dedup-isc-ft-v107-score
float64 0.32
1
| uid
stringlengths 32
32
| text
stringlengths 0
32.5k
| paper_id
stringlengths 1
14
| original_image_filename
stringlengths 5
222
|
|---|---|---|---|---|
0.397998 | a8e4001775be485a8da13a40724b412d | Confidence regions for the K-S test-based fits, along with the supremum of the joint p-values (likelihoods) for the simultaneously estimated parameters α and γ, for the redshift evolutionary models n ∈{0, 1, 2}. Rows correspond to the models n=0 (top), n=1 (middle) and n=2 (bottom) while columns represent the joint p-value contour plots for α vs. γ (left) and the supremum of those p-values for α (centre) and γ (right). The dashed red curve in the right column represents the likelihood distribution of the constraint p(γ|α=1.5). The survey fluence limits for the Parkes and ASKAP telescopes are held fixed at F_0,P = 3and F_0,A = 26during the fitting process. | 2012.15051 | errorpanel-2020-08-12-nAll-Fit-KolmogorovSmirnov.jpg |
0.439235 | 2afadee2159241bf83ccc87962992f21 | The structure of the encoder E and decoder G_L. We employ the pre-trained VGG19 <cit.> as the features extractor and use the features at five layers (i.e., Relu_X_1, X=1, 2, 3, 4, 5). The f_L_Relu_5_1 and f_L_Relu_4_1 are the features of I_L at corresponding layers, respectively. | 2012.15052 | gl.jpg |
0.523805 | 435afc32c7c749e2822c56a7c842780c | The structure of quality attention module (QAM). The QAM enables f_H to adaptively select content-relevant features from the spatial non-local inter-domain similarity and then incorporate the style attributes from the channel-wise statistics. | 2012.15052 | qam.jpg |
0.492812 | cba56a2b09e74e7c973db62e99218556 | v_out vs v_in when α=5 and β=1 with v_c ≈ 0.7295. The one-bounce window is in solid black. Two-bounce windows are in blue and three-bounce windows are in green. | 2012.15060 | vin_vout_alpha_5_beta_1.jpg |
0.500865 | 3ff669e690034aacb2f396e67414af5d | Folded X-ray light curve of FR Cnc in X-ray and optical bands. The colour curve is shown in the bottom panel. The error in single measurement of V is 0.008 mag. The continuous blue curve in the top panel is the best fit sinusoidal curve to the folded X-ray light curves. | 2012.15062 | frcnc_flc.jpg |
0.440248 | 7119b88626f24be492ff0e7e83064660 | B and V bands light curves of FR Cnc. Insets in each panel show the flare light curves in the respective band as observed on March 11, 2019. | 2012.15062 | frcnc_olc.jpg |
0.453142 | 380bfa84e72347a5b55f20c459c15352 | X-ray spectra along with the best fit 2-temperature plasma model apec. | 2012.15062 | frcnc_spec_2T.jpg |
0.47083 | 8be0137b6c8241b5bc86c4ba3255ef8d | Background subtracted X-ray light curve of FR Cnc in 0.5-2.0 keV energy band. | 2012.15062 | frcnc_xlc.jpg |
0.405711 | 7f44cca03bff403392be2084b49318aa | Long term X-ray variation of FR Cnc in 0.5-2.0 keV energy band. | 2012.15062 | frcnc_xlc_longterm.jpg |
0.427357 | c4584078fee44a97b8b563955a5a2322 | Momentum-resolved spectral function of Iron metal in ferromagnetic phase at 300 K. | 2012.15063 | fig3.jpg |
0.47213 | 38f872a93e5e41f79a446ae535f84bd1 | 2012.15063 | fig6b.jpg |
|
0.478761 | 8fd594f1f9254b83bf02270bba452d6b | (a) Scheme of buckling instabilities for the initial bulge and the secondary ring. (b) Wrinkling instability in a disk of contractility Λ=15 with the inner circle of lower contractility Λ=10 and decreased thickness demonstrating different number of wrinkles depending on the initial thickness and size of the inner circle at (b) h=2, R_2=2, (c) h=2, R_2=3, and (d) h=0.25, R_2=3. (e) The profile of the shell middle line in the vicinity of inner circle (only one half is shown due to symmetry) along the radial direction depicted by the white dashed line in (b) where the difference between deflections in two domains is comparatively small. Dotted line is the initial planar shape, while it develops a spherical cup shape in region with low contractility (red line) and small out-of-plane displacements in the outer region (purple line), which bring about the appearance of the secondary ring of radius R_r. | 2012.15064 | Buckling.jpg |
0.461412 | 36d275df472648828fee817c667adf22 | Steady states of a spherical shell colored according to the local concentration at different threshold c^∗ and initial thickness h. Patterns emerge as a consequence of a small initial chemical perturbation at a pole of the sphere (the central point in the top views). Higher concentration regions are associated with greater positive (outward) curvature. | 2012.15064 | D0_h_a.jpg |
