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
|
|---|---|---|---|---|
0.444879 | da621d8fc27147d4b6837095492c667c | Performance of the the wavelet filter on the signal. | 2012.13448 | filtering_onlywavelet.jpg |
0.481235 | 24373bc2c69949bb883cd57cfcd38f9b | The power of the first pilot subcarrier and the number of vehicles on the road. | 2012.13448 | freq_response.jpg |
0.494997 | 5bcb7635466a44ad9cf91c1a1986ae29 | The methodology that integrates the simulators. | 2012.13448 | meth_wo.jpg |
0.429147 | 0bca43371f09421dbd50d7aff76e0f17 | The vehicle collected SPaT messages 200 meters away from the roadside unit. | 2012.13448 | moving.jpg |
0.401058 | 002cdcf2c95641c996a352a36878b740 | The vehicle collected SPaT messages 200 meters away from the roadside unit. | 2012.13448 | moving.jpg |
0.490785 | c9c109ab747d4a9c9bf789919d785ffe | Original and outlier filtered signals. | 2012.13448 | outlier.jpg |
0.475783 | 16da6dfe1d654d0cad069bac1ed7f05e | Problem model. | 2012.13448 | problem_woSpeed.jpg |
0.426916 | 0ef33da3d2444cc0addfdb4abde3a1ef | 2012.13448 | roc.jpg |
|
0.382192 | 07852097af3e4f258b0dfa6851b0e684 | A typical communication scenario in a vehicular network. | 2012.13448 | traffic_figure.jpg |
0.506072 | e80da1835f80426f8b54f2a94ea6d82b | The 12 AOIs in the cockpit | 2012.13449 | AOI_car.jpg |
0.409453 | d12b609e1c9945178781751b81211bc5 | Confusion matrix for fusion of all three modalities for the 12 classes | 2012.13449 | Confusion_Matrix12.jpg |
0.406678 | dd7463951087428d844d16966df48588 | Confusion matrix for fusion of all three modalities for the 7 classes | 2012.13449 | Confusion_Matrix7.jpg |
0.400935 | 67c57c06dd6c4f04b0f506e185e8f2c8 | Confusion matrix for fusion of all three modalities for the 9 classes | 2012.13449 | Confusion_Matrix9.jpg |
0.537412 | 5336836e48684336be4a3d55800dbefa | A multimodal late fusion architecture | 2012.13449 | Methodology.jpg |
0.43159 | ad8830aa21a74b07b5deea0573f3003b | Driver makes a pointing gesture to interact with the car. Image courtesy of <cit.> | 2012.13449 | bmw-natural-inteaction.jpg |
0.406444 | 24fc7e7b31204866bd2bfc7d40762488 | Driver points to AOI 5 (multimedia) in the car. The face is blurred due to the privacy policy. | 2012.13449 | driver_pointing.jpg |
0.415467 | 7d57c580c13444f28582878fabe66cc4 | Model performance on the individual drivers | 2012.13449 | driver_test.jpg |
0.439806 | 83d226780e9847b795d3b776d8b86cec | Change of origin from centre of front axle of the car (left) to the centre of the driver's seat (right) | 2012.13449 | origin.jpg |
0.454898 | ee93230af5bc43ca91fed117d7f9d997 | The modalities used: gaze and finger ray-cast (left) and head pose (right) | 2012.13449 | ray_casting.jpg |
0.400081 | f1fb3a7e14bc49618b5c722ef6330899 | 3D scatter plot of the measured AOI points. | 2012.13449 | scatter_AOI.jpg |
0.416904 | a6e3e5f1c49848c2af1693ecac397b25 | Circuit diagram of the gradual evolution of the circuit architecture optimization the transverse field Ising Hamiltonian, (Left) after 10 generations and (Right) after 150 generations. The notation R_i with i ∈{x,y,z, xx, yy, zz} represents a rotation gate wrt to axis i (for the purpose of clarity the explicit rotation's angles are omitted). Note that gates spanning multiple qubits are only applied to the upper- and lowermost qubit. Note that by generation 150 some previously present gates with redundant parameters are removed. Moreover, the circuit generally reflects intelligence in its design and indeed is nicely aligned the expected useful set of gates for the TFI problem. | 2012.13453 | circuit.jpg |
