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
|
|---|---|---|---|---|
0.488468 | 6935b672d84540458307e5bd2599ad0d | The coefficient λ_k^⋆ (see (<ref>)) as a function of the number of iterations k; it is a random variable, and the solid line is the mean value (the dashed lines are resp. the quantiles 0.25 and 0.75) over 1e3 independent paths. | 2012.14670 | CoeffOpt.jpg |
0.440415 | f1ba1bd3aa6f4473b757d1eeb22c3bc4 | For λ∈{0.25, 0.5, 0.75} and μ∈0.01, 0.9, evolution of the constant C solving (<ref>) | 2012.14670 | ConstantC1.jpg |
0.424716 | 2584d0dd770e4b428ffe8fded3454ce7 | For λ∈{0.25, 0.5, 0.75} and μ∈0.01, 0.9, evolution of the constant C solving (<ref>) | 2012.14670 | ConstantC2.jpg |
0.438692 | 0891d557e654488c8b3dabf26a728ff3 | Value of the control n^^-ℬ given by <ref> (Case 1, with a circle), <ref> (Case 2, with a cross) and Karimi et al. (no markers). The control is displayed as a function of μ∈0.01, 0.9 and for different values of : =n (solid line), = 1e2 n (dash-dot line) and = 1e3 n (dashed line). | 2012.14670 | Controls.jpg |
0.457327 | 39660bfc52f24e859d7079c00493a4da | Estimation of the p=20 eigenvalues of the covariance matrix Σ. The estimator is the value of the parameter obtained at the end of a single path of length 100 epochs. | 2012.14670 | EigenSingleRun.jpg |
0.382731 | ebfa525fd5d5449a95c51b3953c26bb2 | Evolution of the p=20 eigenvalues of the covariance matrix Σ along one path of length 100 epochs. All the paths start from the same value at time t=0. EM (top left), iEM (top right), Online EM (bottom left) and h-FIEM (bottom right). | 2012.14670 | EigenSingleRunPath.jpg |
0.455428 | 856fe1d89f3c42bd852187f12f56c522 | Estimation of the p=20 components of one of the expectation vector μ_ℓ. The estimator is the value of the parameter obtained at the end of a single path of length 100 epochs. | 2012.14670 | ExpectationSingleRun.jpg |
0.532131 | 3cce0962829e400dab87e0c56c1fd26c | Evolution of the p=20 components of one of the expectation vector μ_ℓ along one path of length 100 epochs. All the paths start from the same value at time t=0. EM (top left), iEM (top right), Online EM (bottom left) and h-FIEM (bottom right). | 2012.14670 | ExpectationSingleRunPath.jpg |
0.478838 | 20985d8b2b804facbb52c9ef8b060ba6 | Monte Carlo approximation (over 1e3 independent runs) of k ↦_k+1^-2[^k+1 - ^k^2 ] for Online EM, FIEM and opt-FIEM. | 2012.14670 | FieldH.jpg |
0.450595 | e36e2f10e32b4c2a8a1db313913e1c63 | Evolution of the normalized log-likelihood: average over 10 independent paths of length 100 epochs. The first 14 epochs are discarded. All the paths start from the same value at time t=0, with a normalized log-likelihood equal to -58.31. | 2012.14670 | LogLikelihoodMultipleRunEnd.jpg |
0.496268 | 082da9f51f6946e7ac0ce0879397cea6 | Evolution of the normalized log-likelihood: average over 10 independent paths of length 100 epochs. The first 25 epochs are displayed. All the paths start from the same value at time t=0, with a normalized log-likelihood equal to -58.31. | 2012.14670 | LogLikelihoodMultipleRunInit.jpg |
0.563404 | b336d9c89d7b4fa0a0660e5ef11be795 | Evolution of the normalized log-likelihood along one path of length 100 epochs: the first 14 epochs are discarded. All the paths start from the same value at time t=0, with a normalized log-likelihood equal to -58.31. | 2012.14670 | LogLikelihoodSingleRunEnd.jpg |
0.514842 | 88c55558682f4eddab2c867d4e6097df | Evolution of the normalized log-likelihood along one path of length 100 epochs: only the epochs 1 to 25 are displayed. All the paths start from the same value at time t=0, with a normalized log-likelihood equal to -58.31. | 2012.14670 | LogLikelihoodSingleRunInit.jpg |
0.429851 | 0c7bcf71a2f24ed9a4d308c85267e624 | For k ∈{1e2, 5e2, 1e3, 1.5e3, …, 6e3, 7e3, …, 20e3}, ratio of the expectations (Exp) [ ^k_opt-FIEM- _⋆] / [ ^k_Alg - _⋆] when Alg is FIEM (solid line with circle) and then Online EM (solid line with cross); and the standard deviations (std) std(^k_opt_FIEM- _⋆) / std(^k_Alg - _⋆) when Alg is FIEM (dashed line with circle) and then Online EM (dashed line with cross). The expectations and standard deviations are approximated by a Monte Carlo sum over 1e3 independent runs. | 2012.14670 | RatioErrorTheta.jpg |