0.397174 | 01a43f46fac1485081e30a61edbf3b98 | (a) Chemical patterns arising on a spherical shell of fixed uniform thickness h at D=0, c^∗=0.1, R=30. Shells obeying the volume conservation condition at (b) smaller radius D=0, h=2, c^∗ =0.1, R=15 demonstrate the same patterns obtained at large R, while (c) increasing diffusion leads to a larger spot size D=1.5, h=2, c^∗ =0.1, R=30, than seen in Fig.<ref>a. (d) Concentration and mean curvature at nodes of mesh on sphere depending on polar angle with the initial perturbation at the pole. The secondary peak at a nonzero polar angle represents the secondary ring that forms as a result of the buckling instability around the initial pertubation at the pole. | 2012.15064 | H-R-D-C.jpg |
0.448252 | ce533c5e16474014ae16dfd75c6fbd92 | Simulation snapshots of spontaneous pattern formation on a spherical shell at h=4, R=30, D=0, decreased contractility change Λ_max/Λ_min=1.05 and concentration threshold c^∗=0.05, but increased production rate w=100. The evolution reveals a fast phase of a spotted pattern formation followed by a slow coarsening. At the steady state (t=100000), only a few spots of high concentration with large radius remain, consistent with the theoretically predicted <cit.>, λ∼ε^-1/4, scaling. | 2012.15064 | LargeSpots.jpg |
0.419602 | cf8b2dfb63e74f4fa6887b9d68c32948 | Geometry of discretized shell that is constituted as a collection of nodes with positions 𝐱_i connected by links l_ij, where each node spans an area A_i around it. | 2012.15064 | ModelMesh.jpg |
0.440411 | 49b45c41052b487f88befaced8fa68c1 | Accuracy of users at (75 workers per condition). In the spoken modality, extractive-sent explanations yield the best results and is significantly better than confidence. In contrast, in the visual modality, extractive-long explanations perform best. We observe a statistically significant (p<0.01) difference between extractive-long in visual vs spoken, perhaps due to differences in user's cognitive limitations across modalities. | 2012.15075 | acc.jpg |
0.405445 | f6b19360e7764e82b23506cfe00fbe82 | Confidence before and after calibration. | 2012.15075 | calibration.jpg |
0.421142 | 8c9bc799d0f54db7a744724e706663bb | Voice clarity: Most participants found the voice of the assistant to be good or excellent. | 2012.15075 | clarity.jpg |
0.452714 | f5bd45dfa7644509b4307da56f4ff60f | (Left) Explanations significantly increased participant ability to detect correct answers compared to simply displaying confidence. (Right) However, only extractive-sent in the spoken modality and both explanations in the visual modality decreased the rate at which users are misled. | 2012.15075 | false-true-pos.jpg |
0.44467 | 0497d383c43243a7bee87a0dac396d18 | Helpfulness: Participants indicated how helpful responses were. These results reflect the large differences we see in performance (baseline vs the rest of the settings), but are not able to capture the more subtle differences among explanation strategies and confidence. | 2012.15075 | helpful.jpg |
0.403574 | 69553ada20e441e7979e033335c82051 | Using end-to-end user studies, we evaluate whether explanation strategies of open-domain question answering assistants help users decide when to trust (or reject) predicted answers. | 2012.15075 | landing_naacl.jpg |
0.372115 | 104bd209a5dd47508113757e7677a7e9 | Top: Users perceive the same explanation to be longer in the spoken modality. Bottom: While extractive-sent and abstractive were the same length, participants rate the latter as longer more often perhaps because of they contain more content. | 2012.15075 | length.jpg |
0.475461 | 592b15dfc9574a89b2c71d9d52728be9 | Reward: The scores presented here are out of $ 2.70. Although all explanations are better than confidence, the explanations leading to the highest rewards change across modalities. | 2012.15075 | reward.jpg |
0.423502 | 07671eb2ed8c43a2bcb581bcc814b57d | UI for visual (left) and spoken modalities (right) for extractive-sent explanation type. Users either read or hear an explanation and decide whether to trust or discard the QA system's prediction. | 2012.15075 | ui.jpg |
0.46004 | 9265a5bd9e1446999962006c33d3b9c5 | Adaptive action | 2012.15081 | ada_action-eps-converted-to.jpg |
0.421915 | ffba692b3c22474cbb190dafe6b85fe2 | CDF of performance difference in AUDR | 2012.15081 | eval_avg-eps-converted-to.jpg |
0.376546 | 971641dd2be04ec1ad9c192e053826ed | CDF of performance difference in 5TUDR | 2012.15081 | eval_fairness-eps-converted-to.jpg |
0.443442 | a543c1edc4f34ff2830bf28a2876cc82 | RBG allocated by different agents. | 2012.15081 | game_framework-eps-converted-to.jpg |
0.42597 | 270e54d4fc184976ba26a40ec845d507 | MARL framework for RBG allocation | 2012.15081 | marl_framework-eps-converted-to.jpg |
