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
|
|---|---|---|---|---|
0.524996 | c5a461cfcc824634a23831b00e207ec7 | annotator 3 | 2012.13052 | random_anno_3.jpg |
0.487083 | ec166cf53bd3432595113e2e4f4caddc | common noise | 2012.13052 | random_com.jpg |
0.484226 | 58bd14718d7748d8a5223bc3074d9917 | Training & testing accuracy plots of various approaches on Synthetic dataset with asymmetric noise when common noise strength is 0.7 and average proportion of common noise is 0.6. | 2012.13052 | syn_training_curve.jpg |
0.44191 | 75bae9092fcf471ebc57ff365b2b5c78 | Broadcasting and Receiving Tokens | 2012.13053 | Broadcast_Receive_Tokens.jpg |
0.456139 | a74e14b34db249109ae50633cef07d62 | Querying Risk Score | 2012.13053 | Querying_Risk_Score.jpg |
0.425288 | bbfe657abc1c4ad8ac46aed708a40144 | Remote Attestation Overview | 2012.13053 | Remote_Attestation_All_Components.jpg |
0.451493 | fcd9759cb0fa442fadfe772509497f86 | Reporting Infection | 2012.13053 | Reporting_Infection.jpg |
0.452296 | 1779d8feeb174e38aacfe775e8c47ea4 | The SDIRK method with successive corrections applied to the van der Pol system (<ref>). The half-precision implementation (<ref>)) shown as a dashed line. The mixed precision implementation (<ref>)) with = O(10^-4) is shown with a dotted line. The mixed precision implementation (<ref>) with = O(10^-4) and one (p=2) correction steps is shown with dash-dot, and with two (p=3) correction steps with a solid line. The double precision implementation of (<ref>) is shown for reference with black dots. | 2012.13055 | 2s3pDirkSweepComparisonHalfPrecision.jpg |
0.488113 | f6212b37e50d4184b8fb75389db8b3c7 | Methods 4s3pC and 4s3pA for time evolution of the diffusion equation. The magenta dashed line is Method 4s3pC with F resulting from a Fourier spectral method approximation to u_xx, and resulting from a second order centered difference approximation to u_xx. The dashed blue line is Method 4s3pC with F resulting from a double precision implementation of a Fourier spectral method approximation to u_xx, and resulting from a half precision "chopping" of the Fourier spectral method approximation. The dotted blue line is Method 4s3pA for this same scenario. The double precision implementation with a Fourier spectral method approximation is shown for reference with black circles. | 2012.13055 | ComparingType1andType2TauConditions4s3p3m3nMethod.jpg |
0.444696 | 4e714fd4da6a4f81af50d81a67fe8ac8 | The implicit midpoint rule applied to the van der Pol system (<ref>). The half-precision implementation (= O(10^-4)) is shown as a blue dashed line and the single-precision implementation = O(10^-8) as a red dashed line. The mixed precision implementation with = O(10^-4) is shown with a dotted blue line, and the mixed precision implementation with = O(10^-8) is shown with a dotted red line. The double precision implementation is shown for reference with black circles. | 2012.13055 | IMR_VDP_MP_Comp.jpg |
0.538425 | 76a2e0d1e1b44a86b670ece6983a2b45 | Evolution of search tree T from left to right in a ternary graph with green, yellow, and red classes. builds the optimal tree T by revealing the classes of the frontier edges in the order defined by the priority queue. | 2012.13057 | lemma_fig.jpg |
0.41212 | e6e1e607f82d479ba52dc1b5799218ab | 2012.13057 | range_300.jpg |
|
0.454177 | 1cdd829c9255473b825560797eb2656f | 2012.13057 | range_inf.jpg |
|
0.429061 | d9644fca028f4b55be8b945a3c222da9 | 2012.13057 | range_nconv.jpg |
|
0.40934 | 65f4f2f81b034b478debcc844689bc46 | 2012.13057 | range_zero.jpg |
|
0.444223 | 237cfcdfd46848aa96d233abcbc79797 | A broad classifications based on technologies, architectures and privacy threats | 2012.13061 | fig2.jpg |
0.397811 | 957a7b11b9ef47b19d3ca1eaea7e1215 | The interface of 'The Shield' application | 2012.13061 | shield.jpg |
0.532495 | dc4295975e68460aa85080412592c6a5 | In the grid the boughs are self-similar. | 2012.13070 | grid.jpg |
0.436958 | 6b9712013e2442be961efe8477692f91 | Maps of reconstructed error of Jasper Ridge dataset. From left to right, top to bottom: FCLS, SUnSAL-TV, CsUnL0, SCHU, Pro-H-NLM, Pro-H-BM3D, Pro-H-BM4D, Pro-H-LRTDTV, Pro-A-NLM, Pro-A-BM3D, Pro-A-BM4D and Pro-A-DnCNN. | 2012.13074 | jas_RE_map.jpg |
0.423192 | 44580413600749dea1374754b3c44d97 | RMSE as a function of the regularization parameters for the Pro-H-NLM with synthetic data (SNR = 20 dB). | 2012.13074 | loss_curve.jpg |
0.471773 | 438cfecf3f3e481d9ba2ec01e59b4a9e | Abundance maps of synthetic data (SNR=10dB). From top to bottom: different endmembers. From left to right: ground-truth, FCLS, SUnSAL-TV, CsUnL0, SCHU, Pro-A-NLM, Pro-A-BM3D, Pro-A-BM4D, and Pro-A-DnCNN. | 2012.13074 | map_abu_10db_new.jpg |
0.427873 | 5767bdb4b5e74cc0bbc1c65a251d39be | Abundance maps of Jasper Ridge dataset. From top to bottom: four endmembers, tree, water, soil, and road. From left to right: FCLS, SUnSAL-TV, CsUnL0, SCHU, Pro-H-NLM, Pro-H-BM3D, Pro-H-BM4D, Pro-H-LRTDTV, Pro-A-NLM, Pro-A-BM3D, Pro-A-BM4D, and Pro-A-DnCNN. | 2012.13074 | map_jas_2.jpg |
0.460851 | aa77d971b6714794868a10e4e0c2fa4a | Abundance maps of synthetic data (SNR=5dB). From top to bottom: different endmembers. From left to right: ground-truth, FCLS, SUnSAL-TV, CsUnL0, SCHU, Pro-H-NLM, Pro-H-BM3D, Pro-H-BM4D, and Pro-H-LRTDTV. | 2012.13074 | map_syn_abu_5db.jpg |
0.579072 | 7c0c3b61a2da4de7be3f1b866422dc9a | The reconstructed image maps. From top to bottom: reconstructed images in different channels (50, 100, 150, 200). From left to right: clean image, noisy image, reconstructed images of Pro-H-NLM, Pro-H-BM3D, Pro-H-BM4D, Pro-H-LRTDTV, Pro-A-NLM, Pro-A-BM3D, Pro-A-BM4D, and Pro-A-DnCNN. | 2012.13074 | map_syn_re_image_new.jpg |
0.451705 | ef0a6945733247b489ed864ade626908 | Abundance maps of Urban dataset. From top to bottom: six endmembers, asphalt, grass, tree, roof, metal, and dirt. From left to right: FCLS, SUnSAL-TV, CsUnL0, SCHU, Pro-H-NLM, Pro-H-BM3D, Pro-H-BM4D, Pro-H-LRTDTV, Pro-A-NLM, Pro-A-BM3D, Pro-A-BM4D, and Pro-A-DnCNN. | 2012.13074 | map_urban_2.jpg |
0.419905 | a36071dc174642509c2328e031d9f022 | Color image of real dataset, Jasper Ridge dataset (left), Urban dataset (right). | 2012.13074 | realdata.jpg |
0.402959 | 6927055117424ae3ad4b5fdf902a8269 | Maps of reconstructed error of Urban dataset. From left to right, top to bottom: FCLS, SUnSAL-TV, CsUnL0, SCHU, Pro-H-NLM, Pro-H-BM3D, Pro-H-BM4D, Pro-H-LRTDTV, Pro-A-NLM, Pro-A-BM3D, Pro-A-BM4D and Pro-A-DnCNN. | 2012.13074 | urban_RE_map.jpg |
0.424606 | 8717e559f8fc4f3c9137b549e077bd6f | The 3-D scatter plot shows the RCDs' largest eigenvalues varies dependent on age and nose width. The Z-axis is for the largest eigenvalues; The X-axis and Y-axis are for the normalized ages and normalized nose widths, respectively. The points represent the observations. The fitted surface is fitted by a polynomial regression model. | 2012.13075 | Motivating2.jpg |
0.418511 | 36eca114fd864fb7a311a9ad98a93a95 | The side-by-side boxplots of the Rand indices are displayed. The title of each subplot gives the value of u. Under a certain u, the x-axis is for the values of λ and the y-axis is for the Rand index. The Rand indices produced by a method of all replications are visualized by a boxplot. For each method, the medians of the boxplots are connected by a dashed line. The legends on the right side give the colors for each method. To clarify, the benchmark methods do not rely on λ and u but are only visualized at the locations to compare with the corresponding Rand indices produced by our algorithm. | 2012.13075 | comparison_phi_1.jpg |
0.403209 | 9f00d17e23044856ae2c863d4fcca09c | The exponential component in the weight function is displayed. | 2012.13075 | exp.jpg |
0.435605 | cc72fe6b88fa46218d2e46c4bff606be | We use scatter plot to show the relationship between the MSE and the log-likelihood. The title of each subplot gives the value u. The x-axis is for the values of MSE and the y-axis is for the values of log-likelihood. The legends on the right side give the colors for each parameter. | 2012.13075 | para_phi_1.jpg |
0.397954 | 5bef8558a2cf4458967a123c2bf659b1 | The side-by-side dot-plots of the Rand indices are displayed. The title of each subplot gives the value of u. Under a certain u, the x-axis is for the values of λ and the y-axis is for the Rand index. The Rand index produced by a method is visualized by a dot. For each method, the dots are connected by a dashed line. The legends on the right side give the colors for each method. To clarify, the benchmark methods do not rely on λ and u but are only visualized at the locations to compare with the corresponding Rand index produced by our algorithm. | 2012.13075 | real_data_comparison.jpg |
0.529132 | 7ad29a27ffd64cce825fe98a1e0e7d9b | The side-by-side boxplots of the Rand indices are displayed. The x-axis is for the values of u and the y-axis is for the Rand index. Given a u, the rand indices are aggregated over all λ∈{5× 2^-5, 5× 2^-4, 5× 2^-3,5× 2^-2,5× 2^-1,5× 2^0}. The aggregated Rand indices are visualized by a boxplot. The medians of the boxplots are connected by a dashed line. | 2012.13075 | u_phi_1.jpg |
0.428331 | 1bd0a8da84cb4d15bcd3c46409969ad0 | Three different strategies for λ, which controls the mixing ratio of second-order and first-order methods. | 2012.13077 | figure2.jpg |
0.451152 | 4cd71149d4e54411be3110ea0206b5a9 | Schematic representation of RE-SiamNet typically designed for object tracking. On the template head, multiple equidistant rotated variants of the original template image are used. | 2012.13078 | fin_main_fig.jpg |
0.385585 | cdbfe33aa01a4265b08e3aa1384ead9d | Example frames from 3 sequences of ROB dataset showing the ground truth bounding box (blue), and predictions obtained using SiamFC<cit.>(green) and RE-SiamFC using 8 rotation groups (red). Further, blue and red arrows show the ground truth pose estimate and the prediction obtained using RE-SiamFC, respectively. | 2012.13078 | grid_figure.jpg |
0.441381 | 62c2aee5f960470dba6ac661a745247e | Example demonstrating rotation non-equivariance in regular CNN models used in object tracking, ψ_θ(f(·)) ≠ f(ψ_θ(·)). Here f(·) and ψ_θ(·) denote the neural network encoding function and rotation transform, respectively. | 2012.13078 | img_cnn_noneq.jpg |
0.501599 | 4562c57dc36d44bbb2605d01185444db | Some graphs used in the paper. | 2012.13079 | fig1.jpg |
0.419559 | 01a49e0eae144c2ca9f5818a737e82eb | The graph XY(x,y;n) | 2012.13079 | fig10.jpg |
0.546066 | 75a2f49394b94bf2b6f9142d4a523999 | From left to right, an open quipu, a closed quipu and a dagger | 2012.13079 | fig2.jpg |
0.505127 | e3b3dae026dc42d390ce41f7c3ad606a | The mixed graphs X_i for 1 ⩽ i ⩽ 10 and Y. | 2012.13079 | fig3.jpg |
0.494388 | 53dbfe8b08434be6bb9f80a427d41684 | Maximal signed graphs in 𝒢_𝒮^⩽ 2. Negative edges are depicted by dashed lines. | 2012.13079 | fig4.jpg |
0.504723 | fa903edf006a43fcb53f4f8e4a435a95 | Maximal connected oriented graphs whose skew spectral radius does not exceed 2. | 2012.13079 | fig5.jpg |
0.399747 | 72a9f683b1bd4a9a9116ea2b1b647f8b | Examples of hypergraphs of type E_i,j,k^(3), F_i,j,k^(3) and G^(3)_i,j:k:l,t. | 2012.13079 | fig6.jpg |
0.454641 | 30c0f6b54c084a25bc13d359b82d634a | Some hypergraphs involved in Theorems <ref> and <ref>. | 2012.13079 | fig7.jpg |
0.46396 | a3ea7c7087554a1a9ce24a668c66851d | Some hypergraphs involved in Theorems <ref> and <ref>. | 2012.13079 | fig8.jpg |
0.478862 | d080243e04a14618a266838f5f2896f9 | The two types of internal path | 2012.13079 | fig9.jpg |
0.402696 | 78955ce5972d42c696147d7c399561c5 | Matching Experimental FFDPs with interferometer inferred FFDPs. | 2012.13081 | 4DFFDPComparisonNew.jpg |
0.415902 | 99f299986e124c2b95d7bdf150ed197b | Illustration of the real edge's azimuth orientations with corresponding dihedral angle offsets in scenarios of REE: real-edge pointing East (left), and REN: real-edge pointing North (right). | 2012.13081 | Fig1_RealEdgePointings_modified.jpg |
0.455743 | 1cb8fda676214d0898841a85acf7e806 | Adjusted surface angles in scenarios of REE: real-edge pointing East (left), and REN: real-edge pointing North (right). | 2012.13081 | Fig2_RealEdgePointings_adjust.jpg |
0.417519 | 0dc3e5eb01d44e1982ffd33487e3dd86 | Estimation of LLOs' photon return ratio (in relative scales) with the modified retro-reflector with REE. | 2012.13081 | Fig7_Observtories_REE.jpg |
0.460821 | 092b52825275466c9c8b592c15134cf6 | Estimation of LLOs' photon return ratio (in relative scales) with the modified retro-reflector with REN. | 2012.13081 | Fig8_Observtories_REN1.jpg |
0.488728 | 8d9a6cec657d4a39844292fa7751734e | Offset angle optimization based on averaged photon returns among 7 major LLOs. | 2012.13081 | Fig9_Optimization.jpg |
0.442565 | 2dc8ee831d2e4ebd80372e2ad2ea23a0 | Interferometer measured PEMs with SP and DP configurations. | 2012.13081 | Fig_13.jpg |
0.418245 | f8311974fa48440fa386dcd07e0c1540 | Far field diffraction pattern of the 100 mm wide dielectric retro-reflector with its 2D projected real edge pointing East and incident plane wave East-West linearly polarized. | 2012.13081 | Fig_4_Paper1.jpg |
0.406649 | bd81665fc2204460805a4b0995d5e549 | Far field diffraction pattern of the modified dielectric retro-reflector with REE orientation, modification angle ΔΦ_REE=0.5 arcsecond and East-West linear polarized input. | 2012.13081 | Fig_5_Paper1.jpg |
0.430667 | 2c7d458e58e74de392c21ecc27e3803c | Far field diffraction pattern from a modified dielectric retro-reflector with REN orientation, modification angle ΔΦ_REN=0.5 arcsecond and East-West linear polarized input. | 2012.13081 | Fig_6_Paper1.jpg |
0.438465 | a5fcfdb60cc244258cc6f81d3ed227a5 | Footprints of traced rays with different TIR sequences. | 2012.13081 | Figure3_sector_graph_reflection_footprint.jpg |
0.468069 | 81fa7a900c9e448a9f8406410e5f4ea7 | Preliminary assembled view of the NGLR: UMD-02. | 2012.13081 | IMG-2146.jpg |
0.412596 | 4c1e5f987e7b4efba3080395c2eec20f | Evaluation of UMD-03 NGLR with East-West linear polarization input and 1064 nm wavelength for the ranging laser-pulses. | 2012.13081 | RBAO_returns_1064.jpg |
0.478343 | bf7075a6d4e34ddbb50e7f1f7517e3f9 | Evaluation of the manufactured UMD-03 NGLR with REN orientation in deployment. | 2012.13081 | RBAO_returns_532.jpg |
0.428775 | 4c8d2579449c4bcba110c06a88f52877 | Evaluation of the manufactured UMD-02 NGLR with hypothetical deployment. | 2012.13081 | UMD_02Evaluation.jpg |
0.464738 | ab78b2d7a2744490a28db98da3d33af8 | Simulation results for rate-1/3 PIC-TCs and PPC-TCs on the AWGN channel. | 2012.13082 | BER_AWGN_1.jpg |
0.478852 | 0830b8e263f9482b919d0d0a6d67cef4 | Bit erasure rate of several PIC-TCs, PPC-TCs and shortened TCs with different code rates. | 2012.13082 | PIC_PPC_BER_new_v2.jpg |
0.378575 | 4d3aa2783fd34b4e8d99c71c5dcdedf3 | Bit erasure rate of PIC-TCs, PPC-TCs with random parity puncturing. | 2012.13082 | PIC_PPC_punc_BER_1.jpg |
0.429177 | a4fcad58507042059bfffd773b79f6c0 | Bit erasure rate of PPC-TCs for rate-1/10. | 2012.13082 | PPC_BER_low_rate.jpg |
0.497013 | d07d836b9d9c42aca6f8383f6cf64852 | Thresholds versus the coupling memory with λ= 1/2. | 2012.13082 | cpm_threshold_0p2.jpg |
0.48412 | 9eaca9c5609e402cac62c235e9b1c621 | Thresholds versus the coupling memory with λ = 1/8. | 2012.13082 | cpm_threshold_0p3034.jpg |
0.439922 | 17246f4c0417467c81088280f18b0dac | An illustration of FF-FB decoding. | 2012.13082 | decoder1.jpg |
0.488144 | 6480e27fd1874f6b9cc143d192db3084 | Block diagram of the encoder of a PIC-TC with m=1. | 2012.13082 | encoder_v3.jpg |
0.517425 | bfb412cb909c44188d2be3704698005b | