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
|
|---|---|---|---|---|
0.552029 | 18fd88def8224fd389bc18d97f788c1e | Polarized neutron data with 𝐐∥𝐏 at T = 290 K showing scans which span the magnetic Bragg peaks associated with the two wavevectors, 𝐤_i (i = 1, 2, | 𝐤_1 | < | 𝐤_2 |), discussed in the main text. Each panel displays the two spin-flip channels measured: +- (blue squares) and -+ (orange circles). The top row, (a)-(d), shows Bragg peaks with zone-center +𝐤_i momenta, and the bottom row, (e)-(h), shows the corresponding zone-center -𝐤_i momenta. | 2012.13010 | Fig7.jpg |
0.414693 | 1a681fb3c2924a32ad64a3aa70578b5a | Ratios of the 𝐐∥𝐏 polarized neutron spin-flip channels, I^+-/I^-+, for the (a) 𝐤_1 wavevector and (b) 𝐤_2 wavevector. Ratios from the data, taken at T = 290 K and shown as orange circles, represent the integrated intensities of fits to the data, as described in the main text. The calculated ratios, shown as blue diamonds, come from structure factor calculations for the given chiral domain populations. | 2012.13010 | Fig8.jpg |
0.428904 | 001d816aaaf44fafb7b5daa511cb54f8 | Polarized neutron data with 𝐐⊥𝐏 at 290 K for (a) (0, 0, 2+k_z,n) and (b) (0, 0, 3-k_z,n). For both (a) and (b), there is no difference between the integrated intensity of the SF and NSF data. This implies that the moments trace a circle as they spiral along the c-axis. | 2012.13010 | Fig9.jpg |
0.41757 | 50f90fd18f9045828e69c422080bdeeb | The region plots showing the distribution of the optimal QR protocol in a three-dimensional parameter space at L_ tot=1000 km for (a) three-qubit repetition code and (b) five-qubit repetition code. Other parameters are as in Fig. <ref>. In the encoded repeater case, the secret key rate is calculated for the better of decoders 3 and 4 using our numerical approximation method at N_ top = 20. | 2012.13011 | 3dplot.jpg |
0.473832 | 82963c8515e24af38efe0db53add43aa | Secret fraction r_∞^ opt of QRs encoded with three-qubit repetition code (solid lines) and five-qubit repetition code (dashed lines) for the first three nesting levels as a function of gate error probability β, with initial fidelity F_0=1 and measurement error probability δ=0. We have used our numerical approximation method at N_ top = 20. | 2012.13011 | 3qubit_vs_5qubit.jpg |
0.465344 | 4116afc7ef2d4534a12f549ec4e4a7d7 | The schematic of different decoder structures considered in this paper: (a) the original decoder proposed in <cit.>, where a decoding circuit is used to generate a qubit, on which a QKD measurement, either in Z or X basis, is performed. The decoding circuit would generate syndrome data d, which is used for state classification. (b) A modified version of the decoder proposed in <cit.>, which is very similar to (a) except that the decoder measurement outcome d is not used for classification. They still use the information in m in their key rate extraction. Note that, in both (a) and (b), Eve can control the decoder module, but has to pass some measurement data to users. (c) First alternative decoder, proposed in this work, where users directly measure the three qubits either all in Z basis or in X basis. They use majority (parity) rules, in Z (X) basis, to decode the key bit. (d) Our second alternative decoder, which is very similar to (c), except that in the Z basis a perfect match 111 (000) is mapped to bit 1 (0). In (c) and (d), we assume Alice and Bob have control over the final set of memories in their secure box. | 2012.13011 | DecoderStr.jpg |