0.484043 | 5f34810e170e4ec894114bfa69ff60db | (Left) Outline of evolutionary circuit discovery: Starting with a random initial parent with known loss value, we create copies and mutate them using our custom mutation strategy using operations for instance as shown on the right. The offspring population is then evaluated on the quantum machine. The circuit with lowest loss value (the original parent included) becomes the parent circuit of the next generation. This procedure is repeated until an appropriate convergence criterion is satisfied. (Right) Mutation performs one of 4 actions with certain probabilities, we show these on the right. INSERT: Add random gate at random position; DELETE: Delete gate at random position; SWAP: Replace gate at random position with random new gate; MODIFY: Change parameter of gate at random position. | 2012.13453 | figure_1.jpg |
0.428981 | 08ceddbeaec742ed94b04d39ab08189a | A human providing demonstrations to the robot for a manipulation task through kinesthetic teaching. | 2012.13457 | 2.jpg |
0.479889 | 46595a3dce374ba4be07cd0362f335bc | Top: Structure of the network defining a single map ψ_m in the diffeomorphism chain <cit.>. Bottom: Structure of the network for defining latent subtask metric __k <cit.>. | 2012.13457 | network_new.jpg |
0.351195 | 47519018e62a4b73916ff850fa86a670 | A transform tree with root in the configuration space alongside hand-specified subtask/leaf nodes (grey) and learned subtask nodes (blue). Each learned subtask node is linked to a latent subtask node (green) under a chained map ψ__k→_k = ψ_1∘…∘ψ_M. | 2012.13457 | tree_new.jpg |
0.451678 | 3e710650d5a74ac393b9bda68ef9ad6b | Experimental set-up, photonic dispersion, and optical bistability. (A) Illustration of our experiment, involving an oil-filled optical cavity driven by a continuous wave laser. The cavity length is controlled with a piezoelectric actuator holding one of the mirrors. The transmitted power is time-resolved with a photodetector, or measured in position and momentum space with cameras. (B) White light transmission measurements of our oil-filled cavity. The measured dispersion relation is fitted by a parabola, indicating the cavity photons behave as free particles with an effective mass m^∗= 2.064 × 10^-35 kg. (C) Transmitted power measured at k_∥=0 when ramping the laser power up and down, evidencing optical bistability due to effective photon-photon interactions. | 2012.13463 | fig1.jpg |
0.380152 | cf185c04a3594a03bc6996d75aefbb08 | Superfluid photons above a critical density. Experimental transmitted intensity as a function of position in (A) and (B), and momentum in (D) and (E), for a laser power indicated in each panel. (C) Cuts along the flow axis passing through the center of the laser spot and the center of the defect in (A) and (B). The peaks observed at 20 mW but not at 340 mW are the result of scattering from the defect, which interferes with the incident flow and generates spatial fringes. (F) Cuts along k_y=0 in (D) and (E). The peaks observed at 20 mW inside the solid-line ellipse in (D), and in the shaded region in (F), evidence backscattering from the defect; these peaks disappear at 340 mW. (G to L) Numerical calculations as described in Methods, for two different laser amplitudes F referenced to the cavity loss rate γ. F/γ is such that fluid density is well-below and well-above the threshold for bistability in (G) and (J), and (H) and (K), respectively. Consequently, interference fringes are present in (G) but not in (H), and backscattering states inside the solid-line ellipse are present in (J) but not in (K). As for the experimental data, the cuts in (I) are taken along the flow axis passing through the center of the laser spot and the center of the defect in (G) and (H), and the cuts in (L) are taken along k_y=0 in (J) and (K). Scale bars indicate 5 μm. Solid dot in (A) corresponds to the position of 0 μm in (C). Dashed circles in (A, B) indicate approximate location of defect. Dashed circles in (G, H) indicate the location of the defect in simulations. The cuts in (C, I) are integrated vertically over the laser spot size and in (F, L) over Δ k_y=0.06 μm^-1. In (C) data has been normalized to intensity at 0 μm. In (F, L) data for both powers has been normalized to the intensity at the peak of the incident momentum. Solid and dashed lines in (C, F, I, L) correspond to low and high power, respectively. | 2012.13463 | fig2.jpg |