0.417988 | a4029655bb3c48338475672655b8f363 | Value of the constant step size given by Karimi et al., <ref> (Case 1) and <ref> (Case 2). The step size is shown as a function of μ∈0.01, 0.9. In Case 2, different strategies for are considered. | 2012.14670 | StepSize.jpg |
0.416282 | f5a60985ad654cc48c97889a18eec436 | Estimation of the g=12 weights of the mixture model. The estimator is the value of the parameter obtained at the end of a single path of length 100 epochs. | 2012.14670 | WeightSingleRun.jpg |
0.640572 | bc25b4d42a274629aae383ec92e303b6 | Evolution of the g=12 weights along one path of length 100 epochs. All the paths start from the same value at time t=0. EM (top left), iEM (top right), Online EM (bottom left) and h-FIEM (bottom right). | 2012.14670 | WeightSingleRunPath.jpg |
0.470275 | d24bcff47fba4dbebdb5ecd60db62150 | The data distribution of WebFG 2020. | 2012.14672 | distribution.jpg |
0.463368 | 42c1d8b0033b4d6390cc21d3cefd14e9 | Exampled images of the dataset in the WebFG 2020 challenge. Each sub-figure contains nine images belonging to one sub-category. Note that, there exist data noises in these sub-categories. | 2012.14672 | examples.jpg |
0.413484 | 960b62fd1df64302b6954622891e9fcb | Irrelevant images (noisy data) in WebFG 2020. | 2012.14672 | irrelevant.jpg |
0.439544 | b54968db14c64dd98c217b176ca3e968 | Schematic illustration of a paste in a container. | 2012.14675 | 2Dmodel.jpg |
0.405089 | 499b749d31f742bb9b60eadf96173c90 | A as a function of N_c after the external oscillation. | 2012.14675 | A.jpg |
0.421333 | 84e2da2c531d4da0aea5bd7e1ce26dbe | α as a function of (X,Z) after the external oscillation with N_c = 1/2. | 2012.14675 | alpha_0.5.jpg |
0.462444 | 8e756a75c4df4d0b998d95851cca3a23 | (a) σ_xz^ (EP) against the shear strain γ for σ_ Y/μ=1.0 with different γ̇τ_0. (b) σ^ (EP)_xz against the shear strain rate γ̇ for σ_ Y/μ=1.0 with different γ̇τ_0 in the steady state. | 2012.14675 | s_ga_gad.jpg |
0.46734 | e8e760e085204d7682e8080a5510b98d | s_xx as a function of (X,Z) after the external oscillation with N_c = 2. | 2012.14675 | sigma_2.jpg |
0.567111 | 9639ee5aaf1e45fcb241d2edc4ac50c5 | PDFs of the vertical vorticity ω_z in the numerical simulations at time t= 24 T_0 at angular velocity Ω=1 (empty symbols) and Ω =2 (filled symbols). Symbols as in Figure <ref>. | 2012.14676 | fig10-eps-converted-to.jpg |
0.393899 | 88e63e66879d4df190470054543e5f63 | (color online) Schematic diagram of anti-symmetric Maxwell's demon-like operation supported by a metasurface comprising of dimer unit cells. The metasurface (a) forbids the penetration of photons with frequency f_1 in downward incidence while it (b) allows the penetration of photons with frequency f_1 in upward incidence. The same metasurface also (c) allows the penetration of photons with frequency f_2 in downward incidence while it (d) forbids the penetration of photons with frequency f_2 in upward incidence. | 2012.14677 | Fig1.jpg |
0.434314 | 361b1d0de81c4724b257fbca2c8cdc63 | (color online) (a) Schematic diagram of the metasurface consisting of gyromagnetic (blue) and isotropic dielectric (grey) cylinders, (b) Top view of the unit cell of the periodic system in (a). Periodic condition is applied in x-axis throughout all calculations. Plane wave illuminating from -y and +y direction are denoted as downward incidence in red arrow and upward incidence in blue arrow, respectively. (c) The numerical and analytical transmittance spectra, represented as solid line and unfilled circles respectively, under downward [top of (c)] and upward [bottom of (c)] incidences. The transmittance spectra of lossless system in both directions are shown as black dashed lines. The insets show the E_z field patterns of those cylinders at those transmittance extrema. (d) The transient electric field E_z at f_1 and f_2. The marked arrows indicate the direction of the incident waves. Material parameters are ϵ_m=15, H_0=500 Oe, and M_s=1750/(4π) G, and n_d=10+2.8i. Geometrical parameters are Λ = 50 mm, r = 1 mm and a = 12 mm. | 2012.14677 | Fig2.jpg |
0.473467 | ce8b57af99f7419aa3697873c828de57 | (color online) Comparison of analytical transmittance (T) spectra (unfilled circle solid line), analytical absorption (A) spectra (half filled circle solid line) from the Eq. (<ref>), and the magnitude of scattered field of dielectric cylinder (|s_0^d|) (green solid line) under (a) downward incidence and (b) upward incidence, denoted by the superscript - and +, respectively. The local extrema of the three spectra under two incidences are simultaneously located at f_1 and f_2, marked by the black dashed lines. | 2012.14677 | Fig3.jpg |