0.41759 | f2af24d8154d431e82a728c1aedba816 | Secret fraction for different decoder (Dec) structures versus (a) gate error probability β at F_0=1 and δ=0, and (b) measurement error probability δ at F_0=1 and β=0. In the curves corresponding to perfect decoders, all error parameters assume their ideal values just in the decoder module; the corresponding value in the rest of the system is as the graph shows. | 2012.13011 | alt_decoder.jpg |
0.488893 | 57455a524a89425f9280c4f5403bbbe9 | Secret fraction r_∞^ opt versus gate error probability β for the first three nesting levels under three different approximation methods, with initial fidelity F_0=1 and measurement error probability δ=0. | 2012.13011 | comparison_of_3approx.jpg |
0.419615 | f42678e7dc6d43fe81f58228f09ff532 | Secret fraction r_∞^ opt versus (a) β, at δ = 1-F_0 = 0; (b) 1-F_0, at δ = β = 0; (c) δ, at β = 1-F_0 = 0; and (d) β, at δ =0.01 and 1-F_0 = 0. The curves labelled by good correspond to the output states where no error is detected at the ES stage, whereas the bad curves are for the output states where some errors are detected at the ES stage. The curves labeled by total are the weighted sum of good and bad terms as given by <ref> and <ref>. | 2012.13011 | goodbad.jpg |
0.370501 | a235589128fd4e1a9929013a30e6146a | Secret fraction r_∞^ opt for three-qubit repetition code, as a function of gate error probability β for different nesting levels, using our numerical approximation technique at N_ top = 20, with initial fidelity F_0=1 and measurement error probability δ=0. Here, the errors in the encoding and decoding circuits are included. The exact simulation results for the first three nesting levels are shown (solid yellow lines) as comparison. | 2012.13011 | higher_nesting_levels.jpg |
0.433561 | 5996105cff8441c5b5dd56bd5420a8ea | Normalized secret key rates for the encoded QRs with/without multiplexing for the first three nesting levels as a function of the total distance, with initial fidelity F_0=0.99, gate error probability β=0.01 and measurement error probability δ=0.005. The secret key rate is calculated for the better of decoders 3 and 4 at p=0.5, η_d = 0.9, and L_ att =22 km using our numerical approximation technique at N_ top = 20. | 2012.13011 | multiplexing.jpg |
0.439611 | 2f3e07259d3f4805b2163d28bce9b015 | Normalized secret key rates for QRs with encoding (solid, three-qubit code) and probabilistic QRs (dashed) in the absence of multiplexing for up to six nesting levels as a function of the total distance, with different error parameters: (a) F_0=0.999, β=0.0005 and δ=0.0001; (b) F_0=0.99, β=0.005 and δ=0.001; (c)F_0=0.98, β=0.02 and δ=0.01. Other parameters are as in Fig. <ref>. In the encoded repeater case, the secret key rate is calculated for the better of decoders 3 and 4 using our numerical approximation method at N_ top = 20. | 2012.13011 | probabilistic.jpg |
0.445932 | e82390e6caf744339b137d3c5a5fdf2d | The schematic of the QKD setup on a repeater chain based on the three-qubit repetition code. The small oval pairs represent bipartite entangled state prepared in advance. Using remote cnot gates, an encoded entangled state is generated across elementary links, and stored in memories represented by large ovals. The encoded entanglement is then extended across the entire link by performing entanglement swapping (ES) operations on the middle nodes. The two users will then apply decoding operation on this state to generate their raw key. | 2012.13011 | setup.jpg |