0.46461 | a44339e6f39848bba9e0480d25b0ce24 | From transient scattering to steady-state superfluidity. (A and B) Time-dependent photon density N along the flow axis passing through the center of the laser spot and the center of the defect. The thermal relaxation rate Γ_T is 1.5 ns^-1 in (A) and 0.015 ns^-1 in (B). Consequently, a steady state is reached much faster in (A) than in (B). (C and D) Two-dimensional plots of N at t=0.2 ns, indicated by the dashed lines in (A) and (B). After this time, the system has reached steady state for Γ_T=1.5 ns^-1 but not for Γ_T=0.015 ns^-1. Consequently, (C) displays the absence of interference fringes characteristic of superfluidity, while in (D) interference fringes are visible. (E) Time-dependent blueshift of excitation spectrum for Γ_T=0.015 ns^-1. (F) Spectrum of excitations at times indicated by corresponding circles in (E). | 2012.13463 | fig3.jpg |
0.399001 | 63a0066f742a48e3add8cab46d27678f | Superfluidity of photons below a critical velocity. Experimental transmitted intensity as a function of position in (A and B), and momentum in (C and D), for slightly different cavity lengths corresponding to a 4% change in the magnitude of the flow velocity |v|. The laser power is 360 mW for all measurements. Scale bars in (A) and (B) correspond to 5 μm. (E) Cuts along the flow axis passing through the center of the laser spot and the center of the defect in (A) and (B) as solid and dashed lines, respectively. The peaks observed for the larger |v| (solid line) are due to scattering from the defect, which interferes with the incident flow and modulates the density. (F) Cuts along k_y=0 in (C) and (D) as solid and dashed lines, respectively. The peaks observed for the larger |v| (solid line) inside the shaded region, which correspond to the peaks observed inside the solid-line ellipse in (C), evidence backscattering from the defect; these peaks disappear for the smaller |v|. Red vertical lines in (F) indicate the main incident wavevector, which changes only slightly in the viscous-to-superfluid transition. Solid dot in (A) corresponds to x=0 μm in (C). Dashed circles in (A, B) indicate approximate location of defect. The cuts in (E) are integrated vertically over the width of the laser spot and in (F) over Δ k_y=0.17 μm^-1. In (E) the data for both velocities has been normalized to the intensity at x=0 μm, and in (F) to the intensity at the peak of the incident momentum of the solid-line cut. | 2012.13463 | fig4.jpg |
0.464537 | 77974d8840a04d32818eb0b94568558f | Embedding diagram of the WH in three-dimensional Euclidean space for different constant values of y with a=0.1 and L=5. Here, z and ρ are the usual cylindrical coordinates. | 2012.13465 | geoCC.jpg |
0.45143 | 9c16038b8724428b807b3bd7d8912a1f | Cross section of the singularity σ' in the plane u_1-u_3 for different values of a/L^2, for simplicity we have made L=1. Here, u_3 is the symmetry axis. In the right panel we show a particular case where a=0.1, and hence a/L^2≪1. | 2012.13465 | singularidad.jpg |
0.439245 | 601070a0dd194e08bbc73c518fd295f7 | Illustration of the different configurations of the GRA module. (a) Mode 1. (b) Mode 2. (c) Mode 3. | 2012.13466 | FigX_GRA_configuration.jpg |
0.396676 | 38fe20c27ef54c6ca547904b58cbd792 | Top view of the ISPRS benchmark dataset colored by elevations. (a) Training and (b) test areas. It should be noted that the scales of two color bars are different. | 2012.13466 | FigX_ISPRS_Benchmark.jpg |