0.434408 | d78609d8dca7418d95f9b4589d414c17 | (color online) The effects of k, the imaginary part of the dielectric constant, on (a)numerical transmission difference ΔT and (b) numerical absorption difference ΔA. The white/black circle dots in (a) and (b) represent the peak/depth locations of difference of A in the Eq. (<ref>). (c) The vertical cross section of (b) at f_1 and f_2, showing the effect of k on ΔA at those frequencies. Excellent agreements are obtained between the closed-form (CF) in the Eq. (<ref>) and numerical results. | 2012.14677 | Fig4.jpg |
0.4017 | 127708f2c5cf4194964b8b1f8e647839 | (color online) (a) Schematic diagram of mode hybridization of gyromagnetic (blue) and dielectric (grey) cylinder array. The collective resonance mode of isolated gyromagnetic cylinder array (MD, denoted as blue arrow) and dielectric cylinder array (ED, denoted as grey arrow) are hybridized into absorptive dominant modes 𝐠^A (|ϱ^Y|≤ |ϱ^d|) and transmissive dominant modes 𝐠^T (|ϱ^Y|> |ϱ^d|) by merging two arrays. (b) The magnitudes of component corresponding to the dielectric cylinder |ϱ^d| in two eigenvectors 𝐠_1 (green lines) and 𝐠_2 (orange dashed lines). The crossing point of two lines is found at f_0=3.80 GHz. (c) The magnitudes of excitation coefficient, |η_1,2^±|, to the two eigenvectors 𝐠_1 and 𝐠_2 under upward (or downward) wave excitation. The subscripts 1, 2 of |η| refer to the excitation of the eigenmodes with corresponding indices. The superscripts -/+ denote the downward/upward incidences, respectively. | 2012.14677 | Fig5.jpg |
0.471221 | 20b79079a17247afb0c314afeb94c75a | Left: The structure of a residual network. Right: The residual block. | 2012.14678 | Structure2.jpg |
0.470255 | 79d17d8d7b7e4fa7ba19c150951c649c | Effects of the network depth (left) and width (right). | 2012.14678 | ablation.jpg |
0.430035 | dc196948827e4b05b1279b0148fad2f2 | Seven image pairs suffering from defocus spread effects. | 2012.14678 | dse.jpg |
0.414077 | 904ea40784c240489fe94a94bdfc325b | The difference between pixel-wise and patch-wise fusion. | 2012.14678 | motivation.jpg |
0.490248 | 0d5b796162604ffebbc071c33ad65207 | Quality v/s Latency plot of FReTNA, ReTA, Wait-k models for EnFr language pair with different stability constraints. | 2012.14681 | BLEU-AL_ef.jpg |
0.546146 | 36ffcca726b6450ba2707f276846df4e | Different iterations | 2012.14681 | de-en-iters.jpg |
0.461014 | 312bbdd693004c2ab8fc5a5f3b513284 | Different strides | 2012.14681 | de-en-strides.jpg |
0.478204 | 123593e92a1e4205ab137d34af76dabd | Positional bias versus Non-Positional bias. | 2012.14681 | pos_non-pos_add_enfr.jpg |
0.425044 | cd5170e2cdd246b7a7fa720aee6bd3f1 | Overview of the proposed FReTNA and illustrated using German-to-English example. | 2012.14681 | re-levT.jpg |
0.448313 | 6ca0d9838951490681300cba580a9cd6 | An easy instance with a large word overlap (colored in orange) between the premise and the hypothesis from the MNLI dataset. The classifiers in shallow layers of a dynamic early exiting model cannot predict correctly, while BERT-Complete <cit.>, a small BERT pre-trained from scratch with the same size can make a correct and confident prediction. | 2012.14682 | CascadeBERT-v2.jpg |
0.374864 | 14762efa94424910998143738ac6e1da | DIS (%, higher is better, see Eq. <ref> in Section <ref>) heatmap of different models on the development set of MNLI and SST-2. The DIS of internal off-ramps in the DeeBERT of shallow layers is lower than that of BERT-kl and BERT-Complete, which leads to more wrongly emitted instances. The exiting decisions in shallow layers of DeeBERT thus can be unreliable. | 2012.14682 | DIS_times_mnli_sst2.jpg |
0.473589 | 2fa66551e2d54d8e834d9fc7eac772c0 | Illustration of CascadeBERT. The input is progressively fed into the models in the cascade framework and the result is emitted once the prediction is confident enough. Dashed components denote dynamic forward propagation according to instance difficulty. | 2012.14682 | cBERT.jpg |
0.46939 | f9918f2669094384a97c584194f7e7e6 | Task performance on the validation set and speed-up ratio trade-off curve comparison of a 2-model cascade (orange square) and a 3-model cascade (blue circle) on SST-2 and MNLI-m. | 2012.14682 | more_cascades_v3.jpg |