0.397265 | 754f88d8c5df46c39828dc5b802f49fd | Total transmission time with different number of RBGs | 2012.15081 | pie-eps-converted-to.jpg |
0.528715 | 956bf98f1fe64f588bad77c33b592e7f | (a) Mixing network structure. In red are the hypernetworks that produce the weights and biases for mixing network layers shown in blue. (b) The overall QMIX architecture. (c) Agent network structure. (d) The GRU structure, where h is the hidden information used for recursion <cit.>. | 2012.15081 | qmix-eps-converted-to.jpg |
0.461591 | 13087f94ccf84e669338d677075f78c2 | The increment of 5TUDR between two adjacent TTI in one simulation. | 2012.15081 | r1-eps-converted-to.jpg |
0.421796 | 54c5bb215cfa4ecdb0dfeea46f1eec7a | The increment of AUDR between two adjacent TTI in one simulation. | 2012.15081 | r2-eps-converted-to.jpg |
0.404145 | 5e793f4a1d0a47848284c89ab8e68427 | Coefficient matrix of MARL scheduling | 2012.15081 | rl_corr-eps-converted-to.jpg |
0.427345 | ef7decc7165e4467b32e921e055ed402 | Average 5TUDR of different schedulers in 1000 random simulations. | 2012.15081 | sample_fairness-eps-converted-to.jpg |
0.374814 | 876867af743e435f8f26fdf9ee0b09d8 | Total residence time and total transmission time | 2012.15081 | stay_alloc_time-eps-converted-to.jpg |
0.409774 | 77c99d555b3644aeaaa05385ebe3d015 | Network Mechanisms | 2012.15081 | system-eps-converted-to.jpg |
0.40027 | ba41617efe294c168341340159f3bff1 | Episode reward for epoch | 2012.15081 | train_reward-eps-converted-to.jpg |
0.398845 | 75f1d68b0e2a4042baff8424a9adb7eb | Distribution of the individual p-values for the Fermi 3LAC blazars and for the radio galaxies. | 2012.15082 | Histo-Pvals-IndivFermiRG_v2.jpg |
0.4505 | 304f019da0244047b8386e00de102bbe | Distribution of the sinus of the declination of the 56 high-energy track-like events detected by IceCube considered in this study. The blue filled histogram represents the Bayesian block fit to the data that is used in the likelihood analysis. | 2012.15082 | IC_DEC_BKGPDF.jpg |
0.429743 | f3d2c9feb48d48adb9085db0e819aea7 | Fermi-LAT measured gamma-ray flux for the BL Lac 4FGL J2255.2+2411, as a function of time in Modified Julian Date (black points). The red line represents the model gamma-ray light curve that is used in the time dependent likelihood (obtained via a Bayesian block algorithm). The blue vertical lines indicate the arrival time of ANTARES events located at an angular distance smaller than 2^∘ from the blazar. The vertical green line corresponds to the arrival time of the IceCube track IC-100608A. | 2012.15082 | NewPlotLC_4FGL_v2.jpg |
0.402477 | b285b6acd7d94831ac9e84a730e0f25b | Cumulative probability to reconstruct a track event direction within an angle α with respect to the true Monte Carlo neutrino direction. The thick solid black line corresponds to the average value for an E^-2 energy spectrum, while the thinner lines are obtained for different energy ranges (still with E^-2 weights included). | 2012.15082 | PSF_vsE_v2.jpg |
0.433817 | 7fc9662c4d5f44149f4da8d158f5e56f | Sky map in equatorial coordinates of the 8754 ANTARES track events (small black circles), together with the considered target sources: Fermi 3LAC blazars (blue circles), radio galaxies (orange stars), star-forming galaxies (violet triangles), dust-obscured AGN (red squares), and IceCube tracks (red empty circles). For astrophysical sources, the radius of each marker is proportional to the square root of the weight used in the likelihood. For IceCube tracks, the radius of each circle is proportional to the angular error (displayed angular size not to scale on the figure for clarity). The Galactic Center is indicated by a black star and the Galactic Plane by a blue dashed line. The blazar MG3 J225517+2409 and the radio galaxy 3C403 have their position indicated by their name. The blue shaded area corresponds to declinations out of the field of view of ANTARES. | 2012.15082 | SkyPlotAllTargets_biglabel.jpg |
0.49389 | 03648da4384d41569d5f9d6f9d068e82 | Jittered boxplots for liver weights | 2012.15093 | liver.jpg |
0.451216 | 563ea1ce30d04fcf981624255440d919 | Gradient Boosting Tree Model construction. T(X,Θ_k) is the best Tree built at step k, its parameters Θ_k are chosen in order to minimize the Loss Function between the target variable Y and the Boosted Model of the previous step. The β_k parameter is the weight of the Tree, when added in the Boosted Ensemble, this is also chosen with respect to the Loss Function. | 2012.15103 | GBM.jpg |
0.42232 | bd8afef5a02a487597e13ba9ecc80a31 | LIMEâs modus operandi. Courtesy of Marco Tullio Ribeiro <cit.> | 2012.15103 | Lime_Ribiero.jpg |