0.413018 | 29eb6ec807174519a4202e3444db3559 | Classification results of test area 1 of the ISPRS benchmark dataset. (a) Ground truth. (b) PointNet. (c) PointNet++. (d) HDA-PointNet++. (e) PointSIFT. (f) Ours. | 2012.13466 | FigX_ISPRS_Results1.jpg |
0.472131 | e0a5515925eb4deda6be17ad6cf685c4 | Classification results of test area 2 of the ISPRS benchmark dataset. (a) Ground truth. (b) PointNet. (c) PointNet++. (d) HDA-PointNet++. (e) PointSIFT. (f) Ours. | 2012.13466 | FigX_ISPRS_Results2.jpg |
0.418374 | 08230eeddd034ae2aad01d0d2a34dc5f | Details of classification results of the ISPRS benchmark dataset. (a) Ground truth. (b) PointNet. (c) PointNet++. (d) HDA-PointNet++. (e) PointSIFT. (f) Ours. | 2012.13466 | FigX_ISPRS_Results_Details.jpg |
0.466887 | b4e20c91c7a3498f89020c501442fed1 | The error map of the classification results of the ISPRS benchmark dataset using our method. | 2012.13466 | FigX_ISPRS_Results_ErrorMap.jpg |
0.386414 | 8c3e9fa8fd27434fb40f108d3fc386d4 | Classification results using different conjurations of the GRA module. (a) Ground truth, (b) Non GRA. (c) Only SRA. (d) Only CRA. (e) Mode 1. (f) Model 2. (g) Mode 3. | 2012.13466 | FigX_ISPRS_Results_GRA.jpg |
0.402605 | 90938d7c97c24184a63c05a33c577740 | Classification results using different local spatial encoding methods. (a) Ground truth, (b) Model A. (c) Model B. (d) Model C. (e) Model D. (f) Model E. | 2012.13466 | FigX_ISPRS_Results_LoSDA.jpg |
0.428154 | a5e0393cde344dafba392c0fa7985642 | Illustration of the data division of the ISPRS benchmark dataset. (a) The division of the large training blocks. (b) The further subdivision of each training block. | 2012.13466 | FigX_ISPRS_benchmark_split.jpg |
0.397921 | 76753490807846429d690298641a2605 | Difference between (a) MLS and (b) ALS point clouds of a same street scene. (c) Different points distribution of two point clouds. | 2012.13466 | FigX_Illustration.jpg |
0.448573 | 6b3d594f063b4b8a8cbdaabf1f128414 | Top view of the LASDU dataset colored by elevations. (a) Training and (b) test areas. It should be noted that the scales of two color bars are different. | 2012.13466 | FigX_LASDU_Benchmark.jpg |
0.462434 | eacdfaec3f6f46ec84e48895453f43f3 | Classification results using the LASDU dataset with different semantic labeling methods. (a) Ground truth, (b) PointNet. (c) PointNet++. (d) HDA-PointNet++. (e) PointSIFT. (f) Ours. | 2012.13466 | FigX_LASDU_Results1.jpg |
0.37918 | a618a6ecaf8d4274a8fe15fd9484af20 | Details of classification results using the LASDU dataset. (a) Ground truth. (b) PointNet. (c) PointNet++. (d) HDA-PointNet++. (e) PointSIFT. (f) Ours. | 2012.13466 | FigX_LASDU_Results_Details.jpg |
0.405256 | 2c77b2f0ee7945a09dedc42f6504baa6 | Detailed network architecture. | 2012.13466 | FigX_Method_DetailedArchitecture.jpg |
0.421091 | e7053aa81ba8495fb3f82e4a78e4bc0b | Illustration of the GRA module. (a) is the SRA module. (b) is the CRA module. | 2012.13466 | FigX_Method_GRAModules.jpg |
0.433776 | 557a5bed9e544c4b929bc22c6b066871 | Illustration of the multi-scale network framework. | 2012.13466 | FigX_Method_GraNetStructure.jpg |
0.511199 | 80ffc358c5334c9e9f38243a71e78b5d | Exploited local spatial attention encoding module. | 2012.13466 | FigX_Method_LoSDEModule.jpg |
0.418499 | 854095a154e14f14bd2fe4b64b0dca16 | (a) Schematic diagram of a subretinal implant in the eye. (b) Monopolar stimulation with 0.5 of current at each electrode. (c) Bipolar stimulation with 0.5 at the center electrode, and -0.25 at each of the side electrodes. (d) Virtual electrode created at the center-right by -0.5, 0.3 and 0.2 at the left, center and right electrode, respectively. (b)-(d) are synthesized by linear combination of the three elementary electric fields. | 2012.13467 | Figure1.jpg |