0.455777 | 67e7888ab08f40b682555c1f64f98f8d | Refining lumped element approach. The island degree of freedom φ_m is split into S subparts φ_ms with s=1… S, which are separated by inductances L_0, forming a transmission line. The lumped-element limit is valid when the internal dynamics are very fast, L_0→0. | 2012.13012 | figure_lumped_element.jpg |
0.478262 | 552409d42fe64e0887f5b1dcc7a1161c | Illustration of the main points. (a) Two SQUID detectors measure the currents to the left (I_L) and the right (I_R) in a JJ array with a total phase bias φ, through incoming wave packets (red and blue). The signal of interest is the sum of the current auto and cross-correlations, S_γ=∑_αS_αγ (here, γ=R). (b) The signal S_R depends on the detector properties (c): the lateral position x_S and distance to the circuit d_S, determining the detector fuzziness (represented as the grey bell curve). It reaches an extremal value S_R^(0) when the detector measures the current at a sharp interface. Here, charge is measured in integer units of 2e. For any S_γ/S_γ^(0)<1, the detector is fuzzy and fails to resolve integer charges. | 2012.13012 | figure_main_points.jpg |
0.483915 | 04df9933c2f849a6bb002b3c48ad9cfc | Sketch of the potential energy landscape, i.e., the Josephson energy as defined in Eq. (<ref>), as a function of δφ, where φ_m=φ_m^(f)+m/M+1δφ for m=1,…,M, and φ_0,M+1=0. The deviation δφ is chosen such that for δφ being multiple integers of 2π, we pass through other minima, that is, for instance φ_m^(f)+m/M+1(δφ+2π)=φ_m^(f+1)+m/M+1δφ. In the plot here, we chose for simplicity φ=0, and M=105. In (a) the local minima are denoted as red dots. The energies at the local minima are given by a parabola with respect to δφ (dashed black line), with the focal length [2E_J/(M+1)]^-1. In (b) the local wave functions |f⟩, centered around the local minima are depicted. Quantum phase slips, occurring with energy scale E_S, are tunneling events between different local wave functions. | 2012.13012 | figure_minima.jpg |
0.375589 | 226ef24cdb91412b954884f0c20e7b16 | Random network. ER model created using the variant G(v, p) with v = 1000 and p=0.01 probability for edge creation. | 2012.13013 | erdos.jpg |
0.480103 | cbba096a48dc4893be182380b38e4509 | Kendall rank correlation coefficient τ between GoT and Degree (Blue), GoT and Betweenness (Orange), GoT and Closeness (Green), GoT and Clustering Coefficient (Red) as a function of the Network size, in SF networks (a), SW networks (b), ER random graph (c) and as a bar chart for real networks (d). In the artificial networks, the size is between 1000 and 15000 vertices. | 2012.13013 | kendall.jpg |
0.462956 | f6b1b8e65fb7424c8be24eea2fe1827e | Pearson correlation coefficient r between GoT and Degree (Blue), GoT and Betweenness (Orange), GoT and Closeness (Green), GoT and Clustering Coefficient (Red) as a function of the Network size, in SF networks (a), SW networks (b), ER random graph (c) and as a bar chart for real networks (d). In the artificial networks, the size is between 1000 and 15000 vertices. | 2012.13013 | pearson.jpg |
0.477349 | 5367f3e1c487453089c04ee6aa6034b5 | Real networks. (a): Dolphins social network with 62 vertices (i.e. dolphins) and 159 edges (i.e. frequent associations). (b): High Energy theory collaborations network with 8,361 vertices (i.e. scientists) and 15,751 edges (i.e. connections existing if the scientists have authored a paper together). (c): Internet network with 22,963 vertices (i.e. autonomous systems - AS) and 48,436 edges (i.e. connections between AS). | 2012.13013 | reals.jpg |
