.66225332e-01 -5.54226339e-01 -1.83870457e-03 1.41819644e+00 -3.86376567e-02 -2.66785651e-01 -1.16413224e+00 -1.46375835e+00 3.70079786e-01 7.79944301e-01 -3.29056352e-01 -4.01404887e-01 6.25224352e-01 -3.20663676e-02 -8.51556845e-03 1.03997819e-01 7.96130240e-01 -1.74578190e-01 -1.10980727e-01 -3.24434429e-01 5.57255805e-01 7.56342947e-01 1.33513141e+00 6.41685963e-01 1.86511561e-01 2.65319478e-02 5.18714607e-01 5.52294791e-01 6.10798478e-01 -1.18571602e-01 -1.48456776e+00 9.76262093e-02 3.07213992e-01 4.54659671e-01 -8.08977425e-01 -5.16909778e-01 -1.30477756e-01 -2.92842090e-01 8.78444374e-01 2.55049169e-01 3.14046830e-01 -1.28026974e+00 1.45273840e+00 6.34276330e-01 -1.33390963e-01 -2.93222398e-01 1.15172362e+00 6.79810405e-01 3.96444678e-01 -2.40141511e-01 8.47647488e-02 8.61225188e-01 -7.88301766e-01 -4.14803088e-01 -3.60405296e-01 1.43644094e-01 -6.92790926e-01 1.09459817e+00 4.44943011e-01 -1.50884998e+00 -3.12446892e-01 -1.08493662e+00 -1.78826392e-01 -4.09805477e-02 -4.53740120e-01 6.02978885e-01 1.59779657e-02 -8.14230144e-01 6.29828811e-01 -1.34865308e+00 -5.12160420e-01 5.83384812e-01 3.74479324e-01 -4.48026955e-01 -4.68851119e-01 -1.57260165e-01 1.45906556e+00 2.74514109e-01 -1.14687763e-01 -1.50770724e+00 -6.73562765e-01 -1.02083480e+00 -1.88569307e-01 4.66402560e-01 -8.80343556e-01 1.34608400e+00 -1.41621694e-01 -1.65191531e+00 5.64018011e-01 2.50023961e-01 -2.12394133e-01 5.02497256e-01 -2.79471010e-01 4.69133824e-01 2.26314664e-01 -1.03287064e-01 9.84166324e-01 7.62374103e-01 -1.84905720e+00 -3.20974067e-02 -3.71156901e-01 1.15139619e-01 6.42430067e-01 4.23746616e-01 -5.45675576e-01 -4.17725325e-01 -3.22795779e-01 5.95080554e-01 -9.68913257e-01 -4.88659889e-01 1.04957902e+00 -1.86009571e-01 2.43349120e-01 9.84180689e-01 -3.84050488e-01 2.65128583e-01 -2.11572242e+00 4.92829919e-01 1.40886143e-01 1.01095773e-01 -8.12858790e-02 -1.08203530e-01 3.15108269e-01 4.15787250e-01 -1.09728403e-01 -1.17380463e-01 -3.14942569e-01 -6.29337654e-02 3.05698633e-01 2.90274620e-02 6.60601377e-01 1.78141788e-01 8.68004084e-01 -9.85480607e-01 -3.47714067e-01 7.44898021e-01 4.43857104e-01 -7.67325461e-01 2.49357834e-01 -5.40403545e-01 6.58114612e-01 -4.04498935e-01 7.84172297e-01 5.67353368e-01 -2.47334942e-01 -9.69450697e-02 -2.12627754e-01 2.89430376e-02 2.24195942e-02 -1.15728378e+00 2.04889965e+00 -5.47335625e-01 3.26773554e-01 4.94063467e-01 -5.46082020e-01 8.89807999e-01 -9.85936262e-04 5.36471605e-01 -2.17700377e-01 2.89863765e-01 5.43826520e-02 -8.96862447e-02 -3.35904360e-01 4.07962590e-01 -2.09431708e-01 3.88954468e-02 3.06976199e-01 1.05207056e-01 -1.27370918e+00 -1.73601031e-01 2.46546030e-01 1.26154327e+00 4.58363801e-01 2.40107372e-01 -1.90649301e-01 -3.00120294e-01 3.35136533e-01 1.54615790e-01 7.31746912e-01 -1.19891763e-02 8.19893062e-01 -1.22181252e-01 -2.73173988e-01 -1.27128911e+00 -1.66925228e+00 -2.79361010e-01 2.46369183e-01 7.76019156e-01 -1.08721465e-01 -1.61727846e-01 -7.68466443e-02 5.26372850e-01 6.70535505e-01 -2.53219306e-01 -3.48566890e-01 -3.34696651e-01 -4.33527716e-02 -6.36455789e-02 5.22564650e-01 1.03207551e-01 -9.50091600e-01 -1.16395795e+00 2.52956331e-01 1.22850589e-01 -1.05230165e+00 -2.75698751e-01 3.62376362e-01 -1.20469916e+00 -1.11106670e+00 -4.29298788e-01 -6.10395908e-01 8.26816559e-01 4.74963307e-01 9.14268613e-01 6.54238537e-02 -5.53867161e-01 9.24934208e-01 -3.10567170e-01 -2.90468365e-01 -3.72286707e-01 -4.98934925e-01 2.35186696e-01 -5.92714310e-01 -3.80489856e-01 -6.80225015e-01 -3.59948397e-01 4.38111335e-01 -6.58999920e-01 3.04838151e-01 5.41487992e-01 6.87343299e-01 8.43356907e-01 -3.05805326e-01 3.61910686e-02 -1.95029005e-01 2.78791815e-01 -2.55428582e-01 -8.28397512e-01 -1.55262798e-01 -3.97756040e-01 2.83756247e-03 2.04987407e-01 -8.64404559e-01 -9.00891244e-01 4.43475932e-01 4.11745429e-01 -8.69069219e-01 4.37358255e-03 3.53790969e-01 -5.44399116e-03 -2.80244589e-01 9.83845830e-01 -2.06410393e-01 4.12713557e-01 -5.64691722e-01 5.02024174e-01 3.02090585e-01 5.19166529e-01 -7.26408005e-01 7.11886942e-01 6.62982345e-01 2.44462699e-01 -7.46579885e-01 -3.08118999e-01 -2.55991668e-01 -7.98194230e-01 -4.93238300e-01 6.84225082e-01 -9.50330853e-01 -7.65870273e-01 4.26211059e-01 -1.05930686e+00 -8.64974499e-01 -5.26361525e-01 7.99724817e-01 -1.14656448e+00 4.62746881e-02 -5.59508204e-01 -7.99443185e-01 5.25343567e-02 -1.31746364e+00 1.05811203e+00 -1.11343032e-02 -3.08382481e-01 -5.29817045e-01 -2.58007556e-01 2.08198085e-01 2.75908440e-01 4.69947696e-01 8.83814692e-01 9.24589112e-02 -1.12835169e+00 -2.53471076e-01 -1.10942841e-01 1.29411099e-02 3.02021895e-02 -1.34273589e-01 -5.34721136e-01 -4.60621655e-01 -2.47556016e-01 -5.63413441e-01 5.29158890e-01 4.64783043e-01 1.04026234e+00 -5.15829101e-02 -5.35452008e-01 4.49318826e-01 1.31555402e+00 6.01644032e-02 3.97906929e-01 -1.60643198e-02 6.54559731e-01 3.29959422e-01 6.55624986e-01 6.16944373e-01 2.89753079e-01 7.96061814e-01 9.41336989e-01 2.58628041e-01 -3.78510565e-01 -3.33747834e-01 1.99607968e-01 5.80867112e-01 1.40160203e-01 -3.89352478e-02 -1.06530023e+00 5.67196667e-01 -1.81345212e+00 -6.32093728e-01 6.01922050e-02 2.38219786e+00 6.50668561e-01 3.44156086e-01 -1.03993796e-01 -4.09425169e-01 4.53246504e-01 -2.17751145e-01 -1.11286521e+00 -2.36908048e-02 3.39554191e-01 -6.13744147e-02 4.84256476e-01 6.39603853e-01 -5.87288141e-01 8.91648710e-01 6.58437443e+00 3.58793795e-01 -1.01781642e+00 -9.15233232e-03 -1.09267943e-01 -4.97749120e-01 -2.23186389e-01 2.52678484e-01 -4.97157544e-01 -2.28235335e-03 2.46389240e-01 -1.51829913e-01 7.27645993e-01 8.95748854e-01 1.75720781e-01 -5.79108655e-01 -1.30489767e+00 9.80248272e-01 1.21831477e-01 -1.32660413e+00 -1.66360158e-02 6.75956234e-02 5.33923388e-01 2.16273919e-01 -4.69642356e-02 -3.67777795e-02 6.46825254e-01 -1.01667094e+00 1.26045847e+00 6.96427464e-01 7.66070962e-01 -3.94783407e-01 1.42893925e-01 6.24179721e-01 -9.32830811e-01 1.34344790e-02 -3.20993572e-01 -3.07147088e-03 4.28834558e-01 2.61328548e-01 -1.00616598e+00 1.32690042e-01 8.45507145e-01 8.30179036e-01 -1.31007060e-01 1.49233425e+00 -9.77707431e-02 -3.21349837e-02 -6.09727323e-01 1.93767473e-02 -1.65787756e-01 2.04597920e-01 9.74988222e-01 6.76734626e-01 1.03483334e-01 2.70249039e-01 5.42324364e-01 1.10715985e+00 2.60330200e-01 -4.28412825e-01 -8.34244847e-01 7.64378160e-02 5.77053428e-01 9.50870752e-01 -9.01606798e-01 -3.99678573e-02 3.96439768e-02 7.56794930e-01 3.91432911e-01 3.18538398e-01 -4.95422661e-01 -6.83314800e-02 4.86230284e-01 4.69741791e-01 3.07714731e-01 -1.02497089e+00 -4.59377140e-01 -1.03454006e+00 -9.77868959e-03 -6.11205339e-01 -2.86167741e-01 -1.31900012e+00 -1.09009933e+00 1.42985284e-01 5.18038213e-01 -1.28829181e+00 1.60565469e-02 -7.51749277e-01 -1.00364186e-01 6.47945344e-01 -1.05994701e+00 -9.86974359e-01 -5.28331518e-01 3.28246564e-01 6.01389945e-01 1.83289170e-01 6.96251988e-01 -1.34965956e-01 1.13612577e-01 -1.78616434e-01 -1.32610604e-01 -3.78756136e-01 4.65384513e-01 -9.43325937e-01 3.43827426e-01 4.83996004e-01 -1.89357977e-02 4.17914897e-01 8.18067312e-01 -8.96286011e-01 -1.90139782e+00 -8.18353057e-01 -1.50548697e-01 -7.59500623e-01 1.84822723e-01 -5.08645535e-01 -9.31315124e-01 7.53884971e-01 -3.29152763e-01 3.07100624e-01 2.16472633e-02 -1.30571947e-01 -1.65258944e-01 2.14077935e-01 -1.45688808e+00 6.94594145e-01 1.22104490e+00 -1.92470819e-01 -4.46329594e-01 4.05619621e-01 6.81587398e-01 -8.83529723e-01 -9.60118711e-01 4.37777400e-01 9.12878275e-01 -5.72412491e-01 1.05884290e+00 -4.31854308e-01 2.44799063e-01 -5.93591750e-01 -3.79295856e-01 -1.38431203e+00 -5.25655329e-01 -3.58002752e-01 -4.89748508e-01 2.61353999e-01 2.40361691e-01 -2.74652064e-01 5.53129852e-01 6.99336290e-01 -4.61661965e-01 -6.53868139e-01 -8.89372766e-01 -1.05345726e+00 -5.77692725e-02 -3.22529554e-01 2.68262029e-01 4.81848925e-01 3.64392623e-02 2.30184257e-01 -7.40116984e-02 2.93330640e-01 7.22450912e-01 -3.64650488e-02 1.04595816e+00 -1.18212271e+00 -3.94959003e-01 -4.15732652e-01 -5.85661769e-01 -1.25540614e+00 -1.04222476e-01 -7.90126979e-01 6.57503366e-01 -1.90808904e+00 4.77062315e-02 -6.64482057e-01 2.77808547e-01 5.16201377e-01 4.00413632e-01 5.23865819e-02 4.67092901e-01 3.94737571e-01 -4.79165167e-01 8.22875261e-01 1.68278670e+00 -3.03640403e-02 -3.05326253e-01 -2.59762287e-01 -9.09010619e-02 7.11014628e-01 3.87031227e-01 -3.18384081e-01 -3.03518951e-01 -7.74268270e-01 -1.95920080e-01 2.62016356e-01 7.19636559e-01 -1.19949210e+00 1.81157261e-01 -4.37578350e-01 6.15098119e-01 -6.25960469e-01 1.00912082e+00 -1.14959538e+00 4.73298550e-01 6.75964296e-01 -2.16519520e-01 -1.13809243e-01 4.36014622e-01 7.72969604e-01 4.33720231e-01 -1.01164386e-01 1.02265298e+00 -5.20273030e-01 -6.15588188e-01 3.48477453e-01 -4.16483998e-01 -1.56806484e-01 1.07024252e+00 -5.33480644e-01 -8.68655443e-02 -4.09907788e-01 -8.50096524e-01 3.46625745e-01 1.14422011e+00 5.03534913e-01 1.02713358e+00 -1.31801820e+00 -4.23989892e-01 2.24859655e-01 1.62509620e-01 5.71528494e-01 -2.54605655e-02 6.03510082e-01 -8.38544011e-01 -1.70711681e-01 -5.18483341e-01 -1.11244905e+00 -1.02616715e+00 5.22452295e-01 4.39271152e-01 3.59676898e-01 -8.93084288e-01 7.63902545e-01 8.23033527e-02 -5.86555600e-01 3.26758206e-01 -4.56301808e-01 4.28453743e-01 -6.29651845e-01 1.04033589e-01 3.23315203e-01 -1.98185802e-01 -5.21177113e-01 -3.16921294e-01 5.92178464e-01 1.06760636e-01 -4.02498901e-01 1.35326326e+00 -6.90630749e-02 2.38165155e-01 7.32279956e-01 7.05703020e-01 -5.35907984e-01 -1.92302024e+00 -2.10009232e-01 -2.91213453e-01 -8.96531999e-01 2.57605582e-01 -8.14771354e-01 -6.58916593e-01 8.22662354e-01 5.43249011e-01 -3.41228366e-01 5.64937413e-01 2.69036561e-01 4.50182617e-01 5.77646017e-01 9.71958935e-01 -7.56063342e-01 5.92361748e-01 6.50361180e-01 1.37003100e+00 -1.15493917e+00 4.24504608e-01 -3.90524626e-01 -4.22281533e-01 9.44341302e-01 8.35809588e-01 -2.56477386e-01 7.70462751e-01 5.07828236e-01 -2.39113316e-01 -2.83397168e-01 -7.51113415e-01 -6.62296638e-03 2.24797979e-01 8.91019881e-01 -1.91354334e-01 -3.95578817e-02 3.54021043e-01 -3.02593894e-02 2.37841927e-03 -1.31628841e-01 5.93748987e-01 1.41198707e+00 -8.46175253e-01 -6.46183193e-01 -3.36974680e-01 5.41587234e-01 3.43532294e-01 3.00749302e-01 -1.51983574e-01 7.06879139e-01 -2.99409121e-01 6.43637002e-01 2.78976291e-01 -2.58809894e-01 5.70655167e-01 -3.64506155e-01 1.14759147e+00 -1.12598932e+00 -2.26290390e-01 1.18472397e-01 -7.39271268e-02 -7.89821744e-01 -1.24123223e-01 -8.65407050e-01 -1.45686495e+00 -1.06414668e-01 -5.65456033e-01 -4.26494002e-01 9.22438204e-01 6.80085242e-01 5.26754916e-01 2.96721548e-01 2.02211022e-01 -1.62096047e+00 -6.20183587e-01 -9.26517248e-01 -5.47284603e-01 1.84572741e-01 3.55958700e-01 -1.26528502e+00 -3.78047936e-02 1.01994209e-01]

[5.833354949951172, -0.8431549072265625]

7c3feedd-96c4-4378-b9aa-6c63e3137e5f

branch-ranking-for-efficient-mixed-integer

2207.13701

https://arxiv.org/abs/2207.13701v1

https://arxiv.org/pdf/2207.13701v1.pdf

Branch Ranking for Efficient Mixed-Integer Programming via Offline Ranking-based Policy Learning

Deriving a good variable selection strategy in branch-and-bound is essential for the efficiency of modern mixed-integer programming (MIP) solvers. With MIP branching data collected during the previous solution process, learning to branch methods have recently become superior over heuristics. As branch-and-bound is naturally a sequential decision making task, one should learn to optimize the utility of the whole MIP solving process instead of being myopic on each step. In this work, we formulate learning to branch as an offline reinforcement learning (RL) problem, and propose a long-sighted hybrid search scheme to construct the offline MIP dataset, which values the long-term utilities of branching decisions. During the policy training phase, we deploy a ranking-based reward assignment scheme to distinguish the promising samples from the long-term or short-term view, and train the branching model named Branch Ranking via offline policy learning. Experiments on synthetic MIP benchmarks and real-world tasks demonstrate that Branch Rankink is more efficient and robust, and can better generalize to large scales of MIP instances compared to the widely used heuristics and state-of-the-art learning-based branching models.

['Jun Wang', 'Yong Yu', 'Jianye Hao', 'Mingxuan Yuan', 'Hui-Ling Zhen', 'Furui Liu', 'Chuhan Shi', 'Weinan Zhang', 'WenHao Chen', 'Zeren Huang']

2022-07-26

['variable-selection']

['methodology']

[ 2.66866654e-01 1.95016086e-01 -1.27588952e+00 -4.32601452e-01 -1.13335121e+00 -7.38099933e-01 2.25804336e-02 6.96677119e-02 -3.04507792e-01 1.42745578e+00 -1.52445391e-01 -8.14437866e-01 -5.14623940e-01 -9.87559915e-01 -9.34149325e-01 -8.35514128e-01 -4.01628643e-01 1.12712681e+00 -5.22229224e-02 1.18439734e-01 2.90284097e-01 3.76243532e-01 -9.37094808e-01 4.04404551e-01 1.37549412e+00 1.32096982e+00 2.58928716e-01 4.73556787e-01 -4.06928211e-01 9.33975756e-01 -3.60863090e-01 -2.65445292e-01 5.06243289e-01 -1.69773400e-01 -1.06855154e+00 -3.91143896e-02 2.66713947e-02 -4.32231009e-01 -3.71950231e-02 9.55344677e-01 2.24571034e-01 1.93556696e-01 1.48762912e-01 -1.33394372e+00 -9.37890708e-02 1.02125525e+00 -7.45668411e-01 2.17433825e-01 4.32070434e-01 3.77682298e-01 1.39334011e+00 -3.05191297e-02 7.35000491e-01 1.30997968e+00 7.22454563e-02 4.18051571e-01 -1.44438541e+00 -5.87644875e-01 1.02286267e+00 2.99560815e-01 -6.98844552e-01 2.95283031e-02 7.00546026e-01 -1.22255594e-01 8.26601505e-01 2.97542989e-01 7.32531726e-01 1.09224725e+00 1.80336624e-01 1.42975938e+00 1.33457911e+00 -1.00957043e-01 5.08351386e-01 -7.60554448e-02 -6.91474741e-03 7.35801339e-01 2.29533166e-02 5.26808083e-01 -3.24618012e-01 -2.97728300e-01 5.27610660e-01 -8.22871774e-02 -1.28950551e-01 -6.76239371e-01 -9.26160336e-01 1.08436513e+00 5.80761075e-01 -3.59792292e-01 -5.99078000e-01 3.47529471e-01 5.03546834e-01 4.52424288e-01 2.32194722e-01 7.47938991e-01 -9.55362558e-01 -3.97372425e-01 -8.82584155e-01 5.66161633e-01 1.00734854e+00 9.78985667e-01 7.73766577e-01 -3.02411765e-01 -7.14006364e-01 6.40876949e-01 -5.84370233e-02 3.37232113e-01 -2.34022737e-02 -1.21316803e+00 1.19496834e+00 5.69924772e-01 3.73420984e-01 -6.40516877e-01 -4.11622763e-01 -6.26065016e-01 -4.80040699e-01 1.42412186e-01 3.65112960e-01 -3.43117625e-01 -1.01334214e+00 1.52491546e+00 4.49637234e-01 -2.06173182e-01 -1.06078655e-01 7.86526203e-01 1.31527290e-01 9.93120849e-01 1.55326063e-02 -8.71770084e-01 6.54934049e-01 -1.44204843e+00 -5.38765430e-01 -3.69411111e-01 5.92253387e-01 -1.27772242e-01 6.81046963e-01 8.83017540e-01 -1.30343044e+00 -9.84570533e-02 -7.38437295e-01 3.35244089e-01 -1.05988398e-01 -1.62391767e-01 1.28074718e+00 3.82359266e-01 -4.67489511e-01 8.53224277e-01 -9.90504384e-01 3.03702235e-01 7.17091262e-01 6.08305454e-01 9.39998105e-02 -4.29365814e-01 -9.24438059e-01 6.36655569e-01 6.42896295e-01 1.22401141e-01 -1.18853986e+00 -7.51862228e-01 -4.75345969e-01 2.10876003e-01 1.35159457e+00 -5.46372414e-01 1.41679931e+00 -9.25752997e-01 -1.71646893e+00 4.65924680e-01 -1.64026812e-01 -5.59264183e-01 8.14418435e-01 -1.21120512e-01 1.37219548e-01 -8.26003179e-02 1.33333042e-01 5.80774724e-01 6.77399695e-01 -1.40610719e+00 -1.35706127e+00 -2.32633337e-01 7.00040936e-01 1.78037897e-01 1.92243099e-01 -4.53059912e-01 -4.57318008e-01 -1.58828706e-01 1.17748633e-01 -8.59812438e-01 -7.86497176e-01 -4.66174960e-01 -3.16266596e-01 -1.95481405e-01 1.38725147e-01 -6.26180351e-01 1.57746577e+00 -1.60960031e+00 6.03272736e-01 4.55441356e-01 -1.89555392e-01 9.34905484e-02 -3.19514275e-01 2.43942901e-01 1.77979544e-01 1.94747493e-01 -8.61986876e-02 2.23848447e-01 1.43644452e-01 5.63406587e-01 -4.94056433e-01 1.11560836e-01 7.46655688e-02 7.66355813e-01 -1.29970121e+00 -3.98933023e-01 -1.76242337e-01 -6.20694876e-01 -1.04194748e+00 3.44387561e-01 -8.54510725e-01 5.82979739e-01 -1.00150251e+00 1.14165437e+00 7.17045963e-01 -6.90493509e-02 4.00412589e-01 2.24064365e-01 -4.02768441e-02 3.80080581e-01 -1.18100536e+00 1.58289361e+00 -8.67053032e-01 -2.00570524e-02 3.65678221e-01 -1.49174833e+00 6.27554059e-01 -1.56189442e-01 6.51611507e-01 -1.11977053e+00 -9.62663740e-02 3.95342022e-01 6.88942969e-02 -4.89129305e-01 1.66277915e-01 1.69986308e-01 -7.63560608e-02 2.21831530e-01 -1.07481420e-01 -3.21088254e-01 8.00624609e-01 -1.65463790e-01 1.21515763e+00 3.15894216e-01 2.25671649e-01 -1.10535522e-03 5.52588463e-01 2.30924428e-01 1.10561824e+00 9.92382526e-01 1.83338411e-02 -1.60024092e-01 1.16980565e+00 -6.63658857e-01 -4.09494340e-01 -9.43547547e-01 -3.83334383e-02 1.51118851e+00 2.02033788e-01 1.02757560e-02 -3.79094064e-01 -1.18333709e+00 2.82938749e-01 9.56041276e-01 -5.39942622e-01 4.23002169e-02 -6.71170175e-01 -7.60420322e-01 -2.30185166e-01 5.51908195e-01 3.35524827e-01 -1.17056298e+00 -4.64248002e-01 5.75444996e-01 -1.98418796e-01 -8.20340633e-01 -1.74358666e-01 8.93236101e-01 -1.14310002e+00 -1.07458138e+00 -4.58479255e-01 -5.72132170e-01 6.43742502e-01 -1.24671474e-01 1.46681249e+00 -9.92168556e-04 -1.73964426e-01 1.24721425e-02 -3.47246200e-01 -1.39248580e-01 5.47588393e-02 3.77099454e-01 -3.52481157e-01 -3.77567917e-01 -7.57887661e-02 -3.16922158e-01 -3.75245363e-01 3.41286242e-01 -4.44405496e-01 5.39863184e-02 8.04972351e-01 1.33228874e+00 1.02207577e+00 3.21271360e-01 4.05036539e-01 -1.09652913e+00 6.52148902e-01 -5.01660228e-01 -1.20367825e+00 7.02460885e-01 -7.23848581e-01 3.51420790e-01 7.37775564e-01 -4.49253380e-01 -9.85960722e-01 4.25025187e-02 1.47162065e-01 -1.84816882e-01 2.57443577e-01 7.96248257e-01 -1.72449023e-01 4.79597561e-02 1.72309786e-01 8.13521147e-02 -5.81709802e-01 -1.67883039e-01 2.81435579e-01 2.77653933e-01 2.06172973e-01 -1.46445894e+00 5.27992904e-01 1.16770752e-01 1.57789186e-01 1.50861263e-01 -1.35423791e+00 -3.50320578e-01 -2.64599293e-01 7.33722225e-02 4.28190947e-01 -5.66574931e-01 -1.09753573e+00 -2.08238393e-01 -9.22928929e-01 -8.71389210e-01 -3.16602170e-01 8.50723758e-02 -1.06652331e+00 -3.07878107e-01 -3.38027477e-01 -8.67432058e-01 -1.22170344e-01 -1.57140160e+00 7.83621371e-01 4.99325037e-01 2.10982054e-01 -9.24875796e-01 9.87320244e-02 5.47574639e-01 1.78890377e-02 2.91223943e-01 1.24481046e+00 -2.69753277e-01 -8.85379434e-01 7.60667473e-02 -1.95007861e-01 4.33159955e-02 -1.48457840e-01 1.01645522e-01 -4.16452765e-01 -4.94071990e-01 -2.51806349e-01 -6.70106053e-01 9.36303318e-01 6.75915301e-01 1.77823853e+00 -7.11510181e-01 -6.04542494e-01 1.07746255e+00 1.47348976e+00 6.25971138e-01 3.89682353e-01 7.61220098e-01 2.57631660e-01 4.51532125e-01 1.41696048e+00 6.77633524e-01 2.60836214e-01 3.86521041e-01 8.02008212e-01 1.78384811e-01 5.94423056e-01 -3.22945029e-01 2.86264896e-01 1.79487675e-01 -1.54968724e-01 -2.95747340e-01 -8.68587017e-01 3.13188761e-01 -2.11677814e+00 -9.41449881e-01 4.73333329e-01 2.26171851e+00 1.17200124e+00 6.04436398e-01 2.20002979e-01 -9.50169712e-02 3.29280645e-01 6.71189651e-02 -1.00599897e+00 -9.43407416e-01 3.25898647e-01 2.57672429e-01 8.64199936e-01 7.04725027e-01 -1.18572652e+00 1.10970330e+00 5.93499613e+00 9.87650335e-01 -1.23543990e+00 -2.79539078e-01 1.09528553e+00 -4.26063180e-01 -4.44150150e-01 6.42674789e-02 -9.64613020e-01 3.83470565e-01 5.96950889e-01 -1.75238490e-01 1.27785873e+00 1.23736739e+00 6.44490495e-02 -3.47761005e-01 -1.44849098e+00 6.08119130e-01 -6.72371089e-01 -1.40276134e+00 -2.19915047e-01 3.72614861e-02 1.20073175e+00 -1.52149066e-01 8.87575299e-02 8.79093826e-01 8.68156493e-01 -1.17576957e+00 5.40056050e-01 2.54785717e-01 5.02239466e-01 -1.08509874e+00 7.61819184e-01 5.94372451e-01 -1.00951660e+00 -9.72556114e-01 -4.45411026e-01 5.71600161e-02 1.48285925e-01 5.44686317e-01 -6.43588126e-01 6.93359494e-01 5.91613531e-01 5.52109778e-01 -1.75000563e-01 1.13655770e+00 -5.17922521e-01 5.87639093e-01 -1.40076444e-01 -1.85085326e-01 7.82660306e-01 -5.67862153e-01 3.00002486e-01 8.92734706e-01 -5.48006631e-02 8.84649456e-02 9.04335678e-01 7.79256344e-01 6.75835460e-02 -3.99258584e-02 4.29804586e-02 -5.30687392e-01 3.89264524e-01 1.19958174e+00 -6.84433937e-01 -1.20341957e-01 -1.10331841e-01 5.48993468e-01 7.22009003e-01 4.72497493e-01 -1.08913922e+00 -1.55951917e-01 4.10078496e-01 -2.63657182e-01 5.07448018e-01 5.24968468e-02 -5.44107258e-01 -7.95119703e-01 1.64583728e-01 -1.27027404e+00 6.11146629e-01 -1.57372192e-01 -1.12349188e+00 1.76479593e-01 -2.86043771e-02 -9.88717139e-01 -2.64333397e-01 -7.39070356e-01 -6.47707462e-01 5.83825827e-01 -1.94163275e+00 -7.34473348e-01 2.28908006e-02 2.20430702e-01 6.02370620e-01 1.08638937e-02 5.03600836e-01 -5.75128570e-02 -7.78643131e-01 5.24821281e-01 3.03387910e-01 -2.69660205e-01 2.75484174e-01 -1.36232662e+00 -3.09219152e-01 2.80156463e-01 -3.19540590e-01 3.03751945e-01 6.11862957e-01 -4.38036531e-01 -1.81805468e+00 -7.06252098e-01 1.12258784e-01 2.03026459e-01 8.51300955e-01 -2.12406553e-02 -4.79691595e-01 5.87143064e-01 -1.15510069e-01 -2.24068388e-02 3.31024736e-01 5.62672734e-01 1.22256903e-03 -5.73132098e-01 -1.20340741e+00 4.48196709e-01 1.05175316e+00 1.84919745e-01 -3.69431257e-01 6.35073960e-01 7.53811836e-01 -8.62808943e-01 -6.32113457e-01 5.85796893e-01 4.37829435e-01 -7.05144405e-01 8.75241756e-01 -1.00626361e+00 7.94695258e-01 2.18130752e-01 -1.52393803e-01 -1.49665868e+00 -5.35399199e-01 -9.04700637e-01 -5.33605874e-01 1.05744612e+00 5.88750839e-01 -4.95091408e-01 1.06620932e+00 6.07906163e-01 1.05133854e-01 -1.57804704e+00 -1.05473328e+00 -8.69571567e-01 2.78349936e-01 -7.23095611e-02 6.69040442e-01 6.34164691e-01 1.97972418e-04 -8.12426396e-03 -1.58365309e-01 2.24802628e-01 5.02569258e-01 8.66548240e-01 7.18361497e-01 -1.00176656e+00 -7.67794251e-01 -7.07427561e-01 2.89321303e-01 -1.31786752e+00 3.81841987e-01 -8.69143784e-01 2.64282107e-01 -1.72777355e+00 2.26116300e-01 -1.05755484e+00 -6.38421655e-01 6.07469618e-01 -2.95125484e-01 -8.84558856e-01 2.42863625e-01 -2.43841067e-01 -1.00556076e+00 5.15654802e-01 1.74233854e+00 -5.77353060e-01 -5.73433757e-01 4.97134864e-01 -6.72756672e-01 4.90140140e-01 6.42322600e-01 -6.15418077e-01 -3.88204098e-01 -3.67391199e-01 5.82869172e-01 8.82467389e-01 -3.80118608e-01 -6.43673837e-01 9.94999614e-03 -1.29485500e+00 9.04420838e-02 -7.31234610e-01 -3.87669797e-03 -9.54369068e-01 -3.25722009e-01 7.63047814e-01 -6.24196112e-01 -1.05860658e-01 9.19738486e-02 5.59312999e-01 -4.01815176e-02 -4.55795079e-01 5.25441527e-01 -3.82993609e-01 -5.99722445e-01 6.86868131e-01 2.24326253e-02 4.27858442e-01 1.28469002e+00 1.62551209e-01 -1.94267139e-01 -7.52602965e-02 -5.73111236e-01 1.23231089e+00 2.33899113e-02 1.07378066e-01 2.14664891e-01 -1.03322792e+00 -4.06616002e-01 -1.21314339e-01 -1.41915753e-01 4.16521758e-01 -7.04191672e-03 8.01335275e-01 -4.64115053e-01 6.16407037e-01 -2.90286094e-01 -4.68919098e-01 -6.80554926e-01 9.51567471e-01 1.44085705e-01 -1.37909567e+00 1.90756768e-02 1.03471470e+00 -1.61415458e-01 -6.49564385e-01 7.66739666e-01 -4.63656962e-01 6.96975216e-02 3.41849715e-01 2.98588365e-01 6.34942770e-01 -1.11163318e-01 4.64432359e-01 -2.53646880e-01 1.47358507e-01 -3.14224035e-01 -3.57139099e-04 1.60398293e+00 1.58991843e-01 -1.29958391e-01 1.99880421e-01 7.66193867e-01 -1.81367889e-01 -1.51527035e+00 -1.91992521e-01 1.27946377e-01 -7.52947867e-01 1.71295017e-01 -1.32935703e+00 -1.34142077e+00 7.88533747e-01 2.62248993e-01 2.56660908e-01 1.11747909e+00 -3.51753771e-01 6.09467506e-01 7.70798266e-01 8.82705450e-01 -1.60343039e+00 2.67994385e-02 5.74880660e-01 7.99154282e-01 -1.43325758e+00 3.59237581e-01 -3.43236297e-01 -7.17924833e-01 1.19849718e+00 8.76043916e-01 -1.39367074e-01 2.29225188e-01 3.25464278e-01 -2.02732682e-01 2.48498440e-01 -1.22194469e+00 -1.13773488e-01 -2.05283035e-02 4.63890433e-01 9.70208570e-02 4.38656807e-01 -4.79358375e-01 6.36211216e-01 2.85582859e-02 1.06264323e-01 8.53649974e-02 1.12381399e+00 -4.35379028e-01 -1.31324196e+00 -2.65359908e-01 7.48845577e-01 -3.51778746e-01 2.17675179e-01 -6.54576123e-02 5.24938405e-01 -2.29551829e-02 7.87203074e-01 -2.14386180e-01 -8.09170529e-02 1.53150529e-01 -1.92666307e-01 7.06061959e-01 -5.73084056e-01 -6.18308425e-01 -6.55689463e-02 2.44753331e-01 -1.14639652e+00 -1.36283666e-01 -4.72241521e-01 -1.18649983e+00 -2.07208350e-01 -4.25310522e-01 2.00282618e-01 5.36518633e-01 1.02019405e+00 7.67157273e-03 8.50336432e-01 1.08580554e+00 -9.32880819e-01 -1.09635186e+00 -3.34025443e-01 -3.75767171e-01 3.34606622e-03 2.59562105e-01 -9.12553668e-01 -1.45579249e-01 -5.91869593e-01]

[5.118056774139404, 2.949507236480713]

9f3a3e69-37d9-435a-a9a8-f6c6b7050c67

pointly-supervised-panoptic-segmentation

2210.13950

https://arxiv.org/abs/2210.13950v1

https://arxiv.org/pdf/2210.13950v1.pdf

Pointly-Supervised Panoptic Segmentation

In this paper, we propose a new approach to applying point-level annotations for weakly-supervised panoptic segmentation. Instead of the dense pixel-level labels used by fully supervised methods, point-level labels only provide a single point for each target as supervision, significantly reducing the annotation burden. We formulate the problem in an end-to-end framework by simultaneously generating panoptic pseudo-masks from point-level labels and learning from them. To tackle the core challenge, i.e., panoptic pseudo-mask generation, we propose a principled approach to parsing pixels by minimizing pixel-to-point traversing costs, which model semantic similarity, low-level texture cues, and high-level manifold knowledge to discriminate panoptic targets. We conduct experiments on the Pascal VOC and the MS COCO datasets to demonstrate the approach's effectiveness and show state-of-the-art performance in the weakly-supervised panoptic segmentation problem. Codes are available at https://github.com/BraveGroup/PSPS.git.

['Tieniu Tan', 'Zhaoxiang Zhang', 'Junsong Fan']

2022-10-25

['weakly-supervised-panoptic-segmentation', 'panoptic-segmentation']

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

[ 4.18473780e-01 2.97883768e-02 -4.08846259e-01 -6.15677774e-01 -1.26514328e+00 -9.54976380e-01 3.96985441e-01 6.93458393e-02 -1.81877509e-01 1.00731745e-01 -2.93207765e-01 -2.73578972e-01 3.38334978e-01 -8.09043884e-01 -7.57616520e-01 -6.64537072e-01 -1.13513889e-02 5.54474175e-01 5.13456762e-01 1.93164721e-01 1.48706570e-01 4.03344989e-01 -1.46958756e+00 3.35593969e-01 8.93515408e-01 9.25067127e-01 3.06957364e-01 7.60909677e-01 -1.84330732e-01 4.07683253e-01 -1.58270717e-01 -2.70550609e-01 8.90848517e-01 -4.63808000e-01 -7.51618564e-01 7.46508896e-01 1.16404021e+00 -9.63562503e-02 2.26902813e-02 1.26453555e+00 -8.70607495e-02 8.28705877e-02 5.19287109e-01 -7.64376581e-01 -6.21651895e-02 1.30794629e-01 -1.12103260e+00 1.52657330e-01 -3.59922141e-01 3.49724650e-01 1.54682219e+00 -7.51015007e-01 5.09651899e-01 1.26341701e+00 7.07056522e-01 3.56366694e-01 -1.57903159e+00 -3.58159453e-01 3.17268640e-01 -6.10835314e-01 -1.09276187e+00 -4.04167920e-01 6.17076337e-01 -6.43630981e-01 5.19546270e-01 3.07344258e-01 4.17797327e-01 2.41921067e-01 -2.49605969e-01 8.85805786e-01 1.40857041e+00 -2.58727789e-01 -8.51414725e-03 1.85999597e-05 7.19838887e-02 1.00513601e+00 -3.08581162e-02 2.06320301e-01 -2.33815730e-01 -9.00656357e-02 9.03423309e-01 -2.06327930e-01 -7.62222633e-02 -5.81527293e-01 -1.13300836e+00 9.65387106e-01 6.42823339e-01 -1.55317783e-01 -1.02976888e-01 1.75742179e-01 3.84770148e-02 -2.25778203e-03 1.05615985e+00 4.27574754e-01 -7.28487074e-01 3.69659632e-01 -1.25137413e+00 1.19249113e-01 7.07782388e-01 7.43938088e-01 1.10540223e+00 -2.35384732e-01 4.82778214e-02 9.55799818e-01 4.20519501e-01 6.91838980e-01 -2.47260630e-01 -1.24492502e+00 5.96427560e-01 4.58280206e-01 1.91215098e-01 -6.59539104e-01 -3.15191537e-01 -5.10378778e-01 -2.16142654e-01 3.94388318e-01 6.24224246e-01 -2.05524549e-01 -1.47427070e+00 1.45673776e+00 7.71366835e-01 3.52657109e-01 -1.12443060e-01 1.14278507e+00 5.91727018e-01 9.81559753e-01 6.34605363e-02 1.79244787e-01 1.40752566e+00 -1.45173705e+00 -5.57625033e-02 -6.41120791e-01 5.89459181e-01 -1.01900697e+00 1.27137053e+00 7.85733089e-02 -9.61268961e-01 -5.29494047e-01 -7.31316328e-01 -6.48072958e-02 -1.90856621e-01 3.02714080e-01 5.13967097e-01 3.53403211e-01 -1.01788521e+00 4.87960786e-01 -9.55074847e-01 -3.49091023e-01 4.75008398e-01 5.37317656e-02 -1.01133483e-02 4.82850075e-02 -6.49344742e-01 3.62325937e-01 4.46314275e-01 -3.62321181e-04 -9.96048272e-01 -8.40999961e-01 -9.18143690e-01 5.37628029e-03 5.83539486e-01 -4.00159687e-01 1.45143044e+00 -9.91482913e-01 -1.31133783e+00 1.41698813e+00 -1.14589073e-01 -3.87308896e-01 3.64259988e-01 -2.09724456e-01 1.00394733e-01 5.91554940e-01 4.33525503e-01 1.48187649e+00 8.87643695e-01 -1.56526136e+00 -1.14263284e+00 -2.26842791e-01 3.54480706e-02 5.58151305e-01 1.82478324e-01 -8.68511572e-02 -9.20959711e-01 -5.44679582e-01 3.45634967e-01 -1.24011207e+00 -5.58985651e-01 4.53644186e-01 -5.48887372e-01 5.12228571e-02 8.46753061e-01 -5.71771324e-01 5.64199746e-01 -2.17815995e+00 -5.61602972e-02 5.72342910e-02 6.40163571e-02 1.88158035e-01 -3.33571821e-01 -1.12796783e-01 3.05244058e-01 5.31841032e-02 -1.07200098e+00 -6.17229044e-01 -2.08893657e-01 3.14501911e-01 -3.83925676e-01 4.23242152e-01 4.58167821e-01 7.52613783e-01 -9.28305924e-01 -6.34124815e-01 5.53761542e-01 1.31884784e-01 -4.09568638e-01 3.56903851e-01 -9.88709390e-01 6.71862423e-01 -4.15727109e-01 8.53560209e-01 8.00568819e-01 -3.08652252e-01 -1.17201120e-01 -4.29885462e-02 -4.13629591e-01 3.19765031e-01 -8.61862719e-01 1.74586177e+00 -4.80067670e-01 6.63258433e-01 6.28204823e-01 -7.76726067e-01 8.35716665e-01 -1.29004732e-01 3.56904477e-01 -3.80196333e-01 -2.39941567e-01 2.22378984e-01 -2.78765470e-01 -2.69903392e-01 3.57019037e-01 -1.22671880e-01 -1.38742253e-01 2.59785116e-01 -7.08609447e-02 -9.06131625e-01 2.69876868e-01 3.64480689e-02 5.97681522e-01 4.66845125e-01 -5.59804216e-02 -4.79098499e-01 2.65746742e-01 7.57451117e-01 5.79492509e-01 6.39503896e-01 -1.69438243e-01 9.84505951e-01 4.32475060e-01 -2.49680340e-01 -9.58997488e-01 -1.13728845e+00 -3.58117700e-01 1.18296731e+00 2.82626152e-01 -1.22911505e-01 -8.49406898e-01 -8.37839663e-01 5.27454428e-02 4.66252655e-01 -5.72875917e-01 6.74347460e-01 -4.96885270e-01 -1.01576698e+00 4.06620383e-01 3.26905876e-01 6.57664239e-01 -7.77688563e-01 -6.94343626e-01 -2.74637416e-02 -2.08888412e-01 -1.40648890e+00 -7.40462482e-01 3.98131341e-01 -1.05404258e+00 -1.10990846e+00 -6.73500597e-01 -9.13922608e-01 7.18095303e-01 6.04715228e-01 1.25723219e+00 -2.28666961e-01 -4.04630721e-01 2.19934464e-01 -1.64690197e-01 -6.14867769e-02 -1.22905932e-01 1.41793430e-01 -6.32715881e-01 -4.10014838e-02 2.23199561e-01 -1.89205006e-01 -7.51580536e-01 5.69719851e-01 -8.98439467e-01 2.43326351e-01 7.26203203e-01 5.36742508e-01 1.23100519e+00 -2.06704110e-01 -1.00897692e-01 -1.13404739e+00 -1.75771818e-01 -2.56470650e-01 -1.18870628e+00 9.58474129e-02 -2.84053683e-01 -1.51459202e-01 2.61372775e-01 -2.36575189e-03 -1.15730882e+00 6.22000635e-01 -1.23687550e-01 -4.29062009e-01 -4.91244316e-01 2.89368272e-01 -6.51140362e-02 -2.23510429e-01 5.55458724e-01 9.01895612e-02 -1.71456695e-01 -6.55756056e-01 8.21539462e-01 5.32203674e-01 7.31678009e-01 -4.96509224e-01 1.02809262e+00 9.31208909e-01 -1.31643862e-01 -9.48129892e-01 -1.66862738e+00 -1.01523793e+00 -8.46015453e-01 -2.98968814e-02 1.25187826e+00 -1.19004607e+00 1.08944096e-01 4.19171005e-01 -8.14517856e-01 -9.72003460e-01 -4.38832283e-01 2.68004537e-01 -5.71890175e-01 3.36539000e-01 -6.93458438e-01 -4.61921841e-01 -3.58559936e-01 -1.07741487e+00 1.56494510e+00 2.62357205e-01 1.69420227e-01 -1.08855534e+00 2.23968402e-01 7.07073212e-01 5.02311327e-02 3.24514449e-01 5.78544974e-01 -2.20452666e-01 -8.81916344e-01 2.59562414e-02 -6.64375722e-01 4.77394164e-01 -1.20000303e-01 -8.09905455e-02 -1.06317675e+00 -6.27978984e-03 8.61885175e-02 -4.91191119e-01 1.19799554e+00 5.31960070e-01 1.14652002e+00 -1.16767548e-01 -2.73404002e-01 1.06908476e+00 1.59170818e+00 -2.20387354e-01 2.96895981e-01 1.13074265e-01 9.70702589e-01 9.82148707e-01 9.53163683e-01 -1.04486952e-02 2.99555570e-01 6.80905938e-01 4.84028995e-01 -5.07022560e-01 -3.71775180e-01 -3.67877245e-01 1.50793254e-01 3.01134169e-01 4.62929547e-01 -3.00008714e-01 -1.02399194e+00 7.51122952e-01 -1.81742883e+00 -5.14585376e-01 -5.28796136e-01 2.05888534e+00 8.79455864e-01 6.20663352e-02 5.49426023e-03 -4.37229306e-01 7.72512794e-01 5.54175496e-01 -5.20497024e-01 -3.21075991e-02 -1.88797563e-01 1.04505830e-01 9.57405806e-01 9.37024593e-01 -1.63979340e+00 1.57105792e+00 5.47174978e+00 8.62417579e-01 -1.20861709e+00 2.59278208e-01 1.12556255e+00 5.30417934e-02 -9.36594680e-02 3.27438802e-01 -1.00242245e+00 2.99287736e-01 6.12993598e-01 6.53549492e-01 2.42066041e-01 7.73175836e-01 2.73807198e-01 -3.69826436e-01 -7.69745171e-01 6.72992826e-01 -8.25784206e-02 -1.19193017e+00 -4.42999750e-02 -4.51768674e-02 1.20941103e+00 7.72286534e-01 2.57761795e-02 -2.09899396e-01 4.48947847e-01 -7.70802796e-01 5.50562978e-01 -1.17708631e-01 8.69769931e-01 -2.68089086e-01 2.40704522e-01 8.07075649e-02 -1.42219114e+00 2.70268887e-01 -3.78617883e-01 3.15175653e-01 2.16488302e-01 7.75533020e-01 -5.67739546e-01 2.99184740e-01 8.05446506e-01 7.99016058e-01 -5.30887842e-01 1.06899118e+00 -5.78475177e-01 9.43372667e-01 -7.15959072e-01 5.74760377e-01 9.00173068e-01 -5.49938858e-01 6.35904372e-01 1.50245821e+00 2.71462165e-02 9.67472568e-02 5.39603770e-01 9.56002414e-01 -1.45712942e-01 5.83153069e-02 -3.66882682e-01 1.27422079e-01 1.61224648e-01 1.66012716e+00 -1.20150709e+00 -3.90126199e-01 -4.07730311e-01 9.37364459e-01 1.17765971e-01 3.12134475e-01 -5.82311094e-01 -1.86605260e-01 5.22138715e-01 2.05368355e-01 5.06674528e-01 -2.60913789e-01 -7.27223992e-01 -1.14839339e+00 -2.93986294e-02 -6.46983325e-01 5.59102535e-01 -7.88997233e-01 -1.19177866e+00 4.08349484e-01 1.59871116e-01 -1.24895108e+00 2.73579150e-01 -5.61130226e-01 -8.72734487e-01 8.67933035e-01 -1.89270234e+00 -1.34611535e+00 -3.94524306e-01 1.32462591e-01 8.77085268e-01 4.97393668e-01 4.40661460e-01 2.49027380e-05 -3.27400476e-01 -7.24384487e-02 1.53407991e-01 1.35244280e-01 6.22612476e-01 -1.50603545e+00 5.97408235e-01 1.11555338e+00 3.69469792e-01 -2.50116363e-02 5.64498663e-01 -4.94810820e-01 -8.96781862e-01 -1.46016443e+00 6.71410620e-01 -4.89353955e-01 8.43314052e-01 -4.35493410e-01 -9.03692007e-01 5.14979184e-01 8.57647285e-02 2.07460091e-01 3.62379789e-01 -1.58037260e-01 -2.93368250e-01 -4.13517505e-02 -1.04529965e+00 5.77225447e-01 9.44740236e-01 -3.97398144e-01 -4.81102556e-01 6.87546909e-01 7.34203696e-01 -5.28357983e-01 -5.91034770e-01 5.95118046e-01 2.74693295e-02 -7.50977457e-01 9.06977177e-01 6.02461211e-03 5.48757017e-01 -6.62113249e-01 -2.56612897e-02 -1.28453803e+00 6.41795844e-02 -5.93132138e-01 2.71887749e-01 1.09122801e+00 8.03675830e-01 -4.10320133e-01 1.04095125e+00 1.54879540e-01 -3.19274664e-01 -6.80094123e-01 -7.72562921e-01 -6.13901138e-01 1.41597882e-01 -2.13283151e-01 8.91660750e-02 9.05028462e-01 -6.00316346e-01 4.45965022e-01 -2.93922693e-01 6.21967375e-01 9.49364483e-01 7.91251302e-01 7.85948694e-01 -1.09700382e+00 -4.63003874e-01 -4.85247731e-01 9.61827394e-03 -1.56802225e+00 3.45093459e-02 -9.64170337e-01 6.63120806e-01 -1.54225671e+00 1.00493364e-01 -8.75179946e-01 4.14839238e-02 5.23032606e-01 -3.13546777e-01 6.94334507e-01 4.67187464e-02 5.82763672e-01 -5.43315828e-01 3.02561581e-01 1.28772795e+00 -1.44588545e-01 -4.56871748e-01 9.54382936e-04 -4.53957796e-01 8.57850611e-01 1.19314289e+00 -6.46251559e-01 -3.84249359e-01 -8.25057149e-01 -2.38891184e-01 -3.41372751e-03 5.27996302e-01 -8.36644888e-01 -1.05780266e-01 -3.38455528e-01 9.78209525e-02 -7.07204759e-01 5.02074480e-01 -5.51661611e-01 -3.69577557e-01 3.35253388e-01 -3.52275103e-01 -2.42743626e-01 3.57063442e-01 6.91250682e-01 -2.85163343e-01 -2.39362612e-01 1.19886672e+00 -2.84603298e-01 -7.33087003e-01 5.62596083e-01 3.49153578e-02 3.36679727e-01 9.71808612e-01 1.55827522e-01 -3.59990150e-01 -2.11583413e-02 -6.76644802e-01 5.96739709e-01 7.67030180e-01 2.29288459e-01 2.74399042e-01 -4.63172406e-01 -8.00279915e-01 8.44524428e-02 2.25033090e-01 5.98448455e-01 1.28042921e-02 8.73743176e-01 -1.04570794e+00 3.90250176e-01 2.11654305e-01 -1.08386564e+00 -1.25134706e+00 1.67473897e-01 5.59543610e-01 -1.49318680e-01 -7.83714533e-01 1.04423034e+00 8.79886150e-01 -7.42024302e-01 6.79823384e-02 -4.62460697e-01 1.84377924e-01 -5.63299023e-02 8.55450556e-02 -1.29328862e-01 -2.78642654e-01 -5.94545364e-01 -6.96984828e-02 9.80147541e-01 1.49403512e-01 -5.05987227e-01 1.17491114e+00 -1.94230333e-01 -1.64326951e-01 4.57630873e-01 9.72627163e-01 -6.14143535e-02 -1.85882795e+00 -4.07285511e-01 5.23989983e-02 -6.29646242e-01 2.49386996e-01 -8.58690202e-01 -1.28719866e+00 1.15560865e+00 5.95270216e-01 -7.32569322e-02 8.37828338e-01 3.40145648e-01 8.43141437e-01 2.29144558e-01 -5.25313094e-02 -9.69456434e-01 -7.61651993e-02 3.83099288e-01 4.19393361e-01 -1.51029289e+00 -1.53431579e-01 -9.33079243e-01 -6.92294002e-01 7.79202938e-01 6.85408175e-01 -2.33560741e-01 5.68271935e-01 1.24099411e-01 5.60638011e-01 -3.64692539e-01 -5.10433257e-01 -6.59399807e-01 4.53398734e-01 3.39155942e-01 2.60201186e-01 2.52840042e-01 4.90974933e-02 -3.83122593e-01 1.63539678e-01 -4.29937214e-01 1.75937369e-01 7.99789131e-01 -9.18626308e-01 -9.19808626e-01 -5.28875113e-01 4.88050401e-01 -4.15872335e-01 -4.09426928e-01 -4.88035560e-01 6.07325137e-01 8.37723091e-02 1.05712867e+00 4.31709439e-01 3.98200788e-02 -2.76900567e-02 -7.30917379e-02 2.30027974e-01 -1.03766322e+00 -2.71004260e-01 4.83779907e-01 2.15224907e-01 -3.96342605e-01 -6.64955258e-01 -6.42518103e-01 -1.20288289e+00 1.23831801e-01 -1.30663753e-01 3.36203694e-01 5.12997389e-01 5.73426187e-01 1.79799974e-01 -7.69340759e-03 7.03754425e-01 -1.05544949e+00 -3.51501644e-01 -9.01805580e-01 -2.74482220e-01 3.69979531e-01 2.65742570e-01 -2.49669850e-01 -5.36705256e-01 3.00085336e-01]

[9.459921836853027, 0.4217296540737152]

ff9383ce-601c-4a17-a34c-9eeb753dbee4

user-based-aggregation-for-biterm-topic-model

https://aclanthology.org/P15-2080

https://aclanthology.org/P15-2080.pdf

User Based Aggregation for Biterm Topic Model

['Jinpeng Wang', 'Yan Zhang', 'Hongfei Yan', 'Xiaoming Li', 'Weizheng Chen']

2015-07-01

user-based-aggregation-for-biterm-topic-model-1

https://aclanthology.org/P15-2080

https://aclanthology.org/P15-2080.pdf

ijcnlp-2015-7

['product-recommendation']

['miscellaneous']

[-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]

[-7.406922340393066, 3.593177556991577]

8a8fa4a6-2508-481e-8eba-69240fff43fa

hop-count-based-self-supervised-anomaly

2104.07917

https://arxiv.org/abs/2104.07917v4

https://arxiv.org/pdf/2104.07917v4.pdf

Hop-Count Based Self-Supervised Anomaly Detection on Attributed Networks

Recent years have witnessed an upsurge of interest in the problem of anomaly detection on attributed networks due to its importance in both research and practice. Although various approaches have been proposed to solve this problem, two major limitations exist: (1) unsupervised approaches usually work much less efficiently due to the lack of supervisory signal, and (2) existing anomaly detection methods only use local contextual information to detect anomalous nodes, e.g., one- or two-hop information, but ignore the global contextual information. Since anomalous nodes differ from normal nodes in structures and attributes, it is intuitive that the distance between anomalous nodes and their neighbors should be larger than that between normal nodes and their neighbors if we remove the edges connecting anomalous and normal nodes. Thus, hop counts based on both global and local contextual information can be served as the indicators of anomaly. Motivated by this intuition, we propose a hop-count based model (HCM) to detect anomalies by modeling both local and global contextual information. To make better use of hop counts for anomaly identification, we propose to use hop counts prediction as a self-supervised task. We design two anomaly scores based on the hop counts prediction via HCM model to identify anomalies. Besides, we employ Bayesian learning to train HCM model for capturing uncertainty in learned parameters and avoiding overfitting. Extensive experiments on real-world attributed networks demonstrate that our proposed model is effective in anomaly detection.

['Mykola Pechenizkiy', 'Vlado Menkovski', 'Yulong Pei', 'Tianjin Huang']

2021-04-16

['self-supervised-anomaly-detection', 'supervised-anomaly-detection']

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

[ 5.13678081e-02 1.05472073e-01 -2.39387751e-01 -6.63991272e-01 -1.26341060e-01 -1.14089407e-01 4.03282911e-01 6.61193311e-01 -1.56199142e-01 6.56400859e-01 -8.70016366e-02 -1.95889473e-01 -4.96184975e-01 -1.08020782e+00 -3.98467898e-01 -6.68636084e-01 -4.33189631e-01 3.62913162e-01 6.87153995e-01 -6.36194879e-03 4.35026824e-01 4.48163480e-01 -1.43026853e+00 -2.06310436e-01 1.08331203e+00 1.02713621e+00 -3.01060736e-01 1.16822183e-01 -3.40519726e-01 6.16854250e-01 -5.72875321e-01 -2.98161618e-02 1.31873295e-01 -4.63098317e-01 -7.23420024e-01 1.59825217e-02 -1.15185961e-01 -3.19905251e-01 -2.89436847e-01 1.24868405e+00 -1.97880249e-02 2.41308063e-01 6.10711217e-01 -1.89280808e+00 -3.92836452e-01 6.88048601e-01 -8.13890934e-01 4.68537450e-01 2.19830915e-01 -9.12006423e-02 1.21159995e+00 -4.99006569e-01 6.39203191e-02 1.12119889e+00 5.74933946e-01 2.62401432e-01 -1.09020185e+00 -7.11315155e-01 6.75252557e-01 5.12530386e-01 -1.52656043e+00 -1.64448768e-01 9.51847315e-01 -6.91145882e-02 5.92675149e-01 3.42539966e-01 6.05468810e-01 1.00438321e+00 7.72271072e-03 5.74016571e-01 7.12748528e-01 -4.23557907e-01 3.97940993e-01 1.35912478e-01 4.06229854e-01 5.71309745e-01 4.35136020e-01 -8.95530134e-02 -2.81096667e-01 -5.87251246e-01 4.76860493e-01 5.14464498e-01 -3.49242017e-02 -1.66289851e-01 -8.86177599e-01 8.68746281e-01 3.06094557e-01 3.74276638e-01 -3.63611072e-01 -4.69013192e-02 3.83189261e-01 3.31354380e-01 4.46852684e-01 1.14857323e-01 -2.85254270e-01 1.54384281e-02 -6.16369069e-01 -9.06131938e-02 6.67264462e-01 7.81883240e-01 9.66125727e-01 2.79326830e-02 8.79495312e-03 8.90018821e-01 7.21503317e-01 1.78853080e-01 3.74695301e-01 -6.27951980e-01 2.93923020e-01 1.02455711e+00 -1.48324445e-01 -1.58686996e+00 -2.91331828e-01 -4.25211281e-01 -1.00510335e+00 -1.36626467e-01 2.85647810e-01 4.06928584e-02 -7.07731128e-01 1.70288360e+00 3.16275597e-01 8.35042655e-01 -1.87604755e-01 6.31180108e-01 3.74720991e-01 5.72102249e-01 1.13478862e-01 -3.29438627e-01 1.00246835e+00 -5.18305004e-01 -6.66111112e-01 -1.46160141e-01 8.71464372e-01 -2.12718949e-01 7.75961399e-01 2.41706714e-01 -6.13569856e-01 1.02925394e-02 -9.50908542e-01 7.64747858e-01 -4.49046135e-01 -4.39012289e-01 3.69026899e-01 5.65814257e-01 -5.70363760e-01 5.63774586e-01 -1.05538332e+00 -4.07916516e-01 3.66540730e-01 2.15463281e-01 -1.21360809e-01 1.19466946e-01 -1.31335616e+00 4.16416347e-01 7.20467746e-01 2.09404156e-01 -6.54344857e-01 -2.46864930e-01 -6.83653474e-01 2.04649255e-01 7.91930079e-01 2.94982065e-02 8.29198599e-01 -6.72358692e-01 -9.41482067e-01 2.36590356e-01 -2.13198632e-01 -4.35947180e-01 3.43148112e-01 8.06886479e-02 -9.16028261e-01 1.37359351e-01 1.96548402e-01 -9.33375861e-03 6.36956275e-01 -1.23631120e+00 -1.00398076e+00 -4.80291307e-01 -3.02495748e-01 -1.16425388e-01 -8.19177568e-01 -2.11426362e-01 -3.38717937e-01 -5.17398119e-01 6.14477873e-01 -6.59653068e-01 -2.87248671e-01 -2.63150126e-01 -8.87145281e-01 -4.28439617e-01 1.42900562e+00 -2.20510840e-01 1.61933887e+00 -2.06757140e+00 -5.44668853e-01 1.06690037e+00 4.94874984e-01 1.58175185e-01 1.40385330e-01 3.39776784e-01 1.99317392e-02 2.39244908e-01 -5.13434887e-01 -5.70547804e-02 -1.24848559e-01 6.56875432e-01 -2.93291330e-01 4.07234848e-01 8.37894231e-02 2.55463481e-01 -1.04262888e+00 -6.03289366e-01 1.34359673e-01 9.61513221e-02 -2.78096139e-01 7.40568861e-02 9.75605175e-02 5.23863614e-01 -8.29800844e-01 7.95731068e-01 6.64830804e-01 -2.63308942e-01 9.97203663e-02 1.38548493e-01 1.24894045e-01 -8.18959996e-03 -1.49698210e+00 8.30376148e-01 -9.87543166e-03 2.10953102e-01 -8.93538371e-02 -1.43706226e+00 1.13828385e+00 3.73223126e-01 6.67272627e-01 -4.64803755e-01 -5.52194417e-02 1.94834232e-01 6.83308616e-02 -4.02817339e-01 8.84716064e-02 5.95508367e-02 1.15911644e-02 4.54248160e-01 -2.56381005e-01 6.80733025e-01 2.07839757e-01 2.82815546e-01 1.40462101e+00 -5.60750425e-01 1.88201442e-01 -6.14001490e-02 8.61550927e-01 -3.14216346e-01 1.22549558e+00 8.97892833e-01 -4.13002253e-01 2.57633120e-01 9.24626052e-01 -4.04871464e-01 -5.95331669e-01 -9.86228943e-01 -1.80208862e-01 1.13159990e+00 4.23265398e-01 -5.68772614e-01 -4.66808647e-01 -1.13303673e+00 3.08174472e-02 7.54812956e-01 -4.37462628e-01 -4.36289430e-01 -4.86465335e-01 -9.96593356e-01 7.74707913e-01 5.97080290e-01 5.44507384e-01 -1.02422106e+00 -2.63451189e-01 2.54937738e-01 -8.71727988e-02 -1.00742888e+00 -1.56410038e-01 6.79832697e-02 -9.88228381e-01 -1.17305398e+00 9.35681611e-02 -4.06535268e-01 9.40023661e-01 2.25694731e-01 7.25903332e-01 5.39611518e-01 -1.18973605e-01 3.39267343e-01 -4.74619448e-01 -3.71969402e-01 -2.76080042e-01 5.02775237e-02 2.00450882e-01 3.98313105e-01 6.77814484e-01 -9.45822299e-01 -4.85375702e-01 6.82638526e-01 -9.96339798e-01 -6.08008802e-01 6.65086150e-01 6.24955654e-01 5.40952206e-01 4.55047816e-01 8.33746314e-01 -1.16003799e+00 7.51016259e-01 -8.97384405e-01 -5.43945730e-01 2.40781009e-01 -1.22991943e+00 4.56790673e-03 6.76102459e-01 -1.13445111e-01 -1.03887820e+00 -2.84325033e-01 3.53605375e-02 -2.43001029e-01 -5.16740441e-01 5.99028647e-01 -2.93071330e-01 1.74990937e-01 3.03778857e-01 1.87447861e-01 -1.03399001e-01 -3.31334859e-01 -1.75378010e-01 6.37742400e-01 3.53099287e-01 -6.76124036e-01 8.00957739e-01 4.55059648e-01 3.21075499e-01 -9.04174149e-01 -7.31396139e-01 -6.02698743e-01 -5.43485940e-01 -1.14695728e-01 4.89006490e-01 -3.95682096e-01 -7.46985495e-01 4.44620490e-01 -8.02479267e-01 2.87347585e-01 -2.68846042e-02 3.88985604e-01 6.17730245e-02 7.84348309e-01 -4.82498646e-01 -1.10041428e+00 -3.52155380e-02 -9.10740793e-01 5.12260020e-01 2.36904398e-01 -2.36017063e-01 -1.32321656e+00 -4.95206863e-02 -8.54604095e-02 3.55223507e-01 3.75175923e-01 1.07354891e+00 -1.36310399e+00 -4.92569983e-01 -6.21478796e-01 -3.01596403e-01 3.17391634e-01 2.81398475e-01 1.72636703e-01 -7.11901426e-01 -1.58920094e-01 -4.84049469e-02 9.44385678e-02 6.80691183e-01 6.46957159e-02 1.44484866e+00 -5.18686116e-01 -6.19580746e-01 3.11263174e-01 1.15737331e+00 2.17144340e-01 4.65858549e-01 4.59305018e-01 8.08765590e-01 6.28234625e-01 6.02224171e-01 7.45861590e-01 2.93569505e-01 3.13073844e-01 8.02252948e-01 2.21985713e-01 6.92321658e-01 -3.40793073e-01 2.21578717e-01 7.74711132e-01 -1.11713968e-01 -2.93277949e-01 -1.12725341e+00 5.33416629e-01 -2.14973021e+00 -9.22112048e-01 -4.10628438e-01 2.41248250e+00 3.98790330e-01 5.26793897e-01 1.67423904e-01 4.55325693e-01 1.08372223e+00 1.60183445e-01 -5.94531715e-01 -1.14808589e-01 -2.91579822e-03 -3.86474550e-01 3.42998385e-01 2.52548158e-01 -9.52277601e-01 4.99763936e-01 5.83968019e+00 7.00059235e-01 -6.59075797e-01 -1.99785262e-01 5.80334544e-01 2.26749361e-01 -1.44029900e-01 3.36557448e-01 -7.37981975e-01 8.14039826e-01 8.70521963e-01 2.30527632e-02 3.12038660e-02 8.84891212e-01 2.02438682e-01 2.82250461e-03 -1.04273105e+00 6.54111564e-01 -1.95631489e-01 -7.06362963e-01 7.66142309e-02 2.38005519e-01 4.99504447e-01 -2.45154083e-01 -2.34515756e-01 2.53709346e-01 2.54328191e-01 -7.66873777e-01 1.53472900e-01 5.87110758e-01 6.36264235e-02 -1.02068949e+00 9.92295206e-01 5.35321772e-01 -1.31308424e+00 -2.92585939e-01 -2.88799912e-01 -9.20944735e-02 -9.33191255e-02 1.00500810e+00 -7.64123440e-01 5.07037640e-01 7.44475722e-01 8.69030416e-01 -5.22625327e-01 1.03956473e+00 -2.68109024e-01 8.69613886e-01 -5.45148671e-01 2.58317232e-01 3.13934982e-01 -3.82892609e-01 7.31337547e-01 9.04667795e-01 3.86350602e-01 -6.43304437e-02 4.20511991e-01 7.44592130e-01 2.27716677e-02 1.28339902e-01 -4.36104715e-01 1.21190965e-01 1.08140278e+00 1.23936510e+00 -9.76794660e-01 -1.68486714e-01 -6.19485199e-01 5.47605276e-01 1.01857759e-01 3.44237953e-01 -6.62897050e-01 -7.09573925e-01 5.67802668e-01 2.30860472e-01 -5.32586835e-02 8.99068639e-02 -1.67314991e-01 -8.96568298e-01 9.36714932e-02 -5.33420146e-01 7.56449521e-01 -6.63857087e-02 -1.68709707e+00 3.19662839e-01 9.44549143e-02 -1.30399024e+00 -1.67065188e-01 -3.15112144e-01 -1.19380403e+00 4.56737965e-01 -1.32203114e+00 -7.11782813e-01 -5.46696603e-01 7.38937140e-01 9.51187015e-02 -1.68739960e-01 6.92504406e-01 1.81795627e-01 -1.01566923e+00 7.03865588e-01 1.23604275e-01 5.28149843e-01 7.40713537e-01 -1.21153581e+00 -6.97156563e-02 9.75606263e-01 -3.14054303e-02 5.53436279e-01 6.73145652e-01 -8.07641447e-01 -7.80253232e-01 -1.22789335e+00 5.64780354e-01 -1.62816957e-01 6.96843207e-01 2.32757330e-02 -1.42295480e+00 8.21415842e-01 -3.42418939e-01 3.78079772e-01 7.26415336e-01 4.04032022e-01 -3.33991230e-01 -3.68742704e-01 -1.28597546e+00 4.09730762e-01 1.01929080e+00 -2.06047535e-01 -4.77527380e-01 -7.58528302e-04 3.72700512e-01 1.75948888e-01 -9.03716207e-01 6.62781715e-01 1.80642068e-01 -1.20891559e+00 6.17728591e-01 -3.88481081e-01 -1.21597294e-02 -4.16753888e-01 -8.40498209e-02 -1.23208570e+00 -1.69917077e-01 -1.80125296e-01 -5.27995765e-01 1.54592991e+00 3.05711240e-01 -1.14409161e+00 8.53077173e-01 5.71524501e-01 -7.55063742e-02 -7.80181825e-01 -9.14357841e-01 -8.35581362e-01 -4.39987868e-01 -5.43619037e-01 8.32261622e-01 1.16713905e+00 1.31307080e-01 1.77805871e-01 -3.97263736e-01 5.41805565e-01 7.12603867e-01 -1.06249742e-01 4.36006546e-01 -2.04956102e+00 -6.38253316e-02 -4.83668327e-01 -7.27113366e-01 -7.21186817e-01 2.43652418e-01 -7.03620851e-01 2.76629347e-02 -1.15072620e+00 1.42855644e-01 -6.48288786e-01 -7.75262952e-01 7.39072323e-01 -1.98557302e-01 4.40843478e-02 -4.35857713e-01 5.97296953e-01 -8.50625753e-01 5.59985220e-01 6.55666292e-01 1.53841466e-01 -3.57741117e-01 1.77753270e-01 -6.12763584e-01 1.12486184e+00 8.64049315e-01 -6.89291835e-01 -4.63434160e-01 -4.21838909e-02 2.33038604e-01 -8.34623072e-03 3.51657420e-01 -9.91489351e-01 3.09059858e-01 -2.16943339e-01 2.72259176e-01 -5.97599626e-01 2.89642196e-02 -1.03610337e+00 -4.01055962e-01 2.77138859e-01 -3.29696774e-01 1.35010973e-01 -2.85568535e-01 1.22326505e+00 -4.51274186e-01 -2.76034594e-01 4.79686201e-01 4.53327075e-02 -7.73753464e-01 4.71627206e-01 -4.55161005e-01 3.48001160e-03 1.18910277e+00 -2.61597753e-01 -3.08734179e-01 -4.10618544e-01 -8.03030849e-01 6.08743012e-01 2.53228128e-01 2.13818386e-01 6.93592370e-01 -1.43089151e+00 -4.99463201e-01 3.37293208e-01 4.43057865e-01 9.40130427e-02 7.27236569e-02 1.09162903e+00 -1.57679170e-01 1.77096695e-01 -4.39853854e-02 -7.33121037e-01 -1.12570667e+00 2.96232790e-01 3.92152034e-02 -2.75937229e-01 -6.35332406e-01 3.88590723e-01 -1.16435990e-01 -4.68340427e-01 4.68502104e-01 4.30813991e-02 -3.42099160e-01 -1.17282785e-01 4.42905098e-01 8.48381698e-01 -3.01108547e-02 -5.29189169e-01 -4.27831322e-01 2.90596992e-01 -3.56600910e-01 2.16705784e-01 1.17023826e+00 -3.25873435e-01 -3.83008033e-01 6.60765171e-01 9.06118512e-01 -4.82227765e-02 -8.22323918e-01 -5.37827313e-01 5.46418071e-01 -6.28882945e-01 -7.90221151e-03 -3.04055065e-01 -1.21153736e+00 6.74149334e-01 4.08941388e-01 8.20134640e-01 1.20555007e+00 -4.75489907e-02 7.83820450e-01 5.49679518e-01 1.25234336e-01 -1.22513390e+00 1.40216336e-01 5.25798917e-01 1.25058919e-01 -1.25441444e+00 -1.51986852e-01 -4.92400050e-01 -4.42803800e-01 1.08771908e+00 9.78614688e-01 -9.93242115e-02 7.98849881e-01 -1.01817735e-01 -2.05143839e-01 -1.87506557e-01 -6.16725802e-01 1.54165566e-01 3.20793279e-02 5.04774749e-01 2.57772386e-01 -9.89511609e-02 -1.11566640e-01 5.49312890e-01 2.63881087e-01 -5.51034570e-01 5.76166332e-01 1.11410511e+00 -8.09569478e-01 -1.16746926e+00 -5.17197430e-01 8.39286387e-01 -4.67668593e-01 2.83681333e-01 -2.58837909e-01 5.75506568e-01 1.53186738e-01 1.21279931e+00 2.04109743e-01 -4.31396723e-01 2.90552467e-01 2.66303927e-01 -2.14898065e-01 -5.55475831e-01 7.75101408e-02 -1.32441014e-01 -1.68274269e-01 -5.24902701e-01 -1.98459178e-01 -4.83683765e-01 -1.48731458e+00 -4.34359193e-01 -5.22951722e-01 5.00107467e-01 2.39027187e-01 1.29373813e+00 3.45344514e-01 5.26326537e-01 8.79647732e-01 -1.64234102e-01 -6.12264216e-01 -9.31838274e-01 -1.05334878e+00 5.26800990e-01 2.59782702e-01 -7.75229335e-01 -7.16742933e-01 -4.67208385e-01]

[7.4605278968811035, 2.676189422607422]

a73a1237-ce26-4143-99f2-cdb2de45cb78

acsnet-action-context-separation-network-for

2103.15088

https://arxiv.org/abs/2103.15088v1

https://arxiv.org/pdf/2103.15088v1.pdf

ACSNet: Action-Context Separation Network for Weakly Supervised Temporal Action Localization

The object of Weakly-supervised Temporal Action Localization (WS-TAL) is to localize all action instances in an untrimmed video with only video-level supervision. Due to the lack of frame-level annotations during training, current WS-TAL methods rely on attention mechanisms to localize the foreground snippets or frames that contribute to the video-level classification task. This strategy frequently confuse context with the actual action, in the localization result. Separating action and context is a core problem for precise WS-TAL, but it is very challenging and has been largely ignored in the literature. In this paper, we introduce an Action-Context Separation Network (ACSNet) that explicitly takes into account context for accurate action localization. It consists of two branches (i.e., the Foreground-Background branch and the Action-Context branch). The Foreground- Background branch first distinguishes foreground from background within the entire video while the Action-Context branch further separates the foreground as action and context. We associate video snippets with two latent components (i.e., a positive component and a negative component), and their different combinations can effectively characterize foreground, action and context. Furthermore, we introduce extended labels with auxiliary context categories to facilitate the learning of action-context separation. Experiments on THUMOS14 and ActivityNet v1.2/v1.3 datasets demonstrate the ACSNet outperforms existing state-of-the-art WS-TAL methods by a large margin.

['Gang Hua', 'Nanning Zheng', 'Junsong Yuan', 'Wei Tang', 'Qilin Zhang', 'Le Wang', 'Ziyi Liu']

2021-03-28

['weakly-supervised-action-localization', 'weakly-supervised-temporal-action', 'video-polyp-segmentation']

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

[ 4.09824401e-01 -2.19691485e-01 -7.37894356e-01 -1.05787441e-01 -6.28694355e-01 -3.83989125e-01 5.86004078e-01 -1.59414709e-01 -4.09689277e-01 5.58442295e-01 3.96182239e-01 -9.87955406e-02 1.92332551e-01 -2.78211981e-01 -5.86095929e-01 -1.11820149e+00 9.50635877e-03 -1.58098161e-01 8.36034298e-01 3.13295007e-01 1.85338885e-03 9.27459523e-02 -1.32092309e+00 8.06209683e-01 5.82924783e-01 1.24581635e+00 3.45948339e-01 4.23315763e-01 -3.91843677e-01 1.51137602e+00 -7.98638284e-01 1.41553238e-01 1.37130335e-01 -8.58789682e-01 -7.83567548e-01 3.21044266e-01 4.65510726e-01 -4.25051898e-01 -4.18594986e-01 8.79113138e-01 1.65417627e-01 8.56033638e-02 2.31318995e-01 -1.54791057e+00 -3.30415964e-01 4.45647389e-01 -7.03103364e-01 6.96413815e-01 3.18829805e-01 3.39164644e-01 9.20005322e-01 -9.55197930e-01 6.33570850e-01 1.14176250e+00 3.09904486e-01 5.31528413e-01 -1.05929482e+00 -6.46434307e-01 8.93827319e-01 4.92772311e-01 -1.30685019e+00 -3.69286001e-01 8.34270179e-01 -5.09429395e-01 6.80835485e-01 7.39128068e-02 8.72704029e-01 1.42844319e+00 1.16457678e-01 1.53614795e+00 8.97403777e-01 -2.78185550e-02 3.93225193e-01 -4.29638386e-01 6.62709819e-03 4.59584653e-01 -6.10098988e-02 -3.03711087e-01 -7.52494991e-01 7.59040266e-02 8.97139788e-01 2.56987035e-01 -4.71774250e-01 -4.12969291e-01 -1.49561059e+00 4.97871041e-01 1.88246638e-01 3.38583708e-01 -2.87706167e-01 4.36353505e-01 5.99378943e-01 -1.34783834e-01 3.64849687e-01 -1.48314178e-01 -5.15287280e-01 -2.84754425e-01 -9.56629097e-01 4.91183735e-02 3.95429939e-01 9.94847834e-01 5.19268692e-01 1.39645517e-01 -5.39425433e-01 6.68344915e-01 1.73040092e-01 2.58693069e-01 4.70067620e-01 -9.90978241e-01 6.05332971e-01 7.76792645e-01 4.97659072e-02 -8.03561628e-01 -6.90178946e-02 -2.91054130e-01 -5.79640865e-01 -8.94947350e-02 4.42514062e-01 4.50743688e-03 -1.00786531e+00 1.69811201e+00 4.79458690e-01 9.54823315e-01 -9.00470540e-02 1.06187737e+00 9.38009262e-01 5.82556307e-01 5.10746360e-01 -3.13763469e-01 1.30825806e+00 -1.34636176e+00 -9.84837472e-01 -4.69928473e-01 5.35009265e-01 -5.05722284e-01 1.14882064e+00 2.62124330e-01 -8.96420062e-01 -5.74945211e-01 -9.05336320e-01 -6.33165017e-02 -9.46953893e-02 4.27525401e-01 5.67798853e-01 1.01305889e-02 -7.12852001e-01 3.51257443e-01 -1.10716593e+00 -1.53870463e-01 7.50030339e-01 3.06886695e-02 -4.23391879e-01 -4.21815403e-02 -1.16056716e+00 4.10551399e-01 4.74098265e-01 1.11609697e-01 -1.30013514e+00 -3.77390772e-01 -9.91407990e-01 -1.46126747e-02 1.12201798e+00 -3.03663105e-01 1.26993823e+00 -1.56303513e+00 -1.28739452e+00 6.97644949e-01 -3.19094449e-01 -2.89671659e-01 4.43439364e-01 -2.80517250e-01 -4.94551808e-01 6.42713189e-01 3.70221078e-01 6.14477813e-01 9.54154193e-01 -1.19487834e+00 -1.09815085e+00 -8.42843130e-02 4.03598815e-01 1.75396830e-01 -3.11776996e-02 1.50852248e-01 -1.13344014e+00 -1.05432951e+00 2.45231703e-01 -7.99870312e-01 1.06608599e-01 4.51219417e-02 -2.07244515e-01 -3.66225004e-01 1.18606567e+00 -7.08022892e-01 1.53329444e+00 -2.27363539e+00 1.41767517e-01 -2.73217410e-01 2.82605112e-01 3.03254545e-01 -1.35349095e-01 9.89857465e-02 -2.99563825e-01 3.63836996e-02 -7.54951239e-02 -2.89357841e-01 -2.64056891e-01 4.14697260e-01 -2.38993242e-01 4.53862131e-01 3.30406517e-01 8.80335867e-01 -1.33748412e+00 -7.20496655e-01 2.14869171e-01 3.57542664e-01 -3.87628734e-01 1.10939540e-01 -4.64413643e-01 6.09557569e-01 -6.22029781e-01 1.07732463e+00 2.05510497e-01 -3.86466116e-01 3.35375488e-01 -2.83669323e-01 -4.86812405e-02 3.52969378e-01 -1.25093329e+00 1.73055208e+00 -8.62057954e-02 7.05820918e-01 1.78298488e-01 -1.06636429e+00 2.29593828e-01 4.55402046e-01 7.94200420e-01 -4.52773839e-01 8.86770859e-02 -7.67740235e-02 3.87335718e-02 -7.37721920e-01 1.55601725e-01 7.54040331e-02 1.25693321e-01 2.64005989e-01 2.81499565e-01 6.89413309e-01 4.18223321e-01 3.26735705e-01 1.16686034e+00 7.61474371e-01 3.72708052e-01 -1.14363700e-01 8.03854287e-01 -2.95845836e-01 1.26805413e+00 4.21967387e-01 -8.32980096e-01 6.38626993e-01 9.29615915e-01 -3.42536479e-01 -3.53240907e-01 -8.32549870e-01 3.08605194e-01 1.39364946e+00 5.13284385e-01 -7.53012836e-01 -8.04675221e-01 -1.24554276e+00 -2.60828763e-01 4.84225214e-01 -8.00942242e-01 -1.67849258e-01 -8.35239589e-01 -3.23204696e-01 3.21124583e-01 9.78121340e-01 6.86810613e-01 -1.23727524e+00 -5.93926549e-01 1.75455049e-01 -5.25862634e-01 -1.58010209e+00 -8.65698278e-01 1.67720914e-01 -7.14011073e-01 -1.37214792e+00 -5.02416432e-01 -5.75332463e-01 5.91611922e-01 5.04966199e-01 1.01913130e+00 2.09386379e-01 2.54075900e-02 3.87630492e-01 -5.85053027e-01 -6.26767576e-02 2.13116836e-02 -3.67896557e-01 -7.67080858e-02 4.13171947e-01 5.20756006e-01 -3.89014840e-01 -7.42839754e-01 7.24306881e-01 -9.61710811e-01 2.93395728e-01 5.96673667e-01 6.53485954e-01 8.50694597e-01 4.48769378e-03 3.33795011e-01 -6.00413203e-01 -4.98758852e-02 -4.77273107e-01 -3.11900705e-01 3.44266415e-01 1.15678422e-02 -2.76512355e-01 3.54375213e-01 -7.55127668e-01 -9.23833549e-01 2.32025042e-01 2.58587003e-01 -8.92017126e-01 -2.94551611e-01 3.73546034e-01 -7.58855343e-01 2.17295647e-01 1.68968305e-01 4.49614972e-01 -4.71320659e-01 -2.43436754e-01 1.33733153e-01 2.69280553e-01 5.52893817e-01 -6.62605107e-01 3.24944913e-01 6.21693969e-01 -2.73152918e-01 -7.54946649e-01 -1.15161002e+00 -7.14124739e-01 -8.33953261e-01 -5.04142642e-01 1.39997578e+00 -1.01587641e+00 -4.04364169e-01 4.50290442e-01 -9.78341341e-01 -7.03778565e-01 -3.18070650e-01 4.77207720e-01 -5.13317466e-01 4.73609775e-01 -5.27289927e-01 -6.64954543e-01 1.57549679e-01 -1.28307724e+00 1.32174790e+00 1.25373974e-01 -8.49361196e-02 -8.25377703e-01 -3.10338378e-01 5.20579278e-01 -1.94401011e-01 2.38659218e-01 5.36893487e-01 -4.65583235e-01 -8.24966311e-01 1.32919441e-03 -1.70650318e-01 3.93564135e-01 3.66546899e-01 5.29152527e-02 -9.93132532e-01 -3.84063758e-02 1.51085376e-03 -1.71403900e-01 1.14526474e+00 4.95070755e-01 1.44792497e+00 -1.40062898e-01 -4.41236973e-01 5.29781044e-01 8.87833953e-01 3.57997417e-01 6.56099677e-01 1.30775377e-01 9.55041468e-01 3.59148681e-01 9.24792171e-01 1.61456808e-01 2.16446415e-01 6.96408153e-01 4.61492747e-01 -1.78551041e-02 -3.48842323e-01 -2.71047294e-01 7.94379652e-01 4.47250873e-01 -2.45499760e-01 -2.96747148e-01 -7.62933433e-01 4.60252076e-01 -2.19515634e+00 -1.26883769e+00 4.57886718e-02 2.03137922e+00 9.34052527e-01 3.42743546e-01 3.11252415e-01 1.58266336e-01 8.26445818e-01 5.28708875e-01 -7.22318769e-01 4.58522052e-01 -3.75729054e-02 -2.87850797e-01 2.83323586e-01 2.12077111e-01 -1.54557216e+00 1.00690126e+00 5.05105352e+00 1.02885056e+00 -9.27496254e-01 2.00854853e-01 5.09343088e-01 -3.53146136e-01 2.61871874e-01 1.40554756e-01 -8.50415409e-01 7.26036847e-01 4.77831960e-01 3.37640494e-01 8.44603032e-02 9.20190394e-01 4.03648466e-01 -4.48409945e-01 -1.38983631e+00 1.00827599e+00 2.24046648e-01 -1.12686956e+00 -1.22843524e-02 -9.77355614e-02 6.14173770e-01 -3.39964092e-01 -3.78641695e-01 4.99213666e-01 -7.18940869e-02 -6.01636827e-01 1.07957625e+00 3.12395990e-01 7.92135775e-01 -3.63515407e-01 5.87451160e-01 2.75651872e-01 -1.79758775e+00 -2.94089466e-01 1.11968048e-01 9.08748060e-02 3.03459376e-01 4.65002239e-01 -1.10384054e-01 4.79954660e-01 7.28384912e-01 1.24338186e+00 -4.06420380e-01 7.61530936e-01 -7.23383963e-01 8.69340956e-01 6.73483908e-02 4.18951213e-01 4.78222340e-01 -1.94253623e-01 4.85240042e-01 1.19616640e+00 -1.59287140e-01 3.76561761e-01 7.94976711e-01 7.54799545e-01 5.59260212e-02 -2.37459317e-01 -8.19024369e-02 -2.24249840e-01 2.34108746e-01 1.16163826e+00 -1.06808949e+00 -5.88277280e-01 -7.57119894e-01 1.02106333e+00 4.54514213e-02 6.78844094e-01 -1.20117879e+00 -8.06323811e-03 6.92301631e-01 1.81463882e-01 4.13213223e-01 -2.70966053e-01 9.08107981e-02 -1.43694246e+00 9.06029269e-02 -8.76082778e-01 5.74908555e-01 -7.67710745e-01 -1.00640666e+00 2.80716479e-01 2.26958558e-01 -1.46460605e+00 2.04044059e-01 -5.38635015e-01 -6.27114832e-01 5.18558979e-01 -1.39777553e+00 -1.23233116e+00 -5.82005143e-01 6.92825437e-01 1.01725554e+00 -4.70210472e-03 2.27293700e-01 4.92474794e-01 -1.06571925e+00 3.08529526e-01 -4.71307695e-01 5.94190598e-01 7.07559228e-01 -1.24380386e+00 -2.12263130e-02 1.16110265e+00 1.28933489e-01 5.49124718e-01 2.87391663e-01 -9.39892948e-01 -1.15173030e+00 -1.27309525e+00 5.68309247e-01 -4.55370188e-01 7.98352540e-01 -4.53900844e-01 -9.10937846e-01 8.67711008e-01 -7.72940814e-02 4.63876814e-01 6.10161781e-01 -9.41381156e-02 -1.97193742e-01 -5.27148321e-02 -5.72144687e-01 7.98696816e-01 1.29558158e+00 -6.80605471e-01 -7.92461276e-01 2.35351488e-01 7.79459476e-01 -3.20552200e-01 -3.20630372e-01 3.17585200e-01 3.96931887e-01 -1.00530279e+00 1.00078666e+00 -5.28979599e-01 6.52273238e-01 -6.42413378e-01 -1.46060020e-01 -6.96823537e-01 -2.14353159e-01 -4.40924525e-01 -7.03741789e-01 1.34128892e+00 -1.04312330e-01 -1.61185816e-01 7.92688847e-01 3.17282438e-01 -3.00136954e-01 -9.48449790e-01 -7.69246340e-01 -9.06392634e-01 -4.14344877e-01 -6.23767853e-01 2.26674274e-01 1.06464791e+00 -9.55481157e-02 3.12653601e-01 -3.62083137e-01 6.86338544e-02 3.31017613e-01 5.90384677e-02 5.96800923e-01 -8.67419243e-01 -2.61794478e-01 -6.41918421e-01 -3.58372658e-01 -1.31440687e+00 3.15677494e-01 -6.37407959e-01 2.35340372e-01 -1.49545360e+00 3.59606296e-01 -1.55373722e-01 -5.48741698e-01 8.56211305e-01 -6.04388833e-01 2.49700155e-02 9.60745886e-02 3.41767609e-01 -1.10566950e+00 5.89014649e-01 1.20321524e+00 -1.34639129e-01 -2.14687169e-01 -4.17877883e-02 -3.00206006e-01 1.07308662e+00 5.51485121e-01 -3.39947224e-01 -6.59207106e-01 -1.72092974e-01 -2.29332477e-01 6.26511499e-02 5.58865666e-01 -9.46113825e-01 1.01635136e-01 -5.49275696e-01 3.09443533e-01 -7.49543548e-01 2.29322031e-01 -8.50191176e-01 -2.16804966e-01 2.08407849e-01 -3.74669164e-01 -3.00481588e-01 3.29337008e-02 9.07396555e-01 -3.56684834e-01 -2.70271697e-03 7.35040426e-01 -2.06457525e-01 -1.18536735e+00 5.57966828e-01 -4.47733104e-01 1.97639078e-01 1.29616928e+00 -3.19806516e-01 -3.17176193e-01 -2.14792594e-01 -7.89810181e-01 4.20413852e-01 2.21173123e-01 5.15826941e-01 5.26079655e-01 -1.50188851e+00 -3.26278716e-01 1.76209435e-02 2.07661465e-01 6.93783835e-02 3.30359906e-01 1.29946423e+00 -1.71225667e-01 2.05214307e-01 1.27324566e-01 -7.96229124e-01 -1.46424019e+00 6.35985434e-01 3.81437838e-01 -1.10847168e-01 -7.81770825e-01 8.08481276e-01 8.38619232e-01 3.27459872e-01 5.70567429e-01 -6.92107618e-01 -3.09522957e-01 1.55952051e-01 8.59753489e-01 2.41717085e-01 -3.68859142e-01 -8.13030958e-01 -4.43194866e-01 3.89946043e-01 1.73635751e-01 1.14607513e-01 8.52546990e-01 -5.48689142e-02 -5.23811877e-02 6.13637745e-01 9.29729283e-01 -1.97638050e-01 -1.80644834e+00 -3.07276040e-01 1.47560254e-01 -5.28668642e-01 1.11832052e-01 -7.79429793e-01 -1.31137967e+00 9.57954347e-01 2.97465682e-01 -1.45278107e-02 1.26746428e+00 1.04302987e-01 7.33864367e-01 6.63788095e-02 2.05679730e-01 -1.22451699e+00 5.96428931e-01 5.05121946e-01 7.57217705e-01 -1.28109491e+00 1.07051559e-01 -4.92655754e-01 -7.77681172e-01 8.45259964e-01 1.00140417e+00 1.49209633e-01 3.77091467e-01 2.67205477e-01 -2.81131663e-03 -4.44869436e-02 -7.07721412e-01 -4.34089512e-01 5.06698430e-01 4.25215900e-01 2.87804514e-01 -3.07388276e-01 -2.35873744e-01 1.03787518e+00 8.36981714e-01 9.64398906e-02 2.92624444e-01 1.29563630e+00 -4.62014794e-01 -8.70605648e-01 -3.13617438e-01 2.09984154e-01 -6.92770481e-01 -2.06015874e-02 -4.35222030e-01 8.33700418e-01 6.06889963e-01 9.90576625e-01 -9.35014561e-02 -4.02273357e-01 1.77491337e-01 1.63424477e-01 2.80476868e-01 -7.30855703e-01 -3.76544625e-01 7.23542273e-01 -1.78716350e-02 -1.11484361e+00 -8.60177875e-01 -7.84867764e-01 -1.54942048e+00 2.60711968e-01 -1.61522985e-01 -3.20649818e-02 7.80845582e-02 1.17139125e+00 1.93374917e-01 7.78854549e-01 2.80540824e-01 -1.03232300e+00 -1.97844370e-03 -7.94691026e-01 -6.37194455e-01 4.12296027e-01 4.41588432e-01 -1.00388622e+00 -4.46429610e-01 5.51240921e-01]

[8.514793395996094, 0.6516478657722473]

0119e6a1-6b37-4279-b15b-2d78fc0f9d03

domain-expansion-in-dnn-based-acoustic-models

1910.00565

https://arxiv.org/abs/1910.00565v1

https://arxiv.org/pdf/1910.00565v1.pdf

Domain Expansion in DNN-based Acoustic Models for Robust Speech Recognition

Training acoustic models with sequentially incoming data -- while both leveraging new data and avoiding the forgetting effect-- is an essential obstacle to achieving human intelligence level in speech recognition. An obvious approach to leverage data from a new domain (e.g., new accented speech) is to first generate a comprehensive dataset of all domains, by combining all available data, and then use this dataset to retrain the acoustic models. However, as the amount of training data grows, storing and retraining on such a large-scale dataset becomes practically impossible. To deal with this problem, in this study, we study several domain expansion techniques which exploit only the data of the new domain to build a stronger model for all domains. These techniques are aimed at learning the new domain with a minimal forgetting effect (i.e., they maintain original model performance). These techniques modify the adaptation procedure by imposing new constraints including (1) weight constraint adaptation (WCA): keeping the model parameters close to the original model parameters; (2) elastic weight consolidation (EWC): slowing down training for parameters that are important for previously established domains; (3) soft KL-divergence (SKLD): restricting the KL-divergence between the original and the adapted model output distributions; and (4) hybrid SKLD-EWC: incorporating both SKLD and EWC constraints. We evaluate these techniques in an accent adaptation task in which we adapt a deep neural network (DNN) acoustic model trained with native English to three different English accents: Australian, Hispanic, and Indian. The experimental results show that SKLD significantly outperforms EWC, and EWC works better than WCA. The hybrid SKLD-EWC technique results in the best overall performance.

['John H. L. Hansen', 'Soheil Khorram', 'Shahram Ghorbani']

2019-10-01

['robust-speech-recognition']

['speech']

[ 3.04055184e-01 -1.00832149e-01 -2.86692777e-03 -4.87320364e-01 -5.13678610e-01 -4.37935591e-01 3.97117019e-01 -6.43709600e-02 -9.89115179e-01 8.00789475e-01 2.46179342e-01 -3.64070415e-01 -2.02179607e-02 -4.63120967e-01 -5.26148260e-01 -8.35250020e-01 2.11330965e-01 4.71988559e-01 4.96955305e-01 -1.94374233e-01 -3.46825495e-02 4.67426866e-01 -1.52361500e+00 -1.56843290e-01 1.10975277e+00 7.35063016e-01 6.74901307e-01 4.94164467e-01 -2.93429494e-01 2.45498613e-01 -7.29474545e-01 -3.32502186e-01 2.40359336e-01 -4.58678901e-01 -7.05207765e-01 -7.52098858e-02 8.37882608e-02 -1.77136451e-01 2.96180006e-02 9.96809542e-01 6.50556922e-01 5.99041224e-01 2.65836596e-01 -8.92125010e-01 -6.41932905e-01 6.42444730e-01 -3.34400445e-01 3.06446642e-01 -1.13047520e-02 -8.29733387e-02 4.78772014e-01 -9.21519220e-01 4.28262949e-01 1.01465654e+00 6.41564131e-01 9.26563978e-01 -1.29490829e+00 -7.77677715e-01 4.02356863e-01 1.62164211e-01 -1.30656564e+00 -7.23347962e-01 8.42874587e-01 -6.83400333e-02 1.14270592e+00 1.86490804e-01 4.43090111e-01 1.25897181e+00 -1.23585261e-01 5.10636747e-01 9.55530047e-01 -7.69802928e-01 5.91248572e-01 4.36528921e-01 2.12796077e-01 5.77878505e-02 -1.04096541e-02 1.15855947e-01 -6.08017504e-01 -1.92867860e-01 2.68420875e-01 -2.49806076e-01 -1.62866831e-01 -1.91352606e-01 -8.05910945e-01 6.33227170e-01 -1.02793343e-01 5.26388228e-01 -4.57836270e-01 -3.93490136e-01 2.87344247e-01 6.01910770e-01 4.98180419e-01 3.47990483e-01 -7.93218195e-01 -3.93155307e-01 -8.51710200e-01 9.37030241e-02 7.21474469e-01 7.40857244e-01 8.03209901e-01 4.22766745e-01 4.58862484e-01 1.62884355e+00 7.15035945e-02 5.18869638e-01 1.16568244e+00 -5.06314933e-01 4.86324430e-01 2.77243674e-01 -2.05582991e-01 -3.24288011e-01 -2.95814902e-01 -4.73166853e-01 -7.26271570e-01 2.06121311e-01 3.13995928e-01 -5.20382762e-01 -1.21706593e+00 2.21889925e+00 3.80684257e-01 2.59162635e-01 3.67155612e-01 4.72748965e-01 4.55115080e-01 7.19954729e-01 2.83618391e-01 -4.90102261e-01 8.53265762e-01 -8.21217060e-01 -7.23477483e-01 -6.20081425e-01 6.56525433e-01 -7.87234008e-01 1.22382247e+00 4.10278589e-01 -8.63395095e-01 -8.87455106e-01 -1.12660408e+00 3.18560183e-01 -6.41434610e-01 -1.11023419e-01 1.10205747e-01 8.37526798e-01 -1.01147628e+00 4.26666945e-01 -6.63711727e-01 -3.30502808e-01 -1.18638448e-01 4.38024759e-01 -4.30124432e-01 -1.24531209e-01 -1.42792916e+00 9.68462288e-01 8.83149564e-01 -2.40839928e-01 -3.44238549e-01 -8.21084678e-01 -7.58292437e-01 -8.39025248e-03 3.73270899e-01 -3.32359225e-01 1.23187184e+00 -1.29671264e+00 -1.85903764e+00 4.00961757e-01 -1.72434211e-01 -4.62945908e-01 2.07795277e-01 -2.79604435e-01 -8.70779514e-01 -4.63798016e-01 -3.05314004e-01 5.29013932e-01 8.29974890e-01 -1.22978878e+00 -7.23864079e-01 -3.84909302e-01 -3.89628112e-01 4.26294982e-01 -7.82391310e-01 -3.64972651e-02 -3.34662020e-01 -7.95176148e-01 8.88242647e-02 -1.04382586e+00 1.60318811e-03 -6.97445810e-01 1.88207068e-02 -2.38243192e-01 9.52092648e-01 -9.05432701e-01 1.60229385e+00 -2.44882464e+00 2.54919708e-01 3.03766578e-01 -2.75004774e-01 8.90943348e-01 -3.65050077e-01 1.37564152e-01 -2.22664416e-01 -1.12162083e-01 -3.56605172e-01 -5.73034465e-01 -9.97788087e-02 6.56113446e-01 -2.15540871e-01 -1.96494818e-01 9.83663201e-02 4.32267040e-01 -6.48843408e-01 -2.28016764e-01 8.05261284e-02 3.62776756e-01 -5.93223691e-01 2.10880801e-01 -3.92473564e-02 2.20348358e-01 -3.88726443e-02 2.38281682e-01 7.60791242e-01 3.87838185e-01 3.06589633e-01 1.59858882e-01 -1.01778686e-01 2.71830529e-01 -1.49895525e+00 1.46972394e+00 -6.26425266e-01 2.37790853e-01 2.06355110e-01 -9.91873085e-01 1.24736404e+00 2.69696653e-01 2.54033506e-01 -7.55839586e-01 -3.35396051e-01 4.15551305e-01 2.48466447e-01 -2.36524031e-01 3.96012694e-01 -4.90404189e-01 -6.93862364e-02 2.08962515e-01 2.97417641e-01 -1.29890293e-01 1.02156147e-01 -2.22903714e-01 8.81574750e-01 -7.19431564e-02 2.16882184e-01 -1.51564434e-01 5.44365287e-01 -4.15974349e-01 1.04867518e+00 5.74408472e-01 -2.16543525e-01 2.79929072e-01 1.93967298e-02 -3.73225093e-01 -9.89518642e-01 -1.22180629e+00 -1.05165755e-02 1.41361737e+00 -3.48578542e-01 -2.05820620e-01 -7.96056271e-01 -6.40594959e-01 2.52793133e-02 1.14703393e+00 -3.21142048e-01 -4.81528372e-01 -7.30318129e-01 -8.51539016e-01 4.72069651e-01 5.71918845e-01 6.19496703e-01 -9.66123283e-01 -3.29977989e-01 3.57978791e-01 -1.53345093e-01 -9.08276141e-01 -5.93167424e-01 6.10601544e-01 -7.63114989e-01 -3.30049276e-01 -7.38919556e-01 -8.78097951e-01 5.04303038e-01 -1.05204098e-01 8.03557396e-01 -1.98033705e-01 2.61346757e-01 1.58925146e-01 -4.83824015e-01 -5.47894597e-01 -7.23429203e-01 2.48175412e-01 5.06934345e-01 7.38967732e-02 5.61129689e-01 -6.80669487e-01 1.03503369e-01 2.51542240e-01 -1.03712881e+00 -1.06143482e-01 6.52570426e-01 9.42760348e-01 5.26652277e-01 2.57955581e-01 1.12161756e+00 -8.49829495e-01 8.45173120e-01 -3.45999628e-01 -4.06802535e-01 3.06749225e-01 -8.58958066e-01 1.37911811e-01 8.01644981e-01 -1.01706827e+00 -1.45194590e+00 1.16475590e-01 -6.50635481e-01 -4.88361746e-01 -2.87661463e-01 5.87329209e-01 -3.95319402e-01 2.09874958e-01 5.34408569e-01 4.04718459e-01 3.25619504e-02 -7.58853972e-01 1.77090138e-01 9.21602249e-01 6.04206026e-01 -5.85977256e-01 6.17942631e-01 -1.77588493e-01 -6.47406161e-01 -1.08113110e+00 -4.98592943e-01 -3.70800823e-01 -7.66679585e-01 -3.45210508e-02 6.14591599e-01 -5.24861932e-01 2.49879193e-02 8.15032125e-01 -8.52584839e-01 -5.25827110e-01 -4.83268738e-01 8.91998112e-01 -1.64039254e-01 3.34517837e-01 -2.51096517e-01 -8.92709374e-01 -1.01796933e-01 -9.46453929e-01 3.57039243e-01 3.26129049e-01 -1.97289690e-01 -1.13509929e+00 3.89138073e-01 5.35644665e-02 7.89842784e-01 -1.82869509e-01 1.08596718e+00 -1.19288981e+00 3.33322257e-01 -1.66291874e-02 2.94232070e-01 9.95715737e-01 3.69479179e-01 -2.58905917e-01 -1.08878744e+00 -4.23395574e-01 2.86311984e-01 -1.48233354e-01 7.30946720e-01 2.02037469e-01 8.66458356e-01 -1.51576936e-01 -1.85247406e-01 3.52741271e-01 1.00255299e+00 8.18548501e-01 4.78180587e-01 4.49491262e-01 3.22239697e-01 4.38427091e-01 4.19194520e-01 2.19374150e-01 1.54514104e-01 6.72728717e-01 -1.22904085e-01 -3.74599323e-02 -3.42127383e-01 -1.98219553e-01 5.59038341e-01 1.51311362e+00 -3.63126174e-02 -8.38796720e-02 -9.56290781e-01 5.51131427e-01 -1.54423022e+00 -7.36141503e-01 5.68622231e-01 2.60533023e+00 1.20213556e+00 5.03363431e-01 1.48336127e-01 2.48892948e-01 6.89447641e-01 -8.56467038e-02 -7.96813250e-01 -7.47673810e-01 -1.39905676e-01 5.27672052e-01 1.86387777e-01 6.57143474e-01 -9.74555135e-01 8.76192093e-01 5.47588634e+00 8.08578253e-01 -1.21365678e+00 2.71677643e-01 2.78934032e-01 -1.30323350e-01 -1.74905077e-01 -1.82394665e-02 -9.23538983e-01 6.44145846e-01 1.32074845e+00 -1.92171454e-01 4.80621219e-01 9.45920110e-01 -1.44662425e-01 -2.10948493e-02 -8.50138783e-01 8.14443171e-01 1.11878082e-01 -7.59933054e-01 9.69809759e-03 -1.72208417e-02 4.57643926e-01 6.70911670e-02 6.27957731e-02 7.72204995e-01 4.16929096e-01 -5.07714629e-01 5.74761152e-01 3.85562330e-01 6.87829375e-01 -8.55454326e-01 8.89684796e-01 5.56888282e-01 -8.49213004e-01 -1.94694996e-01 -3.91768813e-01 2.36586392e-01 6.30052835e-02 6.07193708e-01 -8.25656533e-01 4.98120248e-01 7.59334862e-01 1.73351258e-01 -4.94158149e-01 7.39697933e-01 1.72567621e-01 8.26997578e-01 -4.96183842e-01 8.64158720e-02 6.15655743e-02 -1.78208634e-01 5.78294039e-01 1.24737310e+00 4.69133615e-01 -7.92222992e-02 5.67866862e-02 4.20622736e-01 3.72879729e-02 5.73115386e-02 -4.55573171e-01 1.66148812e-01 7.92112172e-01 6.75268173e-01 -3.76755863e-01 -2.14145452e-01 -5.08890450e-01 1.00829792e+00 4.10397619e-01 4.64749157e-01 -6.10601008e-01 -5.84612668e-01 7.19568312e-01 -6.85197115e-02 3.24850678e-01 -3.40129733e-01 -9.73121729e-03 -8.05716872e-01 3.17914300e-02 -9.65772629e-01 4.39198107e-01 -4.99347806e-01 -1.21240425e+00 8.40817034e-01 2.55375952e-01 -8.15110087e-01 -5.26151359e-01 -4.69737291e-01 -4.88833994e-01 1.10847080e+00 -1.51012945e+00 -7.78526068e-01 1.10517982e-02 5.84293425e-01 7.89660513e-01 -5.28576612e-01 9.68599200e-01 5.13472199e-01 -6.19751513e-01 8.13977361e-01 4.15407628e-01 -8.52730870e-02 9.33068097e-01 -1.08753407e+00 3.96902025e-01 8.70998502e-01 1.10168889e-01 6.87498927e-01 5.09728134e-01 -6.85158789e-01 -9.46033120e-01 -9.16397810e-01 1.00851274e+00 -7.91591406e-02 3.90642673e-01 -4.52312559e-01 -1.42606175e+00 6.19112015e-01 2.61951089e-02 -3.69942069e-01 8.24505389e-01 3.26943845e-01 -3.08531761e-01 -4.30557162e-01 -1.06094623e+00 5.21434247e-01 7.18457401e-01 -2.82640845e-01 -8.84273052e-01 -1.35896280e-01 8.79040122e-01 -2.51387417e-01 -7.38605261e-01 4.23471689e-01 4.53999460e-01 -7.79824674e-01 8.24329853e-01 -5.28315485e-01 -2.96456248e-01 -1.15388229e-01 -3.61212820e-01 -1.89891589e+00 -3.61610681e-01 -5.15990376e-01 -4.85604741e-02 1.57692337e+00 5.63549340e-01 -9.54565048e-01 5.01558900e-01 7.02232778e-01 -3.78752887e-01 -5.08872271e-01 -1.24699545e+00 -1.25346494e+00 3.38682532e-01 -5.39915979e-01 8.26624930e-01 1.03337538e+00 -1.24485552e-01 3.01900506e-01 -2.99070805e-01 7.60758203e-03 1.74915373e-01 -5.04607975e-01 6.07614517e-01 -1.21029317e+00 -2.79816836e-01 -3.57405275e-01 -1.23639226e-01 -8.35937858e-01 2.98433244e-01 -6.60741091e-01 8.80378187e-02 -1.02719200e+00 -1.90481573e-01 -6.58472478e-01 -5.91302514e-01 6.35308325e-01 -2.62302607e-01 -3.02563012e-01 2.74785429e-01 8.65160860e-03 -1.05256282e-01 4.90910739e-01 6.57408834e-01 1.39697701e-01 -5.84675848e-01 2.50088990e-01 -5.52566469e-01 7.55526841e-01 9.81804311e-01 -4.50653255e-01 -7.14536786e-01 -4.86165494e-01 -2.40070209e-01 -3.26242268e-01 -2.09925279e-01 -1.22920549e+00 3.21105003e-01 -2.16662005e-01 4.11613286e-01 -3.03416908e-01 4.66020644e-01 -7.97308445e-01 1.43939465e-01 3.64944369e-01 -2.17241466e-01 1.29726902e-01 6.27027512e-01 4.52858537e-01 -2.32922673e-01 -6.20341897e-01 1.07806802e+00 1.49600118e-01 -9.05653596e-01 -1.91168234e-01 -4.30536628e-01 1.29666120e-01 8.97966862e-01 -2.56061584e-01 -1.23795476e-02 -2.24415153e-01 -9.85069871e-01 9.19344649e-02 2.47785076e-01 7.49465942e-01 5.81069589e-01 -1.35640597e+00 -6.77615762e-01 7.67850578e-01 1.99908344e-03 -1.33503124e-01 2.35345900e-01 4.96095031e-01 4.81308177e-02 2.14754894e-01 -5.23716733e-02 -2.90593475e-01 -1.43554389e+00 4.87194449e-01 2.26889670e-01 -3.35178792e-01 -3.12960863e-01 9.59442675e-01 5.39858118e-02 -8.67382288e-01 5.04384935e-01 -3.22185963e-01 -1.06664382e-01 -2.67597828e-02 4.40253496e-01 2.55236387e-01 3.51077110e-01 -4.19116199e-01 -3.85169029e-01 3.03986281e-01 -4.31215703e-01 -2.97205150e-01 1.36296785e+00 -1.76740974e-01 2.74687350e-01 6.12368226e-01 8.90198290e-01 2.09114373e-01 -1.07765400e+00 -6.45444572e-01 2.27761701e-01 -2.67134368e-01 7.22064301e-02 -1.00059283e+00 -6.69149101e-01 7.08343208e-01 7.75994658e-01 1.80979013e-01 1.39159334e+00 -1.76747590e-01 8.28081131e-01 2.46245652e-01 1.80226490e-01 -1.61424887e+00 3.35614868e-02 7.87974000e-01 6.99846804e-01 -8.76097620e-01 -2.96036780e-01 1.72572687e-01 -9.02667880e-01 1.03109050e+00 7.62470901e-01 3.70830357e-01 8.71470749e-01 3.87312055e-01 1.49564773e-01 4.70049649e-01 -8.02499592e-01 -1.99927017e-01 1.50517851e-01 7.64674067e-01 1.32543623e-01 4.93226051e-02 -1.03586249e-01 8.77954721e-01 -2.91884124e-01 -8.24682191e-02 2.53319979e-01 9.73812163e-01 -6.10837758e-01 -1.41494238e+00 -3.30008179e-01 3.85968715e-01 -1.95106313e-01 -1.41024262e-01 -3.56609017e-01 7.40898967e-01 4.55563486e-01 9.51482177e-01 1.04396604e-02 -5.53291380e-01 6.92262173e-01 8.16281676e-01 6.84224069e-02 -6.12309396e-01 -3.74343842e-01 2.07349867e-01 3.12319864e-03 6.23713573e-03 -2.77044296e-01 -7.48886168e-01 -1.11399829e+00 -1.23801358e-01 -5.01571059e-01 3.77958477e-01 8.95139039e-01 9.11098182e-01 3.93416405e-01 4.96897995e-01 5.28936684e-01 -4.56024468e-01 -8.57838571e-01 -1.18543839e+00 -6.10674262e-01 4.03256565e-01 3.14830929e-01 -6.99753165e-01 -5.21734238e-01 2.28902370e-01]

[14.396815299987793, 6.585726261138916]

285338fa-f584-47b9-90fd-83bccfaa67b8

popmag-pop-music-accompaniment-generation

2008.07703

https://arxiv.org/abs/2008.07703v1

https://arxiv.org/pdf/2008.07703v1.pdf

PopMAG: Pop Music Accompaniment Generation

In pop music, accompaniments are usually played by multiple instruments (tracks) such as drum, bass, string and guitar, and can make a song more expressive and contagious by arranging together with its melody. Previous works usually generate multiple tracks separately and the music notes from different tracks not explicitly depend on each other, which hurts the harmony modeling. To improve harmony, in this paper, we propose a novel MUlti-track MIDI representation (MuMIDI), which enables simultaneous multi-track generation in a single sequence and explicitly models the dependency of the notes from different tracks. While this greatly improves harmony, unfortunately, it enlarges the sequence length and brings the new challenge of long-term music modeling. We further introduce two new techniques to address this challenge: 1) We model multiple note attributes (e.g., pitch, duration, velocity) of a musical note in one step instead of multiple steps, which can shorten the length of a MuMIDI sequence. 2) We introduce extra long-context as memory to capture long-term dependency in music. We call our system for pop music accompaniment generation as PopMAG. We evaluate PopMAG on multiple datasets (LMD, FreeMidi and CPMD, a private dataset of Chinese pop songs) with both subjective and objective metrics. The results demonstrate the effectiveness of PopMAG for multi-track harmony modeling and long-term context modeling. Specifically, PopMAG wins 42\%/38\%/40\% votes when comparing with ground truth musical pieces on LMD, FreeMidi and CPMD datasets respectively and largely outperforms other state-of-the-art music accompaniment generation models and multi-track MIDI representations in terms of subjective and objective metrics.