0.423851 | 75238e77a82a44788591c101c96f4210 | Shape of Logistic, Probit and Linear functions, associated with different parametrization. The Probability of Default, ranging from 0 to 1, is modelled against a single independent variable X. | 2012.15103 | Logit_Probit_Linear_Functions.jpg |
0.399706 | 6099f4af103d4bd580d551a22d105e06 | Lorenz Curve Comparison Gradient Boosting vs Logistic Regression | 2012.15103 | Lorenz_Curve_Finale.jpg |
0.431362 | 736ce6fb0fef480ab3e6cc17864e9ac3 | Illustration of CASSI from 3D scene to 2D coded image. | 2012.15104 | CASSI_flow.jpg |
0.450821 | 20d8ebabb5e0438493ca526d9c93b3c6 | Illustration showing the proposed fusion model. | 2012.15104 | Fuse.jpg |
0.440207 | b3098e7c0a8a4bf59b70c4999aed51b1 | (a) 3D patch extracted from the Toy dataset; (b) singular values in logarithm of reshaped patches V.S global image. We select 100 patches and average the singular values in logarithm. | 2012.15104 | SVD_values.jpg |
0.419557 | 09ffd518819447d7826c0f3d0c384403 | Illustration showing the DCCHI system. | 2012.15104 | illustration_flow.jpg |
0.556412 | eed9edfbaf874fc08cf2506dadec50d9 | The magnetic flux Φ(t) (in arbitrary code units) across a hemisphere on the black hole horizon. The initially frozen-in magnetic flux grows (in absolute value) due to accretion of plasma. At later stages the flux starts decreasing as the magnetic intensity decreases and the field eventually escapes to radial infinity. | 2012.15105 | hydro_bondi_flux.jpg |
0.444213 | 39ab0fae3c894a8682020db4c913eb5e | Graph of inverse accretion rate 1/Ṁ exhibits four phases of the magnetized flow evolution. The initial configuration starts with the Bondi spherical inflow from the outer boundary of the integration domain. The central black hole rotates with the Kerr spin parameter a and it is magnetized by the Wald uniform magnetic field B. Spherical symmetry of the inflow is quickly lost by its interaction with the magnetic field but the axial symmetry is imposed in our 2D computations. In case of a rapidly rotating black hole, the magnetic flux vanishes initially (Meissner effect) but it starts growing with accretion of the plasma. Part of the inflowing material is diverted to an outflow along the equatorial plane and accelerated by reconnection events (they are caused by numerical resistivity); the resulting accretion rate thus gradually drops and it exhibits some random glitches at later stages of its temporal evolution (see the text for further details). | 2012.15105 | hydro_bondi_inverse_accretion_rate2.jpg |
0.482269 | d3083fa1c5f64e1184bade66145947f4 | The accretion rate on the horizon of the black hole initially grows and then saturates as the steady state has been reached. The curves are normalized to an arbitrary value Ṁ_̇0̇ and parameterized by three values of the dimension-less spin a. Time is in geometrical units and scaled with the black hole mass M. | 2012.15105 | thermal_accretion_rate.jpg |
0.468562 | 2dc0b4e0795149c891356121f314334f | Intersection of conic hulls C_1,3 , C_3,5 and C_4,5 which corresponds to the series representation in region ℛ_3, see Eq.(<ref>). The orange region with edges along 𝐞_3 and 𝐞_5, is the common intersecting region and corresponds in fact to C_3,5. | 2012.15108 | AppellF1_3_New.jpg |
0.452688 | 11bcfbb8bb4a4c1b927327993dc6ec16 | Convergence regions of the five series representations of the MB integral in Eq.(<ref>). The well-known Appell F_1 double hypergeometric series converges in ℛ_1, its 4 analytic continuations in the other regions. | 2012.15108 | AppellF1_ROC.jpg |
0.427236 | e6a7296c5a2a4d5281b66fcef8219daa | Architecture of a fully-connected network with L hidden layers of neurons of constant size h. Points indicate neurons, connections between them are characterized by a weight. Biases are not represented here. The W's are the weights, 𝐱 is the input, and the a's are the preactivations as defined in Equations (1-3). Taken from <cit.>. In red: Neural representation of the data at different depths (illustrative). Neurons in the first layers respond to local, simple features, such as edges. Deeper in the network, neurons respond to more and more high-level features such as cats or human faces. Adapted from <cit.>. | 2012.15110 | net_features.jpg |
0.469908 | 0f38fd216ca44acfa9ff74f510b0c985 | Relationship between the kernel PCA of the NTK (obtained both at initialization and after learning) and the labels. The mutual information (characterizing the amount of correlation) between the top eigenvectors of the Gram matrix and the labels (A, B) and their projection ω_λ (C, D) get large after learning for both the stripe model (learnt with a FC with one hidden-layer) and MNIST (learnt with a CNN). From <cit.>. | 2012.15110 | omega_i_mi.jpg |