0.349 | d8121acd0c5944ef9d618949b3371e8c | Scale-free network. Extended BA model by Holme and Kim with v = 1000 vertices, e = 5 random edges to add for each new vertex and p = 0.3 probability of adding a triangle after adding a new edge. | 2012.13013 | scale.jpg |
0.358522 | 34aa9e4fcd164e4a9fd7eaefb956aa8c | Small-world network. NWS small-world network model with v=1000 vertices, each of which is joined with its k=6 nearest neighbors in the ring topology, and p=0.6 probability of rewiring each edge. | 2012.13013 | small.jpg |
0.492157 | 2d9ed8f6908f491fad9017912bd315b7 | Spearman rank correlation coefficient ρ between GoT and Degree (Blue), GoT and Betweenness (Orange), GoT and Closeness (Green), GoT and Clustering Coefficient (Red) as a function of the Network size, in SF networks (a), SW networks (b), ER random graph (c) and as a bar chart for real networks (d). In the artificial networks, the size is between 1000 and 15000 vertices. | 2012.13013 | spearman.jpg |
0.437463 | 847510ad906c40f6bcf2caf6af659b43 | Estrada dos Tropeiros. | 2012.13014 | alphavile-litoral-norte0.jpg |
0.395969 | a4f2de34aeda4b94858016065715a2b1 | Estrada dos Tropeiros. | 2012.13014 | alphavile-litoral-norte1.jpg |
0.38207 | 7c522b937f4a47a0b136c5eb5bad8483 | Mining environment | 2012.13014 | ambiente-fora-de-estrada_mineracao.jpg |
0.478014 | 6600481299244eca8a1746367efa975c | Dilated filter with different expansion rate. Adapted from <cit.> | 2012.13014 | atrous.jpg |
0.443652 | 6070f78085f64f848a64806a765313ac | Detection and bounding box. | 2012.13014 | bbox.jpg |
0.386016 | 3195e66ddaa8437a98fb13b3cc8b8991 | The vehicle used for data acquisition and validation of the proposed system. | 2012.13014 | carro-plataforma.jpg |
0.419825 | 42d11168d35d482089f139abfda41473 | Test track at night. | 2012.13014 | ci-noite-pessoa-carro.jpg |
0.426882 | 8fc0db5b95264eb9821413b95588e9ae | Track dusty/night. | 2012.13014 | ci-noite-poeira.jpg |
0.393054 | c28f1cdf229a424091fb296fb8b298b3 | Track with dusty. | 2012.13014 | ci-poeira2.jpg |
0.395864 | 61f8428d7a994c839a74786096f4ddfa | Daytime | 2012.13014 | daytime-offroad-CM2-img.jpg |
0.444206 | 7916b66491b141cbb6a96c756ce96f2f | DeepLab | 2012.13014 | deeplab_xc71_comp_0.jpg |
0.434831 | f04dac20b2434374b33e367a254765e7 | Praia do Forte. | 2012.13014 | estrada-do-castelo-0.jpg |
0.47675 | f3eb837e2ce34083beec5cd91ccab7af | Praia do Forte. | 2012.13014 | estrada-do-castelo-1.jpg |
0.401801 | 6340c65b60804bd7abc65ed7aa87d009 | P. do Forte raining. | 2012.13014 | estrada-do-castelo-chovendo.jpg |
0.442462 | 0643de3f4ab648babe3720e6dafd7997 | Tropeiros raining. | 2012.13014 | estrada-dos-tropeiros.jpg |
0.356831 | b2abaf98169e4cbf90309be92c9cd64d | Bumpy intercity road | 2012.13014 | estrada_esburacada_bahia_foto_mauro_akin_nassor_correio.jpg |
0.467657 | c93f57f5ad424e8c84b250110ffe8ae6 | Fog. | 2012.13014 | fog_img_aug.jpg |
0.398966 | d753e60cc4d24fe08baa982d548a3e30 | Panoptic Ground-truth. | 2012.13014 | ground-truth2.jpg |
0.443463 | 38780ea7dde14228bd08e4b4eaba219d | Image | 2012.13014 | image_comp_0.jpg |
0.436342 | c32f2a64dd3b4af990e54900740a386b | Original image. | 2012.13014 | img.jpg |
0.50583 | bb60a6f949eb46e18d65605894bcb5c5 | Instance segmentation. | 2012.13014 | instancy.jpg |
0.424807 | 9d8dacebf3e74a8884e5ed30e4214be7 | Jauá. | 2012.13014 | jaua-0.jpg |
0.42303 | 440ca8ea4023403d84ff8926e3002920 | Jauá. | 2012.13014 | jaua-1.jpg |
0.489467 | 78dce8e8a54947b7aa71b3f7e1c850d5 | Unpaved road in Brazil | 2012.13014 | jaua-camacari-rm-salvador.jpg |