0.438119 | 60bb50ba5a754468b7564c584db170c6 | Performance comparison utilizing different models with the same number of layers on MNLI-m and SST-2. Complete models capable of extracting semantic-level information clearly outperform models like DeeBERT which overlooks the high-level semantic features. | 2012.14682 | rep_analysis_times_kl_bar_dynamic.jpg |
0.435279 | 0a3aa359cf684c1595aa184bc41afa23 | Phase diagram of the (Pt_0.95Ir_0.05)/Fe/Pd(111) system. (a) Energy per spin in the spin spiral (SS), skyrmion lattice (SkL) and field-polarized (FP) states as a function of magnetic field at T=0 K. B_m and B_s denote the metastability field for an isolated skyrmion and the transition field from the spin spiral to the field-polarized state, respectively. (b) Static susceptibility as a function of temperature at B=0 T. T_c denotes the transition temperature from the SS to the paramagnetic (PM) phase. | 2012.14688 | Fig1-eps-converted-to.jpg |
0.481006 | 9ae54e0b9cb040ecb1f8837fe425d8ed | Time-averaged spin configurations for topological charges (a) Q=1 and (b) Q=0. (c) Time evolution of the topological charge for a sample run with external field strength B=1 T, T=80 K and number of spins N_S = 25 × 25. The spin configurations in (a) and (b) result from a time average for each spin's Cartesian coordinates over the indicated time intervals. | 2012.14688 | Fig2.jpg |
0.427954 | c65057eb435c42429610345f265d2872 | Average value of Q calculated with ensemble average using free-energy values of topological charges Q=-1, 0, 1 (green) and time average over all topological charges (blue). System simulated with N_S=25 × 25 spins and B = 1 T external field. | 2012.14688 | Fig5-eps-converted-to.jpg |
0.414833 | fff2cf2484d54081b83528aff2a9ade5 | The illustration of BMD challenges with an INbreast dataset example: Q_1 - data scarcity and Q_2 - data entanglement. Red stands for the 2D t-SNE <cit.> embedding of malignant masses and blue for that of benign lesions. The four images are corresponding mass examples. | 2012.14690 | Fig1_final.jpg |
0.470941 | 5370d06335284c2dbef63acb566127d5 | Augmented mass ROIs by Conditional GAN <cit.> (first row), and positive and negative neighbors by our proposed adversarial augmentation method in second and third row respectively. | 2012.14690 | Fig2_final.jpg |
0.377499 | f8f17f1590144673af84b52bb339b2e2 | The deep neural network architecture constructed in COIN to extract deep latent features. “DC block" represents a down-sampling convolutional block, “RC block" is a residual convolutional block, and “SConv" is separable convolutions. | 2012.14690 | architecture.jpg |
0.473082 | 81a4a94e0c114ecf9f52ba7fc3fd1ae2 | An illustration of the proposed COIN framework for BMD, which consists of three steps: adversarial augmentation, to build Signed graph, and the joint optimization. In the figure {x_1, x_2} are samples on benign manifold ℳ_b and {x_3, x_4} are on malignant manifold ℳ_m. In the first step (adversarial data augmentation), positive neighbors x_5 and x_7 are created with Eq. (<ref>) for benign and malignant manifold, separately; and negative neighbors x_6 and x_8 are generated with either Eq. (<ref>) for benign and malignant manifold, individually. After that, a signed graph is built upon both original and augmented samples as Eq. (<ref>). Finally, the joint loss as Eq. (<ref>) is optimized in the deep latent space, so that the margin between benign and malignant manifold are maximized. | 2012.14690 | coin_motivation.jpg |
0.404202 | 84ba08cf96524b53bd77ed12220a7de9 | Estimates of monthly state-level Manufacturing sector indices and Financial index (Fin1) mean growth rates (horizontal axis) and volatilities (vertical axis) with the Bayesian Panel Markov-switching model. The horizontal axis represents the mean growth rates and the vertical axis the volatilities. Sample period: February 1981 to December 2016 (month on month). | 2012.14693 | Parameters.jpg |
0.473541 | 203f7301832d4f8fbae0f559a4429023 | Geometry of the f-plane approximation at latitude ϕ with the full Coriolis vector. Here, x, y, and z are directed towards the east, the north, and upward, respectively. The dashed lines parallel to the axis of rotation in the y-z plane represent the characteristics of the thermal wind relation. | 2012.14694 | fig.jpg |