['Tie-Yan Liu', 'Jinzheng He', 'Yi Ren', 'Zhou Zhao', 'Xu Tan', 'Tao Qin']

2020-08-18

['music-modeling']

['music']

[-1.73262209e-01 -6.65112078e-01 -1.08760111e-01 2.54724026e-01 -5.40575743e-01 -9.49423134e-01 4.36751813e-01 -1.82644054e-01 -1.08283311e-01 6.93719745e-01 4.67333674e-01 2.89122164e-01 -3.46377194e-01 -7.52849162e-01 -5.09808421e-01 -6.07612669e-01 6.52177706e-02 4.35652852e-01 1.75518587e-01 -4.27597761e-01 3.16601127e-01 3.18360981e-03 -1.48702562e+00 3.88303787e-01 7.29022026e-01 7.32470870e-01 3.67879331e-01 8.76768172e-01 -2.98032969e-01 9.10715759e-01 -1.04443407e+00 -3.63489121e-01 1.15256481e-01 -8.17632377e-01 -3.89033586e-01 -2.94999093e-01 6.65977076e-02 5.56037202e-02 -1.75890431e-01 6.71964169e-01 8.85407686e-01 2.42930055e-01 6.30211711e-01 -1.06621861e+00 -4.82282758e-01 1.34933662e+00 -6.50525033e-01 -3.70901898e-02 2.08769307e-01 -9.82341096e-02 1.17757249e+00 -5.63290238e-01 5.26172996e-01 9.22049463e-01 9.66225326e-01 3.09179604e-01 -1.13663352e+00 -1.12806737e+00 -3.19424681e-02 5.07414714e-02 -1.58487856e+00 -2.03512445e-01 1.01148272e+00 -4.15791065e-01 5.96617639e-01 6.20150506e-01 9.44538057e-01 1.04261768e+00 1.86667945e-02 9.28870797e-01 7.62185097e-01 -2.31730595e-01 -1.51546374e-01 -4.98168528e-01 -1.76465228e-01 -9.98296030e-03 -4.13425863e-01 3.70722026e-01 -8.53006721e-01 -1.25375018e-01 1.02457690e+00 -2.24030226e-01 -8.89184624e-02 3.96389186e-01 -1.42083418e+00 5.18960297e-01 6.04649857e-02 5.50859809e-01 -2.37812296e-01 2.43967503e-01 5.03442049e-01 2.60518909e-01 1.94563472e-03 6.70567930e-01 -2.16081217e-01 -7.71308541e-01 -1.34453595e+00 6.29964292e-01 7.66349435e-01 9.69031036e-01 4.33102340e-01 3.67217600e-01 -6.26409650e-01 1.20736408e+00 -9.19063687e-02 4.93240029e-01 7.77656019e-01 -8.87581408e-01 3.13351840e-01 1.48715079e-01 1.67221978e-01 -9.11650956e-01 -3.64573210e-01 -7.85819948e-01 -1.06872106e+00 -9.16329995e-02 1.75737038e-01 6.15579337e-02 -5.86952150e-01 2.07261920e+00 -8.08504522e-02 8.14130306e-01 -1.64950073e-01 7.34454870e-01 8.32003474e-01 9.50788200e-01 -1.13861993e-01 -5.66271305e-01 1.08602786e+00 -1.13694060e+00 -8.89198303e-01 1.60228178e-01 1.21576943e-01 -1.43133879e+00 1.26479137e+00 6.66107774e-01 -1.28273022e+00 -1.08345866e+00 -1.07463038e+00 1.49937853e-01 2.94796377e-01 2.56410867e-01 4.06048596e-01 3.70033741e-01 -3.13798189e-01 9.28133130e-01 -5.29050887e-01 4.24632251e-01 -1.51427954e-01 1.53974518e-01 2.87434995e-01 5.51106155e-01 -1.25481045e+00 3.26625049e-01 3.89453262e-01 -5.58274910e-02 -8.77995551e-01 -9.55625176e-01 -2.80079961e-01 -5.27181774e-02 2.14756012e-01 -4.90542978e-01 1.29640210e+00 -6.49701595e-01 -1.74745727e+00 5.23317218e-01 7.21381828e-02 -2.49186158e-01 4.69354957e-01 -4.73294795e-01 -7.47292936e-01 -3.84682894e-01 -5.26060313e-02 3.18934262e-01 8.48693073e-01 -1.09796011e+00 -7.33364105e-01 1.21570319e-01 -2.12781399e-01 2.10163325e-01 -2.99489915e-01 -3.57609950e-02 -8.29580069e-01 -1.53932631e+00 -1.21617094e-01 -1.35891020e+00 1.85497329e-01 -7.58058012e-01 -6.87590718e-01 7.09158257e-02 6.34842873e-01 -5.79500437e-01 2.18866229e+00 -2.35476828e+00 3.77509683e-01 -2.48098793e-03 -1.41896382e-01 2.73113370e-01 -2.35677719e-01 5.20141661e-01 1.15318701e-01 3.81335318e-02 -3.49044651e-02 -5.01693308e-01 1.48638293e-01 5.07724524e-01 -6.22254193e-01 -1.75675809e-01 -4.14374381e-01 9.69272852e-01 -8.02841723e-01 -3.56677920e-01 -6.48152176e-03 4.64868665e-01 -5.23420095e-01 1.38985723e-01 -2.22759143e-01 6.26546741e-01 -1.47065789e-01 3.88452917e-01 4.09392476e-01 1.57881334e-01 1.41086802e-01 -3.55791062e-01 -4.53519434e-01 5.19388080e-01 -1.67164612e+00 2.13257480e+00 -5.88182747e-01 2.10176602e-01 -2.74459571e-01 -2.53722310e-01 1.25099051e+00 6.57887697e-01 4.65043306e-01 -5.12803197e-01 5.28254099e-02 2.72182733e-01 1.66743994e-01 -1.66781411e-01 8.64529610e-01 -4.29380596e-01 -4.14840102e-01 7.11791039e-01 -1.76791087e-01 -3.32720011e-01 4.23768789e-01 -2.48222142e-01 6.98997676e-01 2.66750425e-01 1.97011948e-01 -9.30429474e-02 2.96289027e-01 -3.86783421e-01 9.18721020e-01 6.25577450e-01 1.92258731e-01 9.28342342e-01 7.95110613e-02 -2.19456434e-01 -1.01189244e+00 -1.20547116e+00 2.15218276e-01 1.19850969e+00 1.20569021e-01 -1.00296104e+00 -4.19703335e-01 5.11781033e-03 -5.68009801e-02 5.95613837e-01 -5.43674529e-01 -1.31694496e-01 -9.57204521e-01 -6.99983478e-01 1.23533475e+00 6.52506590e-01 2.50326842e-01 -1.44733298e+00 -2.14890078e-01 7.10207522e-01 -6.25819683e-01 -6.59564912e-01 -1.08273101e+00 3.66578363e-02 -7.49726117e-01 -7.31167078e-01 -7.21922576e-01 -7.03473747e-01 -4.13358778e-01 -1.34876460e-01 1.29099083e+00 -4.96616252e-02 -3.04610014e-01 -2.07512468e-01 -4.58464563e-01 -4.40487027e-01 -4.75100040e-01 2.55010068e-01 2.63186097e-01 -1.18715107e-01 -1.13401487e-01 -1.13523006e+00 -5.36252320e-01 6.06753469e-01 -8.39573443e-01 2.73069888e-01 4.19525236e-01 8.01778555e-01 1.09832335e+00 1.85848951e-01 7.99550891e-01 -6.03964508e-01 8.05540979e-01 -2.00022012e-01 -1.40400276e-01 6.01347648e-02 -3.61162961e-01 -1.81285396e-01 6.74354315e-01 -1.00613272e+00 -7.57300079e-01 -3.84375095e-01 -2.63560802e-01 -7.16534615e-01 3.83529395e-01 7.04836607e-01 -5.64935580e-02 5.25907397e-01 5.55610478e-01 3.54315221e-01 -3.94133210e-01 -8.44759464e-01 3.50484818e-01 5.22804201e-01 1.13609803e+00 -9.56886053e-01 7.42238939e-01 2.99349148e-02 -1.10479653e-01 -5.68807721e-01 -7.27783561e-01 -2.99219787e-01 -5.04125178e-01 -1.34427279e-01 5.24423063e-01 -8.69339526e-01 -7.55054057e-01 6.89684987e-01 -8.88198495e-01 -3.63701642e-01 -6.65395379e-01 4.44237053e-01 -7.74382889e-01 2.09186405e-01 -9.78525102e-01 -7.31480479e-01 -6.16881907e-01 -7.22656429e-01 7.77309656e-01 2.73905069e-01 -8.01623881e-01 -6.60132349e-01 4.75569904e-01 7.93645903e-03 2.13185877e-01 4.40308332e-01 8.67853820e-01 1.30236531e-02 -1.91540301e-01 4.50070761e-02 4.12042439e-01 3.06147605e-01 4.96218026e-01 2.63200670e-01 -8.14032674e-01 -1.51078478e-01 -1.70168579e-01 -1.54066190e-01 6.11363709e-01 3.82900059e-01 8.56770933e-01 -2.06619710e-01 3.00161596e-02 7.61871696e-01 9.56464827e-01 4.47599232e-01 8.01732481e-01 2.57663846e-01 6.67319715e-01 2.09368467e-01 9.66937959e-01 8.36085021e-01 1.04422137e-01 1.09595513e+00 -8.31152275e-02 2.50318050e-01 -5.60522616e-01 -7.36898661e-01 4.66596067e-01 1.71832383e+00 -6.22156024e-01 -3.59388441e-01 -4.69701856e-01 6.00423932e-01 -2.02311754e+00 -1.28348637e+00 -2.11256444e-01 2.30635834e+00 1.28018749e+00 7.83233196e-02 4.75118846e-01 6.18794739e-01 6.60871148e-01 1.38731405e-01 -3.74438554e-01 -2.78859466e-01 -5.61958969e-01 5.20457447e-01 -5.93972951e-02 3.37231010e-01 -9.13940609e-01 9.08673704e-01 6.08323431e+00 1.32376266e+00 -1.16811800e+00 1.86120979e-02 1.77304626e-01 -5.98919094e-01 -3.86577636e-01 -1.36696786e-01 -7.79232681e-01 8.22348535e-01 7.19907463e-01 -3.50896955e-01 7.89757729e-01 4.29260850e-01 2.21734807e-01 3.97502124e-01 -1.01563776e+00 1.34714162e+00 -1.29226893e-01 -1.45014381e+00 1.76465526e-01 -2.20558167e-01 1.02563262e+00 -2.44172662e-01 2.06991747e-01 4.81964618e-01 -1.56437773e-02 -1.06004333e+00 1.23349845e+00 6.44231439e-01 9.61211562e-01 -1.06867623e+00 3.16835880e-01 4.01055366e-01 -1.88454926e+00 2.14717276e-02 -1.60820395e-01 -1.81339324e-01 4.14223820e-01 2.95477182e-01 -4.06011045e-01 8.48876178e-01 5.55728853e-01 8.03822577e-01 -2.70943016e-01 1.11491609e+00 -2.67846286e-01 1.00032854e+00 -3.65134656e-01 2.26742491e-01 -1.30684804e-02 -3.11016232e-01 8.12097490e-01 1.28668928e+00 8.08176160e-01 9.61146038e-03 3.05181414e-01 1.02829051e+00 9.40372869e-02 1.32009476e-01 2.43662149e-02 -3.23043525e-01 8.55025470e-01 1.05910504e+00 -4.09395605e-01 -2.03802869e-01 7.94857889e-02 9.45659459e-01 -1.68183237e-01 1.05534524e-01 -1.01648092e+00 -2.72824764e-01 7.23754883e-01 7.00538605e-02 3.88818026e-01 -2.82226920e-01 -3.51402253e-01 -9.07253206e-01 -2.39317238e-01 -1.06036687e+00 4.16519433e-01 -7.24889278e-01 -1.23766410e+00 7.03513026e-01 -3.88599783e-01 -1.73945749e+00 -6.30023420e-01 -5.50068868e-03 -7.59272456e-01 1.00292456e+00 -1.07863426e+00 -1.20140600e+00 -1.82157680e-01 7.31026113e-01 5.13878703e-01 -1.15310013e-01 1.05506980e+00 6.33818746e-01 -3.10450137e-01 8.16399276e-01 1.28525058e-02 5.09078465e-02 1.08832192e+00 -1.10166931e+00 4.59545672e-01 3.87044519e-01 8.15787196e-01 6.13511503e-01 7.63075352e-01 -6.83836937e-01 -1.00721192e+00 -9.99789536e-01 8.23054671e-01 -2.96031564e-01 5.46507239e-01 -1.98365867e-01 -7.98372567e-01 4.79618907e-01 7.62892365e-02 -5.32654226e-01 1.07552135e+00 3.89318436e-01 -2.57299513e-01 -1.34159446e-01 -3.84500593e-01 7.05956042e-01 1.13356376e+00 -5.76505601e-01 -7.21269846e-01 -2.90250570e-01 6.65512264e-01 -4.54934597e-01 -1.04974794e+00 5.29712141e-01 9.13544118e-01 -1.01645029e+00 1.05269766e+00 -2.42067292e-01 4.28756982e-01 -8.04714441e-01 -2.65325248e-01 -1.35296428e+00 -6.16255462e-01 -1.22609639e+00 -4.45285201e-01 1.69070148e+00 1.92989856e-01 8.91993195e-02 5.21473110e-01 -2.15016022e-01 -4.45092440e-01 -3.83188665e-01 -6.98759139e-01 -1.13213897e+00 -2.96980813e-02 -6.68687224e-01 1.00492430e+00 9.92076278e-01 -7.13174567e-02 4.49161232e-01 -1.15123188e+00 -2.39653587e-01 4.04294431e-01 7.61616170e-01 9.45816576e-01 -1.13907301e+00 -9.34189081e-01 -6.38845682e-01 -9.15967003e-02 -1.02705419e+00 -2.92896092e-01 -8.73821497e-01 -1.32039189e-01 -9.60286617e-01 1.72087938e-01 -7.19039917e-01 -7.59110749e-01 4.23306882e-01 -2.77439266e-01 5.93408704e-01 6.54195249e-01 6.00279689e-01 -1.94681868e-01 6.51027262e-01 1.50082088e+00 -1.84829667e-01 -6.97065830e-01 2.89698899e-01 -4.55824435e-01 7.32080400e-01 5.80201268e-01 -4.69663918e-01 -4.83022660e-01 -3.44133914e-01 4.14291263e-01 3.36423844e-01 -2.07346641e-02 -1.30391741e+00 2.56900489e-01 -8.51347893e-02 2.40553558e-01 -7.98762918e-01 5.81522346e-01 -1.29510239e-01 8.88893485e-01 3.20342332e-01 -2.53598541e-01 6.51043281e-02 4.54462141e-01 3.76033008e-01 -6.28131449e-01 -1.34765118e-01 5.16062260e-01 8.84041116e-02 -4.22215939e-01 3.02253127e-01 -3.24126035e-02 1.77353472e-01 5.30465364e-01 -2.20837519e-02 -9.63687673e-02 -4.16288644e-01 -9.82309103e-01 -1.22120209e-01 2.82354891e-01 8.47020984e-01 3.10643077e-01 -1.85856640e+00 -9.45233643e-01 2.49198228e-02 1.08094744e-01 -3.24553430e-01 6.01606667e-01 6.90847576e-01 -2.47903019e-01 5.42368740e-02 -2.49198198e-01 -3.15915555e-01 -1.50648057e+00 4.65344697e-01 -3.75457630e-02 -4.58018273e-01 -6.29415274e-01 9.70080554e-01 3.39136183e-01 -3.26513201e-01 1.96802586e-01 -3.98829043e-01 -2.13091657e-01 2.68263757e-01 5.95556974e-01 4.32950437e-01 -2.83890992e-01 -6.33222163e-01 -9.18092728e-02 8.37875903e-01 1.35639101e-01 -3.15047294e-01 1.17404974e+00 -5.93687482e-02 -4.64529656e-02 1.13531661e+00 5.78342080e-01 7.76274264e-01 -1.04083025e+00 -6.00112639e-02 -2.60449111e-01 -3.97444159e-01 -2.96966553e-01 -1.16173720e+00 -8.57159257e-01 6.94275677e-01 2.20834568e-01 5.27497940e-02 1.16363049e+00 -3.70030046e-01 1.29321504e+00 -7.18216822e-02 3.25158954e-01 -1.22652829e+00 2.61461943e-01 6.76268458e-01 1.13282466e+00 -3.58306348e-01 -2.58237541e-01 -1.62450522e-01 -8.69527161e-01 1.01425409e+00 3.04982185e-01 -1.33027598e-01 5.08171022e-01 6.74948454e-01 1.51244149e-01 2.48318404e-01 -7.36263335e-01 8.92170295e-02 4.68355358e-01 2.00329125e-01 7.05559433e-01 4.55950081e-01 -4.52944785e-01 1.46821833e+00 -9.43014026e-01 8.54758024e-02 2.26423949e-01 6.15585923e-01 -2.02327594e-01 -1.59252620e+00 -4.84420151e-01 5.94116636e-02 -4.01389986e-01 -3.15306515e-01 -2.65888155e-01 5.75555384e-01 6.47136331e-01 9.01458502e-01 5.19910492e-02 -9.59623873e-01 3.19829613e-01 -4.14248323e-03 5.80239952e-01 -3.74375612e-01 -1.00534785e+00 7.68716574e-01 1.25489131e-01 -1.91553175e-01 -2.18938410e-01 -6.24704003e-01 -1.35773575e+00 -4.80124116e-01 -2.42068440e-01 3.69516671e-01 3.09690595e-01 4.94761705e-01 2.34763280e-01 1.08040619e+00 6.10336244e-01 -8.84791136e-01 -2.64607936e-01 -1.25223553e+00 -1.08871663e+00 5.61409235e-01 -9.69865993e-02 -5.90368986e-01 -7.04149231e-02 1.44862920e-01]

[15.986462593078613, 5.502670764923096]

35a3125d-73e9-4c68-b0b8-40d569fb7e11

using-a-conditional-generative-adversarial

2211.15807

https://arxiv.org/abs/2211.15807v1

https://arxiv.org/pdf/2211.15807v1.pdf

Using a Conditional Generative Adversarial Network to Control the Statistical Characteristics of Generated Images for IACT Data Analysis

Generative adversarial networks are a promising tool for image generation in the astronomy domain. Of particular interest are conditional generative adversarial networks (cGANs), which allow you to divide images into several classes according to the value of some property of the image, and then specify the required class when generating new images. In the case of images from Imaging Atmospheric Cherenkov Telescopes (IACTs), an important property is the total brightness of all image pixels (image size), which is in direct correlation with the energy of primary particles. We used a cGAN technique to generate images similar to whose obtained in the TAIGA-IACT experiment. As a training set, we used a set of two-dimensional images generated using the TAIGA Monte Carlo simulation software. We artificiallly divided the training set into 10 classes, sorting images by size and defining the boundaries of the classes so that the same number of images fall into each class. These classes were used while training our network. The paper shows that for each class, the size distribution of the generated images is close to normal with the mean value located approximately in the middle of the corresponding class. We also show that for the generated images, the total image size distribution obtained by summing the distributions over all classes is close to the original distribution of the training set. The results obtained will be useful for more accurate generation of realistic synthetic images similar to the ones taken by IACTs.

['Anna Vlaskina', 'Elizaveta Gres', 'Stanislav Polyakov', 'Andrey Demichev', 'Alexander Kryukov', 'Julia Dubenskaya']

2022-11-28

['astronomy']

['miscellaneous']

[ 1.45981148e-01 -2.88219657e-02 5.66500723e-01 -1.97084434e-02 -3.40555906e-01 -7.43530095e-01 7.70879388e-01 -2.61316329e-01 -5.87976515e-01 8.97638381e-01 -4.19940919e-01 -3.65460783e-01 8.96172151e-02 -1.35808659e+00 -7.87572205e-01 -1.16741455e+00 1.12419941e-01 7.79385209e-01 4.74505097e-01 -1.34038076e-01 1.00924149e-01 9.34062481e-01 -1.63111782e+00 1.69479072e-01 6.46698654e-01 7.66556144e-01 1.91440210e-01 9.88086164e-01 -1.72009170e-01 5.38583696e-01 -1.09674132e+00 3.72418799e-02 5.55628359e-01 -9.49274957e-01 -6.33216500e-01 1.89959794e-01 1.46587268e-01 1.06546534e-02 -1.84358984e-01 1.24922681e+00 2.25647897e-01 3.16050172e-01 1.18004048e+00 -1.10178649e+00 -2.32518107e-01 1.53295398e-01 -2.75125265e-01 1.29130051e-01 -7.25376531e-02 1.27498791e-01 1.52768895e-01 -2.98544377e-01 7.25889981e-01 9.39608574e-01 3.19363505e-01 4.74645585e-01 -1.16753078e+00 -5.07738292e-01 -6.17453754e-01 7.89994895e-02 -1.28396654e+00 4.36321199e-02 4.20417666e-01 -5.10300756e-01 3.39805454e-01 5.30281305e-01 7.68999279e-01 6.89591467e-01 2.99144179e-01 7.82474652e-02 1.44867468e+00 -9.03790474e-01 4.60179985e-01 4.27645713e-01 -4.66628283e-01 4.90543872e-01 2.70665020e-01 3.72750759e-01 2.65854031e-01 -1.00048430e-01 9.17427540e-01 -2.23298430e-01 -3.24382067e-01 -2.84397453e-01 -1.08789480e+00 9.21120703e-01 6.49003029e-01 7.40820467e-01 -3.74652594e-01 6.64712563e-02 -1.75291494e-01 9.48675424e-02 2.15647548e-01 4.97320354e-01 -8.04010592e-03 3.30824494e-01 -7.59541571e-01 4.99411523e-01 8.05718780e-01 4.92644310e-01 8.21524262e-01 1.62676767e-01 -3.02485768e-02 3.72131616e-01 -8.71596634e-02 8.64924490e-01 7.44832039e-01 -8.50077450e-01 4.29007113e-02 3.59301329e-01 2.61259973e-01 -6.44308984e-01 -7.59608075e-02 -2.52935976e-01 -6.94711864e-01 9.04637754e-01 8.86044621e-01 -2.05458641e-01 -1.09281683e+00 1.45691895e+00 1.95819512e-01 7.65248807e-03 3.28298718e-01 5.44735551e-01 6.51593626e-01 1.01442766e+00 -1.78719357e-01 -2.53627062e-01 1.24383414e+00 -4.11035895e-01 -2.49478251e-01 1.27930105e-01 6.67385906e-02 -8.15907657e-01 7.86090732e-01 3.10861528e-01 -8.29872727e-01 -7.52073109e-01 -1.07058847e+00 6.33734226e-01 -5.25492370e-01 -4.82235104e-03 3.04771423e-01 7.72400022e-01 -7.85347998e-01 5.86308897e-01 -5.80191672e-01 -9.41371620e-02 -2.20798217e-02 -1.81101169e-02 -1.58626482e-01 3.07170898e-01 -1.11482441e+00 7.31183887e-01 6.31246507e-01 -3.03937137e-01 -1.08992565e+00 -3.15572381e-01 -5.17990351e-01 1.41704381e-01 -1.72297925e-01 -4.30418700e-01 1.03556263e+00 -1.19919205e+00 -1.26977348e+00 1.01390231e+00 1.26397699e-01 -5.46280980e-01 5.44299424e-01 6.09849393e-01 -3.43911231e-01 3.35124791e-01 -1.47450909e-01 6.76835239e-01 1.00911355e+00 -1.57118690e+00 -3.68988305e-01 -2.80062798e-02 -5.30062318e-02 -1.36074126e-01 1.77130654e-01 -2.85367761e-02 -7.91446567e-02 -3.82204294e-01 -4.26789969e-02 -1.11989427e+00 -3.50138545e-01 -4.36012626e-01 -4.47434932e-01 9.27078128e-02 7.36372232e-01 -2.80744731e-01 3.86043876e-01 -2.15136671e+00 -7.60674700e-02 4.17276144e-01 -2.23242901e-02 3.14135492e-01 1.06500134e-01 3.82040083e-01 -5.04141212e-01 2.35501993e-02 -4.68148500e-01 1.57814384e-01 -2.66073227e-01 1.38369441e-01 -3.10706943e-01 3.16353917e-01 -1.06545955e-01 6.09755039e-01 -5.43075264e-01 -3.22474182e-01 3.87408286e-01 3.34471077e-01 -7.51561821e-02 4.82774675e-01 -4.39074218e-01 5.57383478e-01 -3.90509933e-01 -9.55546945e-02 8.76083434e-01 1.90990239e-01 -1.57731652e-01 6.42011687e-02 -1.79215103e-01 -4.64310437e-01 -9.24572349e-01 6.26808941e-01 -4.27920073e-01 7.58470714e-01 -1.91711292e-01 -8.55154097e-01 1.06248951e+00 2.98843473e-01 3.05006236e-01 -5.65089285e-01 3.30786318e-01 7.02602789e-02 3.40258777e-01 -4.73184109e-01 2.60267317e-01 -5.71563601e-01 1.07404627e-01 5.48126280e-01 -3.35870199e-02 -9.55028236e-01 4.26689416e-01 1.01371966e-01 7.08559155e-01 -2.07696080e-01 1.11332916e-01 -1.06697850e-01 4.49449092e-01 1.35245383e-01 8.10859799e-02 7.73794949e-01 1.78096116e-01 9.51793909e-01 7.88189292e-01 -4.36420828e-01 -1.49440491e+00 -1.22566783e+00 -2.41799131e-01 1.31295994e-01 6.37355447e-02 2.08831444e-01 -1.18953204e+00 -3.93633038e-01 -3.78730714e-01 7.86289394e-01 -7.29457140e-01 -6.11888692e-02 -3.82651806e-01 -1.20831048e+00 3.74521822e-01 -1.07384399e-01 6.41567588e-01 -1.71330178e+00 -6.27057314e-01 -2.28102267e-01 -4.67919782e-02 -9.65066850e-01 1.78600043e-01 1.06056690e-01 -6.54959202e-01 -1.24616361e+00 -9.70989347e-01 -6.14725769e-01 1.05912292e+00 -4.74535562e-02 1.10776329e+00 -1.18234493e-01 -4.46633279e-01 1.37615830e-01 -4.43118334e-01 -6.60020709e-01 -1.22716892e+00 -2.77983040e-01 -2.03563288e-01 8.96343403e-03 -1.96447983e-01 -4.81551498e-01 -3.92931163e-01 2.84388691e-01 -1.29964697e+00 -1.57621667e-01 2.45627150e-01 5.60671866e-01 5.94983757e-01 7.77390838e-01 1.25084594e-01 -8.65467131e-01 4.54154015e-01 -1.85437754e-01 -1.18895829e+00 2.63181385e-02 -2.18741134e-01 8.41295570e-02 9.58352745e-01 -3.79241854e-01 -1.08301747e+00 5.24896793e-02 -2.32091412e-01 -2.55120218e-01 -4.84527647e-01 -1.41160237e-03 -6.84366822e-02 -3.99250627e-01 9.02199566e-01 3.66081595e-01 8.78879130e-02 -1.94120444e-02 1.14528120e-01 4.90058124e-01 7.54244685e-01 -2.12308586e-01 1.13133109e+00 4.07555193e-01 2.64341742e-01 -1.02839649e+00 -3.41404885e-01 4.63654399e-02 -2.08484635e-01 -3.47636729e-01 9.58753943e-01 -2.49332637e-01 -5.81939042e-01 7.59021223e-01 -1.01043081e+00 -3.14919978e-01 -7.82650650e-01 6.18996322e-01 -6.14148736e-01 3.28442216e-01 -3.68084490e-01 -8.69061291e-01 -1.45314217e-01 -1.17928338e+00 4.80679959e-01 7.19417810e-01 3.77054095e-01 -9.69618380e-01 1.83055803e-01 -8.73174518e-03 2.66990095e-01 5.27452052e-01 1.17434418e+00 -4.10818428e-01 -5.62051415e-01 -5.85769057e-01 -3.03447191e-02 7.37857819e-01 -3.70858610e-03 2.21360356e-01 -8.79642546e-01 -8.77162814e-02 4.97225106e-01 -4.32517678e-02 8.64727557e-01 6.05036318e-01 1.24622571e+00 9.57197919e-02 -1.81679606e-01 4.73372042e-01 1.64603770e+00 7.02868640e-01 1.18488538e+00 2.93985397e-01 2.87646264e-01 3.52990836e-01 3.29188824e-01 1.31488860e-01 -4.11163270e-01 7.35384345e-01 6.05428636e-01 -4.09087747e-01 -1.23198360e-01 1.08737633e-01 2.44855627e-01 7.83144310e-02 -2.92174071e-01 -5.56156099e-01 -5.81772327e-01 3.51725608e-01 -1.24393821e+00 -1.22643185e+00 -4.19476002e-01 2.57812858e+00 4.72210705e-01 1.64943293e-01 5.14805876e-02 2.63804823e-01 8.41398537e-01 5.58559820e-02 -1.33823708e-01 -4.02275681e-01 -2.38933668e-01 5.98500192e-01 6.79560065e-01 6.05488300e-01 -8.51453662e-01 3.44261378e-01 7.07739592e+00 9.32647467e-01 -1.04415774e+00 -1.24727540e-01 5.88683605e-01 2.41275713e-01 -3.14637482e-01 1.00528955e-01 -4.53752398e-01 6.93581283e-01 8.77285838e-01 -3.32795203e-01 4.15327638e-01 4.97546554e-01 2.96106208e-02 -7.46993482e-01 -4.36514348e-01 6.12343490e-01 -1.34041041e-01 -1.11465728e+00 -9.22158435e-02 2.77131230e-01 1.00363922e+00 -7.46570751e-02 -6.60582259e-02 -9.74554792e-02 2.61286288e-01 -9.98418570e-01 4.89777237e-01 6.02250516e-01 6.45986557e-01 -8.94765913e-01 1.03226399e+00 5.64351916e-01 -6.79427862e-01 3.01776230e-01 -6.05621099e-01 2.12751493e-01 1.13169439e-01 7.77641177e-01 -9.65503156e-01 4.90776837e-01 4.61712122e-01 -2.05049232e-01 -6.43612146e-01 1.03788805e+00 -2.72660762e-01 6.59590662e-01 -5.06053209e-01 -2.00778052e-01 2.17372581e-01 -6.96364641e-01 5.58742702e-01 6.81062818e-01 6.34649873e-01 -5.23746721e-02 -3.49720299e-01 1.09165382e+00 7.95608088e-02 -1.04501039e-01 -6.82830155e-01 7.88433827e-04 1.45884261e-01 1.22744405e+00 -1.10887623e+00 -5.08401155e-01 3.67202908e-02 7.15761364e-01 -3.01300436e-01 4.04315233e-01 -8.37478042e-01 -7.30851889e-01 1.17497966e-01 2.20715523e-01 4.00834203e-01 -5.30156586e-03 4.15604115e-02 -7.56068170e-01 -2.60108918e-01 -5.33652306e-01 5.61825931e-02 -1.14010477e+00 -9.18692946e-01 1.05143535e+00 3.24374169e-01 -1.27924252e+00 -5.50606012e-01 -7.53801346e-01 -9.37632620e-01 1.14233041e+00 -8.37363780e-01 -6.60363376e-01 -5.45596242e-01 4.78847414e-01 1.48132056e-01 -3.62535149e-01 9.28373992e-01 -5.09920381e-02 -1.76686630e-01 -4.95019369e-03 4.48953658e-01 8.07900429e-02 1.75329044e-01 -1.23328125e+00 3.65875006e-01 8.92342210e-01 3.21588606e-01 5.15695587e-02 1.06650603e+00 -3.84171158e-01 -4.01645541e-01 -8.18115652e-01 6.32094502e-01 -2.40268096e-01 2.74729043e-01 -1.63129002e-01 -7.70914078e-01 4.64141101e-01 4.45727974e-01 2.20486224e-02 4.43075038e-02 -6.13033891e-01 2.23313332e-01 6.30935133e-02 -1.34448278e+00 2.69904405e-01 1.96594685e-01 -2.14355707e-01 -3.60415846e-01 5.81684589e-01 3.37497741e-01 -4.19353664e-01 -4.78691876e-01 2.07939923e-01 2.50486821e-01 -1.55596066e+00 8.87003183e-01 -2.95896858e-01 4.77097660e-01 -5.00911951e-01 2.82655329e-01 -1.62353218e+00 -3.29658128e-02 -6.66713119e-02 6.90962791e-01 9.97353375e-01 3.08890998e-01 -7.95295537e-01 9.05393839e-01 -3.53489704e-02 2.85208017e-01 -3.02033156e-01 -7.57650197e-01 -9.23393905e-01 1.84600413e-01 -7.66800493e-02 4.56081301e-01 4.90972430e-01 -6.22601569e-01 -3.53418179e-02 -3.50686088e-02 8.72216225e-02 6.86822295e-01 3.04349273e-01 7.21492350e-01 -1.04564941e+00 -5.67091763e-01 -1.89571321e-01 -5.89061618e-01 -2.52998143e-01 -1.26867956e-02 -6.98609233e-01 1.21619470e-01 -1.33152699e+00 1.09318562e-01 -6.56642437e-01 1.85999483e-01 4.26440649e-02 1.13074414e-01 5.97857773e-01 2.24030703e-01 1.15006559e-01 3.40906829e-01 2.54100084e-01 1.37734401e+00 4.72049005e-02 5.25498688e-02 4.65790629e-01 -4.64191251e-02 7.50434697e-01 9.57723081e-01 -7.22823679e-01 -2.99899638e-01 2.35202104e-01 1.05537377e-01 7.36697391e-02 5.55449188e-01 -1.46141040e+00 -3.42008144e-01 -1.62321597e-01 6.24917865e-01 -4.59791303e-01 2.87236810e-01 -8.28753948e-01 7.19348967e-01 6.35120571e-01 9.12435278e-02 -3.13994497e-01 1.17111288e-01 -4.26624678e-02 -3.39033127e-01 -1.05484426e+00 1.21911836e+00 -4.25168365e-01 -3.67680550e-01 -6.90280572e-02 -6.75325215e-01 -9.77741554e-02 1.25679529e+00 -6.33178949e-02 -1.92866176e-01 -7.06578553e-01 -6.85798824e-01 -4.61777836e-01 6.52081907e-01 -1.30742982e-01 2.84816712e-01 -1.20586216e+00 -6.01404548e-01 4.76809323e-01 -4.31516916e-02 3.78380306e-02 2.54815936e-01 2.89550424e-01 -1.06422639e+00 7.56839290e-02 -5.96813500e-01 -3.28693897e-01 -1.14168322e+00 6.11250043e-01 8.37275207e-01 -2.10688025e-01 -2.85152107e-01 6.00046039e-01 4.62869048e-01 -2.50028938e-01 -3.86835933e-01 -1.04404368e-01 -2.78118074e-01 -2.05088675e-01 5.32791674e-01 7.77468681e-02 1.53556913e-01 -4.74560708e-01 1.78109050e-01 6.04763925e-01 3.63165647e-01 -3.83598208e-01 1.13296175e+00 5.13077140e-01 -2.28373170e-01 1.93138719e-01 9.06607389e-01 3.14856112e-01 -1.10168016e+00 3.22766900e-01 -6.37231052e-01 -3.24770600e-01 -1.49489224e-01 -5.92670083e-01 -1.28813267e+00 6.51287377e-01 7.98867881e-01 8.68210018e-01 1.32908940e+00 7.46455118e-02 3.08299601e-01 1.76085625e-02 3.06515723e-01 -6.17152750e-01 -1.61247253e-01 3.75346065e-01 8.60857487e-01 -7.40289807e-01 -6.65128008e-02 -4.36539441e-01 -5.39338112e-01 1.14598024e+00 3.90113086e-01 -5.01044989e-01 3.97895545e-01 3.32423508e-01 2.32713759e-01 -1.73565656e-01 -2.04301685e-01 -1.28962651e-01 7.72274509e-02 7.79887736e-01 9.70607772e-02 1.49217904e-01 -4.38442111e-01 -1.87404230e-02 -3.86294127e-01 -1.49165258e-01 8.84501934e-01 5.44700325e-01 -4.72073883e-01 -1.33533204e+00 -1.02414191e+00 3.36797953e-01 -2.69671679e-01 5.42248897e-02 -3.00780714e-01 9.53512907e-01 4.00900543e-01 5.75526595e-01 3.84384632e-01 -5.55624953e-03 2.78878927e-01 -8.50635860e-03 8.17721784e-01 -3.09342653e-01 -4.56158340e-01 -2.35395595e-01 -1.29459038e-01 6.53718188e-02 -2.67321557e-01 -5.20818651e-01 -1.28769326e+00 -2.51663059e-01 -1.74657539e-01 5.56840777e-01 9.00665283e-01 7.55930603e-01 -4.16470766e-01 5.69353521e-01 9.10693049e-01 -1.08363879e+00 -1.47694930e-01 -1.07335579e+00 -9.04686868e-01 4.49484944e-01 1.22099172e-03 -3.45380902e-01 -7.28487194e-01 2.09636912e-01]

[11.641218185424805, -0.4896637797355652]

bbad3453-0748-4e9b-a74a-0077f8dde352

home-homography-equivariant-video

2306.01623

https://arxiv.org/abs/2306.01623v1

https://arxiv.org/pdf/2306.01623v1.pdf

HomE: Homography-Equivariant Video Representation Learning

Recent advances in self-supervised representation learning have enabled more efficient and robust model performance without relying on extensive labeled data. However, most works are still focused on images, with few working on videos and even fewer on multi-view videos, where more powerful inductive biases can be leveraged for self-supervision. In this work, we propose a novel method for representation learning of multi-view videos, where we explicitly model the representation space to maintain Homography Equivariance (HomE). Our method learns an implicit mapping between different views, culminating in a representation space that maintains the homography relationship between neighboring views. We evaluate our HomE representation via action recognition and pedestrian intent prediction as downstream tasks. On action classification, our method obtains 96.4% 3-fold accuracy on the UCF101 dataset, better than most state-of-the-art self-supervised learning methods. Similarly, on the STIP dataset, we outperform the state-of-the-art by 6% for pedestrian intent prediction one second into the future while also obtaining an accuracy of 91.2% for pedestrian action (cross vs. not-cross) classification. Code is available at https://github.com/anirudhs123/HomE.

['Ehsan Adeli', 'Li Fei-Fei', 'Juan Carlos Niebles', 'Jiajun Wu', 'Adrien Gaidon', 'Anirudh Sriram']

2023-06-02

['action-classification', 'action-recognition-in-videos']

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

[ 2.59288818e-01 2.39658505e-02 -8.13765883e-01 -5.51247180e-01 -8.97643745e-01 -4.43074137e-01 8.30190599e-01 -1.04419664e-01 -2.64966100e-01 6.14071012e-01 6.67102575e-01 -2.99654342e-02 4.09429848e-01 -6.05516851e-01 -9.46293771e-01 -6.08195007e-01 1.30298033e-01 2.21657664e-01 1.76681265e-01 -3.70431133e-02 9.81989577e-02 2.67540086e-02 -1.59800971e+00 7.14071810e-01 5.29381454e-01 7.41920590e-01 -1.75760701e-01 6.05062306e-01 5.29007494e-01 1.18109488e+00 -7.59462193e-02 -5.79285145e-01 2.66975105e-01 -3.53157580e-01 -9.17433619e-01 3.41257781e-01 1.06860471e+00 -5.48856318e-01 -7.47451246e-01 6.82896852e-01 3.57170790e-01 1.43567935e-01 7.24063933e-01 -1.36698067e+00 -8.80162537e-01 3.76095980e-01 -6.65169120e-01 3.43963265e-01 6.71219647e-01 2.14890569e-01 1.13712585e+00 -7.68582344e-01 7.96383023e-01 1.16168332e+00 5.51374316e-01 6.17843568e-01 -1.41434526e+00 -7.29100525e-01 4.17684019e-01 5.97423673e-01 -1.21996868e+00 -7.48424113e-01 7.71602452e-01 -5.70853770e-01 9.78349328e-01 4.00702432e-02 7.24614561e-01 1.67239034e+00 8.49550515e-02 1.26081908e+00 1.18630564e+00 -1.96942687e-01 -3.70953791e-02 2.57598925e-02 1.41944736e-01 7.98045218e-01 1.41462192e-01 2.72816658e-01 -6.01763725e-01 1.78260982e-01 7.76847422e-01 3.66146892e-01 -2.00056270e-01 -8.58659387e-01 -1.24295247e+00 9.21721101e-01 5.46305299e-01 4.14737873e-02 -5.64026833e-02 7.82240480e-02 6.64206982e-01 2.05341011e-01 5.11170566e-01 9.12920609e-02 -3.42176110e-01 -1.56258985e-01 -6.59392595e-01 1.23122744e-01 4.82188851e-01 8.94782603e-01 6.44208193e-01 -1.77267566e-02 2.53588669e-02 7.28715241e-01 1.45424724e-01 4.58574533e-01 6.00989819e-01 -1.05724907e+00 6.83909833e-01 5.10323763e-01 -1.35243401e-01 -9.77988839e-01 -2.31111988e-01 -3.03162724e-01 -9.88927603e-01 1.63749337e-01 4.59453315e-01 5.43156713e-02 -7.41634786e-01 1.71431231e+00 5.23959845e-02 3.10647488e-01 1.52396455e-01 7.75568426e-01 8.41883838e-01 4.98406529e-01 2.01124907e-01 1.17067695e-01 1.14780021e+00 -1.32309508e+00 -2.73159295e-01 -3.45024467e-01 8.46424222e-01 -4.05809909e-01 8.25160801e-01 3.41835856e-01 -7.76873827e-01 -8.02768171e-01 -8.12640786e-01 -3.05256844e-02 -3.58604282e-01 3.33080560e-01 8.32589626e-01 5.37910044e-01 -9.84515488e-01 5.04314303e-01 -8.85394812e-01 -7.12561071e-01 6.77347839e-01 1.07847646e-01 -1.02125168e+00 -4.48765695e-01 -8.74566734e-01 8.39365065e-01 9.22376812e-02 -3.66757959e-01 -1.06529963e+00 -7.26998508e-01 -1.21234632e+00 -3.43201399e-01 4.00770545e-01 -7.05387533e-01 8.96637559e-01 -1.10922980e+00 -1.22544003e+00 1.29722774e+00 -2.36111239e-01 -7.37204611e-01 7.41162419e-01 -5.03626347e-01 -3.67756099e-01 2.86622822e-01 4.90599275e-01 9.97443497e-01 8.02924395e-01 -1.26324868e+00 -6.80817127e-01 -4.88543540e-01 2.58820176e-01 1.45769104e-01 -2.94511825e-01 -1.63754344e-01 -4.17644531e-01 -7.45345354e-01 -4.58001252e-03 -1.15532625e+00 -5.79221211e-02 1.87438220e-01 -1.65329322e-01 -2.59322405e-01 8.41759980e-01 -7.99320102e-01 9.06026185e-01 -2.07003999e+00 2.90775627e-01 -3.49120259e-01 -1.13557400e-02 1.59853712e-01 -2.06240773e-01 3.70423019e-01 -5.50111592e-01 -6.56929314e-02 -7.37921074e-02 -4.91420388e-01 -2.35335991e-01 1.37062311e-01 -4.43354696e-01 7.51358509e-01 1.99042529e-01 8.63669932e-01 -1.06507623e+00 -4.46018040e-01 4.36201662e-01 4.93982255e-01 -7.33543754e-01 3.14199239e-01 1.69396371e-01 5.80221593e-01 -1.28155783e-01 7.29030550e-01 3.29927474e-01 -5.34567773e-01 1.93839431e-01 -4.17075992e-01 -4.47028615e-02 1.73922271e-01 -9.62988555e-01 1.87091303e+00 -4.59445894e-01 7.42841840e-01 -3.41438383e-01 -1.30676961e+00 7.57028282e-01 1.75155431e-01 7.13049293e-01 -6.46529317e-01 -1.76353067e-01 -2.41441846e-01 -3.32482070e-01 -4.60418016e-01 2.29042560e-01 1.70624837e-01 1.29797617e-02 4.09049690e-01 2.87906170e-01 3.56514007e-01 3.11272770e-01 3.33108187e-01 9.14454818e-01 6.42282605e-01 4.53866065e-01 -3.03687509e-02 5.99521935e-01 -1.84204593e-01 5.82547545e-01 6.09952509e-01 -4.35542494e-01 8.14738870e-01 3.43911648e-01 -7.20071733e-01 -1.03639758e+00 -9.95456517e-01 -8.23381320e-02 1.22616065e+00 3.41848396e-02 -6.54970050e-01 -4.19454545e-01 -1.07669473e+00 1.23014256e-01 5.96423984e-01 -8.69675577e-01 -2.52783120e-01 -7.17049360e-01 -4.75020587e-01 3.53290886e-01 1.00262642e+00 8.24781179e-01 -7.19701707e-01 -4.07303989e-01 -2.60461420e-01 -4.07676309e-01 -1.30527472e+00 -4.34149325e-01 -7.31694102e-02 -9.45350826e-01 -1.30888498e+00 -7.85518587e-01 -5.45923531e-01 7.05668986e-01 6.01865053e-01 1.15469027e+00 -1.15863502e-01 -1.09452866e-01 8.27923179e-01 -4.76506382e-01 1.35223836e-01 -1.17476858e-01 2.75517330e-02 2.07979500e-01 1.59696862e-01 3.69713873e-01 -5.40380895e-01 -6.10446692e-01 5.45445979e-01 -4.71471518e-01 2.28806004e-01 5.32903790e-01 1.00019968e+00 5.42786837e-01 -4.22438562e-01 4.23752248e-01 -8.44986856e-01 -3.32159758e-01 -6.35591030e-01 -2.69843638e-01 1.64870888e-01 -4.08459932e-01 -7.25913420e-03 6.29905522e-01 -3.48040223e-01 -1.04248846e+00 3.65460455e-01 6.86994269e-02 -6.19301438e-01 -5.53191543e-01 1.28738089e-02 -1.88723683e-01 8.94199237e-02 6.17843688e-01 4.34910387e-01 7.88731799e-02 -3.09229165e-01 4.18112069e-01 4.16242123e-01 4.29618835e-01 -4.82148796e-01 5.57474911e-01 8.36234868e-01 -1.17268667e-01 -7.99411952e-01 -1.22281623e+00 -5.60660422e-01 -1.22468579e+00 -3.26110661e-01 1.06373572e+00 -1.46530020e+00 -3.98468107e-01 3.54554981e-01 -6.38401091e-01 -3.31366807e-01 -1.32032856e-01 4.98012275e-01 -8.89917135e-01 5.67723334e-01 -4.06887025e-01 -5.96771121e-01 2.47309860e-02 -9.74729538e-01 1.22685635e+00 -1.10621236e-01 -2.88657486e-01 -9.48355138e-01 1.08928233e-01 1.11877203e+00 3.34365256e-02 2.07787693e-01 3.89330596e-01 -5.46310902e-01 -6.11557066e-01 -2.44533211e-01 -2.21925735e-01 4.08443123e-01 1.94339216e-01 -1.59103245e-01 -9.97055292e-01 -5.22383809e-01 -5.20190120e-01 -7.88768530e-01 1.36988437e+00 2.45664179e-01 1.12871456e+00 -3.16935331e-01 -4.35176939e-01 7.10370541e-01 1.05044746e+00 -1.47198498e-01 7.66607821e-01 4.26428020e-01 8.68497729e-01 5.63990653e-01 6.21296108e-01 3.75349998e-01 6.98497474e-01 8.40280235e-01 3.66940528e-01 6.35754913e-02 -2.94541806e-01 -7.23836482e-01 6.52426958e-01 3.21767449e-01 -3.32594335e-01 -4.05762009e-02 -8.30710649e-01 6.22874677e-01 -2.11278486e+00 -1.41488218e+00 1.79400176e-01 2.22250819e+00 5.21961570e-01 6.06085397e-02 5.15546978e-01 1.23564705e-01 6.22626662e-01 5.38308859e-01 -5.93941271e-01 2.06048265e-01 -1.09789871e-01 -3.33962739e-01 4.82355267e-01 2.35005468e-01 -1.81338644e+00 1.09314406e+00 5.98918962e+00 3.85545284e-01 -8.72797132e-01 1.26248719e-02 9.05492783e-01 -6.90935105e-02 1.28653765e-01 -2.20832112e-03 -9.34318364e-01 3.24913412e-01 7.46770680e-01 2.75999963e-01 3.92909758e-02 1.10436416e+00 1.94936916e-02 -3.86263393e-02 -1.30026698e+00 1.20784032e+00 6.50959313e-01 -1.39276731e+00 2.80430168e-01 9.51952673e-03 8.82277489e-01 -2.19937935e-02 8.14013109e-02 4.61587876e-01 2.89845258e-01 -9.23298180e-01 5.21241307e-01 5.65397978e-01 7.03315794e-01 -6.18455529e-01 5.82310021e-01 1.95370659e-01 -1.22393906e+00 -1.97890729e-01 -3.20093244e-01 -1.37711853e-01 1.42482221e-01 9.14335698e-02 -3.95329803e-01 5.11048794e-01 9.25145924e-01 1.73864162e+00 -9.08338904e-01 7.10715830e-01 -3.07520568e-01 5.88362098e-01 3.42780165e-02 4.11066622e-01 9.92935300e-02 -2.34232649e-01 3.20360988e-01 1.24051654e+00 -9.72327292e-02 1.13382116e-01 4.21803594e-01 3.95335495e-01 -8.33685249e-02 1.11230556e-02 -1.06944132e+00 1.48078904e-01 5.29337302e-03 1.03816557e+00 -4.21881676e-01 -4.78372842e-01 -7.70814061e-01 1.13659477e+00 6.01920962e-01 2.64046460e-01 -8.45977247e-01 3.17873359e-01 8.21255922e-01 2.48057500e-01 5.30269146e-01 -2.32170001e-01 -1.80640928e-02 -1.69567561e+00 -6.53449297e-02 -1.07140648e+00 7.51131415e-01 -6.42023861e-01 -1.48447120e+00 2.95898736e-01 1.74850866e-01 -1.55796993e+00 -4.26436841e-01 -7.12792337e-01 -4.80295539e-01 1.52493030e-01 -1.37747717e+00 -1.60262859e+00 -2.82667220e-01 7.27580130e-01 7.07160413e-01 -5.74770808e-01 8.48789096e-01 2.97887832e-01 -7.15149462e-01 7.02647269e-01 -1.02217518e-01 5.29806912e-01 1.08194506e+00 -1.09345913e+00 2.52953947e-01 7.87655890e-01 3.63764912e-01 4.41480339e-01 3.31234276e-01 -5.98471403e-01 -1.40989387e+00 -1.22860980e+00 6.80417478e-01 -8.26102078e-01 7.34649897e-01 -3.15694928e-01 -6.12349212e-01 1.09375489e+00 2.19112769e-01 4.09502238e-01 9.99816298e-01 3.16206753e-01 -9.59731340e-01 -1.53538048e-01 -8.57108295e-01 5.65889597e-01 1.49287689e+00 -6.32935286e-01 -5.46051383e-01 5.11002898e-01 3.12009424e-01 -2.23229349e-01 -8.52307796e-01 4.36870664e-01 6.99110150e-01 -1.25585556e+00 1.26981235e+00 -1.16613400e+00 7.93420374e-01 -1.90717801e-01 -4.72425193e-01 -1.02980149e+00 -5.52254021e-01 -2.08463922e-01 -2.95880318e-01 1.22191036e+00 2.28950679e-01 -4.52475667e-01 9.35294509e-01 3.95263344e-01 7.59056062e-02 -7.02505410e-01 -8.47075582e-01 -7.79690623e-01 -2.22347025e-02 -2.22342223e-01 -2.10045297e-02 1.07564342e+00 -4.63184305e-02 4.80373889e-01 -8.05525780e-01 5.78915887e-03 9.00397241e-01 3.29619586e-01 1.03675401e+00 -9.26445067e-01 -3.25093567e-01 -2.38387257e-01 -8.34074080e-01 -1.22407591e+00 5.58452010e-01 -9.68788862e-01 -4.28667873e-01 -1.32395160e+00 6.36341870e-01 3.38996723e-02 -2.16661379e-01 6.31704509e-01 -1.20041072e-01 4.66931701e-01 2.99995720e-01 2.74513870e-01 -1.07932723e+00 8.04606140e-01 9.38475311e-01 -3.16712201e-01 2.23764688e-01 3.92862074e-02 -6.83525741e-01 8.58783484e-01 8.45345795e-01 -1.84699014e-01 -3.71494710e-01 -4.07190859e-01 -2.48494700e-01 -1.72575295e-01 7.38184094e-01 -1.26067793e+00 9.51268673e-02 -4.61398177e-02 8.01192284e-01 -6.09106481e-01 4.70665514e-01 -5.86058140e-01 -1.16264462e-01 3.58830601e-01 -5.26477098e-01 1.09264277e-01 -8.82351026e-02 8.59244645e-01 -2.47884721e-01 8.95076767e-02 8.68259609e-01 -1.88684076e-01 -1.02585804e+00 4.04854745e-01 -2.32869223e-01 1.88165858e-01 1.16228664e+00 -8.54518861e-02 -3.51781398e-01 -6.12496674e-01 -7.84418702e-01 1.28390029e-01 5.94044685e-01 7.86137402e-01 6.69533551e-01 -1.52253807e+00 -7.62633562e-01 3.13015908e-01 4.05837923e-01 -4.64369446e-01 4.05614167e-01 6.91660821e-01 -7.90006369e-02 5.73542655e-01 -5.30294418e-01 -7.96644747e-01 -1.44465363e+00 5.31293929e-01 2.18082696e-01 -2.83456326e-01 -7.93101251e-01 5.76800704e-01 3.16377103e-01 -4.52871114e-01 2.68649369e-01 8.88555348e-02 -3.85207117e-01 1.92244664e-01 6.41384363e-01 5.44304252e-01 -4.35965925e-01 -1.03564250e+00 -4.06635374e-01 6.96004450e-01 -2.99443454e-01 1.17550962e-01 1.42950678e+00 7.61405602e-02 2.50271678e-01 4.70866293e-01 1.44785058e+00 -3.03315401e-01 -1.61222637e+00 -2.32534245e-01 -2.60790586e-01 -7.62826979e-01 -1.23331346e-01 -5.30223310e-01 -1.16859090e+00 8.32587957e-01 6.99459791e-01 -3.28610897e-01 7.04880536e-01 2.15178296e-01 5.55668294e-01 5.74477255e-01 4.17820305e-01 -9.77108955e-01 5.30885339e-01 5.57331979e-01 1.06748831e+00 -1.88363171e+00 2.94239312e-01 -3.45162898e-01 -1.05350828e+00 9.48964059e-01 9.19130087e-01 -3.65412802e-01 6.49652004e-01 -2.48792931e-01 -1.50458544e-01 7.46495202e-02 -7.62261689e-01 -2.49025986e-01 3.16623896e-01 8.50255311e-01 6.75156116e-01 2.98939622e-03 2.82111079e-01 3.56721282e-01 -4.36484255e-02 -3.04830167e-02 3.02036732e-01 8.96286249e-01 -1.50383890e-01 -1.03943336e+00 -1.27418786e-01 4.50262606e-01 -1.75378963e-01 3.46219428e-02 -3.68411809e-01 8.10803950e-01 -2.28872374e-02 9.16562140e-01 1.62630066e-01 -5.29676914e-01 4.11878020e-01 5.72009496e-02 5.45684934e-01 -5.87386250e-01 -1.96236402e-01 -2.02600852e-01 3.10033858e-01 -9.60421681e-01 -8.41150582e-01 -1.13948381e+00 -7.62245238e-01 -1.72615111e-01 1.93024457e-01 -3.15288752e-01 8.21514130e-02 7.81041026e-01 4.76167113e-01 1.03726827e-01 6.46852255e-01 -1.01250696e+00 -4.82126504e-01 -8.21793914e-01 -2.33615458e-01 7.46617377e-01 3.24036777e-01 -9.99688804e-01 -3.35617304e-01 3.45413983e-01]

[8.584336280822754, 0.8097463250160217]

66fa1d09-3b5b-49b5-b21c-b178c3aece31

kss-icp-point-cloud-registration-based-on

2211.02807

https://arxiv.org/abs/2211.02807v1

https://arxiv.org/pdf/2211.02807v1.pdf

KSS-ICP: Point Cloud Registration based on Kendall Shape Space

Point cloud registration is a popular topic which has been widely used in 3D model reconstruction, location, and retrieval. In this paper, we propose a new registration method, KSS-ICP, to address the rigid registration task in Kendall shape space (KSS) with Iterative Closest Point (ICP). The KSS is a quotient space that removes influences of translations, scales, and rotations for shape feature-based analysis. Such influences can be concluded as the similarity transformations that do not change the shape feature. The point cloud representation in KSS is invariant to similarity transformations. We utilize such property to design the KSS-ICP for point cloud registration. To tackle the difficulty to achieve the KSS representation in general, the proposed KSS-ICP formulates a practical solution that does not require complex feature analysis, data training, and optimization. With a simple implementation, KSS-ICP achieves more accurate registration from point clouds. It is robust to similarity transformation, non-uniform density, noise, and defective parts. Experiments show that KSS-ICP has better performance than the state of the art.

['Baoquan Zhao', 'Weisi Lin', 'Chenlei Lv']

2022-11-05

['point-cloud-registration']

['computer-vision']

[-2.96374381e-01 -5.52261531e-01 1.19348682e-01 -2.52746493e-01 -5.94685674e-01 -3.51871014e-01 5.67598343e-01 1.28873393e-01 -2.09023744e-01 4.43116240e-02 -9.63118598e-02 7.96810389e-02 -2.76674658e-01 -8.53389323e-01 -5.95737100e-01 -6.80640221e-01 3.77374560e-01 7.65289307e-01 7.09894598e-01 -4.71984386e-01 5.22959232e-01 1.03941071e+00 -1.26671469e+00 -4.96221334e-01 9.37636018e-01 8.94361854e-01 2.70357937e-01 -1.06416427e-01 -2.46295616e-01 -4.30757374e-01 -3.27920109e-01 -1.04452409e-01 6.51157200e-01 -3.55858505e-02 -4.91304457e-01 -6.05755523e-02 -1.01165120e-02 6.91235140e-02 -1.10529713e-01 1.08175528e+00 4.90156323e-01 2.42878169e-01 9.28887308e-01 -1.28348827e+00 -7.62482345e-01 -2.76831180e-01 -9.86802399e-01 -1.04194365e-01 3.25232655e-01 -3.35433900e-01 3.77759308e-01 -1.19133008e+00 4.49718535e-01 1.29878068e+00 8.97272229e-01 1.09660529e-01 -8.44502449e-01 -6.81020081e-01 -2.05155805e-01 5.59670851e-02 -1.98409772e+00 -1.86498731e-01 7.57821918e-01 -3.75656813e-01 5.64443946e-01 6.38121486e-01 6.79025948e-01 1.84038952e-01 2.78082401e-01 3.13297540e-01 8.28040302e-01 -4.21228707e-01 1.98572636e-01 -1.83021307e-01 6.10607266e-02 3.71624172e-01 4.53818589e-01 -2.09369078e-01 -9.27409604e-02 -6.64636791e-01 1.20567632e+00 4.32441413e-01 -2.93833584e-01 -6.50846064e-01 -1.24152243e+00 5.56979060e-01 4.75883931e-01 4.73981708e-01 -2.12772846e-01 -2.98814565e-01 9.24051479e-02 8.81722644e-02 5.44205725e-01 9.44801942e-02 -3.71444017e-01 9.94204134e-02 -6.13286555e-01 2.70923495e-01 4.91917670e-01 1.32946861e+00 7.85190701e-01 -2.50623524e-01 1.79841608e-01 9.23084319e-01 7.39284158e-01 1.05795348e+00 5.75439036e-01 -5.59268653e-01 3.67779732e-01 7.98476398e-01 1.46217600e-01 -1.51385093e+00 -4.07069206e-01 -1.55307084e-01 -9.33327615e-01 -1.61169127e-01 -1.34548068e-01 4.67198581e-01 -5.39942622e-01 1.07111001e+00 8.26043963e-01 6.89012945e-01 -1.05675898e-01 1.03262365e+00 8.38963091e-01 6.37680888e-01 -3.98887366e-01 -2.61730194e-01 1.25587845e+00 -4.31506723e-01 -6.09825015e-01 1.91054136e-01 2.45889053e-01 -1.13696265e+00 8.68181229e-01 -1.33777425e-01 -8.98575425e-01 -4.17158186e-01 -9.88403738e-01 7.16078952e-02 -4.79415767e-02 -1.00915797e-01 2.25168064e-01 3.47870499e-01 -9.41874802e-01 6.53614998e-01 -8.92878175e-01 -4.99375612e-01 4.98689003e-02 5.85655630e-01 -6.34322762e-01 1.55176520e-01 -9.38678741e-01 1.03631866e+00 1.07954107e-01 2.45157793e-01 9.39479843e-02 -4.93559837e-01 -8.25366855e-01 -2.31932804e-01 6.03567176e-02 -6.48625255e-01 8.38765085e-01 -2.19856903e-01 -1.57577574e+00 9.62114453e-01 -4.84565824e-01 2.77509272e-01 2.70674407e-01 6.64788857e-02 -4.49536383e-01 -2.46363267e-01 2.38402024e-01 -1.43628150e-01 8.50952566e-01 -1.32266366e+00 -1.11104213e-01 -8.42589498e-01 -6.57169044e-01 4.67274457e-01 -1.97055086e-01 3.53539199e-01 -7.97298551e-01 -5.85012734e-01 1.22680283e+00 -1.07194793e+00 -2.23546818e-01 -7.19450712e-02 -9.16749239e-02 -3.68432254e-01 8.38787615e-01 -4.57720250e-01 8.18473935e-01 -2.38200521e+00 8.87058750e-02 8.58515859e-01 -1.26680881e-01 2.14289099e-01 -1.21443339e-01 3.41223925e-01 -4.36300561e-02 -3.22882459e-02 -2.74783790e-01 -3.32624972e-01 -1.22137032e-01 2.03706130e-01 -2.04581946e-01 9.81607556e-01 3.18404809e-02 7.48782516e-01 -6.79270267e-01 -6.35527730e-01 3.82633060e-01 5.38431644e-01 -3.01975429e-01 1.58556804e-01 4.01552588e-01 5.47344506e-01 -7.78089345e-01 7.21029282e-01 1.25167561e+00 1.27606899e-01 -5.58905542e-01 -4.58097667e-01 -2.66021878e-01 -2.37774134e-01 -1.64386547e+00 1.60103142e+00 -3.74389067e-02 -7.84564167e-02 9.88713354e-02 -8.35609436e-01 1.75798249e+00 8.21332335e-02 8.53786767e-01 -4.95779604e-01 3.06675583e-01 5.49123108e-01 -3.59795630e-01 -2.38521442e-01 5.37252426e-01 -1.52270630e-01 1.30629703e-01 1.48070216e-01 -3.06322813e-01 -6.77674115e-01 -4.90648001e-01 -2.01657057e-01 6.19043052e-01 -6.37012646e-02 4.07688320e-01 -5.03045022e-01 9.51448321e-01 -7.23185092e-02 7.92780519e-01 1.84850693e-01 5.25204418e-03 8.99778008e-01 -3.31712723e-01 -2.75947601e-01 -9.20375705e-01 -1.13253105e+00 -4.69043702e-01 3.02572519e-01 8.87994409e-01 -2.34298393e-01 -5.23330867e-01 -1.44959137e-01 2.18556017e-01 4.59007397e-02 -1.27441615e-01 -3.92173260e-01 -7.23056018e-01 -6.35149837e-01 2.06004634e-01 2.56151766e-01 4.86425221e-01 -6.41734004e-01 1.48080364e-01 6.18042424e-02 4.04690113e-03 -8.46986949e-01 -8.73320043e-01 -5.46456158e-01 -1.26848388e+00 -1.05234420e+00 -9.08822596e-01 -9.40022051e-01 1.07119524e+00 7.00769305e-01 6.56112015e-01 2.37380430e-01 3.32305543e-02 4.54296559e-01 -4.42531526e-01 -4.23052728e-01 -1.57110676e-01 -7.61471167e-02 4.15584147e-01 1.27482936e-01 5.22746742e-01 -6.71985149e-01 -4.90017354e-01 9.85210538e-01 -8.62868309e-01 -4.02361602e-01 4.77063209e-01 5.68253338e-01 1.09600163e+00 1.20283499e-01 1.93675920e-01 -3.53307784e-01 5.93255460e-01 -3.81171852e-01 -7.52616942e-01 2.33050004e-01 -4.08690393e-01 -1.23766147e-01 2.90044457e-01 -2.38711253e-01 -6.63380623e-01 1.90764502e-01 -1.42103761e-01 -6.50936544e-01 -3.31917889e-02 3.17508042e-01 -3.23737085e-01 -8.73259544e-01 5.09751260e-01 4.95853245e-01 3.78232747e-01 -7.97099113e-01 1.20832756e-01 8.95075202e-01 6.05425000e-01 -7.63733089e-01 1.34908652e+00 7.18158841e-01 3.09476465e-01 -8.93781722e-01 -1.23186484e-01 -9.25292134e-01 -9.49578583e-01 6.44389121e-03 4.62587923e-01 -7.69760013e-01 -7.92815328e-01 5.38137138e-01 -1.37020195e+00 5.19811749e-01 -1.94238529e-01 6.36571527e-01 -4.95205432e-01 6.43574834e-01 -2.68746734e-01 -5.81072986e-01 -5.44889390e-01 -1.16028404e+00 1.22938728e+00 2.53586173e-01 -8.22899956e-03 -8.87109339e-01 2.39133388e-01 8.79530534e-02 3.31863344e-01 3.76002729e-01 3.81000191e-01 -6.27377510e-01 -5.17667174e-01 -4.59201068e-01 -8.17833319e-02 1.50010899e-01 4.56897110e-01 2.23748043e-01 -4.24652636e-01 -4.06359345e-01 5.67148328e-01 5.44932246e-01 1.11604914e-01 1.41345561e-01 9.01155174e-01 -1.16217092e-01 -4.06443805e-01 9.10563350e-01 1.38629520e+00 3.89550269e-01 7.16179073e-01 6.26920819e-01 7.25781143e-01 3.39344919e-01 1.07174373e+00 3.06216031e-01 4.70863193e-01 7.93743551e-01 3.84196281e-01 -2.34158188e-01 3.33118111e-01 -7.22488314e-02 -4.51211398e-03 1.52811646e+00 -4.07632679e-01 5.99229515e-01 -1.05082881e+00 2.59500563e-01 -1.92785859e+00 -6.62396550e-01 -5.90697289e-01 2.28796721e+00 5.75239301e-01 -3.43032211e-01 -2.38917485e-01 9.03075337e-02 1.03573346e+00 -1.63068131e-01 -3.39516193e-01 4.33889143e-02 -9.24838260e-02 1.91552043e-01 6.38450503e-01 3.86683315e-01 -9.58123505e-01 8.04604828e-01 5.99111843e+00 9.07755554e-01 -1.07408524e+00 1.00233711e-01 -7.82877058e-02 5.38720608e-01 -3.69373024e-01 9.55918804e-02 -7.43779600e-01 4.94598120e-01 1.65082425e-01 -3.12463671e-01 2.11854130e-01 9.16169584e-01 1.46283656e-01 1.58225268e-01 -7.10844874e-01 1.37994063e+00 2.04618499e-01 -1.00049257e+00 1.18009545e-01 -8.10998008e-02 7.16173172e-01 1.33441864e-02 -4.56525162e-02 -1.60450891e-01 -1.57029361e-01 -6.74593627e-01 5.70066571e-01 6.81447446e-01 6.62732124e-01 -5.49003184e-01 9.16039884e-01 4.18697029e-01 -1.55207467e+00 5.24947703e-01 -8.77922833e-01 2.43200690e-01 1.87405571e-01 5.59423804e-01 -6.34514272e-01 1.09707034e+00 5.86450577e-01 7.72685170e-01 -3.62113804e-01 1.43516648e+00 2.76259810e-01 -7.21858144e-02 -7.19498813e-01 7.07978681e-02 -1.88798159e-01 -9.15904343e-01 7.15958774e-01 7.22820699e-01 7.94434249e-01 6.24378562e-01 3.00438970e-01 6.54815197e-01 3.61881822e-01 6.69305861e-01 -6.46720886e-01 5.08315265e-01 8.74276996e-01 9.82922852e-01 -7.97929585e-01 -2.61585806e-02 -2.89837509e-01 7.85127223e-01 -1.67536035e-01 9.13398862e-02 -5.09826005e-01 -3.33776712e-01 6.39549732e-01 3.79140139e-01 -8.32320824e-02 -6.28525198e-01 -3.79862309e-01 -1.04155445e+00 1.73617676e-01 -5.89353919e-01 1.34765074e-01 -7.97339916e-01 -1.38672614e+00 5.89373469e-01 2.47799754e-01 -1.69986951e+00 2.16444358e-01 -3.10104519e-01 -7.69424498e-01 1.05327928e+00 -1.40520120e+00 -1.12864411e+00 -5.52259684e-01 1.08416200e+00 -8.04311484e-02 -1.94832355e-01 7.31955111e-01 4.23662990e-01 -1.77810103e-01 3.16915363e-01 2.94123322e-01 -1.19266219e-01 8.08602214e-01 -7.45934963e-01 5.71167290e-01 5.40486395e-01 -1.54931679e-01 1.01814580e+00 5.87638795e-01 -9.12457585e-01 -1.57378829e+00 -8.80564809e-01 6.74344897e-01 -2.73516953e-01 4.75260466e-01 6.82506617e-03 -1.05380738e+00 5.77215374e-01 -5.34074605e-01 3.41217935e-01 4.58457708e-01 -1.03521600e-01 -7.56992102e-02 -4.23985034e-01 -1.44711900e+00 3.41684759e-01 1.06305814e+00 -3.95290792e-01 -9.00716901e-01 4.25547510e-01 6.27908409e-01 -8.63408506e-01 -1.19319153e+00 6.80007756e-01 2.61491150e-01 -6.19818687e-01 1.30449545e+00 7.81450700e-03 -4.64605659e-01 -8.09098125e-01 -1.25187650e-01 -1.08270097e+00 -5.45548141e-01 -4.69809115e-01 4.31886047e-01 1.22790158e+00 -2.56114125e-01 -1.09720445e+00 5.43798447e-01 7.07014501e-01 -3.26356828e-01 -4.88348335e-01 -1.34915376e+00 -1.08033645e+00 -1.09921493e-01 7.08367582e-03 1.22802269e+00 1.21296203e+00 -1.65966481e-01 -1.60594910e-01 -4.64021973e-02 5.54572940e-01 7.32992470e-01 1.36952296e-01 1.03686345e+00 -1.83340895e+00 3.67508948e-01 -1.19060174e-01 -9.40826058e-01 -9.44659889e-01 5.96723221e-02 -8.79798710e-01 -8.19761232e-02 -1.62526166e+00 -7.33601525e-02 -9.90235448e-01 -4.97899279e-02 2.44983122e-01 -5.85687272e-02 1.58743992e-01 1.13856755e-01 1.00764096e+00 -3.11049908e-01 8.23422134e-01 1.44001937e+00 1.25346988e-01 -3.59521925e-01 1.77396491e-01 -3.85604173e-01 8.36408794e-01 8.19018066e-01 -4.36768949e-01 -9.85159427e-02 -4.26413119e-01 -7.79527053e-02 -1.76995829e-01 1.14142217e-01 -9.02486861e-01 6.13141358e-01 -4.28592443e-01 1.26816243e-01 -1.01882970e+00 3.78234595e-01 -1.36666930e+00 7.98490405e-01 3.55618447e-01 5.43457448e-01 3.35818827e-01 2.24909615e-02 4.53492016e-01 -3.77610922e-01 -3.23235720e-01 6.83447182e-01 -5.19506633e-04 -3.03604931e-01 7.24437058e-01 4.56233025e-01 -3.80742639e-01 1.25957143e+00 -4.86949176e-01 -6.91680461e-02 7.55010871e-03 -4.53783661e-01 5.53342663e-02 8.61592233e-01 4.93208945e-01 8.59526932e-01 -1.84743822e+00 -6.45918846e-01 5.21737158e-01 1.21542588e-01 6.73591733e-01 8.20870772e-02 8.05031002e-01 -8.71360779e-01 2.17430010e-01 5.39065059e-03 -9.95350361e-01 -1.37709582e+00 2.86596686e-01 1.52638301e-01 3.37760448e-01 -5.90909183e-01 5.56885779e-01 -7.46107623e-02 -8.69482875e-01 -2.79739261e-01 -3.24975282e-01 -2.72461027e-01 -3.42485249e-01 1.77096814e-01 4.06003475e-01 3.06854278e-01 -1.24010217e+00 -5.92536628e-01 1.53651977e+00 2.20976576e-01 1.22619830e-01 1.35796142e+00 -1.06856361e-01 -6.64721310e-01 2.65154421e-01 1.21589971e+00 2.24460557e-01 -6.71272457e-01 -4.21549171e-01 -6.98307827e-02 -1.02828920e+00 -2.34941632e-01 9.74108353e-02 -9.94035721e-01 3.66280288e-01 5.17244935e-01 7.79451802e-02 7.85412729e-01 5.16330525e-02 8.97198558e-01 2.09613666e-01 8.17125976e-01 -8.66803944e-01 -3.22473556e-01 7.22623408e-01 1.26894474e+00 -1.01477623e+00 2.53490657e-01 -8.34894717e-01 -2.73398787e-01 1.06014037e+00 5.08379221e-01 -4.92633253e-01 9.98337269e-01 9.75800827e-02 -3.40513401e-02 -3.08574826e-01 2.29117468e-01 1.23748079e-01 5.58417141e-01 6.38359547e-01 -6.36151880e-02 -4.07826677e-02 -5.49471259e-01 6.14030421e-01 -2.96946585e-01 -2.48574018e-01 -4.01689596e-02 9.11376595e-01 -5.71038604e-01 -1.19062412e+00 -1.11919904e+00 1.83216214e-01 1.52425364e-01 2.14954585e-01 -2.01705292e-01 5.72664559e-01 -6.76726475e-02 5.94706655e-01 3.44208956e-01 -4.32964265e-01 8.08479428e-01 -2.73974359e-01 3.59860688e-01 -4.57478762e-01 -2.34936729e-01 1.43121362e-01 -7.25464523e-01 -5.23070276e-01 -5.33591390e-01 -7.65168488e-01 -1.47898126e+00 -5.60991704e-01 -8.21884811e-01 4.24109012e-01 1.03568995e+00 7.98302948e-01 5.79404473e-01 -4.47398126e-02 9.84894216e-01 -9.58891571e-01 -6.74290240e-01 -6.45240486e-01 -8.50765228e-01 5.74825287e-01 -7.37005845e-02 -9.33150291e-01 -3.29120547e-01 -2.67280966e-01]

[7.712545394897461, -2.9065239429473877]

3ecc5987-9430-46fc-89b9-ee2cd0e7e100

unsupervised-detection-of-ash-dieback-disease

2204.09041

https://arxiv.org/abs/2204.09041v1

https://arxiv.org/pdf/2204.09041v1.pdf

Unsupervised detection of ash dieback disease (Hymenoscyphus fraxineus) using diffusion-based hyperspectral image clustering

Ash dieback (Hymenoscyphus fraxineus) is an introduced fungal disease that is causing the widespread death of ash trees across Europe. Remote sensing hyperspectral images encode rich structure that has been exploited for the detection of dieback disease in ash trees using supervised machine learning techniques. However, to understand the state of forest health at landscape-scale, accurate unsupervised approaches are needed. This article investigates the use of the unsupervised Diffusion and VCA-Assisted Image Segmentation (D-VIS) clustering algorithm for the detection of ash dieback disease in a forest site near Cambridge, United Kingdom. The unsupervised clustering presented in this work has high overlap with the supervised classification of previous work on this scene (overall accuracy = 71%). Thus, unsupervised learning may be used for the remote detection of ash dieback disease without the need for expert labeling.

['James M. Murphy', 'David A. Coomes', 'Robert J. Plemmons', 'Kangning Cui', 'Aland H. Y. Chan', 'Sam L. Polk']

2022-04-19

['image-clustering']

['computer-vision']

[ 8.42728972e-01 -2.56077439e-01 1.14840958e-02 -2.22576894e-02 -3.72887582e-01 -5.99867284e-01 5.68286538e-01 8.68096426e-02 -1.35841042e-01 7.26670086e-01 5.40897064e-03 -5.91680229e-01 -5.67855239e-01 -9.32985604e-01 3.95487607e-01 -1.00277257e+00 -2.06771195e-01 8.36239934e-01 1.36362076e-01 7.54028708e-02 2.77098715e-01 1.11726618e+00 -1.28501034e+00 8.72694924e-02 6.45724654e-01 4.44953620e-01 8.35279346e-01 8.43158007e-01 -5.34671731e-02 5.85394859e-01 -6.58888161e-01 4.51268464e-01 2.71611124e-01 -3.32463086e-01 -8.48158836e-01 6.30635679e-01 4.18321997e-01 -7.12178171e-01 2.58690149e-01 8.30997825e-01 2.97032222e-02 -2.18991831e-01 1.05866957e+00 -5.05564988e-01 2.28237346e-01 6.08875573e-01 -9.33264434e-01 2.63720632e-01 -2.95876861e-01 2.29180127e-01 9.61175442e-01 -4.71095622e-01 7.45427728e-01 9.23090696e-01 3.89957458e-01 8.34469944e-02 -1.44234622e+00 -7.57096291e-01 -2.61134803e-01 3.04024488e-01 -1.41714931e+00 -2.73770988e-01 3.68815988e-01 -7.77565479e-01 7.12297857e-01 3.09320033e-01 1.01488960e+00 2.90769309e-01 -4.40525748e-02 5.03482103e-01 1.37477934e+00 -2.63837606e-01 5.41555762e-01 -4.41357523e-01 2.87996419e-02 5.00532091e-01 3.39442044e-01 4.75645602e-01 -7.88714886e-02 -2.87675351e-01 6.31830215e-01 1.93649650e-01 -1.96363613e-01 -3.28696132e-01 -6.39974773e-01 9.31625605e-01 7.69008756e-01 3.27890486e-01 -5.74713767e-01 -2.58336484e-01 1.72012538e-01 -2.53875941e-01 4.98886645e-01 5.52091837e-01 -2.24829078e-01 -2.41132919e-02 -1.52077806e+00 -8.91080275e-02 4.36678112e-01 -1.97654068e-01 5.92086554e-01 3.07882667e-01 4.41327661e-01 1.06766856e+00 3.64736438e-01 7.25820124e-01 -6.68062940e-02 -1.07208395e+00 -3.78904760e-01 7.89898992e-01 -7.29125664e-02 -5.99836767e-01 -1.79114014e-01 -4.00332630e-01 -6.02458298e-01 9.29590464e-01 2.31283769e-01 3.57892620e-03 -1.18612993e+00 9.28483903e-01 1.78387105e-01 -3.85486752e-01 -1.70482770e-01 8.21895540e-01 -8.37353095e-02 7.59912908e-01 3.47284824e-01 -3.95552576e-01 8.45145524e-01 -3.99481088e-01 -4.40664262e-01 -2.14932904e-01 2.66507447e-01 -3.84266227e-01 4.77950513e-01 6.19929254e-01 -1.86919287e-01 7.69085959e-02 -1.10529256e+00 9.69841778e-01 -7.62394190e-01 1.81160465e-01 6.90355301e-01 7.97793806e-01 -5.38623154e-01 5.12522101e-01 -8.64301205e-01 -8.25185776e-01 7.73024440e-01 2.95082718e-01 -6.96937814e-02 -2.86928505e-01 -3.01187187e-01 8.09618831e-01 3.94575804e-01 3.19972992e-01 -1.10278380e+00 -3.89691919e-01 4.84144725e-02 -1.59377441e-01 5.57350934e-01 1.57470945e-02 1.00826252e+00 -9.18907046e-01 -1.10770094e+00 9.64220107e-01 2.41778865e-01 -4.44036394e-01 3.39838415e-01 -3.19376647e-01 -3.14273030e-01 4.37297225e-01 3.08836371e-01 4.02444929e-01 7.56187677e-01 -1.44611859e+00 -9.09859359e-01 -1.13568676e+00 -5.97378314e-01 5.99840544e-02 5.59065714e-02 -1.22145824e-01 8.03056121e-01 -7.70438850e-01 3.06887627e-01 -8.19227457e-01 -7.56872296e-02 8.96031782e-02 -2.15379357e-01 2.62177527e-01 1.49143147e+00 -8.31603110e-01 6.30940974e-01 -2.12828612e+00 -1.71376735e-01 4.26322818e-01 3.03999364e-01 4.56239462e-01 3.14906448e-01 5.01014292e-01 -5.56669496e-02 2.56409734e-01 -1.19837880e+00 7.32342303e-01 -4.06920075e-01 6.96715593e-01 -2.40136608e-01 4.04503793e-01 -8.65219012e-02 1.28885612e-01 -9.37319338e-01 -2.20010564e-01 5.52053094e-01 2.91911960e-01 2.61964411e-01 -1.03083618e-01 -1.07743874e-01 -1.96825191e-01 -1.32283941e-01 1.09087551e+00 8.05587769e-01 2.55349427e-01 2.98622847e-01 2.46668383e-01 -6.68712735e-01 -8.05040598e-02 -9.05332983e-01 8.44796658e-01 -1.29198087e-02 7.30115414e-01 8.70094061e-01 -8.04940104e-01 8.17991376e-01 1.24725983e-01 6.37705207e-01 1.40362024e-01 -2.96311617e-01 3.88628870e-01 1.66506395e-01 -1.70646667e-01 -8.83566365e-02 -4.21322346e-01 8.64538193e-01 4.15899098e-01 -3.12399596e-01 -3.75575155e-01 1.10896513e-01 2.05355510e-01 1.15951884e+00 -8.34362134e-02 2.54656941e-01 -6.19677126e-01 4.69557829e-02 9.07035887e-01 8.72702077e-02 4.07189250e-01 -3.66797328e-01 1.34677157e-01 -5.63295670e-02 -1.24135368e-01 -6.82741702e-01 -1.24036849e+00 -5.67632556e-01 9.23953891e-01 -3.14698637e-01 -1.42212063e-02 -5.03603458e-01 -4.34812874e-01 1.28195435e-02 9.42209065e-01 -3.53369772e-01 1.89095050e-01 2.76426613e-01 -1.09375489e+00 5.06760955e-01 3.92683268e-01 9.95703876e-01 -1.15971184e+00 -8.08608413e-01 3.74111861e-01 2.47502744e-01 -7.07972825e-01 5.02032876e-01 7.19603658e-01 -1.14172375e+00 -1.33519638e+00 -7.56730199e-01 -3.84554207e-01 3.14515233e-01 4.02954906e-01 5.61825275e-01 -1.59088537e-01 -1.07567692e+00 5.31685889e-01 -5.49467146e-01 -4.24378932e-01 -4.02673125e-01 1.69633999e-01 -4.97307718e-01 -1.98698014e-01 7.70519614e-01 -6.31484628e-01 -4.41894710e-01 1.91169545e-01 -8.20134521e-01 -1.85198188e-01 8.85858774e-01 4.86442178e-01 6.88811183e-01 7.42097914e-01 2.64281034e-01 -9.94090438e-01 4.34462279e-01 -1.28957093e-01 -5.51621079e-01 5.33158556e-02 -7.18446851e-01 -5.85909903e-01 2.48084277e-01 1.40926629e-01 -9.08274233e-01 4.84679163e-01 4.17963684e-01 2.03752041e-01 -1.13672018e+00 5.51479518e-01 -4.81023751e-02 3.44651602e-02 7.43490160e-01 -3.44148576e-02 8.09915140e-02 -4.39767569e-01 1.94411613e-02 1.11786926e+00 8.87910962e-01 1.34074122e-01 7.48581171e-01 7.99876332e-01 7.74303824e-02 -1.84756660e+00 -5.67793727e-01 -8.55605304e-01 -7.86995888e-01 -5.62692702e-01 1.11611938e+00 -5.49342573e-01 -3.27680968e-02 5.78305066e-01 -3.15182626e-01 -5.30334473e-01 -3.74000639e-01 5.26054442e-01 -3.12862635e-01 1.71428680e-01 -1.06405899e-01 -1.25771308e+00 -4.78003442e-01 -4.07394141e-01 8.37762475e-01 -2.89923877e-01 -2.74700284e-01 -5.52076399e-01 4.07916874e-01 7.97437608e-01 2.46855676e-01 3.99688631e-01 9.53328848e-01 -3.25585544e-01 -3.19799781e-01 -1.04567602e-01 -4.76892859e-01 5.08438528e-01 6.29849851e-01 5.49314797e-01 -9.69923139e-01 3.12911719e-02 1.17613025e-01 5.89584112e-02 1.10919356e+00 5.87558746e-01 8.05838346e-01 1.25253931e-01 -3.83873910e-01 2.39572734e-01 1.59139252e+00 5.93485594e-01 3.52991700e-01 6.41590774e-01 4.88854378e-01 7.70752311e-01 7.45021641e-01 5.19508481e-01 -6.96059704e-01 2.99166031e-02 8.11112761e-01 -2.39900395e-01 -7.87046254e-02 3.39813262e-01 3.39038372e-01 1.91075373e-02 -4.99903113e-01 -8.44584182e-02 -1.43750775e+00 5.63768744e-01 -1.33534586e+00 -1.24595511e+00 -3.44607830e-01 1.94988012e+00 1.70969158e-01 -1.70780182e-01 2.08234534e-01 6.39741600e-01 9.24394906e-01 3.42932463e-01 -5.83288491e-01 -2.75403589e-01 -1.42259583e-01 6.62463129e-01 9.17054355e-01 5.18332660e-01 -1.15345490e+00 9.50571656e-01 6.28742838e+00 5.46734810e-01 -9.17477846e-01 -4.61551547e-01 1.70288861e-01 2.13227831e-02 3.95669043e-01 1.85099185e-01 -3.23393762e-01 -4.85905558e-02 4.55719620e-01 2.98724622e-01 6.92739725e-01 5.39838672e-01 6.12417400e-01 -6.84340715e-01 -8.62873197e-02 5.97693443e-01 -3.23067725e-01 -8.35538208e-01 -5.43106645e-02 5.32254994e-01 3.55126053e-01 3.05364639e-01 -3.95134270e-01 -3.82456690e-01 7.98452735e-01 -8.66647363e-01 -1.26043722e-01 1.31276935e-01 7.13123441e-01 -8.03230703e-01 5.45891047e-01 2.92625815e-01 -1.23987031e+00 -2.35951155e-01 -1.06397785e-01 -2.59857923e-01 7.46730389e-03 8.23979378e-01 -1.15593088e+00 1.34261757e-01 7.75364518e-01 8.67690504e-01 -8.17894995e-01 9.38558519e-01 -2.23879382e-01 1.07386899e+00 -7.34136581e-01 2.56352872e-01 4.95371997e-01 -6.98282242e-01 6.42448485e-01 8.13467383e-01 3.05585861e-01 4.68574204e-02 2.23950908e-01 8.12271893e-01 5.08778989e-01 8.44474211e-02 -5.38661718e-01 -7.19737470e-01 1.55007601e-01 9.92392540e-01 -1.07809269e+00 -1.77789152e-01 2.07475349e-01 9.20721531e-01 -4.66024309e-01 1.49286509e-01 -7.17016011e-02 -3.15262526e-01 3.46107811e-01 7.03644812e-01 4.38520402e-01 -4.15525407e-01 -1.90263167e-01 -3.49890769e-01 -5.76201677e-01 -5.19010782e-01 6.49450302e-01 -1.04370987e+00 -1.18480527e+00 -1.56615488e-02 1.46894623e-02 -5.76927781e-01 -1.40928775e-01 -7.40170717e-01 -5.55550337e-01 4.57110941e-01 -1.00034380e+00 -1.06228256e+00 -7.03475893e-01 4.08390105e-01 6.21899903e-01 -3.21451962e-01 1.15407252e+00 -4.83264595e-01 -5.58546841e-01 -7.94722736e-01 6.31962419e-01 -4.73244116e-02 3.06418657e-01 -1.62998354e+00 -4.64742661e-01 5.63391984e-01 -1.27693051e-02 -3.36978316e-01 3.48481297e-01 -1.16969132e+00 -6.64055943e-01 -1.29532754e+00 2.96561867e-01 4.36241329e-01 7.71771550e-01 3.48005801e-01 -5.22322297e-01 2.60239869e-01 2.95762241e-01 -8.75911832e-01 9.37142968e-01 -2.62873232e-01 -1.96724348e-02 -1.77868396e-01 -1.45262206e+00 3.86221915e-01 7.38716364e-01 -4.82862920e-01 -4.03852165e-01 6.29068613e-01 -6.67479873e-01 7.73094416e-01 -7.21870542e-01 4.30062354e-01 6.03119016e-01 -8.88382554e-01 7.74114728e-01 1.01547793e-01 4.12807137e-01 -3.35471570e-01 -5.09994864e-01 -1.10048234e+00 -4.60678548e-01 -1.23197719e-01 3.83559257e-01 1.13570404e+00 3.23408060e-02 -4.66491967e-01 8.63144636e-01 -3.00328732e-01 3.19991410e-01 1.58272952e-01 -5.67618370e-01 -8.40085268e-01 -1.66737475e-02 -1.58216222e-03 -5.83124869e-02 8.13971639e-01 -5.94840407e-01 4.21846658e-01 4.84669149e-01 4.11808074e-01 1.10926163e+00 1.90581068e-01 4.24723536e-01 -1.97370756e+00 2.88920067e-02 -6.68184340e-01 -5.84247470e-01 -2.78852165e-01 7.25824460e-02 -7.49025583e-01 9.06092022e-03 -1.73243809e+00 8.63280073e-02 -6.88448474e-02 3.49079669e-01 4.90988135e-01 8.18214715e-02 6.51656389e-02 -2.54874304e-02 6.19463205e-01 6.35365605e-01 1.71822160e-01 6.46449745e-01 -5.67488909e-01 -3.03921372e-01 1.46977991e-01 -2.36897811e-01 8.21009755e-01 1.19441414e+00 -4.54629540e-01 -3.05445105e-01 1.75908059e-02 -5.70430577e-01 -3.03338945e-01 6.78667486e-01 -1.06876206e+00 -3.37775111e-01 -4.63548869e-01 3.39304835e-01 -9.58333015e-01 2.60610133e-01 -1.53449619e+00 3.54516715e-01 9.77570295e-01 -8.53881910e-02 -6.38890922e-01 -2.05386989e-02 5.37618756e-01 1.88726224e-02 -5.14186442e-01 1.13711703e+00 -5.14723480e-01 -8.73329103e-01 1.51867002e-01 -1.29606962e+00 -4.10608381e-01 9.28835273e-01 -4.99256641e-01 -2.36430794e-01 -2.59003013e-01 -9.62567985e-01 -1.81946233e-01 5.09037077e-01 -3.30782622e-01 3.51112664e-01 -6.02111042e-01 -7.82030225e-01 4.88824584e-02 1.01596698e-01 -1.09093890e-01 -3.49822819e-01 5.42277277e-01 -1.48105633e+00 2.09787980e-01 -6.36649132e-01 -7.47678041e-01 -1.49569082e+00 1.59528151e-01 4.79321063e-01 6.33299798e-02 -5.51397383e-01 3.82249236e-01 -2.02509537e-01 -1.55174453e-02 2.67916489e-02 3.78035098e-01 -1.86222807e-01 3.46969306e-01 2.12037563e-01 8.28201711e-01 -6.03207499e-02 -6.30376995e-01 -3.12316865e-01 2.57343501e-01 6.62106043e-03 -3.22496235e-01 1.63219368e+00 8.56113732e-02 -2.48129562e-01 5.55184782e-01 5.26306152e-01 -3.14064026e-01 -8.70274246e-01 4.76683855e-01 2.06571266e-01 -5.80826938e-01 9.91717100e-01 -1.17186463e+00 -1.14916098e+00 9.15763736e-01 1.19355464e+00 5.06806135e-01 1.22830915e+00 -1.65311992e-01 1.66009501e-01 5.51402390e-01 -2.59950496e-02 -1.28724205e+00 -6.05667055e-01 -2.21174940e-01 8.22172642e-01 -1.03827453e+00 4.29145932e-01 -5.77307582e-01 -2.48193204e-01 1.19919848e+00 2.87123799e-01 -6.35877103e-02 7.10633278e-01 3.68285745e-01 2.69940227e-01 -5.60502172e-01 -2.41770044e-01 -7.22364366e-01 -5.67748070e-01 1.17315757e+00 1.74284503e-01 6.46010816e-01 -2.85635918e-01 -3.64252687e-01 -1.21662781e-01 -1.90789327e-02 2.79490918e-01 1.41773307e+00 -1.05120742e+00 -6.35119021e-01 -6.85479999e-01 1.02526855e+00 -6.08277395e-02 9.35326740e-02 -1.78382528e+00 9.34629321e-01 2.19869483e-02 9.63734865e-01 7.76831433e-02 -1.18814304e-01 -7.87424296e-02 2.54709929e-01 2.85392314e-01 -6.65012419e-01 -5.04771650e-01 6.44925773e-01 -1.11754343e-01 1.31691113e-01 -1.01278174e+00 -7.02237546e-01 -8.65327716e-01 -3.10807526e-01 -6.27465963e-01 -5.98174846e-03 8.92307162e-01 7.92918146e-01 -4.61663008e-01 3.24216783e-01 5.31920433e-01 -4.93969768e-01 -1.10307723e-01 -1.02035272e+00 -1.60049701e+00 -1.41134143e-01 -7.25653246e-02 -4.27695394e-01 -7.61269629e-01 1.18722141e-01]

[9.495720863342285, -1.527036428451538]

36b76d2a-ff85-4138-9434-cb5e1b27b163

shape-from-tracing-towards-reconstructing-3d

2012.03939

https://arxiv.org/abs/2012.03939v1

https://arxiv.org/pdf/2012.03939v1.pdf

Shape From Tracing: Towards Reconstructing 3D Object Geometry and SVBRDF Material from Images via Differentiable Path Tracing

Reconstructing object geometry and material from multiple views typically requires optimization. Differentiable path tracing is an appealing framework as it can reproduce complex appearance effects. However, it is difficult to use due to high computational cost. In this paper, we explore how to use differentiable ray tracing to refine an initial coarse mesh and per-mesh-facet material representation. In simulation, we find that it is possible to reconstruct fine geometric and material detail from low resolution input views, allowing high-quality reconstructions in a few hours despite the expense of path tracing. The reconstructions successfully disambiguate shading, shadow, and global illumination effects such as diffuse interreflection from material properties. We demonstrate the impact of different geometry initializations, including space carving, multi-view stereo, and 3D neural networks. Finally, with input captured using smartphone video and a consumer 360? camera for lighting estimation, we also show how to refine initial reconstructions of real-world objects in unconstrained environments.

['Daniel Ritchie', 'James Tompkin', 'Vikas Thamizharasan', 'James Guesman', 'Loudon Cohen', 'Purvi Goel']

2020-12-06

['lighting-estimation']

['computer-vision']

[ 5.03760636e-01 -3.91821265e-01 8.68634224e-01 -3.44341546e-01 -7.41742551e-01 -6.45374298e-01 3.10484737e-01 -2.89289504e-01 1.51981190e-01 6.85003936e-01 -2.26983190e-01 -1.46708041e-01 1.91518262e-01 -8.14595222e-01 -8.12708080e-01 -5.11604846e-01 4.09704685e-01 5.57044804e-01 2.98781872e-01 7.82213137e-02 2.21170008e-01 1.03090823e+00 -1.75795972e+00 2.33833745e-01 6.93556786e-01 7.98629224e-01 3.14890891e-01 8.80924284e-01 -9.56619605e-02 3.79153013e-01 -4.48472261e-01 -4.20015246e-01 4.72464710e-01 -2.17469648e-01 -3.06881756e-01 5.22638023e-01 8.47649992e-01 -8.59374821e-01 8.33507106e-02 8.27056587e-01 2.98433632e-01 1.66487113e-01 4.42114204e-01 -9.36916530e-01 -2.77100831e-01 -4.82083201e-01 -7.34991968e-01 -3.73708397e-01 6.78063571e-01 2.32499674e-01 5.75548887e-01 -1.11185658e+00 6.59221292e-01 1.23415887e+00 8.92108440e-01 2.73208171e-01 -1.40463066e+00 -3.24202895e-01 9.55731496e-02 -2.77177662e-01 -1.18951404e+00 -4.96992826e-01 9.64717686e-01 -4.25040573e-01 7.56321132e-01 5.31308651e-01 8.93294990e-01 7.77219057e-01 3.34443331e-01 2.43707612e-01 1.39673364e+00 -2.94975668e-01 2.23784167e-02 1.94552004e-01 -1.04944117e-01 8.67822468e-01 -7.74227083e-03 2.97727793e-01 -3.37894678e-01 -5.68043292e-01 1.31935942e+00 7.94584975e-02 -6.54026806e-01 -6.22231662e-01 -9.86005008e-01 3.40924829e-01 1.17023133e-01 -4.41964149e-01 -2.12009266e-01 2.38568068e-01 -3.13930094e-01 2.66355444e-02 6.50038004e-01 3.68815333e-01 -4.13653374e-01 1.03973560e-01 -6.02376461e-01 2.54880756e-01 8.83136809e-01 9.30476189e-01 9.69477892e-01 1.77382126e-01 4.66504693e-01 8.26694191e-01 4.02989715e-01 9.16427612e-01 -6.11322939e-01 -1.79900348e+00 2.89014671e-02 2.13732123e-01 5.50185978e-01 -8.12940300e-01 -2.61370838e-01 -2.05965117e-01 -4.88325387e-01 9.09180880e-01 4.86412376e-01 -9.94248688e-02 -8.51490557e-01 1.16815233e+00 7.41172969e-01 2.18388602e-01 -3.22849780e-01 1.06764746e+00 6.31526053e-01 6.11038983e-01 -5.93125880e-01 -3.02972138e-01 1.06193352e+00 -7.46794403e-01 -5.48347414e-01 -3.61416340e-02 -1.24440990e-01 -1.05405128e+00 1.07782900e+00 9.02346313e-01 -1.66207159e+00 -2.54636556e-01 -7.81470597e-01 -2.22360775e-01 5.73809110e-02 -4.09406833e-02 6.95086002e-01 5.31926930e-01 -1.11122179e+00 7.66296208e-01 -9.03845012e-01 -1.53624639e-01 1.54839933e-01 2.24055231e-01 -8.99256468e-02 -3.80642533e-01 -5.08906186e-01 6.74346924e-01 -4.83708411e-01 -6.94373250e-02 -8.71865451e-01 -1.08840764e+00 -5.76155245e-01 -8.47981796e-02 3.53048682e-01 -1.09743822e+00 1.09185970e+00 -8.56374264e-01 -1.83695269e+00 7.43113697e-01 -2.76677668e-01 3.34315777e-01 4.72281218e-01 -2.53639460e-01 -5.38239442e-02 2.88995385e-01 -2.82509089e-01 1.22526646e-01 8.64078760e-01 -1.99303567e+00 -2.28937194e-01 -4.70511585e-01 2.06296638e-01 6.24393761e-01 2.23218501e-01 -1.32264018e-01 -6.25170708e-01 -2.90486127e-01 4.00969833e-01 -7.18039513e-01 -2.77375817e-01 7.10736096e-01 -3.55139822e-01 6.86655819e-01 8.67453337e-01 -8.13241303e-01 4.00581598e-01 -1.93111455e+00 2.98654158e-02 3.10527384e-01 9.27426592e-02 -3.59726280e-01 2.52909157e-02 2.30338573e-01 1.91335037e-01 -2.44283602e-01 -2.87281901e-01 -6.95226729e-01 -3.42615366e-01 2.28732184e-01 -1.41535163e-01 5.92316628e-01 -3.10005695e-01 4.35176998e-01 -6.19799435e-01 -6.04440160e-02 5.92122316e-01 1.00343239e+00 -7.19278753e-01 3.18059832e-01 -4.05177206e-01 8.44731271e-01 -1.05972290e-01 8.24153125e-01 9.49250877e-01 -4.44573730e-01 1.21920787e-01 -4.56678361e-01 -3.68456513e-01 -8.73982906e-04 -1.50682032e+00 1.58113301e+00 -8.27028096e-01 5.06673872e-01 9.74659443e-01 -1.90539122e-01 6.11454606e-01 1.07324928e-01 6.30888999e-01 -5.24947941e-01 -1.65604681e-04 1.11118235e-01 -5.84269881e-01 -2.16620192e-01 5.59389174e-01 -4.13434476e-01 6.20047808e-01 5.47719240e-01 -4.64396477e-01 -8.44074726e-01 -4.74188060e-01 -3.84211279e-02 8.42563450e-01 4.50784564e-01 -4.04946180e-03 -1.02489181e-02 2.03184769e-01 -1.32253692e-01 2.18451008e-01 3.72679502e-01 5.81020057e-01 1.19017208e+00 -1.10018313e-01 -3.89044285e-01 -1.12093306e+00 -1.45188713e+00 -3.08024317e-01 7.84492910e-01 5.28034627e-01 -1.91210397e-02 -6.11046374e-01 8.69515166e-02 1.23878799e-01 8.40686798e-01 -1.02927066e-01 5.06833196e-01 -8.41183364e-01 -5.51959574e-01 -1.97324917e-01 1.98586106e-01 2.19544694e-01 -6.86177552e-01 -8.58201981e-01 1.44872829e-01 -8.60099196e-02 -1.21246374e+00 -3.57112557e-01 -1.66726097e-01 -1.04346704e+00 -1.27518117e+00 -7.19119072e-01 -1.63520813e-01 8.11564684e-01 7.18133032e-01 1.48172343e+00 3.25946540e-01 -5.11227787e-01 1.04572189e+00 1.09040111e-01 -6.66494742e-02 -3.46349090e-01 -6.82985485e-01 -3.01038772e-01 -4.28293943e-02 -5.44813216e-01 -8.72164071e-01 -6.85622871e-01 5.96791923e-01 -6.83013737e-01 6.02315545e-01 -1.29800215e-02 4.72268820e-01 8.70893419e-01 -4.66964133e-02 -3.53039593e-01 -9.21778738e-01 1.09792851e-01 -1.79919437e-01 -1.07684278e+00 1.35298401e-01 -4.80322391e-01 -4.29567903e-01 5.70411623e-01 -2.20148742e-01 -1.58020556e+00 1.05374632e-02 -1.04069375e-01 -7.10260570e-01 -3.35943013e-01 -1.79463461e-01 -1.25371471e-01 -4.72270399e-01 4.18338686e-01 4.08829562e-02 -1.09524898e-01 -6.52326047e-01 2.76418358e-01 1.24225341e-01 4.27623063e-01 -7.84906149e-01 7.36158669e-01 1.16553020e+00 9.61058438e-02 -1.26641226e+00 -6.87529564e-01 -1.56765684e-01 -3.86290640e-01 -5.10867000e-01 7.17947423e-01 -6.88920796e-01 -9.38975155e-01 4.49165106e-01 -1.37507021e+00 -5.98592103e-01 -2.07751587e-01 3.41997385e-01 -6.19374990e-01 4.26868945e-01 -5.88874936e-01 -8.33425224e-01 -2.85762642e-02 -1.17243576e+00 1.33635437e+00 1.49301529e-01 3.13484296e-02 -9.88274038e-01 -1.42048165e-01 5.42596877e-01 2.94817179e-01 4.34947729e-01 8.78831685e-01 7.22551823e-01 -1.35496318e+00 4.79088426e-01 -2.43333578e-01 9.23729762e-02 3.29980224e-01 2.99602002e-01 -1.19844139e+00 -2.81331748e-01 2.65962392e-01 -8.30332860e-02 3.73429507e-01 6.29646778e-01 1.31045294e+00 -6.10869937e-02 -2.32494310e-01 1.33225489e+00 1.83727360e+00 2.43716225e-01 5.27811706e-01 -1.53337223e-02 1.10138762e+00 6.88171089e-01 4.46912766e-01 5.61989963e-01 4.00416762e-01 7.91500866e-01 6.66396439e-01 -3.10548455e-01 -4.06855732e-01 1.64427981e-01 2.56863628e-02 5.49876213e-01 -4.21074152e-01 -3.41898412e-01 -7.19875813e-01 5.94591536e-02 -1.13869298e+00 -8.23619664e-01 -6.71955824e-01 2.38513350e+00 5.59054852e-01 -2.45113611e-01 -1.90969422e-01 -2.16171399e-01 3.38179857e-01 -1.07575536e-01 -6.65695310e-01 -1.83776036e-01 -8.38850439e-02 1.27373144e-01 5.52001297e-01 1.22001731e+00 -4.83489513e-01 5.85156858e-01 6.58684540e+00 1.81148589e-01 -1.09746110e+00 2.03254059e-01 6.14171922e-01 -2.90959895e-01 -1.05802941e+00 -6.88913837e-02 -5.89186192e-01 2.04964802e-01 3.47552687e-01 3.90891075e-01 9.64438736e-01 4.83786315e-01 2.49298945e-01 -4.93526965e-01 -9.82338727e-01 1.17004359e+00 2.82493144e-01 -1.19330382e+00 -2.11895242e-01 1.75874550e-02 9.21627045e-01 8.35710838e-02 -4.53695655e-02 -5.47262967e-01 3.79605889e-01 -9.09587204e-01 7.02494383e-01 6.79388344e-01 8.92834425e-01 -4.86505181e-01 2.54583154e-02 4.04013038e-01 -1.00780725e+00 3.93628865e-01 -1.28466114e-01 1.37342796e-01 6.57667458e-01 6.73201323e-01 -5.27713299e-01 4.12346750e-01 6.66798770e-01 3.69890690e-01 -1.69457793e-01 9.65158165e-01 4.28779833e-02 5.45084178e-02 -6.28390372e-01 4.04155582e-01 -3.00427824e-01 -7.93257713e-01 6.92560136e-01 8.14026654e-01 5.35206974e-01 5.60784757e-01 2.15003267e-01 1.20151830e+00 2.37416252e-01 -2.77312517e-01 -6.47935271e-01 6.51620448e-01 2.15773042e-02 1.26590192e+00 -7.89958954e-01 -7.16027841e-02 -5.54096401e-01 1.22060144e+00 1.90817043e-01 9.14257824e-01 -6.91367805e-01 2.65796602e-01 7.25322306e-01 6.69028580e-01 1.52711123e-02 -4.63122368e-01 -6.29473388e-01 -1.23564446e+00 -4.08990681e-03 -5.95915675e-01 -2.72864074e-01 -1.39929008e+00 -1.12583697e+00 2.90453643e-01 -1.99430864e-02 -1.05044591e+00 9.56158042e-02 -6.37154639e-01 -3.49026203e-01 1.05380738e+00 -1.57190990e+00 -1.00586700e+00 -5.41439474e-01 5.54076433e-01 5.40643275e-01 5.22441030e-01 8.01827192e-01 2.61353523e-01 3.63496840e-02 -1.08587898e-01 2.66933173e-01 -5.45932889e-01 4.63816643e-01 -1.15020359e+00 2.50361115e-01 5.13483226e-01 -1.08695574e-01 5.73956251e-01 7.51328707e-01 -6.94820106e-01 -1.74412823e+00 -6.40493929e-01 1.43003436e-02 -5.32762587e-01 5.68793081e-02 -3.09959561e-01 -1.10425079e+00 5.87126434e-01 2.17544511e-01 1.94175079e-01 3.72458190e-01 -2.48091176e-01 -6.19137399e-02 -6.12063706e-03 -1.29210603e+00 6.40538394e-01 1.16950226e+00 -4.38488722e-01 7.45489001e-02 2.56262988e-01 3.30157489e-01 -1.00021160e+00 -7.75742352e-01 2.60297418e-01 8.78911316e-01 -1.60333312e+00 1.38950348e+00 1.90583870e-01 2.31344223e-01 -3.93871218e-01 -3.21928591e-01 -1.42786884e+00 -1.37003720e-01 -7.36950159e-01 -1.14761032e-01 9.10932243e-01 1.23955943e-01 -6.63150072e-01 8.11604798e-01 9.37060833e-01 -3.10027629e-01 -6.91180766e-01 -5.48572302e-01 -4.15244132e-01 -3.55562061e-01 -4.88621265e-01 4.76345062e-01 8.35937798e-01 -9.78986263e-01 4.39834036e-02 -2.77723342e-01 5.89482248e-01 1.24067664e+00 6.94047689e-01 8.79541159e-01 -1.35212755e+00 -6.52384758e-01 1.45579688e-02 4.60886151e-01 -1.28746367e+00 -5.24031185e-02 -4.48173970e-01 1.04940377e-01 -1.57850730e+00 9.81768072e-02 -8.11940968e-01 4.15154815e-01 -1.87343299e-01 -3.57767902e-02 2.93640584e-01 1.08553715e-01 -2.06826851e-02 -5.68796359e-02 4.14134860e-01 1.68336689e+00 2.91367382e-01 -2.01532975e-01 1.03788078e-01 -2.76065499e-01 1.18548012e+00 5.16608298e-01 -1.24638908e-01 -4.89901811e-01 -9.98578846e-01 3.75400662e-01 4.94801611e-01 7.48399913e-01 -6.14558876e-01 -1.50047898e-01 -3.75911236e-01 6.47245407e-01 -8.38375807e-01 1.16281664e+00 -1.18663347e+00 8.45964372e-01 -8.17611739e-02 1.30998746e-01 -4.76265103e-02 2.54774809e-01 4.12870139e-01 3.94219249e-01 -8.87093171e-02 1.08328331e+00 -5.62233984e-01 -1.83530867e-01 2.16640949e-01 -8.19204748e-02 -6.95041567e-02 7.20684230e-01 -6.19228065e-01 -1.02799200e-01 -6.05854511e-01 -5.90442359e-01 -3.36556882e-02 1.28611171e+00 -1.95266068e-01 9.51416790e-01 -1.17142141e+00 -3.64530563e-01 4.97544140e-01 -3.90863657e-01 1.51339039e-01 2.89280206e-01 5.07542431e-01 -1.04299307e+00 -1.23143375e-01 1.71201840e-01 -8.55544031e-01 -1.54651356e+00 5.74423969e-02 5.60971677e-01 1.93922177e-01 -9.81468976e-01 7.60096550e-01 6.61027014e-01 -5.82644820e-01 1.89316031e-02 -3.88052374e-01 4.20890212e-01 -4.73874211e-01 3.69107246e-01 6.76938355e-01 8.99559166e-03 -3.61073852e-01 -4.43940721e-02 1.21307826e+00 5.45605421e-01 -3.33108842e-01 1.34695005e+00 -4.70205694e-01 -8.71396884e-02 5.38993299e-01 1.04317558e+00 2.94303179e-01 -1.71377361e+00 1.16631962e-01 -8.04277718e-01 -9.12856638e-01 1.41576961e-01 -6.73207939e-01 -1.10912478e+00 9.82520342e-01 3.22153062e-01 4.11057658e-02 1.05013001e+00 -1.28841683e-01 8.82316291e-01 1.94728866e-01 7.13271320e-01 -7.88768768e-01 -5.38664944e-02 4.24866408e-01 9.02692139e-01 -9.74437952e-01 4.26918238e-01 -1.01007736e+00 -1.50701776e-01 1.26182258e+00 6.97684467e-01 -6.47008717e-02 7.26207674e-01 6.53017819e-01 7.58208409e-02 -3.56895596e-01 -4.91906255e-01 2.45044991e-01 1.38435960e-01 5.35735786e-01 2.01640174e-01 -1.21467054e-01 4.83003139e-01 -3.05284709e-01 -9.16691646e-02 -3.38044554e-01 6.40156567e-01 5.87037623e-01 -2.60360032e-01 -8.12036037e-01 -8.33496034e-01 4.17872280e-01 -2.77495891e-01 -4.25274484e-02 -4.15977873e-02 5.56138515e-01 7.49758631e-03 5.98788679e-01 2.15415642e-01 2.87126422e-01 3.63482714e-01 -3.18371058e-01 9.96004820e-01 -7.63743877e-01 -4.85095352e-01 4.70975250e-01 2.25698829e-01 -8.57268572e-01 -4.17610645e-01 -7.18802929e-01 -1.18486273e+00 -2.51075178e-01 -3.93454790e-01 -3.75851482e-01 7.23367035e-01 5.40288687e-01 1.21158473e-01 4.26294088e-01 6.58322632e-01 -1.32052124e+00 -3.10119838e-02 -3.17885399e-01 -7.33762741e-01 3.39401752e-01 6.05282545e-01 -6.78576708e-01 -6.96122587e-01 2.63373882e-01]

[9.659646034240723, -3.0799920558929443]

aaeb9db3-f0d9-4cd2-adcc-fba7c54eb572

abstract-visual-reasoning-with-tangram-shapes

2211.16492

https://arxiv.org/abs/2211.16492v1

https://arxiv.org/pdf/2211.16492v1.pdf

Abstract Visual Reasoning with Tangram Shapes

We introduce KiloGram, a resource for studying abstract visual reasoning in humans and machines. Drawing on the history of tangram puzzles as stimuli in cognitive science, we build a richly annotated dataset that, with >1k distinct stimuli, is orders of magnitude larger and more diverse than prior resources. It is both visually and linguistically richer, moving beyond whole shape descriptions to include segmentation maps and part labels. We use this resource to evaluate the abstract visual reasoning capacities of recent multi-modal models. We observe that pre-trained weights demonstrate limited abstract reasoning, which dramatically improves with fine-tuning. We also observe that explicitly describing parts aids abstract reasoning for both humans and models, especially when jointly encoding the linguistic and visual inputs. KiloGram is available at https://lil.nlp.cornell.edu/kilogram .

['Yoav Artzi', 'Robert D. Hawkins', 'Wai Keen Vong', 'Alane Suhr', 'Noah Rush', 'Noriyuki Kojima', 'Anya Ji']

2022-11-29

['visual-reasoning', 'visual-reasoning']

['computer-vision', 'reasoning']

[ 2.07255557e-02 3.56783241e-01 -8.97790939e-02 -5.43520190e-02 -6.19654238e-01 -1.14835453e+00 6.88333571e-01 4.18844432e-01 -3.04384559e-01 4.76793498e-01 6.12592936e-01 -4.06050503e-01 -4.41471040e-02 -3.64611238e-01 -6.66234791e-01 -2.99373358e-01 6.28645569e-02 7.65163541e-01 -3.09730396e-02 -1.13202959e-01 6.91158652e-01 3.53684157e-01 -1.14789093e+00 1.01813030e+00 7.25837588e-01 8.20290625e-01 1.36726066e-01 8.57205391e-01 -2.32257679e-01 9.46019769e-01 -5.14445007e-01 -8.30462217e-01 6.64482564e-02 -1.70746148e-01 -1.30011594e+00 9.39606950e-02 7.41980076e-01 -1.91827029e-01 -4.51321334e-01 6.97643399e-01 1.98910072e-01 8.89694542e-02 7.58076906e-01 -1.21134806e+00 -1.49227691e+00 7.46576071e-01 -6.46287560e-01 4.38491911e-01 6.99167132e-01 7.81050265e-01 1.52175474e+00 -9.11339641e-01 7.07465231e-01 1.49792135e+00 4.76478636e-01 5.60568511e-01 -1.52816391e+00 -2.46427357e-01 4.29570556e-01 2.31399849e-01 -8.90537918e-01 -3.68489444e-01 6.41049623e-01 -5.97821653e-01 1.38716686e+00 3.05148631e-01 9.63900208e-01 1.21591890e+00 -1.54516488e-01 1.12576592e+00 1.16113567e+00 -2.79523432e-01 6.24878258e-02 -4.18316662e-01 2.66678840e-01 1.09805393e+00 4.69541937e-01 -1.88627422e-01 -7.61421859e-01 -2.06729770e-01 1.24508274e+00 -2.01523229e-01 -1.77125826e-01 -4.18304503e-01 -1.50776732e+00 4.96094972e-01 5.82264960e-01 1.04750261e-01 -2.05066293e-01 5.14326215e-01 4.54515338e-01 2.39892453e-02 7.68850558e-03 8.19603324e-01 -2.85496980e-01 -1.11804314e-01 -4.42938596e-01 2.90474385e-01 6.55065536e-01 1.04066730e+00 3.86291862e-01 3.44529338e-02 -3.41862828e-01 9.13400829e-01 2.61341244e-01 3.90175134e-01 2.63576150e-01 -1.61778462e+00 7.80413568e-01 8.37391675e-01 2.44601190e-01 -9.61378574e-01 -5.52435994e-01 -2.74098843e-01 -5.41377008e-01 5.22692680e-01 9.10889208e-01 4.70035970e-01 -9.43268001e-01 1.85498762e+00 -6.15832396e-02 -4.50331658e-01 -2.55298108e-01 9.92057920e-01 1.09128129e+00 3.16722721e-01 6.12454653e-01 4.61280465e-01 2.02976441e+00 -1.14308619e+00 -3.50340575e-01 -6.20152950e-01 3.98094803e-01 -3.90491813e-01 1.91046953e+00 5.78580260e-01 -1.57901454e+00 -3.84676039e-01 -9.67262626e-01 -7.46166050e-01 -5.41023135e-01 3.36018577e-02 1.02014184e+00 3.04406524e-01 -1.18222749e+00 2.59729862e-01 -6.12563252e-01 -4.91348505e-01 9.69608605e-01 -1.73038974e-01 -3.70008081e-01 -3.72139812e-01 -6.82322502e-01 1.04418099e+00 3.89295697e-01 -1.83888510e-01 -8.66545618e-01 -6.18264139e-01 -1.07256830e+00 3.01442504e-01 5.16757727e-01 -1.06565630e+00 1.35232484e+00 -5.48031688e-01 -9.14078832e-01 1.26934099e+00 -1.85823515e-01 -5.53829037e-02 3.33302349e-01 -6.02815747e-02 -1.80452608e-03 5.93534946e-01 1.05921872e-01 9.93150234e-01 5.77836156e-01 -1.52078211e+00 -1.84826791e-01 -1.99776322e-01 6.34814441e-01 2.39200145e-01 1.41867980e-01 -2.01205015e-01 -4.65613723e-01 -8.84462237e-01 7.88290203e-02 -5.60208321e-01 8.72666761e-02 6.38637066e-01 -4.06607866e-01 -4.08647150e-01 2.20493898e-01 -9.49698746e-01 7.25396872e-01 -1.86637568e+00 4.10415977e-01 -2.19144523e-01 7.63300717e-01 -1.87787339e-01 -5.35484374e-01 5.96065044e-01 2.94440016e-02 4.59553272e-01 -4.84942496e-01 -4.37030107e-01 6.89778090e-01 2.87391096e-01 -2.71835029e-01 1.27095059e-01 3.15328658e-01 1.66346979e+00 -8.94845068e-01 -7.00213790e-01 3.36488821e-02 1.29289240e-01 -5.75227678e-01 -1.90046787e-01 -6.20865226e-01 3.52826178e-01 -1.66835546e-01 8.86104643e-01 4.55055475e-01 -8.33200812e-01 2.14059949e-01 -1.61152378e-01 2.81105399e-01 3.99331391e-01 -6.88847959e-01 2.07761574e+00 -2.56671458e-01 7.69032896e-01 -1.25725538e-01 -6.14991426e-01 2.95033634e-01 2.69561082e-01 -1.44461647e-01 -8.14252794e-01 -4.91599320e-03 -1.39974341e-01 6.92882836e-02 -5.12601852e-01 3.98052901e-01 -5.77540286e-02 -2.54629642e-01 8.51787329e-01 1.13012075e-01 -4.25234079e-01 6.75211191e-01 6.35350525e-01 9.96875167e-01 3.96257460e-01 4.79875624e-01 -2.44738013e-01 1.70262590e-01 3.27406317e-01 2.75244892e-01 8.05158198e-01 -1.71628907e-01 4.91686374e-01 8.23815763e-01 -6.25600815e-01 -1.36641347e+00 -1.69409525e+00 2.91822225e-01 1.16656363e+00 9.94318426e-02 -5.36971033e-01 -5.18738687e-01 -3.63777816e-01 5.80046810e-02 7.84876049e-01 -7.31634080e-01 1.87051028e-01 -5.99799991e-01 -3.26919377e-01 8.97724450e-01 9.88793135e-01 2.91620582e-01 -1.50000966e+00 -7.10997820e-01 -2.67126620e-01 -3.27152371e-01 -1.11009729e+00 -3.30347359e-01 -2.09737256e-01 -7.06715107e-01 -1.32175040e+00 -7.09754050e-01 -7.34030426e-01 7.69834995e-01 1.49851376e-02 1.77927673e+00 6.63738132e-01 -6.20693207e-01 6.65434957e-01 -2.91772157e-01 -6.75817132e-02 -1.15238413e-01 -4.16504890e-01 -3.44412535e-01 -8.35468233e-01 -9.91036594e-02 -5.56218266e-01 -6.03941679e-01 -2.50260457e-02 -8.61330807e-01 6.38417661e-01 5.67906439e-01 5.94436407e-01 3.75499666e-01 -3.91878009e-01 2.39149362e-01 -6.98682606e-01 8.53166759e-01 -3.15467417e-01 -2.91856676e-01 4.57700402e-01 8.67308863e-03 1.14470422e-01 4.42313522e-01 -4.62165773e-01 -9.51659083e-01 -4.35871899e-01 1.72061563e-01 2.80169230e-02 -3.55829209e-01 3.88906598e-01 4.63578068e-02 1.04921766e-01 6.16916060e-01 1.44886509e-01 -5.39627373e-01 -3.96592081e-01 7.55049586e-01 -3.82380709e-02 7.11025119e-01 -1.25666046e+00 6.10020697e-01 5.07399857e-01 -7.86908120e-02 -6.01491690e-01 -6.76972270e-01 -5.24408417e-03 -6.42942429e-01 -3.35695222e-02 9.61593926e-01 -7.33745337e-01 -1.53967500e+00 1.58708453e-01 -1.37530875e+00 -9.37593699e-01 -2.58100033e-01 1.53212026e-01 -9.42969978e-01 5.78571379e-01 -1.06845152e+00 -5.80718696e-01 6.39478266e-02 -9.32754397e-01 8.96617234e-01 1.76433027e-01 -8.02791834e-01 -1.02138638e+00 -2.09005386e-01 6.86986208e-01 4.36652936e-02 2.62186736e-01 1.69708073e+00 -2.75091767e-01 -7.62369871e-01 3.84842813e-01 -8.32220018e-01 -4.22379941e-01 -2.12117076e-01 -1.95090666e-01 -8.29653919e-01 -4.18004878e-02 -4.22349989e-01 -9.23660815e-01 1.03095448e+00 6.36161566e-02 1.25907886e+00 -3.89138401e-01 -1.18639015e-01 4.48181510e-01 1.31211317e+00 -6.68357417e-04 5.89501083e-01 3.47745538e-01 7.70627975e-01 5.92933774e-01 4.48805355e-02 3.05822700e-01 7.68640518e-01 3.25885534e-01 3.15281898e-01 1.07502118e-01 -4.72195894e-01 -2.44143918e-01 -3.97478603e-02 3.04007530e-01 -3.21340561e-01 -5.22680819e-01 -1.43682349e+00 6.35815680e-01 -1.82719028e+00 -1.17723441e+00 9.07638893e-02 1.50830281e+00 9.67723548e-01 2.18420997e-01 1.40852123e-01 -3.77365239e-02 2.88799107e-01 3.17846537e-01 -7.36985743e-01 -6.22024834e-01 -1.72200292e-01 1.87474951e-01 -2.27472410e-02 6.61164463e-01 -6.52364492e-01 1.07409418e+00 7.28022623e+00 7.02661157e-01 -2.99200833e-01 4.42351028e-03 4.30051774e-01 -1.51622191e-01 -7.36379981e-01 -1.58192143e-01 -1.28039390e-01 8.62152223e-03 3.26782852e-01 4.19399105e-02 8.91560078e-01 2.53213704e-01 -1.35137737e-01 -3.16189557e-01 -1.32452190e+00 1.01044118e+00 2.30391473e-01 -1.51630890e+00 4.29741681e-01 -2.32923031e-01 3.99135411e-01 -2.55191565e-01 2.68326372e-01 2.38195300e-01 7.03191340e-01 -1.56125557e+00 1.12557089e+00 6.16606116e-01 7.02531815e-01 -3.33129406e-01 -8.69081914e-02 2.45273620e-01 -1.33985126e+00 7.00580096e-03 -1.48182765e-01 -4.55227315e-01 2.47598127e-01 -8.77799168e-02 -3.62868637e-01 2.13199094e-01 7.56409049e-01 5.47700286e-01 -9.53857243e-01 9.00318444e-01 -7.04947233e-01 1.31268904e-01 -6.80554332e-03 2.03184057e-02 2.09389001e-01 1.62472222e-02 3.44321758e-01 1.22311270e+00 -1.25596583e-01 3.66372466e-01 1.61341682e-01 1.37537396e+00 -4.88210581e-02 -3.67100745e-01 -3.68629217e-01 -5.90800047e-01 4.78445917e-01 1.04212761e+00 -1.21717286e+00 -5.56410193e-01 -4.06725705e-01 1.15584254e+00 8.92278850e-01 7.42158711e-01 -8.40966821e-01 -2.12257132e-01 7.42154896e-01 -1.63419411e-01 2.48139277e-01 -9.31142092e-01 -9.00814295e-01 -1.11844182e+00 -4.96623144e-02 -7.78252482e-01 5.66729009e-01 -1.71200764e+00 -1.51796484e+00 4.16898310e-01 1.91244273e-03 -5.85583687e-01 4.74153124e-02 -1.19900966e+00 -5.19849777e-01 9.47678268e-01 -1.01437593e+00 -1.50912416e+00 -4.15905446e-01 4.14853185e-01 6.72868431e-01 8.13914612e-02 8.36765051e-01 -2.73749083e-01 -2.85688162e-01 5.06500423e-01 -5.59735119e-01 5.33159077e-01 2.64836431e-01 -1.47735870e+00 1.11754584e+00 4.95807141e-01 4.72678721e-01 8.44594598e-01 6.15030468e-01 -7.19117761e-01 -1.32245731e+00 -2.33184680e-01 5.00736475e-01 -1.03418922e+00 7.97159016e-01 -6.18867576e-01 -8.42854857e-01 1.05393696e+00 5.75346589e-01 -1.69145420e-01 7.02227354e-01 2.96806723e-01 -1.13350880e+00 7.46205568e-01 -1.24850559e+00 1.16337991e+00 1.65033960e+00 -8.70078206e-01 -1.16173220e+00 2.11000368e-01 7.64138997e-01 -4.93813366e-01 -7.29120672e-01 1.89077505e-03 8.26110482e-01 -9.87855911e-01 1.36082864e+00 -1.03641117e+00 6.98572576e-01 -1.67153507e-01 -2.85418779e-01 -1.25966239e+00 -8.86008024e-01 5.39298542e-03 -2.42858708e-01 8.63603115e-01 5.14223993e-01 -6.17761612e-01 6.81864500e-01 6.95966661e-01 -8.04596320e-02 -7.74228334e-01 -6.33181393e-01 -7.02620685e-01 1.84892908e-01 -4.52884585e-01 3.98762882e-01 7.79351890e-01 4.99705911e-01 3.00241321e-01 1.85575262e-01 6.49277046e-02 7.06364691e-01 4.17219549e-01 3.90347123e-01 -1.12304175e+00 -5.11095583e-01 -9.43038583e-01 -9.42114070e-02 -1.08026576e+00 2.19941214e-01 -1.10048413e+00 -2.61960477e-01 -2.20103788e+00 3.30854535e-01 -1.81001991e-01 -4.22108509e-02 8.92556965e-01 -4.17828746e-02 7.52479851e-01 4.71143484e-01 1.90614358e-01 -7.76904225e-01 2.02445596e-01 1.73753452e+00 -3.18662584e-01 4.51825678e-01 -8.46167326e-01 -9.99072909e-01 9.88959074e-01 1.06096160e+00 -1.35293320e-01 -4.29375052e-01 -1.00642741e+00 6.59507513e-01 -5.39568551e-02 9.57593024e-01 -7.04826057e-01 -6.50309026e-02 -5.69149852e-01 8.79494011e-01 -4.07105029e-01 8.17885160e-01 -4.52058673e-01 -3.12149853e-01 4.60176796e-01 -4.82905388e-01 4.33447450e-01 6.43137217e-01 4.56072986e-01 2.29744270e-01 -2.06140261e-02 5.43309808e-01 -7.31232941e-01 -9.91356373e-01 6.90299347e-02 -5.56948423e-01 6.25877202e-01 6.25152409e-01 -3.64708126e-01 -1.06681585e+00 -2.71148145e-01 -9.31206584e-01 2.89078832e-01 7.06405640e-01 2.11190209e-01 7.27983832e-01 -1.42836845e+00 -4.64268774e-01 -1.88897744e-01 3.63162816e-01 -2.32107684e-01 3.51553112e-01 6.67834640e-01 -7.93474972e-01 4.58704591e-01 -6.41966403e-01 -1.18574880e-01 -9.63363469e-01 8.29062879e-01 -7.45386928e-02 6.53128028e-02 -7.93730140e-01 9.73067880e-01 4.89815712e-01 -3.81768584e-01 2.22693965e-01 -7.46645153e-01 -5.98777048e-02 -2.80193925e-01 5.49055874e-01 3.09800178e-01 -6.49292290e-01 -4.06430602e-01 -4.05396402e-01 7.63621628e-01 2.94817030e-01 -3.84408832e-01 1.24988496e+00 -1.50321215e-01 -3.23147357e-01 6.39601052e-01 6.38989151e-01 2.31746852e-01 -1.22842407e+00 -3.36183235e-02 1.65311992e-01 -3.83942455e-01 -7.41624534e-01 -1.33922803e+00 -5.06376684e-01 1.08785605e+00 -3.12877983e-01 2.10769475e-01 7.64182389e-01 4.91909295e-01 6.15545034e-01 5.11466086e-01 4.01493251e-01 -6.20174289e-01 6.57397509e-01 5.92765629e-01 1.36793721e+00 -1.15860629e+00 7.09639639e-02 -3.97909850e-01 -8.20474982e-01 8.92106652e-01 7.73103714e-01 -2.28852823e-01 3.99785936e-02 2.68527597e-01 -8.88149515e-02 -4.99632269e-01 -8.88321996e-01 -2.55697399e-01 4.25268054e-01 8.87028515e-01 4.43551630e-01 1.18796788e-01 -9.75518371e-04 8.02387178e-01 -3.45370352e-01 -2.94379652e-01 3.65242720e-01 8.45593810e-01 -5.89199364e-01 -6.78231359e-01 -4.13718015e-01 2.74408013e-01 5.60294501e-02 -3.47756922e-01 -6.74762964e-01 9.15504932e-01 3.98229696e-02 6.64387226e-01 2.08131537e-01 6.94012567e-02 4.70899455e-02 3.66234154e-01 1.13605738e+00 -6.31646633e-01 -3.75969738e-01 -6.09713972e-01 3.48635465e-01 -6.49668753e-01 -2.02913210e-01 -4.57317799e-01 -1.67055202e+00 -5.87833703e-01 6.54242992e-01 -1.99626744e-01 1.05260432e-01 8.74229729e-01 1.50180370e-01 6.07618332e-01 -6.84194207e-01 -1.15424466e+00 9.65525955e-03 -8.61491501e-01 -4.26485747e-01 5.69685876e-01 2.60732681e-01 -7.46696532e-01 -1.76338285e-01 2.29065254e-01]

[10.797025680541992, 2.012969970703125]

2c3fada0-4301-4dea-9afc-c501177449bf

an-improved-raftstereo-trained-with-a-mixed

2210.12785

https://arxiv.org/abs/2210.12785v1

https://arxiv.org/pdf/2210.12785v1.pdf

An Improved RaftStereo Trained with A Mixed Dataset for the Robust Vision Challenge 2022

Stereo-matching is a fundamental problem in computer vision. Despite recent progress by deep learning, improving the robustness is ineluctable when deploying stereo-matching models to real-world applications. Different from the common practices, i.e., developing an elaborate model to achieve robustness, we argue that collecting multiple available datasets for training is a cheaper way to increase generalization ability. Specifically, this report presents an improved RaftStereo trained with a mixed dataset of seven public datasets for the robust vision challenge (denoted as iRaftStereo_RVC). When evaluated on the training sets of Middlebury, KITTI-2015, and ETH3D, the model outperforms its counterparts trained with only one dataset, such as the popular Sceneflow. After fine-tuning the pre-trained model on the three datasets of the challenge, it ranks at 2nd place on the stereo leaderboard, demonstrating the benefits of mixed dataset pre-training.

['Wenjie Jiang', 'Rui Xu', 'Hualie Jiang']

2022-10-23

['stereo-matching-1']

['computer-vision']

[ 2.28354603e-01 -1.79356396e-01 7.79451989e-03 -2.88886547e-01 -7.79741883e-01 -6.20042264e-01 9.20920968e-01 -2.09655330e-01 -8.12332213e-01 3.93055975e-01 2.24046752e-01 -1.41439140e-01 8.83630887e-02 -4.21303004e-01 -9.79097486e-01 -5.33997595e-01 2.14830309e-01 4.49780136e-01 4.73326921e-01 -5.65517545e-01 3.76757473e-01 4.19278115e-01 -1.93417156e+00 3.33020836e-01 8.91230822e-01 1.02098846e+00 3.53299491e-02 3.77128273e-01 2.91868418e-01 6.83100820e-01 -1.14641726e-01 -7.90274501e-01 9.41661119e-01 1.75865680e-01 -8.92964721e-01 4.91710715e-02 1.39610445e+00 -4.14142817e-01 -4.27326590e-01 1.00047076e+00 5.68537951e-01 -4.10872586e-02 2.85565048e-01 -1.30437028e+00 -2.10511178e-01 3.10862243e-01 -6.09096050e-01 7.24214464e-02 2.14233354e-01 6.67699456e-01 8.63039672e-01 -5.83067656e-01 1.07231772e+00 1.30514538e+00 1.06049705e+00 4.48835760e-01 -1.31846559e+00 -7.63960063e-01 -1.43751400e-02 3.22724670e-01 -1.17913866e+00 -6.02152109e-01 4.59043503e-01 -7.33053625e-01 8.70187163e-01 1.16786666e-01 2.71171808e-01 1.49618614e+00 -9.95435044e-02 7.36114025e-01 1.29296029e+00 -3.44616398e-02 -1.30627421e-03 -1.50928631e-01 1.11783803e-01 3.00064564e-01 3.35575759e-01 7.79929101e-01 -4.78568643e-01 1.09996744e-01 4.76617157e-01 -2.82204360e-01 -2.76604563e-01 -8.53700101e-01 -1.31750214e+00 6.16906524e-01 6.58342183e-01 -4.74689230e-02 -1.16935946e-01 -9.64004323e-02 7.49608040e-01 5.02398372e-01 3.62397254e-01 5.72510302e-01 -4.09053534e-01 -5.03015667e-02 -1.12307894e+00 6.19205415e-01 4.50995147e-01 8.96816611e-01 8.17783713e-01 -2.61365652e-01 -1.86972529e-01 7.60070980e-01 -4.42622118e-02 4.93441910e-01 1.25815213e-01 -1.12210333e+00 7.89166033e-01 6.42964721e-01 1.07864756e-02 -6.99223876e-01 -4.29862529e-01 -5.36857307e-01 -9.41483080e-01 6.10172629e-01 8.56024146e-01 1.87971955e-03 -1.09496522e+00 1.58928657e+00 2.76857793e-01 2.33749747e-01 -6.03573173e-02 1.01607394e+00 9.62408900e-01 1.26696825e-01 -9.76734236e-02 5.44751406e-01 8.40297580e-01 -1.18200028e+00 3.76566425e-02 -4.97572809e-01 5.71107447e-01 -8.81636202e-01 8.90854895e-01 5.64182997e-01 -5.37454963e-01 -8.52852762e-01 -1.18220389e+00 -3.39788437e-01 -4.05788988e-01 -2.09091812e-01 6.81955576e-01 3.69262606e-01 -1.19266462e+00 8.03170562e-01 -4.51097846e-01 -7.19066024e-01 4.68533456e-01 8.73788744e-02 -9.97523069e-01 -3.57577711e-01 -8.85811508e-01 9.77216780e-01 4.60971653e-01 -5.50903939e-02 -1.04179668e+00 -1.09773231e+00 -7.31877983e-01 -4.50412601e-01 1.88320592e-01 -9.05185640e-01 1.11300886e+00 -7.78934658e-01 -1.23200297e+00 1.49985814e+00 3.25665265e-01 -8.55796635e-01 1.44141519e+00 -4.39494431e-01 -9.25471112e-02 -2.50732541e-01 1.55721769e-01 1.21026409e+00 9.51724410e-01 -1.05378664e+00 -7.93872476e-01 -3.44741255e-01 3.70870620e-01 -4.88565899e-02 1.96219638e-01 -1.34921402e-01 -6.04595959e-01 -3.35728437e-01 1.25822842e-01 -1.12635481e+00 -3.85102451e-01 5.96841499e-02 -5.57740748e-01 5.44465557e-02 5.46031833e-01 -5.64742029e-01 5.42360306e-01 -2.29802918e+00 2.01060906e-01 -4.20910791e-02 1.61439627e-01 4.28758711e-01 -3.55248988e-01 3.01597506e-01 -4.17219460e-01 -2.39064902e-01 -2.67172277e-01 -4.48696285e-01 -6.43453747e-02 -8.22007507e-02 -2.13232502e-01 7.81775355e-01 1.36370689e-01 7.54131138e-01 -8.27922106e-01 -3.17548871e-01 6.45228028e-01 2.92567909e-01 -7.65781283e-01 1.45599067e-01 3.00366208e-02 7.59155095e-01 2.10337713e-02 5.22501826e-01 1.08942735e+00 -7.49635175e-02 -2.34768808e-01 -3.41392308e-01 -2.70045429e-01 1.28717586e-01 -1.37211347e+00 2.02653837e+00 -3.49680424e-01 7.59093583e-01 -6.62895292e-02 -7.83492327e-01 7.73404241e-01 -8.59726146e-02 3.48459482e-01 -1.17992985e+00 -8.45518801e-03 2.65650332e-01 5.25887735e-05 -3.07730347e-01 4.89607126e-01 2.66629875e-01 6.07951395e-02 3.99718173e-02 3.95740062e-01 -3.01299125e-01 4.84202951e-01 1.74833760e-01 1.13627768e+00 3.57479870e-01 -1.42386975e-02 -5.35748601e-01 6.71896756e-01 5.79461008e-02 5.11171162e-01 8.70443642e-01 -4.92044657e-01 1.22358620e+00 1.54837981e-01 -7.89406180e-01 -1.24245715e+00 -1.00068891e+00 -3.76615047e-01 9.30867672e-01 2.27997482e-01 -4.40288156e-01 -6.59321427e-01 -7.67229557e-01 2.62503564e-01 3.78682792e-01 -7.46143520e-01 -2.78967321e-02 -4.08048779e-01 -4.76862550e-01 6.81368530e-01 2.82300770e-01 9.50051248e-01 -8.01423371e-01 -7.24637747e-01 -1.37380779e-01 -2.66716450e-01 -1.40127063e+00 -2.68716693e-01 -2.08635330e-02 -7.04834282e-01 -1.49048805e+00 -4.93479699e-01 -2.82905966e-01 1.92553818e-01 5.50553262e-01 1.56357789e+00 9.45521370e-02 -4.12934959e-01 1.99502349e-01 -2.77653426e-01 -3.21411163e-01 -3.32752258e-01 5.04323125e-01 -1.73672438e-01 -6.10044003e-02 1.54456615e-01 -5.23925364e-01 -6.75122380e-01 3.86678964e-01 -9.27048147e-01 2.66426176e-01 4.04936075e-01 7.16549695e-01 3.73921573e-01 -5.20705879e-01 -9.15652886e-02 -8.51451457e-01 -6.41232729e-02 -2.08087966e-01 -1.00250816e+00 1.83788873e-02 -4.57235157e-01 1.30513102e-01 3.20042640e-01 -6.79084007e-03 -8.85824978e-01 2.28443682e-01 -3.17867607e-01 -4.84339178e-01 -4.16722804e-01 -4.87471484e-02 -3.67398597e-02 -3.76244336e-01 1.07470250e+00 -1.33992299e-01 5.05246259e-02 -5.52948177e-01 4.35249656e-01 1.50234431e-01 9.26711380e-01 -4.30477113e-01 1.23974597e+00 7.71818936e-01 1.52436346e-01 -7.02986479e-01 -8.94285500e-01 -6.75702333e-01 -8.09137464e-01 -1.42480791e-01 8.19649875e-01 -1.20300937e+00 -6.31189704e-01 7.69647241e-01 -1.11786687e+00 -2.85964459e-01 -1.49580330e-01 3.39357346e-01 -6.16875350e-01 4.09512937e-01 -1.08679324e-01 -2.02457130e-01 -2.02331468e-01 -1.35865045e+00 1.29285383e+00 -1.75406225e-02 -1.70861065e-01 -8.07816029e-01 4.73273277e-01 7.23332524e-01 5.50520778e-01 5.08957565e-01 2.52694637e-01 -4.04687852e-01 -6.38146222e-01 -1.49678618e-01 -4.79498893e-01 5.03513336e-01 -1.56615838e-01 1.87716141e-01 -1.47434640e+00 -4.58457738e-01 -3.85372460e-01 -7.13581979e-01 1.40971768e+00 8.97405520e-02 9.43324685e-01 3.78401220e-01 -7.69735128e-02 1.30690479e+00 1.42695081e+00 -4.81348336e-01 8.36344063e-01 1.03195357e+00 7.70195305e-01 6.18431926e-01 5.08386672e-01 1.78186864e-01 5.18631697e-01 8.31523478e-01 9.38249588e-01 -3.36568773e-01 -3.66857648e-01 -2.33453855e-01 1.43710583e-01 -5.53223267e-02 -3.38711888e-01 2.85902739e-01 -1.06522346e+00 3.70119691e-01 -1.87320006e+00 -1.18370497e+00 -2.17867926e-01 2.34419727e+00 5.37058234e-01 3.01499963e-01 1.73885867e-01 1.17377788e-01 4.83355045e-01 3.92104626e-01 -5.28052092e-01 4.67856135e-03 -5.43656290e-01 1.01913989e-01 8.37396264e-01 3.64507914e-01 -1.58558583e+00 1.02187634e+00 6.23930883e+00 6.28177643e-01 -1.44119024e+00 -4.84971814e-02 4.76379246e-01 4.56220564e-03 1.25137791e-01 1.08594827e-01 -7.51996398e-01 3.40267450e-01 6.24788284e-01 6.89210966e-02 3.68392080e-01 9.00867403e-01 -5.60881235e-02 -1.95190355e-01 -1.41464436e+00 1.19025993e+00 -1.23534501e-01 -1.51009953e+00 -1.36107046e-04 1.43535770e-02 8.63262236e-01 8.05944800e-01 1.03901751e-01 4.68121350e-01 3.96089405e-01 -1.07318819e+00 8.71823788e-01 3.75880748e-01 6.26172662e-01 -3.95379245e-01 9.37261522e-01 2.87686527e-01 -7.95797527e-01 3.62020247e-02 -4.39850688e-01 3.96975176e-03 -1.26390159e-01 6.14059746e-01 -6.73865855e-01 9.94734764e-01 1.15484321e+00 1.05189359e+00 -1.32131100e+00 1.47263932e+00 -1.14688218e-01 3.05415243e-01 -3.49018991e-01 9.07208502e-01 2.79917359e-01 -1.83874965e-01 5.96872032e-01 1.16534793e+00 -2.07573571e-03 -6.60074472e-01 8.44168440e-02 6.51191592e-01 -7.80706555e-02 -1.71318889e-01 -8.93342733e-01 4.96930897e-01 1.64959371e-01 1.38808632e+00 -2.90655673e-01 -6.59719408e-02 -3.95016283e-01 8.03875983e-01 4.74992514e-01 5.87797798e-02 -8.51542950e-01 5.47677390e-02 8.36548984e-01 1.14135891e-01 2.24664479e-01 -6.28442541e-02 -2.76376426e-01 -1.30711186e+00 3.12545672e-02 -1.32588232e+00 4.74393636e-01 -5.68663657e-01 -1.33807945e+00 6.71182215e-01 -1.56643018e-02 -1.48394108e+00 -2.91910857e-01 -5.98118067e-01 -6.41850114e-01 9.18088734e-01 -1.62817347e+00 -1.40618062e+00 -6.31069899e-01 8.17910254e-01 3.21130276e-01 -2.07816839e-01 4.48128670e-01 3.99060726e-01 -5.24798870e-01 7.20638514e-01 -3.75409089e-02 1.16134241e-01 1.06643260e+00 -1.15649283e+00 9.05353010e-01 1.09833670e+00 1.73929170e-01 3.52676481e-01 1.04476047e+00 -1.42248511e-01 -1.26383018e+00 -1.09740317e+00 6.66908443e-01 -7.10100651e-01 6.43043995e-01 -3.28678906e-01 -7.03474045e-01 6.40282691e-01 3.41817915e-01 1.89549074e-01 7.58802593e-02 1.62055507e-01 -9.31468248e-01 -3.70624840e-01 -1.06445694e+00 5.59323609e-01 1.51394546e+00 -5.32573223e-01 -6.74948633e-01 2.50467718e-01 5.01547515e-01 -6.99678779e-01 -5.70555210e-01 7.43186772e-01 5.75955272e-01 -1.71137357e+00 1.20950234e+00 -4.56015676e-01 5.04334629e-01 -3.81120771e-01 -3.47346127e-01 -1.38190556e+00 -2.44864896e-01 -5.37795365e-01 6.65683329e-01 1.19566071e+00 2.75083661e-01 -5.17277539e-01 8.83979857e-01 2.42245540e-01 -2.18997002e-01 -7.75998607e-02 -1.08987153e+00 -1.06568992e+00 1.84989348e-01 -4.63842928e-01 5.02692938e-01 8.40470731e-01 -6.98552787e-01 2.96708465e-01 -3.24323297e-01 -5.44745065e-02 1.03830433e+00 1.93501320e-02 1.54889667e+00 -1.49651504e+00 -2.22844645e-01 -4.38053042e-01 -7.92425334e-01 -7.44642138e-01 1.86090201e-01 -9.40691650e-01 -5.00688590e-02 -1.24278212e+00 1.56741217e-01 -3.42317700e-01 -2.17944592e-01 3.91571194e-01 -2.12494090e-01 4.35083807e-01 5.05423844e-01 7.26516992e-02 -6.57066226e-01 3.33431572e-01 1.16908014e+00 -3.84620100e-01 9.93535891e-02 -5.36088981e-02 -4.64569658e-01 6.70508206e-01 5.71766317e-01 -2.03648299e-01 -2.52661407e-01 -5.54147303e-01 1.67528138e-01 -3.91275853e-01 8.34824622e-01 -1.55498981e+00 1.53198689e-01 8.60989746e-03 1.84860796e-01 -7.42848635e-01 2.81236351e-01 -6.86286986e-01 4.62425798e-01 4.42857742e-01 -2.12166309e-01 1.75234731e-02 4.07048166e-01 1.57489017e-01 -2.58522362e-01 1.87186539e-01 1.05139196e+00 -5.60113415e-02 -1.24632835e+00 4.37552810e-01 3.44744295e-01 4.34450924e-01 7.18180537e-01 -2.86487490e-01 -6.73071444e-01 -1.35149017e-01 -2.11624831e-01 3.87913316e-01 7.74201870e-01 8.32401693e-01 1.89377904e-01 -1.05975974e+00 -9.33087111e-01 2.05064863e-01 6.81334198e-01 7.42681697e-02 4.76742536e-01 8.22936296e-01 -7.24819303e-01 4.43107098e-01 -7.35084116e-01 -1.06846261e+00 -1.18395042e+00 5.67251146e-01 5.70132732e-01 -3.59206587e-01 -7.38360643e-01 9.93280888e-01 2.79365271e-01 -7.93079793e-01 2.83633679e-01 -2.65390635e-01 -4.31092456e-02 4.44888836e-03 3.98155898e-01 4.38258469e-01 5.57156801e-01 -7.46716559e-01 -5.42643547e-01 6.40815079e-01 -7.96198845e-02 -1.29534109e-02 1.34802628e+00 1.60395559e-02 -9.85275805e-02 9.82014909e-02 1.19267404e+00 -1.95448458e-01 -1.60676980e+00 -2.12384924e-01 2.41932482e-01 -6.73363328e-01 -3.00858216e-03 -6.67657256e-01 -1.26235521e+00 8.08692396e-01 9.05593574e-01 -2.19440296e-01 8.28587532e-01 -2.51006395e-01 3.75283301e-01 5.29354811e-01 7.74428666e-01 -9.66206729e-01 -1.93366528e-01 8.29846621e-01 1.10550070e+00 -1.76070917e+00 3.71737592e-02 -2.27923483e-01 -5.96492231e-01 9.31647420e-01 8.47921312e-01 -3.18858445e-01 5.53405464e-01 2.29800157e-02 4.15137738e-01 -3.04804300e-03 -6.01134241e-01 -5.07857084e-01 3.98857594e-01 7.54186511e-01 2.66268015e-01 -1.95010081e-01 1.51023101e-02 -9.05503109e-02 -5.13993859e-01 7.61678740e-02 2.77917683e-01 6.30548716e-01 -5.67999296e-02 -1.15646076e+00 -4.02841389e-01 1.14689253e-01 -2.55672008e-01 -5.41422814e-02 -4.83553499e-01 1.11386955e+00 4.47343051e-01 9.69847381e-01 -4.73466143e-02 -5.82818270e-01 6.95446253e-01 -2.40025863e-01 5.23520350e-01 -3.49133104e-01 -8.61620188e-01 -5.36437929e-01 1.47948563e-01 -1.18041837e+00 -5.36438644e-01 -5.61952889e-01 -4.50794846e-01 -6.56024933e-01 2.37843782e-01 -3.76381755e-01 5.62794209e-01 9.56219196e-01 3.19315851e-01 2.43410975e-01 5.08080840e-01 -1.25520396e+00 -6.26019120e-01 -8.10299993e-01 -1.48713738e-01 9.86430526e-01 4.15759355e-01 -7.79772043e-01 -3.12044203e-01 -1.92053303e-01]

[8.68610668182373, -2.2377936840057373]

fa0c4f64-6bd6-4d16-9823-ee71e544ab56

scalp-superpixels-with-contour-adherence

1903.07149

http://arxiv.org/abs/1903.07149v1

http://arxiv.org/pdf/1903.07149v1.pdf

SCALP: Superpixels with Contour Adherence using Linear Path

Superpixel decomposition methods are generally used as a pre-processing step to speed up image processing tasks. They group the pixels of an image into homogeneous regions while trying to respect existing contours. For all state-of-the-art superpixel decomposition methods, a trade-off is made between 1) computational time, 2) adherence to image contours and 3) regularity and compactness of the decomposition. In this paper, we propose a fast method to compute Superpixels with Contour Adherence using Linear Path (SCALP) in an iterative clustering framework. The distance computed when trying to associate a pixel to a superpixel during the clustering is enhanced by considering the linear path to the superpixel barycenter. The proposed framework produces regular and compact superpixels that adhere to the image contours. We provide a detailed evaluation of SCALP on the standard Berkeley Segmentation Dataset. The obtained results outperform state-of-the-art methods in terms of standard superpixel and contour detection metrics.

['Vinh-Thong Ta', 'Rémi Giraud', 'Nicolas Papadakis']

2019-03-17

['contour-detection']

['computer-vision']

[ 4.74291772e-01 1.01041786e-01 9.23443809e-02 -3.20353627e-01 -4.29006755e-01 -5.97700894e-01 3.59550685e-01 5.61071157e-01 -6.07730925e-01 4.86920267e-01 -2.79559851e-01 2.60864198e-02 7.03766523e-03 -9.01167750e-01 -4.82352465e-01 -7.70090401e-01 2.83403937e-02 4.37740922e-01 8.42239559e-01 2.69529790e-01 5.15053868e-01 5.59804797e-01 -1.54324245e+00 5.09701595e-02 1.18581307e+00 8.28716815e-01 1.46710441e-01 6.69588089e-01 -1.22797623e-01 1.26971707e-01 -3.58300567e-01 -1.90064028e-01 2.92450905e-01 -4.91949677e-01 -9.03284132e-01 6.78230464e-01 6.18204832e-01 1.05313614e-01 6.86709523e-01 1.51871455e+00 -8.43710266e-03 6.01988509e-02 7.69789577e-01 -8.43495011e-01 -2.09433600e-01 5.20769060e-01 -1.06958091e+00 2.43011594e-01 -1.69711009e-01 -3.09321254e-01 8.42120767e-01 -5.88669360e-01 8.77687991e-01 9.06732619e-01 6.88970685e-01 2.51430869e-02 -1.47283137e+00 -2.43117228e-01 6.64734691e-02 4.74293791e-02 -1.40453959e+00 -1.23736046e-01 8.49209607e-01 -5.38615286e-01 3.71839553e-01 4.27051783e-01 6.93255484e-01 -7.77899176e-02 5.79159036e-02 6.51272416e-01 1.20037401e+00 -5.64660788e-01 6.55245423e-01 1.26529008e-01 7.16655910e-01 5.41976094e-01 5.28630197e-01 -3.95455927e-01 -3.78883667e-02 6.91963434e-02 9.31681812e-01 -2.69221365e-01 -2.36688867e-01 -3.98014188e-01 -8.32793474e-01 5.69275558e-01 4.74325657e-01 8.19533527e-01 -5.97504854e-01 1.69169515e-01 8.83429199e-02 -2.51272649e-01 6.65353060e-01 1.26510695e-01 2.57577449e-02 2.51428574e-01 -1.75495946e+00 -4.33628447e-02 5.05983055e-01 7.73861349e-01 1.05719876e+00 -3.85172188e-01 -5.16357236e-02 5.34581304e-01 2.36105204e-01 -1.95387334e-01 1.83953315e-01 -1.06676078e+00 9.89480391e-02 9.48674619e-01 4.17538434e-02 -1.11788547e+00 -2.64797062e-01 -4.96052414e-01 -9.24445212e-01 6.00533247e-01 6.69019222e-01 4.45176661e-02 -1.03198397e+00 1.25798905e+00 7.00876772e-01 2.03307718e-01 -2.98938043e-02 7.55173981e-01 3.64089370e-01 6.55236781e-01 2.84513682e-02 -5.73130012e-01 1.42987669e+00 -9.30077910e-01 -5.76432168e-01 -2.04420649e-02 2.50204295e-01 -9.31480944e-01 5.89142978e-01 4.58654433e-01 -1.39765537e+00 -7.44012415e-01 -1.05175924e+00 1.10989392e-01 -1.45360157e-01 2.79174894e-01 4.78778124e-01 8.82732451e-01 -1.27687252e+00 7.56739795e-01 -1.02091157e+00 -5.62926829e-01 3.85454565e-01 3.11834008e-01 -1.88978255e-01 3.14140767e-01 -3.44602734e-01 3.65162641e-01 7.48003304e-01 -1.75292790e-02 -4.86464947e-01 -5.34095764e-01 -4.75203872e-01 1.86517403e-01 2.68018812e-01 -5.29345751e-01 7.18177557e-01 -1.25940180e+00 -1.08922768e+00 1.16658151e+00 -1.70949444e-01 -5.14550924e-01 8.08956742e-01 -4.91240586e-04 2.11845621e-01 5.61138511e-01 9.17568132e-02 7.65520096e-01 9.51087713e-01 -1.61400831e+00 -9.87797320e-01 -4.46499228e-01 -3.23182702e-01 8.83758813e-02 -1.30451187e-01 -4.37195376e-02 -8.56809318e-01 -6.91860080e-01 6.31931722e-01 -6.00740016e-01 -3.56752664e-01 -1.49256721e-01 -7.41771042e-01 -3.75620723e-01 9.26310539e-01 -4.92544442e-01 1.31407857e+00 -1.98581016e+00 2.66878039e-01 5.35214484e-01 1.29962042e-01 2.34230578e-01 3.52389783e-01 -8.61884654e-02 -8.68778601e-02 2.49181733e-01 -5.98543465e-01 -6.52939916e-01 -3.60904783e-01 -2.14483827e-01 1.86847806e-01 6.30210459e-01 -2.64461637e-01 3.44994307e-01 -7.07648695e-01 -1.14328742e+00 2.80144125e-01 3.56545001e-01 -3.25843960e-01 1.16446033e-01 -1.80435613e-01 3.08138490e-01 -3.27999622e-01 5.11154711e-01 1.03304768e+00 2.47812252e-02 1.03661545e-01 -5.17009437e-01 -6.18945301e-01 -4.60380197e-01 -1.55259812e+00 1.44521749e+00 1.52982756e-01 3.61323476e-01 2.90347576e-01 -9.57422733e-01 1.02050495e+00 2.04932258e-01 6.06376350e-01 -1.43701807e-01 1.19628295e-01 2.16885790e-01 -4.31707174e-01 -1.04335763e-01 6.11313105e-01 1.93012413e-02 3.84612888e-01 5.31341255e-01 -1.44248530e-01 -3.67856145e-01 8.33892643e-01 5.37161110e-03 5.06221414e-01 1.85566619e-01 3.14220160e-01 -9.73913908e-01 8.07550788e-01 1.87264740e-01 6.22018337e-01 3.72281641e-01 -1.66038871e-01 8.09067726e-01 5.44494390e-01 -2.76673198e-01 -1.15313888e+00 -8.44962895e-01 -2.51410306e-01 5.98267019e-01 3.89738142e-01 -4.71503846e-02 -1.59409201e+00 -2.86823273e-01 -3.26009542e-01 5.20433903e-01 -6.13274932e-01 5.34112394e-01 -5.81098676e-01 -8.87180150e-01 2.17370257e-01 4.23922181e-01 9.89392161e-01 -7.76279390e-01 -1.08015621e+00 2.03834459e-01 -1.11940838e-01 -9.77142155e-01 -6.59209251e-01 1.25818089e-01 -1.35371244e+00 -1.00182700e+00 -9.58217680e-01 -1.06788445e+00 1.28980052e+00 3.53398591e-01 8.89696896e-01 2.94534445e-01 -3.92117143e-01 4.45939377e-02 -4.15095627e-01 1.60621643e-01 -3.81939948e-01 -1.46055192e-01 -3.75729680e-01 2.35282093e-01 -8.39980841e-02 -5.50935566e-01 -8.16543102e-01 4.04234707e-01 -1.05444348e+00 2.19208241e-01 5.29515326e-01 3.90534163e-01 1.16172683e+00 5.84541202e-01 -2.61923730e-01 -9.76546764e-01 4.02518600e-01 -5.37080765e-02 -1.02884018e+00 5.06413937e-01 -6.22012973e-01 5.55668660e-02 2.01657683e-01 -1.57529950e-01 -1.20224249e+00 4.99716848e-01 2.34329909e-01 -1.19658872e-01 -3.60138476e-01 2.68896431e-01 6.89577835e-04 -1.40366703e-01 3.40847611e-01 2.68811673e-01 -3.87579441e-01 -5.25177956e-01 5.47693014e-01 4.55605030e-01 7.17442214e-01 -5.56880355e-01 5.17813861e-01 8.44927609e-01 1.07859030e-01 -1.02289987e+00 -4.04833913e-01 -8.30913126e-01 -1.20683110e+00 -3.25628370e-01 1.32778203e+00 -2.03295946e-01 -4.61936742e-01 6.00577235e-01 -1.16537082e+00 -2.83623397e-01 -7.07607195e-02 1.16071485e-01 -5.02559125e-01 8.71647775e-01 -5.27100325e-01 -1.05426264e+00 -5.25976419e-01 -9.88166153e-01 9.07127619e-01 5.38920283e-01 -1.79453984e-01 -9.29405272e-01 1.13207258e-01 4.43947375e-01 1.06673921e-02 6.21298194e-01 8.85262072e-01 -1.03511833e-01 -5.72305024e-01 -1.14929363e-01 -4.46741402e-01 4.48862910e-01 1.25699058e-01 4.49009806e-01 -7.26930916e-01 -1.32243782e-01 9.80404988e-02 2.81285793e-01 1.08780062e+00 7.61649370e-01 7.26383984e-01 -2.43052721e-01 -5.13571620e-01 6.57544494e-01 1.88567638e+00 2.13040516e-01 6.96667254e-01 2.87256837e-01 5.42677522e-01 7.55228579e-01 7.59839892e-01 2.46425331e-01 -1.09368652e-01 4.55548435e-01 3.47874999e-01 -3.92761588e-01 -1.97276503e-01 1.90915257e-01 -7.49427825e-02 4.43348557e-01 -4.10072565e-01 -9.48528871e-02 -8.68239284e-01 9.09604788e-01 -1.78222811e+00 -7.59503722e-01 -7.32569754e-01 2.27183366e+00 8.76601398e-01 4.98222321e-01 2.00885221e-01 5.52529871e-01 1.09387505e+00 -2.92104892e-02 -1.64595544e-01 -2.91278481e-01 2.62017027e-02 1.13490745e-01 6.22960687e-01 9.26101506e-01 -1.41682684e+00 1.12477040e+00 5.49234056e+00 1.03731847e+00 -9.01668906e-01 1.56659573e-01 1.01200831e+00 5.07699072e-01 8.99957642e-02 6.16780296e-02 -7.36542046e-01 2.38723367e-01 5.21064922e-02 6.87929019e-02 -4.56142016e-02 7.02014029e-01 1.95184469e-01 -8.73011649e-01 -9.50843394e-01 6.88900530e-01 -6.23485334e-02 -1.11802363e+00 -5.94156459e-02 -9.50068831e-02 1.04218006e+00 -5.10614634e-01 -1.43361300e-01 -7.51815498e-01 7.41529092e-02 -8.66957545e-01 9.32109416e-01 2.32757375e-01 4.88511741e-01 -6.73493862e-01 6.66973054e-01 2.69382566e-01 -1.56495512e+00 4.17649508e-01 -3.82882506e-01 2.68196374e-01 2.38729909e-01 9.12009656e-01 -6.64846063e-01 4.87103283e-01 6.87564552e-01 3.73218805e-01 -4.00900930e-01 1.28545856e+00 -3.12095970e-01 5.64158440e-01 -5.58070421e-01 1.44450352e-01 5.04552424e-01 -8.00612628e-01 4.29256648e-01 1.49620187e+00 1.00882679e-01 1.32787049e-01 7.48899281e-02 1.32968104e+00 2.61074513e-01 4.00517017e-01 6.86910981e-03 1.55958742e-01 3.00607622e-01 1.45220792e+00 -1.88830721e+00 -5.50671756e-01 5.10498732e-02 1.16599143e+00 4.68965843e-02 3.34318757e-01 -4.66419876e-01 -2.62132823e-01 2.83853024e-01 1.42221719e-01 4.35275763e-01 -4.14769113e-01 -8.76389563e-01 -4.67627048e-01 2.98440140e-02 -2.60558248e-01 3.55853409e-01 -4.81396914e-01 -6.83555186e-01 8.28590572e-01 1.38055738e-02 -1.05079913e+00 3.14866126e-01 -2.69013703e-01 -8.56870830e-01 6.05757058e-01 -1.28745186e+00 -1.26327169e+00 -4.62135583e-01 4.39785838e-01 6.75443649e-01 4.95732814e-01 4.55012769e-01 -2.64093205e-02 -5.11876941e-01 1.03073880e-01 -2.29741167e-02 1.40854880e-01 2.39697009e-01 -1.43903184e+00 4.69878614e-02 1.46325386e+00 1.53346658e-01 5.39370835e-01 9.63046432e-01 -8.79469812e-01 -4.81899798e-01 -9.41286385e-01 7.77141869e-01 1.47067294e-01 2.28699163e-01 -4.93616238e-02 -7.98195660e-01 3.47645253e-01 4.29783702e-01 -3.11891466e-01 4.44470167e-01 -2.96148509e-01 4.62403633e-02 -9.83663946e-02 -1.19775081e+00 4.73887563e-01 5.18344760e-01 7.47720376e-02 -5.16487062e-01 4.52545993e-02 1.65394500e-01 -2.88291392e-03 -8.74171138e-01 1.90966338e-01 3.85287374e-01 -1.44278836e+00 9.29514527e-01 1.55150190e-01 3.29855114e-01 -6.43708169e-01 3.38401794e-01 -1.07646620e+00 -3.51309836e-01 -7.10464776e-01 4.60942388e-01 1.35714734e+00 2.77280331e-01 -1.46070912e-01 1.16693270e+00 5.02382398e-01 -2.27474961e-02 -5.47329545e-01 -9.41621780e-01 -5.60555875e-01 -6.83020428e-02 -2.53201902e-01 2.59663701e-01 6.90266728e-01 6.49566129e-02 -2.04587150e-02 1.76446855e-01 2.34065935e-01 1.23099852e+00 3.83592427e-01 4.45725203e-01 -1.36605680e+00 -8.71625468e-02 -7.97293425e-01 -4.53762650e-01 -9.70613241e-01 -4.04407650e-01 -5.25080264e-01 1.49445057e-01 -1.74251783e+00 4.94843781e-01 -5.78268290e-01 -1.26797929e-01 3.13442588e-01 -3.71654063e-01 5.33733487e-01 1.67117834e-01 3.92099828e-01 -5.70394814e-01 -1.28782645e-01 1.18668175e+00 -8.96303076e-03 -4.72332507e-01 1.40521675e-02 -2.55981535e-01 9.70420539e-01 8.23705792e-01 -3.21481854e-01 -3.15542698e-01 -2.93130964e-01 -1.29355535e-01 -2.07734898e-01 1.36589989e-01 -1.12742484e+00 5.46099186e-01 5.85360639e-02 1.27594993e-01 -8.93605471e-01 8.26286525e-02 -8.70196462e-01 1.83141023e-01 6.43255830e-01 -4.38434165e-03 -2.96777517e-01 5.10914810e-02 3.95168811e-01 -3.00240815e-01 -5.99825978e-01 1.20366943e+00 -2.92436272e-01 -6.67845130e-01 -1.06658757e-01 -3.83026361e-01 -2.67247945e-01 1.61649477e+00 -7.57947803e-01 3.98937352e-02 1.60664190e-02 -7.67910838e-01 9.68955606e-02 7.46465802e-01 -2.27891162e-01 5.54083526e-01 -7.74891973e-01 -6.35203779e-01 5.10727018e-02 -3.14668864e-01 2.86794722e-01 5.31354547e-02 9.68055904e-01 -1.23619747e+00 2.80630052e-01 -3.45030099e-01 -7.85069585e-01 -1.87643445e+00 6.18905663e-01 3.05354983e-01 -3.08672965e-01 -5.45398414e-01 1.05567276e+00 2.40602389e-01 4.02485400e-01 2.84921348e-01 -8.38078678e-01 -3.64594966e-01 2.68243879e-01 1.87911198e-01 7.43805587e-01 -7.56376758e-02 -6.28936112e-01 -2.62734413e-01 1.11096716e+00 1.99737981e-01 -2.12486506e-01 1.16380990e+00 -2.36835435e-01 -5.62040091e-01 1.53629065e-01 9.08665121e-01 -1.07630782e-01 -1.30077565e+00 8.33164901e-02 3.62929672e-01 -4.47148442e-01 2.56049335e-01 -3.21925223e-01 -1.28863275e+00 7.58046389e-01 7.40376890e-01 4.12106425e-01 1.21832633e+00 -9.97252688e-02 6.31620049e-01 -2.20034644e-01 2.90557086e-01 -1.36813831e+00 -2.68630683e-02 -1.30251601e-01 4.86172110e-01 -9.50529456e-01 2.94078618e-01 -1.11684966e+00 -4.49742973e-01 1.33074057e+00 2.73899436e-01 -3.66876036e-01 7.78611958e-01 3.80425930e-01 -1.64269172e-02 -1.60092160e-01 5.44720590e-02 -3.30380559e-01 1.90150127e-01 3.41250718e-01 4.34158474e-01 9.06274840e-02 -9.04152095e-01 2.92312264e-01 1.29198760e-01 -2.26590764e-02 3.73807967e-01 7.34122217e-01 -8.83391261e-01 -9.96669888e-01 -8.63877535e-01 -4.99162972e-02 -4.70907271e-01 6.70839623e-02 -4.33409691e-01 4.88480002e-01 5.07629633e-01 1.06294060e+00 2.90150493e-01 1.10424787e-01 5.86409532e-02 -1.41426221e-01 4.48134184e-01 -5.04323602e-01 -5.62123656e-01 6.35734141e-01 -1.90344766e-01 -4.99079347e-01 -7.99210548e-01 -8.72997224e-01 -1.49283636e+00 2.63245422e-02 -4.66374218e-01 3.39147389e-01 5.79193592e-01 7.45710611e-01 -5.13905920e-02 3.70420247e-01 3.13142180e-01 -1.04919910e+00 8.86602700e-03 -8.06815028e-01 -8.13643098e-01 5.17367244e-01 -3.98934027e-03 -1.77745149e-01 -1.29827857e-01 7.14944184e-01]

[9.265283584594727, -0.37765565514564514]

d270d3c8-5024-4193-879b-897df97f4f59

second-order-neural-ode-optimizer

2109.14158

https://arxiv.org/abs/2109.14158v2

https://arxiv.org/pdf/2109.14158v2.pdf

Second-Order Neural ODE Optimizer

We propose a novel second-order optimization framework for training the emerging deep continuous-time models, specifically the Neural Ordinary Differential Equations (Neural ODEs). Since their training already involves expensive gradient computation by solving a backward ODE, deriving efficient second-order methods becomes highly nontrivial. Nevertheless, inspired by the recent Optimal Control (OC) interpretation of training deep networks, we show that a specific continuous-time OC methodology, called Differential Programming, can be adopted to derive backward ODEs for higher-order derivatives at the same O(1) memory cost. We further explore a low-rank representation of the second-order derivatives and show that it leads to efficient preconditioned updates with the aid of Kronecker-based factorization. The resulting method -- named SNOpt -- converges much faster than first-order baselines in wall-clock time, and the improvement remains consistent across various applications, e.g. image classification, generative flow, and time-series prediction. Our framework also enables direct architecture optimization, such as the integration time of Neural ODEs, with second-order feedback policies, strengthening the OC perspective as a principled tool of analyzing optimization in deep learning. Our code is available at https://github.com/ghliu/snopt.

['Evangelos A. Theodorou', 'Tianrong Chen', 'Guan-Horng Liu']

2021-09-29

http://proceedings.neurips.cc/paper/2021/hash/d4c2e4a3297fe25a71d030b67eb83bfc-Abstract.html

http://proceedings.neurips.cc/paper/2021/file/d4c2e4a3297fe25a71d030b67eb83bfc-Paper.pdf

neurips-2021-12

['time-series-prediction']

['time-series']

[-2.20468864e-01 -6.50673499e-03 -7.18880147e-02 1.14536304e-02 -4.41687018e-01 -6.15233898e-01 5.00354230e-01 -3.28001559e-01 -5.55537105e-01 8.27548265e-01 1.10182986e-01 -6.94126964e-01 -2.82234579e-01 -4.10738438e-01 -9.74474609e-01 -7.59877026e-01 -1.62959158e-01 1.07982397e-01 -2.93441355e-01 -2.83374399e-01 2.92096645e-01 8.30203593e-01 -1.03954232e+00 -1.71174988e-01 9.51185405e-01 1.18963528e+00 -1.09623238e-01 9.90335286e-01 1.87101632e-01 8.84151280e-01 -7.86567777e-02 -3.02359760e-01 5.97450435e-01 -5.02445757e-01 -9.42941487e-01 -8.77691805e-02 4.84199673e-01 -4.80260193e-01 -6.48664832e-01 7.30937183e-01 4.40110505e-01 5.67119718e-01 5.15124798e-01 -1.11871326e+00 -6.52857363e-01 1.58463836e-01 -1.11517601e-01 2.23018214e-01 -2.34385699e-01 2.72162586e-01 1.16305327e+00 -1.02581120e+00 7.24563122e-01 1.15416682e+00 9.52353537e-01 6.80632710e-01 -1.46086156e+00 -1.42355174e-01 3.29779744e-01 3.66625041e-02 -1.07075703e+00 -4.41444457e-01 7.31101513e-01 -7.36201465e-01 8.65199089e-01 3.09938192e-01 6.88354731e-01 9.24207866e-01 3.00319225e-01 9.35613751e-01 8.65116656e-01 -2.29547143e-01 1.49156511e-01 -7.22182840e-02 6.09972179e-02 9.50312078e-01 -4.06760955e-03 2.89191961e-01 -4.59998250e-01 -7.10563511e-02 1.00282800e+00 -8.94382596e-02 -5.01900136e-01 -2.41817802e-01 -1.15890384e+00 1.12408614e+00 5.97348094e-01 3.08415085e-01 -4.28626627e-01 6.58221543e-01 4.47285920e-01 3.43702257e-01 7.15587854e-01 6.99200928e-01 -7.03095317e-01 -2.90185958e-01 -9.18615580e-01 4.60800290e-01 1.15745819e+00 5.03581941e-01 7.41819382e-01 4.06958342e-01 -1.69334590e-01 5.62888443e-01 4.30428386e-01 2.05627024e-01 2.81312764e-01 -1.70233917e+00 1.38602734e-01 1.06690615e-01 4.78462465e-02 -9.67608571e-01 -4.86656964e-01 -9.01810527e-01 -1.23267543e+00 3.80926609e-01 4.95408088e-01 -5.11495650e-01 -5.18021643e-01 1.88902044e+00 4.21742171e-01 3.16241682e-01 -1.04990035e-01 1.03980255e+00 7.70825371e-02 6.84257388e-01 -4.60817397e-01 -3.82328600e-01 9.51862335e-01 -1.10237157e+00 -5.34841478e-01 1.49934679e-01 9.53915060e-01 -4.21142995e-01 8.56297553e-01 4.86124337e-01 -1.13720679e+00 -2.60832965e-01 -8.47103834e-01 -4.93134260e-01 -2.24038467e-01 2.58107156e-01 8.31387043e-01 2.42673144e-01 -1.55666733e+00 1.54613388e+00 -1.13894343e+00 1.19910859e-01 4.09497678e-01 2.75809228e-01 -1.53323531e-01 4.97819304e-01 -1.12636852e+00 6.90391004e-01 -1.89535990e-01 5.96491039e-01 -8.39412272e-01 -1.36668515e+00 -6.72994196e-01 -4.11063060e-02 2.18390942e-01 -1.09792626e+00 1.33270812e+00 -8.71793509e-01 -2.04310942e+00 4.58916306e-01 -4.20068145e-01 -7.95322299e-01 9.73856866e-01 -6.02011740e-01 3.30943793e-01 2.06887320e-01 -5.64793088e-02 4.73015666e-01 9.54266965e-01 -7.09119320e-01 -9.00023431e-02 -1.65654749e-01 3.02246809e-01 8.71837363e-02 -4.08549577e-01 -3.63630146e-01 -1.17343485e-01 -6.72411084e-01 -1.35925442e-01 -1.17528021e+00 -6.52863026e-01 5.88207841e-01 -2.56040126e-01 6.61574975e-02 4.21963483e-01 -8.62878025e-01 1.34022784e+00 -1.86950994e+00 6.13407910e-01 -3.27061228e-02 5.11417866e-01 3.24423611e-01 1.14497490e-01 4.14323568e-01 -1.78418413e-01 1.22041807e-01 -4.11133766e-01 -7.26403892e-01 1.44066826e-01 3.41439009e-01 -5.50224602e-01 5.16383648e-01 3.69364649e-01 1.09829664e+00 -1.04212344e+00 -2.06265971e-01 9.76384804e-02 8.55632722e-01 -1.07544541e+00 -3.64523269e-02 -7.17582703e-02 6.96688294e-01 -4.52275842e-01 -4.33276929e-02 4.04500276e-01 -5.35037696e-01 -5.65181263e-02 -4.49017733e-01 -5.07432282e-01 3.37281317e-01 -1.05127084e+00 1.82745731e+00 -7.96614110e-01 1.11641228e+00 2.96053559e-01 -1.43135953e+00 2.38058224e-01 1.81076050e-01 7.91757762e-01 -3.47356558e-01 -1.94809716e-02 4.69533652e-01 -1.26148567e-01 -3.68442386e-01 2.87650585e-01 5.98202944e-02 4.60954159e-01 9.89769548e-02 2.24777132e-01 -2.54412331e-02 5.47705114e-01 4.08673584e-01 9.71817315e-01 4.43842560e-01 -1.33536711e-01 -6.09964430e-01 7.15197146e-01 -1.28619850e-01 3.59357387e-01 6.39990389e-01 7.49862716e-02 4.13250953e-01 8.85254383e-01 -4.82204229e-01 -9.79677081e-01 -6.73058331e-01 -1.57013714e-01 1.08737516e+00 -5.49758792e-01 -3.91405612e-01 -7.49484241e-01 -2.94657588e-01 2.00017706e-01 2.30391040e-01 -7.49796093e-01 -1.18806139e-01 -8.28110933e-01 -6.28546059e-01 4.70997721e-01 4.09571201e-01 4.56027210e-01 -4.83731896e-01 -6.03081107e-01 3.47286224e-01 1.24119252e-01 -8.97565246e-01 -7.62581825e-01 2.37119704e-01 -1.21301317e+00 -7.41244316e-01 -1.09400225e+00 -2.93242514e-01 4.96354580e-01 -2.14756519e-01 9.67901587e-01 1.73070788e-01 -2.47711807e-01 5.52276492e-01 1.23843893e-01 7.05954656e-02 -1.52847931e-01 1.37528151e-01 9.87853408e-02 2.76831418e-01 -6.76615834e-01 -8.75927687e-01 -8.64765942e-01 -4.17200290e-02 -7.24071205e-01 1.12195760e-01 4.90917772e-01 1.09427702e+00 5.26941836e-01 -5.05537748e-01 1.63452342e-01 -8.10389102e-01 7.04133213e-01 -2.86201656e-01 -9.17353153e-01 -1.39134414e-02 -7.27738857e-01 7.39966631e-01 1.05508983e+00 -4.63084400e-01 -8.18229020e-01 2.00248379e-02 -3.27144265e-01 -8.56784463e-01 5.09522319e-01 6.95691347e-01 5.17921984e-01 -4.60823119e-01 6.86734498e-01 1.31823555e-01 2.68081903e-01 -6.08917952e-01 6.38683200e-01 -4.68624458e-02 5.68023145e-01 -7.64640629e-01 6.81589305e-01 6.06457770e-01 6.29975557e-01 -8.41738224e-01 -1.00977111e+00 -3.63146037e-01 -5.15953243e-01 -1.81054026e-01 7.57362425e-01 -6.93197608e-01 -1.12164783e+00 6.51466906e-01 -1.37052941e+00 -8.27374220e-01 -3.74224007e-01 4.31445271e-01 -6.98238611e-01 2.62198001e-01 -1.12316012e+00 -8.99341762e-01 -3.70351315e-01 -1.10960066e+00 9.12145913e-01 7.59029388e-02 5.87996915e-02 -1.40580356e+00 1.17067747e-01 -1.09379292e-01 4.97385055e-01 3.10726553e-01 5.96888363e-01 -1.37866050e-01 -6.62988186e-01 2.39003822e-01 -6.14101663e-02 6.80591285e-01 -4.86188948e-01 2.93784171e-01 -7.68575966e-01 -1.26194641e-01 3.46078515e-01 3.65789235e-02 1.05091536e+00 6.24970019e-01 1.04955554e+00 -7.47464359e-01 -3.26395743e-02 1.26493251e+00 1.43716347e+00 -2.24422514e-01 1.77172124e-01 -9.13974717e-02 9.33866680e-01 4.32730436e-01 4.68054749e-02 5.11952937e-01 2.18295485e-01 5.30557871e-01 3.64266992e-01 -3.99915548e-03 -6.70884177e-02 -1.81308500e-02 4.64058399e-01 1.11100101e+00 -4.79328126e-01 2.56457061e-01 -7.43108034e-01 3.27289134e-01 -2.04628444e+00 -8.62453759e-01 -4.47612077e-01 1.93088043e+00 9.21655178e-01 -2.23063096e-01 6.12102747e-02 1.03731029e-01 1.61806688e-01 2.04816952e-01 -9.21698749e-01 -4.80222493e-01 9.95442346e-02 4.81037706e-01 6.96177423e-01 9.93130565e-01 -1.05615759e+00 5.68652749e-01 6.07500362e+00 6.04985416e-01 -1.45376039e+00 3.53932440e-01 7.61168480e-01 -2.20582515e-01 -2.64489561e-01 6.61692992e-02 -6.24146104e-01 3.62595052e-01 9.80738461e-01 -2.27200598e-01 8.65886390e-01 8.91078174e-01 4.58371609e-01 1.97577953e-01 -1.02987576e+00 9.96976137e-01 -3.64988327e-01 -1.73109090e+00 -2.93560535e-01 3.68479788e-01 9.65249777e-01 3.12766671e-01 2.20061451e-01 1.79242760e-01 1.22588374e-01 -7.57072270e-01 7.29585528e-01 7.33945251e-01 4.95489240e-01 -4.55541164e-01 9.21009034e-02 2.58429050e-01 -9.94247079e-01 -1.02509975e-01 -2.03357145e-01 -2.86541134e-01 3.21097970e-01 1.00168312e+00 -6.50433376e-02 5.41275620e-01 3.49323004e-01 1.12900865e+00 -2.25423709e-01 8.23453546e-01 -8.54131132e-02 4.78059888e-01 -5.45865536e-01 -5.18293232e-02 5.17820239e-01 -7.48106778e-01 6.67640865e-01 1.00757527e+00 3.91430348e-01 -1.38366781e-02 -3.26241970e-01 1.08178508e+00 -3.93719710e-02 -8.51735845e-02 -3.04779291e-01 -1.98704988e-01 -1.96599856e-01 1.31979752e+00 -3.65114659e-01 -1.80239409e-01 -2.13396654e-01 1.12049317e+00 4.91418034e-01 7.51888633e-01 -8.62245381e-01 -4.41046476e-01 1.06959200e+00 -2.30093390e-01 3.04843038e-01 -6.93476856e-01 -3.17981690e-01 -1.52299714e+00 3.37140888e-01 -6.01156831e-01 6.04737177e-02 -4.25148964e-01 -9.61539686e-01 4.07016486e-01 -2.07923740e-01 -9.63253736e-01 -5.65278351e-01 -1.07330966e+00 -5.50226510e-01 8.20226312e-01 -1.55680573e+00 -5.95536351e-01 1.29427835e-01 5.74120402e-01 1.97339818e-01 2.64683753e-01 4.43883538e-01 5.75945258e-01 -7.28373110e-01 4.41519797e-01 5.06852984e-01 8.19948036e-03 1.52991191e-01 -1.35702610e+00 2.31117636e-01 1.03397202e+00 1.71665609e-01 7.18759418e-01 6.99175477e-01 -1.69622198e-01 -1.67849350e+00 -7.93886065e-01 7.81461120e-01 -2.64265299e-01 1.12433660e+00 -2.71343261e-01 -8.57274950e-01 5.53554237e-01 4.51780520e-02 3.59742939e-01 2.10883319e-01 -9.40661430e-02 -7.59010911e-02 -2.21558303e-01 -5.17200470e-01 7.30784535e-01 1.25891387e+00 -6.66857898e-01 1.80258378e-01 6.24477386e-01 8.18335712e-01 -5.77857912e-01 -9.38053966e-01 1.72787920e-01 6.76207542e-01 -9.44787800e-01 1.04606283e+00 -7.31976688e-01 7.05070794e-01 -3.97164980e-03 1.41065657e-01 -1.31754363e+00 -1.03943318e-01 -1.54204798e+00 -7.86456347e-01 6.71447277e-01 3.92567366e-01 -1.07653821e+00 5.22191823e-01 7.73077548e-01 -5.11191428e-01 -1.43983757e+00 -8.74126554e-01 -7.65690506e-01 3.41287643e-01 -5.02856255e-01 -6.26108646e-02 7.75297523e-01 -2.80866951e-01 6.98939934e-02 -4.70798194e-01 -5.24337925e-02 3.99179280e-01 -2.16624439e-01 5.24306893e-01 -1.02771521e+00 -6.82555318e-01 -9.29534137e-01 -1.40536174e-01 -1.64152694e+00 3.45817447e-01 -9.22248065e-01 -1.05654970e-01 -1.27956975e+00 -6.23677135e-01 -1.68273687e-01 -1.81790575e-01 2.06251562e-01 -5.79578569e-03 -6.56708181e-02 3.98375094e-01 4.29544933e-02 -1.57783180e-01 6.97159410e-01 1.51648962e+00 2.71162033e-01 -2.03432858e-01 6.77128555e-03 -4.75314289e-01 6.87988400e-01 6.55332744e-01 -2.83163369e-01 -4.48925048e-01 -6.71126842e-01 5.22890389e-01 1.23800047e-01 6.72842920e-01 -1.00331521e+00 4.38443512e-01 -6.56131357e-02 -5.49225090e-03 -1.00061052e-01 3.21616530e-01 -4.06754613e-01 -6.18463717e-02 7.96313524e-01 -5.12073398e-01 1.79895326e-01 7.34513551e-02 2.87198037e-01 -1.33148640e-01 -3.02926153e-01 6.49926364e-01 -1.41236186e-01 -3.43355268e-01 6.05830014e-01 -2.78601885e-01 3.84148717e-01 4.23857331e-01 2.12921202e-01 -1.32231727e-01 -4.14065629e-01 -9.36868072e-01 1.43802956e-01 -1.00866575e-02 -7.40101263e-02 3.18827748e-01 -1.42349792e+00 -5.32490253e-01 8.15395266e-02 -8.22798610e-01 8.61290023e-02 2.75240093e-01 1.51518643e+00 -7.51404464e-01 4.86856461e-01 1.90620840e-01 -5.39626539e-01 -4.37851608e-01 1.45922750e-01 7.60304153e-01 -6.89232767e-01 -3.90048474e-01 1.06899345e+00 3.26662749e-01 -2.85232544e-01 -2.74161845e-02 -5.62704206e-01 3.33865851e-01 1.22079216e-01 2.21386731e-01 6.06379926e-01 9.24701244e-03 -3.43928546e-01 -1.75031677e-01 6.45132720e-01 3.06389272e-01 -4.78845477e-01 1.40268517e+00 -3.00394595e-02 -3.25650901e-01 5.41384995e-01 1.90303802e+00 -2.92811424e-01 -1.75337124e+00 -5.82288466e-02 -1.35684043e-01 -1.77471071e-01 1.85049638e-01 -2.50101119e-01 -1.45731461e+00 1.20181656e+00 1.23964258e-01 1.66979685e-01 1.01358128e+00 -5.58392406e-01 7.44644165e-01 6.86152518e-01 5.59703261e-02 -9.46862340e-01 4.96557355e-02 9.05663073e-01 1.12932432e+00 -8.48285735e-01 -6.47856072e-02 -9.57944244e-02 -2.53755271e-01 1.36396229e+00 2.55261362e-01 -6.60596132e-01 1.01292217e+00 8.22908878e-02 -1.69210896e-01 2.05082968e-01 -9.61552322e-01 -1.12543358e-02 5.63820660e-01 2.12601691e-01 4.87531602e-01 -2.73519248e-01 -5.31512797e-01 8.87518004e-02 -2.32263952e-01 1.08382456e-01 4.83742774e-01 5.96525252e-01 2.37273544e-01 -1.01184571e+00 2.18354240e-01 3.84457558e-01 -5.91937363e-01 -2.22441539e-01 2.50924267e-02 7.11649537e-01 -3.32617998e-01 4.66888547e-01 7.25458656e-03 1.66256540e-02 3.89645435e-02 1.79855153e-01 5.05060196e-01 -1.44313514e-01 -5.05992949e-01 -4.56696786e-02 -1.52959615e-01 -1.00874734e+00 -4.50511247e-01 -6.36946499e-01 -1.11241710e+00 -4.65981990e-01 -4.67380621e-02 5.06023541e-02 7.29218841e-01 9.49473381e-01 6.45416081e-01 4.92643416e-01 5.93656301e-01 -1.29972506e+00 -9.13271904e-01 -6.09413564e-01 -1.93837747e-01 2.92391717e-01 9.75324869e-01 -5.02547503e-01 -6.10830486e-01 1.76725276e-02]

[6.874040603637695, 3.5666449069976807]

d59f5fb7-51d0-4c47-bcb9-aea6ed46e4d9

a-coupled-approach-to-model-the-effect-of

2102.05594

https://arxiv.org/abs/2102.05594v1

https://arxiv.org/pdf/2102.05594v1.pdf

A coupled approach to model the effect of wear on the dynamics of the shrouded bladed disk

This paper deals with modelling the effect of wear on the dynamics of the shrouded bladed disk with frictional contacts at the shrouds and the contact interface evolution. Prediction of fretting wear commonly occurring at the contacts of turbomachinery components, and its impact on the dynamics is increasingly researched due to the components subjected to their structural limits for performance and operating at high loading conditions. Over a lifetime, the fretting wear at these contacts could alter the global dynamic response of these bladed disks from the designed operating point and could lead to high vibration amplitudes. This study implements a coupled static/dynamic harmonic balance method (HBM) with wear energy approach and an adaptive wear logic to study the impact on the steady-state nonlinear dynamic response. Firstly, the methodology is applied to a cantilever beam with a contact patch and then to a shrouded bladed disk with shroud contacts using cyclic symmetry boundary conditions. The novelty of the paper is to show the effect of wear on the dynamics with the changing contact pre-load using a coupled approach. A wear acceleration $v_{w,max}$ parameter is defined, and the impact of the choice of this parameter is discussed in detail on the computation time and the accuracy of results. The test cases demonstrate the impact of wear on the dynamic nonlinear response curves and the contact interface evolution for various scenarios.

['Stefano Zucca', 'Daniele Botto', 'Lakshminarayana Reddy Tamatam']

2021-02-10

['cantilever-beam']

['miscellaneous']

[-4.48101044e-01 -8.02442729e-02 5.80069363e-01 6.08638465e-01 3.25360507e-01 -1.51695356e-01 4.33669835e-01 -1.99284583e-01 -1.33475721e-01 5.62612534e-01 -2.02934355e-01 1.17808349e-01 -7.31310129e-01 -3.64683747e-01 -3.44544232e-01 -9.57007289e-01 -2.16027245e-01 4.66466069e-01 6.48420990e-01 -6.63454950e-01 3.00895661e-01 6.86494887e-01 -1.96218836e+00 -2.08194777e-01 4.01782930e-01 7.54573226e-01 5.26509464e-01 4.11835819e-01 5.11739671e-01 1.30754873e-01 -8.64354908e-01 2.80086398e-01 -5.05385883e-02 -2.53569692e-01 -6.75092578e-01 3.08454365e-01 -5.88141859e-01 -2.70506918e-01 3.24793875e-01 4.82124686e-01 4.63574469e-01 5.57759345e-01 9.49985802e-01 -6.54706955e-01 -2.23283330e-03 -1.40516330e-02 -4.26411480e-01 3.29429090e-01 3.43513697e-01 -6.79049566e-02 4.02504742e-01 -1.22544658e+00 6.74733579e-01 8.45361114e-01 8.52163553e-01 3.17679226e-01 -8.09784412e-01 5.68668023e-02 -7.65597165e-01 4.56915408e-01 -1.40047228e+00 -1.76317677e-01 8.94355476e-01 -8.39930475e-01 9.48641479e-01 4.99743819e-01 7.37121463e-01 3.32728058e-01 1.05877757e+00 -4.45791095e-01 1.03526747e+00 -6.02849126e-01 4.45864469e-01 2.34595761e-02 1.59870341e-01 2.59264201e-01 4.18227941e-01 3.63641411e-01 -2.42756575e-01 2.18386836e-02 5.04439354e-01 -4.17582899e-01 -4.64145571e-01 -7.99899399e-02 -1.20979421e-01 3.82173866e-01 -2.09276363e-01 6.64712250e-01 -3.12146753e-01 -3.75282437e-01 5.84463537e-01 3.64567310e-01 4.50850189e-01 4.68686759e-01 -4.08357263e-01 -4.94771779e-01 -7.47195959e-01 3.70439470e-01 1.14792681e+00 -9.79040936e-02 8.28482360e-02 1.58327408e-02 1.80954456e-01 7.65064180e-01 2.43922263e-01 4.93908137e-01 2.89229006e-01 -5.14821351e-01 -2.77662307e-01 3.34294677e-01 3.16010535e-01 -8.14453542e-01 -5.18834293e-01 -2.22817972e-01 -2.07631692e-01 7.57395267e-01 7.50089139e-02 -5.86987317e-01 -2.43870795e-01 7.90609360e-01 7.57107139e-01 -4.23181415e-01 -6.79115131e-02 1.37848878e+00 3.77077788e-01 6.15644872e-01 -4.13088083e-01 -6.23237073e-01 1.23256230e+00 -2.84881145e-01 -1.11259568e+00 3.77535373e-01 3.45857650e-01 -1.18862307e+00 1.13767374e+00 4.92348582e-01 -1.37878549e+00 -5.94123781e-01 -1.35649729e+00 4.45939809e-01 -6.01563044e-02 2.18034670e-01 -6.15802228e-01 4.26503390e-01 -6.75043166e-01 1.01176894e+00 -7.94447064e-01 -1.89058870e-01 -6.56770706e-01 2.49051541e-01 3.38027067e-02 3.34318221e-01 -1.14367580e+00 1.66520190e+00 -2.76249051e-01 3.85640651e-01 -4.69947904e-01 -8.24142635e-01 -1.70808300e-01 -5.28871231e-02 3.59257996e-01 -3.00357163e-01 1.12474620e+00 -2.41963774e-01 -1.99560058e+00 2.42189214e-01 1.56226590e-01 -4.51430306e-02 9.60965574e-01 -6.18758857e-01 -3.14776272e-01 1.89027771e-01 -2.18200684e-01 -7.72482038e-01 8.34613681e-01 -1.35899198e+00 3.13272387e-01 -3.72186936e-02 -4.32443023e-01 1.93740040e-01 -1.26611754e-01 5.96932508e-02 4.03342783e-01 -5.74146390e-01 -3.68379295e-01 -1.03933525e+00 3.33160996e-01 -5.93404651e-01 -5.31335212e-02 -2.71559596e-01 1.56903172e+00 -9.43998039e-01 1.16276598e+00 -2.13518381e+00 4.47132498e-01 2.83303112e-01 -4.15188044e-01 2.01169878e-01 8.57997835e-01 1.34905708e+00 1.35069996e-01 -4.25862193e-01 -4.37855959e-01 5.77956550e-02 -4.75993365e-01 2.41784230e-01 -1.02997325e-01 6.37131453e-01 1.04684964e-01 -1.12352416e-01 -3.72444451e-01 -1.24287037e-02 -7.60987774e-02 4.50079113e-01 -2.79918194e-01 5.36526442e-01 2.79715866e-01 4.68980744e-02 -2.34905314e-02 2.15475887e-01 4.49337065e-01 6.75216913e-01 2.54373066e-02 -4.57435668e-01 -8.84807467e-01 -2.60146707e-01 -1.15572190e+00 5.83597660e-01 -5.83554685e-01 3.13489765e-01 7.21761584e-01 -7.11542785e-01 1.40384114e+00 6.81238890e-01 4.10999596e-01 -4.16568488e-01 1.70394510e-01 7.72294939e-01 3.28309506e-01 -8.92619729e-01 4.52808142e-01 -4.99455780e-01 4.00679916e-01 2.70543098e-01 -4.03738290e-01 -6.58172488e-01 -1.06609995e-02 -2.91099101e-01 6.55014873e-01 2.76624024e-01 -1.88635647e-01 -1.13197708e+00 7.11755395e-01 -4.86904085e-02 3.17483306e-01 -3.71322304e-01 5.31649292e-01 3.94430101e-01 6.60723746e-01 -1.12671249e-01 -1.13884580e+00 -4.26659882e-01 -6.19258165e-01 4.49998319e-01 3.66349995e-01 9.51519981e-02 -8.92264605e-01 4.71934587e-01 2.45711446e-01 7.32825935e-01 -4.97351408e-01 -7.16143548e-01 -9.48164523e-01 -9.61935341e-01 -3.73855419e-03 1.77932441e-01 2.15630308e-02 -1.05295718e+00 -1.46232378e+00 2.46139437e-01 5.72342157e-01 -7.56568253e-01 -2.09879398e-01 2.52643973e-01 -1.01667464e+00 -1.54379356e+00 -3.33086699e-01 -4.18200642e-01 4.25548226e-01 -2.83405215e-01 8.21627319e-01 6.52728856e-01 -6.84676826e-01 5.32044590e-01 -5.63761652e-01 -4.86938208e-01 -8.70111465e-01 -2.57548064e-01 4.72953111e-01 -1.56711474e-01 -5.70844948e-01 -2.64540970e-01 -5.88153541e-01 9.85653162e-01 -9.80203748e-01 -3.48916560e-01 4.24697734e-02 7.98696637e-01 2.93701112e-01 5.57930768e-01 8.54740739e-01 -6.09758556e-01 1.15284574e+00 -5.03586411e-01 -3.46786320e-01 -2.53204405e-01 -8.06385219e-01 -3.76274705e-01 8.06175947e-01 -3.82300764e-01 -1.35848689e+00 -7.75407910e-01 -3.45114768e-02 -4.84686941e-01 1.12930506e-01 5.96738160e-01 -2.19618812e-01 9.34955403e-02 4.54188347e-01 -2.74422348e-01 7.38364816e-01 -9.40255344e-01 -3.40915263e-01 3.85506541e-01 3.58235151e-01 -7.26542175e-01 7.09509611e-01 1.14294238e-01 5.17719269e-01 -1.36861372e+00 1.67103380e-01 -7.34184533e-02 -4.97589946e-01 -6.95901632e-01 5.91023624e-01 -5.06746843e-02 -8.78552377e-01 8.76419902e-01 -7.88189590e-01 -6.06420338e-01 -4.89850342e-01 3.52746457e-01 -3.13705832e-01 5.77400684e-01 -8.29969585e-01 -9.85058844e-01 -5.92715859e-01 -1.26454127e+00 4.68737245e-01 1.09688386e-01 -3.93550783e-01 -1.06142306e+00 1.07143447e-01 -3.42159271e-02 5.36092758e-01 6.68931723e-01 1.15114391e+00 1.74892358e-02 3.35931271e-01 -1.79885939e-01 6.49645090e-01 4.45486158e-01 1.72429159e-01 4.95566845e-01 -4.14980978e-01 -6.73634529e-01 9.77601647e-01 8.67339689e-03 2.93741643e-01 3.09389919e-01 2.76769310e-01 -1.78453252e-01 -2.85826653e-01 3.41357142e-02 1.70962977e+00 3.86267632e-01 1.79161325e-01 6.37468025e-02 2.08435297e-01 1.01378608e+00 1.01859403e+00 6.72998130e-01 -6.21035755e-01 1.14939702e+00 6.90952718e-01 -2.01426875e-02 -9.82756168e-02 5.71162820e-01 5.51738679e-01 1.18110383e+00 -6.05836093e-01 3.43748666e-02 -7.11813152e-01 4.60753798e-01 -9.04049516e-01 -5.28843343e-01 -5.66043615e-01 2.47850657e+00 6.91135526e-01 1.21516004e-01 4.55018245e-02 6.80886209e-01 6.90791547e-01 -2.89094508e-01 -1.74559385e-01 -1.26676583e+00 1.01452330e-02 3.13149124e-01 2.33172745e-01 8.44570279e-01 -8.08107257e-02 -8.23435038e-02 5.73292303e+00 3.57153922e-01 -1.33876753e+00 2.59167939e-01 -7.70223588e-02 -2.54337460e-01 -7.77103603e-02 -8.74205306e-02 -5.71777344e-01 8.95705879e-01 1.00165176e+00 -1.33419484e-01 2.40820780e-01 2.08372906e-01 6.52698874e-01 -6.44444168e-01 -4.80253965e-01 8.23568106e-02 -4.04848680e-02 -7.87894309e-01 -4.06836003e-01 8.46889094e-02 3.87302458e-01 -5.60808897e-01 -2.50433713e-01 -4.10896182e-01 -7.69392431e-01 -5.50193548e-01 9.91063774e-01 8.70333433e-01 9.56055880e-01 -8.37357879e-01 1.03415358e+00 4.10398185e-01 -9.58531022e-01 -2.48656332e-01 -1.30046040e-01 -4.49370742e-01 6.96136117e-01 8.71812284e-01 -8.54739785e-01 7.05924749e-01 8.03354025e-01 4.55239452e-02 -1.83020040e-01 6.83990777e-01 3.28073293e-01 7.56108940e-01 -4.36420441e-01 -2.54919559e-01 -2.04739600e-01 -7.28925228e-01 1.04122353e+00 5.71568549e-01 4.61761653e-01 2.17307568e-01 -3.62387925e-01 8.54685664e-01 8.11710179e-01 1.82106376e-01 -3.31178844e-01 4.28639919e-01 6.17911696e-01 1.03170037e+00 -5.22940457e-01 3.24259192e-01 6.05812632e-02 3.33301604e-01 -4.96736825e-01 1.69729471e-01 -6.42912507e-01 -3.19094121e-01 6.00600719e-01 1.22077107e+00 1.12970859e-01 -3.99548560e-01 -8.84754434e-02 -6.42861426e-02 2.76959658e-01 -1.74954414e-01 -2.20697615e-02 -6.61114156e-01 -1.03790855e+00 3.33235264e-01 7.51526117e-01 -8.64678502e-01 -1.02775440e-01 -7.37785339e-01 -1.16394091e+00 1.03537822e+00 -5.25199950e-01 -5.98203361e-01 -3.15369479e-02 4.39171791e-02 4.87731159e-01 -9.15898010e-02 3.94062996e-01 1.30157337e-01 -4.82342303e-01 1.30232617e-01 4.46951181e-01 -7.54004240e-01 3.21471363e-01 -9.71002221e-01 -1.06049195e-01 4.81361687e-01 -1.05502033e+00 3.88625115e-01 1.41929495e+00 -1.00323796e+00 -1.68452704e+00 -1.98885128e-01 4.17747915e-01 1.69202209e-01 5.58047533e-01 -2.01597661e-01 -1.34511292e+00 -4.19891417e-01 3.75036567e-01 -3.53120208e-01 2.39771947e-01 -3.54578197e-01 8.23729873e-01 7.06649618e-03 -1.06697416e+00 2.11994484e-01 4.31874126e-01 -4.21837538e-01 -7.47396588e-01 8.02647471e-02 2.31413081e-01 -4.21613365e-01 -1.29377997e+00 7.65094221e-01 5.49145401e-01 -1.00856197e+00 5.98376095e-01 -8.39356035e-02 2.84007251e-01 -1.99729621e-01 5.69558203e-01 -1.22996759e+00 -4.69259992e-02 -6.50191367e-01 -1.39547527e-01 1.29723215e+00 -1.12177311e-02 -7.94311106e-01 -3.23548801e-02 3.62187654e-01 -7.65494585e-01 -1.52924919e+00 -1.34189081e+00 -5.36693871e-01 2.31754765e-01 4.82560158e-01 1.30049899e-01 3.91809374e-01 -7.56401792e-02 -8.09933990e-02 8.65210406e-03 9.30970013e-02 9.28638969e-03 -2.86140382e-01 2.11887255e-01 -1.46095920e+00 -3.99031162e-01 3.51943634e-02 1.34203404e-01 2.64058679e-01 -3.19130093e-01 5.48550859e-02 4.76013683e-02 -1.02665591e+00 -5.85232139e-01 -3.59707087e-01 3.30994755e-01 -4.69593942e-01 -1.16750322e-01 -2.30801433e-01 8.11037123e-02 6.84075654e-01 8.30655277e-01 4.68407214e-01 1.29443431e+00 4.89600837e-01 -4.61786747e-01 1.77483588e-01 5.55420578e-01 3.54631901e-01 3.32797080e-01 -2.15491414e-01 -3.89705658e-01 1.29734054e-01 2.69038677e-01 4.20117378e-01 3.27202916e-01 -1.25717354e+00 -1.33126572e-01 1.89956173e-01 -5.23615628e-02 -3.98554146e-01 1.65266439e-01 -9.39807236e-01 1.00405335e+00 8.77222538e-01 3.36057186e-01 5.55951238e-01 2.59000003e-01 2.20837042e-01 -1.64263189e-01 -7.01756358e-01 9.41015601e-01 2.95942664e-01 -1.61025170e-02 -6.63389564e-01 -8.08190823e-01 -4.73035783e-01 1.13966036e+00 -6.32110000e-01 4.84747104e-02 1.86093688e-01 -7.31051803e-01 -2.25285277e-01 7.62932897e-01 2.71905005e-01 1.01206847e-01 -6.07037246e-01 -4.73996580e-01 5.60347438e-01 -7.70319760e-01 -6.02832921e-02 4.59024668e-01 1.10920298e+00 -9.86771345e-01 -5.14889741e-03 -3.01893532e-01 -3.37197512e-01 -1.29007542e+00 2.62892485e-01 8.91785026e-01 -7.34038725e-02 -5.02850890e-01 6.34991348e-01 -3.24241966e-01 4.60328013e-01 -5.14162362e-01 -1.51206225e-01 -4.44169849e-01 3.52078915e-01 -2.68992841e-01 1.27609599e+00 8.06012988e-01 -7.08279848e-01 -9.71598923e-02 9.93393540e-01 5.70090473e-01 -7.21071437e-02 1.05741072e+00 -2.09500208e-01 -3.81997049e-01 8.09890926e-01 8.05732965e-01 3.05388898e-01 -1.30944943e+00 7.21824646e-01 -1.24640472e-01 -2.47928858e-01 -4.33934703e-02 -6.02177382e-01 -8.28841746e-01 4.68253165e-01 4.12327051e-01 4.09985095e-01 9.37355518e-01 -1.55953154e-01 6.54596329e-01 -7.91371837e-02 1.34374231e-01 -1.44059765e+00 3.59448940e-02 3.82414550e-01 1.27382970e+00 -8.17112476e-02 3.51411343e-01 -6.90442979e-01 -3.81789386e-01 1.38566458e+00 6.31085455e-01 -5.08304298e-01 9.96965349e-01 6.82116210e-01 5.26918434e-02 -3.88139993e-01 -5.57963550e-01 5.48225820e-01 2.77050436e-01 -3.81067842e-02 5.65719604e-01 -4.48418632e-02 -1.28825259e+00 5.11053562e-01 -1.29554376e-01 -1.04634412e-01 5.66269279e-01 1.15749240e+00 -5.07547617e-01 -1.01382840e+00 -9.31661904e-01 3.74448866e-01 -7.01971412e-01 6.22056663e-01 -1.60024226e-01 1.06747448e+00 3.81272703e-01 6.58809841e-01 2.15968326e-01 -1.30167976e-01 8.31541657e-01 3.14433575e-01 1.19615704e-01 -3.73052299e-01 -1.07445240e+00 -5.91017539e-03 2.29230672e-01 1.33264571e-01 -1.87914729e-01 -9.38557029e-01 -1.58260703e+00 -6.12324513e-02 -8.24883163e-01 8.21314275e-01 7.08128691e-01 9.59940195e-01 1.69507578e-01 6.53306425e-01 6.61602080e-01 -9.74335909e-01 -8.19979370e-01 -1.26119936e+00 -1.07856441e+00 5.01115859e-01 2.22622827e-02 -1.50842619e+00 -7.61422575e-01 -1.49209455e-01]

[5.919312477111816, 2.8501393795013428]

f39058cc-2549-40d4-b200-7ea48be24b9a

deep-learning-to-improve-breast-cancer-early

1708.09427

http://arxiv.org/abs/1708.09427v5

http://arxiv.org/pdf/1708.09427v5.pdf

Deep Learning to Improve Breast Cancer Early Detection on Screening Mammography

The rapid development of deep learning, a family of machine learning techniques, has spurred much interest in its application to medical imaging problems. Here, we develop a deep learning algorithm that can accurately detect breast cancer on screening mammograms using an "end-to-end" training approach that efficiently leverages training datasets with either complete clinical annotation or only the cancer status (label) of the whole image. In this approach, lesion annotations are required only in the initial training stage, and subsequent stages require only image-level labels, eliminating the reliance on rarely available lesion annotations. Our all convolutional network method for classifying screening mammograms attained excellent performance in comparison with previous methods. On an independent test set of digitized film mammograms from Digital Database for Screening Mammography (DDSM), the best single model achieved a per-image AUC of 0.88, and four-model averaging improved the AUC to 0.91 (sensitivity: 86.1%, specificity: 80.1%). On a validation set of full-field digital mammography (FFDM) images from the INbreast database, the best single model achieved a per-image AUC of 0.95, and four-model averaging improved the AUC to 0.98 (sensitivity: 86.7%, specificity: 96.1%). We also demonstrate that a whole image classifier trained using our end-to-end approach on the DDSM digitized film mammograms can be transferred to INbreast FFDM images using only a subset of the INbreast data for fine-tuning and without further reliance on the availability of lesion annotations. These findings show that automatic deep learning methods can be readily trained to attain high accuracy on heterogeneous mammography platforms, and hold tremendous promise for improving clinical tools to reduce false positive and false negative screening mammography results.

['Joseph H. Rothstein', 'Li Shen', 'Weiva Sieh', 'Russell B. McBride', 'Laurie R. Margolies', 'Eugene Fluder']

2017-08-30

['breast-cancer-detection', 'breast-cancer-detection']

['knowledge-base', 'medical']

[ 6.79841161e-01 5.44936299e-01 -4.82939243e-01 -7.97251582e-01 -1.31226373e+00 -4.51961279e-01 1.25368059e-01 4.50105220e-01 -5.71325541e-01 2.95205802e-01 -2.41280317e-01 -9.32824612e-01 -1.15832565e-02 -8.58903408e-01 -8.66007686e-01 -5.76629460e-01 -3.01233828e-01 5.19638538e-01 4.81087357e-01 2.65157521e-01 -5.53727686e-01 3.90248239e-01 -7.89915800e-01 8.34873259e-01 3.50718826e-01 1.32032549e+00 1.45421147e-01 1.10189342e+00 4.30151463e-01 8.92262638e-01 -1.79688752e-01 -5.16185045e-01 1.79824769e-01 -3.80448729e-01 -7.13208854e-01 1.30825713e-01 4.70821828e-01 -7.83191741e-01 -4.22734946e-01 9.50976968e-01 6.25686169e-01 -6.46860838e-01 5.53475797e-01 -4.18877900e-01 -3.23161036e-01 5.17849326e-01 -7.23620653e-01 3.87895703e-01 -3.91110629e-01 -4.56557535e-02 5.26151001e-01 -4.95823711e-01 4.10329789e-01 4.85434383e-01 1.22606814e+00 5.49107254e-01 -1.11182094e+00 -5.76362550e-01 -5.62281251e-01 -2.69079387e-01 -1.10968065e+00 -2.74957657e-01 -5.83062544e-02 -4.93298471e-01 6.09398603e-01 4.63705212e-01 8.12449932e-01 5.68468213e-01 6.46260500e-01 6.88139796e-01 8.95148218e-01 -4.42458779e-01 8.30106959e-02 9.37240943e-02 2.69281864e-01 1.17091858e+00 4.83713925e-01 5.45521565e-02 2.72819042e-01 -4.32942986e-01 8.97767305e-01 1.66989729e-01 1.30642846e-01 -3.46448839e-01 -9.53143597e-01 9.41510022e-01 6.21622205e-01 3.65368783e-01 -2.91240156e-01 6.57224208e-02 6.23867810e-01 4.24174145e-02 3.10609370e-01 3.83712083e-01 -1.78222194e-01 3.67887467e-01 -1.14759409e+00 -1.94963962e-01 6.19589686e-01 4.72835630e-01 2.76253164e-01 -5.01239538e-01 -2.83796281e-01 9.33787346e-01 -9.80294645e-02 4.99191046e-01 7.18382299e-01 -9.25884187e-01 -5.23701347e-02 7.04337180e-01 5.62422462e-02 -7.72059739e-01 -9.25416708e-01 -7.74150133e-01 -1.09071338e+00 -6.34089485e-02 3.08830798e-01 -2.19428718e-01 -1.34785473e+00 1.44096589e+00 1.60290107e-01 -3.93568903e-01 1.96636189e-02 7.20631719e-01 8.74213398e-01 1.58413395e-01 2.51477093e-01 -1.13243714e-01 1.64304173e+00 -5.30403912e-01 -2.30994344e-01 -2.35914022e-01 1.27238131e+00 -4.17117149e-01 6.11616671e-01 3.04638833e-01 -1.21330500e+00 -4.17785734e-01 -1.24227297e+00 1.64810836e-01 1.03255086e-01 7.32348084e-01 8.20470214e-01 9.05236185e-01 -1.11922443e+00 3.46154273e-01 -1.51940131e+00 -4.16557819e-01 9.52211678e-01 7.61143267e-01 -5.29343605e-01 -5.10679960e-01 -7.36560106e-01 8.30335200e-01 3.81719410e-01 1.91808734e-02 -1.09403872e+00 -7.58189619e-01 -6.76611602e-01 -9.49044712e-03 2.06498682e-01 -5.96332073e-01 1.81503737e+00 -1.28423500e+00 -9.25468504e-01 1.34825170e+00 7.52484426e-02 -7.20660686e-01 6.38129950e-01 3.06924760e-01 -3.15995634e-01 5.45555294e-01 3.46544087e-01 6.16243958e-01 2.93359518e-01 -7.63507128e-01 -8.70358288e-01 -4.67421889e-01 -1.50611609e-01 -1.55432448e-01 -4.47195709e-01 -8.84392411e-02 -6.08278394e-01 -1.77201450e-01 2.56261170e-01 -1.00959027e+00 -5.70092082e-01 3.05701792e-01 1.98191386e-02 2.94283062e-01 5.26023507e-01 -8.96542072e-01 9.43517148e-01 -2.21331143e+00 -5.63363552e-01 2.47460172e-01 4.05694902e-01 5.42944491e-01 7.39028603e-02 -4.05840069e-01 -1.74224406e-01 1.61835298e-01 -1.91048339e-01 2.32958406e-01 -6.74803913e-01 -2.21816991e-02 5.43041825e-01 5.47776461e-01 5.51335514e-01 1.16110218e+00 -9.64268446e-01 -6.31719291e-01 1.01327792e-01 2.48267755e-01 -5.42883754e-01 -3.83459940e-03 2.44764313e-01 1.13553897e-01 -4.04038161e-01 7.07348764e-01 5.15992105e-01 -9.49175715e-01 6.42979324e-01 -4.78347957e-01 3.29452902e-01 -2.71138400e-01 -6.49763346e-01 1.33092964e+00 -4.16662544e-01 4.95694548e-01 1.52569801e-01 -1.06876564e+00 6.55217111e-01 3.53924513e-01 7.66908228e-01 -8.03073585e-01 2.45249569e-01 5.09178162e-01 5.67531228e-01 -6.55472815e-01 -1.28121540e-01 -2.73877323e-01 3.57184112e-02 2.73553401e-01 9.53340232e-02 2.80764396e-03 5.39998487e-02 5.98169267e-02 1.69579995e+00 -4.85459417e-01 5.60939133e-01 -3.41724306e-01 1.35419443e-01 4.44602549e-01 2.57101089e-01 1.04969633e+00 -1.14265062e-01 7.23948240e-01 5.31960428e-01 -6.97537541e-01 -1.17466974e+00 -9.16302502e-01 -7.58314252e-01 7.89852023e-01 -3.38852733e-01 1.49423182e-02 -6.23490870e-01 -8.89787257e-01 3.43982503e-02 1.92495495e-01 -1.00246251e+00 -2.11893067e-01 -6.05914414e-01 -1.37514377e+00 7.89532661e-01 8.60027611e-01 5.35827041e-01 -4.99746978e-01 -8.10682535e-01 4.05661434e-01 -5.91422804e-02 -9.07591403e-01 1.90275218e-02 5.35597682e-01 -9.52530384e-01 -1.34343398e+00 -1.09809732e+00 -9.60342050e-01 9.23846543e-01 9.82790813e-03 1.04833722e+00 9.73969847e-02 -7.28022158e-01 6.76757395e-02 -1.71392322e-01 -4.53168094e-01 -9.06840563e-01 6.89749420e-02 -5.18837571e-01 -1.41016454e-01 3.30446810e-01 1.05991669e-01 -8.07543337e-01 3.41367960e-01 -1.00365376e+00 1.66699693e-01 1.13202560e+00 1.24860799e+00 8.27007949e-01 -2.20972538e-01 5.65535724e-01 -1.07444608e+00 -1.33778274e-01 -6.44362390e-01 -4.98705685e-01 2.04107136e-01 -1.51239410e-01 -3.44945282e-01 2.36814290e-01 -4.30269361e-01 -8.80411029e-01 4.04396564e-01 -2.43941814e-01 1.11929188e-02 1.52479008e-01 6.83633149e-01 4.55834359e-01 -2.16167405e-01 1.01310456e+00 -4.97266278e-02 2.27096394e-01 -2.47349501e-01 -3.27071816e-01 8.34984243e-01 6.99273229e-01 -4.08191187e-03 1.13074947e-02 6.50724113e-01 3.34196627e-01 -5.96174538e-01 -1.07139790e+00 -5.34507811e-01 -6.01588249e-01 1.03043020e-01 9.35153842e-01 -1.00850689e+00 -3.40595365e-01 2.73996532e-01 -5.54093599e-01 -5.39722979e-01 -2.45675057e-01 6.09180331e-01 -2.95213044e-01 1.05620071e-01 -8.50166738e-01 -3.15060824e-01 -5.22465825e-01 -1.17359483e+00 1.20578957e+00 3.26910056e-02 -2.87989110e-01 -7.96297491e-01 -3.10527891e-01 2.18497396e-01 3.65091860e-01 4.34822500e-01 9.58153367e-01 -8.33160222e-01 -1.39853999e-01 -8.93089712e-01 -6.54651701e-01 4.82449949e-01 4.08796847e-01 -2.78428704e-01 -7.94267952e-01 -4.11705762e-01 -1.18205070e-01 -5.48344314e-01 1.08710539e+00 8.38909268e-01 1.62445593e+00 2.62685716e-01 -8.47065091e-01 7.23641813e-01 1.30491853e+00 3.37837845e-01 3.42491060e-01 2.25972012e-01 2.64557481e-01 8.00272450e-02 3.80171090e-01 1.43320635e-01 8.07634443e-02 2.41998836e-01 2.83395290e-01 -7.26302266e-01 -3.53681237e-01 -8.05384070e-02 -3.63710850e-01 1.14583693e-01 2.45123934e-02 -8.37637857e-02 -1.28105116e+00 5.98775983e-01 -1.48863053e+00 -5.35753071e-01 -1.22248679e-01 1.97817051e+00 8.90791476e-01 3.00937057e-01 -1.65808141e-01 1.21615576e-02 6.22816324e-01 -4.04686719e-01 -6.39277041e-01 -1.11599259e-01 5.95403351e-02 4.08232689e-01 8.55082333e-01 1.62935421e-01 -1.54736853e+00 3.54244143e-01 6.55769157e+00 5.91562152e-01 -1.21675372e+00 3.35242271e-01 1.41016853e+00 8.88061523e-03 3.15991104e-01 -4.93161291e-01 -6.14418387e-01 1.87554449e-01 1.12599206e+00 1.90372765e-01 -4.23782676e-01 8.70444655e-01 -3.11554167e-02 -4.22370583e-01 -1.24999678e+00 5.77259123e-01 -1.69118308e-02 -1.51371920e+00 -2.51718849e-01 6.79179803e-02 8.84711325e-01 1.88575074e-01 3.33816335e-02 1.36248097e-01 2.91989446e-01 -1.14911890e+00 6.20637424e-02 2.88025200e-01 1.53712988e+00 -5.09837329e-01 1.42496836e+00 3.55131179e-01 -7.27023125e-01 -1.21993497e-01 -1.82326198e-01 2.75747746e-01 -3.02334458e-01 6.44722223e-01 -1.45771134e+00 2.77929634e-01 8.42739165e-01 3.29276741e-01 -7.92474627e-01 1.08473766e+00 4.01448876e-01 8.13624978e-01 -3.50294590e-01 7.33909830e-02 2.70527631e-01 7.59788752e-01 -2.14297161e-01 1.56105995e+00 3.78229827e-01 4.27617162e-01 8.34876224e-02 3.09656262e-01 -2.71519840e-01 -2.02945806e-02 -2.13661984e-01 3.21870744e-02 -1.06356917e-02 1.32180882e+00 -9.52027559e-01 -6.18322432e-01 -4.02691185e-01 5.91437101e-01 1.25297919e-01 -1.07028700e-01 -6.76831901e-01 -1.64398283e-01 -4.02228206e-01 3.91145170e-01 2.70468235e-01 3.59522194e-01 -2.96993017e-01 -6.97329760e-01 -1.56725898e-01 -8.72504354e-01 5.75434744e-01 -3.92202109e-01 -1.16529429e+00 4.17889476e-01 -3.38506810e-02 -1.00851083e+00 -2.45317057e-01 -8.50803494e-01 -2.63757020e-01 6.30037427e-01 -1.21484017e+00 -1.24658668e+00 -5.43265402e-01 1.51350141e-01 1.67897880e-01 -6.57279342e-02 1.03133154e+00 3.90993685e-01 -1.88305452e-01 9.73111153e-01 4.11857843e-01 5.91360092e-01 6.25322342e-01 -1.02755773e+00 2.35329308e-02 3.77993315e-01 -6.08897805e-01 1.02441117e-01 -1.98403094e-03 -6.02288544e-01 -1.20386600e+00 -1.48582816e+00 6.82041824e-01 -2.60811985e-01 4.81315941e-01 1.19798973e-01 -6.90498412e-01 8.73309314e-01 -4.24295962e-01 4.98408556e-01 1.00926304e+00 -2.81160682e-01 3.79705988e-02 -1.79086164e-01 -1.38793683e+00 2.00347483e-01 6.29929781e-01 2.90732104e-02 -1.68820202e-01 6.04240417e-01 2.50181407e-01 -7.54029930e-01 -1.08726418e+00 9.84902322e-01 7.63078153e-01 -7.34464943e-01 7.15506256e-01 -5.35761297e-01 6.48649633e-01 2.10649952e-01 -1.78085774e-01 -7.10394561e-01 -5.66169560e-01 3.85837778e-02 2.01289967e-01 4.46943611e-01 6.15095675e-01 -3.72336596e-01 9.62104321e-01 5.65742433e-01 -2.67451286e-01 -1.06881714e+00 -8.92873168e-01 -3.02695572e-01 1.50584340e-01 -1.87981203e-01 1.49906844e-01 6.50265217e-01 -3.02040845e-01 -2.92329967e-01 1.50095880e-01 2.23401457e-01 3.87776732e-01 -9.36715975e-02 2.93383390e-01 -9.92804706e-01 -6.90743029e-01 -2.36950859e-01 -6.34560049e-01 -6.69670522e-01 -4.66621429e-01 -1.15515268e+00 -1.21623449e-01 -1.46643305e+00 9.96209264e-01 -6.50688827e-01 -5.15705049e-01 6.98289454e-01 -3.82034123e-01 8.19031179e-01 -2.42253602e-01 2.09421173e-01 -4.37989801e-01 -3.74472320e-01 1.27571058e+00 -2.72907078e-01 1.15382969e-01 3.33111525e-01 -7.46441662e-01 9.27830398e-01 6.99714303e-01 -5.66164315e-01 2.66957749e-02 -5.55697680e-01 -7.76424780e-02 5.25065005e-01 6.11609340e-01 -1.16093278e+00 -5.27259484e-02 1.63323238e-01 1.20148230e+00 -3.69973689e-01 2.68435657e-01 -5.65017343e-01 1.21759161e-01 1.14611638e+00 -4.74832714e-01 -5.10740697e-01 5.23165762e-01 4.15618211e-01 -1.16211250e-01 -2.52614021e-01 1.10948551e+00 -4.46269542e-01 -3.24674308e-01 2.46663854e-01 -4.54643488e-01 -3.38036686e-01 1.10054100e+00 -2.49729261e-01 -2.64120959e-02 2.54617594e-02 -9.98268485e-01 9.57743227e-02 1.50723904e-01 -3.20386812e-02 2.44947910e-01 -1.19658566e+00 -9.48567510e-01 2.61390686e-01 9.73196328e-02 6.31967038e-02 2.78385133e-01 9.90474820e-01 -6.87716067e-01 5.13331473e-01 -9.73278061e-02 -1.08894873e+00 -1.40944946e+00 5.24322279e-02 9.40075040e-01 -5.70873559e-01 -3.91950548e-01 1.01730978e+00 3.67868662e-01 -2.51004010e-01 4.57063615e-02 -4.17239428e-01 1.76734135e-01 -2.77899355e-01 6.48611546e-01 -5.55118397e-02 5.31460226e-01 -1.61631450e-01 -2.71583408e-01 2.04145581e-01 -5.20574570e-01 1.01512738e-01 1.32243657e+00 4.59772229e-01 2.02133805e-01 7.13739395e-02 1.38090897e+00 -4.46134627e-01 -9.73354220e-01 -2.66677827e-01 -1.23799965e-01 -2.35374704e-01 4.05272990e-01 -1.24598217e+00 -9.48688209e-01 7.69897819e-01 1.10157454e+00 1.27852932e-01 1.25591040e+00 2.84098148e-01 4.83835548e-01 4.79879439e-01 1.43842116e-01 -5.50568104e-01 5.11828847e-02 4.58277799e-02 4.67787266e-01 -1.70075452e+00 1.41031742e-01 -1.77173316e-01 -5.25478601e-01 1.27032340e+00 4.60774004e-01 -7.82220960e-02 7.18838215e-01 5.36761880e-01 3.40425745e-02 -2.64175028e-01 -4.94740754e-01 2.97152251e-01 2.72744656e-01 4.70284015e-01 8.31442177e-01 4.04260397e-01 -2.32892931e-01 1.01721716e+00 9.54898372e-02 5.30454695e-01 2.95225412e-01 1.17174625e+00 -6.63779736e-01 -7.79812634e-01 -1.74552977e-01 1.27939713e+00 -9.17815685e-01 1.22372724e-01 -2.01882333e-01 9.74151850e-01 2.70013213e-01 6.85428083e-01 2.10263416e-01 -4.39951345e-02 2.49417901e-01 -1.59079060e-01 5.03000975e-01 -7.68242359e-01 -5.43128908e-01 2.78947324e-01 2.78401643e-01 -3.36953312e-01 -2.22149611e-01 -5.45635879e-01 -1.18054509e+00 1.07846074e-01 -5.70873320e-01 -9.15532559e-02 5.29536664e-01 7.29211688e-01 1.84521765e-01 7.03621507e-01 4.31511104e-01 -5.63956678e-01 -7.67011106e-01 -9.56443429e-01 -4.58787262e-01 5.68843745e-02 5.77537537e-01 -1.88844159e-01 9.32775214e-02 2.65511572e-01]

[15.181408882141113, -2.522956132888794]

8b27e878-4d1d-4d48-a9d5-784475a8cfb5

attend-to-who-you-are-supervising-self-1

2111.12892

https://arxiv.org/abs/2111.12892v1

https://arxiv.org/pdf/2111.12892v1.pdf

Attend to Who You Are: Supervising Self-Attention for Keypoint Detection and Instance-Aware Association

This paper presents a new method to solve keypoint detection and instance association by using Transformer. For bottom-up multi-person pose estimation models, they need to detect keypoints and learn associative information between keypoints. We argue that these problems can be entirely solved by Transformer. Specifically, the self-attention in Transformer measures dependencies between any pair of locations, which can provide association information for keypoints grouping. However, the naive attention patterns are still not subjectively controlled, so there is no guarantee that the keypoints will always attend to the instances to which they belong. To address it we propose a novel approach of supervising self-attention for multi-person keypoint detection and instance association. By using instance masks to supervise self-attention to be instance-aware, we can assign the detected keypoints to their corresponding instances based on the pairwise attention scores, without using pre-defined offset vector fields or embedding like CNN-based bottom-up models. An additional benefit of our method is that the instance segmentation results of any number of people can be directly obtained from the supervised attention matrix, thereby simplifying the pixel assignment pipeline. The experiments on the COCO multi-person keypoint detection challenge and person instance segmentation task demonstrate the effectiveness and simplicity of the proposed method and show a promising way to control self-attention behavior for specific purposes.

['Wankou Yang', 'Erjin Zhou', 'Yiping Bao', 'Shu-Tao Xia', 'Zhibin Quan', 'Shoukui Zhang', 'YanJie Li', 'Ze Chen', 'Zhicheng Wang', 'Sen yang']

2021-11-25

attend-to-who-you-are-supervising-self

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

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

['multi-person-pose-estimation']

['computer-vision']

[-6.74836859e-02 3.61638144e-02 1.30719498e-01 -4.32129145e-01 -6.86163008e-01 -5.22958815e-01 4.94581223e-01 4.69417840e-01 -6.91533685e-01 2.70966589e-01 1.50437683e-01 5.64598203e-01 -2.04409227e-01 -8.04948032e-01 -9.12929296e-01 -5.09441316e-01 2.60539223e-02 8.02945793e-01 5.47789037e-01 -1.72857657e-01 1.02947235e-01 4.73584384e-01 -1.69924951e+00 2.99629271e-01 7.82180190e-01 9.57670450e-01 2.30516240e-01 6.89638138e-01 2.63560086e-01 2.77050138e-01 -7.89829969e-01 -5.37174165e-01 3.44261855e-01 -1.11148104e-01 -5.95868886e-01 -5.81037626e-02 1.17390156e+00 -2.53644258e-01 -1.90108314e-01 9.44603264e-01 6.21214330e-01 2.84469038e-01 5.15123129e-01 -1.35855055e+00 -5.57569802e-01 2.90104449e-01 -7.17923939e-01 3.72849852e-01 5.02132893e-01 2.11059257e-01 1.29993355e+00 -8.63534272e-01 4.36190456e-01 1.41778636e+00 7.81782687e-01 4.71644819e-01 -1.25030732e+00 -6.21475279e-01 6.27514541e-01 5.85017443e-01 -1.57763505e+00 -1.99577466e-01 6.06040776e-01 -2.90450245e-01 8.94304037e-01 5.55467963e-01 1.14100921e+00 7.03289807e-01 -1.69619232e-01 1.14014983e+00 7.51011610e-01 -2.32022867e-01 -3.13674510e-02 4.06579316e-01 3.51064086e-01 7.28073359e-01 1.52644351e-01 -2.12394238e-01 -7.33412385e-01 -1.00125909e-01 7.26515949e-01 2.19493389e-01 -1.51743203e-01 -5.65054595e-01 -1.37370765e+00 6.77294254e-01 1.03070593e+00 7.48685449e-02 -4.47446018e-01 3.47593695e-01 1.07693098e-01 -2.67756075e-01 3.04793984e-01 4.90294605e-01 -4.18578058e-01 2.47752100e-01 -9.84668076e-01 7.08084881e-01 2.78463691e-01 9.03415143e-01 8.85021627e-01 -8.31997037e-01 -6.70260668e-01 7.22364843e-01 2.69354403e-01 5.00665069e-01 1.94534719e-01 -6.75393999e-01 4.94205505e-01 8.88232470e-01 1.97903365e-01 -1.27923298e+00 -3.74797523e-01 -4.52182591e-01 -3.81191283e-01 -9.05780196e-02 5.02653241e-01 1.81778431e-01 -1.04415321e+00 1.69943988e+00 5.91726661e-01 3.09544653e-01 -5.81966162e-01 1.15051651e+00 7.55923271e-01 4.06319350e-01 1.21177956e-01 2.83675462e-01 1.87819290e+00 -9.46955562e-01 -5.77836752e-01 -4.28296924e-01 3.33708435e-01 -2.42841139e-01 1.11596179e+00 2.34653786e-01 -1.03872097e+00 -7.62580931e-01 -9.25060451e-01 -2.81385034e-01 -5.85011244e-01 1.52217716e-01 7.14392364e-01 5.50023735e-01 -8.14252615e-01 2.82163650e-01 -8.10981035e-01 -4.95063335e-01 5.88629842e-01 5.90558887e-01 -4.53630239e-01 1.20421998e-01 -1.15042901e+00 7.32570887e-01 1.74708039e-01 2.04970747e-01 -5.74689269e-01 -8.09969068e-01 -8.88068438e-01 1.83084235e-01 4.06689674e-01 -8.02156329e-01 8.67404997e-01 -6.27994657e-01 -9.35395360e-01 8.38998616e-01 -3.75443727e-01 -4.54956174e-01 5.79399467e-01 -7.20958233e-01 4.78897505e-02 3.76524746e-01 5.33717036e-01 1.11126041e+00 1.00747454e+00 -1.23540533e+00 -1.05548692e+00 -6.02486372e-01 2.33251169e-01 3.96217793e-01 -5.57346165e-01 -3.15701626e-02 -1.23582923e+00 -5.35091400e-01 1.86786935e-01 -1.05234396e+00 -2.69675314e-01 8.47014319e-03 -6.97805882e-01 -5.58897913e-01 7.00024426e-01 -4.80392516e-01 9.81528640e-01 -1.93564773e+00 2.63056159e-01 4.07781810e-01 3.01527739e-01 2.12901533e-02 -1.51395621e-02 -3.82581912e-02 4.22198847e-02 -2.82817911e-02 5.45127429e-02 -5.67668080e-01 3.56230401e-02 -7.83677213e-03 -3.16958092e-02 4.77463007e-01 4.59247202e-01 1.17264521e+00 -9.32519972e-01 -6.14883900e-01 5.36871910e-01 6.94676816e-01 -7.53796935e-01 1.86686426e-01 -7.15490282e-02 2.31316417e-01 -2.85975248e-01 6.69134617e-01 5.59847593e-01 -1.90222010e-01 -2.55209535e-01 -6.14339948e-01 1.60317928e-01 5.35027161e-02 -1.59263647e+00 1.51638639e+00 -1.64598733e-01 5.20265579e-01 -1.39002144e-01 -5.63328981e-01 3.64986956e-01 -6.67429715e-02 2.86789387e-01 -5.96590281e-01 -2.20132411e-01 -2.94882536e-01 -2.06736922e-01 -2.60817200e-01 6.19365811e-01 2.99058646e-01 -2.47065142e-01 1.87699661e-01 7.15729743e-02 2.58632332e-01 3.07743847e-01 3.67559761e-01 8.44037652e-01 6.47139549e-02 1.47129875e-02 -1.77756045e-02 4.77056503e-01 -8.93934667e-02 5.83022535e-01 9.87088919e-01 -2.56454200e-01 8.76214683e-01 2.84603179e-01 -5.04218340e-01 -7.11025417e-01 -1.04857337e+00 -9.68589932e-02 1.36325002e+00 4.10760552e-01 -7.33524680e-01 -9.10178304e-01 -9.59457040e-01 2.13172451e-01 1.94714710e-01 -9.90155280e-01 9.34562981e-02 -6.51851416e-01 -6.68138981e-01 2.07424641e-01 7.51026154e-01 4.35268998e-01 -9.31142032e-01 -7.41084576e-01 8.00197348e-02 -3.95700991e-01 -1.18659914e+00 -7.82105863e-01 1.19789079e-01 -2.58064568e-01 -1.10994279e+00 -9.75844085e-01 -5.41841865e-01 9.65629756e-01 2.90161639e-01 9.65070724e-01 1.82479486e-01 -4.77251798e-01 8.01784277e-01 -1.26933768e-01 -3.48292261e-01 4.62285519e-01 1.93233252e-01 1.08467057e-01 3.33570540e-01 6.46131933e-01 -2.06593931e-01 -1.05743515e+00 4.92253959e-01 -3.39878172e-01 -2.61145473e-01 6.29331052e-01 6.06081605e-01 8.13953340e-01 2.24735364e-01 5.44593222e-02 -6.69897199e-01 3.45637560e-01 1.25675902e-01 -3.49782139e-01 3.00750047e-01 -1.40213519e-01 -8.42137933e-02 1.04332112e-01 -5.35521388e-01 -6.57758117e-01 4.98909175e-01 5.95675632e-02 -4.70874876e-01 -2.15116173e-01 -3.11923772e-02 -5.41050076e-01 -1.35929525e-01 4.48731124e-01 1.41177297e-01 -3.64915997e-01 -2.76636571e-01 6.18701100e-01 3.75474900e-01 5.62021196e-01 -4.76749957e-01 8.43074799e-01 7.50617445e-01 -3.31511021e-01 -8.05389345e-01 -8.95291448e-01 -7.75568187e-01 -8.77682388e-01 -3.37936819e-01 1.27053273e+00 -1.05124652e+00 -1.13456833e+00 2.63507038e-01 -1.08206499e+00 -5.20923547e-02 -2.03778699e-01 2.50257879e-01 -3.16630274e-01 3.36802721e-01 -4.07269180e-01 -7.79933333e-01 -1.75250322e-01 -1.14265299e+00 1.58326602e+00 3.58860224e-01 -5.51390707e-01 -7.12748230e-01 -2.30090499e-01 5.58017433e-01 -4.54770550e-02 6.10555448e-02 4.09099728e-01 -5.74341953e-01 -9.18173909e-01 -5.43909252e-01 -2.90085107e-01 -5.08124121e-02 -1.01380147e-01 -3.33054036e-01 -9.89007115e-01 -3.53616297e-01 -4.61102635e-01 -6.41968846e-02 1.03046834e+00 4.51951832e-01 1.17603755e+00 -1.08881414e-01 -7.73704708e-01 5.70398629e-01 1.02824616e+00 -4.36946958e-01 4.72121060e-01 3.77376139e-01 1.17602408e+00 8.30668151e-01 6.98629200e-01 2.26793587e-01 6.76457405e-01 1.04915929e+00 3.34464759e-01 -3.57228190e-01 -5.01987003e-02 -4.05006409e-01 4.63064723e-02 -1.41795427e-01 -1.69334948e-01 -2.20062342e-02 -7.03168750e-01 7.55233228e-01 -2.01484942e+00 -1.04950249e+00 -2.80515552e-01 2.22384119e+00 5.23161173e-01 2.10524887e-01 5.29131532e-01 1.66686311e-01 8.51429701e-01 -7.47075453e-02 -4.07686502e-01 3.53506595e-01 7.28320479e-02 1.18376330e-01 5.50121009e-01 4.76129383e-01 -1.56495559e+00 9.78446126e-01 5.80075502e+00 6.61162317e-01 -6.90314829e-01 4.48265374e-02 4.69515294e-01 -4.95412588e-01 9.92086083e-02 -2.27637142e-01 -1.30282128e+00 5.46779811e-01 3.16296667e-01 4.68996167e-01 6.74641505e-02 7.67227530e-01 2.51845643e-02 -1.78017378e-01 -1.43379700e+00 1.25616145e+00 1.61514595e-01 -1.03127503e+00 1.63705140e-01 1.15164436e-01 3.23326528e-01 -4.26397413e-01 2.36688733e-01 1.90377235e-01 -4.86246385e-02 -8.28124940e-01 8.35321844e-01 5.73296070e-01 2.46085331e-01 -1.05973637e+00 6.23620927e-01 1.54885411e-01 -1.45026612e+00 -1.48144171e-01 -3.69890392e-01 1.23001665e-01 1.23737536e-01 4.55495477e-01 -7.72489667e-01 2.57589847e-01 1.23633599e+00 6.21905625e-01 -8.36679876e-01 1.07674706e+00 -4.40854788e-01 1.63459525e-01 -5.76242805e-01 -3.36068086e-02 1.22128259e-02 2.12236896e-01 5.66789627e-01 1.31829607e+00 7.41004915e-05 -8.36093500e-02 2.43516535e-01 9.27068412e-01 3.57247621e-01 -2.11139861e-02 -1.39087606e-02 3.55284631e-01 3.75770301e-01 1.51290846e+00 -1.01406097e+00 -4.13603693e-01 -2.64989883e-01 1.38774288e+00 4.12392467e-01 2.68717319e-01 -8.97743046e-01 -3.94144714e-01 8.94313812e-01 6.30743206e-01 5.90857744e-01 -8.79747123e-02 -2.23162189e-01 -9.54162121e-01 2.93279082e-01 -6.87384725e-01 5.59959471e-01 -7.42530823e-01 -1.37200999e+00 3.01295191e-01 7.74551407e-02 -9.53695893e-01 6.14857562e-02 -6.27623141e-01 -4.76154983e-01 7.71156132e-01 -1.17413652e+00 -1.52680480e+00 -5.24190724e-01 7.51404047e-01 2.61724025e-01 1.47938386e-01 5.76274633e-01 3.77421379e-01 -5.44076324e-01 9.61995423e-01 -7.04170167e-01 4.59573805e-01 7.06639290e-01 -1.59259570e+00 4.78644490e-01 9.35732245e-01 4.23066109e-01 9.72235322e-01 7.06040204e-01 -6.48408115e-01 -1.20744109e+00 -1.01332986e+00 7.24336207e-01 -1.12575305e+00 2.97250032e-01 -8.58581185e-01 -8.16526830e-01 7.01570332e-01 -1.21362843e-01 2.19388828e-01 6.66915655e-01 5.00145197e-01 -2.43244648e-01 -2.35967234e-01 -9.68827426e-01 6.42845690e-01 1.20876575e+00 -4.53298032e-01 -7.07452297e-01 4.70553994e-01 5.06310105e-01 -4.37163651e-01 -5.75163424e-01 2.45296225e-01 4.95876521e-01 -9.25228179e-01 1.49434245e+00 -3.27035129e-01 1.88098270e-02 -6.61259413e-01 1.13353416e-01 -1.15415013e+00 -6.64808929e-01 -2.38646954e-01 -2.50545651e-01 1.37498760e+00 2.33305544e-01 -4.04959053e-01 9.71928596e-01 9.04590189e-01 2.25490406e-01 -5.11830091e-01 -1.02180398e+00 -4.17793959e-01 -4.41306263e-01 -3.18779796e-01 6.80190265e-01 6.54845595e-01 -5.70596270e-02 3.68383586e-01 -3.06592256e-01 6.99234426e-01 8.96956146e-01 -4.92375158e-02 1.01352358e+00 -1.15712821e+00 -4.67775494e-01 -2.75827974e-01 -8.49540174e-01 -1.32205212e+00 -8.85811821e-02 -6.94027007e-01 4.99141999e-02 -1.67247307e+00 4.63111639e-01 -5.31142175e-01 -5.12133300e-01 6.22259021e-01 -7.78546810e-01 6.33019567e-01 2.97813028e-01 1.75082847e-01 -9.13106143e-01 2.42726982e-01 9.92585897e-01 -3.94317716e-01 -3.87959272e-01 1.86459478e-02 -7.02082753e-01 5.92355132e-01 3.34079713e-01 -3.45381618e-01 -2.71632344e-01 -2.25407243e-01 2.56468773e-01 -5.57890177e-01 1.02169192e+00 -1.24902081e+00 5.10712683e-01 2.63583273e-01 9.46272016e-01 -9.49033022e-01 6.00233316e-01 -8.50599289e-01 -2.05170423e-01 3.27206016e-01 -2.63194323e-01 -6.09138049e-02 7.99623132e-02 7.67701805e-01 9.62543413e-02 1.33232966e-01 6.01507008e-01 -2.77772635e-01 -9.20567632e-01 4.89682049e-01 -1.36230113e-02 -1.24114089e-01 1.15325749e+00 -4.06242847e-01 1.26867533e-01 -1.27714425e-01 -9.73337948e-01 4.90128577e-01 3.30106318e-01 5.19586921e-01 6.65467203e-01 -1.30962813e+00 -5.06891310e-01 3.45814973e-01 3.87476355e-01 1.43663704e-01 4.04845566e-01 7.04043150e-01 -5.35902940e-02 2.14574993e-01 -5.06953038e-02 -1.03138673e+00 -1.49965906e+00 6.39241934e-01 5.49954534e-01 -1.09411262e-01 -7.85252690e-01 1.26376975e+00 4.24560875e-01 -2.91832238e-01 3.40485483e-01 -2.72722632e-01 -2.75316596e-01 3.87118429e-01 8.50910902e-01 1.58163786e-01 5.30942641e-02 -6.99744821e-01 -6.90227330e-01 8.27505887e-01 -4.64956820e-01 -7.53794089e-02 1.25036168e+00 -1.17350183e-01 -2.55970787e-02 1.53638512e-01 8.95097017e-01 3.49656641e-02 -1.38086247e+00 -8.21559355e-02 -2.79354721e-01 -6.10829830e-01 -1.02324791e-01 -7.02198207e-01 -9.40693259e-01 8.37073445e-01 8.56531322e-01 1.19328052e-01 8.09415698e-01 3.22833598e-01 7.63785779e-01 2.72541553e-01 2.19606042e-01 -1.30224800e+00 1.91731349e-01 3.85626741e-02 7.88705885e-01 -1.25113845e+00 -2.65844204e-02 -5.77310264e-01 -4.86363500e-01 6.62472665e-01 9.77207661e-01 -1.86640173e-01 3.36316109e-01 1.16423860e-01 -1.80042200e-02 -3.28624517e-01 -3.02787662e-01 -6.06697619e-01 6.49376988e-01 8.31944346e-01 2.07474738e-01 -8.48458242e-03 1.38980597e-01 5.60789466e-01 -3.30708534e-01 -4.59060699e-01 -9.07404646e-02 5.27154088e-01 -4.38499749e-01 -8.79015803e-01 -8.01400304e-01 4.44794953e-01 -2.60208130e-01 2.31904089e-02 -5.42128146e-01 7.32447803e-01 4.33622360e-01 7.74070799e-01 2.52947241e-01 -2.21842885e-01 6.48081899e-01 -2.14561448e-03 5.45806408e-01 -6.38838589e-01 -7.49809802e-01 -7.81844370e-03 -2.46229097e-01 -5.88118255e-01 -3.25864106e-01 -6.84504867e-01 -1.16657734e+00 -1.02146249e-02 -3.09462667e-01 2.10985281e-02 2.24440545e-01 8.30822766e-01 3.53684843e-01 6.18495584e-01 9.78545696e-02 -1.17300546e+00 2.61554234e-02 -7.27680922e-01 -1.88180268e-01 7.63991177e-01 2.81620651e-01 -8.25971842e-01 -8.35493207e-02 -7.10686445e-02]

[7.214339256286621, -0.7632354497909546]

c1964ab9-71ee-4615-bda9-0d0d851e0ae7

where-does-it-exist-spatio-temporal-video

2001.06891

https://arxiv.org/abs/2001.06891v3

https://arxiv.org/pdf/2001.06891v3.pdf

Where Does It Exist: Spatio-Temporal Video Grounding for Multi-Form Sentences

In this paper, we consider a novel task, Spatio-Temporal Video Grounding for Multi-Form Sentences (STVG). Given an untrimmed video and a declarative/interrogative sentence depicting an object, STVG aims to localize the spatio-temporal tube of the queried object. STVG has two challenging settings: (1) We need to localize spatio-temporal object tubes from untrimmed videos, where the object may only exist in a very small segment of the video; (2) We deal with multi-form sentences, including the declarative sentences with explicit objects and interrogative sentences with unknown objects. Existing methods cannot tackle the STVG task due to the ineffective tube pre-generation and the lack of object relationship modeling. Thus, we then propose a novel Spatio-Temporal Graph Reasoning Network (STGRN) for this task. First, we build a spatio-temporal region graph to capture the region relationships with temporal object dynamics, which involves the implicit and explicit spatial subgraphs in each frame and the temporal dynamic subgraph across frames. We then incorporate textual clues into the graph and develop the multi-step cross-modal graph reasoning. Next, we introduce a spatio-temporal localizer with a dynamic selection method to directly retrieve the spatio-temporal tubes without tube pre-generation. Moreover, we contribute a large-scale video grounding dataset VidSTG based on video relation dataset VidOR. The extensive experiments demonstrate the effectiveness of our method.