0.524957 | e0f6184ee63e4ad9a07044ce175c9226 | Jauá raining. | 2012.13014 | jaua-chovendo.jpg |
0.405721 | 3db769d47ee449308a02c3a444f21e1d | Expected | 2012.13014 | label_comp_0.jpg |
0.491115 | fe3795341cbd436a83f98fde944148a9 | Map of the off-road test track. | 2012.13014 | mapa-da-pista-construida.jpg |
0.508286 | 97514d2c828f4911a8c09c875dc3d6bd | Sensor layout and operating region. | 2012.13014 | montagem-de-sensores.jpg |
0.40751 | e00827c3f5aa4cfaa0613932e0b7faa4 | Night-dusty | 2012.13014 | night-dusty-offroad-CM2-img.jpg |
0.406419 | 9267513d3add4a4ea8b9544629c44dea | Night | 2012.13014 | night-offroad-CM2-img.jpg |
0.424014 | 751f933943b84e9693fe436ed564721d | Original image. | 2012.13014 | noise_crop_img.jpg |
0.430864 | bbd0f3dc8fb446dc80a5ff053d087126 | Noise, rotating and crop. | 2012.13014 | noise_crop_img_aug.jpg |
0.428512 | d7ee5f28bdd34b758671f5a1dd6bb670 | Image, segmentation mask. | 2012.13014 | noise_crop_img_lbl.jpg |
0.467833 | 88982d5646d54b1e9319815c67b6880a | Panoptic Segmentation. | 2012.13014 | panotic.jpg |
0.419139 | 56fec671e0c44dbca454b66141881a01 | Slopes and open space. | 2012.13014 | pista-espaco-talude.jpg |
0.455963 | 8655e15b78344bb392a129d9e4941992 | Pickets. | 2012.13014 | pista-piquete-piquete.jpg |
0.422043 | d9a890b1ebe2479cbc4b4ea053945329 | Pickets and slopes. | 2012.13014 | pista-piquete-talude2.jpg |
0.413122 | 8625d80b762d46e087def62ad46cf3f2 | Dusty | 2012.13014 | pista-poeira0.jpg |
0.385774 | 4c16ad6bf57c40489747da60312e3555 | Slopes. | 2012.13014 | pista-talude-talude.jpg |
0.426818 | 4d9a767bf7c247e19ed3def8c2a2a43a | PSPNet | 2012.13014 | pspnet101_comp_0.jpg |
0.47014 | 6047afd3709d4ca78ac10b661ba814a7 | Semantic segmentation. | 2012.13014 | semantic.jpg |
0.43606 | 93a649cc2d684a8b93f78c5db550eb15 | Perception paradigms. Adapted from <cit.> | 2012.13014 | three-paradigms-for-autonomous-driving.jpg |
0.412031 | f016120fd40040a785d211e23c757bef | Road poorly signposted | 2012.13014 | via-mal-sinalizada-lauro-de-freitas.jpg |
0.418505 | 152ec75cf51842a2acbfd3b23e270eca | Evolution in time of 50 trajectories of the cost function ϕ(z(t)) generated by the 50 solutions x shown in Figure <ref>. The solid black line indicates the upper bound T^* on the convergence time. | 2012.13015 | CostRandom.jpg |
0.417959 | 1df96522fcb74354b9e63568f800bc70 | Evolution in time of several solutions x initialized in a neighborhood of the optimizers. The solid black line indicates the upper bound T^* on the convergence time. | 2012.13015 | Random.jpg |
0.461262 | 531acf34940747d18c64790e6334784a | Scheme of Newton-like Fixed-Time Extremum Seeking Dynamics for static maps. | 2012.13015 | SchemeNewton.jpg |
0.400602 | 46a65d1390e64e839cf5c6b8c265c76d | Evolution in time of the state x. Blue line corresponds to x_1, and red line corresponds to x_2. The dotted lines correspond to the trajectories generated by the traditional Newton-based ESC of <cit.>. The solid lines correspond to the trajectories generated by the NFxTES. | 2012.13015 | x.jpg |
0.413621 | a9871e528fc84b21b900bd20c3e475c9 | Weakly-supervised shadow image decomposition. With only shadow mask supervision, our method automatically learns to decompose the shadow effect in the input image patch I^sd into a matte layer α and a relit image I^relit. The matte layer α combines I^sd and I^relit to obtain a shadow-free image patch I^output via Eq. (<ref>). | 2012.13018 | 10-1_png_007_014.jpg |
