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
|
|---|---|---|---|---|
0.446792 | 969c7fc433a84473a05b1f641155e528 | 2012.14371 | conc1da.jpg |
|
0.391641 | 4abfb6a432eb4a6085e4142215599044 | 2012.14371 | conc1db.jpg |
|
0.448125 | b052bc0e56214ec18e5e51eb174bc04e | Spectral detectors | 2012.14371 | det1.jpg |
0.525158 | c87ae295d932414c88f891ef0c6f5c6a | Initial spectral dist. | 2012.14371 | dist1.jpg |
0.464218 | 7276ef7e13af441cbc07b57972bb1e18 | Pushforward (MaxExp) | 2012.14371 | dist2.jpg |
0.430834 | 46362443bd2e46bba9ff2ef7f357ba88 | Pushforward (Gamma) | 2012.14371 | dist3.jpg |
0.40517 | 95befea7b1294f548611ef69c0fa4db3 | Illustration of third-order statistics in SCK. Linearized components (x), (y), (z) and time (t) denoted as (∘, □, ▿, +) are captured in third-order tensor , triplets (∘□▿) and (∘□+). This exposes SVM to rich body-joint statistics. | 2012.14371 | ker1b.jpg |
0.411322 | 2a338441fda74e7e80941768aa65bb8c | 2012.14371 | ker1b_new.jpg |
|
0.442483 | b455788b63f14a70ace5eb30e4be5970 | 2012.14371 | ker2a.jpg |
|
0.534636 | e18a676a3f2a408690e75d533d63a033 | 2012.14371 | ker2b2.jpg |
|
0.402534 | 93da767a5530479fa66886766b2c140b | 2012.14371 | ker2c_new.jpg |
|
0.475984 | 3ab5738dc86a429eb77f36444f3b3f5c | Order r statistics from Eq. (<ref>) can be understood by studying the linearization in Eq. (<ref>). For a given joint i at time s/N (normalized frame number), we embed a 3D joint coordinate _is (all centered hip) via function (·) into a non-linear Hilbert space representing an RBF kernel according to Eq. (<ref>). Similarly, we embed the time s/N via function (·) (also by Eq. (<ref>)). Finally, ⊗_r performs the third-order outer-product on concatenated embeddings aggregated next over frames s (note ∑_s). The interpretation: the Gaussians `soft-divide' the the Cartesian coordinate system along x, y, z direction, resp., and time s/N. Thus, triplets (x,y,z), (x,y,s/N), (x,z,s/N) and (y,z,s/N) assigned into such a `soft-divided' space capture locally three-way occurrences. They factor out one spatial (or time) variable at a time (note invariance to such a variable). | 2012.14371 | physics1.jpg |
0.483118 | 99d25d9f49ab46958d1a2ff881e781d3 | Third-order statistics from Eq. (<ref>) can be understood by studying the linearization in Eq. (<ref>). For a given pair of joints i≤i' at times s/N and s'/N (normalized frame numbers), we embed displacement vectors _is-_i's' of 3D joint coordinates _isand _i's' via function (·) into a non-linear Hilbert space representing an RBF kernel according to Eq. (<ref>). Similarly, we embed the starting and ending times s/N and s'/N via function (·) (also by Eq. (<ref>)). Finally, ⊗ performs the third-order outer-product on concatenated displacement and time embeddings aggregated next over frames s and s' (note ∑_ss'). The interpretation: the Gaussians `soft-divide' the Cartesian coordinate system along x, y, z direction, resp., as well as time direction (s/N and s'/N). We project displacements along x, y, z directions of Cartesian coordinates and assign each projection to Gaussians. Thus, triplets ([x;y;z],s,s') assigned into such a `soft-divided' space capture locally displacements of pairs of joints on the time grid (3-way soft-histogram). For DCK⊕ in Section <ref> we use velocity vectors _is-_i's'/max(1,|s'-s|) (displacement vectors) with short- and long-term estimates depending on s'-s (3-way soft-histogram of short- and long-term speeds). | 2012.14371 | physics2.jpg |
0.468941 | f3e466c48af1426b9decaf860477c1d0 | 2012.14371 | piv1a.jpg |
|
0.523317 | 4578e98df9054ee9bac00026098af2c4 | 2012.14371 | piv2.jpg |
|
0.445284 | f28d25e549ef4429995883406e00b541 | 2012.14371 | piv3.jpg |
|
0.427017 | 4082ec97822d458f9acd74a78de5895f | 2012.14371 | piv4c.jpg |
|
0.487084 | d0008a7c9fb5462191813336189adb57 | 2012.14371 | piv5c.jpg |
|
0.468014 | af151d5e55dd425795b5d71f4f6d8da1 | 2012.14371 | stickman5.jpg |
|
0.492915 | fcbc5756275e46738baa5d045dd4b519 | 2012.14371 | tens-basics.jpg |
|
0.472318 | bb18c705ed334a32a48e9c91bb205409 | 2012.14371 | tens-outer.jpg |
|
0.452881 | 7f045d361f684ba787372d89a7e4fee2 | X-ray surface brightness versus spectral index of the radio halo emission. The Bayesian linear regression is performed for data points with at least a 3σ radio noise level. The red and green dashed lines correspond to the best-fit obtained separately for the H1 and H2 region of the halo. The anti-correlation between these two quantities is evident, indicating that the spectral index is flatter at high X-ray brightness and tends to get steeper in low X-ray brightness regions. | 2012.14373 | IR_index.jpg |