['Qi. Wang', 'Huasheng Liu', 'Zhou Zhao', 'Lianli Gao', 'Zhu Zhang', 'Yang Zhao']

2020-01-19

where-does-it-exist-spatio-temporal-video-1

http://openaccess.thecvf.com/content_CVPR_2020/html/Zhang_Where_Does_It_Exist_Spatio-Temporal_Video_Grounding_for_Multi-Form_Sentences_CVPR_2020_paper.html

http://openaccess.thecvf.com/content_CVPR_2020/papers/Zhang_Where_Does_It_Exist_Spatio-Temporal_Video_Grounding_for_Multi-Form_Sentences_CVPR_2020_paper.pdf

cvpr-2020-6

['video-grounding', 'spatio-temporal-video-grounding']

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

[-2.39572097e-02 -1.57799631e-01 -1.69854537e-01 -2.16623440e-01 -5.59203267e-01 -6.59037232e-01 3.88778120e-01 1.25482725e-02 -1.61291227e-01 4.40613210e-01 2.07516015e-01 -9.42376405e-02 -3.17502588e-01 -8.25990677e-01 -1.01773679e+00 -4.07138288e-01 -8.78333598e-02 3.23838770e-01 7.77759731e-01 -1.34262130e-01 3.44502069e-02 9.26133469e-02 -1.30766475e+00 6.24405861e-01 5.75458765e-01 1.05787086e+00 5.55302501e-01 6.68191969e-01 -1.33076027e-01 1.36462235e+00 -4.76399869e-01 -3.12210768e-01 -1.50950057e-02 -6.01834834e-01 -9.92010474e-01 5.00168920e-01 4.39364672e-01 -4.10665959e-01 -8.37350011e-01 9.43324924e-01 7.80329183e-02 4.80724782e-01 2.40724206e-01 -1.41312528e+00 -6.13752007e-01 6.80524766e-01 -6.02720022e-01 6.48286581e-01 8.13051462e-01 3.15945387e-01 1.02411926e+00 -7.94869184e-01 1.15175426e+00 1.27693415e+00 3.73831093e-01 2.17805967e-01 -7.35499263e-01 -3.63381386e-01 7.51270950e-01 5.67993283e-01 -1.61054337e+00 -2.36884713e-01 1.10555983e+00 -4.79321539e-01 8.56306612e-01 3.52843076e-01 9.30125296e-01 1.01192713e+00 -8.61610174e-02 8.88223410e-01 7.54871130e-01 1.20912187e-01 4.39196527e-02 -3.66556734e-01 -1.14370637e-01 1.06935477e+00 -2.39985377e-01 -3.59355479e-01 -7.32305825e-01 7.92091563e-02 1.00130367e+00 1.02195375e-01 -4.19066340e-01 -1.52170867e-01 -1.66718590e+00 4.78318244e-01 4.13872153e-01 6.97504997e-01 -2.67192602e-01 2.79113054e-01 5.05354047e-01 2.39856273e-01 4.54859018e-01 1.66307017e-01 -3.40639234e-01 -6.58527017e-02 -9.49101746e-01 2.98090219e-01 6.53858602e-01 1.39937413e+00 8.00560653e-01 -2.32391819e-01 -4.79084551e-01 4.15671349e-01 1.19763665e-01 2.00001076e-01 2.24623710e-01 -8.16628635e-01 1.15269232e+00 8.59371722e-01 6.81664888e-03 -1.68023992e+00 -2.94405729e-01 -5.56651093e-02 -6.74667299e-01 -6.82467103e-01 3.16058904e-01 2.22113162e-01 -6.59278631e-01 1.51093566e+00 6.49472117e-01 8.19109976e-01 -1.42166913e-01 1.20851386e+00 1.30919421e+00 9.30623531e-01 -7.14215711e-02 -6.29699528e-01 1.55060804e+00 -1.09550905e+00 -8.78362834e-01 -1.32538751e-01 6.32937908e-01 -3.87705982e-01 1.18591785e+00 7.43008256e-02 -9.23363626e-01 -5.34752667e-01 -8.00374925e-01 -3.48337620e-01 -1.99000672e-01 -5.77101111e-02 2.49425054e-01 -1.66666493e-01 -7.64437199e-01 2.22921804e-01 -6.99496627e-01 -2.67972410e-01 2.97626555e-01 1.88699529e-01 -4.85042572e-01 -4.21612233e-01 -1.18851876e+00 2.58721799e-01 7.54575551e-01 3.51969212e-01 -8.30767632e-01 -3.56208950e-01 -1.10117507e+00 -2.37921745e-01 1.21439445e+00 -7.49585748e-01 8.94983709e-01 -8.20140481e-01 -9.46030736e-01 7.90357351e-01 -4.04625177e-01 -3.05167586e-01 4.62156862e-01 -1.03389576e-01 -4.85683382e-01 8.23513210e-01 4.03543770e-01 2.89775938e-01 7.76230812e-01 -1.07890987e+00 -5.24482727e-01 -3.76531720e-01 5.91325223e-01 4.44932848e-01 -7.74553865e-02 5.95547445e-02 -1.13997114e+00 -8.94275963e-01 4.59278882e-01 -6.27738118e-01 9.23820809e-02 -1.64081588e-01 -6.43401206e-01 -3.32547992e-01 1.23464763e+00 -7.26497650e-01 1.62541127e+00 -2.16072679e+00 4.77860272e-01 -1.77910849e-01 4.02835637e-01 -2.23009244e-01 -9.73916873e-02 3.87167364e-01 4.75667529e-02 1.02920018e-01 -5.68365529e-02 -2.82933712e-01 -2.39338562e-01 5.26329219e-01 -4.54621345e-01 4.03258026e-01 2.85068471e-02 1.23616815e+00 -1.27184093e+00 -1.11390364e+00 2.02482313e-01 9.47836339e-02 -3.40146482e-01 2.46581465e-01 -6.54248178e-01 4.48351532e-01 -8.28710079e-01 7.34635353e-01 3.13238561e-01 -5.69378793e-01 -3.92376930e-02 -7.42822647e-01 -8.71800557e-02 7.57157430e-02 -1.09054160e+00 2.05379033e+00 5.35635948e-02 3.49770129e-01 -3.09455842e-01 -1.04489565e+00 5.50215125e-01 2.63226420e-01 7.38482237e-01 -5.53279579e-01 3.45592648e-02 -5.28036356e-02 -3.90939802e-01 -1.30249488e+00 5.16764998e-01 9.28529426e-02 -1.98921487e-01 2.97610074e-01 1.26681805e-01 -1.14571743e-01 4.91566479e-01 6.43951595e-01 1.28781617e+00 3.34867328e-01 9.98324156e-02 1.15798391e-01 6.18042827e-01 1.34615451e-01 6.02152348e-01 4.78330046e-01 -1.36214182e-01 8.54214728e-01 7.39858568e-01 -5.49567163e-01 -5.85098743e-01 -7.77846098e-01 5.03089547e-01 9.64812696e-01 8.66431117e-01 -8.16256940e-01 -4.81056750e-01 -8.99375141e-01 -4.04633582e-01 3.09893757e-01 -6.03164136e-01 7.58654177e-02 -9.40014422e-01 -2.93863475e-01 2.69379675e-01 4.62995946e-01 8.09946537e-01 -1.03175735e+00 -4.55754638e-01 1.48643672e-01 -9.07322168e-01 -1.80598688e+00 -1.00258529e+00 -4.17194575e-01 -7.10658491e-01 -1.36908424e+00 -3.29142362e-01 -9.36250925e-01 7.50916481e-01 4.61866051e-01 1.07511568e+00 3.49073619e-01 7.49285566e-03 4.82588440e-01 -6.21473908e-01 3.07634175e-01 9.77387354e-02 -2.01386422e-01 -1.58755541e-01 3.37152570e-01 1.15843095e-01 -4.44202840e-01 -6.83293223e-01 5.45451343e-01 -9.61553037e-01 4.58825469e-01 1.41167328e-01 4.64533031e-01 1.00312805e+00 4.01760966e-01 1.78214610e-01 -7.13117898e-01 1.91975564e-01 -5.77493012e-01 -4.39029098e-01 4.41544265e-01 2.47701719e-01 -2.91153222e-01 5.63368440e-01 -7.14572012e-01 -7.20124245e-01 3.31907645e-02 2.80011982e-01 -1.01793337e+00 1.21286221e-01 7.67803729e-01 -2.33292863e-01 2.99143970e-01 3.39378297e-01 4.18770581e-01 -4.55144346e-01 -1.33147746e-01 3.01058352e-01 5.31184971e-02 8.30880880e-01 -6.80026889e-01 9.11536455e-01 6.24274254e-01 2.24008821e-02 -6.29033923e-01 -1.14036870e+00 -6.09009624e-01 -8.41974676e-01 -5.55434108e-01 1.31069255e+00 -9.03196573e-01 -6.92855358e-01 7.47649297e-02 -1.24977481e+00 -4.87167507e-01 -1.64625376e-01 2.72252023e-01 -6.78578436e-01 7.20227361e-01 -5.34005046e-01 -6.10780478e-01 4.05538641e-02 -1.05319738e+00 1.35091174e+00 -1.06398076e-01 -4.25898395e-02 -9.04943049e-01 -3.20047468e-01 6.13096297e-01 -4.16147947e-01 5.96357465e-01 7.17948973e-01 -2.64101356e-01 -1.07994652e+00 9.59822163e-02 -2.99432188e-01 -2.81687915e-01 2.20709920e-01 -8.96256268e-02 -4.58682358e-01 -6.95953965e-02 7.32672140e-02 -1.84621319e-01 6.58284366e-01 2.13587642e-01 1.37865400e+00 -7.64277041e-01 -3.88770133e-01 6.58165395e-01 1.16355002e+00 1.62357673e-01 3.88385981e-01 1.85627848e-01 1.20375240e+00 6.42642140e-01 1.02999079e+00 2.54510909e-01 7.13157952e-01 7.64901936e-01 5.22519648e-01 1.79639995e-01 -1.19861372e-01 -6.63627088e-01 2.74500400e-01 1.07717884e+00 -2.93688744e-01 -4.77605283e-01 -8.83940816e-01 6.40953243e-01 -2.45702553e+00 -1.26563883e+00 -3.07451218e-01 1.63926697e+00 5.50548196e-01 5.95788918e-02 2.43392691e-01 -1.93533581e-02 8.25256050e-01 4.48027521e-01 -5.73347092e-01 6.11478746e-01 -3.65605772e-01 -5.52231312e-01 -5.49804866e-02 3.01953733e-01 -1.25568211e+00 1.03767192e+00 5.15902185e+00 8.41957510e-01 -9.04416680e-01 2.15833753e-01 5.98377585e-01 -2.35368535e-01 -3.67505819e-01 5.99333160e-02 -3.96690339e-01 5.96277475e-01 2.97436923e-01 -1.11102812e-01 4.15104896e-01 5.46267629e-01 3.42018217e-01 -1.19612202e-01 -1.16685092e+00 1.24752200e+00 4.15307969e-01 -1.49125588e+00 1.53970197e-01 -3.57590437e-01 5.39734423e-01 -2.07316205e-01 -1.68456823e-01 1.47350550e-01 -1.91124901e-01 -7.61752605e-01 1.16176236e+00 5.21919250e-01 9.54301000e-01 -4.14682418e-01 3.52024406e-01 5.36491036e-01 -1.85821593e+00 -2.91957501e-02 -1.34274706e-01 1.77282557e-01 2.86967188e-01 2.65186310e-01 -3.80725652e-01 9.58883643e-01 8.23310256e-01 1.11990666e+00 -5.34814775e-01 7.38779068e-01 -2.99071789e-01 3.94047588e-01 -3.66234422e-01 7.41373077e-02 3.83735985e-01 -4.50049758e-01 7.47526824e-01 9.38585162e-01 4.13154691e-01 7.16180325e-01 4.52501923e-01 1.00788331e+00 2.50585433e-02 -1.04695139e-03 -7.88110554e-01 -4.80860546e-02 4.33465868e-01 1.07400727e+00 -1.08218813e+00 -4.18819010e-01 -6.41258419e-01 1.08578682e+00 3.53747606e-01 6.69578016e-01 -1.10626853e+00 -6.65997341e-02 3.77244204e-02 3.84107739e-01 4.50301707e-01 -4.05876696e-01 1.89642772e-01 -1.57252491e+00 4.58842635e-01 -7.15596139e-01 7.68237710e-01 -1.31700218e+00 -1.04662240e+00 6.31077945e-01 3.69944751e-01 -1.40241289e+00 -1.38648778e-01 -1.57839522e-01 -4.25947636e-01 1.93036377e-01 -1.15047240e+00 -1.49227107e+00 -6.16952181e-01 1.07812309e+00 8.25805187e-01 1.19187154e-01 1.76828474e-01 3.12888920e-01 -6.56078160e-01 9.55082029e-02 -7.02806294e-01 3.76604617e-01 2.50184029e-01 -8.85932505e-01 2.88842320e-02 1.07272887e+00 4.48490769e-01 5.32594442e-01 5.82221746e-01 -9.63666081e-01 -1.74166858e+00 -1.49753630e+00 7.31547415e-01 -6.13694608e-01 9.07393217e-01 -6.31690264e-01 -9.96019244e-01 9.89141047e-01 -5.14539368e-02 4.32885975e-01 -1.69089865e-02 -2.81255633e-01 -2.18422681e-01 6.82078823e-02 -7.51921654e-01 5.93781114e-01 1.61780214e+00 -8.24437261e-01 -6.92143857e-01 6.75086975e-01 1.30384171e+00 -9.57831562e-01 -6.22087657e-01 4.20045137e-01 7.11233616e-02 -8.77157629e-01 1.00815010e+00 -5.10304391e-01 6.62397921e-01 -6.73231721e-01 -1.40105888e-01 -6.71827018e-01 -1.40198320e-01 -8.97903919e-01 -5.74662626e-01 1.21522284e+00 3.04690599e-02 -6.96329698e-02 6.10166669e-01 3.60561877e-01 -2.84994602e-01 -9.26947117e-01 -9.97661650e-01 -8.24453115e-01 -6.86431587e-01 -7.48722732e-01 6.47664726e-01 1.08198142e+00 1.35314940e-02 3.49261522e-01 -4.50768262e-01 3.37588131e-01 2.33473003e-01 5.38585305e-01 7.89772034e-01 -5.79575181e-01 -3.22605669e-01 4.36674096e-02 -5.73516250e-01 -1.45890820e+00 1.78863242e-01 -6.84986889e-01 1.25601977e-01 -1.61817777e+00 3.45446944e-01 -5.81680611e-02 -8.57742727e-02 4.42062318e-01 -3.08514208e-01 1.21638671e-01 2.23005652e-01 3.98500144e-01 -1.37616730e+00 5.04818261e-01 1.58126378e+00 -2.89147466e-01 -2.03237459e-01 -3.39810073e-01 -2.98434734e-01 7.75931239e-01 2.30352834e-01 -3.47846866e-01 -8.50453675e-01 -5.80605388e-01 5.12332380e-01 5.94954133e-01 7.45413423e-01 -7.88779020e-01 4.49211031e-01 -4.98958051e-01 8.23588818e-02 -1.03103113e+00 5.59296429e-01 -8.37904513e-01 2.24199012e-01 1.35340601e-01 -6.31765351e-02 3.81343812e-01 -3.79596241e-02 9.75371599e-01 -4.14912224e-01 1.21525548e-01 2.12252751e-01 -2.32645586e-01 -1.09801769e+00 7.98891306e-01 -6.85296282e-02 2.76458144e-01 1.33717144e+00 -3.70037794e-01 -4.93216991e-01 -3.42927217e-01 -8.99460614e-01 4.40012068e-01 3.55482668e-01 5.61061740e-01 1.02452850e+00 -1.52088702e+00 -3.44661146e-01 -1.67758048e-01 3.46662074e-01 5.55983841e-01 6.02086365e-01 1.14682078e+00 -4.86297876e-01 4.87282835e-02 3.68098736e-01 -8.08878660e-01 -1.33282590e+00 9.16594028e-01 2.57746875e-01 -3.01351994e-02 -1.09114540e+00 9.48860824e-01 6.70507193e-01 1.07440047e-01 1.78957209e-02 -6.61995828e-01 -2.96771199e-01 -3.84717481e-03 3.95831913e-01 -1.48204574e-02 -4.28752720e-01 -1.12377584e+00 -3.65275830e-01 8.34719002e-01 1.92511290e-01 3.47958207e-02 1.17604351e+00 -4.12877560e-01 -4.53705728e-01 8.43946636e-01 1.24088585e+00 -2.20226169e-01 -1.13198245e+00 -4.28258359e-01 -2.07274660e-01 -6.25304639e-01 -2.16154128e-01 -2.03255937e-01 -1.10053563e+00 4.42253768e-01 -2.36229077e-01 2.74480969e-01 1.17892337e+00 4.92779702e-01 1.04165721e+00 3.60715359e-01 4.83719558e-01 -9.10195470e-01 7.39013791e-01 4.15369451e-01 1.10061491e+00 -9.90202844e-01 1.46710604e-01 -9.15715396e-01 -6.23660386e-01 1.03185415e+00 8.35648537e-01 9.41103920e-02 4.49014753e-01 -2.35631354e-02 -3.54880035e-01 -7.28178859e-01 -8.11385930e-01 -2.59229630e-01 5.29572785e-01 3.34585398e-01 -2.44209915e-02 -2.80605406e-01 4.80686985e-02 5.67483544e-01 -1.44512787e-01 -1.20114964e-02 2.72565067e-01 8.91733170e-01 -1.52761236e-01 -4.76097584e-01 -2.56465465e-01 3.08887213e-01 -1.66338846e-01 9.83821973e-02 -4.77893591e-01 8.42163742e-01 4.43909824e-01 9.42708910e-01 1.05725266e-01 -5.54064751e-01 3.56688499e-01 -3.16874295e-01 4.47764367e-01 -6.82908416e-01 -3.93668354e-01 3.55181575e-01 3.88203003e-02 -8.76618147e-01 -7.35280156e-01 -6.86477959e-01 -1.48482716e+00 -2.44763494e-03 -1.87377691e-01 8.16784203e-02 -9.95499268e-02 1.16495097e+00 3.46706897e-01 6.77349329e-01 3.92235845e-01 -5.73901176e-01 2.16978326e-01 -5.61132133e-01 -4.62626994e-01 7.14305580e-01 5.01708150e-01 -6.53142273e-01 -1.33810088e-01 4.54676002e-01]

[9.833141326904297, 0.7200238108634949]

c19c3409-c25f-4547-9ec0-1c111274ddfe

group-sparse-coding-for-image-denoising

2212.11501

https://arxiv.org/abs/2212.11501v1

https://arxiv.org/pdf/2212.11501v1.pdf

Group Sparse Coding for Image Denoising

Group sparse representation has shown promising results in image debulrring and image inpainting in GSR [3] , the main reason that lead to the success is by exploiting Sparsity and Nonlocal self-similarity (NSS) between patches on natural images, and solve a regularized optimization problem. However, directly adapting GSR[3] in image denoising yield very unstable and non-satisfactory results, to overcome these issues, this paper proposes a progressive image denoising algorithm that successfully adapt GSR [3] model and experiments shows the superior performance than some of the state-of-the-art methods.

['Fei Wu', 'Luoyu Chen']

2022-12-22

['image-inpainting']

['computer-vision']

[ 3.98971587e-01 -3.78577381e-01 1.61112756e-01 -1.38040677e-01 -5.73292732e-01 1.97008774e-02 1.71270728e-01 -5.06359696e-01 -1.12166777e-01 8.07349205e-01 7.02404320e-01 3.78931791e-01 -1.31937221e-01 -5.62229276e-01 -4.29643005e-01 -9.77838874e-01 -1.01968804e-02 -4.86080498e-01 1.33773565e-01 -5.73909044e-01 6.10824704e-01 4.16017324e-01 -1.48580503e+00 2.29867131e-01 9.88162160e-01 8.53091419e-01 5.52120864e-01 4.60267484e-01 -1.43358827e-01 9.33121562e-01 -4.75012690e-01 1.20777212e-01 3.59181821e-01 -9.16280568e-01 -5.83425045e-01 4.23076242e-01 3.35990250e-01 -1.69197708e-01 -7.19083309e-01 1.53410542e+00 6.20052099e-01 2.98359692e-01 4.24388319e-01 -8.40004146e-01 -1.29430687e+00 2.02031255e-01 -9.60622907e-01 5.06395996e-01 4.65212315e-01 -2.39006802e-01 2.86176771e-01 -7.40029454e-01 6.73321187e-01 1.29772317e+00 9.13971424e-01 4.45090264e-01 -1.04234862e+00 -4.08785939e-01 -1.70794576e-01 4.92986321e-01 -1.61000657e+00 -4.82231468e-01 1.12804389e+00 7.09242523e-02 6.75458968e-01 4.70044494e-01 5.49680471e-01 7.01843441e-01 3.77427965e-01 5.87364554e-01 1.60788190e+00 -4.90038723e-01 3.12484484e-02 -2.77918160e-01 9.97235775e-02 4.86295491e-01 9.88859013e-02 9.07041803e-02 -6.65599883e-01 -1.56968743e-01 1.20231295e+00 2.37722158e-01 -6.83521092e-01 2.47050017e-01 -9.51295555e-01 7.38922238e-01 5.66689134e-01 8.03378880e-01 -6.41268313e-01 2.64883377e-02 2.12220386e-01 5.88506401e-01 6.77876651e-01 1.00912362e-01 2.37086639e-02 6.45041168e-02 -1.19871068e+00 1.60848215e-01 1.84333131e-01 6.15289986e-01 7.97735095e-01 4.99985218e-01 6.68623522e-02 1.19240427e+00 1.68550625e-01 4.36402678e-01 8.58163476e-01 -1.24716020e+00 -3.32486108e-02 1.99129522e-01 -1.26540020e-01 -1.64032793e+00 1.82813615e-01 -4.95732218e-01 -1.60126567e+00 2.20395580e-01 -1.69894800e-01 4.18945491e-01 -9.73199248e-01 1.36315417e+00 9.40236971e-02 4.34232295e-01 1.36016265e-01 1.17653692e+00 1.06804371e+00 1.03739214e+00 -2.63527900e-01 -5.81658483e-01 8.82086873e-01 -1.01781285e+00 -1.31504738e+00 -1.73050404e-01 9.80082620e-03 -1.19540358e+00 7.08570242e-01 6.28795505e-01 -1.21364820e+00 -8.29662979e-01 -8.56297910e-01 -1.83408499e-01 6.98013455e-02 -2.31940731e-01 6.03859007e-01 4.22020078e-01 -1.30066454e+00 8.66895974e-01 -3.86161953e-01 -5.14000237e-01 3.51385504e-01 -3.98529135e-02 -6.61245286e-01 -6.49657726e-01 -1.07162476e+00 8.92127454e-01 6.03970326e-02 3.39463741e-01 -6.46030009e-01 -3.36141914e-01 -6.00715339e-01 -1.97767094e-01 3.79282832e-02 -6.59696758e-01 2.92956948e-01 -1.12510717e+00 -1.48441923e+00 8.21695447e-01 -3.96820128e-01 -2.43020907e-01 1.50257781e-01 -1.29222974e-01 -5.48930466e-01 3.72954756e-01 1.41990334e-01 2.77339637e-01 1.35556364e+00 -1.40197217e+00 -3.44033651e-02 -3.84753466e-01 -5.41338503e-01 2.93313086e-01 -2.30358452e-01 2.65403062e-01 -3.16458732e-01 -1.32895267e+00 8.78272235e-01 -5.50039411e-01 -5.04088223e-01 -2.72749305e-01 -9.28126872e-02 1.27574578e-01 9.84567642e-01 -1.42728662e+00 1.24794865e+00 -2.30410480e+00 4.68104303e-01 -6.40785918e-02 8.18240643e-02 2.95872509e-01 -4.04496640e-01 5.17283440e-01 -3.63141835e-01 -3.94615205e-03 -5.34669161e-01 -1.74631894e-01 -6.82490349e-01 4.86183017e-01 -3.68228555e-01 8.58146250e-01 -3.30628544e-01 5.84234715e-01 -6.54615223e-01 -6.16537988e-01 2.73543328e-01 6.75574958e-01 -2.74955362e-01 2.68518865e-01 4.00252372e-01 8.38047206e-01 -5.47810674e-01 7.37183750e-01 1.11817646e+00 6.31413907e-02 -5.06581485e-01 -7.01122880e-01 -2.82252491e-01 -3.03160042e-01 -1.32412577e+00 1.93794942e+00 -6.09312654e-02 4.87000853e-01 5.35176873e-01 -1.39166963e+00 1.04545438e+00 2.23538324e-01 5.95342755e-01 -7.40995109e-01 -4.38229889e-02 2.14536041e-01 -4.87083673e-01 -8.37101758e-01 4.58943427e-01 -3.31844717e-01 5.16693830e-01 1.44634634e-01 -1.37461379e-01 -3.07189256e-01 -1.36157930e-01 3.59579593e-01 1.17675591e+00 -4.18416550e-03 3.09286028e-01 -4.39869374e-01 8.08089733e-01 -4.55875471e-02 6.04815066e-01 6.89726055e-01 -1.40080929e-01 1.34013200e+00 -2.18602449e-01 -4.97624397e-01 -9.65058744e-01 -8.68443549e-01 1.63226888e-01 4.36632037e-01 5.03125072e-01 -2.73344964e-01 -7.03474104e-01 -2.65712529e-01 -3.74404639e-01 4.43426937e-01 -5.34348369e-01 -9.05597657e-02 -7.55285740e-01 -8.61539662e-01 2.48098925e-01 4.58450317e-02 1.09433210e+00 -1.05867672e+00 5.51959611e-02 2.95301318e-01 -6.25286639e-01 -8.05051744e-01 -5.47059834e-01 -3.37096393e-01 -1.29325354e+00 -9.01892066e-01 -1.18313289e+00 -1.03225422e+00 9.97755587e-01 1.07052815e+00 9.98633206e-01 5.85101426e-01 -4.85421091e-01 3.33994150e-01 -7.75096297e-01 3.37814242e-01 -4.47556645e-01 -4.89577502e-01 -1.04581296e-01 2.52400875e-01 -2.83920854e-01 -9.32035446e-01 -7.71828830e-01 4.52066958e-01 -1.39265299e+00 -1.74822077e-01 5.52512825e-01 9.04568791e-01 8.69539499e-01 4.80890125e-01 4.63451922e-01 -7.94482112e-01 7.29324400e-01 -2.32089177e-01 -1.42396510e-01 9.90225300e-02 -6.87017918e-01 -1.38058677e-01 5.41209161e-01 -2.90546149e-01 -1.08445978e+00 -2.47839168e-01 -3.01007569e-01 -6.22891665e-01 6.44334331e-02 5.74870586e-01 1.05523303e-01 -6.63267851e-01 6.13715529e-01 1.03038371e+00 2.64244288e-01 -7.26125181e-01 2.21895263e-01 5.63352585e-01 6.52121067e-01 -2.92808175e-01 1.00493717e+00 6.41643941e-01 1.18499786e-01 -1.04691577e+00 -7.02089846e-01 -6.44303322e-01 -2.14277044e-01 -1.87705785e-01 7.02957153e-01 -1.05484915e+00 -9.76328179e-02 8.06358337e-01 -1.05420899e+00 -3.24959271e-02 -3.37879270e-01 3.17673564e-01 -4.56419528e-01 1.07028782e+00 -8.09737921e-01 -5.06854951e-01 -5.11004031e-01 -8.99536610e-01 5.90026915e-01 1.63948238e-01 2.72460163e-01 -7.96816766e-01 6.83734119e-02 5.82367182e-01 1.06261384e+00 3.08193356e-01 4.07652974e-01 3.27192336e-01 -6.54614151e-01 -1.66062504e-01 -3.03451926e-01 8.79856765e-01 5.02382874e-01 -4.12633002e-01 -4.84135449e-01 -6.37873232e-01 8.80959451e-01 -1.50532305e-01 9.48156834e-01 6.59240425e-01 1.18031180e+00 -4.68524158e-01 -5.10415882e-02 9.32802975e-01 1.74764526e+00 -7.35267177e-02 1.42400956e+00 4.55238461e-01 5.52381456e-01 3.89430851e-01 6.43703520e-01 2.08768800e-01 -3.16938125e-02 5.33548415e-01 3.65200996e-01 -3.20104808e-01 -5.87337196e-01 -4.57272567e-02 4.25577134e-01 1.17623365e+00 -3.60234380e-01 -2.76994985e-02 -2.00224981e-01 5.18357694e-01 -1.90660012e+00 -1.22020686e+00 -4.84267145e-01 1.73847246e+00 8.74915779e-01 -5.38666964e-01 -4.20785457e-01 7.60604814e-02 7.99256444e-01 7.68776774e-01 -1.59661993e-01 -1.26587242e-01 -7.36495078e-01 2.36120149e-01 7.08135247e-01 5.57275891e-01 -7.83710718e-01 9.86455083e-01 7.13935518e+00 1.34134972e+00 -9.82085526e-01 3.69499803e-01 6.81255877e-01 3.78983915e-01 -3.10769230e-01 9.69999507e-02 -2.51366317e-01 5.22228360e-01 2.95553088e-01 -1.40913025e-01 9.86165762e-01 3.23747188e-01 4.79500383e-01 -2.29787931e-01 -1.84159458e-01 1.52887285e+00 6.14616692e-01 -1.29003704e+00 1.27219066e-01 -1.85653716e-01 1.16158509e+00 -1.80046842e-01 4.45331968e-02 -2.34286919e-01 -6.43972158e-02 -1.05187035e+00 3.24067444e-01 7.39044011e-01 5.35467863e-01 -4.61159736e-01 7.21119463e-01 2.35933915e-01 -9.81977463e-01 1.83419898e-01 -9.17715073e-01 -1.28099658e-02 3.82927626e-01 1.05026162e+00 2.42797002e-01 7.04749882e-01 1.01647377e+00 1.13824785e+00 -5.00202239e-01 1.11032295e+00 -3.32112461e-01 7.61137784e-01 -2.24313021e-01 6.13339007e-01 8.61704201e-02 -8.20525706e-01 9.27562833e-01 8.44399214e-01 6.61896765e-01 6.71707213e-01 -2.05313563e-01 8.03662241e-01 1.28402248e-01 3.20597649e-01 -6.88571036e-01 2.17402250e-01 -1.46867751e-04 1.21423042e+00 -5.79502344e-01 -2.63313234e-01 -2.77180284e-01 1.62118840e+00 -1.49394959e-01 5.32892287e-01 -4.12668616e-01 -2.07266867e-01 2.69641489e-01 2.56474614e-01 3.05142969e-01 -3.29813302e-01 -2.97782868e-01 -1.35772204e+00 -3.15561071e-02 -1.14942431e+00 1.99592531e-01 -1.08442736e+00 -1.49411118e+00 7.07944214e-01 -1.98378533e-01 -1.38397682e+00 1.04880407e-01 -5.97054996e-02 -6.51272774e-01 1.02162516e+00 -1.67068923e+00 -1.20511496e+00 -5.49458742e-01 1.03597176e+00 6.29668832e-01 -3.37045968e-01 5.68965197e-01 2.97885627e-01 -3.24669182e-01 5.18001318e-02 3.12470943e-01 -4.44756031e-01 8.03167939e-01 -7.79825449e-01 1.91783115e-01 1.11142504e+00 -1.61410227e-01 7.04467833e-01 1.01267016e+00 -9.16430116e-01 -1.60655546e+00 -9.08385754e-01 5.87354541e-01 3.01855594e-01 2.55061984e-01 4.87166166e-01 -1.19224167e+00 3.39463949e-01 5.66268325e-01 3.78181368e-01 2.74853468e-01 -5.67652225e-01 -1.76350370e-01 -2.34105781e-01 -1.52560997e+00 4.65501159e-01 1.15283394e+00 -2.89378285e-01 -6.68620706e-01 3.38680774e-01 3.81208181e-01 -1.06215790e-01 -7.44217753e-01 3.52380246e-01 -2.99472716e-02 -1.13417673e+00 1.38510835e+00 7.22944438e-02 3.68514508e-01 -5.72260737e-01 -4.88348335e-01 -1.40301037e+00 -6.98818624e-01 -8.50070655e-01 1.14543289e-01 1.27604735e+00 -3.85007262e-01 -7.37088084e-01 7.08520889e-01 4.81398813e-02 -1.81417346e-01 -2.59356648e-01 -9.93773639e-01 -7.16140568e-01 -3.77365649e-01 2.97657669e-01 1.20660752e-01 1.16949415e+00 -5.00631511e-01 -1.37319297e-01 -8.90050173e-01 1.58634037e-01 1.23601568e+00 1.21432073e-01 4.82092232e-01 -7.61793673e-01 -1.19475730e-01 -5.85330203e-02 -4.91806030e-01 -1.10263097e+00 -1.16480514e-01 -5.96128523e-01 1.15308098e-01 -1.69993973e+00 2.49215975e-01 -2.15257391e-01 -3.35537940e-01 2.78226048e-01 -3.42390150e-01 7.39596665e-01 1.12256117e-01 6.59878790e-01 -5.42351663e-01 7.92043746e-01 1.53372490e+00 -1.74107850e-01 -8.64028484e-02 -3.60651374e-01 -7.56413102e-01 5.68319261e-01 7.21157849e-01 -4.61460233e-01 -1.71997249e-01 -4.55803841e-01 8.80102534e-03 2.13228151e-01 3.25599372e-01 -9.47951198e-01 2.84341663e-01 -1.92684889e-01 2.77773142e-01 -5.10634363e-01 1.98912084e-01 -8.16097140e-01 5.34982681e-01 3.19664925e-01 -1.52823329e-01 -3.67562324e-02 -8.23678747e-02 6.76975727e-01 -9.17416275e-01 -4.13217276e-01 1.05048585e+00 -6.98274672e-01 -7.65356243e-01 1.79288477e-01 -2.84444958e-01 -1.91383302e-01 6.94143116e-01 -5.25092125e-01 -7.06354994e-03 -6.29644752e-01 -5.31496704e-01 -3.62446129e-01 5.78133523e-01 1.48548618e-01 1.17943859e+00 -1.28828740e+00 -1.00338364e+00 5.23337841e-01 -4.58776236e-01 -1.45295307e-01 6.17882073e-01 8.63107502e-01 -8.33728492e-01 -2.37288803e-01 -4.53996152e-01 -3.21157157e-01 -1.34584427e+00 4.77045029e-01 1.02145709e-01 -6.95750341e-02 -8.22797120e-01 8.44171405e-01 1.04680255e-01 7.80389532e-02 -9.00281072e-02 3.60789865e-01 -3.22847605e-01 -3.09986264e-01 6.54198468e-01 5.66428542e-01 -9.65384021e-02 -8.09516013e-01 -4.60383967e-02 1.08203530e+00 -8.29648003e-02 1.41675204e-01 1.74320340e+00 -6.68926120e-01 -8.77700090e-01 -5.01132123e-02 1.38688850e+00 -1.96833108e-02 -9.99980569e-01 -2.55797476e-01 -5.70192993e-01 -1.05703449e+00 6.77616298e-01 -4.67337847e-01 -1.56999409e+00 4.20597464e-01 7.68476903e-01 3.14087391e-01 1.62608957e+00 -3.05682421e-01 1.21717942e+00 2.91612912e-02 5.76529562e-01 -8.35944295e-01 2.46180177e-01 1.61740229e-01 1.47272468e+00 -1.15210795e+00 4.55441624e-01 -7.56359875e-01 -3.21297586e-01 1.01281929e+00 2.24782825e-02 -6.58508182e-01 6.94344103e-01 4.64295372e-02 1.28743201e-01 -8.34912509e-02 -1.37283966e-01 8.90511647e-02 -1.09441634e-02 6.65677845e-01 2.40900695e-01 -5.76355219e-01 -9.08062637e-01 2.49907404e-01 4.11070555e-01 3.30684960e-01 4.97806102e-01 1.01951909e+00 -4.40815628e-01 -1.13796496e+00 -9.98844624e-01 1.77650377e-01 -7.22898304e-01 -2.18660414e-01 1.04968943e-01 2.26879582e-01 -7.43522793e-02 1.08327639e+00 -5.73363125e-01 -2.66302258e-01 2.04984799e-01 -5.54563940e-01 5.67441583e-01 -2.00528279e-01 -4.77401555e-01 3.34345251e-01 -3.98914605e-01 -7.41161406e-01 -8.80812645e-01 -4.59676355e-01 -9.18960869e-01 -3.67588580e-01 -2.11858094e-01 2.53120780e-01 5.02633810e-01 4.60875243e-01 3.42285335e-01 2.84962356e-01 1.00680339e+00 -1.04515553e+00 -4.03212547e-01 -1.07369661e+00 -1.08024240e+00 8.66933942e-01 3.32701892e-01 -2.70643562e-01 -8.92709374e-01 4.11696106e-01]

[11.393348693847656, -2.4571340084075928]

fa86a5f9-e589-4375-8d29-a20b665efc45

an-empirical-investigation-of-3d-anomaly

2203.05550

https://arxiv.org/abs/2203.05550v3

https://arxiv.org/pdf/2203.05550v3.pdf

Back to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection

Despite significant advances in image anomaly detection and segmentation, few methods use 3D information. We utilize a recently introduced 3D anomaly detection dataset to evaluate whether or not using 3D information is a lost opportunity. First, we present a surprising finding: standard color-only methods outperform all current methods that are explicitly designed to exploit 3D information. This is counter-intuitive as even a simple inspection of the dataset shows that color-only methods are insufficient for images containing geometric anomalies. This motivates the question: how can anomaly detection methods effectively use 3D information? We investigate a range of shape representations including hand-crafted and deep-learning-based; we demonstrate that rotation invariance plays the leading role in the performance. We uncover a simple 3D-only method that beats all recent approaches while not using deep learning, external pre-training datasets, or color information. As the 3D-only method cannot detect color and texture anomalies, we combine it with color-based features, significantly outperforming previous state-of-the-art. Our method, dubbed BTF (Back to the Feature) achieves pixel-wise ROCAUC: 99.3% and PRO: 96.4% on MVTec 3D-AD.

['Yedid Hoshen', 'Eliahu Horwitz']

2022-03-10

['rgb-3d-anomaly-detection-and-segmentation', '3d-anomaly-detection-and-segmentation', 'depth-anomaly-detection-and-segmentation', '3d-anomaly-detection', 'rgb-depth-anomaly-detection-and-segmentation']

['methodology', 'methodology', 'methodology', 'methodology', 'methodology']

[ 2.07695439e-01 -2.53248870e-01 1.46361306e-01 -1.08805798e-01 -7.28223026e-01 -7.52306938e-01 6.28227115e-01 2.63552666e-01 -2.48579815e-01 1.13069631e-01 -3.77744615e-01 -7.78746963e-01 7.89691135e-02 -6.18196011e-01 -7.20073283e-01 -6.80589557e-01 -2.45017648e-01 1.57934755e-01 4.75258142e-01 -2.27248698e-01 5.42754292e-01 9.56054509e-01 -1.57448554e+00 4.50809896e-02 9.11646068e-01 1.49567103e+00 -5.59892714e-01 6.94607913e-01 -2.81110495e-01 2.50779957e-01 -5.06257534e-01 -1.83633655e-01 6.72738731e-01 -3.38151217e-01 -4.56615835e-01 2.22418815e-01 9.32625592e-01 -6.09536648e-01 -1.09238684e-01 9.37124193e-01 3.37848306e-01 -2.11746112e-01 7.68539906e-01 -1.34237242e+00 -7.48046517e-01 -3.99663597e-01 -1.03665614e+00 5.06049812e-01 2.94051826e-01 2.57769406e-01 8.33029449e-01 -8.46431375e-01 3.71083647e-01 1.08717787e+00 9.13650870e-01 4.24668700e-01 -1.23752260e+00 -2.64391571e-01 2.56351680e-01 2.68173397e-01 -9.85180616e-01 -6.42790422e-02 9.74636018e-01 -5.63626170e-01 1.02588379e+00 3.41590941e-01 5.90114057e-01 1.14573872e+00 1.35850087e-01 8.49633336e-01 1.40857840e+00 -5.73972344e-01 1.21640906e-01 -4.23232138e-01 2.88545946e-03 8.75379801e-01 7.02219427e-01 2.61829704e-01 -3.18514615e-01 -2.51419276e-01 6.64825797e-01 1.95126794e-02 -6.99529331e-03 -6.11122072e-01 -8.98080528e-01 6.96317911e-01 1.67343080e-01 8.24364275e-02 -1.40619218e-01 2.17673883e-01 3.77957463e-01 5.64755201e-01 7.92730629e-01 5.22435725e-01 -5.69176555e-01 -2.14657396e-01 -8.39703739e-01 1.10718496e-01 2.43151963e-01 4.29348528e-01 6.24346316e-01 3.53877753e-01 7.17633218e-02 7.05873668e-01 1.89020559e-01 6.26791060e-01 2.04512358e-01 -8.86596203e-01 -2.05998272e-02 8.89749765e-01 2.94445804e-03 -1.01039648e+00 -7.59732902e-01 -2.78629184e-01 -6.28137231e-01 9.30731654e-01 9.18093741e-01 1.05989814e-01 -1.49822628e+00 1.29284525e+00 3.16095710e-01 -6.37709582e-03 -3.47534239e-01 8.28963816e-01 4.64791268e-01 4.12047058e-02 -2.84046054e-01 2.03066915e-01 9.81214464e-01 -5.44160843e-01 -2.22299501e-01 -2.18466431e-01 7.69339204e-01 -8.41224551e-01 1.07443750e+00 8.12401891e-01 -7.73002088e-01 -2.47344688e-01 -1.19301677e+00 1.64583251e-01 -7.46577382e-01 -7.20720664e-02 7.01724946e-01 1.04270315e+00 -1.09315014e+00 6.04092658e-01 -7.77545393e-01 -6.10670090e-01 5.91820240e-01 -2.93588098e-02 -6.29133105e-01 -2.51277953e-01 -6.50502622e-01 8.11337888e-01 2.67292447e-02 -5.27547225e-02 -6.19564474e-01 -4.81011599e-01 -8.64034593e-01 -4.68787789e-01 4.65262979e-01 -3.39619845e-01 9.00519967e-01 -9.79045928e-01 -1.18805826e+00 1.31256294e+00 -2.17467025e-02 -4.80414212e-01 7.34003365e-01 -5.07058322e-01 -3.12063366e-01 3.91196132e-01 -5.19425198e-02 2.99243033e-01 1.07784927e+00 -1.46885550e+00 -7.17832386e-01 -5.08987367e-01 -4.95960936e-02 -1.94430470e-01 -1.58104047e-01 -1.66108489e-01 -5.99011302e-01 -7.31451392e-01 5.88967979e-01 -9.04642761e-01 -7.65314922e-02 4.45551842e-01 -7.08525002e-01 -2.33073775e-02 9.70669329e-01 -6.38503134e-01 8.27630997e-01 -2.18432093e+00 -4.97299373e-01 5.50386310e-01 2.58175313e-01 4.74679172e-01 -2.30326772e-01 7.88244903e-02 -2.21196994e-01 3.60749900e-01 -5.37255585e-01 -2.18002692e-01 3.03241927e-02 1.10423207e-01 -2.71474779e-01 7.57802069e-01 6.69591904e-01 6.15892649e-01 -8.03002477e-01 -2.40921348e-01 5.10894299e-01 2.29996487e-01 -4.86216903e-01 -1.03686072e-01 -8.93178806e-02 3.37066114e-01 -2.94025302e-01 1.34478092e+00 1.13079321e+00 4.01451662e-02 -3.92408758e-01 9.77781340e-02 -7.12819844e-02 -1.85952887e-01 -8.47949743e-01 1.56663966e+00 3.88669558e-02 7.60736585e-01 -2.61362702e-01 -1.26731944e+00 1.03477192e+00 -1.68554485e-01 7.33722925e-01 -1.13928294e+00 -1.67591467e-01 5.78396082e-01 -6.90460652e-02 -5.49788654e-01 2.14370966e-01 3.04602742e-01 7.15739504e-02 3.06360245e-01 -1.69418052e-01 -5.65230250e-02 -6.43932968e-02 1.28938004e-01 1.54287183e+00 4.31399465e-01 1.28516659e-01 -6.68472797e-02 3.24057549e-01 6.17352985e-02 4.22823101e-01 1.23904407e+00 -5.97328842e-01 1.09861875e+00 1.00362825e+00 -5.80691218e-01 -1.06698394e+00 -1.25575244e+00 -2.26153091e-01 7.62762964e-01 2.10772038e-01 -1.11413814e-01 -4.74973112e-01 -1.15233660e+00 4.66756403e-01 5.06765664e-01 -9.86922204e-01 -1.93587422e-01 -3.63473892e-01 -9.91410196e-01 5.79268217e-01 5.64278960e-01 4.48990911e-01 -5.10299027e-01 -7.16599882e-01 -1.12453111e-01 3.05569410e-01 -1.18601441e+00 -3.25310975e-02 2.41787866e-01 -1.05109739e+00 -1.28456783e+00 -6.27992570e-01 -1.02110758e-01 6.53904378e-01 3.25095385e-01 1.12831759e+00 3.63021553e-01 -5.28766692e-01 8.55791628e-01 -6.99120820e-01 -4.88200098e-01 -1.12016305e-01 -1.90370575e-01 -2.29254887e-01 4.36259024e-02 5.65567672e-01 -5.27502775e-01 -7.99002349e-01 1.84512317e-01 -7.87885666e-01 -3.48288357e-01 8.83209705e-01 7.95357049e-01 5.37558138e-01 -3.87607247e-01 2.47457370e-01 -7.38290727e-01 2.12656155e-01 -2.97692627e-01 -5.55827200e-01 -7.12227076e-02 -7.78556466e-01 3.06128580e-02 3.31945419e-01 -1.23373970e-01 -5.40544748e-01 2.51331180e-01 -2.35476226e-01 -5.62875390e-01 -6.26834691e-01 4.23053689e-02 2.04436649e-02 -2.46057242e-01 6.84671938e-01 2.00208560e-01 1.24407537e-01 -6.63694322e-01 1.72130540e-01 2.29753926e-01 6.82377577e-01 -5.21828711e-01 1.02061808e+00 9.97261345e-01 4.07649726e-01 -8.96317899e-01 -7.95745492e-01 -5.05706370e-01 -7.58277535e-01 -2.07721561e-01 9.03462589e-01 -4.71693307e-01 -3.45785856e-01 8.72741580e-01 -1.02209210e+00 -2.90663213e-01 -2.30277047e-01 1.06483862e-01 -4.87052500e-01 6.69694185e-01 -2.13567138e-01 -1.01224840e+00 -1.06082894e-01 -9.39894676e-01 1.35366559e+00 1.70259140e-02 9.99068469e-02 -7.95184374e-01 -1.24176890e-02 1.48415659e-03 4.69431967e-01 1.08339393e+00 1.01929474e+00 -9.38246727e-01 -4.27895993e-01 -5.24988294e-01 -6.14218056e-01 4.46232945e-01 1.85335353e-01 3.18666905e-01 -1.13473690e+00 -3.05614080e-02 -3.86482030e-01 -2.86667924e-02 1.26172531e+00 4.67464149e-01 1.45500994e+00 2.78485209e-01 -8.13371465e-02 7.37638354e-01 1.42452860e+00 1.66475803e-01 6.77610993e-01 8.26938391e-01 8.38270724e-01 4.01688457e-01 4.00486737e-01 3.86304587e-01 1.65108725e-01 6.53416336e-01 8.64421248e-01 -5.29317558e-01 -1.41869530e-01 7.20814541e-02 3.00418139e-01 -7.46488199e-02 -2.57349908e-01 -7.01898187e-02 -1.17164350e+00 5.16901374e-01 -1.72101152e+00 -7.14332402e-01 -3.28704119e-01 2.28925419e+00 1.96825892e-01 5.61178148e-01 3.49042922e-01 3.97574693e-01 3.00909042e-01 5.18900603e-02 -7.50499845e-01 -5.01582503e-01 -3.99120450e-01 1.63710907e-01 6.36109829e-01 2.13931371e-02 -1.47132909e+00 6.74676657e-01 6.50154686e+00 5.76274157e-01 -1.14332724e+00 -2.15475738e-01 7.73192823e-01 8.45141187e-02 -2.30794743e-01 -1.17415607e-01 -2.40307108e-01 2.87102759e-01 4.84844446e-01 4.87925917e-01 -6.80919215e-02 7.61401594e-01 1.42238423e-01 -4.52326387e-01 -9.86665785e-01 9.67611492e-01 3.66137415e-01 -1.03090847e+00 5.28534055e-02 1.76193506e-01 5.76666415e-01 1.56172857e-01 2.64241576e-01 -3.20719182e-02 -4.53108503e-03 -1.06700981e+00 7.03128874e-01 6.00984871e-01 7.91441798e-01 -6.99001253e-01 7.32370257e-01 -3.03335845e-01 -8.59564304e-01 -5.67249656e-02 -9.04286355e-02 2.00067714e-01 -1.55272707e-01 8.60286057e-01 -5.00669003e-01 6.29825771e-01 1.15221405e+00 7.63036013e-01 -1.24136436e+00 1.31134593e+00 -3.97521406e-02 6.71416640e-01 -5.57369709e-01 5.06604552e-01 4.79815125e-01 -6.16177879e-02 7.86991954e-01 1.23338449e+00 4.82461452e-01 -4.27497894e-01 2.53558066e-02 7.10077167e-01 3.02597970e-01 -7.75620192e-02 -9.23543334e-01 1.41524225e-01 -2.03198083e-02 1.05119634e+00 -1.00501239e+00 7.34099187e-03 -7.91036189e-01 1.27165508e+00 -3.91725563e-02 4.92034227e-01 -5.48076153e-01 -4.22334820e-01 7.35669851e-01 1.52233690e-01 4.67409968e-01 -4.44453448e-01 -5.60544252e-01 -1.00563085e+00 1.43148854e-01 -8.49767685e-01 5.32932818e-01 -6.64536655e-01 -1.56263793e+00 2.43448749e-01 -1.16847537e-01 -1.35276306e+00 -1.69641256e-01 -1.26267290e+00 -7.78665483e-01 4.55442280e-01 -1.85135067e+00 -1.12569177e+00 -4.58341569e-01 5.38317204e-01 2.27532521e-01 -2.37600669e-01 6.10083461e-01 9.16568264e-02 -6.55166686e-01 7.16205716e-01 2.37423539e-01 3.26974273e-01 8.20406556e-01 -1.73981595e+00 6.83401883e-01 1.31977057e+00 2.68651128e-01 2.94778217e-02 5.95721841e-01 -4.95639861e-01 -1.46829236e+00 -7.97625482e-01 1.12677999e-01 -8.50639701e-01 4.45784539e-01 -2.92905211e-01 -1.03873074e+00 3.26805562e-01 -2.38442779e-01 5.33159256e-01 4.76840287e-01 3.96329015e-02 -7.09738910e-01 -2.72697099e-02 -1.18184960e+00 4.80191857e-01 1.13901639e+00 -2.74728090e-01 -3.59959424e-01 -4.24051471e-02 4.51003432e-01 -3.88603210e-01 -3.30843449e-01 5.89790642e-01 5.05711854e-01 -1.40838063e+00 1.11176336e+00 -5.22453845e-01 2.60157436e-01 -3.54412049e-01 -3.49280745e-01 -1.10064304e+00 8.22347701e-02 -4.89855170e-01 -2.67927408e-01 8.72267902e-01 3.86223584e-01 -8.79407167e-01 7.81519890e-01 3.86649698e-01 -3.88405412e-01 -7.58700788e-01 -1.01793754e+00 -1.15047801e+00 3.64147723e-01 -9.10897732e-01 4.03943419e-01 8.66499066e-01 -4.32988077e-01 -5.88967443e-01 -1.33519113e-01 2.23564208e-01 7.33461916e-01 1.45612612e-01 8.08376312e-01 -1.50273538e+00 1.45758256e-01 -7.62337983e-01 -8.77369523e-01 -7.02012539e-01 -1.55885860e-01 -5.86719036e-01 -2.23808229e-01 -1.34776545e+00 -1.57434866e-01 -3.37267876e-01 -5.22965848e-01 7.62847126e-01 -1.61724985e-01 6.71494305e-01 -7.48267025e-02 -6.79725632e-02 -5.70948601e-01 2.07426310e-01 1.00990140e+00 -7.36934170e-02 -1.13038398e-01 -2.20731571e-01 -6.92706108e-01 9.65203643e-01 8.66279125e-01 -1.90669328e-01 6.11776602e-04 -1.58153594e-01 2.89047599e-01 -7.20432878e-01 7.43509829e-01 -1.10889125e+00 -2.92259634e-01 -1.13057509e-01 8.11488509e-01 -7.25754023e-01 9.71129164e-02 -7.63214767e-01 -7.01694310e-01 3.53883892e-01 1.87888563e-01 2.90322483e-01 6.09229386e-01 7.48274982e-01 -2.22378522e-02 6.69765472e-02 7.15518177e-01 -4.97091077e-02 -1.18018103e+00 3.30955982e-01 -4.13383424e-01 1.76355660e-01 9.35253143e-01 -5.61567783e-01 -3.72820377e-01 -3.33352089e-01 -4.26691711e-01 -6.37333915e-02 7.49095559e-01 4.93908316e-01 7.17901111e-01 -1.20593584e+00 -7.11027145e-01 5.12136698e-01 5.03648460e-01 -1.71335354e-01 1.53986245e-01 1.00649726e+00 -7.92120695e-01 8.97365361e-02 -3.44621927e-01 -1.14132583e+00 -1.02080441e+00 3.87383938e-01 5.42781234e-01 9.37355608e-02 -9.55833673e-01 5.93205214e-01 -6.55203387e-02 -2.58128196e-01 1.52851388e-01 -3.72362047e-01 8.91184285e-02 3.34773138e-02 2.57875383e-01 2.61584729e-01 3.75775099e-01 -3.50567490e-01 -6.09819770e-01 9.07934308e-01 -8.67155492e-02 -6.96292371e-02 1.26408648e+00 -6.51753545e-02 1.60332918e-01 4.45771664e-01 1.04186523e+00 -7.92658981e-03 -1.40950668e+00 -1.14518456e-01 2.25367621e-01 -8.46500814e-01 1.31960392e-01 -8.68505478e-01 -1.32654405e+00 1.00582743e+00 1.06759596e+00 5.09450138e-01 1.31563544e+00 8.57240334e-03 5.95655978e-01 3.01965237e-01 1.69090368e-02 -1.17671525e+00 3.89415860e-01 4.69749600e-01 7.15921223e-01 -1.59783542e+00 1.94609046e-01 -8.75644088e-02 -4.47938651e-01 1.28431249e+00 8.29959691e-01 -2.97143906e-01 6.29936576e-01 6.80679921e-03 2.87624776e-01 -3.24536920e-01 -2.61957496e-01 -6.59028530e-01 5.40294409e-01 9.43405092e-01 1.99723169e-01 -3.17421138e-01 1.47719145e-01 1.70700341e-01 1.67791381e-01 -5.26336908e-01 5.23468792e-01 9.64400768e-01 -4.32049155e-01 -9.70449090e-01 -6.03560090e-01 7.20287561e-01 -5.20542741e-01 2.05972612e-01 -6.74331129e-01 1.08864081e+00 1.84818074e-01 7.15023279e-01 3.81236583e-01 -3.32301885e-01 4.00074571e-01 2.32665420e-01 4.78189141e-01 -1.14760242e-01 2.87967511e-02 1.84950475e-02 -5.86547889e-02 -1.03751040e+00 -2.63980359e-01 -7.77558982e-01 -9.13937628e-01 -2.06805333e-01 -1.93188921e-01 -5.47123611e-01 6.59290075e-01 9.78826523e-01 3.53752524e-01 2.11319789e-01 5.96761286e-01 -8.97178471e-01 -1.36523694e-01 -4.97244418e-01 -5.05676210e-01 7.09653139e-01 9.31625187e-01 -8.25246811e-01 -7.17867672e-01 -2.41487786e-01]

[7.690794944763184, 2.095890522003174]

1724c9cf-6701-4449-865b-2823158d593c

improving-robustness-of-jet-tagging-1

2303.14511

https://arxiv.org/abs/2303.14511v1

https://arxiv.org/pdf/2303.14511v1.pdf

Improving robustness of jet tagging algorithms with adversarial training: exploring the loss surface

In the field of high-energy physics, deep learning algorithms continue to gain in relevance and provide performance improvements over traditional methods, for example when identifying rare signals or finding complex patterns. From an analyst's perspective, obtaining highest possible performance is desirable, but recently, some attention has been shifted towards studying robustness of models to investigate how well these perform under slight distortions of input features. Especially for tasks that involve many (low-level) inputs, the application of deep neural networks brings new challenges. In the context of jet flavor tagging, adversarial attacks are used to probe a typical classifier's vulnerability and can be understood as a model for systematic uncertainties. A corresponding defense strategy, adversarial training, improves robustness, while maintaining high performance. Investigating the loss surface corresponding to the inputs and models in question reveals geometric interpretations of robustness, taking correlations into account.

['Annika Stein']

2023-03-25

['jet-tagging']

['graphs']

[ 1.36462510e-01 1.51606545e-01 -3.20782922e-02 -4.41303134e-01 -1.02701819e+00 -8.47274721e-01 8.62540603e-01 5.22971809e-01 -2.79176444e-01 4.29702848e-01 1.40516222e-01 -6.05043173e-01 -5.58781028e-01 -8.71803403e-01 -8.14022005e-01 -1.08957934e+00 -1.73224151e-01 3.83895785e-01 1.70380726e-01 -1.16263770e-01 2.65690744e-01 9.10887420e-01 -8.31306577e-01 2.24171504e-01 3.22859257e-01 9.80553448e-01 -5.57581842e-01 3.92941773e-01 1.11963347e-01 3.79965603e-01 -5.04357338e-01 -8.36975038e-01 4.80733871e-01 -2.23434076e-01 -7.25056469e-01 -7.31673777e-01 1.70765296e-01 3.09599817e-01 -6.20603323e-01 1.43969917e+00 5.25664151e-01 1.21842004e-01 5.94119310e-01 -1.01063466e+00 -4.05309767e-01 7.36266851e-01 -3.45607996e-01 5.71025133e-01 -6.29501417e-02 4.37755078e-01 1.03773010e+00 -3.68398756e-01 4.62815136e-01 1.23874664e+00 7.09926426e-01 2.23012447e-01 -1.48066282e+00 -7.04712868e-01 -2.30338484e-01 1.28010288e-01 -1.03874969e+00 -3.91916871e-01 8.43021095e-01 -4.28025693e-01 5.79379320e-01 3.74117643e-01 -6.08374253e-02 1.34377098e+00 5.37187755e-01 4.17410880e-01 1.11399567e+00 -5.18209398e-01 3.09055448e-01 3.35377365e-01 2.15007231e-01 1.74523935e-01 2.40120620e-01 7.71549284e-01 -2.68755198e-01 -3.75993371e-01 2.92060196e-01 -5.22478044e-01 -6.47639707e-02 -5.01537383e-01 -7.81955242e-01 1.09289050e+00 7.08678126e-01 4.06295598e-01 8.03423524e-02 6.19958527e-02 7.01575756e-01 4.40428704e-01 4.67085391e-01 1.06479061e+00 -6.51807725e-01 1.61998436e-01 -7.76454329e-01 5.52819073e-01 4.30081397e-01 4.04311866e-01 2.37559542e-01 1.02133997e-01 -3.88619661e-01 5.34486651e-01 4.28160094e-02 2.61578739e-01 -1.12821884e-01 -6.11957669e-01 4.43305224e-01 3.10027421e-01 -1.42388582e-01 -9.01462376e-01 -8.02818537e-01 -1.04611647e+00 -8.11089337e-01 4.80558544e-01 7.56928205e-01 -5.75043894e-02 -7.27320731e-01 1.81018388e+00 2.66233325e-01 -2.99676567e-01 1.05730303e-01 8.11601400e-01 3.59842718e-01 4.66620088e-01 2.12877840e-01 6.15600459e-02 1.24906015e+00 -1.84381783e-01 -1.72036633e-01 -2.53967434e-01 3.67629558e-01 -6.35409594e-01 6.15950763e-01 4.89389241e-01 -9.35444653e-01 -3.70185018e-01 -1.06707335e+00 1.46455273e-01 -5.99862397e-01 -4.04032409e-01 8.06691766e-01 9.99340892e-01 -4.29349601e-01 1.13215363e+00 -6.55528367e-01 8.27712119e-02 2.35605270e-01 4.41058815e-01 -1.15162425e-01 6.54682666e-02 -1.48144567e+00 9.15045917e-01 2.50488877e-01 1.63217202e-01 -7.66813576e-01 -1.10660827e+00 -4.98135835e-01 3.56085807e-01 3.14920515e-01 -3.88195068e-01 1.13609838e+00 -5.47003448e-01 -1.07975924e+00 5.07103622e-01 3.66598845e-01 -6.96509480e-01 7.78966308e-01 1.10862315e-01 -8.79773855e-01 -6.91053271e-02 -1.29290536e-01 2.51483440e-01 8.73566568e-01 -1.07485068e+00 -1.06879391e-01 -1.62215099e-01 3.25890005e-01 -6.14946663e-01 1.17551878e-01 6.21362507e-01 1.80727050e-01 -7.33644247e-01 1.10234864e-01 -1.00005853e+00 -3.46851528e-01 -4.48655754e-01 -6.22636497e-01 1.25295855e-02 5.31657517e-01 -4.86465424e-01 8.85990024e-01 -2.16408014e+00 3.09265442e-02 6.39839232e-01 -3.45567055e-03 2.04103202e-01 3.10381353e-01 4.47772950e-01 -4.89923477e-01 2.49220565e-01 -4.05727059e-01 1.07448228e-01 2.78309166e-01 -1.90348193e-01 -4.48682666e-01 5.37385881e-01 4.12972778e-01 8.79393399e-01 -5.81846893e-01 1.92067146e-01 1.24015454e-02 1.52422115e-01 -6.27400815e-01 -1.83653265e-01 -3.12001258e-01 7.65088558e-01 -4.10948694e-01 4.79241997e-01 7.21948802e-01 -1.83134060e-02 -9.51340050e-02 -2.08308712e-01 -1.97132416e-02 3.67137760e-01 -8.16947103e-01 1.12929296e+00 -4.44653720e-01 5.67478716e-01 1.00378402e-01 -1.19181573e+00 6.50485337e-01 2.73990352e-03 3.47062975e-01 -9.96111989e-01 3.12743276e-01 7.75141716e-02 7.75937676e-01 -1.58609331e-01 4.05551106e-01 -6.45791709e-01 -5.17700732e-01 3.30088079e-01 -7.21693859e-02 2.21104026e-01 -3.33826154e-01 2.33322322e-01 1.22437847e+00 -2.65001476e-01 -1.38652129e-02 -4.69951212e-01 3.13707322e-01 -1.35609835e-01 5.28156817e-01 8.21389616e-01 4.53228503e-03 4.67797965e-01 8.37058723e-01 -4.25863624e-01 -1.17442596e+00 -1.18443489e+00 -6.47131801e-01 5.61820030e-01 -1.75893679e-01 -1.74985319e-01 -5.14642954e-01 -8.52436125e-01 3.83625805e-01 1.09888351e+00 -4.70691830e-01 -7.34512508e-01 -6.17062807e-01 -1.28053129e+00 7.49313951e-01 4.76597220e-01 1.97641224e-01 -8.28625202e-01 -1.79264888e-01 3.53605114e-02 4.42673624e-01 -1.01553261e+00 5.29573858e-02 6.07454419e-01 -5.30538440e-01 -9.57895100e-01 -1.50037542e-01 6.68862984e-02 3.86008561e-01 -2.44951010e-01 1.14083064e+00 -2.11349130e-01 -5.21347702e-01 6.05137236e-02 -2.52349734e-01 -5.56982696e-01 -9.29279327e-01 -1.99962050e-01 4.04949248e-01 -1.60054803e-01 2.55833149e-01 -5.58773339e-01 -3.46026570e-01 1.76526114e-01 -9.77061212e-01 -6.41242146e-01 6.18110359e-01 7.75083184e-01 9.33639854e-02 5.14216781e-01 7.12176204e-01 -8.90148520e-01 4.75032389e-01 -7.84806430e-01 -8.36507797e-01 1.03215039e-01 -5.66892147e-01 3.76751244e-01 8.40821028e-01 -1.58288434e-01 -1.10128999e+00 -4.94745940e-01 -4.23119068e-01 -3.36382180e-01 -3.24654430e-01 2.84151673e-01 -3.83375943e-01 -4.27524090e-01 8.42662275e-01 -3.98981601e-01 -3.81594270e-01 -5.86361110e-01 2.50417680e-01 1.27618805e-01 4.58809376e-01 -8.30548227e-01 1.22676468e+00 5.74361756e-02 4.82969165e-01 -4.61333483e-01 -9.73614991e-01 -1.23029381e-01 -4.88005280e-01 8.96140351e-04 5.32139599e-01 -4.79676664e-01 -6.90876842e-01 2.04549003e-02 -8.87777448e-01 2.17499688e-01 -3.60465199e-01 4.85765278e-01 -2.65272558e-01 1.54815182e-01 -3.37480515e-01 -7.36235261e-01 1.07949153e-01 -1.26555955e+00 7.35993862e-01 1.64802954e-01 -6.43361807e-02 -9.57448244e-01 -1.22446179e-01 1.32593632e-01 4.15644258e-01 4.31998849e-01 1.61580968e+00 -1.02257025e+00 -6.54325783e-01 -4.82926309e-01 -2.01813862e-01 3.64248782e-01 -3.13707858e-01 -2.86475807e-01 -1.33393919e+00 -3.20142329e-01 2.68428206e-01 -1.79487243e-01 8.52458954e-01 3.86594474e-01 1.53244352e+00 -2.02587366e-01 -3.19518924e-01 6.04445279e-01 1.21676385e+00 2.49772325e-01 4.20497000e-01 3.70853931e-01 2.84496397e-01 5.86296141e-01 8.38373080e-02 1.91461146e-02 -4.21855867e-01 8.55457962e-01 5.07457376e-01 -6.80430233e-02 1.43646613e-01 -1.27208680e-01 1.29557073e-01 2.47716427e-01 1.73392639e-01 -2.46980652e-01 -7.99689651e-01 -1.55001925e-02 -1.34228289e+00 -1.17438865e+00 -1.20880775e-01 2.29344916e+00 4.42912102e-01 9.54453766e-01 3.59473489e-02 3.52040738e-01 4.34612036e-01 2.34154359e-01 -7.73847699e-01 -6.58084393e-01 -9.41222757e-02 3.82208407e-01 6.87515438e-01 2.41029814e-01 -9.63523388e-01 4.27794039e-01 6.33038282e+00 1.02881289e+00 -1.10998607e+00 1.13273397e-01 9.00433838e-01 -3.09489310e-01 -4.40184355e-01 2.13995837e-02 -7.19482422e-01 6.03876293e-01 1.10663629e+00 -4.68323939e-02 3.21543515e-01 6.09917343e-01 1.01201557e-01 -7.77230337e-02 -1.22637928e+00 3.86285603e-01 -2.58105636e-01 -1.44037914e+00 -2.95482934e-01 2.33697906e-01 5.19730270e-01 2.28653867e-02 3.20956647e-01 6.17443442e-01 3.21505427e-01 -1.13099670e+00 7.05353498e-01 5.45485020e-01 5.38825393e-01 -1.05054188e+00 7.49711394e-01 3.34308505e-01 -6.25523746e-01 -1.66293398e-01 -3.91562432e-01 7.95725361e-02 9.90279540e-02 1.03453350e+00 -7.28227496e-01 7.97012746e-01 6.78523481e-01 -8.25339556e-02 -6.46838784e-01 9.79911208e-01 2.49276385e-02 8.89032423e-01 -2.36804187e-01 1.87647298e-01 3.31683397e-01 -1.06929436e-01 8.11986506e-01 1.22974324e+00 1.62275940e-01 -3.08778826e-02 -1.09013759e-01 1.12602627e+00 -1.63929224e-01 -3.49332392e-01 -7.74961233e-01 -1.93042010e-01 1.07840635e-01 1.21009064e+00 -8.79792690e-01 2.33505249e-01 -2.62465626e-01 3.79287273e-01 1.16394043e-01 1.78964630e-01 -8.38858902e-01 -2.33280137e-01 8.30927134e-01 5.87642305e-02 -5.43253347e-02 -2.55089611e-01 -3.53066593e-01 -7.06198275e-01 3.69363874e-02 -9.34293628e-01 3.00781608e-01 -3.13375920e-01 -1.38898861e+00 2.64638215e-01 1.52012557e-01 -1.10727835e+00 -1.26913786e-01 -8.69184911e-01 -1.03393066e+00 7.90065587e-01 -1.09791267e+00 -9.20160353e-01 4.55649763e-01 3.22747901e-02 2.91344315e-01 -3.74704540e-01 5.88455319e-01 2.88577855e-01 -5.06683528e-01 7.73875535e-01 4.98291641e-01 2.98134461e-02 5.08544505e-01 -1.25804603e+00 6.86991334e-01 9.12232518e-01 3.62029642e-01 6.64084375e-01 1.21948731e+00 -5.49343228e-01 -1.37047005e+00 -9.41036344e-01 5.82849383e-01 -5.83327115e-01 9.11704242e-01 -8.15154314e-01 -9.98042822e-01 3.21047425e-01 5.31538911e-02 2.24390551e-01 4.98255968e-01 3.18001717e-01 -4.14398551e-01 -4.71771657e-02 -1.43820882e+00 3.76955688e-01 8.95034194e-01 -6.87180519e-01 -3.14127654e-01 5.28233588e-01 6.60978019e-01 -4.09004331e-01 -8.63393068e-01 6.42515361e-01 3.63108367e-01 -1.06452417e+00 1.32903755e+00 -1.11045206e+00 3.21095586e-01 -1.21338323e-01 -8.61067995e-02 -1.59297049e+00 -5.59746027e-01 -6.18699253e-01 3.01047981e-01 1.19000459e+00 5.65270722e-01 -4.73903060e-01 7.01005638e-01 6.36906505e-01 -1.25250056e-01 -6.35475934e-01 -1.10449004e+00 -1.02778840e+00 5.46490908e-01 -8.42586994e-01 6.21942759e-01 8.68545294e-01 -2.32776731e-01 5.86680137e-02 -2.31527865e-01 8.27173412e-01 6.30552590e-01 4.08993252e-02 2.90182948e-01 -1.26182771e+00 -5.30443430e-01 -6.92717791e-01 -4.78991568e-01 -2.19379559e-01 2.10014656e-01 -1.33822036e+00 -2.16870517e-01 -6.09182179e-01 -1.21526726e-01 -7.91603327e-01 -5.30255556e-01 8.37866869e-03 4.75407206e-02 1.61435381e-01 2.25396857e-01 -2.98415750e-01 -9.72917899e-02 3.49221677e-01 6.51919901e-01 -3.80440682e-01 4.39139128e-01 5.52017152e-01 -8.80546331e-01 8.02930892e-01 8.63892496e-01 -7.36988902e-01 3.31847556e-02 -7.37529173e-02 7.47733951e-01 1.14365354e-01 6.66128874e-01 -1.16399801e+00 1.32999957e-01 -4.81800139e-02 8.32764924e-01 -2.68370390e-01 2.77394205e-01 -6.62824690e-01 2.64961541e-01 2.71166712e-01 -5.85866690e-01 4.03125063e-02 4.22327787e-01 5.53201199e-01 -1.25536621e-01 -7.83564985e-01 9.82024610e-01 -1.10986166e-01 -1.87406927e-01 1.98066711e-01 -1.21100187e-01 1.72002465e-02 8.40972841e-01 5.66542983e-01 -2.83128202e-01 -8.72412324e-02 -8.47718000e-01 8.74511153e-02 2.80300736e-01 4.18628097e-01 -5.47121279e-03 -1.31821895e+00 -4.20470566e-01 1.39033541e-01 1.87950693e-02 -3.30327034e-01 3.31905037e-01 6.00869834e-01 1.22385398e-01 5.75747490e-01 5.30989580e-02 -3.94306153e-01 -8.64524901e-01 7.97133446e-01 4.71924096e-01 -3.29091281e-01 -5.83111823e-01 8.45053613e-01 2.30413675e-01 -3.88998806e-01 1.01893321e-01 -3.33728909e-01 1.91938460e-01 -1.82877202e-02 2.82357067e-01 4.89504516e-01 5.85352838e-01 -2.06878722e-01 -3.27629924e-01 3.87789041e-01 -2.55351048e-02 2.91652381e-01 1.18230724e+00 3.91940594e-01 1.61401689e-01 2.44577885e-01 1.22488487e+00 5.54036558e-01 -9.51566815e-01 -1.87697828e-01 5.40012598e-01 -3.37395072e-01 1.12128615e-01 -1.18280482e+00 -1.04995537e+00 1.11984134e+00 6.29562318e-01 6.88804626e-01 8.08389485e-01 2.62922704e-01 4.56306010e-01 3.07452917e-01 3.14540058e-01 -8.21135402e-01 -1.69551983e-01 4.82204467e-01 8.96820247e-01 -1.38222468e+00 -1.09063433e-02 -3.30357850e-02 -3.19419235e-01 1.20075536e+00 2.89810151e-01 -3.03051975e-02 5.40412962e-01 4.67499048e-01 -2.25036174e-01 -2.79262036e-01 -4.18219864e-01 2.00860813e-01 4.94605392e-01 2.53367096e-01 1.75245121e-01 1.50246754e-01 -1.08252177e-02 7.06834853e-01 -4.37188208e-01 -7.05417097e-01 3.47998828e-01 4.37356263e-01 -3.42023015e-01 -1.29210937e+00 -4.84865397e-01 6.16485357e-01 -7.93160737e-01 -2.33114026e-02 -2.35678077e-01 8.09395790e-01 1.83010563e-01 7.79104590e-01 -2.07502991e-01 -2.80606270e-01 3.64173740e-01 3.78010273e-01 3.08216065e-01 -4.15313900e-01 -7.91708350e-01 -3.14952135e-01 2.31029660e-01 -4.88726169e-01 5.82861528e-02 -9.12443638e-01 -9.34055567e-01 -2.30963841e-01 -2.27278665e-01 4.08924431e-01 8.61354828e-01 1.15305853e+00 3.86888310e-02 8.93759489e-01 7.88876235e-01 -7.63426125e-01 -1.10724056e+00 -7.79755592e-01 -4.48318094e-01 1.92342684e-01 3.68878365e-01 -9.16798770e-01 -5.50813615e-01 -7.97395051e-01]

[15.661308288574219, 2.928506374359131]

e2028f88-3292-4546-80e4-49eb5fb9924a

mads-modulated-auto-decoding-siren-for-time

2307.00868

https://arxiv.org/abs/2307.00868v1

https://arxiv.org/pdf/2307.00868v1.pdf

MADS: Modulated Auto-Decoding SIREN for time series imputation

Time series imputation remains a significant challenge across many fields due to the potentially significant variability in the type of data being modelled. Whilst traditional imputation methods often impose strong assumptions on the underlying data generation process, limiting their applicability, researchers have recently begun to investigate the potential of deep learning for this task, inspired by the strong performance shown by these models in both classification and regression problems across a range of applications. In this work we propose MADS, a novel auto-decoding framework for time series imputation, built upon implicit neural representations. Our method leverages the capabilities of SIRENs for high fidelity reconstruction of signals and irregular data, and combines it with a hypernetwork architecture which allows us to generalise by learning a prior over the space of time series. We evaluate our model on two real-world datasets, and show that it outperforms state-of-the-art methods for time series imputation. On the human activity dataset, it improves imputation performance by at least 40%, while on the air quality dataset it is shown to be competitive across all metrics. When evaluated on synthetic data, our model results in the best average rank across different dataset configurations over all baselines.

['Svitlana Vyetrenko', 'Yousef El-Laham', 'Elizabeth Fons', 'Tom Bamford']

2023-07-03

['imputation', 'imputation', 'imputation']

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

[ 4.81205404e-01 -1.18232317e-01 -1.28630236e-01 -4.98564750e-01 -1.16506076e+00 -4.82406735e-01 8.86955082e-01 -9.26435441e-02 -3.79516870e-01 7.14884877e-01 7.02304721e-01 -2.36344784e-01 -4.02925283e-01 -5.02145708e-01 -9.80006278e-01 -3.92290860e-01 -1.60800397e-01 3.35611612e-01 -4.95688856e-01 -2.29064614e-01 5.06281946e-03 1.69342488e-01 -1.42309713e+00 2.81468451e-01 6.31782234e-01 9.32970345e-01 -5.03592670e-01 3.02374810e-01 1.54636964e-01 9.23133016e-01 -5.57637870e-01 -3.20370108e-01 3.79523963e-01 -2.66982704e-01 -4.51275140e-01 -4.43876147e-01 2.08939150e-01 -2.40286946e-01 -4.25220579e-01 3.71302038e-01 6.23301506e-01 1.91515371e-01 5.52864254e-01 -1.28710890e+00 -4.31582510e-01 5.48886657e-01 -4.37636495e-01 1.36973172e-01 3.27117950e-01 2.33616263e-01 7.96590030e-01 -4.19606358e-01 4.04527128e-01 8.39151144e-01 1.55813205e+00 3.27519029e-01 -1.96650946e+00 -7.43383348e-01 -8.77676159e-02 -1.18265167e-01 -1.14388180e+00 -7.08831191e-01 6.70474470e-01 -5.53613603e-01 1.23738968e+00 2.76926666e-01 2.72517920e-01 1.68435287e+00 2.37290651e-01 6.41445518e-01 1.22300053e+00 -5.17388843e-02 2.48694152e-01 -3.91791046e-01 1.29397720e-01 -1.79727942e-01 -1.01711363e-01 3.57705325e-01 -6.64956629e-01 -3.59744489e-01 4.70527530e-01 1.23119004e-01 -9.08033252e-02 -1.05311163e-01 -1.43300533e+00 5.97201645e-01 1.08318247e-01 3.27644940e-03 -6.11598670e-01 4.80529904e-01 5.21112323e-01 3.42780679e-01 8.76591861e-01 4.07404214e-01 -6.81579053e-01 -5.37172496e-01 -1.27990854e+00 6.26145899e-01 9.11643803e-01 4.54107702e-01 2.77426802e-02 2.11613894e-01 -4.21754956e-01 8.67014289e-01 -1.94331817e-02 4.09244835e-01 4.16240335e-01 -1.04674149e+00 6.95698142e-01 3.05486441e-01 2.20307857e-01 -9.14415538e-01 -6.07175529e-01 -6.95106447e-01 -1.13075137e+00 2.05461293e-01 6.60057127e-01 -3.95132661e-01 -9.95833278e-01 2.23296118e+00 -9.75006595e-02 7.22059190e-01 2.89947551e-04 7.06636906e-01 5.41592836e-01 6.65536225e-01 2.13313565e-01 -1.89634606e-01 8.40085387e-01 -4.12969738e-01 -7.17161000e-01 -2.14070275e-01 3.22597772e-01 -5.05895495e-01 6.22351289e-01 6.02837145e-01 -1.01663792e+00 -5.04170775e-01 -9.35200691e-01 -5.67148253e-02 -2.59996235e-01 -2.20619619e-01 7.58107662e-01 5.45713067e-01 -7.88161039e-01 9.74856496e-01 -1.16638780e+00 -1.95846334e-01 3.78904492e-01 5.35725594e-01 -4.25185710e-01 1.13601454e-01 -1.04115868e+00 8.68337512e-01 1.05994917e-01 3.46947104e-01 -5.71746588e-01 -1.19888985e+00 -7.92436719e-01 -1.47159062e-02 4.77498323e-02 -8.81456673e-01 1.29356062e+00 -9.22843993e-01 -1.12490582e+00 3.43529314e-01 -1.75488561e-01 -9.02155042e-01 7.66300559e-01 -3.15109283e-01 -8.15786719e-01 -5.30609846e-01 -4.98058312e-02 4.26222265e-01 6.16539836e-01 -7.27228403e-01 -1.50331214e-01 -2.50328124e-01 -3.13946754e-01 -2.28856280e-01 -3.59742492e-02 -5.23675680e-02 -2.62253787e-02 -1.05905819e+00 -5.88606074e-02 -9.41898108e-01 -4.99348342e-01 -4.71018046e-01 -3.82644057e-01 1.61457136e-02 4.88738954e-01 -9.99937057e-01 1.20162082e+00 -1.93331444e+00 1.91315323e-01 2.43263930e-01 -8.46456911e-04 -1.14293352e-01 -2.19850749e-01 6.91118419e-01 -4.08948183e-01 -9.09606963e-02 -6.77541316e-01 -7.05877602e-01 3.95210266e-01 2.71008849e-01 -7.35239387e-01 5.62421739e-01 2.75305301e-01 9.53245521e-01 -5.54830909e-01 2.68148482e-01 1.46422714e-01 8.30250680e-01 -4.90595311e-01 1.89035133e-01 -2.44551569e-01 7.89273798e-01 -3.38736776e-04 3.54023784e-01 5.51062822e-01 -1.94645986e-01 2.03148305e-01 -6.34996444e-02 -3.80798951e-02 5.39903104e-01 -1.15166068e+00 1.92323136e+00 -4.95656222e-01 5.81176698e-01 -2.54161894e-01 -9.93724048e-01 7.13765979e-01 6.13207459e-01 7.87899375e-01 -9.41695392e-01 -1.17927030e-01 1.55658826e-01 4.91496548e-02 -3.74870151e-01 4.06844676e-01 -1.96999148e-01 -3.23914587e-01 7.86017537e-01 -6.65274039e-02 1.63780779e-01 -4.17514220e-02 -8.50039124e-02 1.30780411e+00 6.69323623e-01 -9.59095135e-02 -3.11925840e-02 2.37973733e-03 -1.33834168e-01 6.83789968e-01 8.93136024e-01 2.44113520e-01 9.22884822e-01 4.04873192e-01 -7.27642357e-01 -1.14404380e+00 -1.07722497e+00 -2.85706013e-01 1.05401552e+00 -4.93114769e-01 -3.80534589e-01 -6.60148323e-01 -2.73744047e-01 2.42106050e-01 9.42789912e-01 -8.34204376e-01 -6.80292994e-02 -5.82970142e-01 -1.41937768e+00 7.77637601e-01 7.76155949e-01 1.01748504e-01 -1.15152478e+00 -4.31630135e-01 5.99836946e-01 -4.28464919e-01 -1.04506946e+00 -1.24625757e-01 3.08934182e-01 -1.07740831e+00 -7.65724838e-01 -3.98019552e-01 -3.86797413e-02 1.22962669e-02 -2.57718503e-01 1.39197385e+00 -3.28723848e-01 -1.82463855e-01 3.62948090e-01 -3.12376749e-02 -7.01629817e-01 -3.81577134e-01 3.98639321e-01 1.34505108e-01 7.28108361e-02 5.99747479e-01 -1.10600793e+00 -5.36066055e-01 -1.83969073e-03 -8.96379888e-01 7.11293370e-02 2.49887079e-01 7.91560590e-01 3.92564893e-01 -2.45722353e-01 1.07018185e+00 -8.18863451e-01 6.72344744e-01 -1.07804215e+00 -5.23917377e-01 -2.09344491e-01 -6.01857007e-01 1.77279413e-01 6.06497467e-01 -2.65325665e-01 -8.03584397e-01 3.21199484e-02 -3.19002271e-01 -2.06803024e-01 -2.28459224e-01 7.73021996e-01 2.67536968e-01 4.46643740e-01 6.62082136e-01 1.80542693e-01 1.79890916e-01 -8.06676984e-01 2.07547784e-01 5.10777831e-01 8.20866644e-01 -5.81208348e-01 8.22184622e-01 4.42794025e-01 1.35338396e-01 -6.44837141e-01 -8.89423728e-01 -1.11276992e-01 -5.03982127e-01 8.38609487e-02 6.70208395e-01 -1.27626288e+00 -5.93358874e-01 5.53684354e-01 -9.93646920e-01 -5.60590565e-01 -3.05972397e-01 5.63993096e-01 -7.06568182e-01 3.13546583e-02 -4.86338675e-01 -8.92976940e-01 -3.08756441e-01 -8.19751859e-01 1.12118530e+00 -3.70945811e-01 -6.35375023e-01 -1.06869340e+00 3.92583460e-01 2.58954793e-01 8.19956303e-01 1.01338494e+00 7.82218218e-01 -6.15146577e-01 -5.84482670e-01 -2.94376761e-01 3.46089453e-02 2.42461056e-01 -1.09669954e-01 -3.47326607e-01 -1.31049585e+00 -2.00578690e-01 -2.76426412e-02 -2.43269965e-01 1.02419639e+00 4.87543821e-01 1.48333800e+00 -1.76855654e-01 -1.53779224e-01 8.44623029e-01 1.22867393e+00 -2.51262486e-01 9.42456305e-01 3.87762696e-01 5.73712051e-01 4.71943647e-01 7.00111762e-02 4.99998719e-01 5.31074047e-01 8.23481858e-01 4.60002095e-01 -2.36214414e-01 1.04788236e-01 -2.48212054e-01 2.93179810e-01 8.24133694e-01 -2.59036183e-01 -1.80584311e-01 -9.39327419e-01 6.93726122e-01 -2.08407831e+00 -1.28663528e+00 -3.54659557e-01 2.40888333e+00 8.75974715e-01 8.10775459e-02 3.38611484e-01 2.87448764e-01 9.12175030e-02 2.66866148e-01 -7.37351656e-01 -3.98842394e-01 -2.09088281e-01 6.20889544e-01 4.78767425e-01 1.01338677e-01 -1.24514592e+00 1.88143253e-01 7.41616201e+00 4.39636648e-01 -8.93624187e-01 1.27861843e-01 6.22704208e-01 -4.35813248e-01 -2.14371964e-01 -3.06435227e-01 -2.68577009e-01 7.29412258e-01 1.64442539e+00 -7.84447789e-02 7.45765924e-01 2.61154652e-01 4.06898052e-01 2.81951666e-01 -1.35101783e+00 8.75463009e-01 -1.76845416e-01 -1.32169175e+00 -4.16797906e-01 1.46983281e-01 7.43356168e-01 5.49657345e-01 1.08033791e-01 6.48481309e-01 4.36084718e-01 -1.61481512e+00 6.79825008e-01 8.56699646e-01 6.64826751e-01 -7.40694106e-01 7.42067933e-01 5.16513407e-01 -7.86504328e-01 -1.36362702e-01 7.78189674e-02 -5.09461701e-01 2.31748983e-01 6.03700042e-01 -4.20269817e-01 6.50736928e-01 6.88956022e-01 9.40078616e-01 -4.19881701e-01 1.04220450e+00 1.47398397e-01 1.14992714e+00 -5.81831038e-01 7.46370077e-01 -2.15490870e-02 -1.90509349e-01 3.78771216e-01 1.26303339e+00 5.01182318e-01 -4.24502492e-02 -3.24482173e-02 8.40734005e-01 -1.62592664e-01 -2.96841949e-01 -6.99045777e-01 1.52628403e-02 2.75257319e-01 8.48071635e-01 -5.04575484e-02 -1.04355402e-02 -5.00243843e-01 5.61470747e-01 2.03263775e-01 5.07154346e-01 -1.06252503e+00 -2.07821712e-01 7.97548413e-01 1.13878414e-01 2.04710871e-01 -3.07324409e-01 -5.28027594e-01 -1.06346571e+00 2.31167331e-01 -1.11265182e+00 5.05290508e-01 -8.59872520e-01 -1.63216984e+00 5.40098071e-01 -2.82131396e-02 -1.34770143e+00 -7.09215105e-01 -2.22760290e-01 -4.78065103e-01 1.33259130e+00 -1.35809159e+00 -1.22656524e+00 -2.55255550e-01 4.06114042e-01 2.34995604e-01 -5.39937057e-02 9.75362778e-01 4.70464736e-01 -4.10078526e-01 4.55468863e-01 5.53207457e-01 1.04301505e-01 6.81147337e-01 -1.13851202e+00 1.07339144e+00 8.67305517e-01 2.73100704e-01 6.67345643e-01 8.43453944e-01 -4.49027807e-01 -1.61572087e+00 -1.35892892e+00 7.42423177e-01 -9.65580106e-01 5.63189030e-01 -5.32762885e-01 -1.13136661e+00 1.05436206e+00 3.43847096e-01 -1.57930255e-01 8.36135983e-01 4.74493235e-01 -6.17974281e-01 -1.13675117e-01 -9.90259826e-01 4.23304945e-01 1.05870914e+00 -5.45665443e-01 -6.97251141e-01 2.28220463e-01 5.95193446e-01 -4.78990853e-01 -1.21865928e+00 5.40971577e-01 8.69847894e-01 -9.49872136e-01 1.15841222e+00 -7.23493099e-01 7.11282909e-01 -3.58263075e-01 -2.83307403e-01 -1.34067905e+00 -3.72615814e-01 -7.21208513e-01 -4.95030522e-01 1.37201476e+00 4.69027400e-01 -6.28105998e-01 5.72437108e-01 9.31531966e-01 -6.30682111e-02 -4.71179932e-01 -9.52459037e-01 -8.42056453e-01 1.50267452e-01 -1.01125729e+00 1.02614284e+00 1.10153377e+00 -3.28082770e-01 2.53273755e-01 -7.90895164e-01 8.39079842e-02 8.74132752e-01 7.90385082e-02 9.89532053e-01 -1.44428468e+00 -5.79052329e-01 -3.16966981e-01 -2.33758166e-01 -7.16793358e-01 2.91875869e-01 -1.04620123e+00 -2.22176582e-01 -1.40576410e+00 2.56706346e-02 -4.96397883e-01 -5.85669637e-01 6.34365916e-01 8.64294544e-02 6.00670934e-01 5.66040277e-02 1.52751043e-01 -2.87160963e-01 5.20295858e-01 4.83232200e-01 -1.89533399e-03 -2.10789025e-01 1.09837085e-01 -8.28503191e-01 4.92356837e-01 8.35610628e-01 -5.53092062e-01 -2.89096236e-01 -7.47785509e-01 2.42727593e-01 3.51864874e-01 7.26043761e-01 -1.19475222e+00 -1.57940816e-02 -4.59860638e-03 7.30560541e-01 -6.25008106e-01 5.60721517e-01 -9.32410717e-01 9.31232393e-01 9.54798833e-02 -5.63671947e-01 3.03070992e-01 5.79692781e-01 8.35155010e-01 1.66055653e-02 3.11386764e-01 3.11692417e-01 3.42416525e-01 -3.37744355e-01 1.60026059e-01 -1.35260284e-01 2.47710660e-01 4.78520244e-01 -2.54316479e-01 -2.14006573e-01 -5.67058802e-01 -6.73432648e-01 1.27303094e-01 2.63731271e-01 7.00208306e-01 7.61852041e-02 -1.40852869e+00 -1.00218976e+00 3.88981104e-01 6.95248991e-02 -3.23886901e-01 1.63920775e-01 1.07476819e+00 7.20319077e-02 1.72648326e-01 -8.81704241e-02 -7.39998460e-01 -7.56525636e-01 3.77081007e-01 3.03751111e-01 -2.55080044e-01 -7.69342124e-01 1.42353222e-01 -1.80243731e-01 -6.77837729e-01 1.93368554e-01 -4.78906900e-01 8.94089267e-02 -1.01078883e-01 5.14932156e-01 4.27877635e-01 3.04471523e-01 -5.36219358e-01 -1.72955453e-01 3.56727481e-01 1.97055578e-01 -1.67386562e-01 1.91125250e+00 2.03709111e-01 -4.06312272e-02 6.85168564e-01 1.06596100e+00 -1.65653631e-01 -1.44688058e+00 -1.72058836e-01 1.57689005e-01 -3.06252301e-01 -9.89252254e-02 -1.29258132e+00 -7.21162736e-01 6.69767439e-01 4.84236091e-01 2.78274447e-01 1.25460839e+00 -5.63671947e-01 8.39899361e-01 9.63564664e-02 4.44368601e-01 -7.64831483e-01 -5.80876231e-01 5.82336426e-01 7.99135149e-01 -1.22354627e+00 4.83238101e-02 1.57233402e-01 -4.48188901e-01 7.89205670e-01 -4.92546596e-02 -2.42623970e-01 4.84209388e-01 3.87496114e-01 4.41991091e-02 3.78475785e-02 -1.04719019e+00 2.20258430e-01 3.56953800e-01 7.26748586e-01 7.49075711e-01 -4.72439378e-02 -1.41318029e-04 8.13555598e-01 -2.97402442e-01 4.98279721e-01 5.24783134e-01 5.21315336e-01 3.23523551e-01 -1.25237930e+00 -2.95594543e-01 7.41971552e-01 -7.89913297e-01 -9.05747190e-02 2.09195428e-02 8.22828650e-01 -8.73915851e-02 9.82186198e-01 1.47621259e-01 -1.48464084e-01 4.93758023e-01 2.78142482e-01 3.20293009e-01 -1.45492196e-01 -1.07097042e+00 -1.22241266e-01 3.39877874e-01 -6.94670200e-01 -5.38309932e-01 -1.06445205e+00 -8.01974654e-01 -5.12876630e-01 2.37879992e-01 4.96308394e-02 7.15171754e-01 8.87669802e-01 7.05069840e-01 7.33434796e-01 2.98903048e-01 -9.59991276e-01 -7.45258272e-01 -9.45152879e-01 -3.10275465e-01 7.49733806e-01 5.52341282e-01 -4.43519235e-01 -2.19832063e-01 1.52041867e-01]

[7.06557559967041, 3.311727285385132]

f284f1ee-38f4-4d0f-9252-2f6418151ab6

extracting-information-from-twitter

2306.08236

https://arxiv.org/abs/2306.08236v1

https://arxiv.org/pdf/2306.08236v1.pdf

Extracting Information from Twitter Screenshots

Screenshots are prevalent on social media as a common approach for information sharing. Users rarely verify before sharing a screenshot whether the post it contains is fake or real. Information sharing through fake screenshots can be highly responsible for misinformation and disinformation spread on social media. Our ultimate goal is to develop a tool that could take a screenshot of a tweet and provide a probability that the tweet is real, using resources found on the live web and in web archives. This paper provides methods for extracting the tweet text, timestamp, and Twitter handle from a screenshot of a tweet.

['Michele C. Weigle', 'Michael L. Nelson', 'Tarannum Zaki']

2023-06-14

['misinformation']

['miscellaneous']

[ 1.84913017e-02 3.92932117e-01 -3.77601683e-01 1.64847858e-02 -6.20731890e-01 -9.11891699e-01 9.07808483e-01 6.83596075e-01 -1.77301019e-01 8.40735674e-01 3.17438394e-01 -4.77140039e-01 7.33993530e-01 -9.56177592e-01 -5.76522708e-01 2.24582344e-01 3.18860441e-01 -3.22979629e-01 6.48441494e-01 -2.72001863e-01 9.23290670e-01 9.05170590e-02 -1.00520122e+00 7.44894803e-01 3.08982193e-01 7.09476054e-01 -1.63710907e-01 8.02656233e-01 -5.02008200e-01 9.14975524e-01 -1.03285766e+00 -6.24605715e-01 -1.01550266e-01 -4.60864097e-01 -4.83636260e-01 -3.23098063e-01 3.32606345e-01 -6.78316891e-01 -5.14511049e-01 1.24881172e+00 5.22694811e-02 -6.19109571e-01 1.27724856e-01 -1.68700135e+00 -2.18500048e-01 8.32594275e-01 -4.86052454e-01 4.73845273e-01 1.09264171e+00 -1.92896515e-01 4.97207791e-01 -6.82273507e-01 8.08565795e-01 8.54697049e-01 9.29517686e-01 4.28227037e-02 -8.06197643e-01 -8.75430226e-01 -7.49781430e-01 -3.22222501e-01 -1.31835628e+00 -2.87759930e-01 5.53118467e-01 -6.75495625e-01 4.68453139e-01 8.14254403e-01 1.01145649e+00 1.34838092e+00 6.25739515e-01 4.79872763e-01 1.06113684e+00 -4.34223443e-01 -7.38244578e-02 8.36918414e-01 1.77740037e-01 6.59610569e-01 7.00645924e-01 -3.56739163e-01 -9.26675975e-01 -8.91283929e-01 4.01795834e-01 3.58450204e-01 -5.45598306e-02 4.85024422e-01 -1.38289213e+00 8.22062671e-01 1.75956666e-01 5.07838309e-01 -2.01080024e-01 7.23680779e-02 4.77163136e-01 7.03806102e-01 1.01290810e+00 5.56213140e-01 9.45913121e-02 -3.60336512e-01 -9.81268287e-01 2.75531411e-01 1.21542370e+00 8.06610227e-01 1.08906484e+00 -5.94746053e-01 1.70301795e-01 2.43863150e-01 4.30237055e-01 7.45991409e-01 7.38404334e-01 -2.57173389e-01 6.94107711e-01 9.05441046e-01 7.86819756e-01 -2.02023768e+00 -1.43173277e-01 2.00061738e-01 1.05277210e-01 -2.93724507e-01 5.30226350e-01 -5.69291472e-01 -2.22779419e-02 5.66224098e-01 2.72734314e-01 -1.15230337e-01 -8.22622538e-01 3.68739963e-01 7.17351794e-01 6.09409630e-01 -3.95825118e-01 -1.83572531e-01 1.21401858e+00 -4.13795918e-01 -1.02281964e+00 -3.01441550e-01 9.19447243e-01 -1.28636003e+00 9.05514896e-01 4.62898389e-02 -6.19865179e-01 1.51275665e-01 -1.05077422e+00 9.71186981e-02 -1.07220531e+00 -2.77509421e-01 2.39125669e-01 1.01242256e+00 -4.55921471e-01 8.71825159e-01 -4.89471585e-01 -6.46456063e-01 3.06089610e-01 -3.18883181e-01 -2.16251463e-01 4.96287018e-01 -1.21145117e+00 7.95663893e-01 6.70509115e-02 -5.35053313e-01 -8.25253054e-02 -5.45287967e-01 -2.36165881e-01 -4.43049997e-01 3.54602337e-01 9.29658189e-02 1.40602434e+00 -1.09078681e+00 -1.15527368e+00 9.91980791e-01 -4.76497971e-02 -4.61633950e-01 1.04855299e+00 -2.12034255e-01 -7.41524518e-01 1.85284019e-01 4.13009942e-01 -4.87008661e-01 1.47824001e+00 -6.99562907e-01 -5.68775713e-01 -2.49623641e-01 -4.02945340e-01 -4.85959917e-01 -6.62254214e-01 3.33687246e-01 9.18024331e-02 -6.32811666e-01 1.29383311e-01 -9.70252991e-01 4.69146520e-01 -1.79312989e-01 -1.19821990e+00 3.27343017e-01 1.31728911e+00 -8.69793832e-01 2.07527351e+00 -1.87015390e+00 -9.55356002e-01 3.00217807e-01 4.50908184e-01 2.65641719e-01 8.45473289e-01 1.40185153e+00 4.35104460e-01 1.06904459e+00 5.13494432e-01 -2.83964090e-02 -7.18982145e-02 -4.40184265e-01 -6.42935753e-01 1.00188839e+00 -5.59195757e-01 6.26928449e-01 -1.28337872e+00 -3.73758525e-01 -6.12837784e-02 1.33181229e-01 -6.35932609e-02 -4.72066440e-02 -1.39171407e-01 1.26459360e-01 -6.25257611e-01 1.96627185e-01 6.14631295e-01 -6.74370527e-01 -4.85726222e-02 1.29979715e-01 -7.01418221e-01 9.50177729e-01 -7.50689030e-01 6.66927814e-01 -1.51627466e-01 1.44706929e+00 -4.17593330e-01 1.82866350e-01 5.05382597e-01 3.67531896e-01 1.40418679e-01 2.22317944e-03 3.92427474e-01 2.50959367e-01 -9.33256626e-01 -4.14430976e-01 1.00860476e+00 2.34989926e-01 -1.89910188e-01 1.37665665e+00 -7.16907859e-01 8.29717517e-02 -7.12646171e-02 5.64247727e-01 1.31431186e+00 -4.73468333e-01 5.54372489e-01 1.43233016e-01 2.31808424e-01 2.88146317e-01 -4.84120756e-01 8.68158937e-01 -2.62234390e-01 4.50789064e-01 6.07999265e-01 -3.02795649e-01 -1.09341776e+00 -4.12605643e-01 1.32337391e-01 1.00015950e+00 2.85755366e-01 -7.60669768e-01 -7.42135704e-01 -8.03889930e-01 2.54160911e-01 7.66415775e-01 -4.93981034e-01 2.85823122e-02 -9.84135047e-02 -4.65511650e-01 6.49920046e-01 -5.50224781e-01 4.95038003e-01 -5.25433958e-01 -4.37663496e-01 2.18447104e-01 -6.28853321e-01 -8.84133935e-01 -7.35805571e-01 -7.66232073e-01 -4.87312526e-01 -1.20792878e+00 -4.96695697e-01 -1.95253462e-01 6.62708282e-01 9.94009256e-01 5.26346743e-01 6.30173326e-01 2.25897536e-01 9.58526358e-02 -5.67026794e-01 -5.77434361e-01 -8.09502125e-01 1.12950265e-01 -1.72334358e-01 2.85953671e-01 4.60088640e-01 -1.87056348e-01 -4.14530009e-01 4.91950750e-01 -1.08271587e+00 1.53405815e-01 -4.09315944e-01 1.55990183e-01 -4.56024051e-01 4.96765366e-03 2.37124324e-01 -1.65332901e+00 8.16002965e-01 -1.28494656e+00 -4.25247550e-01 -4.07521993e-01 -6.73078775e-01 -7.23704696e-01 3.68984401e-01 -4.17568177e-01 -5.77535391e-01 -2.58066744e-01 2.51592785e-01 3.43630612e-01 -6.07134514e-02 5.65754175e-01 7.65294313e-01 -4.04484838e-01 9.99140680e-01 1.82815678e-02 1.55683290e-02 -4.67785984e-01 -4.95724864e-02 1.16645181e+00 -3.49210173e-01 6.34782076e-01 1.04976797e+00 1.10200727e+00 -7.09492981e-01 -1.27912688e+00 -9.28430319e-01 -1.06954265e+00 -2.18798801e-01 -5.19329309e-01 4.13325615e-02 -6.99622273e-01 -7.09837019e-01 6.43692136e-01 -1.22613037e+00 7.03021139e-02 5.08142114e-01 1.67121366e-01 8.58411342e-02 5.46059728e-01 -5.39465249e-01 -7.07822144e-01 1.65496677e-01 -1.80561617e-01 4.12838250e-01 -6.31029084e-02 -1.01983953e+00 -1.25639057e+00 3.26005131e-01 3.98663819e-01 6.59531415e-01 7.31101871e-01 -5.96400350e-02 -1.14450705e+00 -3.94734323e-01 -1.25937974e+00 -4.18932647e-01 -3.31794620e-01 3.80833954e-01 6.42922282e-01 -1.04153943e+00 -5.41118532e-02 -1.70601889e-01 1.76779809e-04 7.23728612e-02 -3.01151037e-01 6.79093897e-01 -1.40243828e+00 -7.21115232e-01 -1.17580339e-01 1.38477492e+00 -3.24082017e-01 7.28158236e-01 7.60609090e-01 6.02659166e-01 4.91769135e-01 2.66812414e-01 1.10732949e+00 1.48031846e-01 3.61138046e-01 4.81214404e-01 3.73672366e-01 4.03040528e-01 -8.96068037e-01 5.27418375e-01 6.45043910e-01 6.92102849e-01 -8.31364036e-01 -8.90619934e-01 5.20594656e-01 -1.54235327e+00 -1.41756034e+00 -7.65788972e-01 2.34047580e+00 4.94640589e-01 4.93863583e-01 5.77745140e-01 2.58238763e-01 1.38010657e+00 4.70752299e-01 1.66881066e-02 -1.52422220e-01 3.93367320e-01 -6.46303654e-01 1.28728163e+00 6.36686742e-01 -7.35638678e-01 6.27388835e-01 6.59090376e+00 3.19375604e-01 -1.28553939e+00 2.19199896e-01 2.16347963e-01 6.48614168e-02 -2.67330766e-01 -2.73511652e-02 -9.33044672e-01 1.14883637e+00 1.24446571e+00 -5.29087722e-01 3.97243261e-01 6.18316352e-01 7.40998387e-01 -5.54327965e-01 -7.15212524e-01 7.45765030e-01 2.51865238e-01 -1.70984674e+00 -1.35663316e-01 4.03840840e-01 5.93776107e-01 2.60210663e-01 1.11025786e-02 -2.99636722e-01 4.66659106e-02 -5.07896543e-01 8.53960216e-01 3.38917285e-01 5.53982675e-01 -3.66556019e-01 6.60163701e-01 6.30820692e-01 -5.18750191e-01 6.37614191e-01 1.74873471e-01 -4.80520487e-01 2.69804090e-01 8.03350270e-01 -1.70866060e+00 -2.41164193e-01 2.77257740e-01 9.95497882e-01 -8.81124675e-01 1.16707301e+00 -1.13835864e-01 7.88418829e-01 -4.98194009e-01 -8.80656421e-01 -3.77960205e-02 1.47596568e-01 1.07921302e+00 1.40018833e+00 4.13581371e-01 -6.07085228e-01 -4.22502577e-01 6.83516979e-01 -3.87998104e-01 -7.50217289e-02 -1.07986748e+00 -1.15182889e+00 7.81850100e-01 1.12603033e+00 -9.54643011e-01 -3.15458834e-01 -3.54876131e-01 1.25097072e+00 -1.35242522e-01 1.30914077e-01 -8.28301907e-01 -7.25930154e-01 1.52820557e-01 1.18515456e+00 -2.08491549e-01 -1.89477637e-01 -2.48819008e-01 -1.33663070e+00 -3.68946344e-02 -4.84763443e-01 -2.75207870e-02 -7.73632109e-01 -1.20056081e+00 2.80342221e-01 -2.85600215e-01 -1.46265268e+00 -2.07314774e-01 -1.44854471e-01 -7.44014621e-01 6.17092490e-01 -8.97381604e-01 -4.33857828e-01 -3.57145786e-01 3.86223704e-01 -2.37998776e-02 2.40997776e-01 4.19650286e-01 4.14640397e-01 -2.19260812e-01 3.09698433e-01 2.23212495e-01 2.98915625e-01 9.42002237e-01 -1.08873165e+00 8.26297462e-01 5.05014360e-01 5.89507408e-02 7.57297099e-01 1.33579063e+00 -1.29671466e+00 -1.42556345e+00 -9.21837568e-01 1.58403194e+00 -1.01996577e+00 1.53399134e+00 -1.85840815e-01 -6.08958304e-01 8.89472067e-01 2.57676449e-02 -3.47300738e-01 9.28746104e-01 -2.28248775e-01 -2.45336860e-01 5.98972023e-01 -1.33570290e+00 6.31422877e-01 4.22323823e-01 -1.12139082e+00 -5.03195167e-01 1.12931001e+00 3.15940589e-01 -4.92579728e-01 -3.25665325e-01 -9.51678455e-01 8.20271254e-01 -1.07591438e+00 5.39722979e-01 -3.88487965e-01 4.06120151e-01 -4.01502438e-02 5.28728664e-01 -1.29708886e+00 4.47969198e-01 -1.50934279e+00 2.20065750e-02 1.20491123e+00 5.58468044e-01 -1.01789582e+00 8.57258141e-01 5.64043999e-01 6.78410590e-01 3.40104669e-01 -3.46084952e-01 -4.28775966e-01 -7.05081165e-01 -4.89365160e-01 3.86836916e-01 1.27076983e+00 6.84213161e-01 7.05375196e-03 -5.07138073e-01 2.22044550e-02 5.26012897e-01 -5.85564017e-01 1.06657338e+00 -1.20876753e+00 3.08312237e-01 -5.85590079e-02 -1.83658451e-01 -6.76857531e-01 -6.96882188e-01 -5.87933540e-01 -3.32960218e-01 -9.13874805e-01 -1.80329289e-02 -5.57038561e-02 6.54476464e-01 -1.01347029e-01 5.06931186e-01 7.07791746e-01 -4.28710654e-02 9.20681596e-01 -7.17931867e-01 -3.04999530e-01 9.67487812e-01 1.95391089e-01 -3.73707533e-01 2.96384186e-01 -6.17299139e-01 1.15870726e+00 7.86445856e-01 -1.16091096e+00 2.56862462e-01 4.54938978e-01 1.39846504e+00 -1.17013941e-03 4.76430207e-01 -6.84108078e-01 2.82775193e-01 -7.47064278e-02 -1.13297015e-01 -7.65750170e-01 -1.98483132e-02 -7.42850065e-01 1.52008414e-01 4.53502089e-01 -4.77187246e-01 1.47989035e-01 -1.60241202e-01 1.17335856e+00 4.67005670e-02 -3.21006417e-01 4.66567427e-01 -3.84345233e-01 -1.13400267e-02 -9.90480408e-02 -1.35580564e+00 9.62904766e-02 9.28004324e-01 -3.33554596e-01 -9.13744032e-01 -8.32538068e-01 -6.32437468e-01 -5.46739995e-01 9.54486847e-01 2.57988125e-01 2.21468598e-01 -1.04527330e+00 -3.05632770e-01 -4.95063215e-02 2.77011573e-01 -1.03471279e+00 -4.43596035e-01 6.38768733e-01 -1.01838779e+00 4.87898320e-01 -2.20245481e-01 -9.33151841e-02 -1.37689090e+00 2.05116272e-02 -2.91471072e-02 2.81525463e-01 -4.50688332e-01 4.10747111e-01 -8.21144521e-01 -1.09565832e-01 -2.60837493e-03 -1.39741659e-01 -6.70567378e-02 4.38183486e-01 1.40963268e+00 9.84005451e-01 2.77227193e-01 -8.54554176e-01 -3.18804801e-01 -4.09216493e-01 -5.46679422e-02 -5.43394327e-01 9.12435770e-01 -8.98264587e-01 -3.31547499e-01 1.20086181e+00 1.58688545e+00 8.04474831e-01 -5.73187590e-01 2.11413559e-02 1.06910318e-01 -1.10195148e+00 2.19861105e-01 -5.13460457e-01 -1.73495322e-01 2.42320091e-01 -1.00710563e-01 1.63231087e+00 -2.11591765e-01 -2.40876481e-01 1.36395717e+00 3.83194566e-01 4.69525695e-01 -1.05141222e+00 2.60470539e-01 3.92832994e-01 7.22708404e-01 -1.48360801e+00 3.93563300e-01 -5.20543516e-01 -3.70067179e-01 1.40130949e+00 -7.28177577e-02 -1.79852381e-01 1.03963625e+00 -1.69919789e-01 -6.28055111e-02 -5.18018007e-01 -2.71296263e-01 3.59985471e-01 6.48462623e-02 6.41521811e-02 6.68362439e-01 3.33064869e-02 -5.82341373e-01 3.07618938e-02 -3.65776062e-01 -1.97214466e-02 1.46333873e+00 1.06291401e+00 -8.16372812e-01 -4.92286682e-01 -5.80538511e-01 6.62117124e-01 -1.20795190e+00 1.23032495e-01 -1.04048777e+00 5.73546767e-01 -3.20115387e-01 1.19900322e+00 -6.74481317e-02 -6.34925008e-01 -5.00036299e-01 -6.29972368e-02 -3.26571018e-01 -8.72987866e-01 -1.01395071e+00 -5.35742640e-01 7.84638405e-01 -6.77461565e-01 -2.06322044e-01 -6.82655394e-01 -9.01044190e-01 -1.15298331e+00 -7.61737764e-01 2.39573449e-01 1.16993022e+00 9.51211095e-01 3.52058351e-01 -3.99624527e-01 7.25259304e-01 -6.67740345e-01 -1.34451330e-01 -9.30426061e-01 -4.02775675e-01 3.15491617e-01 8.14278245e-01 -2.53417850e-01 -1.07682467e+00 2.82959566e-02]

[8.15894603729248, 10.214685440063477]

3cfd6bc6-cd98-4a3b-a7fa-59044678563a

using-motion-history-images-with-3d

2110.12396

https://arxiv.org/abs/2110.12396v2

https://arxiv.org/pdf/2110.12396v2.pdf

Using Motion History Images with 3D Convolutional Networks in Isolated Sign Language Recognition

Sign language recognition using computational models is a challenging problem that requires simultaneous spatio-temporal modeling of the multiple sources, i.e. faces, hands, body, etc. In this paper, we propose an isolated sign language recognition model based on a model trained using Motion History Images (MHI) that are generated from RGB video frames. RGB-MHI images represent spatio-temporal summary of each sign video effectively in a single RGB image. We propose two different approaches using this RGB-MHI model. In the first approach, we use the RGB-MHI model as a motion-based spatial attention module integrated into a 3D-CNN architecture. In the second approach, we use RGB-MHI model features directly with the features of a 3D-CNN model using a late fusion technique. We perform extensive experiments on two recently released large-scale isolated sign language datasets, namely AUTSL and BosphorusSign22k. Our experiments show that our models, which use only RGB data, can compete with the state-of-the-art models in the literature that use multi-modal data.