0.446378 | f3477560d0d84f9392ea0a649cd53a6d | Performance for Simulation 2 using learning rate 5·10^-5 and a batch size of 80. | 2012.15116 | ex3_plt0.jpg |
0.44675 | f981f7af49cc4ffcba351a7440bac49c | The same plots as in Figs. <ref>&<ref>, but given for run 3 (m_i/m_e=100). | 2012.15117 | m100_01_12.jpg |
0.425983 | 13ba03c3a6c14b909be5a9aa8af5f80e | From top to bottom rows: the ion longitudinal phase space, density field, transversely averaged density profile, B_z field, and transversely averaged perpendicular magnetic field profiles. The plots are given at early (left column) and late (right column) stages of the shock evolution in run 1 (m_i/m_e=16). | 2012.15117 | m16_01_13.jpg |
0.444356 | 674ad8077d814e018e1f68acb1646482 | From top to bottom: ion and electron longitudinal phase spaces, transversely averaged density and perpendicular magnetic field profile. The plots are given at the end time of the long run 4 (m_i/m_e=16). | 2012.15117 | m16xl_68.jpg |
0.389303 | ba3ec51b9ef34615b52886cf8d61f19c | The same plots as in Fig. <ref>, but given for run 2 (m_i/m_e=50). | 2012.15117 | m50_03_78.jpg |
0.543426 | dbe1a543d694489a9a4f96bb6aab24fd | The evolution of the downstream ion (red coloring) and electron (blue coloring) spectra in the simulation run 4 (m_i / m_e = 16). Each color-coded line represents the spectra in a different time during the evolution. The darker lines correspond to earlier times, and the lighter lines correspond to later times. On the horizontal axis, the momentum p/mc is given. On the verical axis, the normalized 4 π p^4 f(p) is given, where p is in the units of mc. | 2012.15117 | spec_m16xl_evolution.jpg |
0.434912 | 810e992e984d416d8edf14d4a16f0982 | Sample layout and microscopy images of the T-junction. (a) Schematic illustration of the layout of a T-shaped selectively-grown 3D TI nanostructure with in-plane magnetic field-driven steering effect indicated. (b) Scanning electron microscopy image of a T-junction device. The contacts near (far from) the T are referred to as the inner (outer) contacts. The magnetic field is applied in-plane. The input current I_, which is injected from the outer contact on the bottom leg, splits off into currents I_ and I_, which are measured at the outer contacts on the right and left leg of the T, respectively. (c) Transmission electron micrograph of the cross section of an AlO_x-capped Bi_2Te_3 nanoribbon. The scale bar indicates 50 nm. The thickness of the ribbon is determined to be 14 nm and the average width is about 135 nm. The stripes in the 3D TI layer (see inset) correspond to quintuple layers of the material. | 2012.15118 | T-junction-SEM.jpg |
0.46465 | 18c009c8d0ed4523a5585e8f129162d2 | Steering ratio of the T-shaped tri-junction. (a),(b) The steering ratio () of the T-shaped tri-junction presented in Fig. <ref>b as a function of the in-plane magnetic field orientation θ (with the direction relative to the T-junction indicated by the red arrows at the top) and the field strength ||, as obtained from the transport measurements (see Fig. <ref>) in a, and from the qualitative transmission model, as discussed in the Main Text and derived in Sec. <ref> of the , in b. (c) Line cuts of the steering ratio for increasing magnetic field strengths, obtained from the experimental data (dashed lines) and model results (solid lines) presented in a and b, respectively. The scale indicated on the left (right) is for the experimental (model) values and the curves are shifted up by multiples of 0.5×10^2 (0.2). | 2012.15118 | fig-SR-maps-cuts_Arial.jpg |
0.418634 | 45743c218bac4fe581a6244226d17377 | The current across a Y-shaped tri-junction as a function of the in-plane magnetic field components from the bottom leg to (left column) both output legs, and to (middle column) the left and (right column) the right output leg individually, measured at different temperatures. | 2012.15118 | fig-current-maps-Y_Arial.jpg |
0.448568 | 830fdc314980422ea54971b1a0d8739b | Currents through the T-shaped tri-junction as a function of in-plane magnetic field components. (a)–(c) The current is measured from the bottom leg to both output legs (shown in a), and individually to the left leg (shown in b) and right leg (shown in c). The sketches on top of the plots indicate the current paths. The measurements were carried out at a temperature of 25 mK for the T-shaped tri-junction presented in Fig. <ref>b. The reference frame is also shown in that figure. | 2012.15118 | fig-current-maps_Arial.jpg |