0.406664 | d0006e58020045dba0ee634dd9890401 | Lateral confinement of the electric field at three stimulation depths. Top: potential difference between 45 and 80 from the implant; Middle: 25 to 60; Bottom: 5 to 40. | 2012.13467 | Figure2.jpg |
0.478078 | 8834c150ffd8490ca271019e93c851ba | Structures of protonated and deprotonated water clusters. Five protonated water clusters H_3O^+(H_2O)_n (n = 1–3, and 6) and seven deprotonated water clusters OH^- (H_2O)_n (n = 1–6) are taken from Ref.. The red and white spheres respectively indicate oxygen and hydrogen atoms. The dashed lines represent hydrogen-bonds. | 2012.13469 | fig-1.jpg |
0.478453 | 1a0c8c18408c400198d4818f5bd1200a | Structures of the studied (a) water–halide clusters and (b) water–alkali clusters. Oxygen and hydrogen are represented by red and white spheres. The anions and cations are marked by other colors. Dashed lines represent hydrogen-bonding. | 2012.13469 | fig-2.jpg |
0.447727 | c0d69b9e51e44df1bbb642e3ccfed176 | (a) The basis-set dependence of the total energy of OH^-, plotted with respect to the diffuse-4 basis (see text for the details of basis). Variation of the binding energy (E_b) of the (b) OH^-(H2O) and (c) H3O^+(H2O) molecules with increasing size of the basis. Here DFA@FLOSIC refers to the DFA energy computed at the corresponding FLOSIC–DFA density. At the complete basis set limit, only SIC can bind the full excess electron in OH^- and OH^-(H2O). | 2012.13469 | fig-3.jpg |
0.456735 | 264b9314d7654271942a412e4dbea9eb | Projection onto the OH^- axis of the PBE self-interaction error of the electron density for the hydroxyl group and two hydroxyl-water clusters. For further explanation, see the last two paragraphs of section 3.1. | 2012.13469 | fig-4.jpg |
0.435226 | 0aecf45d7d4e47768babbf3b56b4a7dd | Error in the binding energies of the (a) deprotonated water clusters OH^- (H_2O)_n (n = 1–6) and (b) protonated water clusters H_3O^+(H_2O)_n (n = 1–3, and 6) with various functionals compared to the CCSD(T)-F12b reference <cit.>. Here a negative sign indicates overbinding. Self-interaction correction to the binding energies (E_b^ SIC) of the (c) deprotonated and (d) protonated water clusters. | 2012.13469 | fig-5.jpg |
0.445348 | 2b7a725246a843b9a215262a242520d5 | Isosurfaces (0.02 e/Å^3) of FLO densities from FLOSIC–SCAN for OH^-, H2O, H3O^+, OH^-(H2O), and H3O^+(H2O), showing lone-orbitals in blue and bond-orbitals in green. | 2012.13469 | fig-6.jpg |
0.444033 | 14f124f26d1b4bb99091964da70f1775 | (a) Error in the binding energies and (b) self-interaction correction to the binding energies (E_b^ SIC) of X^-(H_2O)_n with X=(F, Cl, Br) and n=1,2 with various functionals compared to the CCSD(T)-F12b reference. | 2012.13469 | fig-7.jpg |
0.487817 | df8f56b2154f45f0acaf79cb57b7f8d0 | (a) Error in the binding energies and (b) self-interaction correction to the binding energies (E_b^ SIC) of M^+(H_2O)_n (M=Li, Na, K) and n=1,2 with various functionals compared to the CCSD(T)-F12b. | 2012.13469 | fig-8.jpg |
0.455656 | ccc915878d9b42a69c84a4ec0f4921d5 | Mean unsigned error (MUE) given by different functionals for (a) deprotonated water, (b) protonated water, (c) water–halide, and (d) water–alkali clusters, shown from the top to the bottom. | 2012.13469 | fig-9.jpg |
0.551352 | 265edbdb79b44e2a92d1e6103fe03a32 | A rational -parallelogram envelope for n | 2012.13471 | fig1.jpg |
0.493291 | 5d4b98a557f94befb2debcf42ea17c6c | A -parallelogram envelope with angle τ for n | 2012.13471 | fig2.jpg |
0.474004 | 0d8ca3725a6c47a48abed32c1865b787 | The role of w in -parallelogram envelope for n | 2012.13471 | fig3.jpg |
0.53953 | 60f1f82269f64a698e0f095f539ee00a | A -parallelogram envelope with ratio m for n | 2012.13471 | fig4.jpg |
0.377788 | f4a5c7ad69ee4d1eaa258e58856dcc20 | Two -parallelogram envelopes with ratio m for n and m n | 2012.13471 | fig5.jpg |
0.4705 | 84c40d44643a42b29576d106ed794099 | A τ-triangle for 2n n' obtained from a -parallelogram envelope for n | 2012.13471 | fig6.jpg |