0.438227 | 1951831f590842b4ade4db982023fae9 | I_ R-I_ X relation of the halo in MACS J0717.5+3745, extracted in square boxes with of 8 (about 50 kpc). Linmix best-fit relations are indicated by red dashed lines. The upper limits represent cells with data points below 2σ radio noise level. The horizontal black dash-dotted lines indicate the 1σ in the radio maps. The best-fit parameters are obtained for the halo emission (blue dots). The cyan lines show samples from the posterior distribution. The lower panel shows the residuals of logI_ R and logI_ X with respect to the Linmix best fit line. For the halo, the radio brightness correlates well with the X-ray at all three frequencies. The best-fitting slopes at 144 MHz, 1.5 GHz and 3 GHz are 0.67±0.05, 0.81±0.09, and 0.98±0.09, respectively. The strong dependence of the slope on frequency suggests that the halo shows a spectral steepening towards high frequencies. Displayed errors in the radio flux density measurements include the statistical and systematic uncertainties. | 2012.14373 | I_radio_Xray.jpg |
0.411826 | f93455fcc09b4ede82b57041aa4c28bc | Comparison of the halo from 144 MHz to 5.5 GHz. All images are at created at a common 8 resolution. The top-middle panel image (111-182 MHz) is adopted from <cit.>. The colors in all images were scaled manually. It is evident that the halo emission is more extended towards low frequencies. Contour levels are drawn at [1,2,4,8,…] × 3.0 σ_ rms, where σ_rms, 144 MHz=82 Jy beam ^-1, σ_rms, 147 MHz=130 Jy beam ^-1, σ_rms, 575 MHz=23 Jy beam ^-1, σ_rms, 1.5 GHz=7 Jy beam ^-1, σ_rms, 3.4 GHz=3 Jy beam ^-1, and σ_rms, 5.5 GHz=3.2 Jy beam ^-1 The beam sizes are indicated in the bottom left corner of the each image. | 2012.14373 | comparison.jpg |
0.448406 | fd5cd9936f5e43d48ecc85d932f9a64a | Spectral index profiles across the radio halo from 144 MHz and 5.5 GHz. The spectral indices were extracted in boxes with a width of 8, corresponding to a physical size of about 50 kpc. There is indeed a clear radial steepening in the spectral index value of the radio halo for both the H1 and H2 region of the halo. | 2012.14373 | halo_profile.jpg |
0.447987 | 2f10a802c91e464f8622a10b3257edcf | Left: Integrated spectrum of the halo between 144 MHz and 5.5 GHz. Dashed lines are fitted (between 144 MHz and 1.5 GHz) straight power-laws. The overall spectrum of the halo cannot be described by a single power-law spectrum. The halo spectrum is curved and shows a high frequency spectral steepening above 1.5 GHz. Black and red points show the integrated spectrum of the halo obtained by including and excluding polarized patches, respectively, detected in the halo region. The polarized regions are shown with cyan boxes in the right panel of Fig. <ref>. Right VLA 1.5 GHz image depicting the blue region where the integrated flux densities of the halo was measured (excluding source A). The flux densities are extracted from 15 resolution images, created using uniform weighting with a uv-cut at 0.2 k. The red boxes have a width of 8 used for extracting the radial spectral index profile across the halo. | 2012.14373 | halo_spectrum.jpg |
0.465571 | 2140f1dd22ab4aceaf205b62791a9d13 | A completely time-reversible oscillation of the control parameter (horizontal axis, here: swelling ratio) gives rise to non-time-reversible shape cycle of an elastic disk because of hysteresis of the triggered shape transition. The deformation cycle between shapes A and C resembles the opening and closing of a scallop but is hysteretic (compressed shape B and the transition state between B and C are only visited upon swelling). | 2012.14381 | Fig1.jpg |
0.476197 | aef18133a8ce4abe8d673707549c59ad | z-coordinate z_CoM of the disk's center of mass as a function of time for ten full deformation cycles. The simulated disk featured the same parameters as the disk in Fig. <ref> but with n_B= 60 and ε = 5× 10^-8 k_outR_out^2. The spheres in the mesh had a size of a=2π R_out/(10n_B). In each swelling cycle, the swelling factor α varied between 1 and 1.1 in 200 steps. The lower graph shows a zoom into a single (the first) cycle. The net swimming motion is in the direction of the opening. | 2012.14381 | Fig10.jpg |
0.392805 | 5fe4d7fb6067459cbba6b8e8ae5e5029 | Net swimming distance |Δ z_CoM/R_out| per cycle as a function of the shape height Δ z/R_out both for disks and 9-bead swimmer. For the 9-bead swimmer the parameter Δ z is directly prescribed. For disks the increasing shape height Δ z in the snapped state is realized by increasing the thickness in the range h/R_out=0.02-0.14, the other parameters are as in Fig. <ref>. The fit is Δ z_CoM/R_out = 0.022(Δ z/R_out)^3. | 2012.14381 | Fig11.jpg |