0.417239 | f6c747cc6f62426187f52c4dfbf31751 | Evolution of the mean inflaton field for simulations with μ = 5× 10^3 M_pl, λ=1.0 × 10^27, γ_max≈ 3.898 (red solid line), with μ = 5× 10^3 M_pl, λ=3×10^26, γ_max≈ 2.075 (orange dashed line) and canonical case γ = 1 (blue dotted line). | 2012.14697 | phi_evl_all.jpg |
0.472445 | cfb79151c8f94156a8eadce4e91369a9 | Evolution of the variance of the inflaton field for simulations with μ = 5× 10^3 M_pl, λ=1.0 × 10^27, γ_max≈ 3.898 (red solid line), with μ = 5× 10^3 M_pl, λ=3×10^26, γ_max≈ 2.075 (orange dashed line) and canonical case γ = 1 (blue dotted line). | 2012.14697 | phi_var_all.jpg |
0.439168 | 82f6647de9a542ecbf4da24e9db96504 | Four most similar images retrieved by six algorithms. OC: Object Count, GIN: IRSGS-GIN, GCN: IRSGS-GCN. The visual genome ids for the query images are 2323522 and 2316427. | 2012.14700 | qual_all.jpg |
0.4418 | 4442282e26ef41cfbcd6c73f5178587c | Four most similar images retrieved by six algorithms. OC: Object Count, GIN: IRSGS-GIN, GCN: IRSGS-GCN. The visual genome ids for the query images are 2323522 and 2316427. | 2012.14700 | qual_all_2.jpg |
0.50199 | 629a3b8836864f51afd09f8eb9902d6b | Image retrieval examples from ResNet and IRSGS. ResNet retrieves images with superficial similarity, e.g., grayscale or vertical lines, while IRSGS successfully returns images with correct context, such as playing tennis or skateboarding. | 2012.14700 | smir_fig1.jpg |
0.441937 | d7e53cac8a0e4b52bc367b8231a8e4a1 | An overview of IRSGS. Images ℐ_1, ℐ_2 are converted into vector representations ϕ(𝒮_1), ϕ(𝒮_2) through scene graph generation (SGG) and graph embedding. The graph embedding function is learned to minimize mean squared error to surrogate relevance, i.e., the similarity between captions. The bold red bidirectional arrows indicate trainable parts. For retrieval, the learned scene graph similarity function is used to rank relevant images. | 2012.14700 | smir_fig2.jpg |
0.501784 | eee3fba7961c43b889c9b62ec33b4616 | Sparsity pattern of a small FCI matrix, here for H_2O in the minimal basis set sto-3g. | 2012.14702 | H20_sto-3g.jpg |
0.414414 | 88e79389064842339f69132bd9d3f972 | Convergence of IPT in the two-dimensional example <ref>. Left: In the complex -plane, RS perturbation theory (RS-PT) converges inside a circle of radius 1/2 (orange line) bounded by the exceptional points ± i/2 where eigenvalues have branch-point singularities and M is not diagonalizable. Dynamical perturbation theory (IPT) converges inside the domain bounded by the blue cardioid, which is larger—especially along the real axis, where there is no singularity. Outside this domain, the map can converge to a periodic cycle, be chaotic or diverge to infinity, following flip bifurcations (along the real axis) and fold bifurcations (at the singularities). The domain where the map remains bounded (black area) is a conformal transformation of the Mandelbrot set. Right: The bifurcation diagram for the quadratic map f along the real -axis illustrates the period-doubling route to chaos as increases away from 0 (in absolute value). Orange and left vertical lines indicate the boundary of the convergence domains of RS-PT and IPT respectively. | 2012.14702 | fig1.jpg |
0.506793 | 189a04e6796d497eb0051953b7bfd9cf | Timing of full spectrum routines on the GPU for dense perturbative matrices. Here IPT proves to be faster than both Divide and Conquer (SYEVD) and Jacobi (SYEVJ) when is sufficiently small. | 2012.14702 | fig2.jpg |
0.488665 | 5ba2d882ad5e4baeaebfaaa8ee353322 | IPT vs. LAPACK for the complete diagonalization of perturbative matrices with increasing , in double precision. The largest speed-ups are obtained for the non-symmetric problem. | 2012.14702 | fig3.jpg |
0.401701 | 384c537728824282aafcc2c11855d791 | IPT vs. standard iterative algorithms for the lowest-lying eigenpair of perturbative matrices of the form <ref> and increasing . Davidson-type methods with diagonal preconditioning outperform the Krylov-Schur algorithm on these problems, and IPT is yet faster when the perturbation is sufficiently small (or sparse). | 2012.14702 | fig4.jpg |
0.440937 | 49690e08685a44d39863cde058c1d7b5 | IPT as a refinement algorithm for ill-conditioned problems <ref> with increasingly small eigengap. Combining single-precision GEEV with IPT gives eigenvectors with acceptable accuracy at a fraction of the computational cost of double-precision GEEV. | 2012.14702 | fig7.jpg |