0.408171 | f8054100699d4ed4879b206f8be337c3 | A -parallelogram envelope with angle τ for n obtained from a rational τ-triangle for 2nn' | 2012.13471 | fig7.jpg |
0.48899 | a33583bb1ff64a8c845c0e9e7fc8e892 | The eigenvalues λ of the linearized Eliashberg equation in the single-band Hubbard model. For reference, we plot the red (blue) marks represent the case where hopping parameters derived from DFT are used for Hg1201(Tl1201). The arrows indicate the values of λs in Hg1201 and Tl1201 when we use the evaluated U_screened in the constrained-RPA. The red and blue solid lines in the figure are guide for the eye. We write the names of compounds in abbreviated form. | 2012.13472 | FLEXv6.jpg |
0.466849 | 0045e1ed13e0497d9fb8a3104b172a96 | Atomic structures of (a)αBa_2CuO_4F and (b)βBa_2CuO_4 with α=Tl,In,Ga,Al and β=Hg,Cd,Zn. In TlBa_2CuO_5, F of αBa_2CuO_4F is replaced by O. | 2012.13472 | a_b_str_v2.jpg |
0.443155 | 773fe77a2bba4820aa9b91de1a364d20 | Band structures of (a)HgBa_2CuO_4, (b)TlBa_2CuO_5, (c) TlBa_2CuO_4F, (d) InBa_2CuO_4F, (e) GaBa_2CuO_4F, (f) AlBa_2CuO_4F, (g) CdBa_2CuO_4 and (h) ZnBa_2CuO_4. Wannier-interpolated bands are plotted with a red line. Here, the Fermi level is set to 0. | 2012.13472 | band_eight_eps.jpg |
0.490763 | 7ced6d4d93f84325a5dcd8ce5218316c | Log-log plot of autocorrelation time τ_ac / Δτ and phonon frequency ω/t for λ = 0.1 and λ = 1.0. The black dotted lines are linear fits for each curve. For λ = 0.1, the slope is -0.9876 with R^2 = 0.9993 and for λ = 1.0, the slope is -0.8617 with R^2 = 0.9942. Can we scrutinize the λ=1 results and make sure the data are reliable? For any ω where we do not expect the result to be exact within error bars, the autocorrelation time is under-estimated. Again, this might not show up in a 'normal' run, if the semiclassical wf is biasing the sampling SO much | 2012.13473 | autocorr.jpg |
0.419307 | 334b0e1228544387a648a37ff0e39f96 | Born-Oppenheimer potential energy surfaces in units of t for the 2-electron and 2-site Holstein model: (a) ω=t, λ = 0.1t, and g=0.447t and (b) ω=t, λ = t, and g=1.414t. The minimum in (a) is E=-2.40 t at (X_1=0.63, X_2=0.63) while the two minima in (b) are E=-8.25 t at (X_1=3.90, X_2=0.10) and (X_1=0.10, X_2=3.90), respectively. | 2012.13473 | figure1.jpg |
0.44488 | 9fe29229575740b5961b3a23a0c2f120 | Total energy per site in units of t for the 20-site 20-electron 1D Holstein model as a function of ω for various values of λ values: (a) λ = 0.1 results, (b) λ = 0.3 results, (c) λ = 0.8 results, and (d) λ = 2.0 results. Note that all error bars for AFQMC are too small to be seen on the plotted scales, DMRG reference values are unavailable for ω < 0.8 when λ = 0.8 and λ = 2.0, and AFQMC/TP(11) is not presented for λ = 0.1 because the results for this trial wavefunction are nearly identical to those of AFQMC/S. | 2012.13473 | figure_20site.jpg |
0.439144 | fbef361653024d4b95659ec086de6bba | Error in total energy per site compared to DMRG in units of 0.001t for the 20-site 1D Hubbard-Holstein model at half-filling with U/t=4 as a function of ω for various λ values: (a) λ = 0.1 results and (b) λ = 1.0 results. The black dotted line in (a) indicates the electronic CP-AFQMC energy error (i.e., λ = 0.0). In (b), DMRG energies are only available for ω = 1.0, 1.5, 2.0. Note that for (b) the reference DMRG values are potentially unconverged with a maximum phonon number of 40 and the remaining time step error of CP-AFQMC is not negligible on the plotted energy scale. | 2012.13473 | figure_20site_error.jpg |
0.48645 | 65c813b3faf047a3b581cd11ff107b8e | The total energy per site in units of t for the 20-site 1D Hubbard-Holstein model at half-filling with U=4t as a function of λ for various ω values: (a) ω = 0.1, (b) ω = 0.4, (c) ω = 1.0, and (d) ω = 2.0. Note that all error bars of CP-AFQMC are too small to be seen on the plotted scales. Note that the DMRG results are unavailable for ω = 0.1 and ω = 0.4 as well as for λ > 2. | 2012.13473 | figure_20site_hh_new.jpg |