0.458969 | f87fe067d5794fb88eda8cf141211c1d | Illustration of the cyclic deformation in the 9-bead model. First row: corresponding states of the disk in the fixed frame. Second row: three-dimensional sketch of the 9-bead swimmer sates. Third row: two-dimensional side view on the 9-bead model. Each column represents one deformation state (1-2-3-1). | 2012.14381 | Fig12.jpg |
0.50134 | 576e9b16f62e47fda6b88a91f0f5420e | z-coordinate of the 9-bead swimmer's center of mass as a function of time during a single deformation cycle. The bead size is a = 0.1 R_out and Δ z = 0.5 R_out. The sketches illustrate the conformations of the swimmer during the deformation cycle. | 2012.14381 | Fig13.jpg |
0.42184 | 5ea4d9b467684f2ab26cf6daaa7aaa6f | Velocity field u⃗(r⃗) in the (y=0) -plane for each deformation phase. The left side shows the disk with a fixed frame from Fig. <ref> and a discretization of n_B=40. The right side shows a 9-bead swimmer with a=0.1 R_out and Δ z = 0.5 R_out. Colors indicate the absolute |u⃗| in units of k_out/η for the disk swimmer and in units of R_out/T_sw for the 9-bead swimmer. The arrows show the direction of the fluid velocity vectors. | 2012.14381 | Fig14.jpg |
0.486994 | de31b3c761d0490585b7fe3147b5b5be | Schematic demonstration of the swelling pattern. The disk is divided into two parts. The outer annulus (blue) swells with a constant factor α, the dinner disk (red) with 1/α. (a): With α > 1, the inner disk expands and the annulus shrinks. (b): Shrinking in the inner region and material expansion in the annulus with α < 1. | 2012.14381 | Fig2.jpg |
0.462053 | 93b16327743241e9914e35f24e7679da | Illustration of the direct neighborhood of a vertex at position r⃗_i. The area A_i that is associated with this vertex is shown cyan. The angle between the springs to the neighbor vertices at r⃗_j and r⃗_j+1 is called θ_j and is used in the calculation of the Gaussian curvature. | 2012.14381 | Fig3.jpg |
0.473935 | f34eb4aef6e046e5abbfd54beb5d6766 | Example Delaunay triangulation with n_B = 40 vertices on the boundary representing the spring mesh. Left side: flat disk with α = 1. Right side: resulting elliptic shape for α > 1 (swelling of interior disk) and hyperbolic shape for α < 1 (shrinking of interior disk). Swollen springs are shown in red, while shrunk springs are shown in blue. | 2012.14381 | Fig4.jpg |
0.392716 | 6b60ca079e7f4b4da80d5243a834079e | Energies (a), shape's height Δ z and negative lowest Hessian eigenvalue -λ_min (b) in the spring mesh model as functions of the stretch factor α. Circles denote numerical values calculated with decreasing α, while crosses are related to increasing α. The disk is always flat if it is located in the white area and always curved in the cyan regions. The blue shapes illustrate the corresponding conformations of the disk. The red areas mark the regions of pseudo-hysteretic effects. Arrows illustrate the directions inside the hysteresis loops. The simulated disk had a mesh with n_B=120 boundary vertices, R_in/R_out = 0.5, γ_FvK = 600, k_in/k_out = 0.24 and maximum fluctuations of δ_max= 5× 10^-4R_out. | 2012.14381 | Fig5.jpg |
0.414352 | 39190478f0d745df9dbf5b166c1865dd | Critical swelling factor α_c,e as a function of γ_FvK for R_in/R_out = 0.5 and different values of k_in/k_out. The solid black line represents the theory curve given by eq. (<ref>). | 2012.14381 | Fig6.jpg |
0.422411 | 7a58c67e02044fa4b0bfa37ba7eef58b | Illustration of the set-up of a flat disk inside of a fixed frame. The flat disk (a) is placed inside a fixed and undeformable frame (b) with an attractive central region (in red). Disk and frame are compatible in the flat state (c). If the disk swells uniformly, an elliptic dome-like shape results (d). | 2012.14381 | Fig7.jpg |
0.433631 | 4a1c71f74a0d4184b99ba1b916c40de5 | Energies (a), shape's height Δ z and negative lowest Hessian eigenvalue -λ_min (b) in the spring mesh of a disk in a fixed frame with attractive potential as functions of the swelling factor α. Crosses denote numerical values calculated with increasing α, while circles are related to decreasing α. The region with a genuine hysteresis is marked in yellow, while pseudo-hysteretic effects are again marked in red. The disk is always flat if it is located in the white area and always curved in the cyan region. Arrows illustrate the directions inside the hysteresis loops. The simulated system is the same system from Fig. <ref> but with an additional potential energy for each vertex given by eq. (<ref>). The simulated disk had a mesh with n_B=120 boundary vertices, γ_FvK = 1066 and maximum fluctuations of δ_max= 10^-4R_out. The parameters of the attractive potential were set to ε = 7.3× 10^-7 k_in and σ = 0.01 R_out. The potential acted in an inner region A_c with a radius of 0.2 R_out. | 2012.14381 | Fig8.jpg |