0.438816 | f8b8966d2e724d57aee2e567f038eb31 | The convergence domain on the -plane for the first column of (the first eigenvector 𝐳_1) for the 3×3 example. The Mandelbrot-like set (domain where orbits remain bounded) of the iterative scheme is shown in black and grey. The empirical convergence domain is shown in black. Its largest component corresponds to the stability of a steady state (the applicability domain of the iterative method). Small components correspond to stability of various periodic orbits. Various shades of grey show the values of that lead to divergence to infinity (the darker the slower the divergence). In red are the values of where the matrix is non-diagonalizable. | 2012.14702 | figSIn1.jpg |
0.408795 | 492a60030f904e6a89ed9a696d448a5e | Same as <ref> for the second eigenvector 𝐳_2 (the second column of ). Small components correspond to stability of various periodic orbits. Various shades of grey show the values of that lead to divergence to infinity (the darker the slower the divergence). | 2012.14702 | figSIn2.jpg |
0.473054 | ee9fd76dcc954c92933e8a0f49ec66cb | Same as <ref> for the third eigenvector 𝐳_3 (the third column of ) . | 2012.14702 | figSIn3.jpg |
0.424283 | 6eeec37b78a34badb2ef20260fe42d5c | A schematic sketch of the geometrical relations between the original space ℂ^n, the projective space ℂP^n-1, its affine charts, and points in these different spaces that correspond to certain eigenvectors. The projective space ℂP^n-1 (in red) can be associated with the unit sphere in ℂ^n, in which each circle carved out in it by a complex line passing through the origin (a special case of a real plane) is glued to form a single point (here exemplified by two abstract points p and q, the circles are degenerated to pairs of points in real settings). The standard basis {𝐞_k} of ℂ^n (eigenvectors of D) define affine charts U_k of ℂP^n-1 (only two are depicted: for k=i in blue and k=j in green). U_i and U_j are shown as affine subspaces of ℂ^n. Note that 𝐞_j is at infinity of U_i and 𝐞_i is at infinity of U_j. An actual eigenvector and all its nonzero rescalings correspond to a single point in ℂP^n-1 and in each U_k. Two eigenvectors 𝐳_i and 𝐳_j are depicted with a rescaling choice such that they are equal to the i-th and j-th columns of matrix . | 2012.14702 | sketch.jpg |
0.492431 | 7d84f824b14c45419362ad39a5650fa9 | Comparison of CsPbI3 band structure calculated in DFT and in empirical tight-binding method. The DFT calculations are shown in green dashed lines, tight-binding results are in thin black lines. | 2012.14705 | CsPbI3.jpg |
0.456601 | d09c5d2db63e42c192d8f90ad31ce55a | The band structure calculated using the ETB parameters fit to DFT results and corrected for the experimental data. Grey dotted line shows the band structure calculated using parameters from Ref. <cit.>. | 2012.14705 | CsPbI3_expt_corr.jpg |
0.515546 | 9d111f7f79ac4ec79573ad3bbae0e5a2 | Comparison of CsPbI3 band structure calculated in DFT and in empirical tight-binding method using sp^3 model with parameters from Table <ref>. Green dashed lines show the DFT results, black solid lines ETB results. | 2012.14705 | sp3.jpg |
0.526148 | 67ddc64c8ea14d4a876a6a81d0421a29 | Schematic of a curved-wall boundary along one single lattice direction. The thin solid line is the grid line, and the thick curved one represents the boundary surface. White circles (∘): the fluid nodes; Black circle (∙): the intersection point of the boundary with the grid line; Square box (□): the solid node outside the computational domain. | 2012.14707 | Fig1.jpg |
0.433952 | 1fde7bcc82cd43aa9bbb484d82dae6d7 | Schematic illustration of the microcylindrical Couette flow. | 2012.14707 | Fig17.jpg |
0.511739 | c684a7dbb0d04b53b1312663529bb4bb | Schematic of the flow and lattice arrangement with a distance ratio γ. The wall boundary corresponding to the halfway DBB boundary condition is placed with γ=1/2. | 2012.14707 | Fig2.jpg |
0.386542 | acea684d326d4a4f8cae4c7ababb0930 | Schematic of the Poiseuille flow in a microscale channel with aligned straight walls. | 2012.14707 | Fig3.jpg |
0.464548 | 37e61efbdbc4412197b57d387f6f73cf | Schematic illustration of the Poiseuille flow in an inclined microchannel. | 2012.14707 | Fig8.jpg |