0.469602 | 7c3694ffc0bd48baa1c576b20b0efb25 | Total energy per site in units of t for the 4x4 2D Hubbard-Holstein model with U/t=4 as a function of ω: (a) ω = 0.1 results at half-filling, (b) ω = 2.0 results at half-filling, (c) ω = 0.1 results at 1/8 hole-doping, and (d) ω = 2.0 results at 1/8 hole-doping. Note that all error bars of CP-AFQMC are too small to be seen on the plotted scales. | 2012.13473 | figure_4x4_hh_2d_final.jpg |
0.460532 | 3c0dbcfd2b9a487e8e0a2f5415be171a | Total energy per site in units of t for the 4x4 2D Holstein model at half-filling as a function of ω for various λ values: (a) λ = 0.1 results, (b) λ = 0.3 results, (c) λ = 0.5 results, and (d) λ = 2.0 results. Note that all error bars of CP-AFQMC/S are too small to be seen on the plotted scales, DMRG reference values are only available for ω≥ 0.8 when λ < 2.0. | 2012.13473 | figure_4x4_holstein.jpg |
0.399262 | 6c480c4b3614405995f7fd7df41b1a5e | Error in the total energy in units of t for the 2-site 2-electron Holstein model as a function of ω for various λ values: (a) CSMP2, (b) LFPT2 and (c) AFQMC results. For λ = 0.1, LFPT2 energy errors lie outside the plotted range. In (c), for λ≥0.5 AFQMC/TP(11) results are shown while for λ = 0.1 we present AFQMC/S results. Note the different vertical scales in panel (c). | 2012.13473 | lambda_shiwei.jpg |
0.495769 | c7ad914846f84b04bd0860edb7939763 | SimCLR <cit.> can be re-casted as a phylogenetic tree where augmentations are evolution. In the original <cit.> paper, x is an input image, and two image augmentation methods, t_1 and t_2, are sampled from a set of image augmentation methods 𝒯, to produce image augmentation v_1 and v_2, which are then passed into a trainable encoder g(·) (i.e. g_1 and g_2 share parameters entirely). In conceptualizing evolution as an augmentation strategy, x can be viewed as a common ancestor, while 𝒯 denote possible evolutionary trajectories, characterized by different evolutionary distance, mutation and genetic drift, and t_1, t_2 are two example trajectories that lead to v_1 and v_2, sampled from a set of homologs. Note that notations are adapted from the original SimCLR paper for consistency with the current work. | 2012.13475 | simclr_phylo.jpg |
0.426264 | 18468b1273d24d2aa70de0e86aeb7a4a | Maximum errors of the SOE approximations of the Gaussian kernel with the number of exponentials. Data are shown for five SOE methods: the best rational approximation, the parabolic contour, the hyperbolic contour, the modified Talbot contour and the VPMR. The dash-dotted lines with different color indicate the fitting lines of corresponding SOE methods. | 2012.13477 | Compare.jpg |
0.421408 | 709c63b18e6740f7b7da0f084a54aa5b | CPU time as function of t for different time steps h=0.05 (red square), 0.01 (green circle) and 0.005 (blue triangle). The solid lines are linear fitting of these data. | 2012.13477 | Time.jpg |
0.468258 | 95df462a63d641efb7693c2c9762f29b | Applying the effect of action on the OGMs: Left: OGM 𝐢_t and corresponding action at time t. Middle: the output of IOT1, 𝐣^ego_t+1, after applying transformation on the ego-features of 𝐢_t, . Right: the output of IOT2, 𝐣^env_t+1, after applying transformation on 𝐢_t+1. | 2012.13478 | IOTs.jpg |
0.386138 | 45fd05b4859c48b893034eb1f83539b2 | Distribution of actions. | 2012.13478 | actions.jpg |
0.410743 | 2612550f891e48d187b3d791b55d1946 | The prediction model with all of its components. The measurement estimation module updates the position and velocity of the ego-vehicle and provides transformation parameters for OGM transformer modules. IOT1 and IOT2 provide information about the first and second-order effect of action on the OGM, respectively. Prediction module is trained to minimize the cost in Eq. <ref>. | 2012.13478 | apm_train_3.jpg |