0.454814 | b8a1d277691a483899a6fcb7659c3bf9 | Illustration of the sphere mesh. Small spheres (blue circles) are placed on the vertices of a Delaunay triangulation. Example mesh with n_B = 40 boundary vertices. The spheres have a radius a=2π R_out/(10n_B), which is about 10 % of their average distance. | 2012.14381 | Fig9.jpg |
0.478839 | 81666ed3e6674f92be3312a31f8e91e0 | Construction of scattering geodesics. | 2012.14384 | scatgeo.jpg |
0.533106 | 16705fc3dedd422b890d5f3bb815c71e | (a) Two different ways to encode a state of a walking particle on a hypercube graph: one qudit (left), and n separable qubits (right). (b) Quantum circuit for a CTQW implemented on n separable qubits. (c)-(d) Quantum and classical simulation of CTQW on a 3-dimensional hypercube (c) and a 20-dimensional hypercube (d). Each plot shows a probability distribution of particle position on a graph at different times, corresponding to quantum circuit parameters ϕ=-π/2, λ=π/2 and θ = 2Ω t=π/2, π, 3π/2, 2π. The blue circles and orange triangles correspond to execution on the quantum IBM Q device and the classical simulator, respectively. Hamming distance indicates the distance at which particle propagated starting from the initial state labeled as “00… 0”. | 2012.14386 | fig1.jpg |
0.481668 | a6ed684750d94498a2c6e8c12be34702 | (a)-(b) Mapping an n-dimensional hypercube graph to a line graph. Examples for n=3 (a) and n=4 (b) are shown. Weights β are defined in Eq. (<ref>). (c) Quantum circuit for a CTQW implemented on four entangled qubits. (d) Quantum and classical simulation results for mapping a 3-dimensional cube to a weighted line with 4 vertices. Each plot shows a probability distribution of particles position on a graph at different times: t=π/4Ω, π/2Ω, 3π/4Ω, and π/Ω. The blue circles and orange triangles correspond to execution on the simulator and the IBM Q quantum register, respectively. Hamming distance indicates the distance at which particle propagated starting from the initial state corresponding to an excited state of the q_0 qubit. | 2012.14386 | fig2.jpg |
0.511871 | 36765b7ac1b74fa98dee0144aa053bfd | (a) A weighted graph generated from a 4-qubit quantum circuit with U_3 and CNOT gates on which we evolved a CTQW starting from the node “0000” and ending up in “1111” with probability one. Weights are not shown. (b) A 4-qubit quantum circuit consisting of U_3 and CNOT gates with a circuit depth of 5. (c) Quantum and classical simulation results for a quantum walk on the graph (a). Each plot shows a probability distribution of particles position on a graph for different initial conditions: |ψ_16(0)⟩=|0000⟩, |1111⟩, and |1010⟩. The blue circles and orange triangles correspond to execution on the simulator and the IBM Q quantum register, respectively. Vertex labels correspond to vertices in (a). | 2012.14386 | fig3.jpg |
0.449926 | 625f607dc22b4e6491e4a5ca6c35b53e | PACF of loads sequences (on half-hour lags) in UK-SMEC households | 2012.14389 | PACF_UK.jpg |
0.411719 | b8a09b42b0584ecc9bd1a57b5bcdcda6 | Predicted distribution quantiles over instances from UK-SMEC testset | 2012.14389 | dist_quantiles_vert.jpg |
0.396267 | c438847c17c44aae84f708239211a9aa | Schematic representation of the Bayesian MDN | 2012.14389 | netSchema.jpg |
0.464193 | f6f44aeb7df74132bd50d595bc266d36 | Predicted distributions over instances from UK-SMEC testset | 2012.14389 | predDist_MACs.jpg |
0.396204 | de6bb4aac0794791aa24aa01360f4590 | Predicted distribution samples over instances from UK-SMEC testset | 2012.14389 | samples_dist_vert.jpg |
0.485311 | 57b765d6bed34b8f84b5a5d8ca48ee51 | A coherent retroreflective metasurface which is illuminated by two incident plane waves, from top and bottom, and reflects those incident waves fully in the same directions, without any parasitic additional modes. | 2012.14391 | fig1_new_concept.jpg |
0.470286 | 1bf29d31c0c94f2ba935fa787633bf9d | (a) Schematic representation of a coherent retroreflector for TE waves. (b) Normalized reactance profile for TE- and TM-polarized coherent retroreflectors. The impedance for the two polarizations differ by the scale factor cos(θ_i)^2 (θ_i=60^∘ for this illustration), using different values of ϕ_R. (c) The coherent retroreflective sheet exhibits total retroreflection when it is illuminated from both sides; however, additional scattering is produced when it is illuminated from only one side (θ_i=60^∘, ϕ_R=0^∘ and f=10 GHz for this illustration). | 2012.14391 | fig2_sheet_theory.jpg |