['Hacer Yalim Keles', 'Ozge Mercanoglu Sincan']

2021-10-24

['sign-language-recognition']

['computer-vision']

[ 1.87116303e-03 -3.46520007e-01 -6.35525063e-02 -3.54232132e-01 -7.15067565e-01 -1.51575327e-01 6.90645635e-01 -9.84899640e-01 -8.31819713e-01 3.89171988e-01 4.00600374e-01 -6.18649460e-02 1.31139606e-01 -3.28327328e-01 -6.33156776e-01 -8.40256155e-01 4.98708099e-01 1.83045313e-01 8.01860154e-01 -1.04207814e-01 2.69088686e-01 5.49723983e-01 -1.95399857e+00 5.54623008e-01 2.45082632e-01 1.07599890e+00 -7.22779150e-05 9.71485436e-01 -1.24958858e-01 1.39749956e+00 -3.41210127e-01 3.75559144e-02 1.64986357e-01 -6.64085686e-01 -6.08550370e-01 4.70445678e-02 7.49629736e-01 -7.42546499e-01 -7.88680255e-01 6.30922496e-01 5.99997282e-01 1.24677680e-01 4.22659516e-01 -1.33601511e+00 -4.75677460e-01 8.11139643e-02 -4.25583631e-01 -4.27674092e-02 2.76466161e-01 6.21451557e-01 6.93385005e-01 -9.35406148e-01 1.00991750e+00 1.31918561e+00 3.05822760e-01 9.66333210e-01 -6.74985826e-01 -5.78160822e-01 3.82884711e-01 6.80841327e-01 -1.26323962e+00 -5.38784862e-01 9.05267656e-01 -2.94818550e-01 1.08377779e+00 1.20797753e-01 1.19961369e+00 1.11978698e+00 -1.28248349e-01 1.49782252e+00 1.25698400e+00 -5.90200126e-01 1.93284273e-01 -5.59871912e-01 1.72492370e-01 9.27857637e-01 -1.05277620e-01 1.93208277e-01 -9.57296133e-01 1.89696923e-01 9.28518057e-01 7.70839080e-02 -3.27254415e-01 -3.81676525e-01 -1.23716962e+00 3.66832763e-01 5.18357217e-01 4.81728166e-01 -3.70982856e-01 7.01315463e-01 -2.92362645e-04 4.69052531e-02 -1.84717208e-01 -4.28768069e-01 -3.22606504e-01 -2.73205519e-01 -9.87878442e-01 2.93273360e-01 3.95760983e-01 7.01509655e-01 2.96071738e-01 -3.29039395e-02 -9.62999463e-02 3.60961467e-01 8.63785267e-01 6.70271814e-01 8.91461670e-01 -8.88343751e-01 5.42215705e-01 6.40618086e-01 5.09428009e-02 -3.92635494e-01 -4.91358042e-01 1.37114212e-01 -3.08767200e-01 7.21724629e-01 8.28792155e-01 1.80932403e-01 -1.69152069e+00 1.60395074e+00 2.17345685e-01 3.87929469e-01 4.22828831e-02 1.31453025e+00 1.10244226e+00 2.92959660e-01 1.12005539e-01 3.47954333e-01 1.13620698e+00 -1.22081280e+00 -5.58397233e-01 1.40152946e-01 4.99227881e-01 -3.32987398e-01 9.41227615e-01 5.44435740e-01 -1.13673031e+00 -3.27516556e-01 -7.73041070e-01 -2.91038007e-01 -6.83140337e-01 4.43837762e-01 5.32117486e-01 4.29207414e-01 -1.18077099e+00 3.13976258e-01 -1.22274673e+00 -6.26929939e-01 2.68739462e-01 2.82062560e-01 -6.20281994e-01 -2.55026340e-01 -6.85823977e-01 8.82876277e-01 1.36082336e-01 4.25749779e-01 -7.51223564e-01 -3.93438675e-02 -8.10889244e-01 -4.88385379e-01 2.53534336e-02 -5.93167424e-01 1.26697350e+00 -8.16896796e-01 -1.95863354e+00 9.83112097e-01 -4.71780509e-01 -1.22989165e-02 6.58752322e-01 -2.02435449e-01 -3.25197250e-01 3.33776712e-01 -4.70619708e-01 7.87469506e-01 9.55750048e-01 -1.27252710e+00 -7.92699099e-01 -6.39202356e-01 -4.03253175e-02 6.26209006e-02 1.62214905e-01 2.17365891e-01 -7.50205100e-01 -3.23564500e-01 3.98818552e-01 -1.00660276e+00 -5.00395475e-03 4.18882459e-01 -2.42908686e-01 -1.60858840e-01 1.02026105e+00 -9.18479145e-01 8.39926302e-01 -1.88807070e+00 4.51163441e-01 2.22016618e-01 -2.46854238e-02 6.88604593e-01 -3.47411424e-01 -1.26038492e-01 1.61137834e-01 -2.95318395e-01 -1.80743679e-01 -6.64439678e-01 -6.33912981e-02 4.66064930e-01 -3.64762098e-02 6.32388055e-01 2.29105324e-01 1.32270300e+00 -8.39952230e-01 -4.99297708e-01 5.51394463e-01 7.94116020e-01 -3.63420069e-01 9.18920115e-02 -1.80073753e-01 8.05986762e-01 -2.24322647e-01 1.18884957e+00 3.61069530e-01 -1.11253761e-01 -3.74224335e-02 -1.80389941e-01 -3.06868255e-01 -1.23364851e-01 -9.03176904e-01 2.20935106e+00 -2.56656706e-01 6.95968449e-01 -1.10863157e-01 -6.92093074e-01 2.80130595e-01 3.03230107e-01 4.92441744e-01 -1.01524270e+00 4.31670874e-01 5.44228196e-01 -3.69093530e-02 -5.55009902e-01 1.67626932e-01 1.53826803e-01 9.20349136e-02 4.48945224e-01 2.50948727e-01 1.24019168e-01 1.21787265e-01 -1.51220873e-01 1.27015007e+00 7.11831570e-01 -3.89565248e-03 5.77780366e-01 5.97338617e-01 -1.76616281e-01 3.31145436e-01 6.41118228e-01 -5.10621727e-01 1.00417590e+00 1.29417658e-01 -4.36773628e-01 -8.47796440e-01 -1.05094516e+00 2.84266144e-01 7.45773256e-01 4.74206880e-02 -1.96015425e-02 -3.67772907e-01 -7.18085051e-01 1.17787994e-01 3.51226658e-01 -7.84777582e-01 5.27225249e-02 -1.00089610e+00 -4.41535115e-01 5.53899050e-01 9.80656981e-01 7.16679752e-01 -1.31779373e+00 -1.41866279e+00 8.39731544e-02 -3.84152979e-02 -1.20308816e+00 -3.36763680e-01 -8.85837153e-02 -6.94521844e-01 -1.09831667e+00 -1.27216840e+00 -6.75885797e-01 4.79972392e-01 8.51719826e-02 4.56793278e-01 5.24593145e-03 -5.14655113e-01 7.02442348e-01 -6.70844615e-01 -1.30210251e-01 4.18323167e-02 -4.25221831e-01 -1.27780423e-01 3.46273184e-01 3.28607827e-01 -5.83868384e-01 -6.85986221e-01 1.74939945e-01 -9.26345348e-01 1.86235324e-01 7.50593901e-01 8.79995227e-01 4.82478648e-01 -8.29436421e-01 -8.88017491e-02 1.91989660e-01 -7.15762004e-02 -1.32239638e-02 -5.94152033e-01 5.80291629e-01 1.59721240e-01 2.52561212e-01 3.00543122e-02 -5.16528010e-01 -1.02988482e+00 7.07542598e-01 -9.99730229e-02 -7.73861587e-01 -2.76918560e-01 1.35366276e-01 -2.19180301e-01 -5.03229618e-01 9.11032259e-02 5.93811691e-01 9.82788950e-02 -7.28034198e-01 6.75807416e-01 7.49251664e-01 8.66741955e-01 -1.01317521e-02 3.74336541e-01 9.92218018e-01 1.27395868e-01 -8.01432610e-01 -3.76219332e-01 -7.04576135e-01 -1.00397313e+00 -6.36325538e-01 1.01283050e+00 -7.10318625e-01 -9.38766539e-01 1.13614118e+00 -1.34490585e+00 -6.50080144e-01 -2.03142092e-01 7.13325441e-01 -8.24607074e-01 4.37707931e-01 -4.79026854e-01 -1.05279326e+00 -3.05671878e-02 -9.98660386e-01 1.61976969e+00 2.65308291e-01 1.61897074e-02 -6.17160499e-01 1.33645490e-01 2.41525635e-01 3.05337757e-01 4.12129700e-01 4.37548816e-01 -1.49680644e-01 -1.00822675e+00 -3.10804576e-01 -3.08847070e-01 2.39457950e-01 -5.08870706e-02 -1.21908821e-01 -1.09400225e+00 7.60860443e-02 -5.63197196e-01 -5.47252238e-01 1.36350787e+00 4.29691911e-01 9.17490780e-01 -1.07081225e-02 -1.75858423e-01 7.17875183e-01 1.26479673e+00 3.27528685e-01 7.86771536e-01 1.85474366e-01 8.15020502e-01 3.57425749e-01 3.56228262e-01 3.60688210e-01 5.83666325e-01 9.38759983e-01 2.98662096e-01 6.73922012e-03 -7.77573287e-01 -2.85248041e-01 6.27560019e-01 7.72827208e-01 -6.25400364e-01 -9.95489806e-02 -9.74650443e-01 5.38799703e-01 -2.29736280e+00 -1.06320214e+00 8.13986827e-03 1.86566436e+00 3.51370245e-01 -3.61305624e-01 2.49091864e-01 3.60958457e-01 9.60204378e-02 -3.31908837e-02 -6.66483402e-01 8.49363506e-02 -5.75995088e-01 3.34680796e-01 3.65960300e-01 5.20085394e-01 -1.12196636e+00 1.17513323e+00 5.63836384e+00 1.88662276e-01 -1.52511096e+00 2.40253568e-01 -1.37237549e-01 -5.32971323e-01 1.76832139e-01 -9.27390382e-02 -7.20354676e-01 2.90988624e-01 7.43323028e-01 3.77700537e-01 2.26404041e-01 5.43322980e-01 1.73973456e-01 -2.80838430e-01 -1.06784570e+00 1.25569797e+00 4.83393669e-01 -1.03431439e+00 4.88902964e-02 2.12276414e-01 4.32385266e-01 3.51824939e-01 1.83429681e-02 7.44111612e-02 2.27592602e-01 -8.66042197e-01 1.26477349e+00 1.28962195e+00 7.88717151e-01 -7.64436424e-02 5.56174338e-01 2.73908257e-01 -1.31581724e+00 -2.25586951e-01 2.95421898e-01 1.18651174e-01 5.91785908e-01 -1.76754102e-01 -1.51445284e-01 3.56890798e-01 6.65122569e-01 9.33946133e-01 -6.77875996e-01 1.23506546e+00 -4.56637919e-01 3.31061810e-01 -6.11077189e-01 -1.18714020e-01 3.51606488e-01 1.79891720e-01 4.29603338e-01 1.00944328e+00 3.81820500e-01 2.65125483e-01 -2.16390207e-01 7.02392399e-01 5.07806182e-01 -1.40178785e-01 -5.05374432e-01 6.87199906e-02 -3.34486276e-01 6.12640500e-01 -4.58216906e-01 -5.34409463e-01 -6.67176783e-01 1.48731661e+00 9.18874294e-02 6.03659272e-01 -7.33405411e-01 -1.93583265e-01 6.20585859e-01 -2.34414756e-01 6.17349088e-01 -5.23774326e-01 2.58797519e-02 -1.63472259e+00 3.67738992e-01 -5.10477602e-01 3.52227330e-01 -1.14433789e+00 -8.85028780e-01 3.95907730e-01 -1.54011041e-01 -1.39236295e+00 -5.04787922e-01 -1.25631189e+00 -2.42932752e-01 7.42291272e-01 -1.82518733e+00 -1.63431966e+00 -5.82547963e-01 1.09154558e+00 2.55277395e-01 -1.02526076e-01 7.46842027e-01 1.73379555e-01 -3.16446275e-01 5.31129897e-01 -3.50420251e-02 4.28051621e-01 3.86186898e-01 -7.83336997e-01 2.69651592e-01 7.97531307e-01 3.69442999e-01 3.72132719e-01 -1.69190094e-02 -5.58320701e-01 -1.42162192e+00 -9.01630580e-01 9.47490394e-01 -6.87501788e-01 3.84000570e-01 -1.79201558e-01 -3.87342185e-01 6.67488992e-01 -1.64329365e-01 4.67555225e-01 3.16532254e-01 -7.41460502e-01 -4.14432913e-01 1.01929523e-01 -1.10699427e+00 6.24826431e-01 1.49272835e+00 -5.14451206e-01 -5.88788509e-01 -5.51876575e-02 2.52204746e-01 -4.58627701e-01 -4.84079808e-01 3.67824763e-01 1.30892122e+00 -8.53815794e-01 8.37049842e-01 -8.23901713e-01 1.20285228e-01 -5.12395203e-01 -4.90042925e-01 -8.96165609e-01 2.02432156e-01 -1.32744640e-01 -5.38558304e-01 6.94231927e-01 6.62182942e-02 -3.64285827e-01 8.36172581e-01 7.09064722e-01 1.25536561e-01 -4.21456963e-01 -1.45331132e+00 -8.71181428e-01 -1.77723318e-01 -8.16399336e-01 3.88537318e-01 2.95781553e-01 -5.81847830e-03 -2.22948626e-01 -4.31933194e-01 -1.11093037e-01 8.13618243e-01 1.53492123e-01 8.40824783e-01 -1.03431523e+00 -1.67838976e-01 -6.21975124e-01 -1.02286720e+00 -1.13877881e+00 2.44369298e-01 -5.99229217e-01 2.07543612e-01 -1.60263920e+00 -1.28789127e-01 1.89604089e-01 -3.09016258e-01 8.59233260e-01 3.39456439e-01 3.92185539e-01 6.46255314e-01 2.55893618e-01 -7.49945700e-01 7.11412430e-01 1.20189559e+00 -2.06213176e-01 -2.42449209e-01 -3.12331259e-01 2.49029934e-01 9.20826972e-01 4.26704377e-01 -7.72193372e-02 9.18239206e-02 -5.36098421e-01 -3.22256476e-01 6.91062137e-02 9.48448062e-01 -1.33623052e+00 5.46587884e-01 -1.70046557e-02 3.75182599e-01 -7.49074697e-01 8.38914633e-01 -7.60734141e-01 -1.57866478e-01 5.61833739e-01 -1.78985838e-02 -1.76280409e-01 6.99783638e-02 5.32142282e-01 -3.44672441e-01 4.00747538e-01 7.51025259e-01 -4.52069163e-01 -1.31353545e+00 1.88519135e-01 -5.45880437e-01 -5.07957101e-01 7.84216702e-01 -3.17824513e-01 -7.85075352e-02 -4.08013433e-01 -1.06361115e+00 1.42755511e-04 2.28533179e-01 6.50579154e-01 1.10452878e+00 -1.48867977e+00 -3.93649757e-01 4.50480491e-01 3.66594642e-01 -2.10340157e-01 2.72660702e-01 1.02630341e+00 -4.58110690e-01 8.08482051e-01 -3.48467469e-01 -7.87401795e-01 -1.39640784e+00 1.43700331e-01 4.52144444e-01 2.33244319e-02 -8.12665224e-01 7.79694557e-01 -2.93577820e-01 -3.99524659e-01 6.05749786e-01 -9.05973673e-01 -1.01607619e-02 -1.44074872e-01 6.70794964e-01 1.23273596e-01 -1.76979095e-01 -1.28064537e+00 -6.81829453e-01 1.08750749e+00 4.41664636e-01 -8.82086277e-01 1.19531083e+00 5.10759838e-02 4.71625887e-02 5.95020056e-01 1.20353484e+00 -5.85626245e-01 -1.38766646e+00 -4.81700569e-01 -8.02178755e-02 -5.40951073e-01 1.27402470e-01 -1.00270998e+00 -1.23950195e+00 1.10241842e+00 9.52708602e-01 -7.23002493e-01 1.33800852e+00 1.05781198e-01 7.54632592e-01 3.02192986e-01 7.10273504e-01 -1.09456325e+00 1.19598918e-01 6.33254230e-01 1.09317315e+00 -1.21600509e+00 -4.89137709e-01 2.42548898e-01 -5.57848036e-01 1.14701664e+00 5.62014222e-01 7.09364144e-03 8.30227554e-01 -7.58955032e-02 3.98991257e-01 -1.06652174e-02 -5.45003235e-01 -9.55242932e-01 5.09906590e-01 4.51013088e-01 2.14148149e-01 -1.16131172e-01 -2.10303783e-01 5.79524338e-01 1.84667885e-01 5.89132607e-01 1.84921414e-01 1.52042830e+00 -5.97208925e-02 -1.12125254e+00 -4.20009702e-01 -8.01316500e-02 2.84801394e-01 9.25542638e-02 -6.57383442e-01 8.60063791e-01 3.30585361e-01 7.27166593e-01 -6.70581162e-02 -5.67048371e-01 5.37244320e-01 6.79378092e-01 1.11663747e+00 -1.27696291e-01 -2.47834742e-01 -1.46371061e-02 -1.78201497e-01 -1.12558460e+00 -9.28426623e-01 -1.00065494e+00 -1.34537005e+00 1.82002172e-01 -1.22321770e-03 -7.53333211e-01 9.07415211e-01 1.10959160e+00 1.35664791e-01 2.66459316e-01 -1.03917811e-02 -1.34363186e+00 -3.24057400e-01 -9.92981791e-01 -7.33932853e-01 4.63729888e-01 7.71972656e-01 -8.06009412e-01 -1.32117987e-01 1.12816602e-01]

[9.159063339233398, -6.467127323150635]

20a4fceb-a7f9-4012-8be8-ee71c22bbaba

orthographicnet-a-deep-learning-approach-for

1902.03057

https://arxiv.org/abs/1902.03057v3

https://arxiv.org/pdf/1902.03057v3.pdf

OrthographicNet: A Deep Transfer Learning Approach for 3D Object Recognition in Open-Ended Domains

Nowadays, service robots are appearing more and more in our daily life. For this type of robot, open-ended object category learning and recognition is necessary since no matter how extensive the training data used for batch learning, the robot might be faced with a new object when operating in a real-world environment. In this work, we present OrthographicNet, a Convolutional Neural Network (CNN)-based model, for 3D object recognition in open-ended domains. In particular, OrthographicNet generates a global rotation- and scale-invariant representation for a given 3D object, enabling robots to recognize the same or similar objects seen from different perspectives. Experimental results show that our approach yields significant improvements over the previous state-of-the-art approaches concerning object recognition performance and scalability in open-ended scenarios. Moreover, OrthographicNet demonstrates the capability of learning new categories from very few examples on-site. Regarding real-time performance, three real-world demonstrations validate the promising performance of the proposed architecture.

['Hamidreza Kasaei']

2019-02-08

['3d-object-recognition']

['computer-vision']

[-2.72471607e-01 -8.50188881e-02 3.24447080e-02 -5.02594709e-01 4.58678193e-02 -5.46506107e-01 5.89472711e-01 3.29799838e-02 -5.57932198e-01 2.91222066e-01 -3.64098221e-01 -6.44433424e-02 -1.28441170e-01 -5.82251132e-01 -1.01625550e+00 -4.28039014e-01 -3.96638662e-01 8.29658091e-01 3.42676520e-01 -1.87683120e-01 2.58433461e-01 1.37873459e+00 -1.93034828e+00 1.21089496e-01 3.61283511e-01 1.38986194e+00 6.96044028e-01 2.13645622e-01 -1.85547605e-01 5.04739106e-01 -5.63709497e-01 1.06990561e-01 7.45254934e-01 4.05406684e-01 -7.84531832e-01 4.19785202e-01 5.65625131e-01 -4.25355107e-01 -5.87412715e-01 9.21799898e-01 1.53216869e-01 1.52481049e-01 6.84506416e-01 -1.24265683e+00 -5.80604255e-01 2.84828603e-01 5.11664934e-02 -2.54535675e-02 4.22568262e-01 2.96953231e-01 4.60851699e-01 -1.14072275e+00 7.09064245e-01 1.42733526e+00 5.56460619e-01 3.08115304e-01 -7.96202958e-01 -6.18907571e-01 2.54218161e-01 5.95951915e-01 -1.14069664e+00 -1.04536265e-01 5.75612366e-01 -3.74349952e-01 1.17920685e+00 -3.53597671e-01 8.46074045e-01 9.16110992e-01 1.89922735e-01 8.67879272e-01 7.00216949e-01 -3.56065303e-01 4.54922974e-01 1.15927406e-01 4.75558601e-02 4.37883645e-01 5.63148260e-01 2.63680011e-01 -2.66910583e-01 5.35320222e-01 8.55588496e-01 4.27853286e-01 1.40327126e-01 -1.29709160e+00 -1.41956007e+00 3.65293711e-01 9.09293592e-01 4.64678437e-01 -4.30704653e-01 2.03726664e-01 4.67265278e-01 6.12548292e-01 -1.46792591e-01 6.26059771e-01 -4.96451378e-01 -8.89752060e-02 -2.08172321e-01 9.01024789e-02 8.80143821e-01 1.63740706e+00 8.66871715e-01 2.56264001e-01 4.73024428e-01 7.25611269e-01 2.15542451e-01 5.02352595e-01 6.79661095e-01 -6.80657625e-01 3.05849433e-01 8.61507952e-01 2.38736287e-01 -8.09868395e-01 -7.89645553e-01 -5.40557623e-01 -6.49696946e-01 5.73739171e-01 4.10589010e-01 3.07296664e-01 -1.18069625e+00 1.04824233e+00 3.19521099e-01 -1.35264799e-01 3.37881893e-01 1.12419343e+00 9.37703073e-01 4.08427387e-01 -2.35057801e-01 4.56121266e-01 1.26379943e+00 -1.15483558e+00 -3.77123058e-01 -4.75840509e-01 6.01919532e-01 -5.48173904e-01 8.36527467e-01 4.27984595e-01 -5.56114137e-01 -9.51942742e-01 -1.16920364e+00 2.83537626e-01 -7.29617238e-01 3.63965034e-01 6.17336571e-01 3.33948940e-01 -8.21942389e-01 5.15007734e-01 -7.85373330e-01 -9.63315725e-01 3.75920087e-01 5.39028883e-01 -9.32962000e-01 -3.10815305e-01 -6.18670821e-01 1.28585672e+00 7.45295405e-01 1.74680561e-01 -1.39135361e+00 -2.29933098e-01 -8.18612635e-01 6.11306876e-02 5.53920031e-01 -2.09917694e-01 1.50421417e+00 -7.06260979e-01 -1.57395291e+00 8.50515783e-01 4.98924047e-01 -5.45208633e-01 5.35292387e-01 -4.08207387e-01 -4.99849379e-01 1.86263666e-01 3.05376742e-02 6.23601973e-01 8.31349730e-01 -1.49751854e+00 -8.16241622e-01 -5.47855377e-01 3.80121231e-01 2.91016579e-01 -1.26587689e-01 -1.84828728e-01 -2.79546589e-01 -3.62618893e-01 6.50369585e-01 -1.01897860e+00 -1.45603776e-01 3.87939513e-01 2.77669989e-02 -3.10089082e-01 1.13224208e+00 4.61946726e-02 -1.14071727e-01 -2.34379101e+00 -8.22966769e-02 7.96285942e-02 -1.02981091e-01 3.64778817e-01 -2.14993775e-01 3.45571816e-01 -1.17536515e-01 -4.60701168e-01 1.71490341e-01 1.07714366e-02 1.59700632e-01 3.67320210e-01 -8.77656564e-02 5.54095149e-01 8.39858949e-02 7.34273911e-01 -9.08483505e-01 2.65009534e-02 4.99276578e-01 6.20639473e-02 -2.65431613e-01 3.42313260e-01 -3.05231009e-02 4.74552661e-01 -2.56431401e-01 7.74789393e-01 7.33933687e-01 -7.21396580e-02 2.04067305e-01 -1.45445138e-01 -1.62788764e-01 -1.19731262e-01 -1.35930192e+00 1.88179195e+00 -7.81601608e-01 5.91341853e-01 -9.36374515e-02 -1.47617853e+00 1.49734664e+00 -7.55960308e-03 2.12664545e-01 -7.00107634e-01 3.19852024e-01 5.69127381e-01 -2.58288067e-02 -5.75070798e-01 6.95441127e-01 2.58700758e-01 -1.93810403e-01 3.48520875e-02 6.12028718e-01 -4.32337463e-01 2.39926606e-01 -2.76083171e-01 9.08507586e-01 1.74338788e-01 3.85116160e-01 -2.44145289e-01 4.05541718e-01 1.33442655e-01 8.63799006e-02 8.32818866e-01 -3.37413996e-01 4.41249102e-01 -7.74229914e-02 -1.00911641e+00 -9.62214172e-01 -1.09135616e+00 1.85156390e-02 8.98015678e-01 6.66543245e-01 3.86898547e-01 -1.52650729e-01 -6.20065451e-01 3.52466375e-01 4.92908984e-01 -4.89056289e-01 -2.06331089e-01 -7.33146548e-01 1.02175400e-01 1.89075366e-01 6.48655653e-01 6.67047739e-01 -1.40589225e+00 -1.19463074e+00 3.85312021e-01 3.82785201e-01 -1.46477044e+00 1.69831187e-01 6.72635317e-01 -1.14223027e+00 -1.12860692e+00 -8.39936793e-01 -1.32914996e+00 8.02076101e-01 8.39542747e-01 7.85612702e-01 -4.07041907e-01 -3.64535391e-01 6.57869577e-01 -7.08748102e-01 -4.21378165e-01 -4.39873755e-01 1.50898591e-01 3.43611360e-01 -2.79496703e-02 4.70782220e-01 -6.40332758e-01 -3.64205778e-01 6.13728166e-01 -7.16436267e-01 -4.09795821e-01 9.85228598e-01 7.10430264e-01 2.96055436e-01 -2.01400980e-01 6.61731720e-01 -2.02109590e-01 1.96254864e-01 -3.43925416e-01 -8.78869891e-01 2.35670581e-01 -4.28946048e-01 -8.73868093e-02 6.47421598e-01 -7.19816208e-01 -6.82440877e-01 4.60732013e-01 2.29313001e-01 -7.88851380e-01 -7.62340426e-01 3.83104742e-01 -1.21327592e-02 -2.40318775e-01 8.65421057e-01 3.97431195e-01 1.30955800e-01 -7.50993133e-01 4.43429679e-01 7.91856349e-01 7.72449493e-01 -2.54275620e-01 8.69678617e-01 5.09827733e-01 -1.72829717e-01 -8.21289301e-01 -3.82168621e-01 -8.55499685e-01 -1.20286262e+00 -5.10463953e-01 6.04413748e-01 -1.09005308e+00 -8.26003134e-01 8.75738382e-01 -1.27984679e+00 -3.87007207e-01 -4.46696877e-01 7.54946351e-01 -7.71133304e-01 1.34522319e-01 -4.85929064e-02 -5.54529130e-01 7.67217129e-02 -1.18726206e+00 9.65744138e-01 2.71050870e-01 3.10545176e-01 -5.09568036e-01 -3.19856703e-01 -3.80278975e-02 3.95546764e-01 -8.59763771e-02 6.27797186e-01 -9.37273622e-01 -7.73126245e-01 -6.30976677e-01 -3.79341066e-01 2.46193618e-01 1.47151560e-01 -5.98635972e-01 -6.68788075e-01 -4.96705741e-01 -1.68569639e-01 -4.87167835e-01 5.12535691e-01 -2.85512090e-01 1.00376940e+00 -7.54527971e-02 -3.71655613e-01 3.92752200e-01 1.34157097e+00 5.32263219e-01 4.06175882e-01 7.20141768e-01 5.19715011e-01 3.97610456e-01 7.98140526e-01 5.35407066e-01 2.80216813e-01 9.28661227e-01 1.03261483e+00 1.72538012e-01 -1.35665357e-01 -1.17308259e-01 2.33981609e-01 6.11074328e-01 7.40813315e-02 8.16454142e-02 -1.05150783e+00 7.26171374e-01 -1.62300158e+00 -6.11266196e-01 2.31375918e-01 2.08225751e+00 1.34421453e-01 1.73002973e-01 1.84091125e-02 -4.15023826e-02 7.40258157e-01 -2.30615154e-01 -8.91330779e-01 -1.88523158e-01 -5.61772808e-02 -8.42379332e-02 4.89448994e-01 -2.74857998e-01 -1.19287288e+00 9.26567793e-01 5.58859015e+00 4.39647794e-01 -1.45926487e+00 -5.56715988e-02 -1.25638187e-01 2.95982182e-01 5.87253213e-01 -3.30322623e-01 -7.84924805e-01 -2.34891176e-02 3.32504064e-01 -3.09322439e-02 2.30750456e-01 1.74921751e+00 -2.35653177e-01 -1.32918105e-01 -1.50484002e+00 1.23465085e+00 3.16095084e-01 -1.10085237e+00 -6.28010407e-02 -1.31247997e-01 5.50725579e-01 5.04026949e-01 -1.91584751e-01 5.50422728e-01 2.70999402e-01 -6.67773902e-01 1.24745321e+00 2.46315569e-01 6.13665342e-01 -3.58026057e-01 9.03525889e-01 4.25071567e-01 -1.13738441e+00 -5.19990683e-01 -6.62140489e-01 -7.08299726e-02 -6.79272115e-02 1.71235785e-01 -1.24873936e+00 6.03556812e-01 1.08566940e+00 7.49504507e-01 -6.24464631e-01 1.45848787e+00 -2.62588501e-01 -6.02173107e-03 -4.60581571e-01 -3.83691639e-01 4.68188882e-01 7.14049041e-02 4.88797247e-01 1.02098858e+00 5.36617517e-01 -1.73546135e-01 2.08905995e-01 5.80005109e-01 2.89995503e-03 -3.79621834e-02 -1.01511049e+00 2.11890996e-01 3.05919111e-01 1.15393567e+00 -9.44158673e-01 -3.50863397e-01 -3.70658129e-01 8.72393310e-01 3.21344048e-01 2.15105355e-01 -4.09366906e-01 -5.72819829e-01 4.09199834e-01 -2.29502022e-01 7.72798657e-01 -7.07243562e-01 -6.15162291e-02 -9.70274985e-01 3.13357502e-01 -8.70770216e-01 8.85621756e-02 -9.32625711e-01 -1.07223392e+00 6.99974775e-01 1.03677819e-02 -1.90430915e+00 -1.82227865e-01 -1.22974861e+00 -2.55598724e-01 1.67174891e-01 -1.46479881e+00 -9.82629418e-01 -7.43677855e-01 4.78255957e-01 7.29370534e-01 -6.89215660e-01 8.33766818e-01 2.50140458e-01 6.37334287e-02 4.80861068e-01 5.32511830e-01 1.23437867e-01 4.82138991e-01 -8.91804397e-01 4.06406611e-01 5.18893361e-01 2.95577884e-01 3.31488013e-01 4.61851627e-01 -3.48619401e-01 -1.73154855e+00 -1.09534848e+00 3.34954590e-01 3.77548896e-02 5.40914059e-01 -6.10281408e-01 -7.20841467e-01 6.44745469e-01 -1.69802219e-01 2.39623442e-01 6.78730160e-02 -1.19987661e-02 -6.83756828e-01 -4.74435389e-01 -1.16309500e+00 4.73996520e-01 1.34649396e+00 -2.39328533e-01 -8.98151517e-01 4.98088479e-01 6.46459579e-01 -4.77173179e-01 -7.79524922e-01 6.15434051e-01 7.83262134e-01 -8.69469702e-01 8.17549109e-01 -5.45674384e-01 2.09788308e-01 -4.58219975e-01 -3.58017921e-01 -1.49387336e+00 -2.40198061e-01 -9.71120298e-02 -2.10357774e-02 4.92229104e-01 7.88536966e-02 -8.35100770e-01 8.00627470e-01 -4.03262153e-02 -6.85082734e-01 -3.51118833e-01 -9.69968259e-01 -1.59526527e+00 -1.62196487e-01 -4.99552369e-01 7.87783980e-01 6.60661101e-01 4.20998260e-02 2.97756512e-02 -2.22669430e-02 3.05726349e-01 4.09805655e-01 3.45221162e-01 1.39112580e+00 -1.46974683e+00 4.86503989e-02 -3.23363274e-01 -1.29622293e+00 -1.40249681e+00 1.20661847e-01 -9.00146604e-01 4.35177386e-01 -1.56178546e+00 -2.52230167e-01 -5.87772787e-01 -2.32879668e-01 5.94850659e-01 6.27349436e-01 2.52589822e-01 6.39623225e-01 3.97967517e-01 -8.19864929e-01 7.87431717e-01 1.22101486e+00 -4.53774780e-01 -9.47090089e-02 1.27023280e-01 -4.05578427e-02 6.60526574e-01 8.30787420e-01 -2.61269152e-01 -8.60369354e-02 -7.00171828e-01 -2.20906019e-01 -2.58753955e-01 4.09380972e-01 -1.65153110e+00 3.10032040e-01 9.08257440e-03 2.97325939e-01 -8.41646075e-01 4.11547154e-01 -1.36249304e+00 1.36049300e-01 7.78511167e-01 5.55127300e-02 -1.13836475e-01 3.89579415e-01 4.81285840e-01 -3.12029630e-01 -4.92085457e-01 7.90616870e-01 -2.90237099e-01 -1.49689150e+00 2.73274630e-01 -6.09894156e-01 -5.14577150e-01 1.18521643e+00 -6.63615882e-01 -1.84652507e-01 -2.18690798e-01 -8.12318385e-01 7.47031122e-02 4.54818875e-01 8.78568649e-01 6.97757304e-01 -1.37754464e+00 -2.57018089e-01 4.18264449e-01 7.23734736e-01 2.41441280e-01 1.90166846e-01 3.40050578e-01 -8.17234755e-01 6.00075006e-01 -7.94625819e-01 -9.04198587e-01 -7.70078778e-01 7.47249842e-01 2.59598970e-01 1.13945872e-01 -8.24549794e-01 5.12302935e-01 2.69280881e-01 -1.05766499e+00 6.41941190e-01 -4.57564652e-01 -2.39998564e-01 -1.84457526e-01 1.89303264e-01 3.54023337e-01 3.46044183e-01 -6.80537820e-01 -4.31389779e-01 6.47143781e-01 -5.41368220e-03 2.29853928e-01 1.58187246e+00 -1.64646372e-05 9.26978737e-02 5.63605070e-01 1.09534681e+00 -8.76271784e-01 -1.38039696e+00 -4.92093980e-01 1.66591421e-01 -4.00101811e-01 -4.15059745e-01 -8.57506692e-01 -9.02249873e-01 7.75552452e-01 1.07922244e+00 3.66364308e-02 7.06041038e-01 1.32272780e-01 2.83539951e-01 1.48799372e+00 1.19493496e+00 -1.16681433e+00 4.95309085e-01 1.00925457e+00 1.31579590e+00 -1.28183103e+00 -1.01447955e-01 -2.64125198e-01 -2.26639524e-01 1.45899701e+00 7.93420076e-01 -3.83652508e-01 7.27986574e-01 -1.51574999e-01 3.63036156e-01 -1.95078909e-01 -1.76188320e-01 -2.17500314e-01 2.09311843e-02 1.02392578e+00 -4.33487147e-01 5.69500551e-02 2.85617471e-01 1.01337805e-01 -2.77257979e-01 -8.36630389e-02 6.12135708e-01 1.10822582e+00 -6.34201646e-01 -6.70405149e-01 -3.46808463e-01 1.70324832e-01 1.97197363e-01 5.44785023e-01 -1.27020895e-01 1.04405427e+00 4.39396165e-02 7.25288093e-01 2.40226895e-01 -3.91676068e-01 7.98663199e-01 -1.60412509e-02 6.09017313e-01 -7.64062524e-01 -2.49898434e-01 -4.15926486e-01 -1.35586530e-01 -6.48748696e-01 -5.46720088e-01 -5.82975030e-01 -1.30658317e+00 2.21731752e-01 -4.75126922e-01 -1.62839442e-01 1.34683490e+00 9.06487048e-01 5.31977117e-01 3.87330085e-01 7.80203760e-01 -1.56796348e+00 -8.08453202e-01 -1.17014873e+00 -6.31198823e-01 4.28783715e-01 1.41721278e-01 -1.22871733e+00 -1.87909231e-01 6.77899495e-02]

[7.6280598640441895, -1.2136725187301636]

14713f6d-9291-4ddd-a937-e164a8d09ebf

learning-semantic-representations-for-novel

1811.03866

http://arxiv.org/abs/1811.03866v1

http://arxiv.org/pdf/1811.03866v1.pdf

Learning Semantic Representations for Novel Words: Leveraging Both Form and Context

Word embeddings are a key component of high-performing natural language processing (NLP) systems, but it remains a challenge to learn good representations for novel words on the fly, i.e., for words that did not occur in the training data. The general problem setting is that word embeddings are induced on an unlabeled training corpus and then a model is trained that embeds novel words into this induced embedding space. Currently, two approaches for learning embeddings of novel words exist: (i) learning an embedding from the novel word's surface-form (e.g., subword n-grams) and (ii) learning an embedding from the context in which it occurs. In this paper, we propose an architecture that leverages both sources of information - surface-form and context - and show that it results in large increases in embedding quality. Our architecture obtains state-of-the-art results on the Definitional Nonce and Contextual Rare Words datasets. As input, we only require an embedding set and an unlabeled corpus for training our architecture to produce embeddings appropriate for the induced embedding space. Thus, our model can easily be integrated into any existing NLP system and enhance its capability to handle novel words.

['Hinrich Schütze', 'Timo Schick']

2018-11-09

['learning-semantic-representations']

['methodology']

[ 2.50442475e-01 1.14475943e-01 -4.05081600e-01 -3.95500302e-01 -7.72099197e-01 -7.47361422e-01 7.06742883e-01 6.47948384e-01 -9.79704499e-01 3.84438306e-01 4.22070265e-01 -4.88164574e-01 3.53403360e-01 -9.50261652e-01 -6.06483757e-01 -4.33169454e-01 -4.92886603e-02 3.61003548e-01 5.72426878e-02 -3.96373600e-01 1.91376850e-01 4.84771848e-01 -1.39443624e+00 1.20187499e-01 7.56935120e-01 4.93220389e-01 2.74140060e-01 6.60362661e-01 -7.10493565e-01 9.33057535e-03 -5.53246558e-01 -1.95747226e-01 2.71681309e-01 -1.89377964e-01 -6.18835390e-01 -2.84261763e-01 2.10122854e-01 -1.84813127e-01 -3.21786076e-01 8.65207732e-01 3.94651324e-01 5.60557425e-01 8.72615159e-01 -8.19063663e-01 -1.26340067e+00 7.03096151e-01 -2.58248538e-01 5.67179143e-01 2.32133508e-01 2.50698835e-01 1.45083332e+00 -1.45728421e+00 6.71362519e-01 1.20552135e+00 3.11435610e-01 7.90934980e-01 -1.36418116e+00 -5.17942667e-01 3.78458947e-01 1.52374595e-01 -1.19630325e+00 -4.64011490e-01 6.85328603e-01 -3.10899585e-01 1.49349797e+00 -4.01976844e-03 5.42282701e-01 1.36400127e+00 1.10383429e-01 7.11793602e-01 5.63618004e-01 -8.81569088e-01 3.89339715e-01 3.01171720e-01 4.32924867e-01 2.37944856e-01 4.10834938e-01 1.73026830e-01 -3.22512299e-01 -6.47410229e-02 2.45259851e-01 1.92670375e-01 -1.10426702e-01 -1.19518571e-01 -1.17056918e+00 1.09000301e+00 3.32340837e-01 6.80807292e-01 -3.45113307e-01 1.60721213e-01 4.50012505e-01 2.39169270e-01 5.92513084e-01 9.16129947e-01 -6.56577468e-01 -2.72818774e-01 -6.53590202e-01 2.66629487e-01 6.51694715e-01 8.57405782e-01 1.05303967e+00 2.08856608e-03 -1.77811943e-02 8.18579793e-01 2.78445542e-01 3.83013070e-01 1.10398579e+00 -1.01084977e-01 3.88503611e-01 5.76583982e-01 7.20486790e-03 -9.52568114e-01 -2.01760143e-01 -4.72990386e-02 -2.47610599e-01 -1.61627173e-01 2.16758735e-02 -1.45238608e-01 -1.15896130e+00 1.85700405e+00 2.63426185e-01 2.16900080e-01 3.76828402e-01 3.92826349e-01 6.99023604e-01 1.17327082e+00 9.48184282e-02 -1.00140192e-01 1.39045310e+00 -9.70856309e-01 -7.46215701e-01 -8.61337423e-01 9.98892784e-01 -5.52928865e-01 1.44167495e+00 -2.60611549e-02 -6.88074529e-01 -5.20547271e-01 -1.18595660e+00 -3.51544768e-01 -1.06846678e+00 -2.28804141e-01 3.59893918e-01 5.20151794e-01 -7.46592343e-01 4.51086104e-01 -6.30488455e-01 -5.52479744e-01 1.69894710e-01 9.46277678e-02 -4.00705636e-01 -3.38058531e-01 -1.36478174e+00 1.05840743e+00 8.68436515e-01 -3.08381647e-01 -5.35343647e-01 -8.45033109e-01 -1.25449002e+00 1.71551436e-01 1.83647528e-01 -3.19827020e-01 1.02151632e+00 -7.98071980e-01 -1.06895578e+00 7.02694952e-01 -4.22030181e-01 -3.94287825e-01 -3.05319816e-01 -2.97205091e-01 -7.72532523e-01 4.78307530e-02 7.41486922e-02 5.08100867e-01 9.86543775e-01 -1.14480615e+00 -4.67645109e-01 -2.14259058e-01 1.37193978e-01 -3.66764031e-02 -1.05099905e+00 1.10712869e-03 -1.18753746e-01 -8.34356964e-01 -2.47692883e-01 -7.06334531e-01 -2.86615819e-01 6.03690147e-02 1.24084624e-02 -5.57408154e-01 6.47853613e-01 -4.91104215e-01 1.48797607e+00 -2.46505713e+00 -4.81801294e-02 2.13751376e-01 1.25732303e-01 6.08356178e-01 -5.61050951e-01 7.04615653e-01 -2.26386771e-01 4.48448390e-01 -1.36801302e-01 -3.63716781e-01 2.76640177e-01 5.10952294e-01 -7.66732752e-01 3.92976403e-02 6.57130659e-01 1.16094422e+00 -1.28704822e+00 -1.93465501e-01 1.44335017e-01 4.22311872e-01 -6.86253369e-01 2.91706800e-01 -3.36752534e-01 -1.69514358e-01 -2.33811572e-01 2.67162681e-01 3.86627257e-01 -3.72285843e-02 1.82225451e-01 1.53283337e-02 -5.25427274e-02 7.08459854e-01 -9.87278461e-01 1.66715443e+00 -8.70354593e-01 6.36447072e-01 -5.81159055e-01 -8.60304594e-01 1.04983807e+00 3.41480404e-01 -1.29633434e-02 -3.87274653e-01 1.15465000e-01 3.08114141e-01 1.84850067e-01 -5.76677799e-01 8.11778426e-01 -4.52518493e-01 -1.80523574e-01 8.46794128e-01 4.80256170e-01 -4.33831736e-02 3.89463514e-01 2.27392033e-01 1.34929705e+00 -4.60407048e-01 6.33434772e-01 -1.66729838e-01 2.87228405e-01 -1.02291711e-01 4.60152239e-01 5.91675997e-01 -3.25378329e-02 3.88078839e-01 1.49818733e-01 -5.41563511e-01 -1.10168648e+00 -1.20093012e+00 -2.12468371e-01 1.36372244e+00 -2.38535807e-01 -7.81654894e-01 -1.85442060e-01 -9.07237530e-01 2.19753772e-01 9.73357975e-01 -8.04676592e-01 -5.20055175e-01 -7.46033430e-01 -3.60062957e-01 3.86069924e-01 7.63791025e-01 -4.04392242e-01 -1.23198044e+00 -3.40486228e-01 4.32822436e-01 2.45555133e-01 -9.30500865e-01 -6.48283660e-01 4.72205430e-01 -7.80340850e-01 -7.37917185e-01 -3.78318220e-01 -9.95661438e-01 8.87873173e-01 2.96388209e-01 1.16112399e+00 -3.37619446e-02 -2.84378529e-01 3.03992987e-01 -8.08282673e-01 -5.28215826e-01 -3.33640873e-01 1.86999783e-01 4.61100698e-01 -7.56697580e-02 1.05534911e+00 -6.22182906e-01 -3.52266818e-01 -3.01873803e-01 -1.46976531e+00 -4.95472074e-01 6.07928872e-01 1.00442016e+00 4.30049181e-01 -1.15056634e-01 7.87892222e-01 -8.90362978e-01 8.33170712e-01 -7.14245439e-01 -1.01163365e-01 1.32655099e-01 -6.61827147e-01 3.05879414e-01 9.05760229e-01 -8.39932680e-01 -6.17127001e-01 -1.94465175e-01 -3.37901711e-01 -2.05438480e-01 -1.38437450e-01 8.57880533e-01 -8.95158276e-02 3.90533626e-01 8.71171176e-01 1.81193382e-01 -2.51739979e-01 -5.96928358e-01 9.60637271e-01 8.61928880e-01 1.95894271e-01 -7.05205142e-01 1.08434463e+00 2.46034160e-01 -6.67988062e-01 -1.03776312e+00 -8.72726440e-01 -6.37503922e-01 -7.96001673e-01 3.77628118e-01 6.74782991e-01 -6.81647837e-01 1.47746384e-01 -2.03248084e-01 -1.44363356e+00 3.65645960e-02 -8.30865145e-01 5.39939225e-01 5.96070178e-02 2.71023333e-01 -1.38714865e-01 -6.81512475e-01 -3.44446868e-01 -7.29556739e-01 8.90820980e-01 1.96768478e-01 -5.88684559e-01 -1.36847472e+00 4.94296432e-01 -3.10960919e-01 5.37781358e-01 2.64301356e-02 1.32950354e+00 -1.29191589e+00 -9.74513888e-02 -4.79839593e-01 3.65524702e-02 6.39123857e-01 4.90609407e-01 8.37036315e-03 -1.02413571e+00 -4.74929750e-01 -1.41579136e-01 -3.52477044e-01 8.60421896e-01 -1.60685807e-01 8.76099765e-01 -5.36744654e-01 -4.12247270e-01 4.02186364e-01 1.32830071e+00 1.24189965e-01 3.18774730e-01 2.66343117e-01 4.27729160e-01 4.86990184e-01 3.60281646e-01 2.65070796e-01 1.57057881e-01 3.16451669e-01 8.62960890e-03 2.15306252e-01 1.17561474e-01 -6.51447952e-01 6.15223229e-01 1.33957624e+00 5.61555088e-01 -3.05186272e-01 -1.06245589e+00 1.22340930e+00 -1.55522442e+00 -5.59240818e-01 4.69866693e-01 1.89538753e+00 1.22631121e+00 2.34634429e-01 -3.73799324e-01 3.30096669e-02 5.16944051e-01 4.55499291e-01 -6.56754255e-01 -8.36951792e-01 1.11319246e-02 8.61279845e-01 -3.67468074e-02 4.91230339e-01 -8.28430712e-01 1.11326408e+00 6.01791477e+00 6.16881132e-01 -1.09887719e+00 1.96080402e-01 -8.15751310e-03 -2.74504066e-01 -9.19173598e-01 3.00710928e-02 -9.96720970e-01 2.98507392e-01 1.12999845e+00 -6.97595477e-01 3.22323322e-01 8.19357276e-01 -8.44823867e-02 3.84265423e-01 -1.44148564e+00 8.76875699e-01 4.44490194e-01 -1.23682868e+00 5.30264139e-01 6.37852587e-03 7.14038491e-01 1.34735685e-02 2.00680614e-01 6.79771006e-01 3.39616239e-01 -1.14482009e+00 2.09385753e-01 2.15699058e-02 8.83480906e-01 -6.53328359e-01 6.18306398e-01 3.43292654e-01 -1.05590153e+00 -7.07358867e-02 -6.91011906e-01 -1.37746021e-01 1.61917821e-01 7.79438436e-01 -9.60284412e-01 1.61202103e-01 2.28158161e-01 9.00995851e-01 -5.22838891e-01 4.61525500e-01 -6.08579993e-01 6.14333034e-01 -3.04023683e-01 -3.01204950e-01 4.40615594e-01 8.69133025e-02 5.06930768e-01 1.37049580e+00 2.75722831e-01 5.55660501e-02 6.11576065e-02 9.01463270e-01 -4.53454673e-01 4.63355333e-01 -9.90914047e-01 -6.60580397e-01 7.50782311e-01 1.12895811e+00 -2.38624975e-01 -3.98732215e-01 -5.56571722e-01 7.74596095e-01 6.64117932e-01 4.40925300e-01 -4.29323405e-01 -6.67372048e-01 1.15906250e+00 -8.19618180e-02 5.50510168e-01 -4.20346886e-01 -1.33115560e-01 -1.36109483e+00 2.19558805e-01 -5.82315564e-01 3.22469771e-01 -4.15063143e-01 -1.71618891e+00 7.01726377e-01 -1.81989655e-01 -8.74878526e-01 -4.78419036e-01 -9.10471737e-01 -8.13175917e-01 1.05812359e+00 -1.75393569e+00 -9.12909448e-01 1.56338587e-01 1.97460920e-01 6.89592183e-01 -1.69270411e-01 1.24710286e+00 7.47535378e-02 -3.40406328e-01 8.33489716e-01 3.50749761e-01 2.67643183e-01 7.73142397e-01 -1.21141338e+00 9.03353691e-01 1.06172264e+00 7.41619229e-01 1.17962503e+00 4.68964279e-01 -4.58056450e-01 -1.51468098e+00 -1.31237745e+00 1.39875805e+00 -6.91075981e-01 9.83458996e-01 -7.56961584e-01 -1.19591045e+00 9.22495544e-01 1.12818077e-01 2.79774427e-01 1.16209888e+00 3.74723971e-01 -8.07345629e-01 9.04905871e-02 -8.94641399e-01 8.41040313e-01 8.41370761e-01 -8.36363077e-01 -1.37687159e+00 2.16037273e-01 1.30249441e+00 3.20374109e-02 -5.79568863e-01 8.26459303e-02 3.98168892e-01 -1.40779540e-01 8.70864511e-01 -1.26636004e+00 4.85900939e-01 -2.36497551e-01 -3.15463185e-01 -1.75808680e+00 -3.42156410e-01 -5.30099750e-01 -4.37007248e-01 1.21575952e+00 7.93412983e-01 -7.96265841e-01 3.91955912e-01 6.23238802e-01 -4.41606082e-02 -1.01606500e+00 -9.25394177e-01 -1.05836451e+00 5.49188018e-01 -6.38045907e-01 6.89831674e-01 1.01074266e+00 2.61755735e-01 6.36608541e-01 1.92949787e-01 7.81788975e-02 7.27260038e-02 -6.45112321e-02 6.52065516e-01 -9.41549480e-01 -8.67481381e-02 -1.19624503e-01 -5.79941034e-01 -1.28584528e+00 5.49508214e-01 -1.24121344e+00 1.30744919e-01 -1.42725050e+00 -4.86856550e-02 -4.63442594e-01 -5.99575698e-01 5.69135070e-01 -4.91971105e-01 -2.60228454e-03 9.80396122e-02 -9.04486850e-02 -2.56378025e-01 8.49877477e-01 6.54995680e-01 -2.67853677e-01 -2.30639666e-01 -6.89675629e-01 -8.70337486e-01 6.16399944e-01 8.29665422e-01 -6.82661355e-01 -3.82699341e-01 -7.18024135e-01 4.13973272e-01 -8.97619247e-01 -7.88957402e-02 -5.77309728e-01 1.02718756e-01 -1.72730699e-01 1.46510437e-01 -2.85976619e-01 3.02616030e-01 -8.45718563e-01 -6.06751382e-01 1.77176043e-01 -4.43057805e-01 2.48924419e-01 3.69574517e-01 7.05207765e-01 -1.86706990e-01 -6.36062741e-01 6.06641948e-01 -9.83432215e-03 -9.11387682e-01 3.11773866e-01 -3.27910513e-01 5.84935725e-01 8.98272157e-01 -1.01140000e-01 -3.18232954e-01 8.89286697e-02 -4.70291555e-01 -1.04654152e-02 5.51155746e-01 9.56513703e-01 9.23908889e-01 -1.63490844e+00 -6.98902845e-01 5.44991732e-01 7.80989885e-01 -5.72808459e-02 -1.98452309e-01 9.32526290e-02 -1.78739950e-01 2.63688982e-01 1.81544766e-01 -1.79795831e-01 -8.08570445e-01 8.87558997e-01 -2.01430023e-01 -2.58358717e-01 -6.62707567e-01 8.68083060e-01 1.88305497e-01 -6.46377802e-01 -5.56591414e-02 -6.58203006e-01 -1.62711799e-01 2.31652051e-01 9.32162642e-01 -1.94192678e-01 5.16275764e-02 -5.00279903e-01 -3.35235029e-01 2.87613541e-01 -5.34542918e-01 -2.25483283e-01 1.56006277e+00 7.45901838e-02 -6.90001920e-02 9.37927186e-01 1.68084538e+00 4.11229953e-02 -6.90902591e-01 -7.45822430e-01 3.02586198e-01 -6.85707390e-01 -9.06303823e-02 -5.16822219e-01 -5.10660052e-01 9.72128391e-01 3.54781598e-01 8.06168467e-02 6.05657578e-01 1.49034083e-01 1.31856561e+00 6.94443464e-01 2.05126926e-01 -1.14163506e+00 3.17210257e-01 9.28516686e-01 7.95196235e-01 -1.14437473e+00 -2.22141758e-01 -9.54815652e-03 -2.99694330e-01 1.28235090e+00 3.65302950e-01 -2.91730225e-01 8.88604522e-01 8.52785781e-02 1.53220847e-01 -1.47625014e-01 -9.82065618e-01 -2.28075877e-01 3.87120575e-01 7.00539827e-01 5.07209301e-01 1.03007359e-02 -3.25554252e-01 8.15547645e-01 -2.06283495e-01 -3.98293734e-01 4.36252326e-01 1.08364999e+00 -5.17364919e-01 -1.27781463e+00 5.73006682e-02 7.11033762e-01 -1.68702811e-01 -5.94934404e-01 -3.51679087e-01 5.21378517e-01 1.73312217e-01 7.57962227e-01 1.55085713e-01 -2.59044021e-01 4.76925164e-01 5.09267092e-01 1.51363969e-01 -1.54735243e+00 -4.45214599e-01 -5.65233886e-01 -1.21542580e-01 -3.99590790e-01 -5.60078188e-04 -4.08749789e-01 -1.27992165e+00 -8.01773295e-02 -3.84046406e-01 3.04415584e-01 7.47900248e-01 1.03201485e+00 4.52723473e-01 4.50050235e-01 6.79546058e-01 -6.05006516e-01 -6.90562904e-01 -1.07015955e+00 -6.01882696e-01 5.64027786e-01 4.14747447e-01 -4.62482691e-01 -7.19475806e-01 9.53770429e-02]

[10.505314826965332, 8.70571231842041]

f7a0957b-2dcf-4038-a81c-7e273d0f030c

s-nlp-at-semeval-2021-task-5-an-analysis-of

https://aclanthology.org/2021.semeval-1.120

https://aclanthology.org/2021.semeval-1.120.pdf

S-NLP at SemEval-2021 Task 5: An Analysis of Dual Networks for Sequence Tagging

The SemEval 2021 task 5: Toxic Spans Detection is a task of identifying considered-toxic spans in text, which provides a valuable, automatic tool for moderating online contents. This paper represents the second-place method for the task, an ensemble of two approaches. While one approach relies on combining different embedding methods to extract diverse semantic and syntactic representations of words in context; the other utilizes extra data with a slightly customized Self-training, a semi-supervised learning technique, for sequence tagging problems. Both of our architectures take advantage of a strong language model, which was fine-tuned on a toxic classification task. Although experimental evidence indicates higher effectiveness of the first approach than the second one, combining them leads to our best results of 70.77 F1-score on the test dataset.

['Quang Huu Pham', 'Huy Quang Dao', 'Tam Minh Nguyen', 'Viet Anh Nguyen']

2021-08-01

semeval-2021

['toxic-spans-detection']

['natural-language-processing']

[ 1.04879297e-01 1.19909830e-01 -4.06001449e-01 1.04712896e-01 -8.59447420e-01 -7.19149470e-01 9.85814273e-01 4.27061468e-01 -6.81706190e-01 6.84007525e-01 7.17355072e-01 -3.02250326e-01 -2.35117655e-02 -5.61730564e-01 -3.35873216e-01 -5.52238584e-01 -1.88439742e-01 1.36538222e-01 4.00768310e-01 -3.43453646e-01 6.51332617e-01 2.04104930e-01 -1.26464272e+00 5.09324253e-01 7.17164218e-01 8.59050512e-01 -3.33839655e-02 6.10702157e-01 -4.48778272e-01 9.06916857e-01 -4.78666455e-01 -7.64531314e-01 8.22464228e-02 -4.35366809e-01 -7.83635974e-01 -2.78199911e-01 2.85729349e-01 3.06180060e-01 -2.71610439e-01 8.90474498e-01 7.26496279e-01 -1.97630957e-01 7.02901185e-01 -8.47442329e-01 -5.79769969e-01 1.17724276e+00 -4.10887152e-01 5.13610363e-01 6.12378895e-01 1.35350510e-01 1.19338799e+00 -7.29263484e-01 6.99834049e-01 1.03553009e+00 1.01724637e+00 4.51226473e-01 -1.29587340e+00 -5.31690896e-01 -7.82733262e-02 1.67724565e-01 -1.14691377e+00 -2.70951658e-01 8.64278734e-01 -7.11813688e-01 1.04277253e+00 1.80362746e-01 6.27072215e-01 1.65685654e+00 2.00515419e-01 6.36690676e-01 1.41147685e+00 -5.36579132e-01 2.73164272e-01 5.77971578e-01 6.19946063e-01 5.67860603e-01 3.29471081e-01 -9.26604271e-02 -8.75164628e-01 -5.89741707e-01 -1.55882135e-01 -2.88054198e-01 -2.89227992e-01 -3.78628261e-02 -9.52016890e-01 9.64364588e-01 -1.49352755e-02 9.88005996e-01 -4.46077228e-01 -1.04729284e-03 9.50854540e-01 3.35209578e-01 7.17060983e-01 6.59766614e-01 -5.19185722e-01 -2.69419223e-01 -1.12481117e+00 3.46047342e-01 9.94897306e-01 4.73634422e-01 1.92753330e-01 -3.02240886e-02 -6.59761369e-01 6.66563034e-01 3.05110365e-01 1.77974150e-01 8.94541502e-01 -1.61560237e-01 5.32125711e-01 7.01487541e-01 -5.44631667e-02 -7.57141948e-01 -5.29276490e-01 -4.41739857e-01 -1.30743742e-01 1.10469311e-02 5.96245885e-01 -3.73755693e-01 -7.37866998e-01 1.82253861e+00 6.55479589e-03 -1.34492442e-01 -2.18298912e-01 4.04564202e-01 6.08597755e-01 4.06832337e-01 7.58906245e-01 -3.46545875e-01 1.64288294e+00 -8.23711514e-01 -6.87912345e-01 9.34824944e-02 8.68520021e-01 -8.54046464e-01 1.09844160e+00 6.07250035e-01 -8.45326424e-01 -1.13869540e-01 -1.19960558e+00 -3.31436768e-02 -9.97054756e-01 -8.72872397e-02 4.44063187e-01 1.06429279e+00 -1.00585473e+00 7.36243606e-01 -6.58134371e-03 -4.72487062e-01 3.47664535e-01 -5.78333698e-02 -2.75177270e-01 2.10935012e-01 -1.57799780e+00 1.16935039e+00 4.57536489e-01 -5.64899683e-01 -6.24321520e-01 -7.11173892e-01 -6.03601992e-01 1.70406457e-02 3.46439064e-01 -3.10205251e-01 7.52610028e-01 -9.23448384e-01 -1.35241401e+00 1.18221700e+00 1.19407251e-01 -7.17186213e-01 6.12843931e-01 -4.34946746e-01 -4.19484168e-01 1.08835384e-01 -1.64157860e-02 4.81807739e-02 7.34100640e-01 -8.65276098e-01 -6.39586449e-02 -4.65001822e-01 -1.50288120e-01 -2.63214737e-01 -8.63568723e-01 4.23355490e-01 -7.43842497e-02 -6.76790774e-01 -5.39772570e-01 -7.56479323e-01 -1.32930785e-01 -5.39134741e-01 -5.55494547e-01 -6.63437665e-01 7.22107232e-01 -1.00650740e+00 1.83230913e+00 -1.77296281e+00 9.99607071e-02 1.43080533e-01 1.21964328e-01 5.45873344e-01 -9.56624225e-02 9.41774189e-01 -2.58376747e-01 5.61445534e-01 -9.50477123e-02 -3.54169607e-01 2.85808086e-01 -4.69397932e-01 -3.31817299e-01 3.82218152e-01 6.44359589e-02 7.33205438e-01 -9.18836415e-01 -5.77900410e-01 -1.58390641e-01 4.07121986e-01 -1.55663520e-01 -1.92449577e-02 -2.83140093e-01 -5.70813045e-02 -5.61957240e-01 3.92192185e-01 3.49425912e-01 1.99683249e-01 2.85166264e-01 4.02113870e-02 -2.24172488e-01 6.03836656e-01 -9.32708919e-01 1.75423431e+00 -5.11485040e-01 4.35515165e-01 -4.06066477e-01 -8.46373379e-01 7.67157674e-01 5.33782482e-01 5.56663394e-01 -5.73587954e-01 2.82772481e-01 1.55593321e-01 -1.54921696e-01 -9.62130129e-01 4.55072403e-01 -8.85635391e-02 -4.55130637e-01 4.55083907e-01 2.24238336e-01 3.52960825e-01 5.30178547e-01 4.65428710e-01 1.62829936e+00 3.15036595e-01 5.44257402e-01 -5.29087305e-01 7.20781147e-01 -1.60502091e-01 5.03361285e-01 5.87394536e-01 -2.80198604e-01 2.78365761e-01 9.27652299e-01 -3.81272733e-01 -9.90613461e-01 -7.04932332e-01 -1.52648777e-01 1.09069586e+00 -3.97077709e-01 -9.79816139e-01 -8.15664947e-01 -1.23914075e+00 6.39087707e-02 9.26150680e-01 -7.60234177e-01 -2.16020048e-01 -4.86197144e-01 -9.06220436e-01 9.25257623e-01 3.52131009e-01 1.83507055e-01 -1.21781683e+00 -6.03615046e-01 3.52443755e-01 -1.15991428e-01 -8.83619964e-01 -3.64389688e-01 4.86783952e-01 -6.14663541e-01 -1.04584503e+00 -4.70935196e-01 -4.05745029e-01 -1.67599916e-01 -2.04937100e-01 1.37657821e+00 2.05017641e-01 -3.58965695e-01 2.29941368e-01 -5.32011807e-01 -4.96967763e-01 -3.61754835e-01 1.92964807e-01 -3.48450392e-02 -1.22560121e-01 8.45268011e-01 -6.67603314e-01 -3.29150349e-01 -3.42312545e-01 -8.88411164e-01 -4.98425603e-01 6.06374025e-01 6.23094201e-01 4.21734788e-02 -3.84517670e-01 8.50346625e-01 -1.17666924e+00 8.71203303e-01 -8.25016558e-01 -8.26491714e-02 1.99965879e-01 -9.18120742e-01 1.50248960e-01 7.23198116e-01 -6.55280411e-01 -7.07670629e-01 9.93409604e-02 -3.33282769e-01 4.30952907e-02 9.86317638e-03 4.79963988e-01 -3.57476354e-01 1.82462588e-01 8.25830698e-01 1.83861092e-01 -4.57725108e-01 -7.62384832e-01 4.51610297e-01 7.14025378e-01 2.16785688e-02 -3.34404945e-01 7.30344415e-01 -5.88857569e-02 -1.83969796e-01 -7.22512007e-01 -9.26128864e-01 -6.06963396e-01 -6.30852044e-01 -1.00679092e-01 1.02075315e+00 -6.94727063e-01 -4.26448613e-01 2.33865768e-01 -1.03925681e+00 -1.57464236e-01 -1.44094393e-01 3.06158930e-01 -2.98042923e-01 6.01695061e-01 -6.43024206e-01 -9.72218394e-01 -6.46207273e-01 -6.28010929e-01 7.21103072e-01 -7.31703937e-02 -6.22225285e-01 -9.96257246e-01 4.84566361e-01 4.27299172e-01 3.19047391e-01 5.20008028e-01 1.00576687e+00 -1.35359204e+00 4.16303612e-02 -2.79333591e-01 8.85880366e-02 1.80702463e-01 -3.44363987e-01 -1.01353571e-01 -9.86542523e-01 7.43468180e-02 -1.67218596e-01 -6.59638286e-01 1.24978793e+00 -2.02012762e-01 1.05658674e+00 -2.83789426e-01 -2.87814587e-01 -5.78845553e-02 1.65272963e+00 -1.19597822e-01 7.49214292e-01 5.98709702e-01 4.15737182e-01 6.32508516e-01 4.01197821e-01 6.21043980e-01 2.01430902e-01 6.72795951e-01 2.36836880e-01 3.99999291e-01 -2.67628819e-01 -4.29730237e-01 8.52765203e-01 6.31033361e-01 3.22285220e-02 -3.92474145e-01 -8.85168970e-01 6.89215899e-01 -1.67434716e+00 -1.35840750e+00 -4.87909108e-01 2.11400604e+00 9.99218285e-01 4.52007115e-01 5.21350622e-01 3.72348577e-01 4.29577142e-01 4.42376852e-01 -1.82474613e-01 -8.31339478e-01 -1.35593012e-01 3.90143484e-01 4.80787098e-01 1.41887188e-01 -1.09684551e+00 7.94188499e-01 6.38080502e+00 1.19586706e+00 -9.22680199e-01 4.50883657e-01 2.46950582e-01 -6.18211031e-02 -4.71087664e-01 -4.22422588e-02 -7.66728163e-01 9.36039627e-01 1.47705686e+00 -9.05088410e-02 1.07078575e-01 7.09490836e-01 3.00892889e-01 -2.21130461e-01 -9.52082932e-01 5.26249766e-01 4.78221774e-01 -1.25678909e+00 4.32389006e-02 1.37761638e-01 4.23604608e-01 -1.13856584e-01 -3.02273761e-02 5.45441985e-01 2.51106858e-01 -8.69874179e-01 8.67194533e-01 6.53968692e-01 3.45557123e-01 -6.67506039e-01 7.77947485e-01 3.60075384e-01 -8.14222991e-01 -1.59341291e-01 -1.06605485e-01 -7.00412644e-03 8.83824676e-02 9.33351696e-01 -3.43698502e-01 3.21489692e-01 6.55068457e-01 4.33458686e-01 -8.74863327e-01 1.04323232e+00 -3.78385782e-01 9.71119761e-01 1.04480781e-01 -4.74801779e-01 2.19457492e-01 3.79702747e-02 7.33176649e-01 1.81088221e+00 4.31401692e-02 -3.31235141e-01 4.78811562e-02 6.17886007e-01 -1.96128577e-01 6.04581118e-01 -8.89479160e-01 -3.26326013e-01 4.20549870e-01 1.56026268e+00 -6.23557925e-01 -4.41776156e-01 -3.03045362e-01 8.54684949e-01 5.96006334e-01 -1.50710136e-01 -9.10193384e-01 -5.43814003e-01 1.72825292e-01 1.85234696e-01 2.78243214e-01 -1.26394048e-01 -6.22433782e-01 -1.05650318e+00 -4.54070419e-02 -7.95823574e-01 5.05007088e-01 -4.93775368e-01 -1.29288721e+00 5.24926603e-01 -1.54694468e-01 -9.22844827e-01 -2.33325794e-01 -5.72730720e-01 -8.21661055e-01 7.38179505e-01 -1.25055182e+00 -1.20578623e+00 9.04203728e-02 2.78722197e-01 6.53876543e-01 -1.05003677e-01 8.16871524e-01 3.64151806e-01 -5.63741863e-01 5.83929002e-01 -1.10982388e-01 -2.77688578e-02 9.53000844e-01 -1.52120602e+00 -5.42734228e-02 8.87898684e-01 1.47285640e-01 6.44187212e-01 8.79470468e-01 -6.88738406e-01 -1.16402745e+00 -9.88016248e-01 1.71800530e+00 -7.66932726e-01 1.01042080e+00 -5.55708408e-01 -8.26539040e-01 4.32251543e-01 6.78555965e-01 -5.21859884e-01 1.14578104e+00 3.31820518e-01 -6.97892547e-01 2.58811653e-01 -1.18454051e+00 5.26310146e-01 1.21800506e+00 -4.93869752e-01 -1.03483737e+00 4.86744672e-01 6.56403601e-01 3.28144461e-01 -1.00982475e+00 -2.03692522e-02 5.06874502e-01 -1.05215335e+00 7.62786388e-01 -8.21207762e-01 8.06319535e-01 3.56347635e-02 -1.29563674e-01 -1.33525848e+00 -4.35063630e-01 -6.60401523e-01 -1.77067608e-01 1.71787500e+00 5.99510729e-01 -3.59731585e-01 4.79326665e-01 2.80612618e-01 -5.83026707e-02 -8.11695755e-01 -8.16434503e-01 -7.36582160e-01 3.31104398e-01 -3.36099267e-01 1.89670995e-01 8.94373894e-01 3.62819254e-01 6.72554791e-01 -4.34607655e-01 -3.67525578e-01 4.36471641e-01 -1.98946103e-01 4.15922701e-01 -1.26085401e+00 -1.24249995e-01 -5.60749292e-01 -4.22698647e-01 -2.36286610e-01 4.63561535e-01 -9.41907525e-01 -2.78251708e-01 -1.11521423e+00 5.09293914e-01 2.60762330e-02 -4.96516198e-01 4.32210803e-01 -2.42947027e-01 4.42181408e-01 1.05998896e-01 -1.65295273e-01 -6.55437469e-01 3.54657203e-01 4.07667458e-01 -1.10906279e-02 -1.38873383e-01 -1.58021420e-01 -1.04805815e+00 7.95267224e-01 9.02421653e-01 -7.29769289e-01 -2.37459213e-01 1.92051932e-01 3.55252117e-01 -4.15078193e-01 1.64192408e-01 -9.75639582e-01 -9.79203507e-02 8.92682374e-02 2.99158096e-01 -5.27795315e-01 9.66297537e-02 -4.81548995e-01 -2.05591112e-01 6.82411253e-01 -6.15785182e-01 1.22145422e-01 4.26402166e-02 7.36137748e-01 1.41633302e-01 -6.98961377e-01 6.61822736e-01 -3.79259467e-01 -7.00723946e-01 -4.00804058e-02 -6.06185317e-01 2.18853399e-01 1.12344933e+00 -4.24608169e-03 -3.16066474e-01 -5.99883310e-02 -6.27460063e-01 -1.58224195e-01 1.69314831e-01 4.66452450e-01 1.89330399e-01 -1.05543172e+00 -7.70867765e-01 -4.57741827e-01 2.68673807e-01 -1.20545220e+00 -2.89030038e-02 1.05439544e+00 -4.10479568e-02 3.21220219e-01 -9.06881616e-02 1.34145886e-01 -1.30794334e+00 9.95568812e-01 7.54511729e-03 -9.16963756e-01 -4.51294899e-01 6.11755788e-01 -4.38216299e-01 -1.23782776e-01 1.29652441e-01 1.97171539e-01 -6.76480532e-01 5.28152525e-01 7.22648680e-01 5.42122483e-01 2.00190604e-01 -6.12471044e-01 -4.21497613e-01 2.85838664e-01 -5.95147386e-02 -2.39920720e-01 1.34105635e+00 -1.28482310e-02 -1.13444239e-01 6.88825071e-01 1.28472102e+00 5.81774056e-01 -4.88516539e-01 -2.47461706e-01 7.29103446e-01 -1.66789740e-01 1.00554399e-01 -1.22584105e+00 -4.27738309e-01 7.80448973e-01 5.17484963e-01 4.45328861e-01 7.75889635e-01 -8.74434635e-02 6.85566366e-01 1.73833609e-01 1.73333347e-01 -1.35768676e+00 2.77132958e-01 4.08423394e-01 9.10131991e-01 -9.92486894e-01 1.30134866e-01 -2.40947887e-01 -6.57526135e-01 1.24658203e+00 4.38929439e-01 -2.46239066e-01 4.97827232e-01 3.56782600e-02 -1.95098549e-01 -2.88349926e-01 -1.00624382e+00 -4.59323376e-01 2.65123427e-01 3.39364529e-01 7.96562254e-01 -9.53518078e-02 -1.37178433e+00 1.03971100e+00 -1.09937906e-01 -1.19701289e-01 4.62015808e-01 7.74356067e-01 -5.39761364e-01 -1.21131051e+00 -8.90112966e-02 4.64236796e-01 -1.06257236e+00 -3.10815126e-01 -9.38210249e-01 9.06844735e-01 4.67509121e-01 9.66226995e-01 -5.27212560e-01 -6.30615413e-01 3.17377925e-01 8.39684784e-01 2.43050516e-01 -6.09334946e-01 -1.35424972e+00 -1.78856611e-01 6.07011199e-01 -4.99982595e-01 -2.71102041e-01 -8.42459738e-01 -7.05075204e-01 -1.80242151e-01 -2.56333917e-01 2.75387615e-01 7.08370686e-01 8.96977425e-01 2.53223926e-01 3.50627840e-01 6.52546406e-01 -4.51739967e-01 -9.02000487e-01 -1.22833359e+00 -7.15477526e-01 6.75492823e-01 -1.03971682e-01 -5.31684041e-01 -4.29330856e-01 -8.78812149e-02]

[8.910737991333008, 10.61579418182373]

00ff2dea-8739-4b9c-bd03-d8656953d515

maniqa-multi-dimension-attention-network-for

2204.08958

https://arxiv.org/abs/2204.08958v2

https://arxiv.org/pdf/2204.08958v2.pdf

MANIQA: Multi-dimension Attention Network for No-Reference Image Quality Assessment

No-Reference Image Quality Assessment (NR-IQA) aims to assess the perceptual quality of images in accordance with human subjective perception. Unfortunately, existing NR-IQA methods are far from meeting the needs of predicting accurate quality scores on GAN-based distortion images. To this end, we propose Multi-dimension Attention Network for no-reference Image Quality Assessment (MANIQA) to improve the performance on GAN-based distortion. We firstly extract features via ViT, then to strengthen global and local interactions, we propose the Transposed Attention Block (TAB) and the Scale Swin Transformer Block (SSTB). These two modules apply attention mechanisms across the channel and spatial dimension, respectively. In this multi-dimensional manner, the modules cooperatively increase the interaction among different regions of images globally and locally. Finally, a dual branch structure for patch-weighted quality prediction is applied to predict the final score depending on the weight of each patch's score. Experimental results demonstrate that MANIQA outperforms state-of-the-art methods on four standard datasets (LIVE, TID2013, CSIQ, and KADID-10K) by a large margin. Besides, our method ranked first place in the final testing phase of the NTIRE 2022 Perceptual Image Quality Assessment Challenge Track 2: No-Reference. Codes and models are available at https://github.com/IIGROUP/MANIQA.

['Yuan Gong', 'Shanshan Lao', 'Yujiu Yang', 'Jiahao Wang', 'Mingdeng Cao', 'Shuwei Shi', 'Tianhe Wu', 'Sidi Yang']

2022-04-19

['no-reference-image-quality-assessment']

['computer-vision']

[-2.63599493e-02 -3.67742836e-01 2.35730391e-02 -3.29932094e-01 -1.10003972e+00 -1.98256016e-01 3.69294107e-01 -4.01876241e-01 2.50501744e-02 4.05524790e-01 5.56411445e-01 -1.54440328e-01 -1.84062436e-01 -8.05400014e-01 -5.60169518e-01 -6.63841546e-01 1.76541641e-01 -9.29052010e-02 5.10561503e-02 -3.33269924e-01 8.63291696e-02 2.73470283e-01 -1.10101616e+00 3.79232883e-01 1.17162251e+00 1.37853622e+00 1.10228926e-01 7.53617167e-01 3.13036859e-01 9.12728608e-01 -5.09232521e-01 -7.15243816e-01 3.44729781e-01 -8.03093970e-01 -7.40076363e-01 5.09045199e-02 3.48462164e-01 -6.54621303e-01 -6.75525010e-01 1.14367867e+00 9.18640137e-01 -6.62372112e-02 5.05704284e-01 -1.25098777e+00 -1.12762046e+00 4.42073673e-01 -6.88964605e-01 5.59735537e-01 3.86392921e-02 6.07628405e-01 1.04973602e+00 -9.38528061e-01 1.95085689e-01 1.19169009e+00 4.22667563e-01 4.65018690e-01 -9.81057882e-01 -7.48355687e-01 8.42954293e-02 7.86456347e-01 -1.24614310e+00 -5.00308454e-01 7.72592366e-01 -2.96565026e-01 8.59139800e-01 1.27605110e-01 5.38160980e-01 8.16098273e-01 4.12953764e-01 6.84222221e-01 1.09485161e+00 2.68475395e-02 8.42635185e-02 -3.84910345e-01 -3.22824478e-01 5.07168770e-01 -3.68478447e-01 3.64351690e-01 -5.47859907e-01 3.08963180e-01 8.16874027e-01 -3.28569770e-01 -4.40893888e-01 -7.86199272e-02 -1.28441238e+00 6.17829621e-01 9.61620748e-01 2.86992908e-01 -5.54065168e-01 3.32355112e-01 3.85523528e-01 4.18766648e-01 5.35033643e-01 3.48469824e-01 -3.77132088e-01 -2.49474585e-01 -8.19908679e-01 -7.98389241e-02 -1.15570493e-01 7.12528169e-01 4.41599399e-01 2.12064117e-01 -6.52474880e-01 1.06196082e+00 2.77725130e-01 5.18419504e-01 4.90748346e-01 -1.17886949e+00 5.56400776e-01 3.22711885e-01 6.81114569e-02 -1.02018595e+00 -1.37639925e-01 -7.02487230e-01 -1.35175788e+00 4.43107903e-01 6.26898482e-02 2.27662595e-03 -9.54933584e-01 1.64177752e+00 5.31048626e-02 2.07740545e-01 -1.63181677e-01 1.26484132e+00 9.07037318e-01 9.71060872e-01 6.77984115e-03 -1.85829133e-01 1.18855524e+00 -1.26380324e+00 -8.08332503e-01 1.62747562e-01 1.34508818e-01 -7.28767931e-01 1.24308443e+00 6.24978781e-01 -1.54745388e+00 -1.16988158e+00 -1.11966574e+00 -1.59827754e-01 -7.24616945e-02 1.94730043e-01 1.89260989e-01 6.18699074e-01 -1.39558947e+00 4.82649803e-01 -6.03357792e-01 1.16512775e-01 7.18926728e-01 2.83623368e-01 3.99910705e-03 -3.67300391e-01 -1.31287301e+00 6.53946877e-01 -1.57906115e-01 2.23358557e-01 -1.15762269e+00 -7.95438468e-01 -5.65161705e-01 1.69573382e-01 2.49862317e-02 -8.34332407e-01 1.00249660e+00 -1.28047299e+00 -1.61432409e+00 5.81588924e-01 -4.90386132e-03 -4.15087849e-01 3.15737873e-01 -1.29003435e-01 -7.91275620e-01 1.53948143e-01 1.87517665e-02 9.42278028e-01 7.94062376e-01 -1.26559258e+00 -5.88529170e-01 -3.34771037e-01 2.83151835e-01 3.70427340e-01 -2.03659713e-01 -4.88561168e-02 -7.73968697e-01 -9.44156229e-01 -2.60867141e-02 -4.02820706e-01 -6.47935793e-02 3.68982852e-02 -2.70985097e-01 -1.01923719e-01 5.75896800e-01 -1.04368699e+00 1.43768549e+00 -2.24385166e+00 3.25489581e-01 1.04909249e-01 4.59533572e-01 3.58942956e-01 -6.39574766e-01 -5.63264415e-02 -1.00304179e-01 2.65924186e-01 -1.30434558e-01 -1.93054125e-01 6.99913129e-02 -1.42001808e-01 1.71063486e-02 3.21932554e-01 4.44925427e-01 1.17670429e+00 -7.11454391e-01 -2.61983335e-01 3.38654697e-01 6.16577566e-01 -6.88383758e-01 4.22068447e-01 1.10218346e-01 5.67744255e-01 -1.38394967e-01 7.33144283e-01 9.44311142e-01 -3.55853558e-01 -2.10842431e-01 -7.58439839e-01 5.88566661e-02 2.54863143e-01 -9.44395065e-01 1.93872726e+00 -7.58263409e-01 4.83443588e-01 -4.37005199e-02 -6.25080526e-01 8.02027583e-01 3.28427911e-01 4.84481007e-01 -1.59496582e+00 1.97666548e-02 1.22847170e-01 2.29054824e-01 -2.67449498e-01 4.04881418e-01 2.25655138e-02 1.04637235e-01 1.39281869e-01 3.22757661e-01 1.72443300e-01 6.57676673e-03 2.68518120e-01 1.05251610e+00 -1.25409663e-01 7.40603358e-02 -7.59319067e-02 5.71502864e-01 -6.67465508e-01 7.10271955e-01 3.48001957e-01 -4.96270537e-01 9.31853414e-01 3.18241179e-01 -2.94942260e-01 -1.21818030e+00 -1.30229330e+00 3.31366844e-02 8.64260018e-01 4.20626193e-01 -2.73469359e-01 -7.47834444e-01 -5.97427726e-01 -5.15634179e-01 3.90130669e-01 -7.72443891e-01 -3.02727818e-01 -2.93782592e-01 -6.43808484e-01 2.46238291e-01 5.31589270e-01 1.05428267e+00 -1.21803713e+00 -2.18824327e-01 2.14637101e-01 -5.05072594e-01 -7.94572175e-01 -8.58570635e-01 -3.44233811e-01 -4.81888622e-01 -8.85854602e-01 -1.06683028e+00 -4.42077130e-01 2.31071100e-01 2.45828569e-01 1.42378044e+00 1.70828179e-01 -3.37989964e-02 2.20941275e-01 -5.48006892e-01 5.72807007e-02 -1.85712829e-01 1.19910985e-02 -3.34923744e-01 1.93676963e-01 -1.56134054e-01 -7.12863207e-01 -1.17916310e+00 5.41684210e-01 -8.66099834e-01 1.53072014e-01 7.83583879e-01 9.64078248e-01 7.57541656e-01 2.15380669e-01 8.80714118e-01 -2.32167214e-01 5.88656068e-01 -4.49891299e-01 -4.29357886e-01 2.35711306e-01 -6.99325919e-01 -3.22133809e-01 5.88140488e-01 -7.74175301e-02 -1.07220030e+00 -4.91734922e-01 -5.60387790e-01 -4.86876011e-01 1.71952567e-03 2.30088174e-01 -7.86247492e-01 4.46793586e-02 3.86487037e-01 2.98572272e-01 -3.98711264e-01 -3.08416426e-01 4.09175783e-01 5.54359496e-01 5.96651673e-01 -2.37012893e-01 7.14866400e-01 1.18408903e-01 -2.17227057e-01 -5.14050648e-02 -6.21169806e-01 -1.73983753e-01 -3.34920257e-01 -4.50457782e-01 9.77918208e-01 -1.25257409e+00 -5.18145263e-01 8.19543183e-01 -9.83226061e-01 -6.11284852e-01 -3.63453269e-01 3.10159326e-01 -5.76802552e-01 1.38485059e-01 -6.92115664e-01 -4.58875448e-01 -6.33180380e-01 -1.41046739e+00 9.93638158e-01 2.71131605e-01 3.71764332e-01 -7.02880919e-01 -1.00828968e-01 5.95658302e-01 7.63539493e-01 2.05577724e-02 8.26929510e-01 2.07765624e-01 -6.11829102e-01 3.39547008e-01 -6.75694168e-01 7.50324726e-01 1.49480224e-01 -2.82219887e-01 -9.38038528e-01 -4.07480538e-01 -2.23795339e-01 -4.59196627e-01 8.63090277e-01 6.40247226e-01 1.54160929e+00 -3.27207804e-01 4.26290244e-01 8.68071496e-01 1.46787131e+00 4.35731351e-01 1.25544119e+00 1.99955940e-01 7.89588630e-01 4.72952314e-02 6.10871732e-01 3.76535028e-01 5.96089303e-01 9.77720261e-01 6.97721541e-01 -5.11892319e-01 -7.59189606e-01 -1.34616569e-01 4.87690240e-01 9.95978892e-01 -2.48842202e-02 -5.40925682e-01 -6.74368024e-01 6.36456013e-01 -1.39738095e+00 -9.61017370e-01 -6.50855713e-03 1.97253716e+00 7.68929482e-01 8.22883025e-02 1.77444294e-01 2.89222866e-01 4.81261730e-01 1.77558318e-01 -8.83666337e-01 -3.97449166e-01 -3.59183937e-01 2.17673853e-01 2.55063593e-01 4.42704320e-01 -9.45639729e-01 5.34579813e-01 5.32780886e+00 1.26815414e+00 -1.04580605e+00 4.64442730e-01 1.38153219e+00 -1.53734520e-01 -4.39460486e-01 -3.81118506e-01 -1.37341097e-01 7.63895929e-01 9.90059018e-01 -2.25623325e-02 8.20382595e-01 3.50411087e-01 3.90735984e-01 4.56882976e-02 -6.98464930e-01 1.12266719e+00 -2.12680213e-02 -1.23482847e+00 1.38435975e-01 1.01313934e-01 1.02494788e+00 1.10299803e-01 5.18313110e-01 2.94799030e-01 2.03157142e-01 -1.22216380e+00 7.47035146e-01 6.32693410e-01 1.21864843e+00 -9.71032441e-01 9.05815244e-01 -1.63834691e-01 -1.30885255e+00 -2.31521368e-01 -4.16640490e-01 3.60832155e-01 2.37283766e-01 7.63815224e-01 1.17411375e-01 8.59135151e-01 9.79151011e-01 8.74127090e-01 -8.56415451e-01 1.09392214e+00 -2.06236392e-01 7.27914989e-01 2.35642329e-01 6.25191033e-01 2.38441646e-01 -5.47993705e-02 3.20923537e-01 8.61031115e-01 6.36297941e-01 3.81769776e-01 -3.66628438e-01 8.28948438e-01 -4.53066081e-01 -4.77378927e-02 5.85373677e-03 3.90811801e-01 2.11286470e-01 1.15291166e+00 -2.63981223e-01 -3.62459302e-01 -3.94028395e-01 1.25137615e+00 -1.40807247e-02 3.56793702e-01 -1.08439302e+00 -2.89161503e-01 7.41963923e-01 9.14215520e-02 4.43595350e-01 1.42630935e-01 -3.85374397e-01 -1.06796086e+00 6.75118864e-02 -1.26224840e+00 1.52053908e-01 -1.26859510e+00 -1.44612861e+00 8.50325525e-01 -5.35434604e-01 -1.47822666e+00 2.62857497e-01 -2.10313886e-01 -6.70385480e-01 1.04601395e+00 -1.66928232e+00 -1.19198143e+00 -5.69328725e-01 8.53237152e-01 6.78168297e-01 -1.47433057e-01 5.99581182e-01 8.23776245e-01 -5.88573754e-01 9.73595917e-01 6.01576492e-02 1.15515679e-01 7.44433999e-01 -1.15207040e+00 6.66053057e-01 1.00255477e+00 1.88481547e-02 -4.43393253e-02 2.20964268e-01 -3.42349410e-01 -1.15750992e+00 -1.27467775e+00 2.96037942e-01 -4.45998758e-02 3.53304237e-01 -2.75883544e-02 -8.44129622e-01 3.75931337e-02 5.70442736e-01 2.78342813e-01 4.63354796e-01 -1.67479068e-01 -4.91784155e-01 -5.70007205e-01 -1.13521004e+00 2.75155872e-01 1.01881468e+00 -4.84315038e-01 1.81979343e-01 4.89872023e-02 9.28325415e-01 -2.58569479e-01 -1.22345054e+00 4.99865204e-01 4.27379280e-01 -1.32636452e+00 9.66443717e-01 -5.89871220e-02 9.10327792e-01 -6.13242149e-01 -3.88743669e-01 -1.62807000e+00 -7.64441371e-01 -3.03039193e-01 -8.73637497e-02 1.40549934e+00 2.68681645e-01 -2.40725219e-01 3.89099240e-01 9.28517431e-02 -3.82950008e-01 -1.07573831e+00 -1.03982484e+00 -5.63168049e-01 1.96460456e-01 -3.20178747e-01 9.49238896e-01 6.92758083e-01 -4.86014009e-01 3.47380221e-01 -7.10183322e-01 1.80631116e-01 5.84417582e-01 -8.99401233e-02 5.73322833e-01 -5.17220199e-01 -5.42669117e-01 -6.49549603e-01 -5.43506265e-01 -1.20526397e+00 -4.01014328e-01 -6.14950061e-01 -7.83859268e-02 -1.72653353e+00 3.60115200e-01 -4.58234578e-01 -7.44287729e-01 2.05012649e-01 -4.26866651e-01 6.27541244e-01 4.37225401e-01 1.32369816e-01 -9.25069511e-01 1.05011880e+00 1.56749451e+00 -4.36699241e-01 1.31410018e-01 -2.57565707e-01 -8.38748693e-01 1.96342036e-01 9.55063283e-01 -1.29604712e-01 -5.25866032e-01 -8.57309401e-01 1.26270950e-01 2.12494925e-01 4.67392087e-01 -1.33625793e+00 -1.28661375e-02 -2.26778779e-02 6.34246051e-01 -5.15794754e-01 1.82324201e-01 -6.46565318e-01 2.16684297e-01 2.55665839e-01 -3.50972623e-01 2.19309807e-01 6.26947582e-02 3.70466590e-01 -5.04311323e-01 4.79742318e-01 1.12561035e+00 2.12138623e-01 -7.32961118e-01 7.79439211e-01 -2.33007111e-02 9.27402750e-02 6.84938312e-01 2.18629539e-02 -4.19455618e-01 -5.80514073e-01 -5.73680580e-01 2.23267555e-01 4.06420350e-01 4.04788554e-01 8.99321139e-01 -1.82338715e+00 -9.94460821e-01 1.50554359e-01 1.79333344e-01 -2.30789393e-01 1.01320732e+00 8.11018109e-01 -2.30249301e-01 1.20788850e-01 -6.21806860e-01 -4.15320843e-01 -9.73032236e-01 4.86990988e-01 6.17945135e-01 -6.21898711e-01 -2.70940363e-01 9.55751836e-01 4.81422633e-01 -1.23583347e-01 8.80278647e-02 -1.19905859e-01 -1.72569498e-01 -3.14999640e-01 7.23261118e-01 3.91070634e-01 2.48384506e-01 -9.19443965e-01 -1.84297219e-01 5.50812721e-01 1.05171226e-01 -4.10811566e-02 1.17414463e+00 -5.43232143e-01 8.81152749e-02 1.15124613e-01 1.24433386e+00 -2.69382358e-01 -1.58577609e+00 -2.55294859e-01 -6.29300117e-01 -6.83862865e-01 4.02205259e-01 -1.41223443e+00 -1.86543703e+00 1.10461318e+00 1.33551919e+00 6.66737650e-03 2.08669543e+00 -1.17853321e-01 1.04484022e+00 -4.15962607e-01 1.30131230e-01 -7.79427826e-01 4.44501251e-01 1.36779517e-01 1.30831671e+00 -1.25208390e+00 -1.59272164e-01 -7.54286163e-03 -6.85702443e-01 6.46745563e-01 8.82942200e-01 -5.47880027e-03 6.51154101e-01 6.96137920e-02 1.70899004e-01 2.65388470e-02 -8.60410690e-01 -3.53374816e-02 6.20574594e-01 7.66514361e-01 4.13189083e-01 1.34897098e-01 -1.23474494e-01 7.36205518e-01 8.49638414e-03 -5.31951115e-02 3.50741804e-01 1.48762703e-01 -9.98837054e-02 -9.59655523e-01 -1.84360966e-01 5.49644351e-01 -5.35783947e-01 -3.47168714e-01 6.27750829e-02 1.80830002e-01 5.45229197e-01 1.22748673e+00 -3.44230980e-03 -8.83290887e-01 4.49351937e-01 -6.88577890e-01 2.64810890e-01 1.77179631e-02 -6.83108091e-01 1.21805429e-01 -1.80737704e-01 -9.83474076e-01 -4.72792506e-01 -3.51092100e-01 -7.22623587e-01 -5.10863662e-01 -1.05105944e-01 -5.66866919e-02 4.73041207e-01 5.48370361e-01 5.41482508e-01 1.06916451e+00 1.07259452e+00 -7.76132822e-01 -1.01117603e-01 -1.02160454e+00 -5.19128084e-01 3.71677667e-01 3.34201187e-01 -3.71897995e-01 -6.65187910e-02 8.40699971e-02]

[11.834555625915527, -1.7728261947631836]

330ca0f0-7635-4f51-bce8-61e27fa36262

more-like-this-semantic-retrieval-with

https://aclanthology.org/2022.konvens-1.19

https://aclanthology.org/2022.konvens-1.19.pdf

More Like This: Semantic Retrieval with Linguistic Information

['Chris Biemann', 'Seid Muhie Yimam', 'Fynn Petersen-Frey', 'Saba Anwar', 'Gregor Wiedemann', 'Steffen Remus']

konvens-ws-2022-9

['semantic-retrieval']

['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]

[-7.213517665863037, 3.6255335807800293]

e514bf0a-0082-41ee-9028-091f90b232f7

holographic-visualisation-of-radiology-data

1808.04929

http://arxiv.org/abs/1808.04929v1

http://arxiv.org/pdf/1808.04929v1.pdf

Holographic Visualisation of Radiology Data and Automated Machine Learning-based Medical Image Segmentation

Within this thesis we propose a platform for combining Augmented Reality (AR) hardware with machine learning in a user-oriented pipeline, offering to the medical staff an intuitive 3D visualization of volumetric Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) medical image segmentations inside the AR headset, that does not need human intervention for loading, processing and segmentation of medical images. The AR visualization, based on Microsoft HoloLens, employs a modular and thus scalable frontend-backend architecture for real-time visualizations on multiple AR headsets. As Convolutional Neural Networks (CNNs) have lastly demonstrated superior performance for the machine learning task of image semantic segmentation, the pipeline also includes a fully automated CNN algorithm for the segmentation of the liver from CT scans. The model is based on the Deep Retinal Image Understanding (DRIU) model which is a Fully Convolutional Network with side outputs from feature maps with different resolution, extracted at different stages of the network. The algorithm is 2.5D which means that the input is a set of consecutive scan slices. The experiments have been performed on the Liver Tumor Segmentation Challenge (LiTS) dataset for liver segmentation and demonstrated good results and flexibility. While multiple approaches exist in the domain, only few of them have focused on overcoming the practical aspects which still largely hold this technology away from the operating rooms. In line with this, we also are next planning an evaluation from medical doctors and radiologists in a real-world environment.

['Lucian Trestioreanu']

2018-08-15

['liver-segmentation']

['medical']

[-7.94570968e-02 5.37561357e-01 3.15443963e-01 -3.53138000e-01 -4.65537459e-01 -2.71760255e-01 1.87072605e-01 3.04521322e-01 -3.98434639e-01 6.18903823e-02 2.00867951e-02 -9.79053736e-01 1.30177271e-02 -6.33431256e-01 -3.51897299e-01 -4.44984704e-01 -2.31937692e-01 6.34074807e-01 1.31748199e-01 -1.53450429e-01 7.44211674e-02 1.11596978e+00 -1.18033850e+00 4.55462515e-01 2.22928360e-01 8.95018518e-01 3.02708119e-01 1.19302511e+00 -2.12310165e-01 6.74486637e-01 -4.45012510e-01 3.12800974e-01 4.26502109e-01 -1.08679533e-01 -7.06782103e-01 -7.04134628e-02 2.42135033e-01 -3.55616033e-01 -2.08021149e-01 6.50779665e-01 7.66825974e-01 -2.75887430e-01 1.11533239e-01 -6.60775185e-01 -3.86442542e-01 4.88258362e-01 -4.97849077e-01 5.98084331e-01 4.55315828e-01 4.01184291e-01 1.94134176e-01 -7.53802359e-01 6.97979748e-01 9.36706960e-01 7.48484194e-01 5.48016131e-01 -9.55094516e-01 -4.86024648e-01 -3.32392454e-01 -3.44911307e-01 -1.08930171e+00 -1.14312544e-01 2.67597497e-01 -6.37117803e-01 1.08324528e+00 5.44620216e-01 1.07539487e+00 4.00266618e-01 6.47544265e-01 5.01218319e-01 1.16787326e+00 -5.18290341e-01 2.37133831e-01 6.01662397e-01 4.02755380e-01 1.06849146e+00 6.12641163e-02 3.85292768e-01 -1.26673773e-01 1.86919406e-01 1.34517586e+00 3.13870579e-01 -4.10907298e-01 -4.94090378e-01 -1.47909296e+00 8.69117022e-01 9.80632544e-01 5.73393404e-01 -6.11688316e-01 2.03327209e-01 4.99683052e-01 3.12145740e-01 2.74888992e-01 3.18840116e-01 -6.70557678e-01 2.47920439e-01 -9.02694404e-01 -4.75164413e-01 7.32523620e-01 6.37063205e-01 3.32419693e-01 4.89423573e-02 -4.52618450e-02 7.23658800e-02 7.76966095e-01 -1.27972692e-01 6.39208436e-01 -4.79240417e-01 -1.59137174e-01 7.80145884e-01 -1.72744378e-01 -5.07009804e-01 -1.06576824e+00 -5.78362405e-01 -8.11088681e-01 9.31634605e-01 3.80724937e-01 -1.25939503e-01 -1.27817631e+00 7.02915609e-01 6.32475674e-01 2.62423575e-01 1.83563039e-01 1.08470058e+00 1.78230464e+00 1.21489748e-01 2.39629477e-01 2.11012840e-01 1.89258814e+00 -9.63575304e-01 -5.28771520e-01 1.72403827e-01 1.08874631e+00 -7.68277705e-01 1.11921990e+00 6.48997486e-01 -1.23697257e+00 -5.15695870e-01 -1.28843534e+00 -3.05063635e-01 -6.42601788e-01 3.42001528e-01 8.76609206e-01 1.06684017e+00 -1.49154377e+00 4.63271469e-01 -1.29161584e+00 -3.54886204e-01 6.50024951e-01 6.90147460e-01 -5.34948051e-01 5.83310612e-02 -5.14381945e-01 1.15575802e+00 2.82757163e-01 4.13887948e-01 -7.94682980e-01 -1.26684034e+00 -9.47095573e-01 -1.38765037e-01 -4.53596003e-02 -1.07903945e+00 1.23168111e+00 -8.89199615e-01 -1.65743113e+00 1.23134589e+00 3.53318095e-01 -6.34354830e-01 7.89372742e-01 -2.45396774e-02 -1.24215312e-01 3.74208689e-01 -4.24367279e-01 9.33627903e-01 2.80988246e-01 -1.14874876e+00 -5.55188656e-01 -5.57155728e-01 2.68575009e-02 1.76840588e-01 5.65139651e-01 1.82628691e-01 -2.56939530e-01 -1.50631920e-01 4.06009972e-01 -7.10114419e-01 -7.17830420e-01 7.48316795e-02 -3.14393699e-01 1.92224205e-01 9.01247621e-01 -8.42890918e-01 3.75003219e-01 -2.04994416e+00 -1.75948650e-01 1.07816271e-01 4.96267766e-01 2.44280368e-01 3.44310045e-01 -2.71202564e-01 -6.67715847e-01 -6.00017868e-02 2.17755213e-01 -2.97977507e-01 -4.70136940e-01 7.27166841e-03 1.02789931e-01 6.48765743e-01 -3.38323057e-01 1.10411918e+00 -5.86726904e-01 -4.95501190e-01 8.89188409e-01 9.55426037e-01 -5.78636639e-02 7.61384666e-02 9.59965289e-02 1.23990881e+00 -2.57309735e-01 5.36493003e-01 6.26583040e-01 -4.02004153e-01 -6.55386373e-02 -1.96366623e-01 -3.21260840e-01 -1.97527441e-03 -1.03079343e+00 2.24562883e+00 -8.34687829e-01 6.44761801e-01 1.57338873e-01 -5.89898348e-01 7.63687313e-01 5.92775524e-01 6.65655255e-01 -6.94169939e-01 5.15861034e-01 5.92591949e-02 3.96511005e-03 -7.40738034e-01 3.07919651e-01 -9.21078548e-02 4.88430619e-01 4.41122979e-01 -1.13933526e-01 -8.04459453e-02 -2.89472878e-01 1.25552597e-03 1.04352415e+00 2.18705103e-01 5.08446872e-01 -1.79346532e-01 4.81128097e-01 5.02300680e-01 -2.00162023e-01 4.00339395e-01 -1.53238341e-01 7.94724047e-01 2.74346769e-01 -9.82832909e-01 -8.43113422e-01 -9.40677166e-01 -3.19464654e-01 6.81959391e-01 -9.57776010e-02 -9.06757563e-02 -6.67763114e-01 -6.35936737e-01 -4.87125993e-01 5.15030086e-01 -7.81912804e-01 4.79177624e-01 -5.91864526e-01 -4.83697325e-01 2.32644871e-01 5.01623154e-01 2.58784592e-01 -1.33967948e+00 -1.82427442e+00 -2.15309896e-02 5.72541475e-01 -1.07495880e+00 2.97433048e-01 2.32154146e-01 -1.36381209e+00 -1.28777409e+00 -7.16492414e-01 -5.84046423e-01 6.34217620e-01 2.26356968e-01 1.27136803e+00 4.11123812e-01 -1.10239434e+00 7.46573746e-01 -2.00205788e-01 -7.45917618e-01 -3.84311318e-01 -5.72277009e-02 -6.89550102e-01 -8.01159501e-01 2.57602185e-01 -3.21958154e-01 -8.59173715e-01 -1.32427737e-01 -8.88307035e-01 4.79243279e-01 7.11565554e-01 4.82477188e-01 8.96254241e-01 -3.67473990e-01 -6.27069995e-02 -1.12557542e+00 1.41315788e-01 -3.29164892e-01 -8.61438751e-01 1.03570465e-02 -4.25521284e-01 -5.51089942e-01 1.69234082e-01 1.03724718e-01 -6.69139326e-01 5.76618493e-01 -3.28597337e-01 -5.91145515e-01 -8.15557003e-01 3.44116986e-01 3.53461802e-01 -3.03286254e-01 7.63679266e-01 -1.83689177e-01 4.33340043e-01 -2.84604549e-01 5.07497907e-01 4.03475910e-01 4.93132085e-01 3.21762115e-01 3.15982997e-01 7.59330392e-01 1.47717193e-01 -8.53116095e-01 -3.83394808e-01 -5.58303833e-01 -9.34676528e-01 -1.39419302e-01 1.23347414e+00 -9.76078808e-01 -7.22633004e-01 -2.33848885e-01 -1.00052750e+00 -4.68362421e-01 -6.63772583e-01 9.02472436e-01 -5.76009452e-01 4.69786860e-02 -6.88586354e-01 -5.94431460e-01 -7.86796570e-01 -1.61551070e+00 1.01858497e+00 6.13238156e-01 -1.55890211e-02 -1.19594109e+00 -1.28389999e-01 3.19549680e-01 5.23599684e-01 5.96584320e-01 7.73205280e-01 -8.74864042e-01 -9.04304326e-01 -2.13487744e-01 -3.20500821e-01 -4.45893332e-02 -2.84407854e-01 -9.42235291e-02 -1.04949832e+00 -1.67532042e-01 -6.39598593e-02 4.99963127e-02 4.96677309e-01 8.96117747e-01 1.14503324e+00 3.85990024e-01 -4.95583504e-01 8.76995146e-01 1.55367541e+00 7.95081407e-02 8.76683056e-01 4.71068501e-01 5.96349180e-01 4.97571886e-01 4.43411231e-01 2.90234778e-02 2.30484843e-01 2.76790053e-01 9.43901837e-01 -1.02306509e+00 -2.38534704e-01 5.28833210e-01 -8.30845460e-02 2.55914629e-01 1.13925999e-02 4.09652203e-01 -1.14217854e+00 1.56560004e-01 -1.34847844e+00 -2.93522120e-01 -6.53570294e-01 2.18503785e+00 1.82685778e-01 -1.62504777e-01 -1.30655933e-02 7.18830898e-02 2.53023267e-01 -4.89786357e-01 -1.28226638e-01 -7.34421194e-01 6.44767284e-01 6.86915040e-01 5.52871048e-01 3.81600827e-01 -1.04244888e+00 6.47015870e-01 6.17942905e+00 4.86188270e-02 -1.54450715e+00 3.99237037e-01 7.30003417e-01 9.05988812e-02 1.85055733e-01 -3.58406395e-01 -3.96891654e-01 -1.52080134e-01 1.09229946e+00 5.10719597e-01 4.57728989e-02 9.05541539e-01 3.51553053e-01 -4.98403639e-01 -1.17322052e+00 9.71001685e-01 -1.51667997e-01 -1.70594025e+00 -4.27619040e-01 1.27529487e-01 1.30879134e-01 7.17270076e-01 1.90382242e-01 7.60207772e-02 1.97782055e-01 -1.46554399e+00 2.07370400e-01 5.73047876e-01 9.11609471e-01 -6.81728780e-01 9.77040589e-01 3.06114197e-01 -7.97915995e-01 4.17518504e-02 -1.42052501e-01 4.09563690e-01 -5.53990193e-02 4.86220688e-01 -1.50468087e+00 7.10437417e-01 7.90098727e-01 2.96639174e-01 -6.63398921e-01 1.31946909e+00 1.11632915e-02 -6.24561543e-03 -8.94633159e-02 4.31568414e-01 2.86128253e-01 -8.14009234e-02 4.38927054e-01 1.46892858e+00 2.08362550e-01 1.86547518e-01 1.09563828e-01 7.78678298e-01 4.57120270e-01 2.03429952e-01 -6.84197068e-01 3.62045854e-01 -2.69445747e-01 1.98481846e+00 -1.47390187e+00 -3.21707100e-01 -2.79866159e-01 6.89375877e-01 -3.78663987e-01 -4.56927419e-02 -6.35132432e-01 -1.05621494e-01 1.16270103e-01 2.43144780e-01 1.68359563e-01 -1.55160815e-01 -7.31341243e-01 -6.24631584e-01 -6.13584459e-01 -6.37392640e-01 5.31393051e-01 -1.08068192e+00 -6.41086519e-01 9.78133619e-01 -2.14235142e-01 -1.09754646e+00 -1.64145961e-01 -8.90692234e-01 -5.94173729e-01 1.09035444e+00 -1.72703803e+00 -1.37174046e+00 -8.60402107e-01 7.08836794e-01 5.45005560e-01 -1.51017427e-01 1.08510530e+00 1.74463421e-01 -3.32791060e-01 -2.35600416e-02 -3.98908347e-01 1.94281474e-01 2.04871491e-01 -1.59302771e+00 4.94847484e-02 5.13649225e-01 2.00312778e-01 5.56456447e-01 5.82008541e-01 -3.35793376e-01 -1.47793889e+00 -8.45580101e-01 3.30016375e-01 -6.05095506e-01 1.80030867e-01 4.95639667e-02 -6.29271090e-01 8.75781238e-01 4.56672728e-01 5.60331047e-01 1.02281559e+00 -2.10821450e-01 9.94500518e-02 4.46028948e-01 -1.65343428e+00 1.56147704e-01 2.99218565e-01 -7.15749105e-03 -4.43502516e-01 6.07706010e-01 7.49095500e-01 -1.35566628e+00 -9.69207644e-01 6.78036511e-02 4.00916338e-01 -1.15928388e+00 1.19862723e+00 -5.16206622e-01 4.33084488e-01 -4.59447652e-01 1.59952521e-01 -1.08171761e+00 1.99947387e-01 -2.39156917e-01 1.15908459e-01 1.89273641e-01 2.73899972e-01 -6.02194667e-01 8.69286299e-01 5.49245417e-01 -5.96587420e-01 -8.69921446e-01 -9.32357490e-01 2.63191704e-02 -1.97711632e-01 -5.42528272e-01 3.56073290e-01 7.59710014e-01 -3.64319324e-01 1.54768288e-01 6.03890061e-01 4.83183891e-01 2.90032119e-01 -6.15609884e-02 8.25502336e-01 -1.27357483e+00 -1.43872634e-01 -4.66018885e-01 -6.90977573e-01 -3.50355774e-01 -5.40768981e-01 -9.59340632e-01 -3.65695655e-01 -2.03709102e+00 -1.48794532e-01 -6.04425132e-01 -1.72487289e-01 3.86907995e-01 5.19490600e-01 4.65311736e-01 1.65961087e-01 1.23783112e-01 -3.31454068e-01 -3.09257805e-01 1.44890654e+00 2.38625929e-01 -4.50326324e-01 2.29552239e-01 -4.53695297e-01 8.71044576e-01 7.41686404e-01 -3.42365146e-01 -2.40144581e-01 -2.93214023e-01 1.45413503e-01 4.84291553e-01 7.24830985e-01 -1.08925521e+00 4.08443421e-01 5.58914483e-01 9.25553203e-01 -1.03376937e+00 1.73136070e-01 -1.24966776e+00 4.08138245e-01 1.00688863e+00 -1.42230792e-02 3.06502402e-01 4.14767265e-01 2.70593703e-01 1.78135157e-01 -1.57610387e-01 1.05739248e+00 -7.50335038e-01 -4.56801653e-01 2.08310619e-01 -2.14228049e-01 -7.89179146e-01 1.26459873e+00 -5.68328977e-01 -2.29632445e-02 1.20043710e-01 -1.44926691e+00 3.78912501e-02 2.53460586e-01 1.79323852e-02 8.22762430e-01 -6.26909733e-01 -6.22276962e-01 4.87295955e-01 -2.50988305e-01 6.04196131e-01 4.91705865e-01 1.39924443e+00 -1.25681174e+00 6.40014231e-01 -4.51002985e-01 -1.05983853e+00 -1.51737201e+00 5.04291475e-01 9.20807421e-01 -4.67562824e-01 -1.38174236e+00 7.28193939e-01 -1.83761045e-02 -5.05652785e-01 1.70526415e-01 -6.71878934e-01 -8.11367214e-01 -9.62975472e-02 7.91652501e-01 1.81353420e-01 6.64955497e-01 -3.92307192e-01 -1.76517099e-01 2.47360259e-01 -1.28341749e-01 -4.46623974e-02 1.50418079e+00 -4.93120775e-03 -8.86833295e-02 2.98100740e-01 8.29233944e-01 -2.17106462e-01 -7.82573044e-01 4.87235524e-02 -3.81197855e-02 -3.56680840e-01 5.81903815e-01 -1.19985378e+00 -1.64281559e+00 1.06483626e+00 1.37065053e+00 2.40266815e-01 9.83045399e-01 -3.90650928e-02 3.90125304e-01 3.20028625e-02 2.24675760e-01 -4.09973979e-01 -5.89567572e-02 1.13340586e-01 7.86543548e-01 -1.12962377e+00 2.37100095e-01 -4.52122033e-01 -6.08932793e-01 1.70412290e+00 4.34882611e-01 -2.64133573e-01 7.61759222e-01 6.12712860e-01 6.04564130e-01 -9.18709815e-01 -3.85058850e-01 -1.51374951e-01 8.38192552e-02 8.97168636e-01 7.62164891e-01 1.74517751e-01 3.14929076e-02 2.63782412e-01 -2.34133869e-01 3.85128289e-01 8.28190148e-01 9.11138237e-01 -2.70514011e-01 -6.03869557e-01 -4.94836241e-01 2.99683183e-01 -6.64039910e-01 3.59343588e-02 7.55603686e-02 1.25445271e+00 1.64674446e-01 4.67163384e-01 1.63609087e-01 1.18165724e-01 2.89399236e-01 -2.00408667e-01 6.23143494e-01 -8.75053585e-01 -1.49436426e+00 1.57031551e-01 -2.90850937e-01 -7.94394612e-01 -1.80991814e-01 -3.73181969e-01 -1.78790891e+00 8.44237115e-03 -2.11431012e-01 -2.21417382e-01 1.52698910e+00 6.24274969e-01 2.74396926e-01 1.24301636e+00 2.03058779e-01 -9.86428380e-01 -1.14220090e-01 -7.75626540e-01 -4.41270381e-01 -3.38568985e-01 3.27977479e-01 -2.18707189e-01 9.93292034e-02 -1.02262028e-01]

[14.529319763183594, -2.5239267349243164]

6ba1eb79-2ff9-408d-8239-acb01ab44a62

sandwiched-video-compression-efficiently

2303.11473

https://arxiv.org/abs/2303.11473v2

https://arxiv.org/pdf/2303.11473v2.pdf

Sandwiched Video Compression: Efficiently Extending the Reach of Standard Codecs with Neural Wrappers

We propose sandwiched video compression -- a video compression system that wraps neural networks around a standard video codec. The sandwich framework consists of a neural pre- and post-processor with a standard video codec between them. The networks are trained jointly to optimize a rate-distortion loss function with the goal of significantly improving over the standard codec in various compression scenarios. End-to-end training in this setting requires a differentiable proxy for the standard video codec, which incorporates temporal processing with motion compensation, inter/intra mode decisions, and in-loop filtering. We propose differentiable approximations to key video codec components and demonstrate that, in addition to providing meaningful compression improvements over the standard codec, the neural codes of the sandwich lead to significantly better rate-distortion performance in two important scenarios.When transporting high-resolution video via low-resolution HEVC, the sandwich system obtains 6.5 dB improvements over standard HEVC. More importantly, using the well-known perceptual similarity metric, LPIPS, we observe 30% improvements in rate at the same quality over HEVC. Last but not least, we show that pre- and post-processors formed by very modestly-parameterized, light-weight networks can closely approximate these results.