0.517653 | b36f09121c2342f6871d2bc1e7844f93 | The standard deviation of the current through a Y-shaped tri-junction. The data is extracted from current data sets over different in-plane magnetic field strengths and orientations (see Sec. <ref> in for current data), evaluated at different temperatures. | 2012.15118 | fig-current-sigma-Y_Arial.jpg |
0.411535 | b656b2f2613646069ea25b28d37ae443 | A decision tree with the rules of transmission across a tri-junction, according to the transmission model for in-plane magnetic field-driven current steering. | 2012.15118 | fig-decision-tree.jpg |
0.372344 | 0d955c5335ef4b7f87a790700491d03f | The wave function density the single channel at 20 meV above the Dirac point energy in the three different legs of a T-junction, with the geometry of the legs, the 3D TI model Hamiltonian parameters, and the magnetic field strength the same as in Figs. <ref>a–b. The density in the different legs is presented in the three columns (for bottom, left, and right leg, respectively) for the four different diagonal in-plane magnetic field orientations in the different rows (indicated by the red arrow in the inset). The incoming mode is shown for the bottom leg, and the exit modes for the left and right legs. | 2012.15118 | fig-density-T-kwant_Arial.jpg |
0.429896 | bf0a23d3e67e43f7bbd1962b7749914a | (a) False color scanning electron micrograph of the Hall bar. The metallic leads are colored in blue, the top gate in violet, the substrate in dark grey, and the 3D TI is left in gray. (b) Longitudinal resistance R_xx in magnetotransport with magnetic field aligned perpendicular to the substrate plane. (c) Hall resistance R_xy along the outer contacts of the Hall bar. (d) Weak antilocalization conductance peak with the fitted HLN model. | 2012.15118 | fig-hall-data.jpg |
0.464561 | 16819245b30a40518325182945e255ac | Schematic diagram of the frequency-doubled Nd:YAG MOPA laser system. The spatial flat-top beam profile is maintained throughout the laser system using relay lenses (not shown) with the image planes indicated in green. VCO: Voltage-controlled oscillator. VVA: Voltage-variable attenuator. AOM: Acousto-optic modulator. OI: Optical isolator, PBS: Polarizing beam splitter. λ/4: Quarter-wave plate. λ/2: Half-wave plate. LBO: nonlinear crystal for second harmonic generation. | 2012.15119 | fig1.jpg |
0.429174 | a22c50577b864302bdb869a42b39d39a | Temporal pulse shapes and instantaneous power of the laser system for an unshaped cw seed laser (left column), and for seed laser pulses shaped to obtain rectangular second harmonic pulses with length 200 μs (middle column) and 60 μs (right column). From the top to bottom row: time profiles of the seed, the double-pass pre-amplifier output, the main-amplifier output, and the second harmonic (SH) pulses. The pulse energies are given in each panel. The blue lines in the second and third rows indicate the time profile of the current driving the pre-amplifier and main-amplifier pump diodes, respectively. The same drive currents are used for each pulse. Note the change in time scales between columns. | 2012.15119 | fig2.jpg |
0.382562 | 403706ce663b40fdb96963718e1c1f2a | Left: Beat-note signal arising from the superposition of the cw output of the master laser with the pulsed infrared light. The oscillation frequency of ∼80 MHz corresponds to the frequency of the accousto-optic modulator used to generate a seed pulse from the cw output. Right: Beat-note spectrum of the continuous wave seed (red) and amplified (black) laser pulse. The beat-note spectrum of a synthetic pulse is shown in green. | 2012.15119 | fig3.jpg |
0.537439 | 6219765b3a944acd90a263174976932e | Pulse energy of the fundamental frequency (1064 nm) and second harmonic (532 nm) versus time, over a 60 min interval. The insets show the mean temporal profile of the laser pulses and its 1σ range of variation (in grey) over 60 min. | 2012.15119 | fig4.jpg |
0.434648 | c9c083575f39404fa015b7bbaee5c284 | Beam profiles of the FF seed beam (left), amplified FF (middle) and SH (right). Units are in μm and all profiles are taken at relay planes. | 2012.15119 | fig5.jpg |
0.420337 | 29fef4e3b9824db8b39f88f55bf1270e | Illustration of the multi-step framework as a data analysis pipeline. The data and results are denoted in blue, and the statistical models are denoted in green. | 2012.15130 | pipeline-chart.jpg |
0.505179 | 6af0062fb66c40f78b8e74bc3d0d2cb8 | Semi-log plot of the temporal evolution the energy up to t=1000. The solid and dotted plots stand for the time evolution of the original energy functional and the SAV-introduced energy functional, respectively. The plots overlap so that differences are indistinguishable. | 2012.15133 | energy_t1K.jpg |