0.450347 | 4988e72690d247afbf3e71bdefea8304 | Cumulated MSPE for the Benchmark procedure (Bench), the Sieve-PCA with locally stationary prediction (SP) and the AR-fit prediction (SAP), the Local-PCA when L=100 with the locally stationary prediction (PCA100) and the AR-fit prediction (PCAR100), Local-PCA when L=200 with the locally stationary prediction (PCA200) and the AR-fit prediction (PCAR200) | 2012.14708 | cumerror-Dec04.jpg |
0.568307 | cbc4a8e2006044428679a146d09dc9ff | Dimension of loading matrix | 2012.14708 | eigen-dim-Dec04.jpg |
0.495439 | 5411f69cc00545708ceae358eda1d521 | Variance explained | 2012.14708 | explain-var-Dec04.jpg |
0.467194 | c0123897c4244c52a86baf8799729559 | Use of adjacency matrix to transform original self-attention, left-hand complete graph, into structure-induced self-attention, right-hand graph which looks clear-cut. Note that we omit self-circles for concision. | 2012.14710 | adj.jpg |
0.493376 | 168bf032996242e9aabab3d96e6684dd | Overall architecture of Structure-induced Transformer (SiT). | 2012.14710 | arch.jpg |
0.442391 | 26cd0c08007547c093bee24793013da9 | A Python code sample of multi-view graph used in Si-SAN. The code snippet is referred from <cit.>, which is original in Java. | 2012.14710 | ast.jpg |
0.432377 | 8f992f73adde4b619dc3c7918a6fa6d1 | Bolometric luminosity of the fiducial model. The isotropic bolometric luminosity for various viewing angles is also shown. The dashed curve and the data points denote the total (thermalized) heating rate and the estimated bolometric luminosity of AT2017gfo taken from <cit.> assuming the distance of 40 Mpc, respectively. | 2012.14711 | Lbol_angle_fid.jpg |
0.485102 | e3cc531c8fb44407bbb297762e6fd5e6 | Specific heating rate of the entire ejecta and the early/late-time ejecta components for t=0.3–30 days. The thermalization efficiency is taken into account by employing an analytic formula of <cit.>. | 2012.14711 | heat_elf.jpg |
0.447119 | 1772e7b35c8d458fb5f5a885555a894e | Total mass in the computational domain as a function of time. The purple and green curves denote the results for the HD simulations in which the radioactive heating is turned on and off, respectively. The orange dashed curve denotes the total mass of the material which is in the unbound trajectory. The blue dotted curve denotes the total input mass determined by the outflow data of the NR simulation (u_t<-1). The black dotted vertical line denotes the time at which the material injection from the inner boundary is truncated. | 2012.14711 | meje.jpg |
0.374997 | 71da0cdccbdb4efd8883304b38ae4679 | Comparison of the expansion opacity <cit.> of bound-bound transitions between the fiducial model (the blue curve) and the model calculated by omitting the contribution of neutral atoms (the green curve). The opacity is calculated by employing the element abundances in (x/ct,z/ct)=(0.001, 0.4) and by assuming 6× 10^-17 g/cm^3 and 2300K, which correspond to the condition at t= 2 days. | 2012.14711 | opc_ionmod.jpg |
0.405889 | f4e46e130e764e0ba1949097d047c115 | Comparison of v_ ave, v_ int, <|Δ v^(r)|>, <|v^(θ)|>, and <|v^(φ)|> (see the main text for their definitions). The black dotted vertical line denotes the time at which the material injection from the inner boundary is truncated. | 2012.14711 | vcomp.jpg |
0.448171 | ab87735f70d84624a723186ba10d19c1 | Illustration of pairs of intermediation attachments. Black bold and blue dashed lines denote added edges and existing paths in the network. Orange line denote paths of intermediations from randomly selected nodes. | 2012.14712 | 1.jpg |
0.528724 | 1921df581095413ca9abfee3c3a2dd48 | Illustration of our proposed methods. (a) BP Non-Preserving, (b) BP Preserving. The nodes with larger q^0_i and q^0_j and the edges between them are filled with red, while the nodes with smaller q^0_k and q^0_l and the edges between them are filled with blue. | 2012.14712 | 2.jpg |
0.446527 | d9ceb1fca8b54b3bb63335342efbfdb9 | The robustness index, the approximate size of FVS, and the degree-degree correlations vs. the number of rewiring. (Left: a, c, e) Rewirings in Preserving. Violet, green, and light blue solid lines denote the result by Degree, BP, and SP Preserving, respectively. The red dot line indicates a baseline of the conventional best. (Right: b, d, f) Rewirings in Non-Preserving. Violet, green, and light blue solid lines denote the result by Degree, BP, and SP Non-Preserving, respectively. | 2012.14712 | 3.jpg |