0.484854 | 1a4a99bc32fe46cf9eb464870465c070 | Shadow Removal via Shadow Image Decomposition. Our framework progressively removes shadows from the input image. SP-Net removes shadows in the umbra shadow areas via a physical illumination model of shadow. M-Net models the smooth shadow effects on the penumbra shadow areas via a matte layer. I-Net inpaints the missing details on the umbra shadow areas, given the output from SP-Net and M-Net. | 2012.13018 | 116-6.jpg |
0.436082 | 8956390684e842d9b3056b5c950df9a4 | The penumbra area of the shadow. We define two areas alongside the shadow boundary, denoted as _in (shown in green) and _out (shown in red). These two areas roughly define a small region surrounding the shadow boundary, which can be considered as the penumbra area of the shadow. | 2012.13018 | 133-7.jpg |
0.446679 | e76a66c9e94f419ea62f80bbc038c37d | Shadow removal results of our model at different stages. From left to right: input shadow images, shadow-free images obtained using SP-Net with shadow masks from <cit.>, shadow-free images generated by SP-Net and M-Net, and shadow-free image generated by our full system with SP-Net, M-Net, and I-Net. The bottom row visualizes the differences between each image and the ground truth shadow-free image. | 2012.13018 | 134-14.jpg |
0.405934 | 8bf962a21d21450797faa5ab090644ee | An example of our color correction method. From left to right: input shadow image, provided shadow-free ground truth image (GT) from ISTD dataset, and the GT image corrected by our method. Comparing to the input shadow image on the non-shadow area only, the mean-absolute distance of the original GT is 12.9. This value on our corrected GT becomes 2.9. | 2012.13018 | fix_dataset.jpg |
0.433846 | 14af03a3f6fe4995a14b64ebf1645bb0 | A comparison of the ground truth shadow mask and our shadow matte. From left to right: The input image, the relit image computed from the parameters estimated via SP-Net, the ground truth shadow mask, the final results when we use the shadow mask, the shadow matte computed using our M-Net, and the final shadow-free image when we use the shadow matte to combine the input and relit image. The matting layer handles the soft shadow and does not generate visible boundaries in the final result. (Please view in magnification on a digital device to see the difference more clearly.) | 2012.13018 | matte.jpg |
0.431476 | 8c76446d003143bf9d1b016e75aeae70 | Shadow Removal Framework. The shadow parameter estimator network SP-Net takes as input the shadow image and the shadow mask to predict the shadow parameters (w,b). The relit image I^relit is then computed via Eq. <ref> using the estimated parameters from SP-Net. The relit image, together with the input shadow image and the shadow mask are then input into the shadow matte prediction network M-Net to get the shadow matte layer α. The system outputs the shadow-free image via Eq. <ref>, using the shadow image, the relit image, and the shadow matte. SP-Net learns to predict the shadow parameters (w,b), denoted as the regression loss. M-Net learns to minimize the L_1 distance between the output of the system and the shadow-free image (reconstruction loss). I-Net takes as input the shadow image, shadow mask, and the shadow-free image computed from SP-Net and M-Net to output a residual image that is used to obtain the final shadow-free image. | 2012.13018 | pami_framework.jpg |