0.501524 | 2519bde442e1490890036e300c62c6c4 | Sketch of a crack propagating in a material along with notations used herein. (a) LEFM description of a straight crack front propagating at a speed v_0 in a homogeneous brittle solid. The crack front is being loaded in mode I (opening mode). The stress intensity factor is referred to as K. (b) Due to microstructural heterogeneities, the crack front distorts both in-plane and out-of-plane. The approach proposed here neglects out-of-plane roughness. It only considers the first-order in-plane perturbations of the crack front from the reference situation in panel (a) with a straight crack growing at constant speed due to overall loading K_0(). The in-plane projection of crack front is parametrized by F(z,t)= t +f(z,t) where f(z,t) refers to the in-plane perturbations. L is the system size. | 2012.15145 | Fig1-sketch.jpg |
0.427923 | a5140ea4a50347af87e4861caa140485 | Full line: FW normalized amplitude A^*_∞=A(t/λ_z>>1)/γ_0 as a function of crack speed, . Dash line recalls the variations of A^*(t=0=-1/C_v(v_0)) with [Fig. <ref>(a)].Here, ^0, and are normalized by c_S. Poisson ratio is set to ν=0.35, so that c_D=2.082 and c_R=0.935. All speeds are expressed in c_S unit | 2012.15145 | Fig10-Ass_vs_v0.jpg |
0.477455 | 0d2345f0b3064a46b9e5f2deb4fbe92e | Generic shape for the FW pulse generated when a dynamic crack propagates at along a toughness with a single strip of Mexican hat profile (γ^*(z^*)=(1-z^*^2)exp(-z^*^2/2)) or Hanning profile (γ^*(z^*)=cos(π z^*/2) for |z^*|<1/2, γ^*(z^*)=0 for |z^*|≥ 1/2) and different parameters γ_0 and ξ_z. The different data point symbols correspond to the velocity profiles along z at t_c=1700 obtained in the simulations achieved with the parameters provided in the legend, once rescaled via Eq. <ref>. Red solid line and red dashed line are the analytical predictions provided by Eq. <ref> for the Mexican-hat strip shape and the Hanning strip shape, respectively. | 2012.15145 | Fig14-norm_pulse.jpg |
0.572959 | 335f6524caba493cafeafac52c5e4678 | Closed contour along which exp(izt)/(z^2-k^2^2) is integrated. (a) Contour γ^+ used for the positive times; (b) Contour γ^- used for negative times. | 2012.15145 | Fig16-residu_figure.jpg |
0.431567 | 3b1fca1331614506843a9641a2422cfd | Speed of the crack front waves, , along the front as a function of reference crack speed . Vertical dashed line shows Rayleigh wave speed, c_R, which sets the maximum speed for a mode I crack. All speeds are expressed in c_S units. Poisson ratio is set to ν=0.35, so that c_D=2.082 c_S and c_R=0.935 c_S. | 2012.15145 | Fig2-cFW_vs_v0.jpg |
0.445447 | 765cb6a7aed240e498a628196c1227a7 | Inset: v(z,t) vs. t in a typical simulation at three distinct locations along the crack front: The green (solid oscillating curve), blue (dash-dotted oscillating curve), and black (dotted oscillating curve) correspond to z=100 , z=200, and z=400, respectively. Main panel: v(z,t)/sin(2π z/λ_z) vs. t at the same locations. Note the curve collapse, demonstrating the variable separation proposed in Eq. <ref> for z and t. Red (plain non-oscillating) curve shows the amplitude A(t) as obtained by interpolating between the positions and amplitudes of the successive optima. Here, =0.8 c_S, γ_0=0.1, N=1024 and λ_z=128. Poisson ratio is set to ν=0.35, so that c_D=2.082 c_S and c_R=0.935 c_S. | 2012.15145 | Fig5-vcollapsed_v08.jpg |
0.449486 | 97de1e0c05964f7dae7b525acbd915b2 | Inset: Time evolution of the amplitude, A(t) for =0.8 c_S and different sets of parameters {γ_0,N,λ_z }, the value of which are specified in the legend on the right-hand side of the main panel. Main panel: Collapse using the scaling form proposed in Eq. <ref>. Poisson ratio is set to ν=0.35, so that c_D=2.082 c_S and c_R=0.935 c_S. | 2012.15145 | Fig6-Aenv_v08.jpg |
0.392783 | d1376589f6124d7cac38b87b29b32ed4 | FW speed at initiation (solid line), ^0=c(t→ 0), as a function of crack speed, . Dashed line recalls the variations of c_FW with [Fig. <ref>]. Dash-dotted vertical line shows Rayleigh wave speed, c_R. All speeds are expressed in c_S units. Poisson ratio is set to ν=0.35, so that c_D=2.082 c_S and c_R=0.935 c_S. | 2012.15145 | Fig8-c0_vs_v0.jpg |
0.503861 | ff15642ebb274e349e38125ebb239a12 | Main panel: Comparison between v(z,t) as obtained by simulations (black circles) and v_∞(z,t) as predicted analytically by Eq. <ref> in the long-time limit regime (red line). Inset: zoom on a smaller time scale to emphasize the overlap. Here, =0.8, L=1024, γ_0=0.1 and λ_z=0.1. Poisson ratio is set to ν=0.35, so that c_D=2.082 and c_R=0.935. All speeds are expressed in c_S unit. | 2012.15145 | Fig9-asympt-vnorm_vs_t_v08.jpg |