0.487112 | 488026e37e5c48f494e413e6c5a8a4f8 | Change of the degree distributions by the rewirings in Non-Preserving. (a) Degree distributions in original and after rewiring networks, (b) Maximum and minimum degrees vs. the number of rewiring in Degree, BP, and SP Non-Preserving for OpenFlights. The above three lines show the maximum degrees. The below three lines show the minimum degrees. Violet, green, and light blue denote Degree, BP, and SP Non-Preserving. Orange denotes the original degree distribution. | 2012.14712 | 4.jpg |
0.475762 | f119f36efafa4c7c98e1ab4622221130 | GRQC <cit.>. Comparison of the robustness index R_hub, the size approximate of FVS, and the degree-degree correlation coefficient r vs. the number of rewiring. (Left: a, c, e) Rewirings in Preserving. Violet, green, and light blue solid lines denote the result by Degree, BP, and SP Preserving, respectively. The red dot line indicates a baseline of the conventional best. (Right: b, d, f) Rewirings in Non-Preserving. Violet, green, and light blue solid lines denote the result by Degree, BP, and SP Non-Preserving, respectively. | 2012.14712 | GRQC_multi.jpg |
0.41223 | 64b4b7f88d99461d994d763917ad67f6 | GRQC <cit.>. (a) Degree distributions in original and after rewiring networks, (b) Maximum and minimum degrees vs. the number of rewiring in Degree, BP, and SP Non-Preserving. The above three lines show the maximum degrees. The below three lines show the minimum degrees. Violet, green, and light blue denote Degree, BP, SP Non-Preserving. Orange denotes the original degree distribution. | 2012.14712 | GRQC_multi_deg.jpg |
0.464154 | df16058d05ee42dbb2d107eb49289f7e | UCIrvine <cit.>. Comparison of the robustness index R_hub, the size approximate of FVS, and the degree-degree correlation coefficient r vs. the number of rewiring. (Left: a, c, e) Rewirings in Preserving. Violet, green, and light blue solid lines denote the result by Degree, BP, and SP Preserving, respectively. The red dot line indicates a baseline of the conventional best. (Right: b, d, f) Rewirings in Non-Preserving. Violet, green, and light blue solid lines denote the result by Degree, BP, and SP Non-Preserving, respectively. | 2012.14712 | UCIrvine_multi.jpg |
0.464237 | 1062d8b972f64936a9146e16cd8f9741 | UCIrvine <cit.>. (a) Degree distributions in original and after rewiring networks, (b) Maximum and minimum degrees vs. the number of rewiring in Degree, BP, and SP Non-Preserving. The above three lines show the maximum degrees. The below three lines show the minimum degrees. Violet, green, and light blue denote Degree, BP, SP Non-Preserving. Orange denotes the original degree distribution. | 2012.14712 | UCIrvine_multi_deg.jpg |
0.466097 | f675281b85134c149dccc5d4e8e13552 | AirTraffic <cit.>. Comparison of the robustness index R_hub, the size approximate of FVS, and the degree-degree correlation coefficient r vs. the number of rewiring. (Left: a, c, e) Rewirings in Preserving. Violet, green, and light blue solid lines denote the result by Degree, BP, and SP Preserving, respectively. The red dot line indicates a baseline of the conventional best. (Right: b, d, f) Rewirings in Non-Preserving. Violet, green, and light blue solid lines denote the result by Degree, BP, and SP Non-Preserving, respectively. | 2012.14712 | airtraffic_multi.jpg |
0.404261 | b444ee1c0d0f4dfbb259486c4359b4c9 | AirTraffic <cit.>. (a) Degree distributions in original and after rewiring networks, (b) Maximum and minimum degrees vs. the number of rewiring in Degree, BP, and SP Non-Preserving. The above three lines show the maximum degrees. The below three lines show the minimum degrees. Violet, green, and light blue denote Degree, BP, SP Non-Preserving. Orange denotes the original degree distribution. | 2012.14712 | airtraffic_multi_deg.jpg |
0.480358 | 50c85022d3d047d987b2717cebcc3715 | E-mail <cit.>. Comparison of the robustness index R_hub, the size approximate of FVS, and the degree-degree correlation coefficient r vs. the number of rewiring. (Left: a, c, e) Rewirings in Preserving. Violet, green, and light blue solid lines denote the result by Degree, BP, and SP Preserving, respectively. The red dot line indicates a baseline of the conventional best. (Right: b, d, f) Rewirings in Non-Preserving. Violet, green, and light blue solid lines denote the result by Degree, BP, and SP Non-Preserving, respectively. | 2012.14712 | email_multi.jpg |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.