['Philip A. Chou', 'Jonathan Taylor', 'Danhang Tang', 'Onur G. Guleryuz', 'Berivan Isik']

2023-03-20

['motion-compensation']

['computer-vision']

[ 5.49217522e-01 -1.22164607e-01 -2.94380128e-01 -2.97227263e-01 -8.63737822e-01 -2.86394149e-01 3.55295956e-01 -1.03357866e-01 -4.22133356e-01 1.90680146e-01 4.83634740e-01 -4.09118980e-01 -2.37629302e-02 -4.53982353e-01 -1.17128897e+00 -5.40793180e-01 -6.36250138e-01 -6.69244155e-02 -6.45227358e-02 9.51204728e-03 1.93494469e-01 2.71521956e-01 -1.32522142e+00 5.94336450e-01 5.14830351e-01 1.55175602e+00 3.23132902e-01 1.28404415e+00 5.24961293e-01 1.06613123e+00 -2.17069089e-01 -6.66450202e-01 6.37491345e-01 -3.11241359e-01 -5.37816048e-01 3.05444062e-01 6.90616786e-01 -1.03152955e+00 -9.42146599e-01 1.05674171e+00 1.99172065e-01 9.87281352e-02 4.75977898e-01 -7.76956081e-01 -7.03140259e-01 6.22771204e-01 -5.95443785e-01 6.72267750e-02 7.28303492e-02 1.19784601e-01 1.27039969e+00 -8.76058936e-01 3.69848758e-01 1.16116309e+00 8.07943702e-01 3.77032012e-01 -1.24543166e+00 -4.24952686e-01 -2.46268004e-01 2.66036570e-01 -1.30149734e+00 -8.83000553e-01 3.85742515e-01 -2.08822832e-01 1.08403492e+00 1.04153445e-02 4.46132034e-01 6.76343858e-01 3.44127715e-01 4.19595182e-01 4.62866649e-02 -1.66721180e-01 1.51005685e-01 -4.27210271e-01 -3.39062214e-01 6.95549786e-01 2.15835094e-01 3.14918339e-01 -3.75386238e-01 2.02608034e-01 1.03830063e+00 1.00227371e-01 -7.95505822e-01 -3.75090182e-01 -9.83277500e-01 4.36796457e-01 4.92113233e-01 -1.98582057e-02 -3.96163046e-01 7.71023810e-01 6.21604681e-01 4.52259302e-01 1.60228580e-01 8.28815475e-02 -3.09559405e-01 -2.95359194e-01 -1.52591038e+00 1.39758602e-01 6.05270803e-01 1.14168990e+00 3.81204188e-01 3.42355371e-01 -3.90913151e-02 8.17320704e-01 2.74668396e-01 3.58813375e-01 3.68662745e-01 -1.98457634e+00 1.02250326e+00 -2.35844970e-01 -6.75957650e-02 -9.82875705e-01 1.47785902e-01 -6.76086903e-01 -1.25387132e+00 3.36516589e-01 -3.01294848e-02 -1.00769371e-01 -6.57774508e-01 1.65176678e+00 -4.50701207e-01 3.86364073e-01 1.28235117e-01 9.72878695e-01 1.59317091e-01 9.28167522e-01 -2.81077117e-01 -3.60337734e-01 1.10560143e+00 -1.13261592e+00 -6.13979757e-01 -2.67955922e-02 4.45450097e-01 -6.64411247e-01 7.00996041e-01 4.12248373e-01 -2.02578950e+00 -8.56225550e-01 -1.54885268e+00 -2.81538725e-01 4.08022374e-01 1.96947549e-02 -7.40404427e-02 3.08918774e-01 -1.46140528e+00 1.18835914e+00 -7.94651866e-01 1.16135411e-01 4.21736091e-01 3.56679082e-01 -2.05940455e-01 -2.82138884e-01 -7.71067679e-01 3.14347953e-01 3.73696864e-01 -2.50840276e-01 -1.16826940e+00 -9.90750909e-01 -8.50359023e-01 6.12686992e-01 2.46533155e-01 -7.31064379e-01 1.48586452e+00 -1.22035968e+00 -1.41065574e+00 5.43588042e-01 -2.61187911e-01 -1.08911061e+00 4.30191815e-01 -5.03067136e-01 -4.82219845e-01 7.12832272e-01 -2.54988134e-01 7.60922551e-01 1.20077181e+00 -1.09065437e+00 -9.27840471e-01 1.33459046e-01 4.83611114e-02 2.33294025e-01 -3.02960426e-01 3.97576429e-02 -1.02587187e+00 -1.01825309e+00 -1.08972311e-01 -5.48535824e-01 4.57356498e-02 4.87770975e-01 3.08886971e-02 4.24896240e-01 1.02520335e+00 -1.06183219e+00 1.23462856e+00 -2.47981334e+00 2.00784713e-01 7.13727400e-02 5.96011460e-01 3.94701093e-01 -4.26337093e-01 2.26677079e-02 -2.10433260e-01 1.82243615e-01 -3.92895669e-01 -4.98684675e-01 -1.63301781e-01 7.50396177e-02 -4.05924141e-01 4.57486421e-01 2.09863156e-01 7.26459742e-01 -5.45867503e-01 -2.16686592e-01 -1.24059636e-02 7.79969335e-01 -1.08950317e+00 4.64990258e-01 -1.13504857e-01 -2.10831493e-01 2.91448116e-01 3.66776347e-01 5.99739850e-01 -4.07671362e-01 2.35076517e-01 -4.90000069e-01 1.33063272e-01 3.11311573e-01 -7.97508001e-01 1.77375031e+00 -3.33100617e-01 1.13635838e+00 6.10894322e-01 -1.08744633e+00 3.92329514e-01 5.63179612e-01 4.41025823e-01 -7.05877304e-01 2.25200042e-01 2.96495110e-01 -6.58421144e-02 -3.05941492e-01 7.70203829e-01 1.90355942e-01 5.44475436e-01 1.79067507e-01 1.69886038e-01 4.58223112e-02 3.49381447e-01 2.94133604e-01 1.18175793e+00 -3.05769080e-03 1.32051215e-01 -4.14833799e-02 4.42964524e-01 -7.59800792e-01 5.35648048e-01 6.20383024e-01 -2.58137822e-01 9.54403579e-01 4.12031740e-01 -2.14014828e-01 -1.68324471e+00 -1.32092392e+00 9.67366695e-02 1.11231220e+00 2.29354389e-02 -6.04313552e-01 -7.99492121e-01 -1.63903400e-01 -2.50047773e-01 4.51385438e-01 -1.47386745e-01 -1.92792535e-01 -8.70335639e-01 5.76704666e-02 7.76748121e-01 5.87415636e-01 8.66666913e-01 -4.35740232e-01 -6.47250354e-01 3.87940884e-01 -1.63165331e-01 -1.40684390e+00 -9.85031784e-01 2.22422779e-01 -1.20288968e+00 -8.90181184e-01 -1.09081674e+00 -9.71661091e-01 2.65411168e-01 4.87906188e-01 1.34109485e+00 2.57272422e-01 1.94531649e-01 1.83741450e-01 -2.41245672e-01 5.26625514e-01 -6.51986778e-01 -2.28237435e-01 -1.27514735e-01 -5.86628132e-02 -2.94689506e-01 -9.18567002e-01 -8.27570856e-01 1.02069259e-01 -9.76998091e-01 2.06317589e-01 5.20873547e-01 5.81243992e-01 5.59439898e-01 1.49102643e-01 -1.13597058e-01 -1.72398597e-01 3.39814752e-01 -3.59031230e-01 -5.57849884e-01 1.49907768e-01 -6.55441880e-01 1.18716680e-01 9.50768471e-01 -1.00746967e-01 -7.56585300e-01 -3.13807905e-01 -2.64117271e-01 -9.82426643e-01 1.16824441e-01 3.01948279e-01 -5.54520935e-02 -7.83938989e-02 4.81587201e-01 2.33325869e-01 -9.75901708e-02 -3.08073074e-01 3.51699054e-01 5.57482958e-01 1.25997519e+00 -5.95317006e-01 8.93373609e-01 2.82436520e-01 3.70885655e-02 -6.22395813e-01 -2.83861935e-01 -1.37057021e-01 -2.06244543e-01 -2.68696598e-03 9.01508093e-01 -1.32457471e+00 -5.44900835e-01 1.90494671e-01 -1.23071563e+00 -5.03839672e-01 -3.14867079e-01 5.29144645e-01 -9.28463042e-01 6.48023427e-01 -1.17867017e+00 -3.47663194e-01 -3.93932909e-01 -1.20755816e+00 8.20415854e-01 -6.61615506e-02 2.12515727e-01 -8.20554316e-01 -3.07930857e-01 5.46506280e-03 6.48669422e-01 -1.56508744e-01 1.06719553e+00 -1.47515628e-02 -7.86327124e-01 1.91038042e-01 -6.74760401e-01 9.51190829e-01 -1.05421618e-01 1.12421788e-01 -7.56556451e-01 -6.13365352e-01 1.30892411e-01 -2.40854517e-01 1.11814988e+00 6.94438100e-01 1.45918989e+00 -6.07175827e-01 1.92185730e-01 1.49302530e+00 1.69681418e+00 5.54889143e-01 7.65137374e-01 -4.23144586e-02 5.28273523e-01 1.82933226e-01 -2.73404308e-02 5.29212475e-01 1.55612901e-01 5.82551718e-01 4.78322864e-01 7.77982622e-02 -3.82114768e-01 -2.01635003e-01 6.32659793e-01 9.78104293e-01 -4.68315035e-01 -3.02259564e-01 -5.39861619e-01 2.31592998e-01 -1.50528395e+00 -1.26407552e+00 3.86146456e-01 2.24605441e+00 8.52585971e-01 2.80102938e-01 -7.01907724e-02 3.35037291e-01 6.56443238e-01 4.16465223e-01 -5.22004962e-01 -4.48256254e-01 -3.00272584e-01 2.24923790e-01 8.24869692e-01 7.13979900e-01 -1.17604828e+00 3.56480539e-01 6.53877211e+00 7.50630558e-01 -1.12173927e+00 1.97907407e-02 9.24546719e-01 -3.53181988e-01 -8.93124193e-02 -1.86330676e-01 -2.60533780e-01 3.33089709e-01 1.50083089e+00 -2.82663584e-01 9.96504843e-01 6.96936905e-01 4.60474163e-01 3.36950481e-01 -1.48851573e+00 1.31922424e+00 6.03043288e-02 -1.80255854e+00 1.81960881e-01 2.28172958e-01 7.89328098e-01 1.19725987e-01 2.50628293e-01 1.98440269e-01 1.35356963e-01 -9.68502641e-01 1.07373965e+00 1.42748863e-01 1.36133218e+00 -6.79698229e-01 5.10290504e-01 -1.87129036e-01 -1.48745286e+00 -3.29937905e-01 -4.23395514e-01 1.04576595e-01 4.06989843e-01 3.46478194e-01 -3.65433954e-02 3.04082155e-01 7.96988249e-01 9.80820715e-01 -1.95218146e-01 1.14194775e+00 1.21517263e-01 6.98666036e-01 9.39349160e-02 8.98531437e-01 2.64682144e-01 -9.40315947e-02 3.31305772e-01 1.45147014e+00 7.32514858e-01 5.86014166e-02 -2.59054720e-01 6.53076768e-01 -6.54440224e-01 -3.59481454e-01 -3.80509496e-01 -5.23094051e-02 3.67951691e-01 5.66979706e-01 -6.26680031e-02 -5.37934005e-01 -5.14044046e-01 1.28485692e+00 4.02883552e-02 7.91603804e-01 -1.03069377e+00 -4.54622418e-01 1.00005972e+00 1.88052520e-01 7.49696195e-01 -3.03749502e-01 -1.98277816e-01 -1.11108434e+00 4.22779061e-02 -1.16802621e+00 -8.94864351e-02 -7.91230977e-01 -7.75163472e-01 3.98228645e-01 -2.73919016e-01 -1.27968121e+00 -3.27310443e-01 -5.98034859e-01 -5.00175714e-01 6.56221032e-01 -1.43919420e+00 -4.29831743e-01 -1.51834652e-01 6.34172857e-01 8.14362049e-01 -3.56732517e-01 3.83959353e-01 6.78665161e-01 -3.76903743e-01 9.11812305e-01 5.37146807e-01 2.98088402e-01 6.04716301e-01 -8.83211553e-01 4.30147767e-01 1.20206356e+00 -1.07588306e-01 5.28674603e-01 7.22048402e-01 -2.93943584e-01 -1.50265455e+00 -1.31367218e+00 5.56023896e-01 4.68705475e-01 4.65041757e-01 -9.22281444e-02 -9.90903080e-01 7.07050979e-01 5.41911721e-01 1.33973673e-01 3.21282178e-01 -4.45392728e-01 -8.28681648e-01 -4.36026186e-01 -9.96454775e-01 5.78821659e-01 1.05759335e+00 -7.86053658e-01 -3.64483625e-01 -1.18677514e-02 1.17884028e+00 -3.84698570e-01 -9.00370896e-01 3.22053820e-01 6.76264942e-01 -1.36939204e+00 1.16044164e+00 -3.24266851e-01 1.37534308e+00 -1.14459708e-01 -9.42614079e-01 -1.00537455e+00 -5.52414834e-01 -8.18170488e-01 -7.15893805e-01 7.99143195e-01 1.53577462e-01 -7.48635456e-02 8.01260054e-01 4.43985403e-01 -4.15749490e-01 -6.79759383e-01 -8.33714068e-01 -9.68187392e-01 -2.12592091e-02 -7.01157749e-01 2.46871293e-01 3.32863122e-01 -3.01078230e-01 6.75378144e-02 -6.15870059e-01 1.99583977e-01 8.78889620e-01 -4.32441711e-01 3.53300691e-01 -6.56001866e-01 -8.45976770e-01 -7.80287921e-01 -1.99795604e-01 -1.84612703e+00 -1.46872416e-01 -6.49108648e-01 1.25239640e-01 -1.14828777e+00 1.28727242e-01 -4.30550463e-02 -2.50988334e-01 -1.21903911e-01 2.25933775e-01 2.34686151e-01 5.23623288e-01 3.57779384e-01 -5.64425647e-01 6.09353781e-01 9.93090868e-01 -2.92176247e-01 1.08600348e-01 -3.31381828e-01 -5.33720076e-01 6.22292221e-01 5.90667427e-01 -1.10651605e-01 -5.06947339e-01 -1.01928055e+00 -3.09434179e-02 5.04316449e-01 3.29573363e-01 -1.52130473e+00 2.42785692e-01 8.93173739e-02 2.63977170e-01 -2.75467753e-01 4.00817156e-01 -9.03252304e-01 -2.69773100e-02 5.21202326e-01 -7.29496300e-01 3.67799222e-01 -9.67217982e-02 5.99404275e-01 -3.81412059e-01 -1.04503967e-01 1.07740414e+00 1.29507542e-01 -6.58955038e-01 4.51876581e-01 -4.38791275e-01 1.71807900e-01 5.62882245e-01 -3.01551104e-01 -2.60508686e-01 -8.54742527e-01 -4.32013839e-01 -1.71617568e-01 5.48716187e-01 1.93116203e-01 9.54638720e-01 -1.34662449e+00 -7.22888827e-01 3.93235385e-01 -3.58841866e-01 -2.24490657e-01 2.04883367e-01 6.22229993e-01 -8.91719460e-01 5.23190081e-01 -1.58818066e-01 -4.69396532e-01 -1.22122455e+00 5.50145328e-01 4.88642275e-01 -1.86966881e-01 -5.77778816e-01 7.20114946e-01 3.16938668e-01 4.22339827e-01 7.25854099e-01 -4.84402567e-01 3.40447277e-01 -5.97921193e-01 7.29032815e-01 4.12225604e-01 -1.37680680e-01 -5.47565103e-01 1.23535953e-01 5.67737937e-01 9.78106782e-02 -2.18176961e-01 1.19730079e+00 -4.07679290e-01 1.31945193e-01 -7.63901100e-02 1.87589216e+00 -1.69507861e-01 -1.92092597e+00 -3.04907769e-01 -3.69405210e-01 -4.49190974e-01 2.08030999e-01 -3.79591256e-01 -1.54082334e+00 1.00786769e+00 7.53322721e-01 -8.33588988e-02 1.53314936e+00 -4.57306087e-01 1.03787386e+00 5.01274884e-01 1.39176890e-01 -1.04385066e+00 7.28224292e-02 6.05815947e-01 9.24164355e-01 -9.16127563e-01 -9.33059603e-02 -6.24420308e-03 -2.29497418e-01 1.34965837e+00 1.77012965e-01 -3.47716331e-01 6.54255748e-01 4.39284384e-01 -2.99648345e-01 3.47317994e-01 -1.07954776e+00 3.76108825e-01 2.20688805e-01 4.89568710e-01 5.49929559e-01 -1.52814180e-01 8.70452374e-02 1.84011802e-01 -5.78594804e-02 1.15027368e-01 5.52210033e-01 5.40393353e-01 -6.53462768e-01 -7.33922541e-01 -2.23174259e-01 4.25195813e-01 -6.46489263e-01 -3.91222090e-01 4.78781164e-01 4.57232445e-01 1.15682662e-01 1.01018989e+00 5.10448158e-01 -7.52397120e-01 3.95991728e-02 -3.45788747e-01 2.77706891e-01 3.82527225e-02 -3.42187315e-01 2.89418638e-01 -1.89462826e-02 -1.05005407e+00 -2.54993349e-01 -2.07062885e-01 -1.10828507e+00 -9.04599369e-01 3.43827456e-01 -3.28633972e-02 3.73922169e-01 7.23872125e-01 4.71870750e-01 5.42472124e-01 6.46756768e-01 -1.14400303e+00 -5.75729311e-01 -4.85763490e-01 -4.13960457e-01 5.18796980e-01 8.33544731e-01 7.49827623e-02 -3.29708070e-01 6.00928009e-01]

[11.351752281188965, -1.5925259590148926]

b7fc36bc-d03c-4aab-9ed6-ab031ea21226

scene-coordinate-and-correspondence-learning

1805.08443

http://arxiv.org/abs/1805.08443v4

http://arxiv.org/pdf/1805.08443v4.pdf

Scene Coordinate and Correspondence Learning for Image-Based Localization

Scene coordinate regression has become an essential part of current camera re-localization methods. Different versions, such as regression forests and deep learning methods, have been successfully applied to estimate the corresponding camera pose given a single input image. In this work, we propose to regress the scene coordinates pixel-wise for a given RGB image by using deep learning. Compared to the recent methods, which usually employ RANSAC to obtain a robust pose estimate from the established point correspondences, we propose to regress confidences of these correspondences, which allows us to immediately discard erroneous predictions and improve the initial pose estimates. Finally, the resulting confidences can be used to score initial pose hypothesis and aid in pose refinement, offering a generalized solution to solve this task.

['Nassir Navab', 'Mai Bui', 'Slobodan Ilic', 'Shadi Albarqouni']

2018-05-22

['image-based-localization']

['computer-vision']

[ 1.90989912e-01 -4.70821559e-02 -4.47183885e-02 -4.39853787e-01 -7.05359817e-01 -4.61779237e-01 5.01367211e-01 2.24637240e-01 -7.37730086e-01 7.19072521e-01 -1.66708931e-01 8.82872716e-02 -1.24596683e-02 -6.13917887e-01 -8.13827932e-01 -5.43746173e-01 5.01484692e-01 6.61853790e-01 3.38536501e-01 3.06742731e-02 6.10203624e-01 8.31666172e-01 -1.48980045e+00 -3.12521219e-01 8.24425876e-01 9.78914320e-01 1.90741301e-01 4.85375494e-01 6.39460981e-02 3.42819422e-01 -2.90791184e-01 -3.72821987e-01 9.79101062e-02 -1.34483844e-01 -5.15166402e-01 2.21029818e-01 6.57145798e-01 -4.37716007e-01 3.27005517e-03 1.14129019e+00 1.75586462e-01 1.08924150e-01 4.52993691e-01 -9.83532548e-01 2.15708598e-01 7.51251280e-02 -7.22039700e-01 -3.17068338e-01 8.03928494e-01 -2.66374230e-01 8.46023023e-01 -9.82037485e-01 5.12709022e-01 8.58273029e-01 1.04742765e+00 1.14862196e-01 -1.29771650e+00 -6.79062068e-01 2.30456963e-01 4.52295274e-01 -1.72820127e+00 -3.26782197e-01 9.74232554e-01 -3.30690295e-01 6.12365544e-01 1.32722095e-01 8.51207972e-01 1.02004087e+00 6.29432872e-02 6.71769738e-01 1.15946245e+00 -5.81410825e-01 1.73810348e-01 -2.17296761e-02 -2.74536222e-01 8.32916439e-01 3.20285141e-01 -8.02607015e-02 -6.17747426e-01 -3.34436148e-02 9.56654251e-01 2.27590516e-01 -2.49331668e-01 -9.10710990e-01 -1.29340124e+00 7.93433547e-01 8.79291654e-01 -1.41532883e-01 -5.74734747e-01 2.16422215e-01 -1.09915286e-01 -3.80921304e-01 4.31526184e-01 5.65591335e-01 -3.85233372e-01 -2.82503478e-02 -1.10091412e+00 2.13745713e-01 6.26328766e-01 7.08547592e-01 1.15389764e+00 -3.34738791e-01 1.77691221e-01 5.90828836e-01 5.04063070e-01 4.42661822e-01 2.21947238e-01 -1.08538640e+00 2.36059949e-01 6.53594911e-01 2.24509358e-01 -1.28323174e+00 -5.71204662e-01 -4.00901854e-01 -8.06809008e-01 1.89958230e-01 3.15777421e-01 9.81479660e-02 -1.09598267e+00 1.23280454e+00 4.51857060e-01 4.46465433e-01 -2.77384818e-01 9.44915950e-01 2.67528623e-01 3.05563569e-01 -2.21417278e-01 -5.47831282e-02 8.21267664e-01 -8.13487232e-01 -3.59192938e-01 -3.91091585e-01 2.37551868e-01 -9.60608661e-01 6.36775196e-01 7.30652988e-01 -5.95956326e-01 -5.91967106e-01 -1.03157246e+00 -1.14011606e-02 -2.44336128e-01 3.68068248e-01 4.12936866e-01 4.11500335e-01 -1.01542151e+00 7.96238601e-01 -1.02587271e+00 -4.88523096e-01 2.24908978e-01 6.15048051e-01 -5.34316659e-01 -1.09437875e-01 -7.03352034e-01 1.23922074e+00 5.40832877e-01 4.84794050e-01 -6.50335729e-01 -1.91689387e-01 -8.61731648e-01 -3.06643635e-01 5.88418841e-01 -8.17815959e-01 1.13154244e+00 -6.21438503e-01 -1.76690531e+00 6.54092491e-01 -3.07950169e-01 -2.81374156e-01 7.29281247e-01 -7.12804019e-01 2.80873239e-01 -2.06046235e-02 1.81620810e-02 8.45032573e-01 1.01635647e+00 -1.56242478e+00 -8.29025984e-01 -4.37553674e-01 -6.24046959e-02 2.47609407e-01 6.31047487e-02 -3.53375554e-01 -7.62673497e-01 -2.66147792e-01 1.03057444e+00 -1.17416620e+00 -5.87459564e-01 1.44355565e-01 -5.71473658e-01 -4.56337370e-02 5.13281703e-01 -7.14480579e-01 7.48556614e-01 -1.60498559e+00 6.47545576e-01 5.32460451e-01 1.62001446e-01 -2.48326194e-02 1.63598970e-01 3.86122242e-02 1.23282686e-01 -2.86193788e-01 -1.97708383e-01 -7.40040600e-01 -3.73827994e-01 2.09407762e-01 -4.07852866e-02 8.01702976e-01 1.76982746e-01 4.79787618e-01 -7.88579643e-01 -4.86582547e-01 7.21596658e-01 6.14454567e-01 -5.98832130e-01 2.60480702e-01 -2.35529602e-01 9.17671442e-01 -3.16293895e-01 3.96474421e-01 6.71507657e-01 6.99803084e-02 1.62999686e-02 -4.39715028e-01 -2.38630340e-01 2.96537429e-02 -1.30361390e+00 2.00101519e+00 -4.85645831e-01 4.80245173e-01 -1.85414165e-01 -7.95803308e-01 1.09522736e+00 -1.00457780e-01 6.46519184e-01 -2.27537349e-01 2.96660990e-01 1.44555256e-01 -4.34645891e-01 -9.54220593e-02 7.13095546e-01 -7.42907003e-02 6.86022341e-02 -1.11109443e-01 1.31405130e-01 -4.60918665e-01 -1.95370317e-01 -2.69031614e-01 8.10430825e-01 9.19442534e-01 6.21208429e-01 4.59157199e-01 8.85322928e-01 2.08574340e-01 4.72686827e-01 4.57379669e-01 6.30857423e-02 1.15397739e+00 1.41424779e-02 -6.05992317e-01 -1.13865256e+00 -1.02123940e+00 -7.13017881e-02 5.74269831e-01 5.05042017e-01 -4.19837683e-01 -6.28518999e-01 -7.10914314e-01 -9.61426198e-02 5.98480999e-01 -1.77733377e-01 -1.01651631e-01 -6.36108816e-01 -5.03586769e-01 1.04811974e-02 5.05413115e-01 5.95152438e-01 -5.26012421e-01 -6.61708474e-01 1.49066895e-01 -4.42848563e-01 -1.17878544e+00 -1.92768890e-02 4.04279709e-01 -9.93104458e-01 -1.20171785e+00 -6.18181944e-01 -3.63367468e-01 9.47436392e-01 2.15496808e-01 7.53109276e-01 2.52354946e-02 -1.07546993e-01 3.12031537e-01 -3.75933290e-01 -1.50382847e-01 -1.02834463e-01 1.96870446e-01 2.68935502e-01 8.60427618e-02 4.45087627e-02 -4.77935016e-01 -5.28079689e-01 3.59522492e-01 -4.41908598e-01 2.46299639e-01 8.50730002e-01 8.25153470e-01 1.05998385e+00 -1.99507311e-01 -7.15510026e-02 -8.51938605e-01 6.28658980e-02 -5.81666939e-02 -8.72282505e-01 2.52246618e-01 -6.56860948e-01 2.71179467e-01 5.39307594e-01 -1.45438701e-01 -9.18742895e-01 1.03164589e+00 -4.08058673e-01 -7.17214048e-01 -4.51100886e-01 4.96864587e-01 2.34380253e-02 -4.26663756e-01 4.14849669e-01 1.61421642e-01 -1.26060382e-01 -5.96372724e-01 4.69395906e-01 3.73721212e-01 7.71096766e-01 -3.63904327e-01 1.04224026e+00 4.27645326e-01 2.07541570e-01 -6.01237416e-01 -9.31128979e-01 -7.45237887e-01 -1.57015800e+00 -4.32687491e-01 9.69748139e-01 -1.00528681e+00 -6.89728200e-01 3.07551384e-01 -1.51292658e+00 2.43163392e-01 9.65114608e-02 7.01267183e-01 -7.13599563e-01 6.69312239e-01 -1.47892581e-02 -8.18202555e-01 3.95763554e-02 -1.22152674e+00 1.44826436e+00 3.30094069e-01 -3.25526297e-01 -6.46482944e-01 3.08857381e-01 2.51758337e-01 -1.34546816e-01 3.15520763e-01 3.58772755e-01 -2.64143795e-01 -7.79089272e-01 -5.62105298e-01 -9.01993737e-02 1.28467575e-01 -4.66700345e-02 1.04904547e-01 -1.00458014e+00 -3.48680288e-01 -6.04957268e-02 -7.44860212e-04 7.27540910e-01 3.10953915e-01 1.11424410e+00 6.07121401e-02 -5.95270038e-01 1.03357494e+00 1.41569340e+00 -1.75100297e-01 5.95251262e-01 7.78384984e-01 8.18824947e-01 3.86078775e-01 9.84545648e-01 5.38145125e-01 3.23204726e-01 8.97801638e-01 8.05615544e-01 -5.41700870e-02 9.79985371e-02 -6.41878843e-01 4.42371778e-02 7.29975164e-01 -3.87508303e-01 1.36367843e-01 -9.79761124e-01 8.18096567e-03 -2.08733177e+00 -4.17961031e-01 -1.11588314e-01 2.35957479e+00 4.79783982e-01 1.64095223e-01 -5.13023846e-02 1.85031060e-03 6.77784979e-01 -3.67286690e-02 -5.79546988e-01 2.50497013e-01 2.16113731e-01 8.53033811e-02 8.53264034e-01 4.62789267e-01 -1.22422945e+00 1.19744337e+00 6.61878252e+00 3.59268188e-01 -1.10749912e+00 -1.66015700e-01 1.76832080e-01 1.70584947e-01 2.00145319e-02 2.88396358e-01 -8.91152084e-01 1.21439904e-01 5.06139100e-01 2.43837163e-01 3.75038445e-01 1.08618581e+00 -1.68719366e-02 -6.53178930e-01 -1.16943705e+00 1.25315654e+00 3.08645964e-01 -1.01854408e+00 -4.83002998e-02 4.00869548e-03 6.10852063e-01 -9.71849635e-02 -1.84093431e-01 3.55314128e-02 2.43472248e-01 -6.98191106e-01 7.87363350e-01 7.49478579e-01 6.50539696e-01 -8.12112093e-01 8.59346271e-01 5.64652622e-01 -9.33173120e-01 -2.79051978e-02 -5.64109981e-01 -1.81339294e-01 2.10108414e-01 4.23848480e-01 -1.35568988e+00 7.74102569e-01 7.80849040e-01 8.66140783e-01 -9.47381854e-01 1.46728742e+00 -5.78758597e-01 4.82843183e-02 -5.06233454e-01 5.37787825e-02 -6.25385270e-02 -4.25584286e-01 4.79707003e-01 6.60347998e-01 5.40644646e-01 -2.39569947e-01 3.67745429e-01 8.40215147e-01 2.61693418e-01 -4.19730879e-02 -5.15762031e-01 5.10840952e-01 2.65284389e-01 1.37224758e+00 -1.07747626e+00 -1.40082007e-02 -1.66954428e-01 1.33648050e+00 5.95880449e-01 1.14630200e-01 -8.40763330e-01 -1.01363286e-02 5.47121108e-01 9.26555991e-02 2.37895995e-01 -7.15210259e-01 -3.34207058e-01 -1.21242034e+00 -1.10213764e-01 -5.27035475e-01 -7.51280878e-03 -1.03971851e+00 -8.21699381e-01 5.72242022e-01 1.87882055e-02 -1.50976348e+00 -3.92150789e-01 -5.72411895e-01 -3.02082598e-01 6.68876410e-01 -1.56028628e+00 -1.21315122e+00 -6.46863103e-01 4.48212504e-01 5.37721694e-01 6.80106580e-02 7.59614587e-01 -2.79318169e-02 -4.92912978e-01 1.28847584e-01 5.98117821e-02 1.90081447e-02 8.50951850e-01 -1.19361699e+00 1.06087990e-01 1.02764559e+00 4.54679519e-01 6.16949201e-01 7.97134638e-01 -7.95458972e-01 -1.24156630e+00 -9.22784328e-01 4.97735769e-01 -6.46053314e-01 3.12101036e-01 -1.44726455e-01 -5.60441375e-01 6.50742292e-01 -2.45471507e-01 4.54349592e-02 2.20978305e-01 2.36617967e-01 -1.30027592e-01 -3.36534858e-01 -8.44706655e-01 5.68914473e-01 9.12010133e-01 -4.13710266e-01 -4.72816110e-01 2.31376439e-01 3.44138324e-01 -7.90967464e-01 -6.34328187e-01 6.55362785e-01 6.50627196e-01 -1.13878226e+00 1.10777640e+00 -1.83593199e-01 1.34424582e-01 -6.44933939e-01 4.02524360e-02 -1.41296756e+00 -3.71306896e-01 -9.14174020e-02 1.65386051e-01 1.08696616e+00 1.66915447e-01 -1.92034781e-01 1.15466368e+00 4.67009306e-01 -8.70911181e-02 -3.64205420e-01 -1.15361977e+00 -3.07800889e-01 -5.76187670e-01 -5.13164520e-01 4.35292453e-01 4.27594602e-01 -4.25463736e-01 2.50188529e-01 -5.74598193e-01 5.88805676e-01 8.01414609e-01 1.56044513e-01 1.07319593e+00 -1.56002104e+00 -5.13917655e-02 -2.13485345e-01 -6.66675687e-01 -1.18869996e+00 3.32951456e-01 -5.69047570e-01 6.84703767e-01 -1.52153909e+00 1.83265522e-01 -4.95488584e-01 -1.79648668e-01 3.91412795e-01 -2.65729219e-01 4.69021946e-01 1.43744394e-01 4.09216523e-01 -7.27262080e-01 4.81417120e-01 8.54259610e-01 7.51009285e-02 -1.46206334e-01 3.96381825e-01 -2.70024717e-01 1.13243663e+00 5.86137414e-01 -4.59790349e-01 9.67590287e-02 -3.06800693e-01 2.73695022e-01 -9.81675163e-02 6.03193879e-01 -1.38864875e+00 6.04181707e-01 -1.16180755e-01 9.10758615e-01 -1.07225752e+00 5.41526973e-01 -1.27322233e+00 2.51258105e-01 4.38623756e-01 -9.58780050e-02 7.84724876e-02 -1.42359450e-01 8.62152338e-01 -3.43861580e-01 -3.65457386e-01 6.87809825e-01 -1.82068288e-01 -1.08711541e+00 2.17917234e-01 -5.59235662e-02 -7.86748350e-01 1.09087265e+00 -4.81690437e-01 2.60084033e-01 -4.21906501e-01 -7.14544356e-01 -8.59185383e-02 6.98519766e-01 3.46051067e-01 9.20516849e-01 -1.19527590e+00 -3.66883636e-01 3.03040445e-01 1.81161001e-01 4.92789328e-01 -1.32668093e-01 9.89817917e-01 -8.72842371e-01 2.91215897e-01 -2.39609256e-01 -1.09225571e+00 -1.27351189e+00 5.04163861e-01 2.87385911e-01 1.99470930e-02 -3.83541197e-01 7.94814050e-01 -1.79404840e-01 -4.33579832e-01 2.78137892e-01 -3.56722593e-01 -2.70481557e-01 -1.09551966e-01 -6.95851399e-03 2.76216298e-01 2.02424452e-01 -8.77443671e-01 -4.86700982e-01 1.21849060e+00 2.02220187e-01 -1.02614656e-01 1.43674231e+00 -3.55759799e-01 -1.16343156e-01 3.69299293e-01 9.77734923e-01 8.35473835e-02 -1.53199267e+00 -1.87315300e-01 2.71036774e-01 -6.87888682e-01 8.41759071e-02 -4.91271347e-01 -8.58362973e-01 8.90929341e-01 6.55286074e-01 -3.96917462e-01 9.22435760e-01 -7.13948756e-02 3.94792676e-01 4.28048223e-01 7.40314841e-01 -1.03539360e+00 5.38501516e-02 4.58423525e-01 8.01095843e-01 -1.41768098e+00 4.87290263e-01 -5.16955972e-01 -3.99683952e-01 1.59914708e+00 7.04148114e-01 -2.35606447e-01 3.56431037e-01 7.71977678e-02 -1.42801434e-01 2.52340972e-01 -2.36122534e-02 -4.95348871e-01 4.01037872e-01 7.42742777e-01 3.42741668e-01 -2.54539102e-01 1.26780003e-01 5.35597801e-02 -3.13377678e-01 -8.61082748e-02 2.71659076e-01 6.06658101e-01 -5.62506855e-01 -1.28921843e+00 -7.84845591e-01 7.94553310e-02 1.30633190e-01 2.26112887e-01 -6.55254245e-01 6.76464081e-01 2.64927000e-01 6.11672759e-01 4.59802710e-02 -6.30353749e-01 4.93127108e-01 1.19541511e-02 6.87746227e-01 -7.16257572e-01 -2.12159574e-01 2.62208432e-01 -2.72798896e-01 -7.61045933e-01 -6.89937711e-01 -7.46394753e-01 -9.76163089e-01 4.11196835e-02 -7.82937586e-01 -1.24818780e-01 9.25822794e-01 9.73838747e-01 -4.55272235e-02 2.45046705e-01 5.45782626e-01 -1.56212032e+00 -4.65472311e-01 -7.75474787e-01 -2.32978776e-01 2.04267889e-01 3.20990831e-01 -8.76002729e-01 -1.44463360e-01 8.32758993e-02]

[7.8181986808776855, -2.299677610397339]

532786fd-4918-4a6e-8667-be670f6ea3c3

reinforcement-learning-based-counter

2303.06433

https://arxiv.org/abs/2303.06433v1

https://arxiv.org/pdf/2303.06433v1.pdf

Reinforcement Learning-based Counter-Misinformation Response Generation: A Case Study of COVID-19 Vaccine Misinformation

The spread of online misinformation threatens public health, democracy, and the broader society. While professional fact-checkers form the first line of defense by fact-checking popular false claims, they do not engage directly in conversations with misinformation spreaders. On the other hand, non-expert ordinary users act as eyes-on-the-ground who proactively counter misinformation -- recent research has shown that 96% counter-misinformation responses are made by ordinary users. However, research also found that 2/3 times, these responses are rude and lack evidence. This work seeks to create a counter-misinformation response generation model to empower users to effectively correct misinformation. This objective is challenging due to the absence of datasets containing ground-truth of ideal counter-misinformation responses, and the lack of models that can generate responses backed by communication theories. In this work, we create two novel datasets of misinformation and counter-misinformation response pairs from in-the-wild social media and crowdsourcing from college-educated students. We annotate the collected data to distinguish poor from ideal responses that are factual, polite, and refute misinformation. We propose MisinfoCorrect, a reinforcement learning-based framework that learns to generate counter-misinformation responses for an input misinformation post. The model rewards the generator to increase the politeness, factuality, and refutation attitude while retaining text fluency and relevancy. Quantitative and qualitative evaluation shows that our model outperforms several baselines by generating high-quality counter-responses. This work illustrates the promise of generative text models for social good -- here, to help create a safe and reliable information ecosystem. The code and data is accessible on https://github.com/claws-lab/MisinfoCorrect.

['Srijan Kumar', 'Mustaque Ahamad', 'Bing He']

2023-03-11

['response-generation']

['natural-language-processing']

[ 2.85602361e-02 7.94968367e-01 -4.02507842e-01 -1.81628093e-01 -8.33683729e-01 -8.27445626e-01 1.07176578e+00 3.96400958e-01 -1.83015108e-01 8.54562819e-01 1.04126453e+00 -6.94972456e-01 3.82794648e-01 -9.99733090e-01 -5.83935618e-01 -3.20923589e-02 7.10293889e-01 4.35475856e-01 -9.23887938e-02 -1.02536452e+00 5.00875056e-01 -2.68031597e-01 -1.08519185e+00 6.74118340e-01 1.28016007e+00 1.15604110e-01 -4.11912441e-01 8.55559587e-01 -3.28117460e-01 1.50585675e+00 -9.65168238e-01 -1.10871577e+00 -9.66147333e-02 -6.95435047e-01 -1.00916028e+00 -3.22900146e-01 6.74467802e-01 -4.94059891e-01 -2.03632236e-01 1.28779697e+00 5.83314776e-01 -1.83551237e-01 3.76509756e-01 -1.06426156e+00 -1.27395535e+00 1.34969032e+00 -8.81214589e-02 1.30085707e-01 1.05548322e+00 5.11708856e-01 8.16977859e-01 -2.95135587e-01 8.85787964e-01 1.66713345e+00 6.55902207e-01 9.09764707e-01 -9.74165201e-01 -7.80555367e-01 -2.16199055e-01 -1.15673110e-01 -7.19180048e-01 -5.07596076e-01 6.96435452e-01 -8.38634253e-01 4.09597874e-01 6.82001293e-01 6.59562588e-01 2.00391078e+00 2.63765901e-02 4.52227920e-01 1.75220227e+00 -1.94764733e-01 1.61793768e-01 3.76810759e-01 2.54836410e-01 5.39666235e-01 5.11048675e-01 3.30889493e-01 -4.45776254e-01 -5.45860767e-01 2.34693542e-01 -1.97329283e-01 -1.65256098e-01 7.76499093e-01 -9.58995759e-01 1.21570551e+00 4.06633288e-01 5.86966336e-01 -4.09573406e-01 -1.39855698e-01 3.56645994e-02 4.62423652e-01 8.86184335e-01 7.38427401e-01 6.78992888e-04 -5.95743597e-01 -4.73866105e-01 7.46481895e-01 1.49234128e+00 5.43739080e-01 4.67349529e-01 8.24249387e-02 -3.45155239e-01 8.66471648e-01 1.33476958e-01 9.69301283e-01 5.98184049e-01 -1.03229809e+00 6.00118756e-01 8.94819915e-01 5.79490542e-01 -1.76759374e+00 -2.35192791e-01 -4.48010921e-01 -6.08033359e-01 -2.03210592e-01 7.51627982e-01 -4.89217877e-01 -1.56180054e-01 1.43390715e+00 4.50324386e-01 -1.59974054e-01 -1.18166069e-02 9.97091413e-01 9.55966413e-01 5.65654874e-01 2.47488469e-01 1.26799839e-02 1.41236305e+00 -3.24085385e-01 -8.89171839e-01 -6.00802302e-01 8.32790732e-01 -9.93386447e-01 1.10865235e+00 1.67193964e-01 -1.02483320e+00 -4.45709676e-02 -6.90604150e-01 -1.33537101e-02 -3.92510206e-01 -5.26834786e-01 2.36269116e-01 1.12218058e+00 -6.52412295e-01 4.64855373e-01 -6.87603951e-02 -1.01262838e-01 4.90726113e-01 -5.49984396e-01 -1.33440960e-02 -4.32694554e-02 -1.67926788e+00 9.38041747e-01 -2.71504074e-02 -4.00497109e-01 -6.70135438e-01 -8.08452368e-01 -7.17742562e-01 -2.83914477e-01 5.48365355e-01 -6.67391419e-01 1.70766997e+00 -1.19617772e+00 -1.65290439e+00 8.95960450e-01 2.95457959e-01 -3.45475405e-01 9.54012156e-01 -2.14294598e-01 -4.34335649e-01 2.02957238e-03 3.51080894e-01 1.73011035e-01 7.01765537e-01 -1.30833852e+00 -1.26326859e-01 -1.42731786e-01 2.33201474e-01 -8.45723599e-02 -7.65003487e-02 2.01113477e-01 8.02959859e-01 -6.84777260e-01 -3.73098582e-01 -7.96499908e-01 -1.93213269e-01 -7.67695487e-01 -6.46154523e-01 -3.22829247e-01 3.03148508e-01 -9.02318001e-01 1.34060884e+00 -1.65553570e+00 -7.93227911e-01 3.31529751e-02 6.02976799e-01 6.71850026e-01 -7.64054134e-02 8.51137817e-01 5.24807498e-02 9.75479960e-01 2.18974143e-01 2.14827508e-01 2.10498646e-01 -3.05348001e-02 -6.03009224e-01 3.79631341e-01 -4.28408384e-02 1.02494717e+00 -1.43975401e+00 -1.72301441e-01 -2.70088702e-01 2.73650140e-01 -7.34999478e-01 1.58891156e-01 -4.98803973e-01 5.29583693e-01 -5.61710894e-01 1.79171115e-01 5.93852699e-01 -4.76350039e-01 2.67837644e-01 4.21677709e-01 -2.12221831e-01 1.00451756e+00 -7.58318126e-01 6.73402667e-01 -4.39260244e-01 5.39639354e-01 8.76532197e-02 -3.04738432e-01 7.69343197e-01 2.17525110e-01 -3.83819491e-02 -1.01809907e+00 4.34760094e-01 3.32781374e-01 -2.74898415e-03 -8.75721633e-01 7.28121877e-01 -2.30230048e-01 -4.48421925e-01 1.13511634e+00 -4.64610130e-01 -4.08353746e-01 -3.47406790e-02 6.88244522e-01 8.82275581e-01 -4.56973672e-01 5.64718723e-01 -9.27737951e-02 3.38855177e-01 2.04432428e-01 2.99018413e-01 1.11214864e+00 -2.11688280e-01 7.52860829e-02 6.36753142e-01 -3.79293084e-01 -7.64102995e-01 -5.36941886e-01 3.56934130e-01 1.34165347e+00 -7.00708926e-02 -3.36163878e-01 -1.16249537e+00 -7.99102306e-01 -5.04522175e-02 1.26247621e+00 -3.72248501e-01 -6.16681315e-02 -3.86659533e-01 -4.05784160e-01 1.01059616e+00 -3.08331877e-01 7.07214534e-01 -7.49714553e-01 -2.34171569e-01 4.10198063e-01 -1.13482022e+00 -1.02054751e+00 -4.93062168e-01 -9.94228244e-01 -2.50156224e-01 -1.32226789e+00 -2.57547259e-01 -5.25914319e-02 5.16623318e-01 4.42194194e-01 1.15057230e+00 7.30294406e-01 2.71134406e-01 3.80541980e-01 -4.88240838e-01 -5.94993114e-01 -1.51731753e+00 -3.66995990e-01 -1.55750483e-01 -6.02503940e-02 3.89731318e-01 -4.14182186e-01 -5.52208900e-01 2.17207953e-01 -9.65600610e-01 -7.31705129e-02 1.10060655e-01 5.66732049e-01 -5.16433537e-01 -6.87123418e-01 7.86501527e-01 -1.37798989e+00 1.35756338e+00 -1.00902104e+00 -2.49387324e-01 -1.03608862e-01 -6.32538915e-01 -1.85098827e-01 6.18271530e-01 -4.64954138e-01 -1.19678748e+00 -7.92794883e-01 -3.79238725e-01 6.29947901e-01 -1.68846086e-01 4.34083939e-01 5.30760586e-01 -1.60103850e-02 1.75954723e+00 5.30850999e-02 1.18243203e-01 -2.08923295e-01 6.47420704e-01 1.10487700e+00 1.57246619e-01 -5.67608118e-01 7.64582634e-01 1.75781697e-01 -6.89566493e-01 -8.61006796e-01 -1.25508153e+00 -3.41078281e-01 3.59404981e-01 -5.82589805e-01 4.88370776e-01 -8.30111384e-01 -1.38009787e+00 3.76174629e-01 -1.44883227e+00 -4.10420448e-01 9.23502445e-02 5.31246811e-02 -2.71211654e-01 5.76993883e-01 -8.92068386e-01 -9.77766812e-01 -3.70088160e-01 -5.48018932e-01 4.36588436e-01 2.13253781e-01 -7.76703596e-01 -1.00127113e+00 2.78896093e-01 1.06731081e+00 8.76309931e-01 2.34824553e-01 5.17612159e-01 -1.30284226e+00 -3.22657019e-01 -2.68272102e-01 -1.35847423e-02 1.18672818e-01 -1.47617698e-01 -5.59953749e-02 -9.33319092e-01 2.81956345e-01 1.02168331e-02 -7.32370555e-01 4.09953177e-01 -1.41121224e-01 3.84780645e-01 -1.70341253e+00 1.33248895e-01 -2.98893213e-01 9.44029510e-01 -3.73774260e-01 5.86759269e-01 1.46118954e-01 2.41641641e-01 9.44410324e-01 1.05654374e-01 6.09596312e-01 1.02820218e+00 9.43967178e-02 2.40310013e-01 4.43533152e-01 -1.40022263e-01 -9.80875552e-01 5.63866258e-01 7.51988769e-01 6.91113025e-02 -3.95973742e-01 -1.03733027e+00 4.58896399e-01 -1.53320920e+00 -1.64454365e+00 -6.69897437e-01 1.73801780e+00 1.36775756e+00 1.24465667e-01 2.71269530e-01 -1.14982314e-01 5.11844158e-01 3.03316265e-01 2.88492560e-01 -7.64156282e-01 -1.59275848e-02 -1.99115336e-01 2.29185566e-01 1.29015970e+00 -5.61126113e-01 9.41212893e-01 5.52957344e+00 5.26998997e-01 -1.01999450e+00 4.85179007e-01 7.35890627e-01 4.17611361e-01 -1.17008269e+00 1.16801724e-01 -7.53816664e-01 7.14291871e-01 1.10265052e+00 -4.26027298e-01 5.90598524e-01 6.70933425e-01 5.00625908e-01 -1.32207736e-01 -5.87310612e-01 5.56000829e-01 1.68872923e-01 -1.63214016e+00 -4.61248308e-02 -1.21115670e-01 1.06800854e+00 -1.61211029e-01 -7.93313887e-03 4.63095754e-01 1.25512171e+00 -8.48830760e-01 8.37424099e-01 5.95933974e-01 2.97871768e-01 -1.83789618e-02 5.60888708e-01 1.01929128e+00 3.49874906e-02 -2.17480406e-01 8.39961618e-02 -6.41243339e-01 3.00451517e-02 8.42970371e-01 -1.41470623e+00 9.76115018e-02 -1.80323627e-02 3.78863484e-01 -6.61415994e-01 5.46884716e-01 -6.60731077e-01 9.96156871e-01 1.14948168e-01 -6.94264352e-01 2.08416998e-01 7.21450821e-02 6.75795138e-01 1.44622886e+00 8.62426385e-02 3.93448174e-01 2.54561961e-01 1.29878831e+00 -3.56931448e-01 1.50019377e-01 -8.25140119e-01 -6.18369102e-01 8.24493408e-01 1.16255653e+00 -7.99034387e-02 -4.88357216e-01 -5.87688386e-03 3.71496618e-01 2.00142995e-01 4.48648393e-01 -4.28946465e-01 1.66776836e-01 3.86132121e-01 5.06447732e-01 -5.88314295e-01 1.24752238e-01 -6.13166869e-01 -1.03829002e+00 -4.23684567e-01 -1.66216969e+00 2.69471318e-01 -6.56672657e-01 -1.65290201e+00 1.22251004e-01 -2.54647076e-01 -6.83150828e-01 -7.35511899e-01 -2.35735446e-01 -5.65043926e-01 8.45662117e-01 -1.29418683e+00 -9.86707449e-01 -4.12571132e-01 1.29640281e-01 3.66819143e-01 2.38796234e-01 4.80940282e-01 4.99274135e-02 -1.63266674e-01 4.31720972e-01 -6.05323434e-01 3.06895912e-01 7.71822691e-01 -9.19245541e-01 5.35853684e-01 6.15583837e-01 -3.32324803e-01 1.04492617e+00 1.13957179e+00 -1.17974639e+00 -9.08159435e-01 -8.15373361e-01 1.64690304e+00 -9.15171266e-01 1.03773725e+00 -2.03889668e-01 -9.43658471e-01 6.26009881e-01 5.68983972e-01 -5.57187080e-01 9.97980058e-01 -1.42830968e-01 -7.11559236e-01 4.68448132e-01 -1.22151625e+00 8.99371505e-01 8.40939581e-01 -6.43485367e-01 -7.56551683e-01 7.10498214e-01 7.57196069e-01 -3.85232449e-01 -4.48040098e-01 -5.70574343e-01 3.42511803e-01 -1.13030291e+00 6.99479282e-01 -8.15444708e-01 1.00015652e+00 2.25042567e-01 1.40067682e-01 -1.65216374e+00 -2.07720995e-01 -1.38302994e+00 2.12184012e-01 1.20891333e+00 3.90648305e-01 -7.92991757e-01 3.54783088e-01 7.87831545e-01 5.64074442e-02 -1.86327934e-01 -6.26574218e-01 -4.05411810e-01 4.01870310e-01 -3.93130541e-01 7.28659153e-01 1.51848376e+00 4.39470828e-01 4.15509433e-01 -3.37386250e-01 -1.50762826e-01 5.59991419e-01 -2.69656897e-01 9.61031199e-01 -1.06284249e+00 -1.27938822e-01 -5.54900646e-01 3.62048864e-01 -8.26628625e-01 1.30226523e-01 -9.28975344e-01 -1.08859148e-02 -1.39286661e+00 9.26740915e-02 -2.84478217e-01 9.02234733e-01 4.48607117e-01 -4.42153513e-01 -2.96110241e-03 1.34160608e-01 -1.46448035e-02 -3.16606134e-01 1.01474442e-01 1.61009061e+00 -8.98052529e-02 -1.91388205e-01 -5.29926419e-02 -1.47564709e+00 8.21413219e-01 1.00895083e+00 -7.49283195e-01 -5.63271232e-02 -9.52531546e-02 1.24135935e+00 2.59639025e-01 8.03138793e-01 -3.67714792e-01 2.07556978e-01 -7.59101510e-01 -2.07215965e-01 1.07515439e-01 -1.53604358e-01 -1.76991835e-01 -1.22035421e-01 5.84095776e-01 -8.18110108e-01 -1.72514059e-02 -3.24339390e-01 5.63807249e-01 2.46031031e-01 -3.38053465e-01 5.01885116e-01 -5.83541155e-01 -6.54804930e-02 -2.60392994e-01 -8.45754921e-01 8.14544022e-01 4.91667807e-01 1.47265911e-01 -1.50294888e+00 -1.18972921e+00 -1.74040064e-01 -1.90392181e-01 2.36628160e-01 5.04064262e-01 4.93326545e-01 -1.12851274e+00 -1.35481930e+00 -6.94636106e-02 6.48499653e-02 -7.25897551e-01 2.91756541e-01 7.89731383e-01 -5.17301261e-01 2.18351081e-01 8.16340297e-02 5.39974608e-02 -7.73661852e-01 2.85281301e-01 2.96416253e-01 -1.24835782e-01 -8.95661041e-02 7.16816008e-01 -4.67162758e-01 -7.55690813e-01 -3.70151371e-01 -4.32849564e-02 -3.70048225e-01 1.23731270e-01 9.47974026e-01 7.26634443e-01 -1.29980266e-01 -7.23876476e-01 1.16358429e-01 -4.25540090e-01 1.18109092e-01 -2.89906025e-01 9.40610170e-01 -8.23874474e-02 -2.37774536e-01 1.58608809e-01 8.66882324e-01 7.87679970e-01 -4.87881064e-01 -3.18912804e-01 2.19725911e-02 -9.73931909e-01 -3.41674387e-01 -1.35781872e+00 -1.94172531e-01 4.39764500e-01 -2.62645900e-01 8.42693686e-01 1.44916505e-01 -1.19911507e-01 8.95164371e-01 1.67619079e-01 4.14982364e-02 -1.15446568e+00 6.77252710e-01 6.68756187e-01 1.23962188e+00 -1.18913496e+00 -2.80194163e-01 -5.24998367e-01 -9.42232728e-01 9.08481419e-01 5.64370453e-01 1.04200289e-01 3.18644583e-01 -5.69170108e-03 4.56595063e-01 -2.44121939e-01 -8.35034490e-01 -7.84290780e-04 1.47436485e-01 6.65395856e-01 7.86146164e-01 5.53627372e-01 -8.57078671e-01 8.49538088e-01 -8.37715864e-01 2.59144232e-04 1.25989616e+00 5.52057505e-01 -7.44218051e-01 -8.50565791e-01 -6.64772689e-01 1.95847481e-01 -8.32163811e-01 -1.02250427e-01 -1.18037236e+00 3.40832174e-01 -7.00423941e-02 1.75513518e+00 -4.83391583e-01 -5.54433048e-01 1.64358750e-01 3.58965360e-02 7.37245474e-03 -6.03527308e-01 -1.23895764e+00 -4.18734342e-01 9.48018968e-01 -4.79998559e-01 -1.79998860e-01 -2.97714859e-01 -9.19865847e-01 -1.17321205e+00 9.69584063e-02 1.42840862e-01 5.01312971e-01 9.78922844e-01 6.02048874e-01 -9.86548960e-02 7.19602823e-01 -6.48866892e-02 -1.22849560e+00 -1.10744262e+00 2.29555577e-01 8.82702112e-01 5.73259652e-01 -1.33137435e-01 -4.14649189e-01 -2.34123677e-01]

[8.618947982788086, 10.149832725524902]

f7bd6611-5f24-48ca-b4d8-9bdeb0a6b3c2

self-supervised-multimodal-versatile-networks

2006.16228

https://arxiv.org/abs/2006.16228v2

https://arxiv.org/pdf/2006.16228v2.pdf

Self-Supervised MultiModal Versatile Networks

Videos are a rich source of multi-modal supervision. In this work, we learn representations using self-supervision by leveraging three modalities naturally present in videos: visual, audio and language streams. To this end, we introduce the notion of a multimodal versatile network -- a network that can ingest multiple modalities and whose representations enable downstream tasks in multiple modalities. In particular, we explore how best to combine the modalities, such that fine-grained representations of the visual and audio modalities can be maintained, whilst also integrating text into a common embedding. Driven by versatility, we also introduce a novel process of deflation, so that the networks can be effortlessly applied to the visual data in the form of video or a static image. We demonstrate how such networks trained on large collections of unlabelled video data can be applied on video, video-text, image and audio tasks. Equipped with these representations, we obtain state-of-the-art performance on multiple challenging benchmarks including UCF101, HMDB51, Kinetics600, AudioSet and ESC-50 when compared to previous self-supervised work. Our models are publicly available.

['Jeffrey De Fauw', 'Relja Arandjelović', 'Adrià Recasens', 'Jean-Baptiste Alayrac', 'Sander Dieleman', 'Rosalia Schneider', 'Lucas Smaira', 'Jason Ramapuram', 'Andrew Zisserman']

2020-06-29

http://proceedings.neurips.cc/paper/2020/hash/0060ef47b12160b9198302ebdb144dcf-Abstract.html

http://proceedings.neurips.cc/paper/2020/file/0060ef47b12160b9198302ebdb144dcf-Paper.pdf

neurips-2020-12

['self-supervised-action-recognition']

['computer-vision']

[ 3.73257875e-01 -1.28224090e-01 -4.15843874e-01 -2.13586241e-01 -8.22414637e-01 -7.16043890e-01 8.63796711e-01 -1.55725494e-01 -5.17704427e-01 6.15163743e-01 7.00265527e-01 -4.91932109e-02 2.31788859e-01 -3.97957385e-01 -9.92853820e-01 -5.41622281e-01 -1.03778042e-01 -4.03681844e-02 1.59900531e-01 -2.38800377e-01 -2.72762030e-01 1.68562949e-01 -1.94212198e+00 7.73774683e-01 1.45803913e-01 1.14873850e+00 5.53513244e-02 8.61698687e-01 2.17725262e-01 9.45654631e-01 -1.95552960e-01 -4.26891327e-01 5.93189411e-02 -2.73989022e-01 -8.59092057e-01 2.77596116e-01 6.20155990e-01 -4.95994955e-01 -7.49524832e-01 5.99889994e-01 5.53454459e-01 1.09187588e-01 5.94685316e-01 -1.40777075e+00 -6.26215041e-01 6.19109690e-01 -4.25244868e-01 1.31683543e-01 6.91341937e-01 2.32328281e-01 1.17312098e+00 -9.33722913e-01 8.90131354e-01 1.35992229e+00 2.94116199e-01 5.94538569e-01 -1.21061993e+00 -5.35493970e-01 2.89015859e-01 3.99932355e-01 -1.07848668e+00 -9.02393460e-01 6.42850280e-01 -4.29538399e-01 7.81021118e-01 1.76226154e-01 4.56687480e-01 1.88744509e+00 -1.41742259e-01 1.03545666e+00 7.61812985e-01 -4.65535134e-01 -1.06562234e-01 1.10985935e-01 -2.49124587e-01 6.54238582e-01 -3.90706718e-01 2.85307877e-02 -1.24547446e+00 7.93157741e-02 5.36868989e-01 2.43598491e-01 -5.71224630e-01 -4.15067762e-01 -1.62004936e+00 7.41963327e-01 3.81961286e-01 2.74935991e-01 -7.50397667e-02 1.99051097e-01 8.32306147e-01 6.81609333e-01 4.22758728e-01 7.84706138e-03 -3.69675189e-01 -2.62847185e-01 -8.11879098e-01 -1.76021397e-01 6.38817728e-01 8.47798944e-01 6.69862568e-01 -9.41600185e-03 -2.28750855e-01 7.58212805e-01 4.56141055e-01 5.31134546e-01 6.46571994e-01 -9.31171775e-01 7.32012153e-01 2.98615575e-01 -2.70904303e-01 -6.72296345e-01 -7.25470632e-02 -8.21590573e-02 -1.04025745e+00 -1.81563608e-02 1.11152783e-01 -9.43392292e-02 -1.06321716e+00 1.93943238e+00 1.25334233e-01 5.10461450e-01 2.85416871e-01 7.12196171e-01 1.15928853e+00 6.06274307e-01 7.64722154e-02 -1.19859025e-01 1.31415343e+00 -1.21695185e+00 -7.03521192e-01 -1.03169598e-01 3.97490591e-01 -5.07161438e-01 8.30946445e-01 3.37554634e-01 -1.04673970e+00 -7.48549700e-01 -9.92777705e-01 -9.21563432e-02 -4.45892543e-01 3.75694260e-02 2.52637714e-01 1.21261306e-01 -1.21181369e+00 5.03680348e-01 -7.83884525e-01 -5.43661773e-01 4.02801454e-01 1.37672871e-01 -1.00815248e+00 -3.31145912e-01 -1.21687877e+00 6.11728847e-01 4.25103873e-01 -7.64052644e-02 -1.41839433e+00 -4.68876868e-01 -1.25840151e+00 -1.07002802e-01 3.80844057e-01 -8.24923337e-01 1.09769070e+00 -1.39882529e+00 -1.46541524e+00 1.00809455e+00 -1.55587137e-01 -4.76108670e-01 4.90070403e-01 -3.14102381e-01 -6.46672964e-01 8.08333457e-01 6.29501492e-02 8.87878060e-01 1.30362439e+00 -1.23261952e+00 -6.43525720e-01 -2.38870963e-01 3.36924046e-01 2.74100333e-01 -7.44096339e-01 5.59958890e-02 -6.24925673e-01 -6.00195527e-01 -2.93813914e-01 -9.31184947e-01 2.58355707e-01 9.46625546e-02 -5.43578088e-01 -4.16141339e-02 1.10414851e+00 -3.64111692e-01 1.03090751e+00 -2.56846666e+00 7.55123854e-01 -1.23487242e-01 2.24991515e-01 8.36808234e-02 -4.25413579e-01 6.22151077e-01 -2.11373761e-01 1.15901880e-01 8.44353363e-02 -8.15361559e-01 5.76466434e-02 3.09380502e-01 -2.89435804e-01 4.85338330e-01 3.33155900e-01 9.54091907e-01 -8.85477901e-01 -5.31295717e-01 2.62874961e-01 7.55314529e-01 -4.71913546e-01 5.10314107e-01 -1.48630723e-01 5.03752947e-01 -5.27234748e-02 8.33662510e-01 2.04608902e-01 -5.10212362e-01 7.33444095e-02 -4.68824655e-01 1.69987261e-01 1.27212405e-01 -1.04406595e+00 2.23387194e+00 -4.83701468e-01 7.15222597e-01 3.66657883e-01 -1.09915149e+00 3.76543432e-01 7.41690457e-01 3.54055196e-01 -4.19787526e-01 6.41384572e-02 -8.01408011e-03 -5.45512557e-01 -6.68243885e-01 3.97217423e-01 -2.07107052e-01 -9.77281854e-02 4.65019107e-01 8.62194836e-01 2.50490785e-01 3.71087670e-01 4.99949694e-01 9.61828887e-01 1.49798974e-01 2.82232892e-02 3.39579284e-01 3.71133715e-01 -3.95581603e-01 5.62652797e-02 5.54657638e-01 -5.03361672e-02 7.33214557e-01 4.27274406e-01 -4.33884561e-02 -7.42576182e-01 -1.16432798e+00 1.80220976e-02 1.80798388e+00 -5.54624759e-02 -7.04363644e-01 -2.19655827e-01 -7.20416248e-01 5.34987338e-02 6.83444068e-02 -8.46142590e-01 -1.80169672e-01 -1.73307851e-01 -2.41394207e-01 5.99686742e-01 6.17190659e-01 2.32630536e-01 -9.86852765e-01 -2.89164275e-01 -1.59114192e-03 -2.79819757e-01 -1.38618565e+00 -3.47197860e-01 3.37299436e-01 -6.14402056e-01 -1.04652810e+00 -8.56405079e-01 -8.21573138e-01 2.00237870e-01 4.49989259e-01 1.10528207e+00 -3.74471582e-02 -4.38706614e-02 9.97072041e-01 -6.29505038e-01 1.11095412e-02 -3.29206079e-01 3.88583094e-02 5.65228425e-02 4.91599053e-01 -2.59640396e-01 -8.13943803e-01 -3.90114456e-01 1.41884521e-01 -1.27505922e+00 5.24102747e-02 4.14245546e-01 1.00734210e+00 4.37008768e-01 -4.78022009e-01 5.37381649e-01 -7.29339719e-01 8.23198110e-02 -9.13131833e-01 1.02494940e-01 2.74256885e-01 1.56530023e-01 -1.12062976e-01 6.14560723e-01 -5.70806444e-01 -8.98398280e-01 1.55092284e-01 -1.91824865e-02 -1.02136230e+00 -3.46315265e-01 6.70089006e-01 -1.93294823e-01 2.42917538e-02 5.42640209e-01 2.38492921e-01 2.08825886e-01 -4.97794718e-01 8.28249454e-01 7.43376374e-01 8.21792483e-01 -4.54900563e-01 7.39153087e-01 8.21357489e-01 -1.28673509e-01 -8.20783675e-01 -8.53424966e-01 -6.52227879e-01 -7.69219160e-01 -3.33427012e-01 1.02254653e+00 -1.42355502e+00 -4.41022009e-01 4.13517535e-01 -9.85167801e-01 -1.36952773e-01 -1.69755921e-01 3.75424474e-01 -6.53863847e-01 4.22191679e-01 -8.13713729e-01 -4.19396579e-01 2.75224075e-02 -1.11884046e+00 1.37042725e+00 1.13324672e-01 -2.48201769e-02 -1.12813091e+00 -4.16434444e-02 4.61520821e-01 2.43641406e-01 3.67452055e-01 3.83856118e-01 -7.77193606e-01 -4.54776675e-01 -4.54181731e-02 -1.16907567e-01 4.33604062e-01 1.11319721e-01 1.47013947e-01 -1.35816503e+00 -5.49691379e-01 -4.30144995e-01 -1.19950438e+00 1.31522715e+00 -7.53902644e-02 1.04572225e+00 -2.23410651e-01 -2.52421767e-01 9.01682675e-01 1.24662256e+00 -4.24113631e-01 3.67072731e-01 4.29180443e-01 9.04350221e-01 4.45016533e-01 2.52775043e-01 4.60015476e-01 4.84571725e-01 5.42836666e-01 7.13442743e-01 -2.11192697e-01 -6.22883290e-02 -4.18407887e-01 6.31341696e-01 8.88525009e-01 -8.79281312e-02 -3.92514735e-01 -5.58446109e-01 5.85770786e-01 -1.93895698e+00 -1.24677539e+00 2.95425475e-01 1.86737490e+00 8.31687391e-01 -1.30315982e-02 3.20659697e-01 2.28995383e-01 5.31917751e-01 4.64181155e-01 -3.52783144e-01 -2.03314633e-03 -3.44201177e-01 9.34208836e-03 9.55432355e-02 1.47943214e-01 -1.53531039e+00 5.37413061e-01 6.42221069e+00 5.79945505e-01 -1.22575724e+00 2.90837079e-01 3.91498625e-01 -4.65768039e-01 -2.87981570e-01 -3.91078264e-01 -5.85969985e-01 4.19479251e-01 1.30015898e+00 4.07705717e-02 3.48904252e-01 5.78163207e-01 -1.15803339e-01 1.72291607e-01 -1.54099822e+00 1.04762483e+00 2.56115794e-01 -1.44888937e+00 2.31124252e-01 -2.73352325e-01 5.28757095e-01 2.92701393e-01 2.42040306e-01 4.28315192e-01 1.67350978e-01 -1.21253681e+00 8.70145619e-01 3.89409631e-01 1.20193768e+00 -4.05172229e-01 4.56976116e-01 1.97483927e-01 -1.29312718e+00 -1.46382928e-01 1.14840046e-01 2.95496821e-01 3.69618177e-01 1.82869136e-01 -3.21732312e-01 9.70577776e-01 1.01962757e+00 1.46063280e+00 -5.01807928e-01 6.81651413e-01 -1.72040209e-01 4.91040528e-01 -2.55547285e-01 5.59296966e-01 2.28187710e-01 3.22917581e-01 4.04928923e-01 1.56428528e+00 1.55115485e-01 -2.96267420e-01 2.56248534e-01 3.31653625e-01 -3.74676973e-01 -1.33964077e-01 -8.78509402e-01 -2.40055427e-01 2.96623558e-01 1.17865407e+00 -1.91860452e-01 -4.77081865e-01 -8.00809383e-01 1.13266098e+00 4.24560934e-01 5.28257608e-01 -7.63784587e-01 -3.30691859e-02 6.08041227e-01 -1.48644105e-01 5.28487504e-01 -8.02030845e-04 4.95098114e-01 -1.52086544e+00 -3.19303200e-02 -1.12569773e+00 6.65957630e-01 -9.15950239e-01 -1.50633132e+00 7.07542777e-01 -2.15689968e-02 -1.38702631e+00 -4.01921034e-01 -6.44425809e-01 -4.02811885e-01 5.12137771e-01 -1.77313995e+00 -1.43032718e+00 -3.09686333e-01 1.25594783e+00 4.45375532e-01 -3.37147385e-01 8.74785244e-01 5.00205576e-01 -4.26985383e-01 6.22955382e-01 1.26932725e-01 2.65081048e-01 1.09042776e+00 -1.04824495e+00 -1.34320423e-01 4.88171577e-01 4.30807441e-01 3.29125583e-01 3.61042857e-01 -1.77955747e-01 -1.66315949e+00 -1.08773327e+00 4.63795871e-01 -4.64342058e-01 1.09167826e+00 -4.68889892e-01 -8.09446335e-01 9.48252082e-01 5.68080246e-01 4.52099264e-01 9.19343948e-01 1.21863395e-01 -8.23413730e-01 -4.70160991e-02 -7.18043089e-01 3.70158434e-01 1.15808213e+00 -1.08948219e+00 -6.68985128e-01 3.10126215e-01 1.05993474e+00 -4.00965452e-01 -8.99990916e-01 3.43123764e-01 5.14091313e-01 -1.04760849e+00 1.16535008e+00 -9.83822703e-01 7.80234873e-01 -1.09428748e-01 -5.07322848e-01 -1.31961107e+00 1.41226128e-02 -7.50159204e-01 -3.35022062e-01 1.41802454e+00 3.74331176e-01 -3.11431825e-01 2.95859903e-01 -9.23833698e-02 -1.27238765e-01 -5.27684033e-01 -1.15718389e+00 -6.03180349e-01 -2.67806292e-01 -4.41861153e-01 2.16511592e-01 1.00481880e+00 2.45495170e-01 5.43941081e-01 -7.68107712e-01 -1.60705261e-02 4.43393916e-01 -4.29502549e-03 7.36895859e-01 -9.58352327e-01 -4.54307050e-01 -1.62817806e-01 -5.45889080e-01 -1.11625433e+00 4.81598556e-01 -8.93782794e-01 -2.13384300e-01 -1.18857968e+00 4.25680220e-01 1.97012544e-01 -5.60877621e-01 6.42259717e-01 7.02007115e-02 5.49336493e-01 5.08199215e-01 2.35093459e-01 -1.17433059e+00 7.47041523e-01 1.15704262e+00 -2.73935378e-01 1.96607754e-01 -3.63507152e-01 -6.91876948e-01 5.34103930e-01 4.23923701e-01 -1.42107099e-01 -4.16465312e-01 -5.81409216e-01 8.15519467e-02 3.06569338e-01 5.37429571e-01 -7.69940078e-01 1.41239554e-01 1.12506136e-01 3.77332330e-01 -2.95640081e-01 7.42994308e-01 -8.48601878e-01 1.00125968e-02 -6.08839318e-02 -6.25989199e-01 -2.48433389e-02 2.91325867e-01 9.77905691e-01 -6.92557633e-01 1.74528018e-01 5.60177386e-01 -6.25159219e-02 -7.40292609e-01 3.28298301e-01 -2.69398987e-01 1.30872041e-01 9.70804989e-01 8.62033442e-02 -5.48712611e-01 -6.13086760e-01 -1.00385380e+00 3.30830991e-01 4.49966520e-01 7.24757671e-01 6.11168981e-01 -1.63825047e+00 -6.14981532e-01 2.49952346e-01 3.67446840e-01 -1.99667543e-01 3.17887396e-01 9.17130172e-01 6.54181615e-02 2.54012764e-01 -2.96152383e-01 -9.76587474e-01 -1.33337283e+00 6.01764679e-01 9.07325596e-02 -9.62613374e-02 -4.69725937e-01 7.75981724e-01 1.60793006e-01 -2.36072972e-01 5.97216666e-01 -6.05283529e-02 -3.26571196e-01 4.43645567e-01 6.62670076e-01 -7.91191906e-02 -1.10617727e-01 -9.50299740e-01 -3.28869402e-01 4.07472610e-01 5.75388968e-02 -2.27596253e-01 1.58574748e+00 -3.34719956e-01 7.89520070e-02 8.65100861e-01 1.55949056e+00 -1.72684610e-01 -1.43227565e+00 -4.72508848e-01 -3.36290538e-01 -3.55135649e-01 -4.86997031e-02 -4.94380355e-01 -1.18557715e+00 1.02137494e+00 4.58336860e-01 4.11500752e-01 1.22622132e+00 2.96172172e-01 4.69965428e-01 3.22221428e-01 7.45678246e-02 -9.43168223e-01 5.37457585e-01 5.41857839e-01 9.49862242e-01 -1.54211664e+00 -3.35884780e-01 -6.28502220e-02 -9.05684888e-01 1.20465589e+00 3.87824506e-01 1.26668006e-01 6.92446232e-01 2.12557957e-01 -4.19468097e-02 -3.18818763e-02 -1.17242265e+00 -3.46721977e-01 4.80839014e-01 6.05808914e-01 5.20836294e-01 -1.92765564e-01 5.61380446e-01 5.02766490e-01 1.73620731e-01 7.10965469e-02 4.26310986e-01 1.14124072e+00 -2.16674358e-01 -1.00423837e+00 -3.05930257e-01 3.35108846e-01 -6.04941785e-01 -5.22166453e-02 -2.77113557e-01 7.90984869e-01 -1.94894709e-02 1.03013492e+00 -3.97697762e-02 -4.90169138e-01 1.80070922e-01 2.09448129e-01 3.19493681e-01 -6.65541708e-01 -3.81113708e-01 1.70896307e-01 2.16965869e-01 -7.56340802e-01 -1.02152431e+00 -5.63113272e-01 -7.28547871e-01 -1.91687062e-01 1.29621821e-02 -1.40355110e-01 4.14669603e-01 9.61523771e-01 3.35739076e-01 4.69719529e-01 5.89057982e-01 -1.36337566e+00 -3.58816177e-01 -9.38238680e-01 -5.23889363e-01 7.12828994e-01 9.03185606e-01 -7.55188227e-01 -6.93668783e-01 4.87443328e-01]

[10.159567832946777, 1.1123274564743042]

d9f8f50a-b740-49ac-aca8-06814bce6d21

stochastic-gradient-bayesian-optimal

2306.15731

https://arxiv.org/abs/2306.15731v1

https://arxiv.org/pdf/2306.15731v1.pdf

Stochastic Gradient Bayesian Optimal Experimental Designs for Simulation-based Inference

Simulation-based inference (SBI) methods tackle complex scientific models with challenging inverse problems. However, SBI models often face a significant hurdle due to their non-differentiable nature, which hampers the use of gradient-based optimization techniques. Bayesian Optimal Experimental Design (BOED) is a powerful approach that aims to make the most efficient use of experimental resources for improved inferences. While stochastic gradient BOED methods have shown promising results in high-dimensional design problems, they have mostly neglected the integration of BOED with SBI due to the difficult non-differentiable property of many SBI simulators. In this work, we establish a crucial connection between ratio-based SBI inference algorithms and stochastic gradient-based variational inference by leveraging mutual information bounds. This connection allows us to extend BOED to SBI applications, enabling the simultaneous optimization of experimental designs and amortized inference functions. We demonstrate our approach on a simple linear model and offer implementation details for practitioners.

['Elliot E. Hui', 'Vincent D. Zaballa']

2023-06-27

['experimental-design']

['methodology']

[ 9.52640846e-02 -4.00439113e-01 -9.34909359e-02 -2.33015478e-01 -6.98418379e-01 -4.18893516e-01 4.81555879e-01 -1.57438472e-01 -4.67799187e-01 1.15244949e+00 -1.87754452e-01 -7.12358356e-01 -7.34167874e-01 -6.00878119e-01 -9.51978564e-01 -8.49113107e-01 -8.93843397e-02 4.37825441e-01 -1.19699031e-01 1.43900849e-02 4.45951343e-01 7.28810966e-01 -1.48673141e+00 -6.16532922e-01 9.35775220e-01 7.64280498e-01 3.00463021e-01 4.98278409e-01 8.10409561e-02 1.97387815e-01 -3.15210670e-01 -7.48168528e-02 -1.06938697e-01 -5.89657605e-01 -3.12124044e-01 -5.56718290e-01 -4.96270992e-02 -2.47656316e-01 -6.48042047e-03 9.27048266e-01 7.51426160e-01 4.77027982e-01 9.83550072e-01 -1.45969021e+00 -3.38794030e-02 2.92583019e-01 -5.09970903e-01 5.25646470e-02 2.79262185e-01 1.10348709e-01 9.20521915e-01 -6.92672491e-01 4.73158300e-01 1.27178347e+00 6.63916945e-01 1.63746566e-01 -1.75753486e+00 -7.13150442e-01 -9.45761055e-02 1.79759875e-01 -1.52151954e+00 -3.42912167e-01 6.95123613e-01 -5.50747931e-01 5.93062103e-01 3.24367344e-01 5.88213742e-01 8.83579195e-01 3.44539315e-01 6.41066432e-01 1.27302325e+00 -3.27389300e-01 7.04519272e-01 -2.04147380e-02 6.90250322e-02 4.97436911e-01 4.51352715e-01 5.01772404e-01 -5.12152791e-01 -2.11658657e-01 8.21684062e-01 -8.04132819e-02 -1.37803495e-01 -4.32808042e-01 -1.00699341e+00 9.91997838e-01 1.26837790e-01 1.75673693e-01 -2.66603231e-01 3.33862960e-01 7.80949146e-02 7.83060268e-02 1.90549895e-01 7.78322339e-01 -3.22606623e-01 -2.04067484e-01 -1.07111812e+00 7.44775057e-01 8.97802293e-01 6.80877864e-01 6.12125278e-01 -3.69440503e-02 -1.60089761e-01 6.60136521e-01 6.96275711e-01 9.12456155e-01 -1.00622125e-01 -1.15908659e+00 1.25682309e-01 2.76402235e-02 4.27547544e-01 -1.03477776e+00 -5.25002480e-01 -5.38451314e-01 -9.68180954e-01 3.19014788e-01 6.57536924e-01 -4.81846817e-02 -5.55003166e-01 1.87002325e+00 5.71327209e-01 -6.57194778e-02 -2.82659173e-01 8.21983159e-01 2.18587577e-01 6.27270877e-01 -2.45824177e-02 -4.45656240e-01 1.16359985e+00 -2.02589422e-01 -7.17591584e-01 2.77728550e-02 4.69965994e-01 -5.63644588e-01 1.05155694e+00 4.80657250e-01 -9.28568661e-01 -8.88064355e-02 -1.12232685e+00 1.12192184e-01 -1.96774662e-01 -1.75708011e-01 5.98674059e-01 8.24358344e-01 -7.38266051e-01 8.29006016e-01 -7.74642408e-01 1.10084414e-01 4.99262184e-01 5.20617664e-01 6.96760491e-02 -1.43004701e-01 -9.35013652e-01 7.68910110e-01 -1.53203264e-01 4.40185577e-01 -8.97695303e-01 -1.30120814e+00 -6.69442534e-01 1.30100334e-02 7.39486277e-01 -8.45397472e-01 9.72664773e-01 4.53794412e-02 -2.04577136e+00 1.57790959e-01 -3.57611775e-01 -2.03640431e-01 5.30460536e-01 2.08063666e-02 2.20761031e-01 -2.09676251e-01 2.48046257e-02 8.68872628e-02 7.20003247e-01 -8.80585372e-01 -8.87383372e-02 -6.34887040e-01 -6.79597035e-02 -1.62580088e-01 1.84665635e-01 -2.08307132e-01 4.34776880e-02 -4.16794926e-01 1.34033531e-01 -9.51279879e-01 -5.26449680e-01 1.57636590e-02 -4.90973830e-01 -2.42201295e-02 4.96306360e-01 -4.68875974e-01 1.16676497e+00 -1.94181383e+00 3.68537664e-01 3.36961627e-01 9.35272500e-02 -1.32781276e-02 3.54859419e-02 3.44947547e-01 2.62937218e-01 5.07060811e-02 -3.65800738e-01 -1.79995969e-01 3.55424464e-01 1.00517981e-01 -1.99630499e-01 5.55922151e-01 3.56694311e-02 8.48981678e-01 -9.30988789e-01 -4.62903291e-01 4.94999707e-01 4.92678821e-01 -8.86720657e-01 3.02619725e-01 -3.12604249e-01 8.19002450e-01 -6.94447339e-01 2.90400684e-01 8.50841999e-01 -3.14983875e-01 2.56992221e-01 -2.88465500e-01 -4.88197684e-01 5.88426776e-02 -1.43067157e+00 1.39047074e+00 -8.21688712e-01 3.25965285e-01 3.85476202e-01 -1.38377166e+00 5.52871525e-01 -1.08301759e-01 4.55476344e-01 -4.99583781e-01 1.12843283e-01 2.94077605e-01 -1.25717863e-01 -3.80923867e-01 1.43474132e-01 -5.36476672e-01 -1.12846240e-01 3.74958307e-01 -1.79113612e-01 -6.67825699e-01 4.65102382e-02 -1.39937520e-01 6.75839245e-01 4.13779348e-01 2.84303904e-01 -6.94298089e-01 3.32325399e-01 -2.69298702e-01 5.29148042e-01 1.04664469e+00 1.31899953e-01 2.26620555e-01 5.49415112e-01 7.22794458e-02 -9.11155045e-01 -1.22882080e+00 -4.62444365e-01 5.85822344e-01 3.89589705e-02 5.00989296e-02 -4.75506902e-01 -1.15834773e-01 3.04441571e-01 9.83085454e-01 -4.68566895e-01 4.12390344e-02 -1.86524570e-01 -1.20406306e+00 2.39060313e-01 2.37377733e-01 2.55859137e-01 -2.57144094e-01 -4.17394400e-01 3.03803653e-01 5.85855059e-02 -6.78766012e-01 -2.94284284e-01 8.66024867e-02 -8.69686007e-01 -7.81906605e-01 -8.49448979e-01 -6.31347820e-02 3.67110819e-01 6.55358359e-02 8.98482800e-01 -2.50197947e-01 -6.09034240e-01 1.97804913e-01 2.49226421e-01 -4.52706426e-01 -4.34848577e-01 -1.46087974e-01 1.13046110e-01 -3.03088218e-01 1.20702609e-02 -8.73980105e-01 -7.29753494e-01 6.00769699e-01 -6.68536067e-01 -9.24006999e-02 4.31105226e-01 1.04399419e+00 6.57492757e-01 9.76705328e-02 6.14580274e-01 -5.64858019e-01 4.29577589e-01 -4.40225929e-01 -1.49723101e+00 2.79310018e-01 -9.09058452e-01 5.81584871e-01 4.72251534e-01 -3.31135869e-01 -1.00242651e+00 -3.50970238e-01 -2.92494118e-01 -7.79542252e-02 2.07326293e-01 8.10072660e-01 -2.71580696e-01 -1.48603395e-01 3.29234511e-01 -1.78346723e-01 3.08998317e-01 -4.69792187e-01 2.61319757e-01 5.99385798e-01 2.42019296e-02 -8.96660030e-01 4.63243604e-01 3.38645548e-01 7.25139022e-01 -1.06411457e+00 -7.99989820e-01 -8.92604142e-02 -1.11249089e-01 -1.88156873e-01 6.79080904e-01 -5.12244344e-01 -1.48159230e+00 2.85567969e-01 -6.91298962e-01 -3.96087021e-01 -2.03923136e-01 8.24322999e-01 -7.25120485e-01 3.77647012e-01 -2.40428284e-01 -1.24714601e+00 6.53016120e-02 -1.57906914e+00 9.48775232e-01 2.10622862e-01 -2.46685445e-01 -1.09117866e+00 8.81942138e-02 1.67121708e-01 4.10842508e-01 2.95921266e-01 1.02796459e+00 -1.08199395e-01 -6.45694673e-01 -1.53412849e-01 -1.70829728e-01 6.63383752e-02 -7.31647685e-02 1.54591471e-01 -7.26313114e-01 -1.64653763e-01 2.23501921e-01 3.67629677e-02 4.77752924e-01 1.13623893e+00 1.08429480e+00 1.06082847e-02 -4.75006312e-01 6.33893251e-01 1.51234102e+00 7.46475905e-02 4.42245811e-01 -8.12043250e-02 4.70848441e-01 4.98420507e-01 5.54033160e-01 7.49874771e-01 7.93041363e-02 9.64034379e-01 1.38546363e-01 2.09713280e-01 2.80375898e-01 -3.10065120e-01 -1.22496262e-02 6.45741761e-01 -3.77667882e-02 -8.54960904e-02 -6.21670127e-01 2.51383334e-01 -1.64673066e+00 -7.82344699e-01 -3.24957132e-01 2.74254441e+00 8.54999661e-01 -1.45604923e-01 6.31354451e-02 -1.03851510e-02 5.09667099e-01 -2.41087615e-01 -7.53600299e-01 -3.12318116e-01 8.62245709e-02 3.46044302e-01 6.49967790e-01 6.17025137e-01 -6.03723764e-01 2.22542360e-01 6.87762737e+00 1.17105246e+00 -6.85844004e-01 1.20221205e-01 5.50861359e-01 -3.57419848e-02 -4.52586651e-01 1.75212845e-01 -1.06373930e+00 5.79243302e-01 1.10593414e+00 -3.70727867e-01 7.28018522e-01 5.45257270e-01 7.05030024e-01 -6.26687646e-01 -1.13645196e+00 9.06698525e-01 -5.86781681e-01 -1.31595707e+00 -4.51453954e-01 4.25201535e-01 6.07182026e-01 -2.21471548e-01 2.00670376e-01 1.94945738e-01 3.22821438e-01 -1.04661262e+00 4.58010107e-01 5.93726993e-01 5.82017183e-01 -7.32952654e-01 5.45607030e-01 2.66088963e-01 -6.52082682e-01 9.45300702e-03 -2.67820120e-01 -2.19402984e-01 3.63486350e-01 1.19624543e+00 -5.95842779e-01 5.82096219e-01 4.70451474e-01 3.48825216e-01 1.56821579e-01 1.05709994e+00 -2.32372992e-02 6.40483499e-01 -9.65124905e-01 -4.98938918e-01 -5.30342087e-02 -8.21169913e-01 6.04852438e-01 7.70468712e-01 5.12826085e-01 -1.65311664e-01 -2.07919553e-01 1.41527402e+00 3.06976289e-01 -3.93922664e-02 -5.42792201e-01 -3.35226923e-01 5.54362595e-01 7.97649682e-01 -5.86426377e-01 4.67225946e-02 -2.12095246e-01 3.24245274e-01 -9.86438096e-02 4.21211839e-01 -9.25634205e-01 -3.30720395e-01 8.81580234e-01 4.87346910e-02 4.40240234e-01 -4.90077078e-01 -4.59986538e-01 -9.07543838e-01 -2.01946899e-01 -8.70512366e-01 2.17584163e-01 -5.21325469e-01 -1.11630666e+00 -2.22216249e-01 4.93614912e-01 -8.10408533e-01 -5.24901509e-01 -8.20992053e-01 -1.38116762e-01 1.05228221e+00 -1.31728673e+00 -7.05826461e-01 2.60652661e-01 2.56696433e-01 1.52919173e-01 3.47959578e-01 6.24048412e-01 3.37880254e-01 -8.83456111e-01 4.41794366e-01 8.50752592e-01 -5.70254505e-01 2.57604480e-01 -1.07414055e+00 3.29541378e-02 5.81074655e-01 -3.58351827e-01 7.25715697e-01 1.28109157e+00 -5.75732708e-01 -2.02325773e+00 -4.07089949e-01 4.50480849e-01 -2.64166147e-01 8.76236022e-01 -3.29485446e-01 -6.53626144e-01 1.69252932e-01 -4.78451848e-01 -2.78887540e-01 5.24069667e-01 5.11223078e-01 6.80592284e-02 -1.44419566e-01 -1.02177262e+00 8.93056095e-01 8.63773465e-01 -3.71060520e-01 -1.16428398e-01 3.73425364e-01 3.11252058e-01 -1.47472590e-01 -1.25011659e+00 4.84858990e-01 6.78662419e-01 -6.86923921e-01 1.13064063e+00 -3.45333785e-01 2.22163588e-01 -4.37617332e-01 -9.71148536e-02 -1.20526254e+00 1.49227768e-01 -7.81825304e-01 2.32946202e-02 1.08109438e+00 4.19314176e-01 -8.59970152e-01 5.24068952e-01 8.33754063e-01 2.86240131e-01 -6.18275702e-01 -1.10842538e+00 -1.09596193e+00 3.27114075e-01 -7.20693290e-01 5.12824237e-01 3.89660180e-01 -9.95849594e-02 2.70570636e-01 -3.40455890e-01 8.36814716e-02 1.18239963e+00 2.49431685e-01 7.93207467e-01 -1.13746798e+00 -6.99312210e-01 -5.78209996e-01 -2.64768153e-01 -1.12340641e+00 1.17357694e-01 -4.79484141e-01 4.12388027e-01 -9.61109817e-01 2.23176807e-01 -5.35098374e-01 5.38362712e-02 -1.83629438e-01 -3.00326884e-01 -2.10974906e-02 -3.17245603e-01 -2.52020270e-01 -1.34609863e-01 8.06724906e-01 1.11956370e+00 6.95712194e-02 -2.25570977e-01 3.31844479e-01 -3.92362148e-01 3.54360253e-01 5.07423997e-01 -6.33639991e-01 -5.64484119e-01 -5.72171202e-03 6.01276755e-01 2.64757842e-01 5.67152619e-01 -7.29817331e-01 5.59644066e-02 -3.91421199e-01 1.20014191e-01 -4.90473121e-01 1.70077860e-01 -5.73808670e-01 6.03320837e-01 4.38224554e-01 -2.13727310e-01 -4.47399706e-01 2.64978439e-01 7.07634687e-01 8.80045593e-02 -4.76270944e-01 6.31896138e-01 1.30593613e-01 -4.23691757e-02 1.50587007e-01 -6.70780182e-01 1.45967454e-01 6.00124836e-01 7.10505545e-02 1.96643963e-01 -4.34111536e-01 -6.71795368e-01 2.45950937e-01 3.78617436e-01 -3.33079487e-01 4.39240992e-01 -9.53046679e-01 -5.52998900e-01 9.83972996e-02 -3.33470047e-01 -1.21380510e-02 5.65764546e-01 1.21331370e+00 -3.06554437e-01 3.23302627e-01 2.60436118e-01 -7.31431067e-01 -8.97280753e-01 6.75090909e-01 2.23061860e-01 -3.62975061e-01 -3.65145415e-01 7.01654434e-01 2.53677636e-01 -2.99920112e-01 -9.95706245e-02 -2.55191863e-01 3.65133226e-01 -1.14081204e-01 4.83677179e-01 6.83164597e-01 -4.94015180e-02 -3.97293791e-02 -2.65381366e-01 4.72947747e-01 1.62197337e-01 -4.70377982e-01 1.28814256e+00 -2.83374459e-01 -9.68353078e-02 6.19132280e-01 1.27244544e+00 -2.34389514e-01 -1.30639088e+00 -2.78860121e-03 3.33264433e-02 -4.63576883e-01 5.72661340e-01 -4.71906662e-01 -5.63193619e-01 8.93041849e-01 4.04053509e-01 7.88630825e-03 7.97297060e-01 -3.05349380e-01 3.58105004e-01 5.08033037e-01 5.11498928e-01 -1.12785828e+00 -3.74091625e-01 1.48430929e-01 7.45199025e-01 -1.05512464e+00 4.18525726e-01 -4.62462038e-01 6.22087829e-02 7.83981562e-01 -8.44132379e-02 1.73199251e-01 1.07255793e+00 5.31385779e-01 -4.73369747e-01 -1.17569854e-02 -2.92470694e-01 -1.69759970e-02 7.80478269e-02 3.57335925e-01 2.31363535e-01 -5.56774437e-02 -5.30179560e-01 4.23922151e-01 3.81247252e-02 4.64010924e-01 2.45394111e-01 8.65811408e-01 -1.78504080e-01 -1.01701987e+00 -5.05756736e-01 4.67027783e-01 -3.22803497e-01 -9.22194347e-02 3.48307997e-01 9.21462297e-01 -6.01831257e-01 9.90933955e-01 -5.31925894e-02 8.44443366e-02 1.83271065e-01 -3.77105549e-02 7.79451668e-01 -2.48119552e-02 4.82223406e-02 6.81577697e-02 -1.05859689e-01 -6.75986886e-01 -3.44315052e-01 -8.79685938e-01 -7.45466173e-01 -6.51847720e-01 -6.28024399e-01 4.23151881e-01 1.20116436e+00 1.14455438e+00 4.15835649e-01 4.77529794e-01 6.65815592e-01 -9.26293015e-01 -9.44436729e-01 -6.39199018e-01 -8.00008893e-01 -1.38576895e-01 2.14384109e-01 -1.08446801e+00 -5.32021224e-01 -5.05425513e-01]

[6.59635066986084, 3.9344887733459473]

21874581-6cff-4605-a860-94adf87e62b5

task-optimized-adapters-for-an-end-to-end

2305.02468

https://arxiv.org/abs/2305.02468v3

https://arxiv.org/pdf/2305.02468v3.pdf

Task-Optimized Adapters for an End-to-End Task-Oriented Dialogue System

Task-Oriented Dialogue (TOD) systems are designed to carry out specific tasks by tracking dialogue states and generating appropriate responses to help users achieve defined goals. Recently, end-to-end dialogue models pre-trained based on large datasets have shown promising performance in the conversational system. However, they share the same parameters to train tasks of the dialogue system (NLU, DST, NLG), so debugging each task is challenging. Also, they require a lot of effort to fine-tune large parameters to create a task-oriented chatbot, making it difficult for non-experts to handle. Therefore, we intend to train relatively lightweight and fast models compared to PLM. In this paper, we propose an End-to-end TOD system with Task-Optimized Adapters which learn independently per task, adding only small number of parameters after fixed layers of pre-trained network. We also enhance the performance of the DST and NLG modules through reinforcement learning, overcoming the learning curve that has lacked at the adapter learning and enabling the natural and consistent response generation that is appropriate for the goal. Our method is a model-agnostic approach and does not require prompt-tuning as only input data without a prompt. As results of the experiment, our method shows competitive performance on the MultiWOZ benchmark compared to the existing end-to-end models. In particular, we attain state-of-the-art performance on the DST task of 2.2 dataset.

['Myoung-Wan Koo', 'Jeehyun Lee', 'Namo Bang']

2023-05-04

['response-generation']

['natural-language-processing']

[-1.42040253e-01 4.33975726e-01 1.32511765e-01 -6.04608536e-01 -7.65116930e-01 -6.47486985e-01 7.00459957e-01 -2.20503896e-01 -3.82458895e-01 7.61099935e-01 3.40551615e-01 -3.55330884e-01 2.25637555e-01 -5.73651671e-01 -1.04555726e-01 -3.04737210e-01 2.64501512e-01 1.03170383e+00 2.84672678e-01 -9.21222925e-01 4.28454541e-02 -1.73548743e-01 -1.00836539e+00 5.68094909e-01 1.08202839e+00 8.32343280e-01 3.33712697e-01 1.00416362e+00 -4.01219606e-01 1.23547971e+00 -8.97494018e-01 -4.14726824e-01 -1.97050989e-01 -6.54171050e-01 -1.51024079e+00 9.67012048e-02 1.30362332e-01 -6.83211982e-01 -1.80608004e-01 4.18650508e-01 8.96394968e-01 3.10975760e-01 3.48494411e-01 -1.25475681e+00 -4.24230129e-01 7.07201421e-01 2.13377222e-01 -4.63335425e-01 6.50023699e-01 6.27898276e-01 1.05203509e+00 -6.44725084e-01 3.75454158e-01 1.36434770e+00 3.89848500e-01 1.31558585e+00 -1.05460560e+00 -1.99944720e-01 1.15906961e-01 -7.17595126e-03 -5.77920675e-01 -7.21355915e-01 7.10672677e-01 -2.39660889e-01 1.21459556e+00 3.47799093e-01 3.31492126e-01 1.57946861e+00 -1.78941444e-01 1.11638987e+00 9.63939250e-01 -3.43275547e-01 7.08241686e-02 3.24987292e-01 8.50915462e-02 5.98618627e-01 -7.67334878e-01 -4.28379238e-01 -4.72241879e-01 -9.27532315e-02 4.20397073e-01 -3.11260581e-01 -1.82789296e-01 -7.04232529e-02 -1.13438773e+00 8.98442507e-01 8.94409791e-02 3.19718808e-01 -2.41398305e-01 -1.12355389e-01 7.39159644e-01 6.03262424e-01 4.19010162e-01 6.94551826e-01 -7.17085540e-01 -7.70145178e-01 -2.70219684e-01 3.80632937e-01 1.29226387e+00 1.19599795e+00 3.74345273e-01 -1.90564878e-02 -8.17448795e-01 1.30051160e+00 8.48018825e-02 1.76698267e-01 5.34039319e-01 -1.06480813e+00 8.99680853e-01 8.41997206e-01 3.40942442e-01 -2.48727396e-01 -7.16221631e-01 -7.58342147e-02 -9.01134074e-01 -1.57444298e-01 8.42042446e-01 -6.84931397e-01 -3.00470829e-01 1.85638165e+00 3.62700701e-01 -2.23703548e-01 4.57794458e-01 8.20688307e-01 1.07894635e+00 7.76958108e-01 1.63256183e-01 -2.54023783e-02 1.44599485e+00 -1.48300123e+00 -9.08122420e-01 -3.72275263e-01 1.03539133e+00 -6.84937716e-01 1.68712115e+00 3.88240099e-01 -1.32652235e+00 -5.90772331e-01 -5.46317697e-01 -1.93505526e-01 -1.04437806e-02 8.57163407e-03 6.23819172e-01 3.91918808e-01 -9.42649841e-01 4.01007980e-01 -3.75773460e-01 -2.87770212e-01 -9.28228498e-02 1.92065150e-01 -1.72968015e-01 3.34087729e-01 -1.46719742e+00 1.17752326e+00 3.19690704e-01 5.32113537e-02 -9.17940497e-01 -5.31638026e-01 -7.79348850e-01 2.00933158e-01 4.51578200e-01 -6.78030074e-01 2.31272650e+00 -6.44939423e-01 -2.51611471e+00 6.35196149e-01 3.54959629e-02 -4.11404788e-01 7.18318343e-01 -2.83579469e-01 -4.51947153e-02 -1.91032998e-02 -2.02307716e-01 5.10980487e-01 4.19777721e-01 -8.00895333e-01 -6.83951557e-01 6.16952442e-02 6.53744161e-01 3.65628958e-01 -4.95888382e-01 1.94457442e-01 -2.59329230e-01 -1.48795294e-02 -5.30108929e-01 -8.07539940e-01 -1.74297988e-01 -4.55823153e-01 -3.90527248e-01 -9.31924880e-01 6.72958136e-01 -5.55701852e-01 1.25009775e+00 -1.82579136e+00 2.04445243e-01 -5.92466414e-01 2.18596905e-01 6.44061625e-01 -3.85989249e-01 9.18612361e-01 4.05815572e-01 -2.54673734e-02 -4.81438413e-02 -5.89514613e-01 2.83502489e-01 9.24953353e-03 -7.12464750e-02 -1.96464717e-01 3.67579609e-01 9.20116603e-01 -1.17289770e+00 -3.91033530e-01 1.92589939e-01 -4.15193886e-02 -6.78922415e-01 1.26080072e+00 -7.13264585e-01 7.91281521e-01 -5.28597295e-01 8.38918462e-02 2.76432723e-01 -2.52935618e-01 2.28606060e-01 1.58973381e-01 -2.99735758e-02 1.04453504e+00 -6.73990905e-01 1.91133428e+00 -1.12654626e+00 3.08559150e-01 2.39936218e-01 -8.13575387e-01 1.14788759e+00 8.55610609e-01 1.96584463e-01 -6.59368277e-01 1.89630210e-01 1.76628053e-01 3.22555602e-01 -8.18146288e-01 4.07113016e-01 1.20740831e-01 -4.71832216e-01 8.35467994e-01 4.69161004e-01 -4.16473180e-01 2.64184892e-01 2.33166233e-01 1.04463792e+00 1.00663550e-01 9.49179381e-02 1.10281393e-01 7.89887369e-01 -6.64450228e-02 2.88241953e-01 6.78948343e-01 -1.48910835e-01 3.80305350e-01 8.23979318e-01 -3.83854121e-01 -8.54104936e-01 -4.87033814e-01 2.89849579e-01 1.61187935e+00 -2.81991094e-01 -4.08418477e-01 -1.12439644e+00 -1.00089431e+00 -4.56576586e-01 9.93269980e-01 -2.82623142e-01 -2.19918311e-01 -7.08247304e-01 -4.76208538e-01 7.16656804e-01 1.84076250e-01 9.69852388e-01 -1.38796377e+00 -3.82992715e-01 6.09057903e-01 -7.13505030e-01 -1.25763083e+00 -7.68302202e-01 3.30321491e-02 -5.76140046e-01 -9.73603010e-01 -6.38580501e-01 -7.41855085e-01 2.71891713e-01 1.78916216e-01 1.30477309e+00 1.35552943e-01 3.31174731e-01 2.35978499e-01 -5.29300630e-01 -3.04555267e-01 -8.55855227e-01 6.06443882e-01 -2.05359086e-01 3.52283046e-02 2.15847239e-01 -3.60460579e-01 -4.88812000e-01 6.49095416e-01 -3.67849171e-01 4.31568503e-01 2.16441453e-01 1.11885607e+00 -3.44850272e-01 -5.55599093e-01 1.03637969e+00 -1.10840845e+00 1.38524055e+00 -3.04858357e-01 -3.62072349e-01 3.05735826e-01 -4.14160877e-01 1.43133951e-02 1.14155281e+00 -5.63956976e-01 -1.40038872e+00 -1.48493409e-01 -4.53347802e-01 9.26448777e-02 -2.19385430e-01 3.05775762e-01 -1.86384678e-01 3.07884336e-01 7.60915577e-01 2.49340221e-01 1.85623854e-01 -5.91247976e-01 3.59981656e-01 1.25151896e+00 2.54485011e-01 -8.56971681e-01 4.59480405e-01 -2.91766226e-01 -6.27024055e-01 -5.92890322e-01 -1.00631607e+00 -3.81904662e-01 -4.93900776e-01 -2.30966270e-01 7.18507528e-01 -7.95505583e-01 -1.21326256e+00 6.86232924e-01 -1.53575480e+00 -1.13339984e+00 -3.17670195e-03 1.65386930e-01 -8.30499947e-01 1.41118333e-01 -8.44872892e-01 -9.87166107e-01 -9.03404236e-01 -1.08133543e+00 8.32432866e-01 3.10998321e-01 -6.23601675e-01 -1.12502718e+00 1.58137083e-01 5.98662078e-01 8.17213237e-01 -2.23313391e-01 6.86031461e-01 -1.09159887e+00 -1.63011059e-01 2.94743967e-03 -4.96834256e-02 4.00605619e-01 1.75477758e-01 -1.93508983e-01 -1.15020347e+00 -2.05107883e-01 9.86694917e-02 -1.11028075e+00 1.90674827e-01 -2.32004076e-01 9.19416964e-01 -6.90991223e-01 2.08294258e-01 2.22431138e-01 5.52701592e-01 1.86751097e-01 3.25885773e-01 2.02491447e-01 4.73761559e-01 1.06797290e+00 7.56246269e-01 3.83662462e-01 8.81053686e-01 9.61685538e-01 2.68025994e-01 -1.24074459e-01 3.93172279e-02 -3.02784085e-01 5.25308371e-01 8.12301993e-01 4.55070920e-02 -3.97215009e-01 -7.27973938e-01 5.22451997e-01 -2.19026279e+00 -8.59032750e-01 -1.91948578e-01 2.00849104e+00 1.47922599e+00 1.42002806e-01 5.46760261e-01 -3.24516624e-01 3.21086913e-01 2.46598408e-01 -4.19125199e-01 -8.55543852e-01 3.58325481e-01 2.05514375e-02 -3.06070805e-01 7.32762754e-01 -8.10166538e-01 1.32789183e+00 5.60463381e+00 5.91443300e-01 -1.02245891e+00 2.35454425e-01 5.01583934e-01 5.29901497e-02 1.91501044e-02 -1.02430247e-01 -8.52978170e-01 3.72735232e-01 1.29049075e+00 -7.36829340e-02 5.21428645e-01 9.37582254e-01 5.55742145e-01 1.69655666e-01 -1.28960431e+00 5.44434309e-01 -3.84816051e-01 -9.34440911e-01 -2.39337653e-01 -4.37806159e-01 2.79092640e-01 -2.04395905e-01 -3.07835340e-01 9.33190167e-01 7.44320095e-01 -8.01690221e-01 4.07109380e-01 2.15260491e-01 6.51499569e-01 -4.46439207e-01 7.83523023e-01 8.75324667e-01 -6.43005669e-01 1.76612418e-02 -3.08115840e-01 -3.19913387e-01 2.48363122e-01 -2.13990156e-02 -1.43598509e+00 2.08707213e-01 2.93927044e-01 2.39366993e-01 -1.99783236e-01 6.63279831e-01 -6.04052007e-01 5.58752894e-01 5.99869825e-02 -5.15227258e-01 4.72510248e-01 -4.34698649e-02 2.28710875e-01 1.27387655e+00 -1.90973766e-02 8.13639238e-02 3.50823909e-01 7.37653613e-01 -2.36587122e-01 1.92477286e-01 -4.02880847e-01 4.74941507e-02 5.05320966e-01 1.59271610e+00 1.33319691e-01 -3.37035507e-01 -4.22993422e-01 9.73546445e-01 6.95825875e-01 1.45147651e-01 -6.47158623e-01 -5.21571159e-01 6.23512268e-01 -1.46494478e-01 -2.65162379e-01 -1.29691228e-01 -5.49378209e-02 -1.14838135e+00 -1.50356919e-01 -1.35160935e+00 2.15890869e-01 -5.72127283e-01 -1.31591690e+00 8.92069101e-01 -1.89918831e-01 -1.12220562e+00 -9.48631883e-01 -5.08271515e-01 -9.07538354e-01 1.19366908e+00 -1.40801144e+00 -1.11688602e+00 -4.10395682e-01 7.17551529e-01 1.06426537e+00 -1.96300671e-01 1.22567070e+00 7.88196847e-02 -7.13449895e-01 7.78956175e-01 -2.80061603e-01 4.09316599e-01 1.13094270e+00 -1.57638860e+00 6.42400622e-01 2.89477110e-01 -4.82595295e-01 5.55272460e-01 6.23276830e-01 -2.48633593e-01 -1.32589054e+00 -9.18517888e-01 1.20507348e+00 -5.94927251e-01 6.47256017e-01 -7.96991646e-01 -9.41686034e-01 5.60703874e-01 7.23681986e-01 -4.91425127e-01 5.47502697e-01 4.46642578e-01 -1.11780316e-01 5.54435467e-03 -1.05688012e+00 7.41305947e-01 8.75599325e-01 -4.75861102e-01 -5.40764809e-01 6.63441181e-01 9.67150867e-01 -7.94576049e-01 -1.00756669e+00 -4.24460806e-02 3.19369346e-01 -9.42851484e-01 6.54520094e-01 -9.39240634e-01 5.23503304e-01 2.04094410e-01 4.51897860e-01 -1.70042598e+00 -1.75604120e-01 -1.34155524e+00 -1.17460497e-01 1.47140574e+00 5.99259138e-01 -6.09547615e-01 5.75414419e-01 9.84294772e-01 -4.20675635e-01 -6.78723693e-01 -6.02976501e-01 -5.32510400e-01 2.08430123e-02 -1.21243939e-01 7.13224292e-01 8.30123305e-01 6.49457991e-01 1.18998754e+00 -7.68195868e-01 -3.65918338e-01 1.00552395e-01 8.34187120e-02 1.36631310e+00 -1.12881362e+00 -3.81901830e-01 -4.60867286e-01 6.69705987e-01 -1.41674912e+00 1.43653542e-01 -6.40341878e-01 2.87529707e-01 -1.51514471e+00 -1.90774620e-01 -5.99625528e-01 2.25619927e-01 8.14004600e-01 -5.26941240e-01 -4.46016043e-01 3.03734452e-01 7.59470388e-02 -8.78991663e-01 7.94080734e-01 1.63440537e+00 -2.04587936e-01 -6.08864903e-01 5.27752697e-01 -7.68439114e-01 4.87764806e-01 1.05894041e+00 -2.40778938e-01 -6.27847016e-01 -5.22401452e-01 -9.82101485e-02 5.21140814e-01 -1.05220757e-01 -6.09917462e-01 3.21496010e-01 -3.46698254e-01 -3.24075758e-01 -1.31706655e-01 4.44102705e-01 -5.06602585e-01 -5.41286588e-01 2.78648078e-01 -7.33874559e-01 -2.01275885e-01 1.49886934e-02 -5.75906895e-02 -1.80249229e-01 -5.98657131e-01 7.16328502e-01 -2.84809232e-01 -3.80089581e-01 1.70994282e-01 -5.74800611e-01 3.27895135e-01 7.07793474e-01 2.68039882e-01 -6.54499531e-01 -9.68412042e-01 -3.91171992e-01 8.28494251e-01 -4.47890125e-02 5.78082681e-01 3.64366651e-01 -1.02883613e+00 -9.53976214e-01 -1.33867696e-01 2.36112669e-01 2.78402478e-01 3.22123408e-01 6.96121156e-01 -2.36018106e-01 5.94499052e-01 -2.45183170e-01 -4.10967201e-01 -1.06339550e+00 2.74561346e-01 5.25688231e-01 -8.81716847e-01 -4.01556611e-01 7.55617976e-01 2.00436618e-02 -1.22167575e+00 3.98856640e-01 -1.29852623e-01 -4.41425651e-01 1.23781353e-01 6.16096854e-01 1.35729074e-01 1.23861074e-01 -5.93697652e-02 1.07226409e-01 1.47467072e-03 -3.17483515e-01 -1.95634604e-01 1.23566484e+00 -6.74554408e-02 3.66043858e-02 3.96658570e-01 8.77270341e-01 -3.31638038e-01 -1.34691072e+00 -4.59500074e-01 8.49089585e-03 -4.62443233e-02 -3.43735993e-01 -1.13716459e+00 -5.07879615e-01 1.16918898e+00 2.06845291e-02 6.19946718e-01 8.25400174e-01 -3.15547168e-01 1.08381438e+00 7.89317787e-01 3.81407976e-01 -1.28058696e+00 5.10601640e-01 1.07416320e+00 1.13367748e+00 -1.39472008e+00 -6.65129244e-01 -1.12513669e-01 -1.06418312e+00 1.21359718e+00 1.26393974e+00 1.80518717e-01 -8.78302902e-02 -4.02695760e-02 3.39880943e-01 -1.36928894e-02 -1.31713510e+00 -1.13286637e-01 -2.19196081e-02 5.85945487e-01 7.63409138e-01 -9.12746564e-02 -3.25680584e-01 7.13043571e-01 -2.88082272e-01 -7.23326877e-02 5.48388481e-01 6.76346898e-01 -3.56574267e-01 -1.52236557e+00 1.90586299e-02 8.68539885e-02 -2.59360105e-01 -9.38501433e-02 -6.54865980e-01 7.26515174e-01 -7.13862240e-01 1.44382906e+00 -2.89077133e-01 -4.94736701e-01 8.36968958e-01 5.29823661e-01 2.56955653e-01 -8.27071965e-01 -1.33254015e+00 -1.05192870e-01 8.29977572e-01 -5.55913448e-01 -9.01352614e-02 -1.34610683e-01 -1.09908664e+00 -4.20284361e-01 -2.45576322e-01 4.84829187e-01 2.93574303e-01 9.62042868e-01 4.10457611e-01 6.32186413e-01 1.11350334e+00 -6.69220150e-01 -1.12738121e+00 -1.73166227e+00 -5.51682152e-02 4.38143611e-01 1.65810809e-01 -2.93866515e-01 -8.85610804e-02 -2.75232792e-01]

[12.816556930541992, 8.047701835632324]

55955f85-c386-43c2-b548-ed86797911f3

meta-learning-for-code-summarization-1

2201.08310

https://arxiv.org/abs/2201.08310v1

https://arxiv.org/pdf/2201.08310v1.pdf

Meta Learning for Code Summarization

Source code summarization is the task of generating a high-level natural language description for a segment of programming language code. Current neural models for the task differ in their architecture and the aspects of code they consider. In this paper, we show that three SOTA models for code summarization work well on largely disjoint subsets of a large code-base. This complementarity motivates model combination: We propose three meta-models that select the best candidate summary for a given code segment. The two neural models improve significantly over the performance of the best individual model, obtaining an improvement of 2.1 BLEU points on a dataset of code segments where at least one of the individual models obtains a non-zero BLEU.