0.465144 | 7a988d9965c5415985a25c54cef1c6af | Visualization of attention scores averaged over all heads and all layers. This case is selected from the SST-2 dev set. The red rectangle indicates higher scores on the right side but lower scores on the left side. | 2012.15150 | attention-crop.jpg |
0.488634 | 53a86427df1b4dbe9930bc444caf8672 | Gate values in different layers on SST-2 and FCE datasets. The blue polyline means that we incorporate the window-based local attention with global attention, and the red polyline corresponds to the syntax-aware local attention. | 2012.15150 | gate2-crop.jpg |
0.456675 | 862a06a75dbf478a8e4dc74d8845abfc | An overview of our model. | 2012.15150 | method-crop.jpg |
0.492673 | d78789d30c4d498588128d149ed6c731 | Performance of low-trigger and high-trigger policies. In each panel, we vary the number $̨ of daily contacts (x-axis) and consider the performance (cost) of the low-trigger policies (=, blue) and of the high-trigger policies (=, green), when the patience parameter is low (=days, left column) and high (=, right column). The dotted red line shows the number^̨⋆of daily contacts that corresponds to the epidemiologicalR_0equal to1(when no individuals have yet recovered). Generally speaking, it is beneficial to have the trigger valuelow (blue curves are below green ones), to impose severe rather than moderate lockdown (all curves go up), and to be patient (the curves in the right panels are lower). Forand, the key is to have the trigger valuelow. Forand, the key is to have the patiencehigh. | 2012.15155 | fig4-20.jpg |
0.457953 | 172e6de0a8b143b1a94d70f0a252c755 | Role of patience in and . The patience parameter d is key in determining the total health damage (total number of critical cases per 1000) and the total economic damage (total duration of the lockdown, in days). The two panels are incomparable (due to different lockdown severity). The low-trigger policies (blue) and the high-trigger policies (green). The patience increases along the arrow. Features: High patience always helps in . High patience helps in , unless high-trigger and moderate (then it does not matter). In the right panel, the diamond and the triangle marks coincide. The high-trigger policy with low patience (d=7, green disk) is off the chard: It would yield D≐ 270 days and ≐ 30 cases per 1000. | 2012.15155 | fig4_damages-ipe.jpg |
0.387034 | 50a55381531e485e8b0404e249427b33 | Activity centroids and a visualization of COVID-19 spreading in Wuhan. (A) An illustration of calculating the activity centroids of smartphone users based on their activity locations. (B to C) The cumulative distributions of the mean and variance of the distance between activity centroids of each smartphone user in 1-6 months and the average centroid. (D) Visualizing the spatial-temporal spreading of COVID-19 in Wuhan between January 22, 2020 and February 27, 2020. | 2012.15161 | figure1.jpg |
0.538983 | db82941b5d714e049b3a7c17f8ce1296 | The temporal spreading pattern of COVID-19. (A to F) The cumulative spreading centroid and spreading radius in Wuhan, Beijing and Urumqi, respectively. (G) Relation between mean travel distance of people in each city and corresponding COVID-19 growth rate. | 2012.15161 | figure2.jpg |
0.435265 | 27edcd1b6dd64b2ab8a2d4e3849fbec6 | The spatial spreading pattern of COVID-19 in Wuhan, Beijing and Urumqi. (A to C) A visualization of the number of confirmed cases in discretized grids in Wuhan, Beijing and Urumqi, respectively. (D to F) The spatial distributions (dots) as a function of distance from overall spreading centroid and the fitted regression lines for these distributions. | 2012.15161 | figure3.jpg |
0.517626 | 4d5f81188d684e619b9d98a3bd320152 | The spatial-temporal model for COVID-19 spreading pattern. (A to C) The fitting performance with respective to spatial dimension. (D) Relation between travel willingness η and Simpson divergence. (E and H) Relation between travel willingness η and daily reported confirmed cases. (G) Relation between travel willingness η and the growth rate of spreading radius. (F and I) Relation between travel willingness η and spreading radius. | 2012.15161 | figure4.jpg |
0.483659 | e817f542916540c29f3f933749d91839 | (Color online)(a) Schematic of the system under consideration: Three identical neutral atoms with internal states |g⟩ and |r⟩ (gr-type qubits) and constant RB energy V, which are subject to four external laser pulses. (b) Pictorial illustration of the effective Hamiltonian H_eff in Eq. (<ref>). | 2012.15169 | WtoGHZstate_Fig1.jpg |