0.422856 | 522eb934e207447abdc0606e1413ed8a | (a) A parallel-plate waveguide with allows propagation of TEM, TE_1 and TM_1 modes. However, a coherent retroreflective sheet placed between the plates acts as a mode filter, preventing propagation of a given mode (TM_1 mode in this proof-of-concept example). (b) and (c): Simulation results for a coherent retroreflective metasurface reveal that TM_1 can be suppressed at 10 GHz, without affecting propagation of other modes for (b) using six-discretized surface impedance stripes and (c) a metasurface implementation based on meandered slots and dipoles (see Summplementary Information). | 2012.14391 | fig3_stop_mode.jpg |
0.407911 | ade9afed6611466380d9ec35cb5fc604 | An example of adversarial attack in wireless communications. An adversary adds perturbation to a test sample to create an adversarial example. Then, the pre-trained ML model wrongly predicts the test sample. | 2012.14392 | AdEx_June2020.jpg |
0.389857 | ea17cfc9499247658c1cd760b78ceb09 | Categorization of adversarial machine learning. | 2012.14392 | adex_attack_types_new.jpg |
0.508094 | 86a24e6ed5414bce88dd3470f9948e01 | Properties of the Resonance Model transition: Panels (a) and (d): In the MBL phase and at the RM transition (1/ζ≥ 1/), the spectral function diverges at low frequencies ∼ω^-1+θ. Panels (a) and (d) summarise the behaviour of the exponent θ in the Floquet and Hamiltonian cases respectively. Both panels show θ=θ_0 → 1 deep in the MBL phase (1/ζ→∞), and θ→ 0 as the transition is approached. At the Floquet RM transition, θ jumps to a finite value θ = θ_c (orange point, panel (a)), while θ_c = 0^+ (indicating the presence of log corrections) at the Hamiltonian RM transition. Panels (b) and (e): In the vicinity of the RM transition, the correlation length |ξ| sets the cross-over frequency scale ω_ξ∼exp(-1/|θ_0|). The low-frequency behaviour (ω≪ω_ξ) is determined by the phase, while the intermediate frequency behaviour ω≫ω_ξ≫ J^-1 is determined by the transition. The two other frequency scales are set by the system size: the Heisenberg scale ω_H is the inverse level spacing, while ω_c is the scale of the smallest off-diagonal matrix elements. The thermal-region I crossover occurs when ω_ξ∼ω_H∼ω_c. In region I, only the exponent controlling the ω > ω_ξ decay is visible. This exponent is continuously varying and is significantly corrected from its value at the transition in region I (as quantified by the O(θ_0) term). The smallest value of the exponent is however set by θ_c. Panel (c): The exponent θ_0 may be directly extracted from ϱ(v), the distribution of off-diagonal matrix elements of a local operator. In the localised phase, there are exponentially many off-diagonal matrix elements which are exponentially small in range, so ϱ(v) diverges as a power-law at small v. The exponent defines θ_0. Panel (f): The time averaged correlator [C_zz] serves as an order parameter for the MBL phase. [C_zz] goes to zero smoothly as 1/ζ→ 1/ is approached from the MBL side, faster than any power law in both the Hamiltonian and Floquet cases. | 2012.14393 | Fig_BigFig3.jpg |
0.469966 | ec4ad81a14d24db5a080f4d57bac0496 | Spectral function data from Ref. <cit.>: Disorder averaged spectral function data for the random XXZ model from Fig. 2a of Ref. <cit.> (same colour scheme). Different series correspond to different disorder strengths W (legend above). Here we plot (ω W)^-2 as a function of (W ω)^-1 so that the pure 1/ω divergence predicted by Ref. <cit.> appears as a horizontal line (black solid, C = 0.0179) whereas the form predicted in this work, (<ref>), appears as line of constant gradient (black dashed). Agreement with (<ref>) is seen for 1.4 decades for (W ω)^-1∈ [1.7,40]. | 2012.14393 | Fig_DriesData.jpg |
0.416238 | 031ca967b71d45f9a15eb449b4265a74 | Drift of the critical disorder strength W_c(L) with L at small sizes: The main plot shows the RM probe spin entanglement entropy [S_P] averaged over all eigenstates vs the scaled disorder strength for L = 5, 10, 15, … 60 (coloured solid lines). The dashed black dashed line indicates the L →∞ limit. Inset: Numerically extracted W_δ with δ = 0.01 (green solid line) vs L, and the corresponding analytic curve from (<ref>) (black dotted line). The red dotted line is a linear fit at small L. We see that W_c(L) ∝ L at small L. Parameters: 1/ξ = log(W/), = 10, λ as given by (<ref>), and J=1. | 2012.14393 | Fig_FiniteSize.jpg |