['Michael Pradel', 'Sebastian Padó', 'Moiz Rauf']

2022-01-20

['code-summarization']

['computer-code']

[ 3.15919340e-01 5.75180531e-01 -8.04494321e-01 -3.09689254e-01 -1.30952144e+00 -6.23213768e-01 3.34865063e-01 5.83316088e-01 4.51209173e-02 3.93534243e-01 7.88545847e-01 -4.07811821e-01 2.77215540e-01 -4.03543919e-01 -8.32596481e-01 -1.06737785e-01 -4.22045663e-02 1.27737850e-01 1.60510302e-01 -3.58716846e-01 6.79055572e-01 -1.48428068e-01 -1.54861164e+00 8.51326048e-01 1.23172235e+00 1.71556115e-01 3.17538887e-01 9.83408272e-01 -6.39233470e-01 1.07609749e+00 -7.35216737e-01 -5.40891111e-01 -1.51290774e-01 -8.38880837e-01 -1.08308077e+00 -2.56962091e-01 7.50085592e-01 1.51570458e-02 -1.79079399e-01 1.23994219e+00 2.29781076e-01 -3.36436868e-01 6.29572332e-01 -1.12646461e+00 -5.40378928e-01 1.43205285e+00 -7.14378715e-01 1.36965334e-01 3.13787818e-01 -2.05212206e-01 1.29332674e+00 -6.58973932e-01 8.24698448e-01 7.62527287e-01 8.67998242e-01 9.26636159e-01 -1.29299104e+00 -2.56597608e-01 -9.08763260e-02 -2.52749979e-01 -1.05374551e+00 -6.45121872e-01 6.79973364e-01 -7.65973091e-01 1.63887513e+00 2.23326027e-01 3.77100408e-01 7.29946375e-01 5.43773353e-01 9.49955821e-01 2.18930796e-01 -5.02351046e-01 2.47648925e-01 -9.12363231e-02 8.16683888e-01 9.24148142e-01 7.18008518e-01 -5.91511905e-01 -2.97321439e-01 -5.97653925e-01 -4.78660055e-02 -2.92005062e-01 -5.88879064e-02 -3.60608935e-01 -1.21330488e+00 9.14530039e-01 2.98083633e-01 4.63455468e-01 -9.89092588e-02 7.00441599e-01 8.39973569e-01 2.03902379e-01 2.99009353e-01 9.80535924e-01 -5.49929976e-01 -5.58031440e-01 -1.33612001e+00 5.58793068e-01 9.35179591e-01 1.58995247e+00 7.12452352e-01 2.49882966e-01 -3.60152841e-01 9.01469350e-01 1.86238900e-01 2.11192265e-01 7.10548341e-01 -1.11032093e+00 9.08183336e-01 8.37259054e-01 -8.34399313e-02 -7.43675888e-01 -3.94418687e-01 -5.58606923e-01 -4.87081051e-01 -2.46650889e-03 -1.41625673e-01 -2.45173186e-01 -7.74756432e-01 1.60202193e+00 -4.79112774e-01 -6.16569877e-01 1.28855884e-01 -1.25946209e-01 1.12260187e+00 7.74422884e-01 -7.87833631e-02 -1.60250738e-01 1.16784108e+00 -1.41052330e+00 -4.27454203e-01 -6.34370685e-01 9.25872803e-01 -7.06997871e-01 7.96225131e-01 1.14275895e-01 -1.40989780e+00 -5.06287813e-01 -1.20819354e+00 -2.18855962e-01 1.10998258e-01 3.77897203e-01 4.47185725e-01 5.85258007e-01 -1.36530077e+00 4.25847352e-01 -7.00384378e-01 -3.99062961e-01 3.49734634e-01 1.41333818e-01 -1.39255390e-01 2.04944074e-01 -3.41734946e-01 6.68188870e-01 6.59276724e-01 -6.01565599e-01 -8.56213450e-01 -6.35671020e-01 -9.74352539e-01 3.45590651e-01 1.20281972e-01 -7.88254201e-01 1.86498153e+00 -1.11612034e+00 -1.08556259e+00 8.51391971e-01 -4.98235285e-01 -5.90653718e-01 1.27501309e-01 -1.98217258e-01 -1.75797403e-01 -2.77584672e-01 2.52318025e-01 6.50811493e-01 5.22593260e-01 -1.44563663e+00 -7.20529377e-01 1.66448727e-01 3.17468554e-01 -2.49469593e-01 -1.82079002e-01 1.71033993e-01 -5.61671376e-01 -7.05680072e-01 -3.16056192e-01 -9.79893386e-01 -3.71743739e-01 -7.38887846e-01 -6.55273736e-01 -1.60080999e-01 1.24598950e-01 -7.76266456e-01 2.01788116e+00 -2.08771658e+00 4.05016899e-01 -2.94194102e-01 3.70332599e-01 5.04629873e-02 -6.54712260e-01 7.26125658e-01 -2.31869042e-01 6.54692590e-01 -6.68878913e-01 -2.94796586e-01 3.60413417e-02 -2.58854032e-01 -4.59934980e-01 1.13307394e-01 1.91034555e-01 1.00141191e+00 -8.48782539e-01 -4.60738599e-01 -6.28704727e-01 -1.45606115e-01 -1.16468751e+00 2.33143881e-01 -5.40629745e-01 -3.13988686e-01 -4.14850354e-01 4.86728251e-01 3.24244499e-01 -2.87026763e-01 -2.47685798e-02 2.04481319e-01 -2.13847890e-01 6.01873934e-01 -3.86812985e-01 2.12623763e+00 -6.64768934e-01 1.01716089e+00 -2.62096614e-01 -4.74902749e-01 9.08705711e-01 1.94401607e-01 3.51589859e-01 -2.78917253e-01 -7.96911940e-02 3.70046407e-01 3.19162369e-01 -7.73930430e-01 1.00753379e+00 3.63717884e-01 -6.24251366e-01 7.70208359e-01 1.54331908e-01 -2.77867049e-01 7.80842602e-01 4.87287879e-01 1.44057965e+00 4.18464318e-02 7.65043914e-01 -6.27347946e-01 3.84432375e-01 4.29801136e-01 4.56282198e-01 1.11194694e+00 1.63475797e-02 8.10208201e-01 1.01082993e+00 -4.61097717e-01 -1.45379674e+00 -5.88450015e-01 1.76484659e-01 1.22896874e+00 -1.58002079e-01 -1.01405430e+00 -1.21324456e+00 -9.67280269e-01 -1.18523359e-01 1.15091181e+00 -7.37262607e-01 -4.15428847e-01 -9.54614341e-01 -7.15201557e-01 9.53525543e-01 4.02260572e-01 1.75941631e-01 -9.96539652e-01 -9.24015999e-01 2.43354008e-01 -4.61562097e-01 -4.21858847e-01 -6.29012048e-01 3.64367902e-01 -9.27761078e-01 -8.59209478e-01 -4.97488022e-01 -8.96895707e-01 7.81583846e-01 2.77147502e-01 1.72054398e+00 4.70601618e-01 2.11740077e-01 3.05754244e-02 -4.47979182e-01 -4.30922866e-01 -1.38097703e+00 8.84958684e-01 -6.20987236e-01 -8.06884885e-01 2.68399596e-01 -2.49191970e-01 8.01154077e-02 -3.15284669e-01 -9.14587677e-01 2.61826545e-01 6.88883245e-01 7.06429839e-01 1.13859549e-02 -4.15438682e-01 6.79783225e-01 -1.08641756e+00 1.00922120e+00 -8.77206028e-01 -2.32508346e-01 5.03250957e-01 -5.04837334e-01 5.44698834e-01 6.21115148e-01 -7.37258047e-02 -1.04254329e+00 1.90046251e-01 -2.24367440e-01 4.34404254e-01 -4.28338274e-02 9.12441492e-01 1.89333975e-01 2.19557211e-01 1.11133218e+00 3.12313437e-01 -2.52268940e-01 -4.94164348e-01 4.83653814e-01 8.03756833e-01 5.32160044e-01 -6.64636135e-01 4.84441489e-01 -6.80151135e-02 -5.81063807e-01 -5.52440405e-01 -5.11275291e-01 -2.40452692e-01 -6.90452635e-01 1.64094985e-01 6.20139778e-01 -7.72330403e-01 1.56874031e-01 2.53319293e-01 -1.67077494e+00 -2.46268854e-01 -2.55852342e-01 1.57791730e-02 -7.78319061e-01 4.34914649e-01 -4.62773383e-01 -3.05120826e-01 -5.00999451e-01 -1.47598457e+00 1.07253861e+00 9.27492082e-02 -7.83836007e-01 -5.81003249e-01 5.31309664e-01 -1.92652345e-02 5.76760769e-01 2.15737075e-01 1.44872463e+00 -7.84594536e-01 -4.11326289e-01 -1.53871119e-01 3.05044055e-02 4.66982871e-02 9.68133733e-02 6.08642101e-01 -6.48885787e-01 -3.50862801e-01 -1.05933480e-01 -2.00072080e-01 1.16084456e+00 4.37170386e-01 1.00949645e+00 -4.53367114e-01 -3.82145047e-01 4.93801236e-01 1.58179855e+00 5.01892865e-01 4.50804710e-01 3.06853473e-01 6.46655738e-01 4.37386930e-01 6.55577183e-02 3.71222615e-01 4.08887506e-01 4.98825490e-01 6.22529089e-01 3.67157757e-01 -3.01100999e-01 -1.87057704e-01 6.83827698e-01 1.31187117e+00 3.05955619e-01 -4.06534195e-01 -1.40070486e+00 1.08114171e+00 -1.78668535e+00 -1.02431917e+00 -4.60318565e-01 1.90126061e+00 1.01414299e+00 7.59340376e-02 2.02441111e-01 -4.55299169e-01 6.51162922e-01 2.93702424e-01 -4.75211263e-01 -8.65005791e-01 4.80193309e-02 -2.75611520e-01 5.11983931e-01 4.64488566e-01 -8.43226314e-01 6.43746793e-01 7.47791529e+00 7.39973187e-01 -7.22238064e-01 3.80279124e-03 3.17864418e-01 -1.31157950e-01 -9.52210844e-01 1.32901207e-01 -7.01751232e-01 4.58249211e-01 1.29038358e+00 -8.77759516e-01 4.20602381e-01 1.15801001e+00 -1.45308495e-01 -1.32026508e-01 -1.29499269e+00 5.30510187e-01 4.45114970e-01 -1.66981566e+00 2.85188556e-01 -9.55406055e-02 1.25134802e+00 5.84260702e-01 -3.25358510e-01 6.71532154e-01 6.74793780e-01 -8.27882349e-01 1.14494944e+00 3.22096229e-01 6.23064876e-01 -8.14439952e-01 6.28454745e-01 4.58184183e-01 -9.78952944e-01 -3.83819342e-01 -2.12098956e-01 1.16685815e-01 -6.25055730e-02 3.87625366e-01 -7.04113722e-01 3.98162007e-01 3.30763459e-01 6.79674625e-01 -1.16631174e+00 1.39191723e+00 1.81325339e-02 6.66746795e-01 2.96987265e-01 -2.39799231e-01 2.84596384e-01 5.15853047e-01 6.48038864e-01 1.78596854e+00 4.73674983e-01 -4.54019666e-01 -9.14127603e-02 1.15502238e+00 -4.22658741e-01 1.07646435e-01 -7.93482900e-01 -1.74180165e-01 4.92104799e-01 9.03921962e-01 -6.01451039e-01 -6.64289176e-01 -5.58201551e-01 6.58335149e-01 4.29512352e-01 1.52697265e-01 -8.20434868e-01 -9.61984873e-01 4.79215413e-01 -3.56277883e-01 2.56804496e-01 3.44218221e-03 -5.98955393e-01 -1.27680278e+00 1.85884550e-01 -1.14542294e+00 1.47156298e-01 -5.81004977e-01 -6.56515718e-01 9.63418543e-01 1.40426561e-01 -1.12875235e+00 -6.64056182e-01 -7.94965699e-02 -8.82048368e-01 5.63182175e-01 -1.12075233e+00 -8.70051622e-01 -3.53242189e-01 -2.92409450e-01 1.05170584e+00 -4.50643748e-01 6.29393160e-01 1.34484604e-01 -5.24518073e-01 5.98299563e-01 3.87299389e-01 2.02779830e-01 4.38010842e-01 -1.46977103e+00 1.33037400e+00 1.31463981e+00 1.67747922e-02 1.23019648e+00 8.98562968e-01 -7.31567144e-01 -1.08216119e+00 -1.22668624e+00 1.26313436e+00 -5.94211996e-01 4.17848468e-01 -1.72617912e-01 -8.64293218e-01 8.31608176e-01 7.47561157e-01 -6.09499574e-01 5.55334032e-01 -1.26999170e-02 -5.56845903e-01 3.31309646e-01 -8.20160627e-01 5.77218592e-01 8.54610860e-01 -6.95305943e-01 -8.76753151e-01 4.43523340e-02 1.03600752e+00 -4.05836046e-01 -6.62349999e-01 1.52168557e-01 4.82233256e-01 -1.17713416e+00 4.88097399e-01 -8.21244538e-01 1.28518093e+00 -2.78676152e-01 -8.18165466e-02 -1.45766163e+00 -4.49992537e-01 -6.30957186e-01 -2.12269783e-01 1.34226978e+00 6.42042518e-01 8.21999609e-02 5.69055200e-01 3.66622120e-01 -6.54240191e-01 -5.83950162e-01 -5.33056021e-01 -6.96587324e-01 4.97487187e-01 -2.63337135e-01 7.20116615e-01 7.18124211e-01 6.05135381e-01 4.41899002e-01 -7.35065117e-02 -3.68259013e-01 3.53759259e-01 2.42005587e-01 9.23261225e-01 -1.08571482e+00 -2.81561822e-01 -1.05342519e+00 -3.53900343e-02 -1.13064194e+00 4.79207516e-01 -1.23958504e+00 3.67636085e-01 -1.84214997e+00 8.74642491e-01 1.63134158e-01 1.72317401e-01 4.43342090e-01 -2.42858201e-01 -3.34185660e-01 1.21707782e-01 2.60131836e-01 -6.60538375e-01 2.06466153e-01 4.09787655e-01 -5.31975746e-01 -1.95967838e-01 -5.64098209e-02 -1.20627046e+00 6.11576378e-01 9.78902698e-01 -9.00769472e-01 -5.13067245e-01 -9.95470166e-01 6.58185422e-01 2.49699250e-01 -1.69037595e-01 -1.04676008e+00 1.70665339e-01 -9.98953879e-02 -5.04306495e-01 -5.73397219e-01 -5.23177266e-01 -4.32181835e-01 1.63239077e-01 8.34770918e-01 -1.07017303e+00 4.48095530e-01 4.63068336e-01 2.69224733e-01 -3.10780168e-01 -1.04148722e+00 8.13365817e-01 -3.69196743e-01 -7.39580452e-01 -1.51981086e-01 -7.10796118e-01 4.31503475e-01 8.53082895e-01 -1.30282879e-01 -7.60951877e-01 -1.68707684e-01 -7.61988945e-03 -4.30654734e-02 9.22064900e-01 8.99218440e-01 3.24528694e-01 -1.19453466e+00 -1.01860189e+00 3.78814787e-02 7.54493833e-01 -2.15721592e-01 -9.54252481e-03 4.70936656e-01 -7.38409758e-01 5.54889619e-01 -3.87598574e-01 -2.57686973e-01 -1.20898616e+00 4.45242882e-01 3.90025139e-01 -2.68941879e-01 -2.90176213e-01 6.98978007e-01 2.50611976e-02 -4.27614152e-01 -1.10320583e-01 -6.98622227e-01 -1.63609445e-01 -7.11378222e-03 6.60543859e-01 4.74581510e-01 2.49393567e-01 -7.36884534e-01 -2.30910912e-01 4.70326543e-01 -2.89634824e-01 1.15827113e-01 1.36940718e+00 4.71765436e-02 -6.27053678e-01 5.36062539e-01 1.57799804e+00 3.66688579e-01 -8.66302311e-01 -1.95452236e-02 5.42222261e-01 -1.61274761e-01 -2.47790024e-01 -7.10653961e-01 -8.50256264e-01 6.66409910e-01 -1.15846843e-01 3.63395661e-01 7.75390506e-01 3.09346057e-02 5.18616915e-01 6.99838221e-01 3.18141013e-01 -7.94666529e-01 -5.70874363e-02 8.79230261e-01 8.77664447e-01 -9.72612143e-01 -3.60216689e-03 1.02947384e-01 -5.88494718e-01 1.39328945e+00 6.82317913e-01 -1.28823251e-01 1.64065957e-01 3.15026343e-01 -1.39652908e-01 -3.57921422e-01 -1.13375878e+00 8.30524340e-02 2.58691251e-01 4.16407049e-01 9.31195676e-01 -1.80950582e-01 -4.86401916e-01 5.23601651e-01 -2.67769933e-01 -1.20680809e-01 1.30925560e+00 1.00509632e+00 -8.28809023e-01 -8.73018622e-01 6.47221282e-02 7.31182635e-01 -5.25665939e-01 -4.02498424e-01 -6.52300715e-01 6.86768889e-01 -1.75558746e-01 8.78802836e-01 -7.19298273e-02 -5.10536611e-01 3.78803670e-01 6.07923865e-02 8.71219486e-02 -1.13473797e+00 -1.13744807e+00 -3.30196470e-01 2.64919519e-01 -4.91893977e-01 -2.72653580e-01 -7.03910947e-01 -1.31395638e+00 -3.07856917e-01 -2.18181372e-01 3.97180974e-01 5.47132730e-01 5.69918752e-01 5.50214350e-01 6.31959677e-01 4.22107399e-01 -6.31860197e-01 -4.57956940e-01 -9.05160487e-01 -8.86680633e-02 1.42681941e-01 7.88127244e-01 -1.65564343e-02 -1.37889102e-01 4.60294187e-01]

[7.607944011688232, 7.954588890075684]

fa48d6e3-4beb-45b5-b8db-8516fdb2f36b

uncertainty-informed-optimal-resource

2307.00032

https://arxiv.org/abs/2307.00032v1

https://arxiv.org/pdf/2307.00032v1.pdf

Uncertainty Informed Optimal Resource Allocation with Gaussian Process based Bayesian Inference

We focus on the problem of uncertainty informed allocation of medical resources (vaccines) to heterogeneous populations for managing epidemic spread. We tackle two related questions: (1) For a compartmental ordinary differential equation (ODE) model of epidemic spread, how can we estimate and integrate parameter uncertainty into resource allocation decisions? (2) How can we computationally handle both nonlinear ODE constraints and parameter uncertainties for a generic stochastic optimization problem for resource allocation? To the best of our knowledge current literature does not fully resolve these questions. Here, we develop a data-driven approach to represent parameter uncertainty accurately and tractably in a novel stochastic optimization problem formulation. We first generate a tractable scenario set by estimating the distribution on ODE model parameters using Bayesian inference with Gaussian processes. Next, we develop a parallelized solution algorithm that accounts for scenario-dependent nonlinear ODE constraints. Our scenario-set generation procedure and solution approach are flexible in that they can handle any compartmental epidemiological ODE model. Our computational experiments on two different non-linear ODE models (SEIR and SEPIHR) indicate that accounting for uncertainty in key epidemiological parameters can improve the efficacy of time-critical allocation decisions by 4-8%. This improvement can be attributed to data-driven and optimal (strategic) nature of vaccine allocations, especially in the early stages of the epidemic when the allocation strategy can crucially impact the long-term trajectory of the disease.

['Saurabh Amin', 'Samarth Gupta']

2023-06-30

['bayesian-inference', 'stochastic-optimization', 'gaussian-processes']

['methodology', 'methodology', 'methodology']

[ 1.38736680e-01 -8.04349110e-02 -1.15920693e-01 1.75806969e-01 -5.26632667e-01 -6.04327977e-01 3.94820571e-01 3.25942606e-01 -5.39200544e-01 1.09166348e+00 2.59562969e-01 -7.14914203e-01 -8.81660700e-01 -6.19173527e-01 -4.02028233e-01 -5.98478675e-01 -4.20334041e-01 1.20800805e+00 4.24087569e-02 -2.14555070e-01 -1.44706830e-01 5.61435997e-01 -8.50158036e-01 -3.86522800e-01 1.10855567e+00 5.45493484e-01 2.76219379e-02 7.70943284e-01 -7.68840953e-04 2.63727099e-01 -6.77423537e-01 -1.60438016e-01 2.59411573e-01 -2.41109893e-01 -2.88057566e-01 -2.05908477e-01 -1.01762199e+00 -4.90691364e-01 1.09493189e-01 6.59577191e-01 6.34456515e-01 -1.69554636e-01 1.31840599e+00 -1.22701967e+00 -2.75571793e-01 3.84243667e-01 -4.89583999e-01 3.43345135e-01 9.16922539e-02 3.85792106e-01 7.01924413e-02 -8.96131024e-02 6.51634216e-01 1.18214023e+00 8.12948942e-01 4.46043462e-01 -1.31046891e+00 -2.66390264e-01 2.80572474e-01 -5.53249538e-01 -1.56040716e+00 -2.63461590e-01 1.74930692e-01 -8.06719720e-01 8.26281130e-01 2.99600482e-01 7.16320276e-01 8.64219189e-01 6.97618604e-01 1.64942116e-01 1.17723238e+00 2.26930529e-02 4.91195232e-01 6.94490820e-02 -2.65342407e-02 9.87076610e-02 6.41877711e-01 2.32609734e-01 1.75102532e-01 -8.46352875e-01 8.66649568e-01 1.48341998e-01 -1.01494044e-01 -9.33440700e-02 -7.75361538e-01 1.02802825e+00 -1.35637417e-01 -9.77121517e-02 -1.04712033e+00 1.44485384e-01 -2.06102803e-02 1.87508211e-01 8.70924950e-01 5.03527761e-01 -6.66779578e-01 1.03501923e-01 -7.66514480e-01 6.20409369e-01 1.09688377e+00 5.74605763e-01 -7.76121318e-02 7.06715286e-02 -4.59573507e-01 4.96124476e-01 5.23252368e-01 1.20037997e+00 -4.86412942e-01 -9.27154660e-01 1.60674468e-01 -5.24094999e-02 8.35927010e-01 -5.60309291e-01 -6.56320333e-01 -4.20048714e-01 -7.54062891e-01 -1.08592279e-01 6.66710734e-01 -1.02366769e+00 -9.40163255e-01 1.88329422e+00 5.20503223e-01 8.44752267e-02 -4.68868911e-02 4.94780451e-01 1.93418115e-01 8.42612088e-01 4.58720833e-01 -9.15645897e-01 1.53900158e+00 1.02335609e-01 -8.74316037e-01 1.14309020e-01 3.10954064e-01 -6.13565683e-01 1.22373462e-01 2.63257138e-02 -1.21733749e+00 4.66119915e-01 -4.61784005e-01 8.65090072e-01 -3.18160176e-01 -4.12884325e-01 3.63951653e-01 6.59269750e-01 -1.14314711e+00 2.00883031e-01 -7.44695127e-01 -3.78845215e-01 2.33930632e-01 5.25787532e-01 2.86602855e-01 -5.67620108e-03 -1.60541201e+00 1.06178725e+00 -6.96116313e-02 2.28796989e-01 -8.39766860e-01 -1.12161589e+00 -5.74905872e-01 2.35473253e-02 6.57198012e-01 -1.21751511e+00 9.39303160e-01 -1.96918890e-01 -1.34787714e+00 2.33473808e-01 -1.62474304e-01 -4.68387276e-01 6.84077084e-01 3.57626200e-01 -1.12777337e-01 1.56601872e-02 -9.21764150e-02 1.69320390e-01 5.76090574e-01 -1.29713476e+00 -9.40532237e-02 -3.15239191e-01 -1.31216392e-01 3.57090712e-01 2.35884279e-01 6.79168880e-01 -2.24752501e-01 -7.22224236e-01 -4.87956047e-01 -1.04491413e+00 -8.01879108e-01 -2.03218594e-01 -1.30085111e-01 2.77350307e-01 1.23272203e-01 -1.09135747e+00 1.31200981e+00 -1.54951465e+00 3.94608706e-01 4.05316055e-01 1.03370026e-01 1.41277879e-01 3.60037126e-02 8.12148392e-01 5.07758677e-01 3.44824880e-01 -6.97586894e-01 -4.01389822e-02 -3.74043956e-02 2.70435780e-01 -2.32904583e-01 4.92081225e-01 3.06583256e-01 8.48568797e-01 -6.41655803e-01 -2.49805301e-01 8.24920461e-02 7.06080616e-01 -6.05913937e-01 1.84079111e-01 -5.21801472e-01 6.99278355e-01 -7.64599860e-01 4.55401391e-01 8.06494415e-01 -2.07432732e-01 3.22732478e-01 3.30969155e-01 -1.06974542e-01 -4.21525866e-01 -1.20431066e+00 4.81430978e-01 -3.88011485e-01 1.13969697e-02 5.74083149e-01 -6.80175424e-01 2.96331763e-01 3.48444790e-01 8.77395034e-01 -5.11284992e-02 3.24210882e-01 1.51824921e-01 2.16307521e-01 -3.61388177e-01 2.36104578e-01 -4.27382439e-01 -2.59853244e-01 7.77108550e-01 -4.43723440e-01 -2.77391404e-01 -8.24918970e-03 7.34302998e-02 1.09942555e+00 -2.98056394e-01 3.88964742e-01 -7.49678075e-01 2.73044296e-02 3.24586987e-01 5.35283804e-01 7.44310081e-01 -2.27647811e-01 1.61080495e-01 9.02571201e-01 -2.28172075e-02 -9.96148884e-01 -1.13864636e+00 -4.64339495e-01 6.46490216e-01 -8.36740136e-02 2.41480812e-01 -8.26198995e-01 -4.54862304e-02 3.14819187e-01 6.42723918e-01 -7.51391470e-01 2.02792063e-01 -5.07245719e-01 -1.68994069e+00 2.38721296e-01 2.02244833e-01 -1.11327529e-01 -6.21174395e-01 -5.47771215e-01 5.29346943e-01 1.87042311e-01 -7.71206617e-01 -4.55483168e-01 1.19233038e-02 -6.13072455e-01 -1.16318047e+00 -1.33385730e+00 4.36170176e-02 8.31019223e-01 -3.01379204e-01 7.73243546e-01 -1.42158523e-01 -3.26235145e-01 7.81349719e-01 1.22906186e-01 -7.28721917e-01 -6.33323014e-01 -2.66670823e-01 2.15988904e-01 -3.65802079e-01 7.54406005e-02 6.98153228e-02 -7.64546156e-01 4.24081594e-01 -1.06107199e+00 -4.69217926e-01 1.51867419e-01 5.21004677e-01 4.75903898e-01 9.09083560e-02 8.95929754e-01 -8.20647001e-01 1.27536738e+00 -1.04691410e+00 -8.30200851e-01 5.22519410e-01 -4.05386329e-01 2.74693161e-01 8.78498852e-02 -5.21105349e-01 -1.22242236e+00 -3.25191766e-01 4.40151151e-03 3.14531475e-02 2.01392189e-01 8.01226616e-01 2.04168037e-01 1.61176085e-01 1.49747044e-01 -8.10225531e-02 4.76213396e-01 -3.09362769e-01 1.89790472e-01 8.10640633e-01 -7.97653198e-02 -8.32336009e-01 4.31552678e-01 5.07556856e-01 2.35877797e-01 -8.15607131e-01 -5.46195619e-02 -1.43049225e-01 1.32847503e-02 -1.03919290e-01 8.90065074e-01 -8.71969700e-01 -9.70497787e-01 8.12015951e-01 -9.70615864e-01 -8.27525616e-01 -2.00811207e-01 5.56835711e-01 -7.32454777e-01 2.73730084e-02 -5.62841356e-01 -1.45260656e+00 -2.02827945e-01 -1.35163581e+00 9.19785380e-01 -9.74923372e-02 -2.23927677e-01 -1.51098156e+00 4.93655294e-01 5.35440771e-03 1.10175657e+00 5.81939042e-01 8.79140079e-01 -5.11980414e-01 -3.89652967e-01 1.08048156e-01 9.01677236e-02 -3.70524108e-01 -8.23026057e-04 1.79777071e-01 -2.94884890e-01 -3.03333044e-01 2.42804244e-01 3.04276288e-01 5.50142109e-01 1.22649562e+00 4.63578790e-01 -5.46665430e-01 -5.91540217e-01 3.19404364e-01 1.30434847e+00 3.28607827e-01 2.06622377e-01 -1.71239689e-01 -3.96135040e-02 1.11109185e+00 4.26670074e-01 8.19059551e-01 6.46215975e-01 7.82991648e-01 9.83080268e-02 1.64936408e-01 4.35761005e-01 6.53643906e-02 1.54685110e-01 4.33640242e-01 -1.06631510e-01 -6.78139448e-01 -1.22596216e+00 5.75341344e-01 -1.74937057e+00 -7.00915933e-01 -1.02628946e-01 2.40050077e+00 8.94678056e-01 -1.89944416e-01 5.42962313e-01 -6.53101146e-01 7.72402227e-01 -1.53622776e-01 -6.13621771e-01 -5.17283261e-01 -1.02078868e-02 -2.33760715e-01 9.50573027e-01 9.19911802e-01 -7.54628658e-01 4.05872077e-01 7.68884039e+00 6.27042651e-01 -6.88423634e-01 1.59996822e-01 9.41604018e-01 -2.70963430e-01 -5.21883607e-01 -2.83078551e-01 -8.64812613e-01 6.87196195e-01 1.33677804e+00 -4.18938220e-01 4.61904883e-01 -9.88250822e-02 7.21886098e-01 -2.75434643e-01 -5.72295964e-01 1.68092579e-01 -3.29264939e-01 -1.22270858e+00 -3.54630262e-01 4.28039432e-01 9.54207659e-01 -7.02636093e-02 3.12973440e-01 6.33561909e-02 1.00704205e+00 -1.08214831e+00 1.09926105e-01 9.34918582e-01 7.02468395e-01 -8.31900358e-01 8.85126233e-01 3.98730040e-01 -8.22992444e-01 -4.18935902e-03 3.06078047e-02 1.96144879e-01 7.74010777e-01 6.59410179e-01 -5.48224092e-01 2.69072711e-01 3.16948235e-01 -1.12532169e-01 1.36271417e-01 1.16122377e+00 2.43210524e-01 5.53373396e-01 -8.36225092e-01 6.90132156e-02 9.03866887e-02 -2.86050200e-01 8.02060127e-01 8.35441113e-01 3.84390354e-01 4.90279496e-01 -2.36870766e-01 8.39753091e-01 4.07695860e-01 -9.36104953e-02 -5.70096672e-01 -2.14567333e-01 6.20162487e-01 4.91380870e-01 -8.19573820e-01 1.39987897e-02 1.65099561e-01 3.80264252e-01 -2.07738653e-01 8.35767984e-01 -7.32022822e-01 1.64583344e-02 8.53454232e-01 2.39583448e-01 2.44449824e-01 -1.59199804e-01 -7.48129860e-02 -9.30285573e-01 -3.69305193e-01 -7.63416529e-01 4.26133901e-01 -2.87199348e-01 -1.27399015e+00 3.15194696e-01 8.51663470e-01 -3.44569892e-01 -8.18297327e-01 -5.48358083e-01 -6.14113629e-01 1.29917049e+00 -1.27984810e+00 -5.77289581e-01 6.39584541e-01 2.49753103e-01 1.41325876e-01 1.64658621e-01 6.41713977e-01 -9.40556675e-02 -7.17402875e-01 6.76786751e-02 8.26440036e-01 -5.66035032e-01 2.95182377e-01 -9.59473550e-01 1.83207795e-01 5.04949570e-01 -1.04580128e+00 8.43953311e-01 1.11340010e+00 -1.28641915e+00 -1.42682934e+00 -9.16016400e-01 6.50028825e-01 -7.47121036e-01 7.82756627e-01 -2.85159051e-01 -7.41943121e-01 3.19367617e-01 5.25937304e-02 -5.18568277e-01 6.48544610e-01 -1.92960784e-01 1.65528610e-01 3.83542567e-01 -1.70632064e+00 7.32395530e-01 5.85807145e-01 -6.80642053e-02 -1.61343619e-01 5.68982005e-01 8.18752766e-01 -1.72928035e-01 -1.17382312e+00 5.10122478e-01 4.18444663e-01 -1.06297918e-01 1.08653212e+00 -8.46663535e-01 -1.21731736e-01 -7.68475011e-02 6.59848703e-03 -1.51171327e+00 3.26587223e-02 -1.07668447e+00 -2.48637706e-01 8.98128569e-01 3.58971149e-01 -1.05370414e+00 1.91789970e-01 9.88893151e-01 6.35037482e-01 -1.05500102e+00 -9.66462612e-01 -8.51107359e-01 5.89447021e-01 -2.65644819e-01 7.85297036e-01 4.67246115e-01 -4.62856650e-01 -4.59296554e-01 -4.98626888e-01 3.16342592e-01 8.98698688e-01 -4.16725338e-01 2.10732713e-01 -1.09497201e+00 -4.13157701e-01 -3.03422123e-01 1.68648154e-01 -3.90788794e-01 -7.99735263e-02 -1.81939915e-01 1.22208133e-01 -1.58645189e+00 4.08366829e-01 -7.12645888e-01 1.55835617e-02 -1.97936292e-03 -4.46075559e-01 -3.08847696e-01 1.49958387e-01 1.40686303e-01 -1.90028504e-01 4.61001545e-01 7.64423430e-01 1.78938866e-01 -4.52066928e-01 3.17497164e-01 -6.97532415e-01 4.58246768e-01 7.97966957e-01 -6.62081420e-01 -5.63917994e-01 -1.86679274e-01 3.70897144e-01 6.34322166e-01 2.99083143e-01 -1.81067228e-01 1.08351365e-01 -8.49139869e-01 -7.60979354e-02 -4.45997655e-01 1.88337162e-01 -8.00883353e-01 9.58876550e-01 9.47815180e-01 -9.09902975e-02 9.83507857e-02 6.17865860e-01 8.83715570e-01 5.04956186e-01 -1.69570759e-01 4.74440873e-01 4.09672856e-02 2.26658970e-01 4.03626204e-01 -1.08158875e+00 4.30360645e-01 1.26440537e+00 1.82801008e-01 -5.87329686e-01 -6.44895971e-01 -7.75255620e-01 6.72951996e-01 4.25386548e-01 7.93027878e-02 1.41795725e-01 -7.29196191e-01 -1.06421018e+00 -1.77475423e-01 -4.37599361e-01 -3.26008081e-01 4.84200597e-01 9.83925760e-01 -6.30300105e-01 7.18910992e-01 6.86044618e-02 -3.37644637e-01 -7.88519800e-01 7.02976108e-01 5.99829435e-01 -6.20792389e-01 1.53575286e-01 4.72897649e-01 2.50758469e-01 -4.63398814e-01 -1.32963076e-01 5.74807674e-02 -5.05792908e-02 1.75732136e-01 4.68216807e-01 6.42742336e-01 -5.86521447e-01 -6.41604245e-01 -6.03277802e-01 5.63470900e-01 1.13525376e-01 -3.90463293e-01 1.37879050e+00 -3.12579066e-01 -8.14130604e-02 1.08411431e-01 6.07038081e-01 -1.07285284e-01 -1.48004007e+00 2.42401450e-03 -2.63975888e-01 1.42748132e-01 2.02265941e-02 -9.82271135e-01 -6.95819914e-01 4.50474769e-01 4.70681518e-01 3.02285671e-01 9.14335072e-01 -7.93365091e-02 2.66267598e-01 -3.66495666e-03 2.43172050e-01 -7.10334599e-01 -6.15550280e-01 3.41851026e-01 9.74339962e-01 -9.01097655e-01 1.71184123e-01 -3.79541367e-01 -6.87959433e-01 4.35770333e-01 -7.07312152e-02 1.20342791e-01 1.22038555e+00 7.35534132e-01 -1.62302673e-01 -2.95841932e-01 -8.93311858e-01 -2.29465693e-01 1.94583535e-01 9.14892733e-01 -1.14244364e-01 2.69419104e-01 -6.78505957e-01 6.77282035e-01 4.89551991e-01 2.07316026e-01 3.83292407e-01 8.05836320e-01 -3.69004399e-01 -1.03177989e+00 -7.46703327e-01 6.66171014e-01 -6.59151375e-01 -2.56273091e-01 -2.73528665e-01 7.52653122e-01 -1.19399615e-01 1.01644409e+00 1.60305440e-01 4.26528007e-01 5.47735170e-02 -1.14448957e-01 2.46984452e-01 -3.53410870e-01 -6.00692391e-01 2.22793654e-01 1.70611426e-01 -6.36127144e-02 -1.36764422e-01 -8.40946019e-01 -9.28023934e-01 -6.03144526e-01 -2.13707805e-01 2.54184961e-01 7.28979945e-01 9.22770202e-01 6.30957246e-01 5.02132058e-01 3.54103446e-01 -5.67150056e-01 -9.44046617e-01 -5.25891840e-01 -5.79134583e-01 -3.53497893e-01 1.79230645e-01 -7.67054856e-01 -5.59221804e-01 -4.84673709e-01]

[6.0094828605651855, 4.371821880340576]

362fbdb6-f32d-4e79-b675-6bddc4a8d553

triggering-dark-showers-with-conditional-dual

2306.12955

https://arxiv.org/abs/2306.12955v1

https://arxiv.org/pdf/2306.12955v1.pdf

Triggering Dark Showers with Conditional Dual Auto-Encoders

Auto-encoders (AEs) have the potential to be effective and generic tools for new physics searches at colliders, requiring little to no model-dependent assumptions. New hypothetical physics signals can be considered anomalies that deviate from the well-known background processes generally expected to describe the whole dataset. We present a search formulated as an anomaly detection (AD) problem, using an AE to define a criterion to decide about the physics nature of an event. In this work, we perform an AD search for manifestations of a dark version of strong force using raw detector images, which are large and very sparse, without leveraging any physics-based pre-processing or assumption on the signals. We propose a dual-encoder design which can learn a compact latent space through conditioning. In the context of multiple AD metrics, we present a clear improvement over competitive baselines and prior approaches. It is the first time that an AE is shown to exhibit excellent discrimination against multiple dark shower models, illustrating the suitability of this method as a performant, model-independent algorithm to deploy, e.g., in the trigger stage of LHC experiments such as ATLAS and CMS.

['Maurizio Pierini', 'Nadezda Chernyavskaya', 'Benedikt Maier', 'Simranjit Singh Chhibra', 'Luca Anzalone']

2023-06-22

['anomaly-detection']

['methodology']

[ 3.46167475e-01 1.20718971e-01 4.09455709e-02 -7.25810111e-01 -9.19957995e-01 -4.46356207e-01 1.34346437e+00 4.31347609e-01 -4.03494745e-01 3.49299431e-01 1.67828068e-01 -6.32298291e-01 -7.09245130e-02 -5.65559626e-01 -8.99316549e-01 -8.17466974e-01 -8.87860656e-02 1.00825489e+00 1.45438641e-01 -2.71870643e-01 1.62278742e-01 9.07435954e-01 -1.21153378e+00 2.04925999e-01 1.43386677e-01 1.16161299e+00 -8.46803710e-02 6.15428150e-01 2.66202569e-01 7.05945432e-01 -4.60989118e-01 -5.00301957e-01 2.86444366e-01 -4.94741976e-01 -5.04523575e-01 -2.65382499e-01 3.09404641e-01 -6.09193929e-02 -8.41272414e-01 1.13576031e+00 5.33965707e-01 -4.20195088e-02 6.26177013e-01 -1.05281246e+00 -1.16047084e-01 4.42474633e-01 -3.49106431e-01 9.00011897e-01 4.50990908e-02 3.53818357e-01 9.99482870e-01 -8.12992156e-01 6.73135161e-01 1.13250184e+00 4.75958914e-01 2.31901497e-01 -1.48020005e+00 -6.63058639e-01 -2.45582610e-01 3.06991071e-01 -9.42054331e-01 -5.36130548e-01 7.39571154e-01 -5.02110898e-01 9.80362773e-01 2.76328027e-01 4.03206378e-01 1.72756851e+00 4.77255732e-01 5.75742602e-01 1.09980643e+00 -5.62308609e-01 5.18800914e-01 1.40946776e-01 -1.24261202e-02 4.45091695e-01 2.60803849e-01 7.59424686e-01 -7.76282191e-01 -5.09232342e-01 4.80572790e-01 -3.88480425e-01 -1.87323522e-03 -6.61293566e-01 -1.30347753e+00 1.06828189e+00 2.94163585e-01 3.98224145e-01 -2.27657378e-01 1.18955314e-01 7.01171041e-01 3.33474100e-01 5.09703100e-01 7.21735120e-01 -5.43003023e-01 -1.32710963e-01 -1.15101123e+00 4.35455561e-01 5.92398167e-01 5.57574034e-01 3.03465933e-01 3.16930652e-01 -3.75337839e-01 2.27469489e-01 2.74500489e-01 5.78522861e-01 4.01798248e-01 -4.16626334e-01 1.41326487e-01 1.55620471e-01 -5.16896658e-02 -5.97988129e-01 -5.07133305e-01 -9.32662308e-01 -5.38725436e-01 2.62129813e-01 6.21760525e-02 -9.82135087e-02 -1.00068760e+00 1.75603688e+00 4.74504530e-01 2.91796654e-01 -5.27793281e-02 5.29836357e-01 3.81089121e-01 4.77608293e-01 5.66017181e-02 -1.48988411e-01 1.12629998e+00 -3.35949719e-01 -4.50798422e-01 -4.73345250e-01 5.48380852e-01 -7.64205992e-01 2.71210939e-01 5.99886417e-01 -8.71252000e-01 -3.95845622e-01 -1.40196955e+00 6.82194382e-02 -1.46454722e-01 -1.41034722e-01 1.22883487e+00 6.89651132e-01 -6.21412158e-01 7.28682876e-01 -1.16444504e+00 -2.28787199e-01 -3.75983231e-02 1.05702825e-01 -2.01354310e-01 3.31048638e-01 -1.06353521e+00 1.02941895e+00 5.33179641e-01 -7.43122175e-02 -1.32916594e+00 -8.42534542e-01 -7.61470675e-01 4.40193973e-02 1.65389657e-01 -4.42674816e-01 1.47214401e+00 -4.93199736e-01 -1.13369012e+00 1.02625120e+00 -6.07463010e-02 -1.15132034e+00 3.55016023e-01 -2.23133825e-02 -1.21811962e+00 1.22557148e-01 3.17366421e-02 2.92001646e-02 9.88974214e-01 -8.43474150e-01 -3.78029853e-01 -2.00553566e-01 -1.73169896e-01 -2.98245490e-01 3.33381444e-01 6.42620206e-01 -2.71381885e-01 -6.81176245e-01 4.21008348e-01 -9.59126413e-01 -4.69400734e-01 -2.96975762e-01 -5.05705416e-01 8.08424875e-02 5.09097278e-01 -4.86264855e-01 9.34499264e-01 -2.28001142e+00 1.28091097e-01 5.91026545e-01 3.74986589e-01 1.34663749e-03 3.98948133e-01 4.38270241e-01 -4.98437017e-01 -3.49156737e-01 -3.67625266e-01 -5.79310432e-02 3.68779689e-01 1.83623001e-01 -5.34448445e-01 6.23634756e-01 3.61782163e-01 7.10329711e-01 -7.89191663e-01 1.63802374e-02 3.51337433e-01 1.98237196e-01 -5.27499735e-01 6.46141395e-02 -4.06359673e-01 8.59536529e-01 -4.78459924e-01 4.35330719e-01 5.91772318e-01 -1.73457921e-01 -1.21049225e-01 -1.82822585e-01 -1.59897998e-01 5.35592377e-01 -8.96066487e-01 1.69583333e+00 -2.32157409e-01 5.15549600e-01 1.95937261e-01 -1.35745406e+00 7.19264984e-01 2.01877743e-01 5.42592824e-01 -9.35820878e-01 1.41863927e-01 4.39391136e-01 4.81445372e-01 1.88037828e-02 3.04859370e-01 -4.87136096e-01 -4.26048279e-01 3.23933870e-01 6.59154356e-01 -1.31516665e-01 1.34225905e-01 5.17241061e-01 1.60910618e+00 -2.17923135e-01 2.04316795e-01 -3.34525764e-01 3.99305791e-01 -1.88303202e-01 5.48027217e-01 1.38547111e+00 5.27780131e-02 3.75829369e-01 3.94146204e-01 -5.71686983e-01 -1.27802241e+00 -1.32462609e+00 -7.83231854e-01 4.80164975e-01 -2.90312797e-01 -8.37652028e-01 -8.12452585e-02 -5.42756796e-01 -4.21012565e-02 1.16626823e+00 -2.96161532e-01 -5.96809506e-01 -4.67864126e-01 -1.33767200e+00 5.02747297e-01 2.79646277e-01 6.58491701e-02 -9.06631589e-01 -3.38175237e-01 3.05509299e-01 2.52992809e-01 -1.29558206e+00 3.05618733e-01 8.89683664e-01 -5.52524984e-01 -8.07272553e-01 -5.49609512e-02 1.19055972e-01 3.09491992e-01 -5.18216372e-01 1.31471252e+00 -4.82951164e-01 -6.11211061e-01 4.21270370e-01 -1.35307148e-01 -6.54490709e-01 -8.31324458e-01 -5.47933638e-01 5.41066766e-01 5.74208423e-02 6.95661604e-01 -5.59966385e-01 -5.38194835e-01 -1.71295833e-02 -7.42783427e-01 -2.32705906e-01 8.18408191e-01 7.55240202e-01 5.18774152e-01 1.67094097e-01 5.39810099e-02 -8.43982697e-01 5.47769628e-02 -7.74469137e-01 -7.86122918e-01 -1.22080810e-01 -4.84091580e-01 4.62900430e-01 2.76399314e-01 3.47608291e-02 -1.02442265e+00 -5.41989654e-02 -4.48962599e-01 -4.07334059e-01 -3.78864825e-01 2.13207975e-01 -2.07843155e-01 -3.62907916e-01 8.01285863e-01 -2.68061701e-02 -5.80267489e-01 -6.98069036e-01 3.49151164e-01 2.18304485e-01 8.05187464e-01 -6.72982037e-01 1.18741465e+00 5.16391754e-01 3.53626698e-01 -5.32499075e-01 -1.04024422e+00 -6.60286844e-01 -4.72783655e-01 1.49660334e-01 7.96243131e-01 -1.03369594e+00 -1.15660205e-01 6.52750432e-02 -8.45625103e-01 2.91907579e-01 -4.80752945e-01 8.54503751e-01 -5.40527284e-01 4.31890547e-01 -3.90730858e-01 -6.65237844e-01 9.69314873e-02 -9.55507636e-01 1.05130184e+00 -1.47640258e-01 -1.93475205e-02 -6.96410537e-01 1.49427503e-01 2.81828512e-02 4.66579705e-01 4.16859388e-01 9.16340113e-01 -1.19882095e+00 -5.18590093e-01 -5.34088433e-01 -4.03855331e-02 3.69607896e-01 -3.53929073e-01 -3.53811026e-01 -1.13468623e+00 -5.33278525e-01 9.06839311e-01 -2.58429110e-01 1.04982328e+00 2.15252817e-01 1.47495437e+00 3.98065560e-02 -5.50723433e-01 7.13155866e-01 1.18604028e+00 1.09077975e-01 3.17372918e-01 1.43690407e-01 3.53831649e-01 3.19469906e-02 2.45149970e-01 5.36064148e-01 -3.32663834e-01 8.94935250e-01 2.00127140e-01 -1.47613883e-01 7.85216168e-02 -1.51743859e-01 4.88212466e-01 8.17341208e-01 4.19783503e-01 2.01887954e-02 -9.17746782e-01 3.09534103e-01 -1.43491840e+00 -1.12086284e+00 -1.66720003e-01 2.39041615e+00 4.88907516e-01 8.46395850e-01 -3.49211276e-01 -7.81439245e-02 3.24114233e-01 1.99256673e-01 -6.32277012e-01 -6.74921632e-01 -2.55815238e-01 1.00035799e+00 7.39516795e-01 2.17620030e-01 -1.19750643e+00 4.10079688e-01 6.62041712e+00 8.33653033e-01 -9.52087283e-01 4.02045339e-01 4.57660556e-01 -3.80990535e-01 -4.08841521e-01 3.47356349e-01 -8.69846225e-01 5.61784387e-01 1.45478511e+00 -1.61507711e-01 1.72354057e-01 6.37484491e-01 -1.45713225e-01 -5.60570247e-02 -1.56589580e+00 9.40831423e-01 8.79722163e-02 -1.20675552e+00 -3.45089138e-01 7.62970448e-02 5.50930738e-01 8.40158701e-01 -6.95420653e-02 6.87330306e-01 4.70750362e-01 -8.27274680e-01 7.31694400e-01 5.52662969e-01 5.32461643e-01 -4.26471055e-01 6.42471075e-01 3.29719454e-01 -5.51280379e-01 1.75279342e-02 -3.10300440e-01 2.36178786e-01 4.60454881e-01 1.12497890e+00 -7.00962842e-01 8.20227385e-01 7.22229838e-01 2.11846784e-01 -7.18450487e-01 1.02810848e+00 -5.17440774e-02 1.11079872e+00 -8.58157933e-01 5.00158072e-01 3.86069804e-01 -1.48752509e-02 1.05128837e+00 1.24797690e+00 4.94501829e-01 -2.21692175e-01 1.50036231e-01 9.33458507e-01 3.13743502e-02 -3.15467715e-01 -7.32410967e-01 -3.24213415e-01 -1.34273440e-01 1.14858639e+00 -6.42039895e-01 -1.52386248e-01 -6.16723895e-01 7.50218332e-01 -1.01065919e-01 9.18061435e-02 -7.99582124e-01 6.82622194e-02 3.28668773e-01 -6.53867200e-02 4.46574211e-01 -1.94506347e-01 6.55656634e-03 -1.43232358e+00 1.20588087e-01 -9.21558857e-01 5.30531168e-01 -5.95880032e-01 -1.47411561e+00 5.77740252e-01 1.72042221e-01 -1.05541503e+00 -5.22838652e-01 -9.13953364e-01 -7.40927458e-01 8.67350876e-01 -1.16448367e+00 -8.60342562e-01 2.86911935e-01 2.47914270e-01 2.82119542e-01 -6.32656515e-01 8.67587030e-01 2.35244498e-01 -2.42323175e-01 1.60538495e-01 5.29014349e-01 -2.44905025e-01 5.69081545e-01 -1.50057220e+00 7.01446295e-01 1.18162489e+00 7.65409529e-01 3.88457656e-01 1.38329196e+00 -8.27054262e-01 -1.36551797e+00 -6.09843552e-01 6.05566621e-01 -7.37712085e-01 1.01550710e+00 -1.04084349e+00 -7.99489915e-01 8.43409181e-01 1.58724427e-01 4.27325696e-01 6.39452577e-01 5.10818839e-01 -1.86011255e-01 1.23309217e-01 -1.01664579e+00 8.51040632e-02 8.96753490e-01 -7.60496855e-01 -7.12508738e-01 8.45671773e-01 2.94986129e-01 -5.33935905e-01 -5.38762450e-01 8.98457050e-01 -8.44818577e-02 -9.82550144e-01 1.14833987e+00 -7.99407959e-01 8.53359550e-02 -2.26946250e-01 -2.22037375e-01 -1.22694540e+00 -4.97849435e-01 -8.53288472e-01 -4.99990761e-01 9.70927656e-01 3.27127457e-01 -3.62514764e-01 4.91113991e-01 2.54488349e-01 -1.31708920e-01 -3.34707379e-01 -1.35348463e+00 -7.22422481e-01 1.83315188e-01 -9.26512599e-01 3.47065270e-01 7.74424076e-01 -3.87610018e-01 2.85742015e-01 -3.38806868e-01 7.63833523e-01 6.43171310e-01 -1.17979161e-01 4.08104688e-01 -1.35472941e+00 -9.05279040e-01 -2.96568513e-01 -8.23427439e-01 -7.43432105e-01 -1.46150932e-01 -1.27072823e+00 -1.58258855e-01 -6.74412966e-01 -7.14849010e-02 -4.45708185e-01 -9.91692245e-01 -1.65720209e-01 2.97461420e-01 -1.11112498e-01 -2.09405631e-01 -4.74485643e-02 -6.01348102e-01 6.76343143e-01 3.85753632e-01 -1.95155829e-01 6.51180744e-01 1.85099751e-01 -4.46478277e-01 9.22485471e-01 6.29382610e-01 -7.23186255e-01 -1.42289862e-01 -7.06578046e-02 5.83084106e-01 -3.10648549e-02 6.33963227e-01 -1.48967135e+00 1.15437619e-01 2.87631333e-01 8.77580166e-01 -5.96335113e-01 4.06398565e-01 -4.35809195e-01 1.66288480e-01 1.97555274e-01 -3.18968862e-01 1.41315833e-01 4.43721831e-01 8.62335861e-01 -2.82834738e-01 -7.70613194e-01 5.71870744e-01 -2.59311706e-01 -6.94984019e-01 2.73030728e-01 -2.69615442e-01 7.87666440e-02 8.75775278e-01 6.42436206e-01 2.22698048e-01 -1.68449879e-01 -8.28097105e-01 7.19573796e-02 4.97771055e-02 5.18689871e-01 2.91843805e-02 -1.25752807e+00 -6.91805601e-01 4.44768459e-01 3.33702624e-01 -4.67180848e-01 2.65526980e-01 7.04101980e-01 -2.94445813e-01 7.55285084e-01 -4.37471904e-02 -6.58149719e-01 -7.51681089e-01 7.82307506e-01 2.75503486e-01 -4.39475149e-01 -1.01875079e+00 8.06280315e-01 3.27358186e-01 -3.73546690e-01 -4.48008664e-02 -1.38735563e-01 4.10997897e-01 -3.34127426e-01 4.98543769e-01 -2.22835481e-01 6.11111522e-01 -5.50950646e-01 -2.41036937e-01 -1.97560206e-01 -3.98045748e-01 -1.62837192e-01 1.25974524e+00 3.23082656e-01 9.64031741e-02 5.37436485e-01 8.39878082e-01 3.33344221e-01 -9.39636171e-01 -5.54091752e-01 4.64938909e-01 -3.35474670e-01 5.65722644e-01 -1.00012589e+00 -6.79038942e-01 9.20391500e-01 9.03501034e-01 3.73766541e-01 7.21290946e-01 5.51948309e-01 6.11252427e-01 3.79638046e-01 3.61015350e-01 -9.28003192e-01 -2.91878641e-01 3.08794409e-01 1.01128030e+00 -1.24827707e+00 5.25292419e-02 7.59370849e-02 -4.61588293e-01 1.00983071e+00 3.48872006e-01 -2.73423027e-02 6.31343305e-01 4.69719648e-01 -4.21029508e-01 -8.00342739e-01 -9.13490832e-01 -1.36677608e-01 4.56898451e-01 2.04952419e-01 1.23036318e-01 2.15143878e-02 -1.09948330e-01 4.13096726e-01 -3.53844136e-01 -5.73195517e-01 3.68415654e-01 7.76654363e-01 -2.19093770e-01 -1.34644723e+00 -2.16293857e-01 8.89942408e-01 -7.42354989e-01 -1.47495493e-01 -3.75842629e-03 5.29113531e-01 1.27413541e-01 5.35652220e-01 -5.04940636e-02 1.62109807e-01 4.90033850e-02 5.97186923e-01 6.36984348e-01 -6.75009310e-01 -4.07576591e-01 -1.00901790e-01 1.16321467e-01 -6.63950026e-01 -1.28547072e-01 -9.47163880e-01 -6.53032482e-01 8.58987495e-02 -3.83150876e-01 3.12577605e-01 9.66274917e-01 1.11571980e+00 1.52765602e-01 7.03987539e-01 5.12812674e-01 -7.07365394e-01 -8.25342774e-01 -8.28290761e-01 -3.02274913e-01 5.16169190e-01 3.42698127e-01 -1.03641081e+00 -5.66339493e-01 -5.68587124e-01]

[15.656211853027344, 2.929722785949707]

e37a560d-0740-44fb-a7b4-1f47ba79e276

real-time-interface-control-with-motion

2201.01755

https://arxiv.org/abs/2201.01755v1

https://arxiv.org/pdf/2201.01755v1.pdf

Real-time Interface Control with Motion Gesture Recognition based on Non-contact Capacitive Sensing

Capacitive sensing is a prominent technology that is cost-effective and low power consuming with fast recognition speed compared to existing sensing systems. On account of these advantages, Capacitive sensing has been widely studied and commercialized in the domains of touch sensing, localization, existence detection, and contact sensing interface application such as human-computer interaction. However, as a non-contact proximity sensing scheme is easily affected by the disturbance of peripheral objects or surroundings, it requires considerable sensitive data processing than contact sensing, limiting the use of its further utilization. In this paper, we propose a real-time interface control framework based on non-contact hand motion gesture recognition through processing the raw signals, detecting the electric field disturbance triggered by the hand gesture movements near the capacitive sensor using adaptive threshold, and extracting the significant signal frame, covering the authentic signal intervals with 98.8% detection rate and 98.4% frame correction rate. Through the GRU model trained with the extracted signal frame, we classify the 10 hand motion gesture types with 98.79% accuracy. The framework transmits the classification result and maneuvers the interface of the foreground process depending on the input. This study suggests the feasibility of intuitive interface technology, which accommodates the flexible interaction between human to machine similar to Natural User Interface, and uplifts the possibility of commercialization based on measuring the electric field disturbance through non-contact proximity sensing which is state-of-the-art sensing technology.

['Yingshu Li', 'Nahom Ogbazghi', 'Jaya Krishna Mandivarapu', 'Hunmin Lee']

2022-01-05

['gesture-recognition']

['computer-vision']

[ 7.96554625e-01 -7.66066551e-01 -1.35366455e-01 1.39105335e-01 -3.01392972e-01 -5.64602315e-01 1.21448383e-01 -2.15655729e-01 -6.74278259e-01 2.37155601e-01 -2.78240860e-01 -2.43375614e-01 -5.60637144e-03 -7.00720131e-01 -1.89399882e-03 -7.66840279e-01 2.67998368e-01 -1.29352808e-01 5.07874310e-01 3.32703739e-02 6.68197036e-01 6.62668109e-01 -1.60770249e+00 2.71170974e-01 6.26130760e-01 1.38543224e+00 3.52345705e-01 6.88121915e-01 -8.81986693e-02 -9.10292864e-02 -7.50968933e-01 3.99551332e-01 -1.72884222e-02 -2.63346791e-01 -1.25018060e-01 -6.40421152e-01 -4.75035936e-01 -5.44633567e-01 1.33568712e-03 7.35986233e-01 9.24006343e-01 4.92946915e-02 6.26272142e-01 -9.69746590e-01 -3.54216635e-01 3.37061435e-01 -7.44636238e-01 2.91119665e-01 1.08700109e+00 6.15901947e-02 1.76366597e-01 -9.69372034e-01 4.20303553e-01 7.73523152e-01 6.76605225e-01 7.32686996e-01 -6.58647478e-01 -9.57344055e-01 -1.48418725e-01 2.35853359e-01 -1.38705444e+00 -5.35288826e-02 9.31453466e-01 -4.45956856e-01 1.21961045e+00 7.10109532e-01 5.20888746e-01 1.13759410e+00 3.67329240e-01 3.93921435e-01 9.00217712e-01 -7.63703525e-01 1.00856628e-02 -4.51515242e-02 2.25446671e-01 4.09062117e-01 2.93471944e-02 1.63463950e-01 -4.21318024e-01 -1.05905980e-01 1.04718733e+00 6.60172343e-01 -6.15056455e-01 3.83905828e-01 -1.25300801e+00 -7.65044838e-02 3.69204655e-02 8.86126280e-01 -3.48219246e-01 -1.13091759e-01 1.71049222e-01 -4.62707952e-02 -3.79065089e-02 1.50140822e-01 2.24164519e-02 -7.99615383e-01 -6.70618355e-01 -3.80883783e-01 7.10583985e-01 6.47006810e-01 -1.77432925e-01 4.74698357e-02 -2.47130811e-01 4.43551630e-01 4.92832601e-01 1.13670671e+00 5.53272665e-01 -2.10036799e-01 3.56624007e-01 7.85591006e-01 4.50370908e-01 -1.01230288e+00 -4.76202846e-01 3.33700687e-01 -8.02115023e-01 2.59729236e-01 3.45576614e-01 -4.43377793e-01 -5.90371370e-01 1.16711533e+00 2.27283686e-01 1.48489803e-01 -3.54072154e-01 1.11607599e+00 6.01519585e-01 6.95282102e-01 9.97192040e-02 -4.82721537e-01 1.41532052e+00 8.03534165e-02 -1.09167719e+00 3.36432189e-01 -2.77384520e-02 -8.95410299e-01 1.41340137e+00 4.86575007e-01 -6.18294477e-01 -8.35881650e-01 -1.23223042e+00 2.90461481e-01 -1.80019200e-01 5.94908476e-01 4.21866775e-01 8.94693255e-01 -2.84267932e-01 3.31916392e-01 -1.14928603e+00 -4.05226082e-01 -1.74988434e-01 4.46920514e-01 2.62485415e-01 4.82060313e-01 -9.69412088e-01 4.88239288e-01 -1.66201249e-01 3.47198546e-01 2.33192921e-01 -5.11266470e-01 -2.28910133e-01 -4.28158343e-02 2.19357479e-02 5.61237261e-02 8.62975359e-01 -4.69147205e-01 -2.00707340e+00 4.59592938e-01 -2.02049658e-01 2.57637143e-01 5.20869970e-01 -3.89907569e-01 -1.00763428e+00 2.93199688e-01 -3.14339638e-01 -1.20091595e-01 9.68426168e-01 -5.69304228e-01 -4.25238609e-01 -4.06499237e-01 -6.10882461e-01 -2.12262571e-02 -4.47746277e-01 2.90591270e-01 -4.94034737e-02 -4.68545496e-01 4.18827295e-01 -7.77150273e-01 3.85407567e-01 1.60499047e-02 -3.15820277e-01 -3.49806100e-01 1.26399088e+00 -5.21888673e-01 1.77286470e+00 -2.47960329e+00 -4.22362506e-01 5.76286614e-01 -2.25789487e-01 5.66261292e-01 3.25770378e-01 5.51352561e-01 3.98633808e-01 -1.12415418e-01 -1.67879283e-01 4.20507699e-01 -4.89104278e-02 -4.24776137e-01 -4.16322470e-01 4.88762587e-01 1.75551996e-02 6.66764557e-01 -7.67723262e-01 -1.55713037e-01 5.28045237e-01 5.68303645e-01 1.43411860e-01 2.63724715e-01 5.09127378e-01 4.88013148e-01 -8.53847623e-01 1.05186319e+00 5.45561492e-01 2.74945021e-01 1.85543358e-01 -5.42515576e-01 -3.77890944e-01 3.05330418e-02 -1.57094634e+00 1.18300354e+00 -2.49562949e-01 5.97331166e-01 1.17093593e-01 -6.71575785e-01 1.43677974e+00 4.53187943e-01 7.15785742e-01 -8.92637908e-01 4.81824040e-01 4.04293746e-01 -2.51383334e-01 -1.20229721e+00 1.86631098e-01 1.48573890e-03 1.15292564e-01 5.04943728e-01 -6.84065223e-01 2.04632878e-01 -4.05267239e-01 -3.29801202e-01 8.55415344e-01 4.02099580e-01 1.11812085e-01 -5.14120050e-02 3.88598025e-01 -1.57283932e-01 7.73670301e-02 4.76852357e-01 -1.60497606e-01 2.71446139e-01 -4.56631333e-01 -9.69027281e-02 -3.09014678e-01 -1.01336193e+00 -2.64290541e-01 8.94938350e-01 7.39123404e-01 7.58759230e-02 -6.50615573e-01 -3.65557708e-02 4.85255532e-02 1.47366837e-01 -4.08609882e-02 -7.79716447e-02 -8.52974296e-01 -4.87543941e-01 6.71267092e-01 8.08949292e-01 6.23474598e-01 -1.33091927e+00 -1.57002389e+00 4.29127365e-01 -2.03011349e-01 -1.02544761e+00 -4.06696886e-01 -1.32163838e-01 -8.53185356e-01 -1.10878205e+00 -7.24427462e-01 -8.06924045e-01 4.92439896e-01 2.06146657e-01 3.83323818e-01 2.16545314e-01 -7.63265312e-01 7.55844951e-01 -4.41287935e-01 -5.26335120e-01 7.39194527e-02 -1.16231151e-01 2.37259045e-01 1.94174349e-01 7.85731494e-01 -3.03407729e-01 -8.48210871e-01 4.46157247e-01 -5.53551018e-01 -2.77822882e-01 5.13791084e-01 3.58068168e-01 4.70022351e-01 -2.89112955e-01 4.84500200e-01 -3.38161141e-01 9.03909802e-01 -1.97305530e-03 -2.35661924e-01 2.43548557e-01 -4.59567010e-01 -3.62888932e-01 4.62660372e-01 -1.02178705e+00 -1.23915243e+00 2.66530886e-02 -1.38564020e-01 1.74386483e-02 -2.02217206e-01 2.37994969e-01 -1.40106127e-01 -6.36441335e-02 7.51125097e-01 6.84498698e-02 6.25607148e-02 -4.13160175e-01 -7.50988200e-02 1.41531372e+00 7.98141897e-01 -3.64414096e-01 4.74801809e-01 2.65329063e-01 -3.59921008e-02 -1.10286188e+00 4.95277941e-01 -5.94734073e-01 -5.18935740e-01 -6.17906809e-01 8.42497706e-01 -4.25633967e-01 -1.34088349e+00 7.77856469e-01 -1.10913146e+00 -1.11252747e-01 2.56463319e-01 1.06190813e+00 1.94502864e-02 3.29197735e-01 -7.38716722e-01 -1.55722392e+00 -7.73379743e-01 -7.51851797e-01 9.77021873e-01 5.40979683e-01 -6.04458213e-01 -3.44689101e-01 -3.00658762e-01 -9.92144719e-02 4.16446894e-01 5.59931636e-01 6.68314040e-01 9.29568186e-02 -3.27491432e-01 -6.80398226e-01 1.20756496e-02 -3.42094868e-01 4.79638517e-01 2.28858843e-01 -1.02357423e+00 -1.27071217e-01 9.71743539e-02 1.28830135e-01 4.55836415e-01 4.04070050e-01 1.04459620e+00 2.48812377e-01 -9.09332037e-01 2.17572421e-01 1.50345528e+00 9.66323197e-01 8.38171840e-01 -2.33497381e-01 5.51156580e-01 2.29621351e-01 8.41539919e-01 4.95117188e-01 -4.41190064e-01 5.93700111e-01 -7.86093250e-02 -1.86284080e-01 2.83448100e-01 -2.21208513e-01 4.05178130e-01 5.56498468e-01 -6.77906871e-01 -1.93561003e-01 -6.40153468e-01 3.34239118e-02 -1.29839432e+00 -1.06558204e+00 -4.23374236e-01 2.20496035e+00 6.01536214e-01 3.91713018e-03 6.82775006e-02 6.05177879e-01 9.45329010e-01 -3.32440704e-01 -7.89322555e-01 -5.63171923e-01 2.75531262e-01 5.63894689e-01 3.31913233e-01 2.39737719e-01 -7.01293707e-01 3.12341690e-01 5.85332584e+00 5.40797114e-01 -1.77398658e+00 -1.39594018e-01 1.92853145e-03 -6.54266775e-02 -1.13807067e-01 -7.93145120e-01 -6.93059385e-01 7.08482325e-01 3.70897502e-01 3.56799990e-01 2.63041437e-01 4.44031924e-01 3.54246408e-01 -4.02437806e-01 -1.04630303e+00 1.42542529e+00 -2.51078904e-01 -8.57500196e-01 -1.39381826e-01 -2.15778053e-01 6.12980910e-02 -6.38353944e-01 -6.80931881e-02 -1.45967856e-01 -9.37964559e-01 -7.69623935e-01 5.50763965e-01 7.09102631e-01 1.50176668e+00 -2.68288523e-01 5.40103555e-01 3.08752477e-01 -1.65520453e+00 -5.18254451e-02 4.83915862e-03 -6.16062880e-01 2.73644775e-01 5.12821972e-01 -5.62682092e-01 1.70336187e-01 6.91802025e-01 2.32148647e-01 2.05136761e-01 4.77659225e-01 1.74941152e-01 8.22643578e-01 -6.82770729e-01 -8.25786591e-01 -4.25380230e-01 -2.29982585e-01 4.90636081e-01 1.40136063e+00 7.20310569e-01 7.62402415e-01 1.37810456e-02 1.02171659e+00 4.41489100e-01 8.52892771e-02 -5.56987047e-01 -1.71754062e-01 9.46051359e-01 1.04357255e+00 -9.21906888e-01 -1.42913714e-01 -2.32554168e-01 1.01496387e+00 -6.80211306e-01 3.85449648e-01 -7.53561854e-01 -1.15237236e+00 1.73128307e-01 1.26752302e-01 -1.46961555e-01 -3.35677326e-01 -7.13404536e-01 -7.46509850e-01 6.37332618e-01 -2.63260692e-01 9.74421799e-02 -5.61074018e-01 -8.87111306e-01 2.00361997e-01 -2.72352576e-01 -1.50232613e+00 -2.72477478e-01 -7.86840796e-01 -9.46476698e-01 1.12847316e+00 -8.77191544e-01 -6.13517106e-01 -6.70199633e-01 7.70422637e-01 1.69042289e-01 1.93087965e-01 9.27236438e-01 2.47199267e-01 -2.72732586e-01 8.02450895e-01 -6.17301539e-02 2.11240977e-01 7.04012036e-01 -7.30417669e-01 1.59400836e-01 8.66477251e-01 -4.10213828e-01 7.78172970e-01 2.13401288e-01 -7.68933833e-01 -1.91856468e+00 -2.73616701e-01 6.74699724e-01 -2.26748571e-01 2.15258181e-01 -4.39539015e-01 -8.65596294e-01 -1.15588680e-01 -4.01505202e-01 -3.28280389e-01 6.65996552e-01 -5.72297394e-01 1.02504745e-01 -2.79186279e-01 -1.51278710e+00 5.33277750e-01 1.22662103e+00 -8.07832837e-01 -4.60813314e-01 -3.56463790e-01 2.63154626e-01 -4.41134125e-01 -9.81591940e-01 4.58835393e-01 1.39209390e+00 -7.36406863e-01 8.95068169e-01 3.38134885e-01 -2.35614717e-01 -3.46948266e-01 -7.19621703e-02 -4.92708892e-01 -3.04375768e-01 -7.50748396e-01 9.43450555e-02 1.32202578e+00 2.01165885e-01 -8.83958638e-01 4.59589064e-01 9.03526068e-01 1.49556220e-01 -6.54477954e-01 -9.09450948e-01 -4.19958770e-01 -7.06849813e-01 -4.56423759e-01 5.48941731e-01 7.06582665e-01 8.00637960e-01 2.34568510e-02 3.41445692e-02 1.33340880e-01 1.91168457e-01 2.36892909e-01 2.37101182e-01 -1.34949291e+00 -1.03419714e-01 -3.95413458e-01 -3.47900659e-01 -1.20641088e+00 -6.19557261e-01 -2.59787828e-01 1.87280729e-01 -1.19425523e+00 -3.16689819e-01 -3.44167799e-01 -3.23823363e-01 2.15943471e-01 -2.69662112e-01 4.86957133e-01 2.08310559e-02 3.00545007e-01 1.29928440e-01 -1.27573282e-01 1.26309061e+00 -1.63957223e-01 -7.43790865e-01 1.57269180e-01 -3.60791944e-02 3.07052821e-01 6.22074127e-01 1.96939290e-01 -2.76239336e-01 -5.41091636e-02 1.87759474e-02 1.45391881e-01 3.44968885e-01 -1.16280699e+00 3.22574347e-01 -2.20214009e-01 7.40656734e-01 -4.39507067e-01 2.31624264e-02 -1.09541750e+00 1.36539534e-01 9.01027918e-01 2.91166216e-04 -2.13246718e-02 8.44141096e-02 2.99558133e-01 2.23258704e-01 1.36807198e-02 4.16489780e-01 1.48958474e-01 -9.10133362e-01 -1.76949188e-01 -6.85522199e-01 -3.65297467e-01 8.10817182e-01 -1.00921154e+00 -6.18299581e-02 1.50346518e-01 -4.21674043e-01 -5.25310278e-01 9.11611225e-03 2.89856702e-01 7.41269708e-01 -1.17862272e+00 -2.55068410e-02 9.59271312e-01 -1.66782048e-02 -4.50930536e-01 2.53316134e-01 7.76773989e-01 -3.20560277e-01 2.72407681e-01 -2.21127972e-01 -8.90799761e-01 -1.32603753e+00 1.73731416e-01 2.17581540e-01 4.24421191e-01 -7.74729729e-01 4.49184626e-01 -5.98234057e-01 4.69355375e-01 3.74369889e-01 -8.86676133e-01 -4.45377141e-01 -1.70227900e-01 8.00037265e-01 9.79362488e-01 1.55822001e-02 -9.67449471e-02 -6.78724051e-01 1.36371279e+00 6.71878576e-01 -2.62306333e-01 5.99119127e-01 -1.58549756e-01 2.67575800e-01 6.77490354e-01 8.41480732e-01 1.44692823e-01 -1.08093023e+00 1.79985538e-01 -1.75773054e-01 -3.93044800e-01 -3.41224194e-01 -1.08177149e+00 -5.66388845e-01 9.28817451e-01 1.27335405e+00 3.42464983e-01 1.37071741e+00 -4.90363747e-01 1.02249956e+00 2.92480588e-01 7.60104716e-01 -1.23160696e+00 -1.14519730e-01 2.73913369e-02 5.97859085e-01 -7.24344015e-01 -3.13630641e-01 -4.49838638e-01 -1.33231297e-01 1.32860756e+00 5.22284210e-01 -5.70823550e-02 7.61837125e-01 8.84612679e-01 3.95148486e-01 1.33880034e-01 6.33747280e-02 1.60610035e-01 3.02989632e-01 9.43355143e-01 6.15204513e-01 4.61170733e-01 -7.20626175e-01 8.29489648e-01 2.26069212e-01 6.28608584e-01 3.03047467e-02 1.13593316e+00 -5.82395613e-01 -7.53097653e-01 -6.18335724e-01 5.25600016e-01 -5.18847406e-01 2.76642114e-01 -3.08037639e-01 4.86688882e-01 1.08375609e-01 1.29099381e+00 2.11843759e-01 -8.04941058e-01 6.28669977e-01 1.96717441e-01 6.00661159e-01 -5.51314726e-02 -6.31308436e-01 2.68521786e-01 -4.41639721e-01 -6.34673834e-01 -5.12917280e-01 -2.75084138e-01 -2.05459762e+00 -1.62180483e-01 -4.30125356e-01 -1.56349316e-02 8.79676640e-01 1.02111733e+00 5.35735548e-01 4.02038515e-01 5.18697023e-01 -1.03752971e+00 -5.00028372e-01 -1.20073533e+00 -7.24235654e-01 3.99653077e-01 1.64602771e-01 -8.25079918e-01 -1.79018497e-01 -1.13140188e-01]

[6.4946794509887695, -0.20103038847446442]

88cc8d0b-d938-4ca6-9f5f-5fd6af40a649

selfvi-self-supervised-light-field-video

http://openaccess.thecvf.com//content/ICCV2021/html/Shedligeri_SeLFVi_Self-Supervised_Light-Field_Video_Reconstruction_From_Stereo_Video_ICCV_2021_paper.html

http://openaccess.thecvf.com//content/ICCV2021/papers/Shedligeri_SeLFVi_Self-Supervised_Light-Field_Video_Reconstruction_From_Stereo_Video_ICCV_2021_paper.pdf

SeLFVi: Self-Supervised Light-Field Video Reconstruction From Stereo Video

Light-field (LF) imaging is appealing to the mobile devices market because of its capability for intuitive post-capture processing. Acquiring LF data with high angular, spatial and temporal resolution poses significant challenges, especially with space constraints preventing bulky optics. At the same time, stereo video capture, now available on many consumer devices, can be interpreted as a sparse LF-capture. We explore the application of small baseline stereo videos for reconstructing high fidelity LF videos. We propose a self-supervised learning-based algorithm for LF video reconstruction from stereo video. The self-supervised LF video reconstruction is guided via the geometric information from the individual stereo pairs and the temporal information from the video sequence. LF estimation is further regularized by a low-rank constraint based on layered LF displays. The proposed self-supervised algorithm facilitates advantages such as post-training fine-tuning on test sequences and variable angular view interpolation and extrapolation. Quantitatively the LF videos show higher fidelity than previously proposed unsupervised approaches for LF reconstruction. We demonstrate our results via LF videos generated from stereo videos acquired from commercially available stereoscopic cameras. Finally, we demonstrate that our reconstructed LF videos allow applications such as post-capture focus control and RoI-based focus tracking for videos.

['Kaushik Mitra', 'Oliver Cossairt', 'Sushobhan Ghosh', 'Florian Schiffers', 'Prasan Shedligeri']

2021-01-01

iccv-2021-1

['video-reconstruction']

['computer-vision']

[ 6.09169185e-01 -4.13221836e-01 -1.56209007e-01 -4.54794437e-01 -9.17600989e-01 -5.57480216e-01 2.39300668e-01 -5.20489395e-01 -3.35044801e-01 8.90075564e-01 4.42761153e-01 6.09145127e-02 -1.82543710e-01 -2.71855384e-01 -1.00990462e+00 -7.11371779e-01 2.56207764e-01 7.15939403e-02 2.94289112e-01 3.16451132e-01 4.50596273e-01 4.87231791e-01 -1.97985780e+00 5.03371596e-01 8.32616985e-01 1.15769148e+00 8.58513355e-01 7.18609393e-01 2.88131833e-01 6.72378957e-01 1.87505186e-01 -5.75838191e-03 3.59048843e-01 -2.48342410e-01 -4.09299076e-01 2.21283585e-01 1.10071552e+00 -8.62274408e-01 -3.93293679e-01 8.32565606e-01 4.79593158e-01 9.84705091e-02 4.16456223e-01 -7.78598905e-01 6.33276105e-02 -1.30597353e-01 -5.21001577e-01 3.38354930e-02 9.91776645e-01 4.14287806e-01 6.46901488e-01 -1.13733006e+00 1.01368189e+00 7.79549003e-01 4.38311458e-01 3.98638278e-01 -1.43371582e+00 -5.88920116e-01 -3.76023561e-01 3.74144167e-01 -1.15341091e+00 -8.68794799e-01 8.44537795e-01 -7.66438365e-01 6.91298842e-01 2.04631031e-01 7.95943439e-01 8.32454503e-01 2.77986974e-01 5.38182080e-01 1.29163408e+00 -5.26623249e-01 7.09260032e-02 -2.00048778e-02 -2.06047535e-01 6.83639348e-01 -1.55778565e-02 3.86775017e-01 -8.70896697e-01 6.67618066e-02 1.03460717e+00 1.63001597e-01 -9.07003820e-01 -5.51954448e-01 -1.31885743e+00 3.87938768e-01 1.23180799e-01 3.65139209e-02 -1.80774316e-01 -6.88866675e-02 7.86395073e-02 6.96261600e-02 4.98028368e-01 4.46020246e-01 -7.15869963e-02 -4.17208262e-02 -1.18940461e+00 -1.85464844e-01 2.75558919e-01 1.08340859e+00 8.47179651e-01 4.37197499e-02 -9.35289040e-02 7.24246383e-01 3.53331596e-01 7.29364216e-01 2.04151750e-01 -1.53995633e+00 3.76453131e-01 9.06993821e-02 2.31181964e-01 -5.97729623e-01 -2.37481683e-01 -1.17906190e-01 -7.71846533e-01 2.92231083e-01 3.28784078e-01 1.08682416e-01 -4.59786594e-01 1.43870437e+00 2.73096502e-01 4.54381198e-01 -3.72337371e-01 1.15630913e+00 6.81832314e-01 6.41352117e-01 -5.69116771e-01 -9.22137797e-01 7.22761631e-01 -5.51274836e-01 -8.40970695e-01 1.10851645e-01 1.61997095e-01 -1.02146053e+00 1.15849209e+00 8.12793911e-01 -1.30933166e+00 -6.22126162e-01 -8.13272178e-01 -3.84540170e-01 5.45149446e-01 2.82873273e-01 4.67409700e-01 3.95444244e-01 -9.99660373e-01 4.05340344e-01 -6.54053748e-01 -2.18879014e-01 3.53366792e-01 5.36525607e-01 -4.76300478e-01 -4.12562132e-01 -5.96436560e-01 2.93179959e-01 1.42041696e-02 -7.59527683e-02 -8.07515621e-01 -1.00282514e+00 -8.10991645e-01 -2.06387848e-01 4.48476881e-01 -8.81458879e-01 8.55333447e-01 -6.94624186e-01 -1.82273293e+00 1.01102412e+00 -5.47832906e-01 -2.80482531e-01 3.77723485e-01 -4.16085809e-01 -2.67247468e-01 7.64835954e-01 1.43359303e-01 5.93576670e-01 1.36506069e+00 -1.12381864e+00 -6.66061401e-01 -7.73703530e-02 2.03282554e-02 2.62390047e-01 -1.69389024e-01 -1.09817073e-01 -5.46066403e-01 -3.42229158e-01 1.09973699e-01 -8.61603379e-01 1.30907401e-01 3.39108199e-01 -2.29030058e-01 5.53536236e-01 9.32426929e-01 -5.37050545e-01 1.14761043e+00 -2.17066073e+00 2.26529941e-01 -1.31598979e-01 3.69607180e-01 3.44973318e-02 1.76061645e-01 2.13036507e-01 -1.01632714e-01 -7.21516371e-01 1.78518135e-03 -1.96975186e-01 -6.77352250e-01 -2.41019830e-01 -2.92489558e-01 8.00834596e-01 -2.26442948e-01 5.99815249e-01 -9.06452119e-01 -4.65548724e-01 7.51600683e-01 4.76633847e-01 -1.18668175e+00 4.75540787e-01 -1.41480044e-01 1.16444147e+00 -1.53986409e-01 6.13257110e-01 7.28956223e-01 -3.62992764e-01 1.93195924e-01 -8.91507030e-01 -5.67012548e-01 7.31283650e-02 -9.64255929e-01 2.04020715e+00 -6.55862272e-01 9.72622275e-01 3.61219287e-01 -6.60328686e-01 3.78667921e-01 9.17475224e-02 7.42611170e-01 -7.27837622e-01 -1.59204435e-02 3.34257841e-01 -4.68848020e-01 -7.72245467e-01 4.77830738e-01 -2.80605137e-01 4.77259099e-01 4.68106925e-01 2.52539963e-01 -3.97219598e-01 2.36707166e-01 2.26754904e-01 6.72509074e-01 3.96569997e-01 1.57424599e-01 -3.34785491e-01 7.30198979e-01 -3.46868277e-01 2.13661224e-01 4.57109272e-01 2.48002619e-01 1.10176635e+00 -1.14990346e-01 -1.99532479e-01 -1.11135995e+00 -1.14298260e+00 -4.38749313e-01 5.48739493e-01 2.96663404e-01 -4.86388922e-01 -5.58347881e-01 -2.91012228e-01 -2.72507757e-01 2.22081259e-01 2.71480586e-02 3.27490240e-01 -7.30944812e-01 3.16746645e-02 -2.98610002e-01 1.37135804e-01 4.48660225e-01 -5.70707381e-01 -6.22558534e-01 4.41203825e-02 -4.52833116e-01 -1.45811069e+00 -5.87261736e-01 -2.50418007e-01 -9.59126949e-01 -1.13871777e+00 -7.38652527e-01 -7.46552765e-01 6.92952514e-01 8.00885141e-01 8.49596500e-01 -1.67024031e-01 -4.07833338e-01 9.39126492e-01 1.14362381e-01 3.37957859e-01 -7.78804272e-02 -4.91289258e-01 5.10062993e-01 4.32607681e-01 -2.39746809e-01 -9.37038958e-01 -1.03521335e+00 5.34316003e-01 -8.77396047e-01 5.41034877e-01 3.23839992e-01 9.46057975e-01 8.91976595e-01 -1.12081587e-01 -4.49150428e-02 -8.16291273e-01 4.51865867e-02 -3.82527374e-02 -9.62305486e-01 3.49237360e-02 -1.68109626e-01 -1.96488008e-01 8.20526898e-01 -3.40263039e-01 -1.45057487e+00 2.09920004e-01 -1.39813751e-01 -8.40619683e-01 4.59720235e-04 3.32121514e-02 -8.56422931e-02 -5.21629691e-01 7.07726359e-01 2.83779860e-01 2.00273648e-01 -2.95892358e-01 1.42313048e-01 5.99033892e-01 7.15346754e-01 -4.51855749e-01 7.69714594e-01 9.51096892e-01 2.29220644e-01 -1.31840074e+00 -8.09393287e-01 -7.29270756e-01 -5.49032092e-01 -8.19780409e-01 8.35605502e-01 -1.18869519e+00 -1.04704034e+00 3.15782547e-01 -9.73619640e-01 -2.69479543e-01 -2.32445136e-01 9.94617522e-01 -9.15430725e-01 5.43502748e-01 -6.45835936e-01 -5.30868351e-01 2.62189507e-02 -1.27307129e+00 1.35019743e+00 2.98766810e-02 4.79422435e-02 -8.81216109e-01 -7.37069026e-02 7.46442974e-01 3.04467916e-01 -1.71523452e-01 5.85194170e-01 7.73239911e-01 -1.30991673e+00 2.49480844e-01 -1.17808357e-01 4.67952281e-01 1.68282390e-01 -5.12401797e-02 -1.14900744e+00 -4.91673440e-01 4.28689510e-01 -3.83641273e-01 6.23630345e-01 9.88808393e-01 1.19270277e+00 -1.87262103e-01 -2.57635713e-01 1.17997122e+00 1.75550187e+00 -3.97124439e-02 7.14716911e-01 -7.36939907e-02 9.50435281e-01 6.75296903e-01 8.56014132e-01 3.84024620e-01 -1.86790377e-01 1.05754650e+00 1.94031715e-01 1.14922278e-01 -3.30408245e-01 -3.86573017e-01 4.04305696e-01 8.38024139e-01 -2.91730553e-01 -3.94035019e-02 -5.78754306e-01 1.41812205e-01 -1.45661950e+00 -1.19400907e+00 -2.96950072e-01 2.59542942e+00 6.54921889e-01 -4.73250709e-02 -8.32431316e-02 1.25965208e-01 5.55573881e-01 -2.46811304e-02 -4.98891681e-01 4.03037488e-01 -1.29715040e-01 1.91858605e-01 3.48900884e-01 9.32528317e-01 -8.68382275e-01 5.60366213e-01 6.23848677e+00 8.95800412e-01 -1.44972241e+00 1.61579967e-01 4.81673270e-01 -6.02912068e-01 -5.39489388e-01 3.35143832e-03 -8.57765377e-01 6.38734043e-01 3.25178951e-01 2.89995167e-02 4.68802959e-01 3.72793525e-01 5.75622201e-01 -6.27400756e-01 -1.33775246e+00 1.56779742e+00 1.11370333e-01 -1.70571291e+00 -1.99567061e-02 1.20890029e-01 9.14035499e-01 -2.92954352e-02 1.89915329e-01 -4.22101200e-01 -5.16000628e-01 -6.14682078e-01 5.25997221e-01 7.14862108e-01 1.61406803e+00 -4.35718983e-01 1.64419003e-02 3.14722061e-01 -9.84493494e-01 -9.89039615e-02 -2.86380172e-01 -1.92102827e-02 6.61666572e-01 9.36174989e-01 -4.29986596e-01 4.23448145e-01 7.63770819e-01 1.20097446e+00 -2.11648181e-01 9.57328618e-01 1.10410191e-01 3.34561855e-01 -3.71815801e-01 4.89784151e-01 -3.92752349e-01 -6.91316068e-01 8.95088613e-01 9.29566026e-01 5.19496143e-01 2.77297229e-01 -9.21980068e-02 5.99078476e-01 3.06097984e-01 -1.33672237e-01 -9.48644102e-01 3.41244549e-01 2.55939234e-02 1.18506050e+00 -3.54772985e-01 -3.59317437e-02 -6.21459424e-01 8.22512090e-01 8.74742679e-03 5.14992356e-01 -5.49241424e-01 7.55175203e-02 4.12801981e-01 9.64433730e-01 1.18072107e-01 -3.97779703e-01 1.17492452e-01 -1.75725114e+00 1.56417519e-01 -5.74700475e-01 1.90243125e-02 -1.33083045e+00 -9.91896272e-01 2.93451905e-01 7.63340965e-02 -2.03838038e+00 -3.96974683e-01 -6.69587672e-01 -3.07380438e-01 4.51330036e-01 -1.70421422e+00 -7.84598410e-01 -4.16403711e-01 1.04979992e+00 7.17662215e-01 -1.36723697e-01 3.81279647e-01 5.46287298e-01 -8.49455670e-02 1.43540293e-01 4.01528001e-01 -6.08397901e-01 9.27973926e-01 -7.05515862e-01 -4.93506193e-01 8.63566399e-01 -4.69203144e-02 7.20529497e-01 6.01306438e-01 -6.28195703e-01 -1.79368651e+00 -7.61784375e-01 4.32719380e-01 -2.50118375e-01 3.38630140e-01 -4.40880597e-01 -5.32709122e-01 4.42075670e-01 1.98680624e-01 1.92383334e-01 6.18427753e-01 -3.31251889e-01 1.27220780e-01 -5.18836319e-01 -1.05934024e+00 4.90069687e-01 1.32360935e+00 -9.34423923e-01 -9.64888185e-02 6.02923214e-01 3.91466230e-01 -6.33790195e-01 -5.41027069e-01 3.55702698e-01 8.09115887e-01 -1.61678445e+00 1.04521060e+00 1.22254565e-01 7.43368924e-01 -4.81559992e-01 -1.18839659e-01 -9.18618977e-01 -3.07695121e-01 -1.02470744e+00 -5.85068241e-02 9.51858342e-01 -1.24377452e-01 -4.22068119e-01 9.00508642e-01 3.24504644e-01 -1.30418465e-01 -4.37425941e-01 -7.97681928e-01 -6.67939484e-01 -9.29781020e-01 -5.31444490e-01 -2.84217209e-01 6.24779582e-01 -3.97081785e-02 3.78417403e-01 -6.22553527e-01 9.78388488e-02 1.15167773e+00 4.06361431e-01 7.78010190e-01 -9.87714648e-01 -6.95034742e-01 2.75593221e-01 -3.05357397e-01 -1.71767533e+00 1.76552102e-01 -7.07479656e-01 -7.89386258e-02 -9.80849504e-01 2.79236406e-01 -2.08775327e-01 3.79267693e-01 -4.19101775e-01 3.03791553e-01 6.17243767e-01 4.03297842e-02 4.86927956e-01 -5.72174430e-01 4.41327184e-01 1.39725876e+00 2.01305866e-01 -1.53673008e-01 2.26654056e-02 -9.72205922e-02 7.42765844e-01 2.75381058e-01 -4.47530262e-02 -7.00667441e-01 -5.06417930e-01 3.98449630e-01 6.01215422e-01 3.98609847e-01 -1.12699187e+00 5.32634079e-01 1.07943267e-02 4.10270900e-01 -5.35703301e-01 6.30748987e-01 -1.00445139e+00 4.77702737e-01 5.86516820e-02 -8.80309641e-02 -4.51457411e-01 -1.70253944e-02 6.37030065e-01 -4.06911880e-01 -1.54833496e-01 9.92303550e-01 -5.29559851e-02 -4.47602630e-01 3.06674212e-01 -4.60680515e-01 6.71117976e-02 8.36708248e-01 -5.88988781e-01 -1.27525911e-01 -7.01028287e-01 -5.97317934e-01 -6.10096231e-02 7.22904205e-01 -1.07171878e-01 9.74711061e-01 -1.25258124e+00 -3.09142143e-01 8.73620093e-01 1.42215695e-02 -1.30875623e-02 7.44174242e-01 1.08582723e+00 -7.64338434e-01 5.30295908e-01 -4.25005078e-01 -1.23733389e+00 -1.44484937e+00 4.21372503e-01 8.18937346e-02 2.81951204e-02 -7.70272672e-01 5.48706055e-01 8.84587109e-01 9.53776091e-02 6.65088594e-02 -2.68783063e-01 -1.43956378e-01 -2.58962721e-01 6.01874828e-01 3.93361062e-01 -7.73456926e-03 -5.93512833e-01 -1.47489551e-02 1.25950682e+00 9.88925844e-02 -2.44366139e-01 1.37436819e+00 -4.97095227e-01 4.09872755e-02 4.01202410e-01 1.33246493e+00 5.56660891e-01 -1.60000002e+00 -1.20211653e-01 -7.08640397e-01 -1.08155775e+00 1.44683093e-01 -2.05419570e-01 -9.92974639e-01 8.61702681e-01 4.98145580e-01 -2.67400265e-01 1.46610641e+00 -2.24571079e-01 6.25971556e-01 1.77590460e-01 8.89438808e-01 -8.87271881e-01 1.41625628e-01 2.41795063e-01 8.07002962e-01 -1.19932604e+00 2.04673409e-01 -9.41061318e-01 -2.40543365e-01 1.20197356e+00 2.86012352e-01 1.97011624e-02 6.25800431e-01 2.86206812e-01 -3.26824725e-01 -6.14342950e-02 -6.84614897e-01 6.78399131e-02 4.56316382e-01 6.65711105e-01 2.59309262e-01 -4.55186963e-01 -2.51114279e-01 -1.08118594e-01 1.23543896e-01 3.77463222e-01 6.53920472e-01 5.42207837e-01 -2.79545605e-01 -9.76758897e-01 -2.83976614e-01 5.98026574e-01 -3.88809323e-01 -1.10172212e-01 3.44886005e-01 3.82566631e-01 -1.23207523e-02 7.43709683e-01 -4.89225462e-02 -1.42471030e-01 1.35212481e-01 -4.30054516e-01 1.06779158e+00 -6.34120584e-01 -5.93634136e-02 5.61610997e-01 -3.31035517e-02 -1.00189066e+00 -8.46204638e-01 -7.11837590e-01 -6.78187788e-01 -5.84406294e-02 -1.70734018e-01 -1.24543071e-01 2.83775091e-01 6.05025291e-01 3.73525023e-01 -1.13169141e-01 8.51280868e-01 -1.40883982e+00 2.34939277e-01 -4.82537210e-01 -6.99678361e-01 4.36133742e-01 6.71730578e-01 -7.36083329e-01 -6.50071144e-01 4.06173974e-01]

[9.692842483520508, -2.5810649394989014]

39624cfb-dfdd-49a2-89d1-14dd1d0e5144

low-complexity-acoustic-scene-classification-2

2306.02054

https://arxiv.org/abs/2306.02054v1

https://arxiv.org/pdf/2306.02054v1.pdf

Low-Complexity Acoustic Scene Classification Using Data Augmentation and Lightweight ResNet

We present a work on low-complexity acoustic scene classification (ASC) with multiple devices, namely the subtask A of Task 1 of the DCASE2021 challenge. This subtask focuses on classifying audio samples of multiple devices with a low-complexity model, where two main difficulties need to be overcome. First, the audio samples are recorded by different devices, and there is mismatch of recording devices in audio samples. We reduce the negative impact of the mismatch of recording devices by using some effective strategies, including data augmentation (e.g., mix-up, spectrum correction, pitch shift), usages of multi-patch network structure and channel attention. Second, the model size should be smaller than a threshold (e.g., 128 KB required by the DCASE2021 challenge). To meet this condition, we adopt a ResNet with both depthwise separable convolution and channel attention as the backbone network, and perform model compression. In summary, we propose a low-complexity ASC method using data augmentation and a lightweight ResNet. Evaluated on the official development and evaluation datasets, our method obtains classification accuracy scores of 71.6% and 66.7%, respectively; and obtains Log-loss scores of 1.038 and 1.136, respectively. Our final model size is 110.3 KB which is smaller than the maximum of 128 KB.

['Qianhua He', 'Wenfeng Pang', 'Qisheng Huang', 'Wei Xie', 'Wenchang Cao', 'Yanxiong Li']

2023-06-03

['acoustic-scene-classification', 'scene-classification', 'model-compression']

['audio', 'computer-vision', 'methodology']

[ 4.54385132e-01 -1.47393748e-01 2.35739425e-01 -1.75434589e-01 -8.57603908e-01 -1.93558902e-01 -3.38370986e-02 6.70307279e-02 -6.15501165e-01 3.85167837e-01 1.74124852e-01 -3.28385681e-01 1.38492927e-01 -4.68997657e-01 -9.86730456e-01 -5.07976949e-01 9.82429311e-02 -4.13045175e-02 2.24265441e-01 -1.65338919e-03 -4.55516465e-02 3.35530788e-01 -1.72090924e+00 4.44074988e-01 5.74289382e-01 1.73853767e+00 3.88689846e-01 8.85524511e-01 -2.78481338e-02 6.84397042e-01 -1.04006219e+00 -3.59100997e-01 1.73471615e-01 -1.43000424e-01 -7.00316131e-01 -2.11701155e-01 5.17600060e-01 -7.48090446e-01 -3.35242152e-01 9.51981425e-01 1.00085711e+00 -1.19727496e-02 2.81279087e-01 -1.20212507e+00 -6.85710758e-02 9.82132554e-01 -6.05139911e-01 2.45511994e-01 -1.03619490e-02 -5.33872098e-02 7.99151242e-01 -7.98476100e-01 -2.12698892e-01 1.13074625e+00 9.76674199e-01 4.54590142e-01 -1.01367843e+00 -1.39460218e+00 8.86785910e-02 5.43730259e-01 -1.70292139e+00 -8.80024493e-01 6.93456769e-01 -2.55227655e-01 1.14516342e+00 4.92057562e-01 6.37998283e-01 9.75646079e-01 -1.65983155e-01 4.53721732e-01 5.58119178e-01 -4.49926704e-01 3.33192706e-01 -2.34195143e-01 7.16296062e-02 2.68716644e-02 -7.53974766e-02 -3.60203803e-01 -8.89452338e-01 -2.76297629e-01 5.70076644e-01 -1.45802543e-01 -2.19611779e-01 5.41374862e-01 -8.94226313e-01 3.86793941e-01 1.21604674e-01 1.68977790e-02 -1.93948135e-01 3.78790796e-01 9.16205168e-01 1.98455498e-01 3.80009651e-01 1.97016001e-01 -6.32897794e-01 -4.03378159e-01 -8.39260876e-01 -5.79607766e-03 3.25829893e-01 1.17131114e+00 3.54958862e-01 3.39314431e-01 5.72973192e-02 1.07081795e+00 2.07063109e-01 6.09834313e-01 8.10120404e-01 -5.91049016e-01 9.99547422e-01 1.09329648e-01 -2.38734573e-01 -9.14630115e-01 -3.97053957e-01 -4.26387578e-01 -1.36337781e+00 -5.22116005e-01 8.62436667e-02 -2.39445820e-01 -8.76090229e-01 1.81360316e+00 1.58135608e-01 9.82009828e-01 2.49846689e-02 6.46904826e-01 9.11715746e-01 8.75538290e-01 9.79316980e-02 -1.82356209e-01 1.45291877e+00 -1.07539165e+00 -9.31070685e-01 -8.17132667e-02 7.40970790e-01 -9.19710159e-01 1.26567411e+00 6.88638508e-01 -1.11026061e+00 -9.55745518e-01 -1.44347262e+00 5.95850730e-03 -1.78850014e-02 2.32994020e-01 2.68574506e-01 8.74640703e-01 -7.78271556e-01 4.56899256e-01 -5.69942296e-01 1.63990200e-01 3.07299018e-01 5.56616306e-01 -1.35841407e-02 5.04776649e-02 -1.29123485e+00 4.75225644e-03 3.90937507e-01 2.20420480e-01 -7.68972874e-01 -9.76796389e-01 -5.88740647e-01 2.98352867e-01 3.20160776e-01 -1.63594708e-01 1.33323407e+00 -6.17397249e-01 -1.61325777e+00 1.71900958e-01 2.67841846e-01 -8.12213480e-01 2.77265042e-01 -6.36501789e-01 -8.72857332e-01 -6.85548084e-03 -3.11360806e-01 3.63082200e-01 7.68971980e-01 -6.34524226e-01 -7.82933891e-01 -2.19081447e-01 -2.77285222e-02 1.61316320e-01 -1.02235806e+00 2.75524139e-01 -7.39179015e-01 -8.96045983e-01 7.61170685e-03 -1.01196647e+00 -1.05589263e-01 -3.79470646e-01 -6.57332420e-01 1.63715228e-01 7.24178851e-01 -9.81036484e-01 1.62026477e+00 -2.66688943e+00 -3.47628087e-01 1.76341653e-01 2.18986526e-01 4.12455708e-01 -1.52790785e-01 -6.21822774e-02 -1.90674603e-01 3.48561972e-01 -5.42201251e-02 -6.60367310e-01 -1.62826523e-01 -2.13863298e-01 -4.69661236e-01 1.98424309e-01 -1.22191504e-01 4.10306484e-01 -2.05942407e-01 -2.46993721e-01 1.17380202e-01 5.55602014e-01 -7.05416143e-01 3.24507296e-01 1.75583050e-01 1.78120911e-01 -9.99960750e-02 4.45710689e-01 1.07972860e+00 -1.97830275e-02 -6.11342713e-02 -6.90333188e-01 -2.42006183e-02 7.38613307e-01 -1.59265792e+00 1.96331453e+00 -7.61485636e-01 4.59054112e-01 1.82744205e-01 -5.64064980e-01 7.29374826e-01 4.54241842e-01 3.28473926e-01 -7.72542655e-01 2.70837784e-01 2.81961977e-01 -1.50061011e-01 -2.58790225e-01 5.06129205e-01 1.91790119e-01 -2.05213025e-01 2.72924334e-01 -1.41915858e-01 -7.07061738e-02 -3.13655108e-01 -1.57104790e-01 1.26854920e+00 -5.25893033e-01 -1.67166263e-01 4.82011288e-02 3.80365759e-01 -6.11268163e-01 7.18368948e-01 5.94738245e-01 1.11284986e-01 7.56774545e-01 1.37247607e-01 -1.84109434e-01 -1.07365775e+00 -6.45775437e-01 -1.28667310e-01 1.06947386e+00 -1.64395645e-02 -9.83846843e-01 -9.54645038e-01 -1.08753212e-01 -3.57499003e-01 2.47848317e-01 -1.98276281e-01 -4.12049741e-01 -8.59859228e-01 -7.95380533e-01 1.22947216e+00 6.99785054e-01 9.09667969e-01 -7.51698315e-01 -5.44658542e-01 3.09754997e-01 -2.57955492e-01 -1.46029735e+00 -5.81984520e-01 3.50192636e-01 -6.66214824e-01 -6.53310955e-01 -4.03133899e-01 -7.93783605e-01 1.56617075e-01 2.25506142e-01 7.95945883e-01 7.14837313e-02 8.52984004e-03 -2.01267451e-01 -5.93741059e-01 -5.39893389e-01 -1.46239161e-01 3.90046239e-01 3.72898012e-01 1.75616235e-01 -6.41613305e-02 -8.09046328e-01 -6.44243717e-01 2.67933488e-01 -9.95380402e-01 2.15076566e-01 5.66094935e-01 6.47699952e-01 6.76268816e-01 1.33587331e-01 5.55837035e-01 -5.08297384e-01 5.19341409e-01 -2.49099672e-01 -4.58202302e-01 5.32822404e-03 -3.25553596e-01 -5.28845072e-01 1.05069947e+00 -9.29194152e-01 -7.25498319e-01 2.26411536e-01 -5.94609320e-01 -5.67445040e-01 -7.36613583e-04 3.89092237e-01 -5.70737302e-01 1.15834456e-02 3.16810429e-01 1.19160831e-01 -4.25378382e-01 -8.05568874e-01 1.49254456e-01 1.38236678e+00 6.01757526e-01 -2.38868251e-01 5.19062340e-01 7.96262845e-02 -3.16593260e-01 -9.87136543e-01 -3.79671305e-01 -1.54058382e-01 -2.69207060e-01 -7.72409746e-03 5.97700894e-01 -1.36589777e+00 -7.89351285e-01 9.49818432e-01 -1.16030085e+00 -4.34804916e-01 -3.92552674e-01 7.50056207e-01 -2.88542390e-01 2.96767384e-01 -9.53600943e-01 -7.09954441e-01 -7.08743632e-01 -9.79484499e-01 9.01470423e-01 -8.68304968e-02 4.00354490e-02 3.38456184e-02 -3.87075156e-01 3.92435670e-01 6.44084156e-01 -2.64386892e-01 8.13551903e-01 -7.35483527e-01 -3.09131712e-01 -1.40531003e-01 -2.08937272e-01 7.49965727e-01 6.96834503e-03 -2.84082413e-01 -1.48165238e+00 -3.84562194e-01 3.14765334e-01 -4.39799905e-01 5.86184740e-01 2.25429878e-01 1.99490321e+00 -4.66810524e-01 -7.96280280e-02 8.46685112e-01 9.98088181e-01 5.35527349e-01 7.95982301e-01 -3.15026268e-02 8.54743183e-01 2.94571698e-01 4.16653484e-01 8.04351270e-01 2.98536122e-01 9.61357892e-01 4.09717113e-01 -7.25344568e-02 -4.52231169e-01 -2.36219168e-01 1.51610941e-01 1.63759136e+00 2.70294636e-01 -2.84886718e-01 -8.94747853e-01 3.73561293e-01 -1.50014234e+00 -3.45690221e-01 -4.85338718e-02 2.30223250e+00 9.23196733e-01 3.72289270e-01 4.88718087e-03 8.26897681e-01 8.36636245e-01 4.35268953e-02 -5.99498868e-01 -2.63088077e-01 -6.20690994e-02 4.06428933e-01 5.86092293e-01 1.21566899e-01 -9.44166064e-01 5.63065112e-01 5.55403423e+00 1.52586460e+00 -1.27388358e+00 3.56437176e-01 9.51800048e-01 -4.66995746e-01 1.73839405e-01 -5.47442436e-01 -9.65164483e-01 8.39206398e-01 1.66936243e+00 2.18981609e-01 4.97814566e-01 6.17784023e-01 -4.36851531e-02 4.43897575e-01 -8.86464119e-01 1.45548606e+00 8.70824307e-02 -1.14950180e+00 -1.91454709e-01 -4.82492261e-02 2.95517087e-01 6.07696958e-02 1.13042504e-01 3.98224801e-01 -3.29578280e-01 -9.32950318e-01 1.07331085e+00 -5.70239462e-02 1.35622501e+00 -9.13305521e-01 9.05041575e-01 1.30408391e-01 -1.72516835e+00 -3.19368839e-01 -3.04219663e-01 -1.29519820e-01 2.04234645e-01 8.63713443e-01 -5.00138700e-01 3.07140857e-01 1.28374100e+00 3.29448700e-01 -3.93276960e-01 1.06322098e+00 3.44252020e-01 8.49401236e-01 -7.79583275e-01 5.28760962e-02 -3.34696591e-01 4.66173589e-01 9.54642445e-02 1.23500156e+00 4.72552150e-01 -2.24544760e-02 -1.76230639e-01 1.83677599e-01 -4.66611773e-01 2.65194118e-01 -1.93811487e-02 3.41732949e-01 9.11159337e-01 9.16950881e-01 -2.81192929e-01 -2.91422993e-01 -2.24274293e-01 8.29798639e-01 2.33610924e-02 -1.68530585e-03 -1.27291358e+00 -6.11031532e-01 6.29912317e-01 1.11340946e-02 2.35422298e-01 2.39771116e-03 -3.20853680e-01 -8.53488386e-01 4.31788206e-01 -9.51101124e-01 -9.44098085e-03 -6.30066335e-01 -9.01381314e-01 1.00184464e+00 -1.63672432e-01 -1.30654562e+00 6.72149435e-02 -1.50486410e-01 -2.52783149e-01 7.66998708e-01 -1.29425359e+00 -9.22697127e-01 -4.51410472e-01 6.20434225e-01 5.63500345e-01 -1.39364034e-01 9.08602595e-01 1.24773860e+00 -7.35254586e-01 1.19133425e+00 -1.92649718e-02 -6.94523975e-02 4.90569532e-01 -6.13876641e-01 9.01024520e-01 6.04103625e-01 -6.57136515e-02 1.92089036e-01 3.55349630e-01 -4.07199144e-01 -1.18758106e+00 -1.28461301e+00 6.78138375e-01 2.69519955e-01 5.19263983e-01 -9.92833793e-01 -8.76275003e-01 2.34013632e-01 -1.17512368e-01 1.13097720e-01 8.33175123e-01 -7.01815560e-02 -4.49366212e-01 -5.54938674e-01 -9.63516176e-01 5.49912870e-01 1.07577538e+00 -7.23675907e-01 1.30843788e-01 2.38129318e-01 1.45256996e+00 -5.79707086e-01 -9.34345126e-01 6.32113993e-01 6.39839768e-01 -5.91235995e-01 9.45647776e-01 -2.25435749e-01 1.56159580e-01 -1.97052076e-01 -5.59909105e-01 -9.02108312e-01 6.87927604e-02 -7.44403362e-01 -3.08428675e-01 1.61449277e+00 4.15574223e-01 -3.37378174e-01 8.15725505e-01 4.19597059e-01 -3.90071839e-01 -8.37703049e-01 -1.37143004e+00 -8.07062447e-01 -3.47770929e-01 -9.13189828e-01 1.13397384e+00 7.00893223e-01 -2.83273578e-01 3.53624314e-01 -8.03414583e-01 1.74657255e-01 1.77817062e-01 -5.62727094e-01 7.13792682e-01 -1.10463297e+00 -4.87000346e-01 -1.35535344e-01 -3.85551870e-01 -1.39824069e+00 -1.09193325e-01 -3.08993459e-01 4.27203812e-02 -7.85601676e-01 -1.22305073e-01 -8.33702743e-01 -4.86923933e-01 5.20672858e-01 3.19676876e-01 4.50873464e-01 3.88207465e-01 1.77336022e-01 -4.76451784e-01 6.83355033e-01 6.59446001e-01 -2.11207643e-01 -4.18181092e-01 6.55977577e-02 -6.51504159e-01 7.66687572e-01 1.11695194e+00 -5.32043695e-01 -3.41214538e-01 -6.42458081e-01 2.61159986e-01 9.44016222e-03 6.51707873e-02 -1.47692955e+00 3.16265553e-01 2.39880234e-01 1.16433457e-01 -6.05967283e-01 7.93053925e-01 -9.96082067e-01 4.73752081e-01 3.66066515e-01 -2.88965672e-01 1.74905330e-01 7.41875708e-01 4.84640121e-01 -1.54510424e-01 -8.78017098e-02 5.21739125e-01 4.30739760e-01 -2.60548592e-01 1.45697773e-01 -1.65951446e-01 -1.51175112e-01 6.15002394e-01 -3.30760479e-02 -4.41578060e-01 -1.72290161e-01 -5.58624446e-01 -1.58262998e-01 -2.22492561e-01 4.00502026e-01 6.81684136e-01 -1.49400926e+00 -6.02328002e-01 4.69862401e-01 1.71654262e-02 3.45850825e-01 7.50521898e-01 6.64588928e-01 -5.56541026e-01 2.50608146e-01 8.37064013e-02 -3.43199193e-01 -1.48496711e+00 2.30956569e-01 2.18103036e-01 -1.87692404e-01 -5.96978605e-01 1.01808524e+00 1.81141749e-01 -1.51372269e-01 7.67060339e-01 -6.56084716e-01 -1.67066857e-01 -1.62037872e-02 8.85882974e-01 5.49015999e-01 6.34018660e-01 -5.74449599e-01 -3.55262935e-01 5.14248967e-01 -1.81364924e-01 7.36449808e-02 1.28846192e+00 -1.62706017e-01 1.62038177e-01 2.77203888e-01 1.27044725e+00 5.92682771e-02 -1.05571902e+00 -2.24011168e-01 -6.28593445e-01 -3.05106580e-01 1.66828528e-01 -7.09731460e-01 -1.31517279e+00 9.64178443e-01 1.03974044e+00 3.59597564e-01 1.68843710e+00 -1.31855816e-01 1.30388165e+00 1.92308083e-01 1.64756313e-01 -1.12019527e+00 -5.31389043e-02 5.13806343e-01 8.94212544e-01 -6.08219564e-01 -2.05101430e-01 -6.30820274e-01 -1.92109436e-01 6.47736728e-01 7.23929167e-01 2.93376267e-01 7.93638468e-01 7.14277804e-01 -2.43570372e-01 3.52423668e-01 -5.71966171e-01 2.92123407e-01 1.06086321e-01 3.66465956e-01 1.21180452e-01 5.21550775e-02 1.48695722e-01 1.14141071e+00 -5.51724792e-01 -2.28362948e-01 3.79285365e-01 7.73339212e-01 -1.95321858e-01 -8.13922167e-01 -4.43339139e-01 6.28555655e-01 -6.86870158e-01 -5.33818901e-01 4.57778573e-02 1.57296300e-01 2.73524314e-01 1.17051983e+00 3.06080997e-01 -1.31886482e+00 5.67648649e-01 -2.98303336e-01 -1.57756489e-02 -4.03942525e-01 -8.09946597e-01 3.52731854e-01 5.06448373e-02 -4.37306851e-01 -1.62775919e-01 -1.91809878e-01 -1.10399091e+00 -5.39912522e-01 -7.35170245e-01 -1.12827741e-01 9.40320730e-01 5.85415602e-01 6.87750757e-01 9.91106272e-01 6.69210494e-01 -7.58864880e-01 -4.97203171e-01 -1.14639342e+00 -7.38619268e-01 5.04230671e-02 4.02927846e-01 -2.66199201e-01 -4.99808639e-01 1.35180326e-02]

[15.050661087036133, 5.313083171844482]

d367a971-d979-42bb-87a2-d18e9e92408a

is-bert-robust-to-label-noise-a-study-on

2204.09371

https://arxiv.org/abs/2204.09371v1

https://arxiv.org/pdf/2204.09371v1.pdf

Is BERT Robust to Label Noise? A Study on Learning with Noisy Labels in Text Classification

Incorrect labels in training data occur when human annotators make mistakes or when the data is generated via weak or distant supervision. It has been shown that complex noise-handling techniques - by modeling, cleaning or filtering the noisy instances - are required to prevent models from fitting this label noise. However, we show in this work that, for text classification tasks with modern NLP models like BERT, over a variety of noise types, existing noisehandling methods do not always improve its performance, and may even deteriorate it, suggesting the need for further investigation. We also back our observations with a comprehensive analysis.

['Dietrich Klakow', 'David Ifeoluwa Adelani', 'Fangzhou Zhai', 'Michael A. Hedderich', 'Dawei Zhu']

2022-04-20

https://aclanthology.org/2022.insights-1.8

https://aclanthology.org/2022.insights-1.8.pdf

insights-acl-2022-5

['learning-with-noisy-labels', 'learning-with-noisy-labels']

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

[ 3.50369066e-01 2.99413711e-01 -1.06219448e-01 -6.53489113e-01 -6.79130495e-01 -7.72265971e-01 5.35201669e-01 5.75712264e-01 -5.96286833e-01 9.26994085e-01 -1.73500087e-02 -5.30981123e-01 -1.37730300e-01 -4.39994007e-01 -5.87774813e-01 -5.37755609e-01 4.73900855e-01 7.06993341e-01 -7.75779188e-02 -7.95027055e-03 1.13585241e-01 3.02464843e-01 -1.56460392e+00 1.74804837e-01 8.36720586e-01 7.37917006e-01 -1.69519231e-01 5.91568470e-01 -6.37144446e-01 9.39077139e-01 -8.39001417e-01 -7.75154710e-01 2.54707068e-01 -1.85553342e-01 -1.10069680e+00 4.07664329e-02 4.57718015e-01 1.91783473e-01 1.03352658e-01 1.25678194e+00 3.72496575e-01 1.72519341e-01 4.98404473e-01 -1.21570754e+00 -4.65944886e-01 9.03387487e-01 -7.41264224e-02 1.77363213e-02 2.79447556e-01 -2.33611718e-01 9.07097220e-01 -9.23270106e-01 7.04569042e-01 1.20055056e+00 1.23584282e+00 7.34239519e-01 -1.60129488e+00 -5.64363599e-01 3.37857842e-01 4.03408669e-02 -1.30338371e+00 -3.64515096e-01 5.13589680e-01 -3.10682297e-01 9.00585592e-01 5.08607090e-01 5.80452159e-02 1.34707642e+00 -5.43740615e-02 7.10940123e-01 1.28985775e+00 -8.61789405e-01 2.24524379e-01 4.09796774e-01 5.37796378e-01 3.39024037e-01 5.63286602e-01 -1.56972483e-01 -4.30891931e-01 -5.18415868e-01 8.41336474e-02 -2.29131460e-01 -2.96733320e-01 -1.73978284e-02 -7.17350841e-01 6.01631522e-01 7.08258301e-02 4.82780635e-01 -8.41796026e-02 -8.24921355e-02 5.83022594e-01 6.47577226e-01 7.76055753e-01 7.56850064e-01 -1.04426003e+00 -2.88950264e-01 -9.56379831e-01 2.42456332e-01 1.29179156e+00 1.04724264e+00 7.76476145e-01 -2.80081898e-01 -6.96784183e-02 8.61658216e-01 5.44174667e-03 2.27191418e-01 3.50992352e-01 -9.79477704e-01 5.26406109e-01 5.42739511e-01 4.57366943e-01 -9.07577395e-01 -8.00229788e-01 -3.89922589e-01 -9.05676305e-01 4.24084030e-02 8.08137238e-01 -2.19593987e-01 -1.06766760e+00 1.68703413e+00 1.72018170e-01 -4.34866659e-02 -9.65791941e-02 5.86525202e-01 5.94171822e-01 4.00473289e-02 5.78204572e-01 -5.41461647e-01 1.06723595e+00 -7.70811677e-01 -1.42814720e+00 -4.65347260e-01 1.06787086e+00 -1.03893828e+00 1.13916695e+00 9.11380351e-01 -8.00692797e-01 -4.40032780e-01 -6.80114269e-01 -2.22114578e-01 -6.14769995e-01 -8.98097530e-02 6.48410380e-01 7.95838594e-01 -6.38497293e-01 1.00186205e+00 -6.50476575e-01 -4.30525839e-01 3.60196173e-01 1.30263135e-01 -4.69041139e-01 -3.48032564e-01 -1.19136548e+00 1.17195773e+00 3.45014274e-01 2.75850534e-01 -3.50551158e-01 -5.20717323e-01 -7.84045100e-01 -4.64101322e-03 7.32192397e-01 -4.39960361e-01 1.41254759e+00 -8.99366021e-01 -9.95289385e-01 8.35941076e-01 -3.03197384e-01 -3.43998820e-01 7.46187568e-01 -5.53674996e-01 -4.19485182e-01 -4.41768527e-01 1.21265128e-01 2.11010665e-01 5.20404458e-01 -1.49001813e+00 -4.49344248e-01 -4.75832433e-01 -2.62786150e-01 -5.41592687e-02 -2.11909786e-01 2.73398459e-01 -3.80120985e-02 -8.95859480e-01 3.69219810e-01 -8.90816689e-01 -5.52854657e-01 -3.77895206e-01 -6.60235405e-01 -4.97844845e-01 6.13281727e-01 -4.58854049e-01 1.39077437e+00 -2.02036619e+00 -1.68250978e-01 4.13291693e-01 6.98611364e-02 4.41479653e-01 5.24080321e-02 3.65871608e-01 -4.13830638e-01 8.63079369e-01 -2.68610477e-01 -8.68743062e-01 1.03749856e-01 8.82813096e-01 -1.21606357e-01 3.59958291e-01 2.02042550e-01 5.88204205e-01 -8.61080170e-01 -4.37000334e-01 7.72828311e-02 3.15416366e-01 -5.46653606e-02 7.68648982e-02 -3.11568707e-01 2.62116581e-01 -6.69387132e-02 4.57658112e-01 7.32588947e-01 -2.51132309e-01 1.56518698e-01 9.73847806e-02 1.99198693e-01 5.28492093e-01 -1.56464946e+00 1.38082910e+00 -3.37533027e-01 5.44888258e-01 3.39277327e-01 -1.03001237e+00 6.21978998e-01 5.12271047e-01 1.20968208e-01 -2.75861740e-01 2.54762173e-01 3.17428291e-01 -2.04386890e-01 -8.55501115e-01 3.01130891e-01 -3.08030456e-01 -5.98962978e-02 3.29741240e-01 1.35386556e-01 -4.97486293e-02 3.68035555e-01 2.76343264e-02 1.28367364e+00 -1.26777992e-01 2.50158280e-01 -1.81488857e-01 3.18979025e-01 2.67087936e-01 8.33912432e-01 1.34899223e+00 -3.72309506e-01 7.26396799e-01 5.52958727e-01 -4.30117399e-01 -9.89419699e-01 -3.00821602e-01 -5.85230649e-01 1.16450131e+00 -4.88397688e-01 -6.41836941e-01 -8.71324003e-01 -1.23545313e+00 -6.77938089e-02 9.09969747e-01 -5.71008623e-01 4.71189767e-02 -4.61913615e-01 -1.01507711e+00 6.14944696e-01 5.50224245e-01 -2.58928984e-02 -9.22189832e-01 1.65986583e-01 4.28494960e-01 -2.31570169e-01 -1.29704320e+00 2.45176945e-02 1.10936725e+00 -7.49867380e-01 -1.25808442e+00 -1.16573721e-01 -6.48212671e-01 8.64283025e-01 1.35752670e-02 1.42747712e+00 6.74722791e-01 3.11374851e-03 6.29732385e-02 -6.45849824e-01 -6.01256430e-01 -7.95774102e-01 1.32340252e-01 1.60255089e-01 -5.37693501e-01 9.24682140e-01 -3.74459326e-01 1.43909141e-01 3.84055853e-01 -1.03522766e+00 -3.26222211e-01 1.73308358e-01 1.10803974e+00 5.40753007e-01 4.94683117e-01 5.33953846e-01 -1.59693682e+00 7.71444261e-01 -3.60246181e-01 -3.44224334e-01 3.01735103e-01 -8.89975488e-01 1.29688248e-01 8.81812930e-01 -4.55841482e-01 -1.03045380e+00 1.47133306e-01 -4.21619028e-01 -3.71863507e-02 -7.86782980e-01 5.32417536e-01 -1.62368864e-01 4.60447632e-02 1.25699174e+00 -2.89952785e-01 -2.82034010e-01 -8.30585599e-01 2.22354919e-01 8.52139235e-01 1.64024755e-01 -4.15665120e-01 6.87594175e-01 3.18951577e-01 -1.33462667e-01 -3.65710288e-01 -1.63800740e+00 -5.83637476e-01 -7.12276638e-01 2.37418488e-01 4.99480009e-01 -7.27315307e-01 -2.95645475e-01 2.65100658e-01 -1.22882283e+00 -2.95977563e-01 -3.28071237e-01 2.89098680e-01 -2.06244230e-01 3.85837108e-01 -6.90546632e-01 -1.05255485e+00 -3.68130989e-02 -8.92529845e-01 7.87189186e-01 -4.33445424e-02 -6.17191613e-01 -1.07783508e+00 -1.83908150e-01 3.91639531e-01 2.33767062e-01 -6.60616606e-02 8.20083141e-01 -1.27757442e+00 1.14113018e-01 -3.88940096e-01 -2.05475334e-02 7.57507682e-01 1.93994895e-01 -6.17385730e-02 -1.24588168e+00 5.37781082e-02 2.23248273e-01 -4.01318610e-01 7.93892264e-01 2.11669207e-01 1.42867172e+00 -2.39601046e-01 -2.88413107e-01 1.72511667e-01 1.30856764e+00 -2.10470915e-01 3.81956369e-01 4.30457830e-01 7.01631308e-01 7.80519783e-01 6.72351480e-01 2.02601373e-01 -4.67846096e-02 3.07549417e-01 2.46202186e-01 -8.16843286e-02 9.93654057e-02 -1.91310316e-01 -2.20709354e-01 4.38109428e-01 1.49021178e-01 -3.79090548e-01 -9.88245189e-01 4.33342665e-01 -2.05844402e+00 -6.20514691e-01 -7.82705247e-01 1.92597544e+00 1.21673310e+00 3.46238434e-01 -2.43367642e-01 7.45674193e-01 6.26884699e-01 -3.49858582e-01 -1.02968782e-01 -3.18944424e-01 -2.11870313e-01 1.32787332e-01 5.96930087e-01 7.06695318e-01 -1.19167542e+00 9.02201176e-01 7.40903664e+00 8.69161189e-01 -4.20673549e-01 2.52735049e-01 6.36469722e-01 1.10009603e-01 -2.72495955e-01 -3.67332958e-02 -8.03955674e-01 3.63357037e-01 8.82726789e-01 2.94099778e-01 3.61777991e-01 9.62748289e-01 3.32847834e-01 -2.09879354e-01 -1.16022599e+00 8.35926771e-01 3.36725973e-02 -6.98374748e-01 -4.35841441e-01 -3.34455997e-01 6.57460451e-01 -1.55491441e-01 -4.50383484e-01 4.64295715e-01 7.25461602e-01 -1.19375646e+00 4.85678822e-01 3.02394986e-01 2.78836451e-02 -7.42637813e-01 1.31767952e+00 7.86415637e-01 -4.23188508e-01 -9.20609310e-02 -4.95803267e-01 -3.31918418e-01 1.88501760e-01 1.31873441e+00 -7.30966806e-01 3.99865061e-01 8.46710443e-01 3.20314705e-01 -7.92812169e-01 1.04809558e+00 -5.18502235e-01 9.55883086e-01 -4.12052900e-01 2.11888984e-01 -5.50757870e-02 1.85080543e-01 3.30508173e-01 1.41162848e+00 -5.51279970e-02 -1.29775414e-02 2.23401293e-01 5.00607848e-01 -1.13310270e-01 1.24446511e-01 -6.84217632e-01 2.39488795e-01 4.47274297e-01 1.17109883e+00 -6.94032788e-01 -4.90463763e-01 -3.12506378e-01 6.87671244e-01 5.52696466e-01 4.67289597e-01 -4.13673997e-01 -1.63990960e-01 4.86429811e-01 -1.32004796e-02 -1.09059647e-01 -5.37087172e-02 -8.92183423e-01 -1.09393680e+00 1.77194804e-01 -9.51400399e-01 4.09929752e-01 -6.75313354e-01 -1.73837316e+00 4.71147597e-01 -2.93882638e-01 -7.49135077e-01 -3.34455147e-02 -7.56527424e-01 -2.02279150e-01 6.23951137e-01 -1.35199499e+00 -6.01245701e-01 -1.31732196e-01 2.39254445e-01 3.86803836e-01 2.43313566e-01 9.75424349e-01 3.30220580e-01 -6.90996826e-01 4.50221896e-01 -1.65446326e-02 8.60793889e-02 9.88264561e-01 -1.69284797e+00 3.34722996e-01 7.74239063e-01 2.55381793e-01 5.90099275e-01 1.29783630e+00 -7.68285394e-01 -7.00405121e-01 -1.09015119e+00 1.58335650e+00 -7.80804932e-01 6.56484783e-01 -6.63272560e-01 -1.46866095e+00 8.34767461e-01 1.60564557e-01 1.56575531e-01 9.80636954e-01 5.82105458e-01 -3.85590702e-01 2.91228712e-01 -1.31863725e+00 2.91358948e-01 1.30152392e+00 -3.60170782e-01 -7.57572353e-01 8.07288349e-01 5.65065622e-01 -3.32182944e-01 -8.04641485e-01 3.45177323e-01 -2.39544977e-02 -7.03135014e-01 4.58411306e-01 -1.00902414e+00 -5.95541783e-02 -2.02254340e-01 -6.31243456e-03 -1.53554010e+00 -2.52553821e-01 -5.37512839e-01 1.15537532e-01 1.49625003e+00 7.59052873e-01 -2.78711289e-01 7.44117737e-01 1.22145092e+00 -4.79092114e-02 -1.22002058e-01 -9.65957940e-01 -8.15581322e-01 1.19496159e-01 -1.17008865e+00 2.85129786e-01 1.32750869e+00 1.28919572e-01 2.83267289e-01 -4.19925928e-01 1.21003784e-01 4.50115830e-01 -6.07391596e-01 5.64747453e-01 -1.63681769e+00 2.24534303e-01 -1.67702392e-01 -1.32184792e-02 -7.48070180e-01 4.61431772e-01 -5.88182569e-01 4.08999324e-01 -1.48076928e+00 -1.15208857e-01 -8.31130922e-01 -1.56864434e-01 9.04344261e-01 -4.99768645e-01 2.14898065e-01 -1.54297307e-01 6.68772087e-02 -7.56601930e-01 8.37115273e-02 9.94073987e-01 7.98233524e-02 1.13116518e-01 2.25004479e-01 -9.75043476e-01 9.31090415e-01 8.23887706e-01 -1.20045197e+00 -6.62836879e-02 -3.18592370e-01 8.08856249e-01 -4.77789313e-01 2.35719115e-01 -6.95394814e-01 2.38417044e-01 -9.68815237e-02 3.64703745e-01 -2.80721128e-01 -8.50745812e-02 -1.40580070e+00 -7.17036501e-02 1.32451624e-01 -5.90094507e-01 3.31824156e-03 1.43083602e-01 6.89517796e-01 -1.51971668e-01 -9.33800757e-01 4.92822289e-01 -1.96029484e-01 -1.15008980e-01 -2.66042650e-01 -6.09184206e-01 2.06121355e-01 5.94394386e-01 2.25550011e-01 -2.54080266e-01 -1.31670222e-01 -1.39168453e+00 2.17541620e-01 4.50381100e-01 3.90644521e-01 -1.03073753e-01 -1.13378978e+00 -5.98920226e-01 1.52941808e-01 1.98127106e-02 3.62144411e-01 -3.67142893e-02 7.68430650e-01 -1.03926986e-01 2.34287068e-01 6.04915202e-01 -4.43992645e-01 -1.38338459e+00 5.70449054e-01 2.77355760e-01 -4.04957920e-01 -3.93532932e-01 8.14073801e-01 -5.39499104e-01 -7.49210238e-01 7.08815634e-01 -5.90878308e-01 -2.29259551e-01 9.75465924e-02 5.57401836e-01 3.39850605e-01 6.44720256e-01 -3.51923764e-01 -1.76927522e-01 1.61775336e-01 -1.57657295e-01 3.29173982e-01 1.21072888e+00 -2.65795052e-01 -1.91232055e-01 6.47553980e-01 7.24449933e-01 -7.77326641e-04 -7.15411067e-01 -4.12468046e-01 5.75817943e-01 -5.01894712e-01 1.00948341e-01 -1.05870879e+00 -6.46650553e-01 6.28146052e-01 2.70229489e-01 9.66952980e-01 8.71698737e-01 -5.75548187e-02 2.31376141e-01 7.83116639e-01 1.55563965e-01 -1.55904424e+00 -3.84937704e-01 7.11870790e-01 6.00951910e-01 -1.76529956e+00 -1.82345614e-01 -6.74417734e-01 -4.68458712e-01 8.93382072e-01 6.13215387e-01 2.59272665e-01 8.11602652e-01 7.27598071e-01 4.90148991e-01 -1.23039462e-01 -9.01749790e-01 -3.01807851e-01 -7.33443350e-02 6.97029591e-01 5.54473519e-01 -1.06660970e-01 -4.01846141e-01 7.75672019e-01 -1.80375159e-01 4.64777164e-02 6.03049755e-01 1.05595899e+00 -4.72929686e-01 -1.44130921e+00 -6.50304377e-01 6.18824303e-01 -8.10695291e-01 -2.09797531e-01 -6.01236582e-01 8.18647921e-01 4.60460931e-01 1.48606920e+00 -1.60083205e-01 -1.01804547e-01 6.47879839e-01 7.46261477e-01 1.08595408e-01 -9.17774498e-01 -9.66607213e-01 -1.18068837e-01 4.88283217e-01 -3.35757345e-01 -6.03551507e-01 -6.32139504e-01 -1.05251098e+00 -3.19111586e-01 -9.35915947e-01 3.26960355e-01 7.15307176e-01 1.31910908e+00 9.58328843e-02 3.76871377e-01 1.37831450e-01 -3.06489110e-01 -7.35937834e-01 -1.48830271e+00 -5.79433978e-01 7.65139520e-01 3.25046360e-01 -6.33213520e-01 -1.02509272e+00 1.26409993e-01]

[9.548697471618652, 4.395499229431152]

6165305f-1ce8-4d2c-b853-ca465d448d15

a-mobile-food-recognition-system-for-dietary

2204.09432

https://arxiv.org/abs/2204.09432v1

https://arxiv.org/pdf/2204.09432v1.pdf

A Mobile Food Recognition System for Dietary Assessment

Food recognition is an important task for a variety of applications, including managing health conditions and assisting visually impaired people. Several food recognition studies have focused on generic types of food or specific cuisines, however, food recognition with respect to Middle Eastern cuisines has remained unexplored. Therefore, in this paper we focus on developing a mobile friendly, Middle Eastern cuisine focused food recognition application for assisted living purposes. In order to enable a low-latency, high-accuracy food classification system, we opted to utilize the Mobilenet-v2 deep learning model. As some of the foods are more popular than the others, the number of samples per class in the used Middle Eastern food dataset is relatively imbalanced. To compensate for this problem, data augmentation methods are applied on the underrepresented classes. Experimental results show that using Mobilenet-v2 architecture for this task is beneficial in terms of both accuracy and the memory usage. With the model achieving 94% accuracy on 23 food classes, the developed mobile application has potential to serve the visually impaired in automatic food recognition via images.

['Hazim Kemal Ekenel', 'Marwa Qaraqe', 'Şeymanur Aktı']

2022-04-20

['food-recognition']

['computer-vision']

[ 1.85061708e-01 -2.84640342e-01 -2.13348255e-01 -2.31609374e-01 -2.58049935e-01 -1.51981309e-01 -5.22307381e-02 6.72300041e-01 -6.17926419e-01 4.58299577e-01 1.03695668e-01 -1.80542171e-01 1.17219679e-01 -9.55100715e-01 -5.27200222e-01 -7.08746076e-01 -6.20769821e-02 -6.87480420e-02 -1.79370835e-01 -9.73141044e-02 -1.40538169e-02 1.57696996e-02 -1.82849979e+00 2.68313169e-01 1.05784249e+00 9.49360251e-01 5.71253419e-01 5.38068175e-01 -3.03188115e-01 5.57769418e-01 -3.38777095e-01 -4.01946679e-02 1.81123942e-01 -3.23905587e-01 -2.03441858e-01 -1.64893895e-01 3.73597234e-01 -4.53661978e-01 1.23948686e-01 1.20654106e+00 6.28702879e-01 -5.02195545e-02 6.58496380e-01 -1.13410974e+00 -8.41995358e-01 5.82090735e-01 -4.23259139e-01 6.65815696e-02 3.82456273e-01 -2.00630748e-03 4.76613343e-01 -6.55058265e-01 -2.16707475e-02 9.22049880e-01 9.19174194e-01 3.64102632e-01 -7.46461093e-01 -7.02172160e-01 1.72110066e-01 2.96303928e-01 -1.32820737e+00 -2.52171576e-01 5.84961832e-01 -4.82815415e-01 1.02486384e+00 1.30629793e-01 1.01836252e+00 8.07833076e-01 1.83134109e-01 8.85793746e-01 8.13422680e-01 -5.85430324e-01 9.19108912e-02 3.69674861e-01 5.02675116e-01 5.22368550e-01 7.61505246e-01 -1.50942281e-01 -2.18705893e-01 3.16025466e-01 2.45742977e-01 2.44580567e-01 -5.88111952e-02 -3.56145054e-01 -1.09907937e+00 1.07054639e+00 8.88231158e-01 4.65734333e-01 -6.64809048e-01 -4.96217646e-02 4.89189774e-01 1.08166978e-01 5.12225568e-01 2.38104369e-02 -5.94983816e-01 1.32384151e-01 -5.57566285e-01 3.91496113e-03 6.96005404e-01 7.44284630e-01 3.34094286e-01 1.64800465e-01 1.29505411e-01 8.69099140e-01 6.63180470e-01 8.83093417e-01 5.88800073e-01 6.28567487e-02 4.61044878e-01 9.62355733e-01 1.42403916e-01 -1.22526348e+00 -8.11639249e-01 -5.59146583e-01 -1.06167781e+00 1.44660696e-01 4.96778190e-01 -5.18746004e-02 -1.09498215e+00 1.47198987e+00 3.69188756e-01 -4.05919313e-01 3.55099648e-01 1.02945805e+00 1.19944990e+00 4.51634735e-01 8.10150981e-01 1.02387622e-01 1.73291492e+00 -8.82614434e-01 -6.36161268e-01 -3.25666130e-01 8.58448148e-01 -8.62798095e-01 1.21063292e+00 2.18019828e-01 -4.62721467e-01 -6.42122328e-01 -1.30935681e+00 -3.40781771e-02 -8.17339003e-01 5.63366830e-01 7.84403920e-01 1.09131122e+00 -6.96961999e-01 1.40021667e-01 -8.13814819e-01 -9.53828275e-01 5.50189972e-01 4.51087862e-01 -2.25788683e-01 -4.10972327e-01 -1.08917654e+00 8.12717080e-01 5.49342394e-01 2.54120439e-01 -2.38114819e-01 -4.75966990e-01 -1.03165877e+00 7.41990581e-02 -9.33082178e-02 -4.63135809e-01 9.12775934e-01 -1.24922407e+00 -1.09465647e+00 6.65262163e-01 3.76404703e-01 -4.65675771e-01 1.66922659e-01 -3.01899731e-01 -7.04987109e-01 -1.15647368e-01 1.77852184e-01 9.23046887e-01 2.93036968e-01 -5.45121968e-01 -8.38163316e-01 -6.43738687e-01 -1.23547725e-01 3.78680915e-01 -5.62229574e-01 -1.55777052e-01 7.71248266e-02 -3.94766688e-01 7.64559656e-02 -9.30663943e-01 5.98221421e-02 1.76338866e-01 -1.21975936e-01 -1.01460733e-01 6.00764215e-01 -9.32278693e-01 1.19687891e+00 -2.19157386e+00 -5.78721941e-01 8.61400831e-03 -2.47572258e-01 6.40683115e-01 -1.33047327e-01 -1.28424048e-01 8.94373059e-02 -2.50466377e-01 5.54791465e-02 2.35331297e-01 -1.31311864e-01 -1.04352988e-01 6.75861478e-01 5.52323699e-01 1.90940991e-01 6.72681868e-01 -7.53929079e-01 -2.64787436e-01 5.17597079e-01 8.12344790e-01 -3.70214432e-01 -2.26091623e-01 -1.39675945e-01 2.22669467e-02 -5.82231283e-01 1.05454898e+00 9.26284671e-01 -2.07419004e-02 5.25263660e-02 -5.46638787e-01 -2.40027964e-01 -1.03378080e-01 -1.13566709e+00 1.66584873e+00 -3.37837219e-01 1.02837339e-01 5.94510473e-02 -8.96066964e-01 7.91271627e-01 1.92550972e-01 5.17609358e-01 -1.05009186e+00 5.03806353e-01 4.27855223e-01 1.71335265e-01 -8.90661001e-01 3.82823676e-01 2.04898000e-01 2.18833596e-01 -4.66312058e-02 -2.82543898e-01 7.46251583e-01 3.65686119e-01 -3.85013372e-01 4.73519683e-01 3.76622707e-01 6.33004069e-01 -5.17410815e-01 3.99797827e-01 4.51309919e-01 1.63969502e-01 3.54193538e-01 -6.25565410e-01 2.07872570e-01 -5.10296881e-01 -6.29209638e-01 -9.13867235e-01 -5.75744033e-01 -1.31314829e-01 1.03983200e+00 -4.15338622e-03 -2.93054320e-02 -8.21366072e-01 -4.50663477e-01 -1.12616852e-01 5.41369438e-01 -3.56169552e-01 -1.77272156e-01 -3.17708820e-01 -1.12194550e+00 2.03510061e-01 6.42082691e-01 1.11592078e+00 -1.15423429e+00 -1.27547050e+00 5.50463319e-01 -2.51582950e-01 -5.63179076e-01 -5.79403698e-01 3.15320283e-01 -6.12248421e-01 -1.46162438e+00 -9.96645272e-01 -1.18369198e+00 1.73077658e-01 8.07367086e-01 1.04042161e+00 -3.06544136e-02 -3.14878851e-01 2.55907416e-01 -6.23051941e-01 -9.47950959e-01 -3.17322046e-01 4.79969591e-01 -2.27528259e-01 -5.51303774e-02 1.35274076e+00 9.51469243e-02 -9.96560514e-01 -3.53059620e-02 -7.70794511e-01 2.68597566e-02 4.62721080e-01 6.18376851e-01 4.17293340e-01 -8.63645375e-02 8.00895810e-01 -4.66842920e-01 2.64118075e-01 -8.00884843e-01 -3.64158362e-01 2.40909129e-01 -5.99562347e-01 -1.81669697e-01 3.45220596e-01 -8.70111167e-01 -6.40743434e-01 6.40326202e-01 -4.62114155e-01 4.63977873e-01 -2.67752260e-01 6.92591190e-01 -2.77426392e-01 3.41012739e-02 6.93033814e-01 2.28634048e-02 1.52083412e-01 -4.23732072e-01 -1.21830963e-02 9.43423688e-01 1.26575559e-01 1.07154064e-01 -1.63984269e-01 -1.82318643e-01 -1.87216535e-01 -1.12802994e+00 -5.06598830e-01 -6.13317430e-01 -1.80381075e-01 -3.30450833e-01 1.21397710e+00 -1.26872873e+00 -7.94475257e-01 7.27231741e-01 -6.78310812e-01 -7.10272565e-02 1.97678059e-01 8.56879771e-01 3.99691090e-02 3.60153839e-02 -1.78219885e-01 -9.18748498e-01 -1.00038397e+00 -1.15883076e+00 7.96493888e-01 4.41537917e-01 -3.46158743e-01 -6.50406241e-01 -2.19143614e-01 1.78835869e-01 5.85782707e-01 4.11097437e-01 9.49083209e-01 -4.42479163e-01 -1.73082039e-01 -1.90871000e-01 -3.82877082e-01 1.32768556e-01 4.24576819e-01 -3.60966176e-01 -8.62435520e-01 -1.45231292e-01 -2.48770058e-01 -1.62842423e-01 7.10333586e-01 5.56635678e-01 4.22159612e-01 -2.62733642e-02 -3.58657002e-01 2.98460990e-01 1.60579669e+00 7.38343060e-01 5.73618114e-01 6.46959424e-01 8.16392183e-01 3.88227046e-01 7.28326261e-01 1.71590075e-01 6.89971805e-01 5.12757540e-01 6.94470167e-01 -2.79641420e-01 -4.11980659e-01 -7.67051205e-02 3.59120101e-01 5.17679930e-01 5.44508696e-01 -2.39798665e-01 -6.89009190e-01 8.72991264e-01 -1.62130094e+00 -6.32307351e-01 -5.32062352e-01 2.12249589e+00 3.98665190e-01 -4.05273706e-01 4.05097604e-01 5.18725216e-01 7.23841727e-01 -4.34233010e-01 -8.46277714e-01 -1.74069002e-01 -1.57993838e-01 4.72814403e-02 6.08376086e-01 -4.83505428e-02 -1.52849329e+00 4.43952411e-01 5.31308031e+00 4.51989681e-01 -1.33069670e+00 8.38476717e-02 5.96915483e-01 7.99854770e-02 3.50140750e-01 -9.47752535e-01 -6.98529661e-01 5.49301207e-01 9.99749780e-01 6.53330863e-01 3.29779476e-01 9.00133014e-01 1.25893369e-01 -2.87324190e-01 -7.10400939e-01 1.17597485e+00 1.38835251e-01 -7.72435486e-01 -2.59160906e-01 -2.15871081e-01 2.21859083e-01 1.54269546e-01 -5.22270910e-02 3.05581868e-01 -7.92341754e-02 -7.81653225e-01 9.03919578e-01 1.66131124e-01 6.94134653e-01 -8.42925012e-01 8.12653542e-01 2.05615550e-01 -1.45202279e+00 -2.89259940e-01 -3.98992985e-01 -1.12718321e-01 -3.52232635e-01 4.85952556e-01 -4.83893663e-01 2.93255270e-01 1.01346827e+00 5.50793588e-01 -6.34009838e-01 1.21056223e+00 3.72684836e-01 3.47096592e-01 -7.78074026e-01 -5.12132049e-01 1.86857834e-01 -1.44934356e-01 -2.33756781e-01 1.04044282e+00 9.54263568e-01 8.54359344e-02 3.78562033e-01 3.14608544e-01 2.00957939e-01 7.86621928e-01 -6.59612417e-01 -1.00377083e-01 -1.01747721e-01 1.16344607e+00 -1.03683376e+00 -5.96971139e-02 -5.91082633e-01 7.80101061e-01 -6.16692714e-02 -1.10071585e-01 -6.23364270e-01 -2.20354363e-01 6.70825243e-01 5.52388746e-03 2.80835390e-01 4.43129018e-02 -2.64741808e-01 -1.08679020e+00 -1.22927129e-01 -9.52303827e-01 5.13048232e-01 -6.14718378e-01 -8.92437696e-01 5.64042509e-01 -3.47332895e-01 -1.08986390e+00 2.35370219e-01 -8.36663246e-01 -2.29697395e-02 7.51072407e-01 -1.49593949e+00 -1.47244489e+00 -6.17128730e-01 4.70582455e-01 6.76446319e-01 -1.42554089e-01 1.11905086e+00 9.15251672e-01 -7.21789002e-01 6.49865091e-01 6.47607371e-02 -5.47573566e-02 4.77107376e-01 -9.80098665e-01 2.35860161e-02 6.68036282e-01 8.80317483e-03 3.36627126e-01 6.57961607e-01 -7.47403324e-01 -1.44932330e+00 -1.41356504e+00 8.52817714e-01 5.80140539e-02 -2.31535658e-02 -1.27836153e-01 -6.84560657e-01 1.52671814e-01 2.42508873e-01 -3.68408680e-01 9.44659770e-01 -2.57267386e-01 -5.60836904e-02 -2.97156513e-01 -1.70463026e+00 3.12122345e-01 7.50249386e-01 -6.90443665e-02 1.25982344e-01 4.12202656e-01 3.70896161e-01 -6.26677349e-02 -7.63257682e-01 1.79055259e-01 8.06704462e-01 -6.58849537e-01 9.70043242e-01 -1.82927951e-01 1.86804473e-01 -3.66450548e-01 -3.64027411e-01 -1.33101225e+00 -4.00124907e-01 4.52226162e-01 2.67089065e-03 1.20805585e+00 3.00700366e-01 -5.54032087e-01 8.70709777e-01 5.22841096e-01 -1.46800950e-01 -3.73103619e-01 -4.28776413e-01 -3.09408426e-01 -9.16042104e-02 -1.11417525e-01 8.24054658e-01 8.11540365e-01 -2.74781585e-01 4.41717237e-01 -3.84289771e-01 2.11112738e-01 3.70665878e-01 -1.64854079e-01 3.79450649e-01 -1.30560374e+00 2.62779206e-01 -5.03149033e-02 -4.30400014e-01 -5.62156081e-01 -3.71728420e-01 -8.45391691e-01 1.49954125e-01 -1.72018313e+00 1.03421375e-01 -4.90466803e-01 -3.80793840e-01 6.23260617e-01 -2.91623026e-01 6.39896572e-01 9.31439251e-02 -2.33453020e-01 -1.46681547e-01 1.84841916e-01 9.27944481e-01 -6.25609338e-01 -3.66671175e-01 -5.44978082e-02 -9.58028495e-01 4.15877700e-01 1.06183362e+00 -2.91300505e-01 -6.57170117e-01 -5.38082600e-01 2.14301080e-01 -4.90908802e-01 2.30693087e-01 -1.37598574e+00 5.23525402e-02 7.88286552e-02 6.34685934e-01 -6.75056398e-01 -2.89472449e-03 -1.24722803e+00 5.17812252e-01 9.62509155e-01 7.37571493e-02 -1.65907275e-02 3.72096330e-01 2.07664162e-01 -3.70414630e-02 -2.06006795e-01 6.87336504e-01 -5.40878177e-02 -8.45928788e-01 2.49302946e-02 -4.55971539e-01 -5.87135434e-01 1.10016847e+00 -3.80821437e-01 -2.99156904e-01 1.67311326e-01 -5.28557301e-01 9.54411924e-02 1.13151170e-01 4.64729995e-01 4.36980516e-01 -1.47772610e+00 -5.86762726e-01 4.08523291e-01 4.92638648e-01 -3.66864502e-01 5.36594510e-01 6.87381327e-01 -9.47844207e-01 6.00873888e-01 -4.55026805e-01 -4.94526327e-01 -1.38133276e+00 7.47127414e-01 2.86819965e-01 4.40803133e-02 -6.04224443e-01 3.62574875e-01 -6.45427629e-02 -3.06366950e-01 4.05106902e-01 -8.62199008e-01 -8.00525606e-01 2.92900056e-01 8.03546607e-01 4.46618736e-01 3.20371926e-01 -6.64653838e-01 -3.97269011e-01 5.31585217e-01 4.24384736e-02 8.91229928e-01 1.39467168e+00 -3.23855102e-01 4.21569854e-01 2.12298155e-01 1.02545726e+00 -6.42069817e-01 -1.00579274e+00 1.98549524e-01 -4.24386971e-02 -1.16358660e-01 2.37164244e-01 -1.15367341e+00 -1.30838275e+00 7.39455819e-01 1.74691284e+00 2.91360438e-01 1.41594708e+00 -6.27393663e-01 8.30713868e-01 2.88506389e-01 4.51479673e-01 -8.52663100e-01 -5.30629635e-01 2.09783897e-01 4.97025996e-01 -1.67874038e+00 -3.04470539e-01 -4.19624031e-01 -2.70711482e-01 9.40208614e-01 4.47581470e-01 9.12422165e-02 6.64200783e-01 6.60193786e-02 4.57591563e-01 -4.29944992e-02 1.15599990e-01 -3.53583902e-01 1.24550089e-01 1.09903538e+00 8.21286142e-01 3.97046238e-01 -5.06599009e-01 6.02682471e-01 8.16478580e-02 4.31906253e-01 1.85223624e-01 8.93266797e-01 -5.11491239e-01 -8.89823020e-01 -5.67252219e-01 7.25684285e-01 -5.13313591e-01 -2.27877527e-01 8.86723399e-02 7.02038944e-01 7.29694545e-01 1.38865554e+00 -8.02823305e-02 -2.40827769e-01 5.17102420e-01 1.52251482e-01 4.37735528e-01 -3.14761430e-01 -9.77619171e-01 8.51981789e-02 3.04742903e-01 -1.31673217e-01 -8.74960482e-01 -5.30154467e-01 -9.15895402e-01 -3.61543953e-01 -2.88751453e-01 -2.14131549e-01 1.07925165e+00 7.91999936e-01 2.85688549e-01 5.81284285e-01 1.61927622e-02 -6.86580598e-01 -3.38863730e-01 -1.26145506e+00 -6.99119568e-01 3.77559870e-01 1.91479176e-01 -7.52205908e-01 2.59549081e-01 5.17364033e-02]

[11.563359260559082, 4.402523994445801]

3bbc780b-14e5-44c1-a414-a7653e8fd458

fusionmotion-multi-sensor-asynchronous-fusion

2302.09585

https://arxiv.org/abs/2302.09585v1

https://arxiv.org/pdf/2302.09585v1.pdf

FusionMotion: Multi-Sensor Asynchronous Fusion for Continuous Occupancy Prediction via Neural-ODE

Occupancy maps are widely recognized as an efficient method for facilitating robot motion planning in static environments. However, for intelligent vehicles, occupancy of both the present and future moments is required to ensure safe driving. In the automotive industry, the accurate and continuous prediction of future occupancy maps in traffic scenarios remains a formidable challenge. This paper investigates multi-sensor spatio-temporal fusion strategies for continuous occupancy prediction in a systematic manner. This paper presents FusionMotion, a novel bird's eye view (BEV) occupancy predictor which is capable of achieving the fusion of asynchronous multi-sensor data and predicting the future occupancy map with variable time intervals and temporal horizons. Remarkably, FusionMotion features the adoption of neural ordinary differential equations on recurrent neural networks for occupancy prediction. FusionMotion learns derivatives of BEV features over temporal horizons, updates the implicit sensor's BEV feature measurements and propagates future states for each ODE step. Extensive experiments on large-scale nuScenes and Lyft L5 datasets demonstrate that FusionMotion significantly outperforms previous methods. In addition, it outperforms the BEVFusion-style fusion strategy on the Lyft L5 dataset while reducing synchronization requirements. Codes and models will be made available.

['Diange Yang', 'Yunlong Wang', 'Jiusi Li', 'Ke Wang', 'Kun Jiang', 'Yining Shi']

2023-02-19

['motion-planning']

['robots']

[-2.00607583e-01 -1.90899402e-01 -3.93069029e-01 -3.80427718e-01 -7.38405466e-01 -1.28197208e-01 8.07807624e-01 -5.31649776e-02 -4.24696654e-01 9.69792306e-01 6.89338967e-02 -2.51687050e-01 -2.25907639e-01 -7.61104822e-01 -6.20987236e-01 -9.65110719e-01 -1.43437818e-01 4.16363955e-01 4.38046247e-01 -3.67176980e-01 2.04288781e-01 7.01864898e-01 -2.14786983e+00 -4.88307737e-02 2.59737104e-01 1.32287121e+00 7.12994874e-01 8.16337407e-01 3.23892206e-01 9.81322587e-01 -8.49677399e-02 4.53340918e-01 1.67567834e-01 1.79498658e-01 -2.65902787e-01 -5.88171370e-02 -8.58891681e-02 -3.09806734e-01 -7.24698901e-01 3.97908837e-01 2.12506786e-01 5.89615107e-01 4.83603030e-01 -1.56984437e+00 2.14196622e-01 2.59669453e-01 -5.08934438e-01 4.37431306e-01 2.47758497e-02 1.88064545e-01 5.23812830e-01 -1.03463352e+00 4.95948523e-01 9.63923812e-01 8.30822885e-01 1.74996063e-01 -7.76855111e-01 -6.57813430e-01 2.29826972e-01 6.38261676e-01 -1.60562670e+00 -7.61704206e-01 5.29866159e-01 -4.34673280e-01 1.41386783e+00 4.04377818e-01 8.26347709e-01 9.15327668e-01 1.23457313e+00 7.77178824e-01 7.26388574e-01 1.97007060e-01 4.46235299e-01 -2.01644763e-01 7.64400512e-03 3.75970006e-01 -1.24406330e-01 4.54656899e-01 -8.21854472e-01 -2.55379491e-02 4.19224650e-01 4.46562827e-01 1.77007064e-01 -3.24431747e-01 -1.26877403e+00 6.98408544e-01 1.36631086e-01 7.07020685e-02 -8.92952323e-01 6.21718049e-01 4.12350118e-01 1.36092365e-01 4.49461371e-01 2.66499072e-02 -3.10198575e-01 -3.16512167e-01 -8.31529319e-01 5.07913947e-01 5.66742837e-01 9.27156031e-01 1.09555244e+00 3.09926152e-01 -1.30916208e-01 1.65376157e-01 3.51618320e-01 9.43258762e-01 5.04256904e-01 -1.25220799e+00 1.07416563e-01 5.71775958e-02 5.74149787e-01 -9.14747059e-01 -8.06246340e-01 -1.64768517e-01 -1.07703316e+00 2.10750565e-01 -3.02607208e-01 -3.38739425e-01 -7.99770951e-01 1.44546294e+00 5.37753701e-01 6.29112124e-01 4.09607112e-01 7.35295773e-01 3.22182000e-01 1.02931142e+00 -2.95796424e-01 -4.68929410e-01 1.00194108e+00 -8.51743817e-01 -1.07670450e+00 -4.88865048e-01 6.26955569e-01 -4.39748853e-01 -1.96916550e-01 2.56129503e-01 -8.94247651e-01 -8.07345271e-01 -1.17549157e+00 6.48760721e-02 -5.30421972e-01 -1.70780271e-01 4.78970319e-01 6.65812939e-02 -1.37160659e+00 3.59099805e-01 -1.35433459e+00 -8.72209594e-02 -2.46284977e-01 5.72485209e-01 -1.96873263e-01 1.24452055e-01 -1.47948492e+00 1.13671482e+00 4.99776512e-01 3.47888976e-01 -1.11232281e+00 -4.59467351e-01 -1.25008118e+00 -2.32345819e-01 4.11893189e-01 -2.73703605e-01 1.45486665e+00 -8.48774165e-02 -1.22697496e+00 -6.38221726e-02 -6.71007991e-01 -1.03777730e+00 2.11067200e-01 -6.80415109e-02 -4.08289939e-01 -2.69491613e-01 2.58104384e-01 8.06750953e-01 7.83676863e-01 -8.53570044e-01 -1.18119693e+00 -2.21159384e-01 -3.62313688e-01 4.83023703e-01 2.23574743e-01 -8.61083686e-01 -2.16791779e-01 1.66896418e-01 1.55952901e-01 -1.12942159e+00 -6.75046980e-01 -3.93115371e-01 2.18715131e-01 -3.67506206e-01 1.16582274e+00 -4.07255858e-01 1.23498023e+00 -2.17529345e+00 9.56527365e-04 2.81466842e-01 2.44482294e-01 -1.42242238e-01 2.49443769e-01 6.80141270e-01 3.85647625e-01 -6.33567154e-01 2.56303638e-01 -6.20990217e-01 3.80126573e-02 5.37986040e-01 -5.29389858e-01 8.28815997e-01 -2.30276093e-01 1.02069807e+00 -7.82536089e-01 -2.49538124e-01 1.00553048e+00 6.79129183e-01 -1.99089080e-01 -7.65545890e-02 -2.68525314e-02 3.39982390e-01 -4.49649781e-01 1.68121174e-01 2.94789970e-01 -1.74342394e-01 -1.32119671e-01 2.13468522e-01 -8.96661222e-01 -1.92542169e-02 -1.06912172e+00 1.49889171e+00 -5.65831125e-01 9.52203214e-01 1.16648458e-01 -7.01160431e-01 1.11745787e+00 3.96109551e-01 9.51079428e-01 -9.99099791e-01 2.86839336e-01 1.89461485e-01 -6.15921915e-01 -1.10573165e-01 1.36999047e+00 7.69234821e-02 -6.73473537e-01 6.98551461e-02 -5.48813462e-01 -2.19732940e-01 3.61078829e-02 -1.36639535e-01 1.16974235e+00 -1.49798214e-01 4.13006246e-01 -1.25068605e-01 3.69578272e-01 7.09071010e-02 7.47276545e-01 8.07500601e-01 -5.49397469e-01 4.98592108e-02 -2.13780716e-01 -5.56886196e-01 -1.10261524e+00 -8.81936789e-01 -5.33246025e-02 8.14930677e-01 5.89733541e-01 -3.45499635e-01 -9.53064784e-02 2.27650762e-01 1.83199719e-01 8.10974956e-01 -6.44129932e-01 -6.39302656e-02 -5.49197912e-01 -2.26923838e-01 9.50218141e-02 6.33626878e-01 4.99604315e-01 -8.17283988e-01 -1.31410074e+00 5.51321745e-01 -3.18398863e-01 -1.06332958e+00 -1.88968286e-01 4.49721575e-01 -2.89329410e-01 -6.00336909e-01 -3.17676932e-01 -5.27094007e-01 -3.90871353e-02 7.54458070e-01 5.82863390e-01 -3.57694626e-01 -1.25023216e-01 4.21517730e-01 -7.48529732e-02 -4.69665349e-01 -1.01112746e-01 -1.13972519e-02 4.00561929e-01 -2.16321230e-01 4.54680830e-01 -4.16119635e-01 -5.69053233e-01 4.35153604e-01 -4.96310622e-01 4.88675594e-01 2.66971350e-01 6.11197412e-01 9.54362035e-01 2.24878415e-01 4.65547651e-01 -4.91548888e-02 5.41385598e-02 -8.46124113e-01 -9.94642913e-01 -1.64971620e-01 -4.44318712e-01 -1.51252553e-01 3.01762342e-01 7.63529986e-02 -9.80012774e-01 3.29562336e-01 -1.03787899e-01 -8.08380842e-01 -2.26079673e-02 4.61099327e-01 4.52667564e-01 3.31222653e-01 2.36969814e-01 3.02328706e-01 -2.13518869e-02 8.49956498e-02 2.12172076e-01 7.94865727e-01 8.69563162e-01 2.49298029e-02 3.00959319e-01 7.72886157e-01 7.90989175e-02 -1.06723285e+00 -3.41986418e-01 -1.04470813e+00 -4.15129602e-01 -7.25751877e-01 1.02706397e+00 -1.27985728e+00 -1.02054095e+00 4.55063194e-01 -1.13603878e+00 -4.59385455e-01 -3.31924975e-01 5.02305627e-01 -1.07091963e+00 -9.43590403e-02 -5.59165299e-01 -1.28000653e+00 -2.92684492e-02 -1.08566880e+00 1.35754812e+00 1.39502212e-01 -2.46864676e-01 -8.79891098e-01 3.25179875e-01 3.57272401e-02 3.07055712e-01 4.20790195e-01 -8.40747431e-02 -1.61561862e-01 -9.02776539e-01 -3.53864461e-01 1.64411753e-01 -2.63871372e-01 3.48382890e-02 -3.60454619e-01 -8.41754794e-01 -3.98337662e-01 1.97999552e-01 6.16244003e-02 8.78969371e-01 7.12831736e-01 8.11392069e-01 -1.50878102e-01 -8.14007699e-01 2.81886667e-01 1.33800864e+00 5.34196854e-01 4.60732430e-01 2.05632478e-01 3.54828984e-01 5.22153974e-01 1.17022932e+00 9.54817057e-01 9.12467778e-01 6.01854026e-01 6.71105146e-01 2.36041844e-01 3.02390009e-01 -1.19243234e-01 4.41705346e-01 9.75386739e-01 2.45974094e-01 -4.30915296e-01 -1.00555861e+00 8.17305923e-01 -2.47037768e+00 -1.14147270e+00 -1.01121336e-01 1.87545383e+00 3.45112383e-03 9.32245031e-02 -2.62683004e-01 6.87365308e-02 5.77037752e-01 4.27842915e-01 -9.10949290e-01 -3.15628529e-01 2.20473353e-02 -4.42923009e-01 1.10415304e+00 7.41999567e-01 -1.16387939e+00 8.70867431e-01 6.38895416e+00 6.75879419e-01 -1.02274799e+00 2.36188427e-01 4.40294653e-01 -2.38679871e-01 -1.40695527e-01 -3.61720204e-01 -1.28587639e+00 2.99939603e-01 1.52684963e+00 -2.17055097e-01 3.08457404e-01 8.12904716e-01 5.53060293e-01 -5.58246791e-01 -6.56447947e-01 1.22400510e+00 7.03435950e-03 -1.60724699e+00 -8.29933584e-01 3.28634679e-01 8.44716609e-01 7.56705999e-01 1.14275753e-01 3.34726751e-01 5.97210169e-01 -8.72083008e-01 8.35910976e-01 1.00804329e+00 3.71864885e-01 -1.07935953e+00 9.09427047e-01 8.00207913e-01 -1.74097168e+00 -3.39670867e-01 -2.63929546e-01 -5.08471429e-01 8.56738985e-01 3.65353316e-01 -9.11138952e-01 5.33471942e-01 7.56756663e-01 1.15189481e+00 -9.00617684e-04 5.35352051e-01 4.91682947e-01 -1.33944973e-01 -5.88594079e-01 -9.68844518e-02 5.76954842e-01 9.76060629e-02 4.15550351e-01 6.57278538e-01 7.11945832e-01 7.73569718e-02 4.94972438e-01 2.48817369e-01 7.43629813e-01 -5.37620366e-01 -1.12722933e+00 2.09290802e-01 6.57382846e-01 9.85467494e-01 -4.20885414e-01 -2.87942380e-01 -3.87410820e-01 6.13315403e-01 8.34368989e-02 2.92997479e-01 -9.84988570e-01 -3.09054535e-02 1.00073755e+00 2.31623482e-02 7.35151410e-01 -6.51282847e-01 -1.42813161e-01 -6.51580095e-01 -3.74090910e-01 -8.96139741e-02 6.84435591e-02 -8.48457873e-01 -5.50124586e-01 4.81353134e-01 1.80993319e-01 -1.20123124e+00 -9.78022695e-01 -2.35947341e-01 -1.24984294e-01 5.96751690e-01 -1.52108979e+00 -9.04763341e-01 -1.21677250e-01 4.48141336e-01 8.48741055e-01 -8.74148245e-05 5.69227159e-01 1.60676595e-02 -4.40672487e-01 -3.76383871e-01 4.63643730e-01 -6.26554251e-01 2.82951325e-01 -7.73705006e-01 7.19019234e-01 6.20201707e-01 -4.23708379e-01 -1.04943879e-01 1.20537663e+00 -9.97198105e-01 -1.76261270e+00 -1.25933754e+00 1.05098140e+00 -4.43369597e-01 6.57593906e-01 -1.24901921e-01 -6.33903742e-01 8.50082159e-01 1.47593915e-01 -1.54542297e-01 2.26748765e-01 -3.53841305e-01 4.09936875e-01 -7.99330994e-02 -7.37093091e-01 4.73665297e-01 7.38890409e-01 -5.53650796e-01 -3.09445620e-01 1.25146434e-01 6.93870127e-01 -6.80033803e-01 -8.90315354e-01 7.03679919e-01 6.41488791e-01 -1.05596256e+00 7.86152720e-01 2.22888529e-01 -2.95976341e-01 -4.59654808e-01 -6.06142700e-01 -1.06820524e+00 -4.11998361e-01 -6.14661038e-01 -4.39412475e-01 4.30887610e-01 9.25278291e-02 -7.23775387e-01 8.86115193e-01 6.01006269e-01 -5.76311767e-01 -6.75842404e-01 -1.51024127e+00 -6.14641488e-01 -5.02600670e-01 -8.94817173e-01 5.71841061e-01 3.36134911e-01 7.97008425e-02 3.83503884e-02 -8.44452322e-01 3.68807793e-01 4.50599521e-01 -1.96066871e-01 8.81074548e-01 -9.08029974e-01 2.02254176e-01 -1.69283934e-02 -6.17684066e-01 -1.26836467e+00 3.92983466e-01 -4.02304858e-01 8.03796887e-01 -1.49494636e+00 -2.44635656e-01 -1.23349242e-01 -3.36105257e-01 2.44567364e-01 4.61776406e-01 3.02449744e-02 -1.44630056e-02 1.54049993e-01 -1.00175953e+00 9.52181876e-01 8.77243459e-01 -7.75989443e-02 -1.74745575e-01 -1.12119287e-01 2.26442978e-01 4.71800953e-01 9.54890490e-01 -6.05760962e-02 -6.78956628e-01 4.22754474e-02 3.43393415e-01 8.21605682e-01 2.33866960e-01 -1.42156160e+00 7.73684978e-01 -3.46998960e-01 2.09385693e-01 -1.74759114e+00 1.19720089e+00 -9.54394996e-01 7.69117951e-01 6.63797796e-01 1.84576511e-01 6.06801569e-01 4.39643651e-01 9.90302801e-01 -1.47155508e-01 3.22507352e-01 3.51910442e-01 1.26791736e-02 -1.42334318e+00 4.91972864e-01 -1.18330312e+00 -6.68503225e-01 1.45199955e+00 -4.75807637e-01 -2.39911117e-02 -5.61488450e-01 -5.78425944e-01 9.59432304e-01 4.15555775e-01 6.22138739e-01 8.21423888e-01 -1.39271200e+00 -3.83781940e-01 4.10354882e-01 -4.50624898e-02 1.11376502e-01 8.07144523e-01 9.70369458e-01 -1.64150611e-01 9.30551887e-01 -1.10996731e-01 -9.42237675e-01 -1.06546819e+00 5.52447855e-01 2.84681767e-01 -1.84284359e-01 -6.52303517e-01 3.54171693e-01 2.16053307e-01 -2.53948420e-01 6.05686493e-02 -2.69298464e-01 -2.26965681e-01 9.70836431e-02 8.62818241e-01 7.50415981e-01 -1.39494896e-01 -1.16978478e+00 -2.45455980e-01 1.10671580e-01 8.67573246e-02 -4.02557105e-01 1.19429767e+00 -1.02109146e+00 2.60084152e-01 1.16691458e+00 1.00605309e+00 -8.04016232e-01 -1.78715634e+00 -1.67876586e-01 -1.25465155e-01 1.93543658e-02 3.19404542e-01 -6.58332836e-03 -6.35432363e-01 5.19737184e-01 6.33298755e-01 1.68517634e-01 6.79284990e-01 -7.77563155e-02 1.03085756e+00 6.09144151e-01 9.13698733e-01 -1.33114636e+00 -2.34069780e-01 1.02816355e+00 7.67766356e-01 -1.14704156e+00 -2.61340916e-01 -3.40716317e-02 -6.08332932e-01 7.84664452e-01 5.41456938e-01 -6.04343368e-03 9.73097026e-01 4.66882676e-01 -1.58170164e-01 -1.70069516e-01 -1.42848289e+00 -2.56913096e-01 -5.92570454e-02 3.70825917e-01 -1.79969430e-01 3.38064909e-01 3.38123858e-01 4.20846269e-02 -3.65087599e-01 -1.01201683e-01 3.86380106e-01 1.22358620e+00 -8.63094807e-01 -3.71118188e-01 -5.11728108e-01 5.70945442e-01 1.84889153e-01 4.21595424e-01 4.42766905e-01 8.11718285e-01 1.07245885e-01 1.12621903e+00 6.11238778e-01 -5.90311706e-01 -1.09853493e-02 -1.07870139e-01 2.61407928e-03 -2.68224806e-01 -8.36655051e-02 2.45194033e-01 9.79127139e-02 -9.20572460e-01 -3.25561076e-01 -1.09349823e+00 -1.53342760e+00 -7.04624057e-01 -1.78208172e-01 1.15675069e-01 8.59796584e-01 1.29071522e+00 3.80989015e-01 7.43586600e-01 6.17036521e-01 -1.41625333e+00 -2.01962233e-01 -6.65572047e-01 -7.08398521e-01 -3.03492576e-01 9.04354513e-01 -8.94465744e-01 -1.28776774e-01 -3.51412505e-01]

[5.900545597076416, 0.788245677947998]

b05466f3-17bd-40d5-8e11-15bf5f62fd7a

efficient-automatic-punctuation-restoration

https://aclanthology.org/2020.iwslt-1.33

https://aclanthology.org/2020.iwslt-1.33.pdf

Efficient Automatic Punctuation Restoration Using Bidirectional Transformers with Robust Inference

Though people rarely speak in complete sentences, punctuation confers many benefits to the readers of transcribed speech. Unfortunately, most ASR systems do not produce punctuated output. To address this, we propose a solution for automatic punctuation that is both cost efficient and easy to train. Our solution benefits from the recent trend in fine-tuning transformer-based language models. We also modify the typical framing of this task by predicting punctuation for sequences rather than individual tokens, which makes for more efficient training and inference. Finally, we find that aggregating predictions across multiple context windows improves accuracy even further. Our best model achieves a new state of the art on benchmark data (TED Talks) with a combined F1 of 83.9, representing a 48.7{\%} relative improvement (15.3 absolute) over the previous state of the art.

[]

2020-07-01

acl-2020-7

['punctuation-restoration']

['natural-language-processing']

[ 2.45574549e-01 2.83170938e-01 -2.10358366e-01 -4.48571175e-01 -1.34627891e+00 -8.25774431e-01 5.86050153e-01 -1.36424094e-01 -3.02720010e-01 9.99346852e-01 5.85023999e-01 -5.98924160e-01 4.78026122e-01 -3.57636899e-01 -7.11132109e-01 -3.13380331e-01 1.75457150e-01 1.97231740e-01 1.99914619e-01 -3.13401401e-01 -4.87782955e-02 1.04687244e-01 -1.07376027e+00 6.12536728e-01 9.26190436e-01 6.55085206e-01 4.04168367e-01 6.67729795e-01 -3.75074923e-01 1.03889000e+00 -1.10908771e+00 -5.43252289e-01 -1.63812622e-01 -4.87066239e-01 -1.09139681e+00 9.50304121e-02 4.27010775e-01 1.98914520e-02 -2.03701973e-01 6.51498854e-01 4.27657813e-01 6.62573576e-02 2.45598152e-01 -6.24649227e-01 -6.26373887e-01 1.21174502e+00 -1.47032723e-01 4.45007205e-01 4.00428891e-01 -7.84665197e-02 1.31195760e+00 -8.08926463e-01 4.14143741e-01 1.15286541e+00 6.21961057e-01 6.89167023e-01 -1.19344091e+00 -4.73410428e-01 4.03182238e-01 -8.11308399e-02 -1.38878214e+00 -8.04561913e-01 4.48430836e-01 6.28352985e-02 1.58918190e+00 5.09526312e-01 2.34846309e-01 1.27169776e+00 -2.90060975e-02 1.07859886e+00 1.24423254e+00 -7.11691797e-01 -5.48948050e-02 2.75380820e-01 5.86978532e-03 4.50673759e-01 -1.30497530e-01 -3.00511628e-01 -6.94195986e-01 -2.18622983e-02 2.61839777e-01 -2.33416215e-01 -1.64015174e-01 6.35362089e-01 -1.15866220e+00 5.15794098e-01 -5.77734560e-02 5.57844996e-01 -1.56992659e-01 1.10512629e-01 2.76414871e-01 3.94193709e-01 6.13537192e-01 5.81645668e-01 -8.18488300e-01 -7.98268735e-01 -1.28716922e+00 1.82895809e-01 9.44323242e-01 1.06661952e+00 5.23858488e-01 3.29533011e-01 -3.70093644e-01 1.15955591e+00 7.67390132e-02 5.93622983e-01 5.92202604e-01 -9.59976435e-01 7.83834934e-01 1.41377971e-01 1.34030744e-01 -2.56979346e-01 -1.07769087e-01 -7.49044597e-01 -3.54231685e-01 -5.82987130e-01 4.67627108e-01 -4.70495313e-01 -8.64215374e-01 1.78508353e+00 -2.20484823e-01 4.89373729e-02 1.18855871e-01 4.22358483e-01 4.39332873e-01 1.02789176e+00 2.22602561e-01 -3.04978698e-01 1.39039803e+00 -1.07728243e+00 -1.14143121e+00 -5.39027512e-01 7.98858643e-01 -1.10459733e+00 1.41211808e+00 5.07471502e-01 -1.25289488e+00 -4.13393766e-01 -9.49585378e-01 -8.30525085e-02 -2.76458651e-01 2.28145182e-01 4.66667533e-01 9.01868105e-01 -1.20709920e+00 6.21709049e-01 -7.20822394e-01 -1.70862943e-01 9.87271499e-03 2.27282912e-01 2.11951360e-02 3.28643978e-01 -1.41535926e+00 9.50363696e-01 8.11133161e-02 -2.41449758e-01 -4.49175626e-01 -8.45103025e-01 -6.96007073e-01 1.58594057e-01 4.09068346e-01 -2.30956718e-01 1.96434593e+00 -9.74831402e-01 -1.79412854e+00 7.81154454e-01 -8.93655598e-01 -7.37229705e-01 4.48925883e-01 -6.70798898e-01 -6.88141048e-01 -1.39439747e-01 -2.52389498e-02 4.74295735e-01 5.31653941e-01 -8.12275767e-01 -7.79726386e-01 2.02731863e-01 -2.12553039e-01 2.15311155e-01 -4.54721212e-01 4.66320366e-01 -3.24681640e-01 -7.62596726e-01 -4.08335686e-01 -7.93514371e-01 -9.50942039e-02 -7.26507664e-01 -4.50802058e-01 -5.63958466e-01 7.20948517e-01 -8.55309963e-01 1.72513247e+00 -2.00637269e+00 -3.13565016e-01 -1.32430673e-01 -1.40872121e-01 5.72537720e-01 1.07636161e-01 6.10481918e-01 -3.10375188e-02 4.66487288e-01 -1.48532137e-01 -8.94060612e-01 6.85919374e-02 2.73462147e-01 -8.03848326e-01 -1.12459682e-01 4.87500846e-01 1.02016044e+00 -9.29585516e-01 -3.02563101e-01 1.29279941e-01 3.66327107e-01 -3.59979212e-01 2.27473438e-01 -2.24279061e-01 2.71819755e-02 -1.67395249e-01 4.43085641e-01 1.83779761e-01 -2.09455758e-01 3.39554846e-01 5.73161960e-01 -4.08460945e-01 1.40418279e+00 -9.41408217e-01 1.63160062e+00 -7.43007660e-01 7.75832355e-01 -2.04095051e-01 -7.83596218e-01 9.54160154e-01 8.03510129e-01 1.16024271e-01 -5.55347681e-01 -1.41622782e-01 3.48268211e-01 -1.15897335e-01 -2.02619433e-01 8.08396637e-01 -1.69399455e-01 -1.37852043e-01 3.27991486e-01 1.22765712e-01 -3.30709189e-01 1.70713365e-01 1.77231610e-01 1.13217688e+00 -3.11831236e-02 2.56814301e-01 -2.86916256e-01 2.64059216e-01 -1.92421749e-01 4.56492603e-01 9.11402345e-01 -9.78479981e-02 7.33841896e-01 3.72335255e-01 -1.06989846e-01 -1.13097215e+00 -1.13285708e+00 1.84849873e-02 1.06419158e+00 -5.12121558e-01 -9.17960644e-01 -1.02987170e+00 -7.01896131e-01 -3.81120533e-01 1.11644483e+00 -1.41477317e-01 9.88990664e-02 -9.49178815e-01 -5.27018666e-01 8.20321500e-01 6.88795507e-01 1.92958936e-01 -1.18493497e+00 4.75096777e-02 5.84227085e-01 -7.66238272e-01 -1.53094959e+00 -7.77627110e-01 3.28894079e-01 -8.84301305e-01 -3.19175512e-01 -5.81238389e-01 -8.86103034e-01 7.87338540e-02 2.64938802e-01 1.40027785e+00 1.00822695e-01 8.73087347e-02 -2.13015899e-02 -4.52442288e-01 -5.52466333e-01 -7.38519311e-01 5.11863232e-01 1.10157385e-01 -1.98841378e-01 5.27362108e-01 -5.35826206e-01 -1.54850602e-01 1.09609380e-01 -5.65349221e-01 -1.34875983e-01 6.09163940e-01 7.36588061e-01 3.55549335e-01 -2.53661990e-01 7.39369273e-01 -8.90577912e-01 6.26263559e-01 -2.61592329e-01 -2.33828902e-01 2.00640678e-01 -6.27391219e-01 1.71903998e-01 9.90513265e-01 -3.58883679e-01 -1.18934202e+00 -3.71935852e-02 -4.30331379e-01 -1.83859076e-02 -2.67292768e-01 3.56160194e-01 6.46778643e-02 5.30580759e-01 5.81374526e-01 5.09286463e-01 -2.42414266e-01 -5.81389368e-01 3.36246789e-01 9.89728987e-01 4.69821215e-01 -5.40895224e-01 6.50916457e-01 3.71257924e-02 -6.58058047e-01 -1.22492194e+00 -1.16196573e+00 -5.20866990e-01 -3.76991808e-01 9.00498256e-02 5.27531862e-01 -9.66755331e-01 -5.34981310e-01 1.84790924e-01 -1.30018830e+00 -5.23687124e-01 -2.47912794e-01 2.10241348e-01 -4.96314704e-01 6.20181203e-01 -8.22750449e-01 -1.03237593e+00 -5.79828560e-01 -8.46016228e-01 1.02919197e+00 1.55176833e-01 -7.50439763e-01 -9.12843227e-01 -3.57781686e-02 4.04652148e-01 4.51333076e-01 -4.08174187e-01 3.94795150e-01 -6.75148785e-01 -4.22953755e-01 1.42241776e-01 8.51613209e-02 6.91390753e-01 4.45393026e-01 2.56048441e-02 -1.16691506e+00 -4.29941006e-02 -1.41872451e-01 -2.96701461e-01 7.14461863e-01 1.54804915e-01 1.00335121e+00 -6.07473612e-01 -2.66224116e-01 1.25878245e-01 9.33593690e-01 2.02078119e-01 7.75615096e-01 2.11405247e-01 3.34127963e-01 3.95648003e-01 5.17563641e-01 3.27698052e-01 3.59797210e-01 7.13463068e-01 -2.46056378e-01 4.15304117e-02 -3.49981934e-01 -5.63732803e-01 8.66158009e-01 1.16859710e+00 1.84282467e-01 -5.31734586e-01 -8.31789732e-01 9.60107446e-01 -1.52113259e+00 -1.12150252e+00 -4.03969511e-02 1.99559140e+00 1.47653270e+00 4.73949224e-01 1.29714146e-01 1.95021838e-01 6.74696803e-01 2.44460270e-01 1.79641262e-01 -8.12149823e-01 -2.60916591e-01 7.78650522e-01 4.12816823e-01 7.91347444e-01 -1.10639668e+00 1.47435534e+00 7.54829693e+00 9.80388999e-01 -1.12528062e+00 9.12786350e-02 5.71650445e-01 -5.88500500e-02 -4.06874120e-01 4.09278423e-02 -1.29345667e+00 6.67681992e-01 1.63689661e+00 -2.79676497e-01 5.15405357e-01 8.26189339e-01 5.55844545e-01 1.04977936e-01 -9.38336313e-01 7.51862466e-01 2.72800419e-02 -1.41138017e+00 1.03089977e-02 -1.81016847e-01 7.68860400e-01 1.72019020e-01 2.30717734e-02 4.49668229e-01 5.33845842e-01 -1.11190569e+00 9.72240031e-01 9.61705670e-02 6.27302587e-01 -7.89772153e-01 4.24781322e-01 4.41340864e-01 -1.06143963e+00 2.09121391e-01 -1.04001753e-01 -3.07350814e-01 4.17490304e-01 5.27237296e-01 -1.32098174e+00 4.99366552e-01 5.61951756e-01 4.76337492e-01 -3.17096919e-01 7.66209602e-01 -5.92130780e-01 1.47322226e+00 -5.03558695e-01 -3.94806087e-01 5.08550942e-01 1.38130844e-01 6.15891457e-01 1.95434678e+00 2.61805952e-01 -9.45680495e-03 2.03612506e-01 7.22228944e-01 -1.59554541e-01 5.38800396e-02 -4.76703882e-01 -1.38369918e-01 8.64107013e-01 1.05287337e+00 -4.39967602e-01 -5.24394155e-01 -4.89592373e-01 1.01943350e+00 5.09750903e-01 1.87536806e-01 -7.76343882e-01 -5.89327514e-01 8.52782130e-01 3.77006717e-02 5.56541741e-01 -5.68329036e-01 -3.71894270e-01 -1.09846509e+00 4.02443230e-01 -1.00917470e+00 1.90146431e-01 -5.46881080e-01 -1.04791260e+00 7.58831382e-01 -2.31174022e-01 -1.01248717e+00 -6.10047877e-01 -4.95055437e-01 -6.41634881e-01 8.92033219e-01 -1.77217376e+00 -9.52613950e-01 2.77124882e-01 1.73809692e-01 9.82099950e-01 7.98671171e-02 9.99780357e-01 3.38656425e-01 -4.05710995e-01 1.01276040e+00 1.18899889e-01 3.92455786e-01 8.21957290e-01 -1.59532988e+00 9.92263377e-01 1.17421269e+00 5.80552697e-01 7.81336427e-01 8.39819491e-01 -4.23091382e-01 -1.06673753e+00 -1.06782818e+00 1.95256579e+00 -8.42756152e-01 9.27432716e-01 -5.87598562e-01 -7.78018415e-01 9.11524892e-01 4.87528056e-01 -4.16159213e-01 6.18823826e-01 4.22901005e-01 -3.71100664e-01 -7.34038949e-02 -5.93468428e-01 8.70676041e-01 1.00050926e+00 -7.61877596e-01 -9.66755688e-01 2.34233662e-01 1.06481647e+00 -5.03689468e-01 -6.62159324e-01 6.76591471e-02 2.71295190e-01 -6.48696482e-01 7.03437686e-01 -5.91512144e-01 2.66008735e-01 -1.82810396e-01 -2.30968714e-01 -1.37342954e+00 -1.94334924e-01 -1.41409218e+00 -2.44030863e-01 1.51604962e+00 9.09413874e-01 -5.02225995e-01 6.33180678e-01 5.21621943e-01 -6.42605305e-01 -6.70789897e-01 -9.19308543e-01 -1.11078858e+00 1.72690034e-01 -6.89310610e-01 4.56227034e-01 7.83463776e-01 2.96456963e-01 5.04786909e-01 -4.79923189e-01 -6.14376180e-02 1.64278090e-01 -1.88447744e-01 3.58605713e-01 -7.32636929e-01 -3.40309173e-01 -3.21961582e-01 6.69532865e-02 -1.50398386e+00 2.35562757e-01 -8.06267977e-01 1.57578602e-01 -1.29770780e+00 -1.47288628e-02 -4.33423430e-01 -2.83367395e-01 7.15707362e-01 -4.12764132e-01 2.11885750e-01 2.23678231e-01 -2.13232577e-01 -6.67471468e-01 4.17500257e-01 8.95061672e-01 2.53078751e-02 -2.93463618e-01 1.19742401e-01 -8.70119154e-01 3.92734408e-01 1.06221437e+00 -4.47469711e-01 -1.17129683e-01 -5.82576871e-01 -1.02433950e-01 -1.71573386e-01 -3.24288197e-02 -9.36255634e-01 -1.32920574e-02 -7.76483268e-02 1.40505210e-01 -6.31266356e-01 4.80995744e-01 -2.79798597e-01 -2.66212344e-01 1.23056762e-01 -4.60967749e-01 2.48692900e-01 2.38936216e-01 1.87739506e-01 -3.01623881e-01 -9.91432443e-02 6.20300889e-01 -1.64495677e-01 -4.57978010e-01 -1.21493347e-01 -9.12781715e-01 4.16145295e-01 3.92870873e-01 -1.53386444e-02 -3.03546607e-01 -4.84779835e-01 -5.19272566e-01 -2.26650992e-03 -1.41041307e-02 7.60250270e-01 2.25621656e-01 -1.03494155e+00 -7.26717472e-01 2.63165068e-02 -1.86656892e-01 -2.41202161e-01 -1.08903520e-01 5.38406193e-01 -7.75212720e-02 8.51956427e-01 5.82742155e-01 -3.92286837e-01 -1.36516356e+00 2.22487122e-01 2.69990325e-01 -4.32594270e-01 -5.68691075e-01 8.57972622e-01 -2.74693400e-01 -2.59115100e-01 3.33688319e-01 -6.06374860e-01 1.14020541e-01 -1.83345094e-01 7.26146638e-01 1.74293131e-01 4.55618411e-01 -4.44279790e-01 -3.45432758e-01 8.44307020e-02 -3.70014429e-01 -4.02418971e-01 1.10252869e+00 -1.95116520e-01 7.72550851e-02 7.27987647e-01 9.76304352e-01 4.76794332e-01 -1.09509170e+00 -4.43700105e-01 3.82510155e-01 -7.61665925e-02 -1.85129330e-01 -1.08205140e+00 -4.37539697e-01 7.86934376e-01 -1.17256209e-01 4.39141512e-01 7.30439901e-01 1.93844199e-01 1.12511432e+00 7.28408337e-01 1.43314019e-01 -1.40618169e+00 5.87258600e-02 1.10195518e+00 6.25331640e-01 -9.59345222e-01 -3.43656778e-01 -5.07544279e-01 -8.29875410e-01 9.06344652e-01 3.16597044e-01 -9.05276462e-03 3.84442896e-01 5.68662524e-01 3.70095968e-01 3.70914757e-01 -1.12972486e+00 -6.04779199e-02 1.66205809e-01 3.20269227e-01 9.89053607e-01 3.22153002e-01 -2.95146972e-01 8.36772382e-01 -7.66666472e-01 -2.96808332e-01 4.03549194e-01 8.78612518e-01 -6.85407281e-01 -1.51498008e+00 -1.80599079e-01 1.83407530e-01 -1.01690590e+00 -5.75625837e-01 -4.91322070e-01 5.04383445e-01 -3.71359497e-01 1.41997921e+00 2.78141312e-02 -2.35327467e-01 2.99579948e-01 5.52968442e-01 2.61493891e-01 -7.76807845e-01 -8.54348600e-01 7.15782195e-02 5.93664110e-01 -3.94265950e-01 -1.38008863e-01 -7.97368050e-01 -1.24975586e+00 -2.47134700e-01 -2.36525029e-01 5.42064130e-01 5.22187948e-01 1.11493087e+00 4.73688394e-01 4.77874219e-01 8.52647722e-01 -3.50490779e-01 -7.24320054e-01 -1.13244033e+00 -2.63862073e-01 1.55463755e-01 3.74276489e-01 -1.49289936e-01 -3.32527190e-01 2.51487762e-01]

[14.235575675964355, 7.060207366943359]

a1b515cc-f502-4a78-9517-fbd8c14e844a

all-in-one-multi-task-learning-for-rumour

1806.03713

http://arxiv.org/abs/1806.03713v1

http://arxiv.org/pdf/1806.03713v1.pdf

All-in-one: Multi-task Learning for Rumour Verification

Automatic resolution of rumours is a challenging task that can be broken down into smaller components that make up a pipeline, including rumour detection, rumour tracking and stance classification, leading to the final outcome of determining the veracity of a rumour. In previous work, these steps in the process of rumour verification have been developed as separate components where the output of one feeds into the next. We propose a multi-task learning approach that allows joint training of the main and auxiliary tasks, improving the performance of rumour verification. We examine the connection between the dataset properties and the outcomes of the multi-task learning models used.

['Arkaitz Zubiaga', 'Maria Liakata', 'Elena Kochkina']

2018-06-10

all-in-one-multi-task-learning-for-rumour-1

https://aclanthology.org/C18-1288

https://aclanthology.org/C18-1288.pdf

coling-2018-8

['rumour-detection']

['natural-language-processing']

[-1.23948045e-03 2.62443960e-01 -1.12774521e-01 -2.79689938e-01 -5.86711109e-01 -3.81090045e-01 1.10663402e+00 5.29692292e-01 -1.22447178e-01 5.94911933e-01 4.67270255e-01 -3.88316691e-01 4.75012958e-02 -4.79701459e-01 -5.66955984e-01 -3.43371391e-01 -1.56736970e-01 6.44418955e-01 5.36813736e-01 -2.81377614e-01 7.80739665e-01 8.30087215e-02 -1.38761032e+00 7.20035553e-01 3.88105750e-01 9.99652624e-01 -5.26421130e-01 7.23916054e-01 -5.14634605e-03 1.71995592e+00 -7.61464357e-01 -2.35951141e-01 -3.38724256e-01 -3.69267344e-01 -1.22003329e+00 2.15351969e-01 1.02915466e-01 -3.28890800e-01 2.38793999e-01 6.51193619e-01 1.22826278e-01 -7.02663884e-02 5.58630943e-01 -1.28077137e+00 -3.77260029e-01 7.77023137e-01 -3.57696027e-01 2.78034508e-01 3.85816097e-01 -2.68357217e-01 9.24812675e-01 -9.69150245e-01 3.77922744e-01 1.14016497e+00 9.06310558e-01 6.42551295e-03 -1.01734626e+00 -6.46452904e-01 -2.61959910e-01 4.17908311e-01 -7.95340300e-01 -6.74662590e-01 5.89752674e-01 -9.28951859e-01 6.87176108e-01 1.42734982e-02 3.65970105e-01 1.20801473e+00 2.83115536e-01 7.13255405e-01 1.34119594e+00 -2.23497495e-01 1.91052079e-01 5.47850013e-01 6.40614927e-01 5.79372823e-01 6.18374161e-02 2.42502578e-02 -7.92255759e-01 -4.63465005e-01 3.12510073e-01 -1.65893778e-01 -6.61272556e-02 -2.05556393e-01 -8.46240044e-01 1.22005486e+00 3.31514597e-01 2.14683503e-01 -5.85471988e-01 -3.36140990e-01 8.71063948e-01 7.68056810e-01 1.04041147e+00 4.19580817e-01 -2.00232193e-01 -4.14337404e-02 -1.56642818e+00 5.94838440e-01 1.31022310e+00 3.96557778e-01 5.15132189e-01 -1.74546316e-01 -2.28304088e-01 8.64940822e-01 2.36848563e-01 -2.44819596e-01 4.05907601e-01 -5.40799499e-01 3.63742933e-02 4.91691679e-01 4.64015901e-01 -8.02618146e-01 -8.57227743e-01 -5.42136073e-01 -1.52222872e-01 1.70057431e-01 3.37022394e-01 -2.33570769e-01 -6.64498210e-01 1.22140479e+00 5.73956370e-01 9.11693871e-02 -2.26984546e-01 9.39539135e-01 8.35835576e-01 2.21908525e-01 -2.05346681e-02 -8.73358399e-02 1.54970038e+00 -1.24830806e+00 -7.34398603e-01 -1.80365950e-01 4.76004630e-01 -1.07111514e+00 5.95968843e-01 4.78834927e-01 -1.07197154e+00 -2.07713813e-01 -1.24882841e+00 -1.69283003e-01 -3.10427129e-01 -1.89019382e-01 4.10507083e-01 4.41156209e-01 -6.54666662e-01 6.06100619e-01 -5.02166390e-01 -6.19860515e-02 3.20121408e-01 -2.12843597e-01 1.01332881e-01 1.06381297e-01 -1.49531174e+00 1.43955433e+00 1.87185436e-01 -1.06126286e-01 -1.23533130e+00 -5.52722156e-01 -5.84007740e-01 -1.43638179e-01 4.30423915e-01 -5.49334347e-01 1.50328684e+00 -7.80437052e-01 -1.39147806e+00 1.08576965e+00 4.70712446e-02 -5.51044941e-01 1.00312555e+00 -4.32765394e-01 -2.10840940e-01 -2.94917494e-01 1.27876073e-01 -2.26717234e-01 1.42974770e+00 -1.05787373e+00 -7.28751838e-01 -4.07930076e-01 -1.56714976e-01 -1.89334974e-01 2.10754931e-01 6.62045121e-01 1.21181861e-01 -3.85371357e-01 -2.04643220e-01 -8.64040852e-01 2.74685085e-01 -5.73539615e-01 -2.59570122e-01 -2.77279079e-01 7.37005532e-01 -9.39106464e-01 1.07220876e+00 -1.88126993e+00 6.52052313e-02 -2.33037695e-01 2.91890383e-01 7.80771151e-02 3.33869517e-01 4.67128277e-01 -1.15079150e-01 -4.86765057e-02 5.59323430e-02 -7.12195277e-01 -1.42901903e-04 -2.70269543e-01 -5.86424530e-01 7.65991449e-01 2.97470391e-01 6.90053225e-01 -7.35200226e-01 -2.10108250e-01 -2.05344558e-01 1.87591583e-01 -1.92666262e-01 3.50525886e-01 -2.25743398e-01 3.23070228e-01 -1.54183328e-01 4.07703429e-01 5.00154555e-01 -3.08310032e-01 1.06780179e-01 1.33154675e-01 -2.84866780e-01 1.20567203e+00 -8.13363850e-01 1.01087809e+00 -4.56956625e-01 6.02855861e-01 4.01682734e-01 -8.56907964e-01 1.00908172e+00 7.29727626e-01 3.62074107e-01 -3.60729992e-01 3.13638419e-01 3.30193877e-01 -2.26411611e-01 -6.74700379e-01 9.45272207e-01 -6.13661051e-01 -2.50586212e-01 1.20454121e+00 7.01782405e-02 3.95747982e-02 -1.53200224e-01 2.59889901e-01 9.38547730e-01 1.96040034e-01 6.82547569e-01 -1.12600267e-01 4.59235519e-01 3.54135305e-01 2.62492150e-01 6.40974402e-01 -2.69851029e-01 4.39636111e-01 9.48184967e-01 -4.32129681e-01 -1.23109460e+00 -5.64467788e-01 -2.71155328e-01 1.59126532e+00 -3.09212446e-01 -1.49670467e-01 -5.39621174e-01 -8.15330744e-01 2.58408308e-01 9.32590604e-01 -1.11120558e+00 -8.39324072e-02 -5.06198227e-01 -1.01179028e+00 7.09041595e-01 1.10443816e-01 -3.46147940e-02 -1.20601273e+00 -7.53524125e-01 2.33574882e-01 -2.97615051e-01 -1.17021346e+00 5.27203679e-02 3.54911774e-01 -5.20911336e-01 -1.15147209e+00 -2.77887523e-01 -2.55319774e-01 -7.12718070e-02 1.48387477e-01 1.20313740e+00 4.10655558e-01 1.58247948e-01 -2.41807938e-01 -5.33944666e-01 -5.93744755e-01 -9.93386149e-01 9.07946751e-02 -1.10677317e-01 7.17382282e-02 1.44399822e-01 -6.29082382e-01 -1.19315550e-01 2.36608684e-01 -7.33172178e-01 7.21397698e-02 3.72666478e-01 7.29465842e-01 -1.01717561e-01 -3.98589104e-01 9.33263958e-01 -1.21265292e+00 9.17366326e-01 -1.18052423e+00 -1.92707688e-01 2.51429658e-02 -6.03913844e-01 -9.88765135e-02 4.45377469e-01 -2.50098228e-01 -7.07109213e-01 -4.63071167e-01 -1.26270801e-01 -2.29531378e-01 -4.89236936e-02 7.05444694e-01 4.80819434e-01 2.06531510e-01 7.30926037e-01 1.44067705e-01 2.21271276e-01 -7.62520671e-01 3.44026387e-01 1.11817729e+00 3.65376770e-01 -9.11864340e-02 4.96140391e-01 3.95248204e-01 -3.11488330e-01 -6.10964775e-01 -1.47967029e+00 -7.43525267e-01 -7.89428115e-01 -3.87869328e-01 4.07461226e-01 -8.75816107e-01 -5.33004165e-01 5.50594330e-01 -1.29869246e+00 -2.31914178e-01 2.55488098e-01 1.20103419e-01 -4.88728762e-01 2.65900135e-01 -8.44006538e-01 -8.02036941e-01 -5.98536253e-01 -8.23033452e-01 6.97947681e-01 -3.73231590e-01 -6.53932273e-01 -9.52126861e-01 4.07783210e-01 8.99785578e-01 6.05454862e-01 1.11499548e-01 8.81748199e-01 -1.36343026e+00 -1.10197112e-01 -1.99187830e-01 -3.77699971e-01 1.75963759e-01 -2.02971280e-01 -1.76695019e-01 -1.17904067e+00 -2.81764269e-01 6.65651441e-01 -7.91671097e-01 1.17067134e+00 3.81831378e-02 2.60624766e-01 -4.09145743e-01 1.06058657e-01 2.71193147e-01 9.60763931e-01 -5.53269744e-01 3.10596228e-01 8.63373399e-01 4.13937747e-01 7.93736935e-01 7.25839674e-01 5.72596610e-01 4.86603856e-01 6.59517944e-01 5.43791175e-01 8.38188305e-02 -8.93290713e-02 1.04453072e-01 8.55477870e-01 5.82604170e-01 2.55438894e-01 1.35641426e-01 -8.82879496e-01 7.08596349e-01 -1.91024303e+00 -9.39551473e-01 -4.89136010e-01 1.90756190e+00 8.57981443e-01 2.79466689e-01 5.02010763e-01 2.98696220e-01 4.53130424e-01 5.04421353e-01 -1.80744827e-01 -1.00177574e+00 4.15706486e-01 -2.71167368e-01 1.90604627e-01 7.40219474e-01 -1.09493041e+00 7.83689439e-01 7.32900620e+00 9.63704586e-02 -1.41881180e+00 5.79948008e-01 -7.77050555e-02 -3.46475065e-01 7.39097502e-03 3.15328613e-02 -6.52011395e-01 3.72833818e-01 1.16391420e+00 1.73838362e-02 3.61906379e-01 7.33182788e-01 4.52697456e-01 -2.75979340e-01 -1.08158791e+00 -4.88889553e-02 4.01995301e-01 -1.15912068e+00 -9.07352120e-02 -2.44304389e-01 4.21964228e-01 5.94809532e-01 -2.91254461e-01 4.57057953e-01 4.85567272e-01 -8.56577158e-01 1.23268676e+00 1.21935099e-01 2.46852726e-01 -5.66837311e-01 8.30492496e-01 7.12268233e-01 -2.59149075e-01 -1.83220714e-01 -6.10384531e-02 -3.13107193e-01 4.22961414e-01 9.02763724e-01 -1.64888585e+00 6.39849305e-01 3.55502367e-01 5.73021173e-01 -3.04345280e-01 8.68770301e-01 -4.37642217e-01 7.46001720e-01 1.61892906e-01 3.49937558e-01 2.06841826e-01 3.29978317e-01 7.13394821e-01 1.44166923e+00 -2.03290045e-01 -4.38145757e-01 4.38127041e-01 1.03585756e+00 -1.09133199e-01 -3.84772941e-02 -2.63199992e-02 2.02051461e-01 2.78236717e-01 1.24647427e+00 -2.62072027e-01 -3.57056022e-01 -3.25813949e-01 7.06109285e-01 6.55543029e-01 -3.07822764e-01 -1.03032708e+00 2.15606153e-01 4.57871616e-01 3.47497284e-01 4.45232242e-01 3.85829136e-02 -5.16703546e-01 -1.11110377e+00 -1.58170164e-01 -1.16887701e+00 5.34284830e-01 -4.34600800e-01 -1.13530517e+00 3.80460441e-01 -4.23088223e-01 -7.52132773e-01 -3.14181089e-01 -1.32586122e-01 -8.29224586e-01 9.89926815e-01 -1.95821381e+00 -1.42154455e+00 -4.24121261e-01 3.10098141e-01 4.86624092e-01 2.57998675e-01 8.48392606e-01 1.76247388e-01 -8.10045123e-01 8.33880007e-02 -2.33525366e-01 3.10421530e-02 1.04830933e+00 -1.06770563e+00 4.99414861e-01 6.84880137e-01 -2.72712409e-01 3.69367063e-01 9.49601293e-01 -7.70408273e-01 -7.59332418e-01 -1.04172397e+00 1.18982100e+00 -6.37076259e-01 1.29578376e+00 -2.91907609e-01 -1.35060596e+00 8.52927744e-01 8.27700738e-03 -3.03105175e-01 4.36394453e-01 6.30605340e-01 -7.35572398e-01 5.79898953e-01 -8.57512295e-01 -3.09469461e-01 1.40716195e-01 -8.22805643e-01 -1.13851678e+00 3.71105701e-01 2.69547313e-01 -3.71342421e-01 -9.28055346e-01 3.60025987e-02 5.81911206e-01 -1.10358572e+00 6.05474532e-01 -9.14220929e-01 1.01182771e+00 1.67401060e-02 3.15641582e-01 -1.25058377e+00 -4.05505121e-01 -4.73430395e-01 -3.86836559e-01 1.10445237e+00 5.00100315e-01 -2.88244307e-01 2.50916421e-01 -1.49574608e-01 8.01879317e-02 -5.46099663e-01 -5.36532521e-01 -3.44889343e-01 1.25416666e-01 -4.00470376e-01 2.64757335e-01 1.27146673e+00 4.66502041e-01 9.32835817e-01 -9.10405517e-01 2.66536623e-02 6.04224145e-01 5.43371677e-01 7.36697853e-01 -1.39859748e+00 -1.70408055e-01 -4.12716568e-01 1.88971177e-01 -7.59538889e-01 -1.77374855e-02 -1.03152001e+00 2.47519851e-01 -1.43898153e+00 5.61321974e-01 -3.62245858e-01 -2.71906197e-01 4.64390963e-01 -4.59114403e-01 1.07490316e-01 4.67953645e-02 8.79083753e-01 -7.24759817e-01 3.48479807e-01 7.71089911e-01 2.73237646e-01 -2.09317029e-01 2.59555846e-01 -6.52860880e-01 7.60078847e-01 6.52331531e-01 -8.78023028e-01 1.90912113e-01 -2.61286139e-01 4.11965489e-01 3.08458626e-01 6.58954501e-01 -6.51767373e-01 1.75827146e-01 7.89367184e-02 -8.39413553e-02 -6.18700385e-01 1.96304232e-01 -2.37774134e-01 -1.84782237e-01 3.11429650e-01 -5.07684469e-01 -1.15033098e-01 -6.08541742e-02 2.79811025e-01 -2.00665697e-01 -2.02752829e-01 1.10757434e+00 -1.36456117e-01 -2.87423074e-01 -3.21941495e-01 -4.93161976e-01 -6.17186576e-02 9.22908723e-01 1.73464239e-01 -6.18231535e-01 -4.04099941e-01 -8.33046675e-01 1.47715688e-01 4.16863143e-01 7.35758185e-01 2.26791084e-01 -5.64180732e-01 -1.37568891e+00 5.50509244e-02 -1.37093991e-01 -5.07420182e-01 -2.01489493e-01 1.53552771e+00 -5.67817241e-02 4.77267772e-01 -3.30614358e-01 -3.09536338e-01 -1.30673289e+00 6.45457149e-01 4.71212476e-01 -6.16140842e-01 -5.64857066e-01 3.80078316e-01 -3.63465607e-01 -3.60676765e-01 9.78666544e-02 2.40149379e-01 -7.66048312e-01 7.30946422e-01 8.42811465e-01 6.15149796e-01 4.70876485e-01 -6.77596092e-01 -2.15577334e-01 -2.98873603e-01 -5.12793422e-01 -3.75646949e-02 1.73632729e+00 -3.42342108e-01 -6.16675973e-01 7.64791608e-01 8.89215648e-01 7.41652921e-02 -6.95413589e-01 -4.09334064e-01 6.50095165e-01 -1.75943933e-02 4.15834993e-01 -9.77962494e-01 -4.32684392e-01 5.75324535e-01 -1.27887517e-01 5.89419246e-01 4.92302001e-01 1.30850583e-01 7.52424359e-01 -2.19445765e-01 6.02314919e-02 -8.98570001e-01 1.06655620e-02 9.80968297e-01 9.53193426e-01 -1.08203650e+00 4.38734442e-01 -2.78583169e-01 -6.89419985e-01 9.49647546e-01 2.88625918e-02 -1.36013806e-01 5.06226778e-01 3.06369632e-01 1.77200466e-01 -6.27847314e-01 -1.21494746e+00 1.93863600e-01 3.11018646e-01 -9.34589207e-02 6.71030760e-01 -8.76114219e-02 -5.11876404e-01 8.48462880e-01 -2.32259214e-01 2.71578431e-01 7.31178701e-01 7.73661733e-01 -8.18450689e-01 -6.08450055e-01 -5.75901449e-01 2.23837584e-01 -9.16015506e-01 -3.43454406e-02 -8.25900018e-01 4.20764536e-01 -1.59392774e-01 1.01701951e+00 -1.67380035e-01 -5.10387540e-01 2.63971746e-01 5.04186869e-01 4.43028063e-02 -6.73214316e-01 -1.16791403e+00 -3.84680748e-01 7.33906746e-01 -5.07348418e-01 -4.87884641e-01 -8.15165460e-01 -8.78671467e-01 -5.97323775e-01 -4.96665955e-01 2.30455592e-01 6.16025090e-01 1.41092670e+00 4.77607436e-02 6.82022989e-01 8.17532957e-01 -7.92764068e-01 -9.88461673e-01 -1.45228422e+00 -3.75532717e-01 5.25370598e-01 7.85927296e-01 -6.72270656e-01 -5.28011739e-01 -1.97018310e-02]

[8.224037170410156, 10.125519752502441]

1b936b02-1bad-45d9-bd84-d1b69c3d3185

tarn-temporal-attentive-relation-network-for

1907.09021

https://arxiv.org/abs/1907.09021v1

https://arxiv.org/pdf/1907.09021v1.pdf

TARN: Temporal Attentive Relation Network for Few-Shot and Zero-Shot Action Recognition

In this paper we propose a novel Temporal Attentive Relation Network (TARN) for the problems of few-shot and zero-shot action recognition. At the heart of our network is a meta-learning approach that learns to compare representations of variable temporal length, that is, either two videos of different length (in the case of few-shot action recognition) or a video and a semantic representation such as word vector (in the case of zero-shot action recognition). By contrast to other works in few-shot and zero-shot action recognition, we a) utilise attention mechanisms so as to perform temporal alignment, and b) learn a deep-distance measure on the aligned representations at video segment level. We adopt an episode-based training scheme and train our network in an end-to-end manner. The proposed method does not require any fine-tuning in the target domain or maintaining additional representations as is the case of memory networks. Experimental results show that the proposed architecture outperforms the state of the art in few-shot action recognition, and achieves competitive results in zero-shot action recognition.

['Georgios Zoumpourlis', 'Mina Bishay', 'Ioannis Patras']

2019-07-21

['zero-shot-action-recognition', 'few-shot-action-recognition']

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

[ 5.37092090e-01 -1.53259650e-01 -4.54366505e-01 -2.94330746e-01 -7.29480445e-01 7.22615346e-02 8.50785375e-01 -7.88935497e-02 -6.73902869e-01 3.41514349e-01 4.37772125e-01 1.81448199e-02 -2.84602821e-01 -6.54495835e-01 -6.43022478e-01 -6.83858335e-01 -1.17924869e-01 2.90260047e-01 4.01802719e-01 -7.00637773e-02 4.95498925e-01 2.24014089e-01 -1.65091312e+00 4.76660848e-01 3.60383511e-01 1.18190551e+00 1.21298224e-01 8.57004881e-01 -2.26274524e-02 1.47431374e+00 -3.98581505e-01 -2.36648723e-01 2.72539020e-01 -9.61676300e-01 -1.05564737e+00 3.71936828e-01 4.37570274e-01 -4.37266886e-01 -6.82369411e-01 7.91689754e-01 4.27505493e-01 7.68584549e-01 6.39043212e-01 -1.19254351e+00 -6.99629545e-01 3.56900275e-01 -3.94562870e-01 7.23007321e-01 4.07824039e-01 4.58976120e-01 7.99491763e-01 -7.70578623e-01 6.40895426e-01 1.21582198e+00 3.32607329e-01 5.53968489e-01 -7.94399917e-01 -1.96980610e-01 3.27415168e-01 8.16848457e-01 -1.06613040e+00 -7.31413782e-01 6.17277980e-01 -4.48908180e-01 1.41079056e+00 -1.72953039e-01 6.36207521e-01 1.40278172e+00 3.45441878e-01 7.34934509e-01 5.88301122e-01 -4.59501326e-01 4.14837033e-01 -6.34126961e-01 5.71359955e-02 4.99002099e-01 -3.22860926e-01 1.35033578e-01 -4.93662745e-01 2.09642127e-01 7.23232806e-01 6.07565284e-01 -8.26884210e-02 -4.58487540e-01 -1.31772304e+00 7.21976876e-01 4.14185107e-01 7.11623311e-01 -6.83269978e-01 4.64381129e-01 7.78177559e-01 5.76423466e-01 4.09469515e-01 2.81776130e-01 1.87712740e-02 -4.53244686e-01 -8.72263312e-01 -8.34880248e-02 4.46833730e-01 5.05686939e-01 5.10976315e-01 2.91928858e-01 -6.04549527e-01 6.12474501e-01 -4.56297249e-02 -4.85486351e-02 1.15656543e+00 -9.31227088e-01 6.06412470e-01 3.54834855e-01 1.15574785e-01 -8.05349290e-01 -5.64413555e-02 8.76007006e-02 -7.33960748e-01 6.46787882e-02 2.25329757e-01 7.33007044e-02 -1.16237795e+00 1.51968539e+00 -5.86970672e-02 8.79123390e-01 3.63397628e-01 1.00343120e+00 6.64748728e-01 6.61198556e-01 2.36478418e-01 -4.41955090e-01 1.18295813e+00 -1.41976666e+00 -7.69192815e-01 -3.08511376e-01 7.63313472e-01 -2.77912438e-01 7.28837788e-01 -4.74248342e-02 -1.11051941e+00 -8.11776280e-01 -1.06882811e+00 4.24835496e-02 -5.03654420e-01 -3.79747272e-01 2.42299363e-01 1.37754619e-01 -8.07382762e-01 9.49471772e-01 -1.00971603e+00 -7.24178910e-01 4.20841247e-01 6.94268243e-03 -4.67889279e-01 -1.70315906e-01 -1.18614376e+00 1.04169559e+00 6.50976479e-01 -3.02924421e-02 -1.19311798e+00 -3.00930291e-01 -1.18693531e+00 1.61760688e-01 7.18590081e-01 -5.74309111e-01 1.29922795e+00 -1.41664457e+00 -1.66007888e+00 7.03363776e-01 4.35640477e-02 -8.71767879e-01 2.76508451e-01 -2.39758685e-01 -4.96244222e-01 3.97924870e-01 -1.21343173e-01 4.79169071e-01 9.03073430e-01 -4.18256819e-01 -5.43050230e-01 -3.35861593e-01 3.93717974e-01 2.72475690e-01 -1.59656882e-01 1.74492091e-01 -3.12883615e-01 -6.99605465e-01 -1.18786924e-01 -7.03899920e-01 -2.69572407e-01 -8.37247819e-02 1.28014326e-01 -3.43347102e-01 8.48056555e-01 -4.12215650e-01 9.16594982e-01 -2.25874877e+00 4.59768265e-01 -3.02730352e-01 -1.72816798e-01 5.82364142e-01 -6.03404522e-01 6.58863604e-01 -3.28200877e-01 -3.04814219e-01 -2.68102109e-01 -2.11812407e-01 -1.06330700e-01 3.75898808e-01 -1.31547615e-01 5.39845586e-01 2.68781126e-01 1.06366026e+00 -1.18149853e+00 -4.54685122e-01 5.26346385e-01 3.59413445e-01 -1.74045995e-01 5.06629050e-01 -1.85040697e-01 3.78000885e-01 -2.84988195e-01 4.65586036e-01 -5.13875000e-02 -1.41489133e-01 2.45587811e-01 -9.17168781e-02 -9.66106802e-02 2.48640150e-01 -9.93729234e-01 2.11704683e+00 -6.07034028e-01 5.63801587e-01 -6.45222008e-01 -1.47378612e+00 8.04095089e-01 6.30277634e-01 6.12403214e-01 -1.09949064e+00 2.80889750e-01 -1.32253706e-01 1.21675991e-01 -7.60151565e-01 3.16558093e-01 -2.28162661e-01 -2.37503797e-02 6.19892538e-01 4.57941383e-01 4.63335872e-01 4.49297845e-01 -7.02277347e-02 1.31011879e+00 3.54641706e-01 7.79336929e-01 2.90737808e-01 6.07791722e-01 -2.23537192e-01 6.10759676e-01 6.50518298e-01 -3.54895473e-01 6.38164639e-01 3.16592664e-01 -6.80704117e-01 -8.90323818e-01 -6.19815648e-01 5.10470212e-01 1.31725597e+00 1.38338745e-01 -3.45860392e-01 -5.03168464e-01 -6.50875628e-01 -2.74810106e-01 6.80213392e-01 -8.86655688e-01 -4.74853724e-01 -6.86431050e-01 -2.29754567e-01 2.59080648e-01 1.06327569e+00 6.29050493e-01 -1.51181400e+00 -1.22975743e+00 3.90767246e-01 -8.70352164e-02 -1.10279512e+00 -5.03566027e-01 2.57650942e-01 -8.70411575e-01 -1.06963193e+00 -8.70388985e-01 -7.86750734e-01 2.47863531e-01 3.19977641e-01 8.97681773e-01 -1.97331965e-01 -3.30236763e-01 6.76878572e-01 -7.93009341e-01 2.71703601e-02 -5.65153956e-02 -1.94523022e-01 -1.66228488e-01 3.88640434e-01 4.37398314e-01 -6.40294254e-01 -7.15528429e-01 3.35768223e-01 -9.90853608e-01 -3.28786731e-01 6.45350873e-01 8.74042392e-01 5.62618017e-01 -2.59956688e-01 4.88200516e-01 -5.08737385e-01 3.42432648e-01 -6.29060388e-01 -2.77146846e-01 5.51974535e-01 -2.32285947e-01 8.11130032e-02 6.52580917e-01 -6.90071404e-01 -8.23438823e-01 7.83837885e-02 1.13187455e-01 -1.07170212e+00 -2.95572132e-01 4.76524800e-01 9.82878059e-02 1.53256044e-01 4.36073303e-01 3.95792544e-01 1.73413213e-02 -3.92847598e-01 4.97333348e-01 5.32954752e-01 6.29639924e-01 1.63266491e-02 3.21224183e-01 4.30800647e-01 -1.97189912e-01 -6.26400411e-01 -8.11545491e-01 -6.82600200e-01 -1.09065604e+00 -3.36585999e-01 1.26779449e+00 -7.70681858e-01 -4.83415961e-01 4.25242782e-01 -9.61488605e-01 -5.07450104e-01 -5.98546326e-01 6.64216340e-01 -1.19253898e+00 3.03215474e-01 -5.06525815e-01 -7.73553252e-01 -2.01360598e-01 -8.52368176e-01 9.58959460e-01 2.04130262e-01 -2.09347144e-01 -8.17321122e-01 4.08201218e-01 2.11779714e-01 2.68474370e-01 3.25130999e-01 5.64975023e-01 -7.71927476e-01 -4.26772237e-01 -2.04292640e-01 -3.83380428e-02 1.70305312e-01 1.72144145e-01 -2.70444661e-01 -6.69548929e-01 -3.76621693e-01 8.07823539e-02 -7.78716445e-01 1.14745736e+00 2.97483534e-01 1.11028600e+00 -2.33101621e-01 -1.18150845e-01 3.91399652e-01 1.40954018e+00 7.22105563e-01 1.14708495e+00 2.07850337e-01 5.72511971e-01 4.52732533e-01 9.35441494e-01 5.18980145e-01 1.80272106e-02 9.66114521e-01 5.06680548e-01 2.56564885e-01 -9.97888520e-02 -5.60808554e-02 4.46055651e-01 5.74188411e-01 -4.28803712e-01 -3.09109986e-01 -7.01760232e-01 7.10057557e-01 -2.38864708e+00 -1.71703744e+00 5.90572894e-01 2.19573379e+00 2.23599449e-01 7.10774735e-02 3.52846920e-01 5.96104749e-02 7.35743165e-01 8.65035951e-01 -6.54648423e-01 -6.93428814e-01 2.94385523e-01 2.27006957e-01 1.41920418e-01 1.89376399e-01 -1.29711425e+00 9.09333706e-01 5.40172338e+00 5.90709984e-01 -1.12846339e+00 3.23133677e-01 3.26386511e-01 -3.74012768e-01 5.37491918e-01 -1.36072949e-01 -2.53992885e-01 3.94941896e-01 1.29529548e+00 -1.38192773e-01 2.57427871e-01 8.16141188e-01 -2.71953084e-03 7.51477527e-03 -1.32628167e+00 1.09247267e+00 5.51499069e-01 -1.25353205e+00 -2.42562462e-02 -2.65157074e-01 4.60863769e-01 -3.89100052e-02 -1.33834556e-01 6.29193366e-01 7.91691467e-02 -9.71958876e-01 6.05560601e-01 8.21768582e-01 6.29462659e-01 -7.02956200e-01 7.84085393e-01 2.84914464e-01 -1.37244284e+00 -4.14852053e-01 -4.87188846e-01 -2.46125966e-01 3.40521574e-01 -3.11438382e-01 -3.49174201e-01 6.06480360e-01 5.04206002e-01 1.21015966e+00 -3.32726717e-01 1.05179501e+00 8.58467221e-02 2.23992154e-01 3.90283823e-01 2.38418374e-02 7.25423157e-01 -2.82307658e-02 3.65301490e-01 1.10945857e+00 2.52207547e-01 3.99310112e-01 4.20370191e-01 2.28116259e-01 -1.72438961e-03 -5.83334528e-02 -8.77108812e-01 -2.99297720e-01 4.15522903e-02 8.65081251e-01 -6.62446558e-01 -6.01054251e-01 -7.20220566e-01 1.38114405e+00 4.14973259e-01 2.14610457e-01 -7.65954852e-01 -4.27494168e-01 5.68612933e-01 -1.76308706e-01 9.58933055e-01 -1.30993025e-02 5.87872326e-01 -1.15148532e+00 -8.16971809e-02 -6.85948133e-01 6.42506421e-01 -8.19287777e-01 -1.19239771e+00 6.60902023e-01 -1.01638548e-02 -1.65069246e+00 -7.72665739e-01 -3.21011603e-01 -8.83262455e-01 3.94964486e-01 -1.27233493e+00 -8.86884868e-01 -1.75560489e-01 7.75227249e-01 1.18124914e+00 -2.31525436e-01 8.41504037e-01 2.46108517e-01 -5.54272354e-01 4.11814302e-01 3.42436247e-02 2.89569080e-01 5.56047559e-01 -8.22115064e-01 4.34625328e-01 8.22211802e-01 4.02323723e-01 2.40816161e-01 4.35171008e-01 -4.60353732e-01 -1.28956318e+00 -1.22951329e+00 7.65995979e-01 -8.00003782e-02 7.17868030e-01 1.15613574e-02 -9.54081178e-01 7.30922163e-01 3.89474183e-01 4.52443838e-01 6.76183343e-01 -1.74571738e-01 -4.45493460e-01 -7.74292201e-02 -8.50615442e-01 4.32371408e-01 1.30414617e+00 -6.94553256e-01 -1.00693524e+00 2.96354145e-01 5.96419394e-01 -1.78433880e-01 -7.79987335e-01 3.02050233e-01 7.44902134e-01 -1.06706774e+00 9.02003586e-01 -1.22559357e+00 5.39125264e-01 -1.75424710e-01 -2.87794232e-01 -1.18891060e+00 -6.14111960e-01 -2.54961252e-01 -3.46613854e-01 8.05400133e-01 -2.23471094e-02 -2.47573435e-01 4.76317257e-01 1.81307465e-01 -2.88074374e-01 -8.64245594e-01 -9.87520933e-01 -1.17989862e+00 -3.47189486e-01 -2.00067297e-01 2.97387868e-01 7.20183790e-01 6.39256462e-02 3.88865352e-01 -7.54258752e-01 -2.05798075e-01 2.58347481e-01 4.39038962e-01 4.77542788e-01 -9.13095772e-01 -4.53445822e-01 -3.63912314e-01 -1.04813600e+00 -7.76179314e-01 4.08697903e-01 -7.09789872e-01 2.23238871e-01 -1.42127311e+00 2.71449983e-01 5.49103439e-01 -8.30279291e-01 5.03809154e-01 1.00280568e-01 4.24578756e-01 4.59078431e-01 2.21846879e-01 -1.28335118e+00 7.97753155e-01 8.90522778e-01 -3.80937576e-01 -6.41958788e-03 -1.47565454e-01 -9.67554152e-02 5.55154800e-01 5.56850672e-01 -4.44511026e-01 -5.75098276e-01 -3.25222760e-01 -3.79688054e-01 5.15844047e-01 5.68191350e-01 -1.25079894e+00 3.03217292e-01 -2.05368742e-01 2.02752694e-01 -4.54108953e-01 5.09916544e-01 -7.28914976e-01 -5.58804087e-02 6.12931490e-01 -6.13855839e-01 6.01481274e-02 -2.32873037e-01 9.33532059e-01 -4.57055598e-01 -3.08251768e-01 8.70978653e-01 -3.83615911e-01 -1.28652501e+00 3.39187831e-01 -4.82073188e-01 1.73903089e-02 1.31829441e+00 -5.24732709e-01 -2.49320909e-01 -2.97002405e-01 -8.98294508e-01 6.99480250e-02 2.39825502e-01 6.20392740e-01 8.60509336e-01 -1.50538802e+00 -4.45239037e-01 -9.10456479e-02 5.13706803e-01 -6.40777171e-01 4.43631649e-01 9.38953757e-01 -4.70895693e-02 4.09969926e-01 -7.48502254e-01 -4.02415603e-01 -1.10987282e+00 9.36296701e-01 3.73761892e-01 -3.17605525e-01 -8.78915668e-01 5.82419932e-01 3.05980891e-02 1.82344481e-01 4.43121105e-01 3.76317576e-02 -4.88594979e-01 3.75528783e-01 8.81490588e-01 3.83859396e-01 -1.49409190e-01 -8.10026407e-01 -3.45775723e-01 5.56667507e-01 -2.45481327e-01 -3.34542170e-02 1.47009909e+00 2.82000490e-02 3.11591566e-01 9.64716733e-01 1.31941831e+00 -1.09142780e+00 -1.36844456e+00 -3.64416331e-01 2.22964175e-02 -6.28192425e-01 -4.82315198e-02 -4.40685868e-01 -1.12684119e+00 8.37543249e-01 8.37091982e-01 -5.25512509e-02 1.06477416e+00 -4.50127237e-02 9.58430350e-01 5.46445787e-01 3.38773370e-01 -1.13142705e+00 5.57349324e-01 5.83465636e-01 7.86083043e-01 -1.37009752e+00 -1.83610559e-01 3.13774258e-01 -8.25375676e-01 1.04390240e+00 6.82094634e-01 -4.46594089e-01 4.30292398e-01 -2.65664339e-01 -1.36095956e-01 -3.09581041e-01 -1.10789561e+00 -6.16660058e-01 2.98488259e-01 5.46584666e-01 2.92894483e-01 -3.29037458e-01 -2.90270805e-01 2.05240965e-01 6.50091052e-01 2.08189383e-01 3.02819997e-01 1.22058713e+00 -4.30485755e-01 -8.34973395e-01 6.25227988e-02 4.14839178e-01 -2.47989818e-01 1.52770907e-01 -3.63356680e-01 6.18212283e-01 -9.26499292e-02 8.34736943e-01 5.14310300e-01 -4.00980234e-01 4.60923404e-01 4.09590989e-01 6.01046979e-01 -7.57867098e-01 -6.24753475e-01 -1.78768411e-01 -7.67919347e-02 -1.03038847e+00 -9.20822680e-01 -6.28661156e-01 -8.88616502e-01 9.66725722e-02 -6.81271553e-02 -2.33373176e-02 2.21474245e-01 1.30895591e+00 3.08329195e-01 6.48022830e-01 6.00881398e-01 -1.08080339e+00 -4.42385495e-01 -1.03460431e+00 -4.87421930e-01 6.83762610e-01 2.92153537e-01 -8.45042825e-01 -5.30313700e-02 5.99327683e-02]

[8.339781761169434, 0.8017737865447998]

122d15b9-a6c1-4f82-998c-279c776bf5d4

active-class-selection-for-few-shot-class

2307.02641

https://arxiv.org/abs/2307.02641v1

https://arxiv.org/pdf/2307.02641v1.pdf

Active Class Selection for Few-Shot Class-Incremental Learning

For real-world applications, robots will need to continually learn in their environments through limited interactions with their users. Toward this, previous works in few-shot class incremental learning (FSCIL) and active class selection (ACS) have achieved promising results but were tested in constrained setups. Therefore, in this paper, we combine ideas from FSCIL and ACS to develop a novel framework that can allow an autonomous agent to continually learn new objects by asking its users to label only a few of the most informative objects in the environment. To this end, we build on a state-of-the-art (SOTA) FSCIL model and extend it with techniques from ACS literature. We term this model Few-shot Incremental Active class SeleCtiOn (FIASco). We further integrate a potential field-based navigation technique with our model to develop a complete framework that can allow an agent to process and reason on its sensory data through the FIASco model, navigate towards the most informative object in the environment, gather data about the object through its sensors and incrementally update the FIASco model. Experimental results on a simulated agent and a real robot show the significance of our approach for long-term real-world robotics applications.

['Alan R. Wagner', 'Sarah M. Rajtmajer', 'Harsh Tyagi', 'Ali Ayub', 'Christopher McClurg']

2023-07-05

['class-incremental-learning', 'few-shot-class-incremental-learning', 'incremental-learning', 'navigate']

['computer-vision', 'methodology', 'methodology', 'reasoning']

[ 2.35850155e-01 2.70519525e-01 1.38331940e-02 -3.78812850e-01 -2.74453789e-01 -5.66782892e-01 7.62603343e-01 2.68633604e-01 -7.03618407e-01 8.59578371e-01 -2.72275656e-01 1.87686339e-01 -4.99334097e-01 -9.67026889e-01 -7.47847795e-01 -6.84720218e-01 -4.41288203e-01 9.32204187e-01 1.12421691e+00 -5.93455672e-01 5.07060826e-01 5.59694707e-01 -2.25332904e+00 -1.04270987e-01 8.93328607e-01 6.32082343e-01 9.85767722e-01 5.39597154e-01 -9.55401510e-02 1.06825852e+00 -1.83535546e-01 4.78579342e-01 3.38473469e-01 -1.92496240e-01 -8.80743384e-01 1.35987287e-03 6.97468743e-02 -1.94161296e-01 1.71426237e-01 9.59184110e-01 5.32715738e-01 6.97857797e-01 4.62012261e-01 -1.26043189e+00 1.64306071e-02 6.98636711e-01 7.32626319e-02 -2.95348376e-01 6.58457816e-01 2.50387102e-01 5.01568913e-01 -1.09591484e+00 1.17218649e+00 1.25184965e+00 5.63173294e-01 7.49946475e-01 -8.86049032e-01 -2.71411121e-01 6.59091234e-01 6.76520586e-01 -8.66384089e-01 -4.96711254e-01 8.06385756e-01 -2.21362665e-01 9.08893704e-01 1.33047197e-02 1.15795302e+00 7.72893071e-01 1.50687382e-01 1.07229662e+00 9.82868433e-01 -6.82551086e-01 1.09909165e+00 1.43959105e-01 2.62736887e-01 7.68381000e-01 1.72596008e-01 1.59181729e-01 -7.08997726e-01 1.73195720e-01 4.93253291e-01 -1.13964505e-01 1.61715031e-01 -1.26268053e+00 -1.18839395e+00 7.32942998e-01 5.68203211e-01 4.02416021e-01 -5.19704223e-01 1.20249070e-01 6.02069981e-02 3.54499400e-01 5.13960272e-02 6.99754655e-01 -4.26559925e-01 -2.06221342e-01 -2.56315738e-01 4.19798374e-01 9.41348732e-01 1.09230232e+00 1.02601147e+00 -2.31941804e-01 9.07419622e-02 9.43092585e-01 5.75230479e-01 2.65367001e-01 5.24642289e-01 -1.46880269e+00 -4.17459570e-02 7.18409657e-01 2.92609960e-01 -4.40063268e-01 -5.90779543e-01 -1.72400177e-01 5.37590422e-02 1.12256098e+00 -1.00601301e-01 -1.53369501e-01 -1.04170406e+00 1.54497278e+00 6.61377192e-01 -1.21154569e-01 2.99396455e-01 6.93138421e-01 4.63829964e-01 5.34462929e-01 -2.58610044e-02 -2.68557578e-01 7.31389761e-01 -1.29636109e+00 -5.38116097e-01 -6.48367286e-01 5.88813245e-01 -1.16571791e-01 8.97769094e-01 6.96113229e-01 -8.16512346e-01 -8.07486951e-01 -1.31776237e+00 3.42419058e-01 -7.55688369e-01 -4.11060423e-01 7.98543513e-01 2.37856373e-01 -1.10738945e+00 5.95530152e-01 -1.15167046e+00 -9.71480787e-01 3.09312761e-01 4.89212841e-01 -2.68211067e-01 -2.13962734e-01 -7.43789434e-01 1.20838022e+00 8.70629251e-01 -1.20941363e-02 -1.16510069e+00 -1.24651745e-01 -8.27382326e-01 -4.49256986e-01 1.03760302e+00 -7.44117677e-01 1.59814394e+00 -6.67593181e-01 -1.96322858e+00 -1.11108180e-02 1.22518241e-01 -4.81918067e-01 2.91406542e-01 -3.64435881e-01 2.24264637e-01 4.34462912e-02 3.20735216e-01 1.01710558e+00 3.54967713e-01 -1.75849998e+00 -1.15238714e+00 -4.41377759e-01 3.53282094e-01 7.31033683e-01 -7.94839486e-02 -6.32204533e-01 -3.13829780e-01 -6.03077002e-02 4.75780696e-01 -1.18490386e+00 -7.29164481e-01 1.40374333e-01 1.69239864e-01 -5.71368337e-01 1.01040339e+00 3.98842394e-01 4.98207897e-01 -1.87181282e+00 4.37548518e-01 -2.66721081e-02 -1.60849884e-01 1.55317664e-01 -1.15506835e-01 6.39942408e-01 5.30654967e-01 -5.58547378e-01 -3.08913916e-01 -4.60331768e-01 -7.14500621e-02 6.68154955e-01 -7.33997626e-03 9.50697064e-02 -2.65932262e-01 6.72114909e-01 -1.50073278e+00 -3.89087766e-01 7.57771432e-01 1.53002605e-01 -5.96180558e-01 1.01412907e-01 -5.00632346e-01 4.73121405e-01 -6.26672626e-01 5.48181772e-01 4.01344866e-01 2.84410119e-01 2.42456663e-02 3.63287300e-01 -5.62485278e-01 -8.81111920e-02 -1.42605007e+00 2.25255585e+00 -4.38130736e-01 4.65137929e-01 1.45575345e-01 -9.29443836e-01 7.38104165e-01 -2.24829484e-02 5.37323415e-01 -7.35839784e-01 -4.24279422e-02 4.23687547e-01 -1.12538360e-01 -8.04055095e-01 5.46229124e-01 9.68824252e-02 5.71208708e-02 1.27586171e-01 3.90239745e-01 -4.18171585e-01 5.72198093e-01 2.21297055e-01 1.13488877e+00 7.38749027e-01 4.28617001e-01 -1.65630251e-01 4.81579602e-01 3.28677624e-01 2.78504968e-01 1.46179593e+00 -4.24001157e-01 1.97165981e-01 -2.59863704e-01 -4.74586040e-01 -5.30913413e-01 -1.05355203e+00 1.65036559e-01 1.42135203e+00 7.37242162e-01 -1.73741266e-01 -5.72219968e-01 -7.34620750e-01 -3.41179699e-01 9.80623066e-01 -6.72283113e-01 -1.88003797e-02 -6.54729187e-01 -4.31800663e-01 -3.03982347e-01 3.42340261e-01 4.55308527e-01 -1.54138970e+00 -1.62557530e+00 6.47646725e-01 2.38268360e-01 -6.39625847e-01 4.72975791e-01 7.45204389e-01 -7.90635467e-01 -1.01053965e+00 -3.26854736e-01 -1.00929141e+00 5.83742678e-01 3.71517628e-01 6.17995262e-01 -3.67094159e-01 -2.28515014e-01 8.96413505e-01 -9.80304956e-01 -7.71262944e-01 -3.35890532e-01 3.22756059e-02 1.64369479e-01 -2.41643071e-01 -7.54547492e-02 -7.01631248e-01 -3.43179673e-01 2.13172272e-01 -7.19579101e-01 1.11174203e-01 7.20230341e-01 4.25448298e-01 5.18320084e-01 1.41299322e-01 6.81836069e-01 -9.32030797e-01 3.86228740e-01 -5.61529398e-01 -6.38804138e-01 1.09217271e-01 -8.15016210e-01 9.17998031e-02 2.79904574e-01 -6.39258564e-01 -1.20916426e+00 6.35323584e-01 -4.29486968e-02 8.52077603e-02 -2.37625450e-01 6.99884534e-01 7.33672008e-02 -2.50451177e-01 8.29641938e-01 2.90615171e-01 -1.00491472e-01 -3.59076291e-01 5.78832567e-01 4.74176764e-01 6.39054298e-01 -3.90413970e-01 4.97677684e-01 5.23218393e-01 1.16323559e-02 -6.44707620e-01 -6.51129127e-01 -7.06298709e-01 -8.71085525e-01 -6.90856695e-01 4.64907527e-01 -5.71599960e-01 -5.71257710e-01 6.61997974e-01 -8.87980938e-01 -7.58052111e-01 -8.15842628e-01 6.09573364e-01 -9.91870821e-01 1.77641675e-01 -5.25536649e-02 -1.17326176e+00 1.27746493e-01 -1.10331595e+00 9.00375247e-01 5.12897849e-01 -3.61473523e-02 -7.07178295e-01 5.76923668e-01 -2.63800174e-01 5.52728176e-01 9.71887410e-02 3.64189088e-01 -6.81831896e-01 -7.69219041e-01 -1.13461383e-01 4.46321189e-01 -2.72665143e-01 9.93753076e-02 -4.33750480e-01 -7.56321430e-01 -4.09155220e-01 1.67714003e-02 -4.16120201e-01 7.95681179e-01 1.86062351e-01 5.79142213e-01 9.46169347e-02 -7.21600354e-01 1.15469225e-01 1.47896385e+00 8.15075755e-01 4.18824822e-01 8.28073323e-01 3.21586370e-01 6.94950819e-01 1.20898652e+00 4.46664482e-01 4.77686584e-01 7.80167818e-01 9.23090458e-01 5.08393228e-01 -2.99775273e-01 -1.12964213e-02 3.37573916e-01 6.35643423e-01 -1.49362087e-01 -2.01594636e-01 -7.54869759e-01 6.57619119e-01 -2.40068364e+00 -8.52214396e-01 1.21895470e-01 2.02370930e+00 5.26526392e-01 4.00141835e-01 -1.11871786e-01 1.38789430e-01 3.44007015e-01 -1.84849203e-01 -8.73998165e-01 -3.93881910e-02 1.35764703e-01 -3.91268507e-02 2.82178462e-01 8.53108704e-01 -1.12727559e+00 1.11459708e+00 5.66576529e+00 6.82400048e-01 -7.97765672e-01 1.36849776e-01 -2.42045447e-01 2.57637594e-02 1.30304173e-01 2.26725057e-01 -8.05217445e-01 1.43214270e-01 6.18346572e-01 4.68721054e-02 4.93199855e-01 1.35262787e+00 -7.64843449e-02 -8.82076800e-01 -1.17763579e+00 6.90951586e-01 1.49692491e-01 -1.03335738e+00 -2.11765826e-01 -8.52320269e-02 6.98933423e-01 3.76501113e-01 -2.03507990e-01 6.53466940e-01 7.52312362e-01 -3.74013811e-01 1.11625588e+00 6.88684344e-01 1.81288183e-01 -4.04090613e-01 7.84364402e-01 8.26557100e-01 -1.20561433e+00 -6.76527083e-01 -3.02801460e-01 -3.29348266e-01 2.46865764e-01 5.03971688e-02 -1.27137232e+00 5.34843624e-01 7.54952013e-01 7.43966699e-01 -7.61606395e-01 1.46458459e+00 -1.60903439e-01 8.16084668e-02 -5.29062212e-01 -5.89420974e-01 3.46943378e-01 5.73560130e-03 8.26416790e-01 6.76516533e-01 3.18921387e-01 1.87953249e-01 4.88729894e-01 3.83134544e-01 5.09008348e-01 -1.34468436e-01 -6.30466759e-01 4.51952338e-01 6.76881611e-01 1.07029140e+00 -1.16194475e+00 -5.32534480e-01 9.53144953e-03 8.96355987e-01 4.45001990e-01 2.01491416e-01 -2.41210863e-01 -5.76944828e-01 2.62949746e-02 -2.87773877e-01 4.20264333e-01 -3.54110628e-01 -3.70405316e-02 -7.52795160e-01 -8.76018405e-02 -4.26934749e-01 1.83714122e-01 -9.40199733e-01 -8.74629736e-01 6.11470461e-01 2.01920614e-01 -1.36230981e+00 -5.94334364e-01 -4.32075739e-01 -2.47670919e-01 9.28908661e-02 -1.33775175e+00 -1.11535001e+00 -6.25551045e-01 5.39550066e-01 1.29644656e+00 -3.26404899e-01 8.75359237e-01 -2.13877261e-01 9.34276357e-03 -2.64250726e-01 6.51805550e-02 -5.42239010e-01 3.36949229e-01 -1.26009238e+00 1.33079350e-01 6.66505456e-01 -2.93323137e-02 5.79382896e-01 9.60414648e-01 -9.03864384e-01 -1.55058086e+00 -8.17952693e-01 2.71825194e-01 -5.27717292e-01 3.16990405e-01 -3.79905611e-01 -6.50698304e-01 6.78842068e-01 7.32745230e-02 -2.31671706e-02 1.39978185e-01 -4.68770675e-02 1.84564337e-01 -4.52037267e-02 -1.21664846e+00 5.70991099e-01 1.37591374e+00 1.35300308e-01 -8.04217994e-01 1.83899716e-01 6.85232580e-01 -2.16266781e-01 -2.22346991e-01 7.89336681e-01 7.61889458e-01 -1.07927847e+00 8.35331738e-01 3.27177043e-03 -3.45024168e-01 -6.36192620e-01 -2.33732641e-01 -1.57813489e+00 -4.49607164e-01 -3.52555275e-01 -2.06598788e-01 8.26735258e-01 3.21417451e-01 -7.32312381e-01 8.46289158e-01 2.34521240e-01 -5.63755572e-01 -4.95786428e-01 -9.63716328e-01 -7.94316530e-01 -3.77414286e-01 -6.07260108e-01 3.61374617e-01 4.65613395e-01 2.08144471e-01 3.16532463e-01 -5.10468148e-02 5.67984842e-02 6.35245562e-01 -9.58217457e-02 8.52436900e-01 -1.69028234e+00 -4.98370752e-02 -2.13304624e-01 -3.66717279e-01 -9.17113721e-01 -1.21484831e-01 -5.50155222e-01 9.19163108e-01 -2.10526085e+00 6.62119463e-02 -8.66678357e-01 -1.97813258e-01 5.22175491e-01 7.19060376e-02 -6.17282726e-02 4.88501698e-01 3.49313200e-01 -1.43078470e+00 6.95682347e-01 9.25145686e-01 -5.63734099e-02 -7.60325074e-01 -1.26247099e-02 -2.41642341e-01 8.81320000e-01 6.00748479e-01 -4.29814219e-01 -7.95158744e-01 -2.51706362e-01 4.49463814e-01 -1.84405014e-01 -2.83365827e-02 -1.72595561e+00 9.19009209e-01 -2.61558175e-01 2.05426380e-01 -6.72116220e-01 7.72546411e-01 -1.13812280e+00 3.12256217e-02 8.64552259e-01 -3.55737895e-01 -5.49671292e-01 -3.87225039e-02 8.14984262e-01 1.76292971e-01 -7.44521201e-01 4.44468170e-01 -6.09363019e-01 -1.53237045e+00 -1.96695374e-03 -9.95896816e-01 -5.35555422e-01 1.36698854e+00 -4.95110363e-01 -1.85028955e-01 -1.82731241e-01 -9.82542098e-01 4.57085341e-01 5.24025202e-01 4.34206903e-01 7.34274805e-01 -1.07588041e+00 -2.55497068e-01 2.65117496e-01 4.13408458e-01 3.52177083e-01 6.11515902e-03 3.73446703e-01 -5.22079349e-01 3.38732839e-01 -4.72397298e-01 -6.60009325e-01 -8.85064483e-01 7.44184434e-01 1.09190695e-01 2.47799292e-01 -5.10949314e-01 8.67847800e-01 -1.17806531e-01 -9.86772895e-01 5.14962435e-01 -6.46089390e-02 -6.98729515e-01 2.55573869e-01 3.53003174e-01 6.49821758e-01 -4.97065410e-02 -3.51532340e-01 -3.42955291e-01 7.36290038e-01 1.94324106e-01 -6.05894327e-01 1.55828416e+00 -4.56903189e-01 9.00123417e-02 1.01182723e+00 6.15270972e-01 -7.09479228e-02 -1.56428397e+00 -2.45616466e-01 8.48650634e-02 -1.96566746e-01 -1.28018752e-01 -9.47998166e-01 -2.90477842e-01 3.43022048e-01 1.06852436e+00 2.41538182e-01 8.96923006e-01 2.68281192e-01 1.63979188e-01 1.15507972e+00 1.35829139e+00 -1.42377162e+00 3.89310747e-01 7.27393091e-01 7.65245676e-01 -1.25106835e+00 -4.22128290e-02 -3.49114597e-01 -4.69788551e-01 1.14872074e+00 7.26276100e-01 -1.59220397e-01 6.11184180e-01 1.08066447e-01 2.73074005e-02 -2.57111609e-01 -8.38762522e-01 -6.64266646e-01 -3.11428159e-01 1.02404928e+00 -3.67265970e-01 -2.17388704e-01 -1.56804442e-01 6.05947301e-02 -8.99202079e-02 3.02686095e-01 7.05999553e-01 1.72777283e+00 -1.30929565e+00 -1.04498315e+00 -1.97160184e-01 1.01439923e-01 3.86032701e-01 4.75970447e-01 -3.55600208e-01 7.09010541e-01 4.54404652e-01 1.11204565e+00 -1.45312294e-01 -3.21319014e-01 3.56211901e-01 4.08836156e-02 6.57982290e-01 -1.00342178e+00 -1.67057648e-01 -4.32102866e-02 7.76502565e-02 -7.34176219e-01 -8.20492208e-01 -9.86357450e-01 -1.56003785e+00 6.39005840e-01 -2.95596182e-01 2.20237240e-01 1.00958312e+00 9.65365827e-01 2.74757683e-01 5.21116555e-01 4.64155674e-01 -1.47217429e+00 -1.42635986e-01 -1.09363425e+00 -1.62050113e-01 -7.62373507e-02 3.07637513e-01 -1.14629507e+00 -8.98099765e-02 -1.71268582e-01]

[4.712668418884277, 0.6207226514816284]