0.402026 | 6b96823413b64f018132e5246fae682e | OGM prediction using PA-DL model. The top row is the target sequence and the middle row shows the predictions. The bottom row shows the difference between two consecutive frames, learned by the model, where the red areas are erased from the frame and the green areas are added to build the next frame. | 2012.13478 | argoverse_2.jpg |
0.443187 | 9e88d8e721ab427fa2fd34a31e64d02f | Difference learning module | 2012.13478 | diff_model_2.jpg |
0.484043 | 3daa3f59e0a74d918907f80a8e8ca603 | Graphical model at time t: Left: Generative links, p(.). Right: Variational links, q(.). Observable variables are gray. | 2012.13478 | gm.jpg |
0.426473 | 1c20bde86066424e981fefd352ce87e5 | Computing the effect of actions on the future position, velocity and change in the direction of the car. | 2012.13478 | measurements.jpg |
0.417776 | 35d97ff57e0e46b8b7ebf1ba83273978 | Predictions of 10 frames (1 sec) by different models. | 2012.13478 | ngsim_1.jpg |
0.557182 | f0739d1315ef4b42b5c5fe5cf995671c | Effect of constantly applying rare actions on the prediction of PA and FM-MPUR models. | 2012.13478 | ngsim_2.jpg |
0.450266 | c70d51c0c7f44b47bdc2e5072c53218c | The prediction module at the training time. The measurements are encoded using fully-connected networks, while we use convolutional neural networks to encode and decode OGMs. | 2012.13478 | pm_high_level.jpg |
0.433415 | 79c9d540a9ae4b6faa0a4975705875ff | Detailed structure of the prediction module. | 2012.13478 | prediction_module_2.jpg |
0.453474 | b159a48505b7404d9338456d537e16ff | Flow and occupancy measured in the month of August by detector 508302. | 2012.13479 | detector_508302_august_data.jpg |
0.468074 | bc32a28be5aa4168a29fdb1290d04667 | Test predictions of DCRNN, GRU, and ARIMAX on flow measured in the month of December by detector 508306. The model was trained with a window size of 1hr and a horizon of 6 data points, or half an hour. Here we show the predictions with a horizon of 6, where DCRNN improves the most compared to the other methods. | 2012.13479 | detector_508306_baselines.jpg |
0.443011 | 4d2f7bd26f7e4cc6aff431129ff31338 | Test predictions of DCRNN on flow measured in the month of December by detector 508306. The model was trained with a window size of 1hr and a horizon of 6 data points, or half an hour. | 2012.13479 | detector_508306_predictions.jpg |
0.391397 | 3e1386f760df40f18ab0d61a8038f476 | Flow-occupancy diagram for detector 508306, one of our detectors of interest. | 2012.13479 | fundamental_diagram_508306.jpg |
0.410901 | 0c64ce7b70a7433b8fc72733a65883da | Large eddy simulation of the vortex induced vibration of a flexible riser in linearly sheared flow. Top figure, instantaneous vortical flow patterns; bottom figure, the corresponding riser displacement (red curve). Note that, the magnitude of the displacement has been enlarged in order to visualize it more clearly. | 2012.13481 | riser_VIV.jpg |
0.40951 | 9a9c929652334af3ad6d443da33a8ecd | Transmission dynamics of the Corona virus | 2012.13483 | MarkovChain.jpg |
0.450842 | 33f13b0dd7184cf99dc425fb2fc6d499 | An illustrative example of a heterogeneous contact network. | 2012.13483 | network.jpg |
0.495153 | 5e54987602384078a6481a873e4472af | An example random walk of length 6 on a network. | 2012.13483 | random_walk.jpg |
0.443237 | 15c0fedaecb0403d9968bafad3504222 | Graphical illustration of the box strategy with increment 3 shown for player number 4 and total 10 boxes. The player starts with box 4 and opens every 3^rd box till the key is found or maximum number of allowed attempts is reached. | 2012.13484 | box_strategy.jpg |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.