0.474482 | 4aec79ced569459ba5b75ce625919ae4 | Local density of states: the local density of states at range r and energy , ρ_σ(,r), is plotted versus the local energy for the σ = (red) and σ = sectors of the probe spin. These distributions have the same width s_(r) but are offset from each other due to the probe spin energy ± h/2. The probability of a state in the sector finding a resonant partner is proportional to the density of states in the sector (see (<ref>)). We illustrate this with an arbitrary cut-off: the states at energies ∉_ + [ - 2 s_(r), 2 s_(r)] (red shaded area) have a much reduced probability of resonating versus those in the bulk of the distribution. | 2012.14393 | Fig_LocalDOS.jpg |
0.455818 | 3de75c9514c8443ba5907fcb822bed12 | Logarithmically growing light cone: the Heisenberg operator σ_P^z(t) (in (<ref>)) is localised to the probe spin site time t=0. Under time evolution, the support spreads and defines a light cone. After a time t, this light-cone has width r(t) ∝log J t (green). | 2012.14393 | Fig_LogCone.jpg |
0.494245 | 4a80f6f5bfce496da9d66c8eb527e8a6 | a) The resonance model (RM) predicts a continuous transition (orange point) between a localised (blue) and a thermal (red) phase, and an inverse correlation length |ξ|^-1 (orange lines) that vanishes with exponent ν = 1 at the transition. In region I at system sizes smaller than a resonance length λ (purple), typical eigenstates have no resonances and spectrally averaged properties resemble those of the localised phase. b) The MBL-thermal finite-size crossover: At large L in the vicinity of the RM transition (hatched region), localisation is inconsistent due to overlapping resonances. The RM is however self-consistent in the blue regions. The RM thus describes the MBL-thermal crossover in small system numerics (horizontal line), even though it does not describe the asymptotic transition (black point). | 2012.14393 | Fig_PhaseDiag2.jpg |
0.408301 | 9b15d5066d6d46fba63313ed6985a058 | Organising resonances by range: a) The many-body spectrum of _̋0 in a small quasi-energy window is divided into two sectors labelled by the state of the probe spin σ = ,. |ϵ_a ⟩ labels a specific reference state. b) The l-bit configuration corresponding to reference state (red spectral line) is shown. The states |ϵ_d ⟩ in the opposite sector (green lines) can be grouped according to their range r from the reference state (ranges r=0,1,2 shown); states at range r differ only on the l-bits with index |n|≤ r (highlighted in orange). A state |ϵ_d ⟩ at range r is resonant with |ϵ_a ⟩ if its quasi-energy separation is less than the matrix element size v(r) (i.e. if it lies within the magenta region). In the plot, the first resonance occurs at range r=2. | 2012.14393 | Fig_Range.jpg |
0.398217 | 1214d194fb564b439bc0c6bcdc5d4375 | Cartoon of approximate eigenstates: For the purposes of calculating the spectral function, the resonant eigenstates may replaced with cat states. Here the resonance is of range r=2, so that only l-bits with indices n ∈{ -2,-1,0,1,2 } (red box) are reconfigured. | 2012.14393 | Fig_cat.jpg |
0.391702 | 2eb8fbd9319b41b2817a02ccec15af21 | Set-up in the physical and l-bit bases respectively: a) A “probe” spin-12 (orange) couples to a strongly disordered chain (blue) at the site n=0 (magenta). b) Transforming to the l-bit basis renders the Floquet unitary of the chain diagonal and the probe-chain coupling quasi-local. The coupling strength decays exponentially with distance from n=0. | 2012.14393 | Fig_pbitlbit.jpg |
0.521648 | e18706dde59f4928a3093843aa7842c7 | Resonances: The spectral function calculation in the RM is self-consistent if the eigenstates in the RM-MBL are well characterised as l-bit configurations dressed with local resonances. a) A l-bit state dressed with two resonances of range r=1 centred at sites n=-3 and n=2. Each resonance is represented by an arc encompassing the patch of rearranged spins. Resonances typically rearrange a patch of size ξ and have density ξ/λ, and thus are well separated for ξ^2 ≪λ. b) For ξ^2 ≳λ, these resonances typically overlap forming large resonant patches that destabilise MBL. | 2012.14393 | Fig_resonances.jpg |
0.47166 | 323d81ca6c834205b2003762d433f8bf | Decay in F(r) for energy independent matrix elements: Values of λ |F̣ / ṛ| are plotted versus r/λ, these are obtained by numerically solving (<ref>) and (<ref>). The point r/λ = 1 is marked with a vertical grey line. For r/λ < 1, the behaviour is consistent with F(r) = exp( - √(r/λ)) (dotted line). For r / λ > 1, the decay is slower F(r) ∝ (λ/r)^2 (dashed). Different series correspond to different values of λ (legend inset). | 2012.14393 | SuppFig_pde.jpg |