0.45009 | 7075952743f34123ab38d8a0c37bbc50 | Weakly-supervised shadow decomposition. We train our SP-Net and M-Net together with an adversarial network, D-Net, using only two sets of unpaired shadow and shadow free patches. These patches are cropped directly from the shadow images (left panel). SP-Net and M-Net predict the shadow parameters (w,b) and the matte layer α respectively to jointly remove the shadow. D-Net is the critic function distinguishing between the generated image patches and the real shadow-free patches. The only supervision signal is the set of shadow-free patches, which can be easily obtained using the shadow masks. The four losses guiding this training are the matting loss, smoothness loss, boundary loss, and adversarial loss. | 2012.13018 | pami_wframework.jpg |
0.42586 | f219c8477c2c40309fdba9f363e7c7a1 | Shadow Removal via Shadow Image Decomposition. A shadow-free image I_shadow-free can be expressed in terms of a shadow image I_shadow, a relit image I_relit and a shadow matte α. The relit image is a linear transformation of the shadow image. The two unknown factors of this system are the shadow parameters (w,b) and the shadow matte layer α. We use two deep networks to estimate these two unknown factors. | 2012.13018 | teaser.jpg |
0.429865 | d2429e8d05c646278c5166b0722fc3be | Logarithmic plot of the time-domain evolution of scalar perturbations γ=0.5, k=0, ℓ=0 (left); r_+=1. | 2012.13020 | 5over10.jpg |
0.421681 | 8a9bba8fc51b49bf8332d858c674742f | Semilogarithmic plot of the time-domain evolution of scalar perturbations γ=0.2, k=10^-4, ℓ=1; r_+=1. | 2012.13020 | Scalar2over10R100.jpg |
0.471453 | ee68291c039c4fc49a6852730dca1294 | The classification error η as a function of the number of representative vectors Q per class. | 2012.13021 | fig1.jpg |
0.469486 | 2157f324af3e48abbed1d6bf033716b0 | The classification error η as a function of the number of representative patch vectors Q per class. | 2012.13021 | fig2.jpg |
0.406691 | ccc1148dde0b4720ac1bf11d18f42e99 | Evolution in time of the actions of the players under the FxTNES dynamics (solid lines) and under the gradient-based NES dynamics considered in <cit.> (dotted line). | 2012.13022 | Fig.jpg |
0.422005 | fd18fb8718f840ec83afe0e5e8ef6ab8 | Comparison between the reachable sets in the time interval [0,6] of the NES dynamics of <cit.> and the FxTNES, with û(0)∈[-15,15]×[-15,15]×[-15,15]. | 2012.13022 | Reachable.jpg |
0.394042 | 2b453ad94c2f43218f5afa69f77d8a95 | Model architecture | 2012.13024 | arch.jpg |
0.422312 | c184d5dd269241e29901060e729a38e7 | 2012.13024 | celeba_prior_Heavy_Makeup_0.01.jpg |
|
0.424853 | c16342fe6422438089ea11689e8a4084 | 2012.13024 | celeba_prior_Male_0.01.jpg |
|
0.430417 | 27e3c7af68bb46ecaea1bef463218bc2 | 2012.13024 | celeba_test_Heavy_Makeup_0.01.jpg |
|
0.418382 | aeff82349a2044f5982cfedd02685b19 | 2012.13024 | celeba_test_Male_0.01.jpg |
|
0.419084 | 4f75d8d4c4984ddbbf882a6344af5a59 | 2012.13024 | celeba_test_gt.jpg |
|
0.479001 | 7d33d1540b8946e19bef65e26e5b295d | 2012.13024 | fmnist_test_cross_0.002.jpg |
|
0.411944 | f61a0253368245889f4c165ccc995558 | Full-modality | 2012.13024 | gr1.jpg |
0.581608 | fc9f7aba25574056907ccf6a69441e53 | Missing-modality | 2012.13024 | gr2.jpg |
0.45938 | 5d114f9acb6f4ae9bfceb91249a82ca2 | Bimodal data | 2012.13024 | intro1.jpg |
0.436893 | 991df5e6b5144a3ab1e6bb6bf6ee9c01 | Private and shared spaces used by DMVAE | 2012.13024 | intro2.jpg |
0.356481 | d3e8d74acbb74797a701d2aa46cbf641 | 2012.13024 | mnist_cross_0.002.jpg |
|
0.418851 | 1742c473333a4978a7cf39dd96f3bb3d | T = 0.01 | 2012.13024 | poe1.jpg |
0.422551 | c810455afbf34b769458780eaa39f994 | T = 0.66 | 2012.13024 | poe2.jpg |
0.397943 | 1e7fab08e0a94c5481fb0aef0227736a | T = 1 | 2012.13024 | poe3.jpg |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.