0.387766 | 29f1f315dc1b4462bbbb0c8a127bc925 | Decay in F(r) for energy independent matrix elements: the distributions f_σ(,r) are plotted for log-spaced intervals of r, using the same numerical solutions to (<ref>) and (<ref>) as Fig <ref>. In each case it is clear that at large ranges the distribution is depleted at energies ≲ s_(r). | 2012.14393 | SuppFig_pde2.jpg |
0.463044 | f8165d06df4f47b7b02d83291f3d7f68 | λ=1/10. | 2012.14394 | nonscalar_k10r10.jpg |
0.429228 | 1e5fd952e8434f458f0938f2620d0c00 | λ=1/50. | 2012.14394 | nonscalar_k10r50.jpg |
0.507777 | d8c715d9e18d43288e891aeaf3b558ea | λ=1/10. | 2012.14394 | nonscalar_n15k6L23_yma.jpg |
0.454795 | 0639493bdbfa454fbdbbf3f56bbaba37 | λ=1/15. | 2012.14394 | nonscalar_n15k6_yma.jpg |
0.423557 | d3328e1655784951a1f2129b440eb07d | Visualizing effects of different regularization terms. Columns (a), (b) and (c) represent saliency maps of original and perturbed image considering only ℒ_CACR, only ℒ_WACR and both regularizers respectively in outer minimization (Eqn <ref>). We note that all models predicted correctly while perturbing the original image using this attack (as in <cit.>) | 2012.14395 | AB1.jpg |
0.366447 | 2ab3823ea51d49c5aee83108c2d9b469 | Visualizing effects of different regularization terms. Columns (a), (b) and (c) represent saliency maps of original and perturbed image considering only ℒ_CACR, only ℒ_WACR and both regularizers respectively in outer minimization (Eqn <ref>). We note that all models predicted correctly while perturbing the original image using this attack (as in <cit.>) | 2012.14395 | AB2.jpg |
0.419583 | 438bb49c114b4f0e9c3b78011b3d090b | Visualizing effects of KL(P_ME || P_NCA) | 2012.14395 | KL2.jpg |
0.490264 | ab3811067a56400b84f1bb7244374025 | (Best viewed in color) Variance in attribution robustness metrics, Top-k intersection score and Kendall's correlation, for proposed method and RAR over all test samples on different datasets. Blue=Top-k intersection score; Green=Kendall's correlation. Our method achieves significant improvement on both metrics over RAR for all datasets, while the variance is fairly similar to variance in RAR | 2012.14395 | Result_box_all_datasets.jpg |
0.404629 | d281ec5b07c24204b7347a360e90249f | Visualizing effects of different regularization terms. Columns (a), (b), (c) represent saliency maps of original and perturbed image considering only ℒ_CACR, only ℒ_WACR and both regularizers respectively in outer minimization (Eqn <ref>). All models predicted correctly while perturbing the original image using this attack (as in <cit.>) | 2012.14395 | ablation_main.jpg |
0.42002 | a40b57c3a7e643c5be6544813e4dfc9c | Original image and human-imperceptible attributional attacks (same predicted class) and their corresponding saliency maps under: (Left) adversarial training <cit.>; (Middle) RAR <cit.>; and (Right) our proposed robust attribution training for a test sample from Flower dataset. Note that the attacks/perturbed images in each column are different, since the attack is carried out based on the model trained on that particular method. | 2012.14395 | intro_hr.jpg |
0.525155 | 2915188d11c84a0bb7f2df1ce80e1be3 | Original image and human-imperceptible attributionally attacked image (both correctly predicted) and their corresponding saliency maps under RAR <cit.> and our method for a test sample from Flower dataset. | 2012.14395 | sup_flower1.jpg |
0.422665 | b2cdb734795d4d7fbde2eb57c58c67e6 | Original image and human-imperceptible attributionally attacked image (both correctly predicted) and their corresponding saliency maps under RAR <cit.> and our method for a test sample from Flower dataset. | 2012.14395 | sup_flower2.jpg |
0.428176 | 67ce30546e274abd80848cd6a0fdfc05 | Original image and human-imperceptible attributionally attacked image (both correctly predicted) and their corresponding saliency maps under RAR <cit.> and our method for a test sample from Flower dataset. | 2012.14395 | sup_flower3.jpg |
0.479707 | 7adde119701e4452812366853ba4bd9f | Original image and human-imperceptible attributionally attacked image (both correctly predicted) and their corresponding saliency maps under RAR <cit.> and our method for a test sample from Fashion-MNIST dataset. | 2012.14395 | sup_fmnist.jpg |
0.4605 | 1124c166c4f9408aa4f1ad5dd825eef4 | Original image and human-imperceptible attributionally attacked image (both correctly predicted) and their corresponding saliency maps under RAR <cit.> and our method for a test sample from Fashion-MNIST dataset. | 2012.14395 | sup_fmnist1.jpg |
0.407293 | 24f212f8ed954a7d9d1d7cfa886b3cbd | Original image and human-imperceptible attributionally attacked image (both correctly predicted) and their corresponding saliency maps under RAR <cit.> and our method for a test sample from Fashion-MNIST dataset. | 2012.14395 | sup_fmnist2.jpg |
0.45299 | 3b54a472452e4e989a5b5da9a7a8b71a | Original image and human-imperceptible attributionally attacked image (both correctly predicted) and their corresponding saliency maps under RAR <cit.> and our method for a test sample from GTSRB dataset. | 2012.14395 | sup_gtsrb.jpg |
0.431266 | 47302bad17494233bc310e13972bb1b1 | Original image and human-imperceptible attributionally attacked image (both correctly predicted) and their corresponding saliency maps under RAR <cit.> and our method for a test sample from GTSRB dataset. | 2012.14395 | sup_gtsrb1.jpg |
0.478238 | 64ea7e2b2e304425af6131646811386c | Original image and human-imperceptible attributionally attacked image (both correctly predicted) and their corresponding saliency maps under RAR <cit.> and our method for a test sample from GTSRB dataset. | 2012.14395 | sup_gtsrb2.jpg |
0.462716 | cfa9788421704c738021f828f6ecdb1a | Original image and human-imperceptible attributionally attacked image (both correctly predicted) and their corresponding saliency maps under RAR <cit.> and our method for a test sample from MNIST dataset. | 2012.14395 | sup_mnist.jpg |
0.439734 | b704220fd928409190ce7249ac40de0e | Original image and human-imperceptible attributionally attacked image (both correctly predicted) and their corresponding saliency maps under RAR <cit.> and our method for a test sample from MNIST dataset. | 2012.14395 | sup_mnist1.jpg |
0.469561 | 7e05f491f5fe47ccafe0298a9e2116fa | Original image and human-imperceptible attributionally attacked image (both correctly predicted) and their corresponding saliency maps under RAR <cit.> and our method for a test sample from MNIST dataset. | 2012.14395 | sup_mnist2.jpg |
0.46179 | 50ed1a5521f349dcba7763d8dc709a8a | Six example objects for category learning that vary among numerous features (e.g. shape, texture, color, size, and position). | 2012.14400 | catdog.jpg |
0.528215 | d848af83f4b440c39d334a1b62e1e871 | Color Constancy Index (CCI) for the 5 readout models tested with the normal and no patch image sets. Each readout take input from all units of the designated layer. By extension, the value of CCI reflects the degree of color constancy at the different layers of DeepCC. | 2012.14402 | CCI_fost_readout.jpg |
0.49125 | fd7dc8e23023492a8cbb0327351073d4 | DeepCC's evaluation results obtained for all measures and all conditions. Panel A: Performance under the CC and D65 conditions (Section DeepCC and Deep65 evaluation); Panel B: Performance of DeepCC under the no patch, wrong back and no back (Section Impoverished visual scenes); C: Performance of DeepCC compared to classical approaches to color constancy (Section Classical approaches). | 2012.14402 | fig_measures_3_grey.jpg |
0.408325 | 932b563e39194f0fb7b7351534ff6c9e | Plots for the neutralino species in the μ-M_B̃ plane. All masses plotted here refer to the values at the low scale. The color coding is the same as Figure <ref>. The diagonal line shows the solutions in which the Higgsinos and Bino are degenerate in mass (μ = M_B̃). | 2012.14411 | HMGUT_splitNUHM2_muMA_mu_M1.jpg |
0.540288 | 8d2df915a8ec445aa049fce3d6d0466b | Lifetime of the proton in correlation with the phase angle ϕ_2 with ϕ_3 set to zero. Green, blue and red dashed curves represent the proton decay into ν_e, ν_μ and ν_τ along with K^+, respectively. The solid curve shows the total lifetime of the proton. | 2012.14411 | point1_tau.jpg |
0.546092 | 9076561ae07d4b57bf57369e24bf38fb | Comparison of computational complexity, number of parameters, frames per second and classification accuracy from different networks for the image classification task on the ImageNet dataset. Values are shown in proportion to the ResNet-50. | 2012.14426 | ImageNet2.jpg |
0.434986 | d8ac22e6c3a54e33b62088fef7150edd | JPEG compression and decompression process <cit.>. | 2012.14426 | jpeg.jpg |
0.530527 | 88731b7e2ced4ff7a334492d575a19ad | 2012.14431 | 2d1.jpg |
|
0.551163 | 03e8aa60b2bd4c09973088ea51cf98ba | 2012.14431 | 2d2.jpg |
|
0.511305 | 98ea9cd0eb52480ea2df5e1eb9c435e2 | Π^ 5μνδ | 2012.14431 | 4d1.jpg |
0.533729 | c79f91d5409640aa92bf02d4e834f693 | Π^2 5μνδ | 2012.14431 | 4d2.jpg |
0.477038 | 6c97b2b40a7644cabd49a3838f3480b9 | Π^3 5μνδ | 2012.14431 | 4d3.jpg |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.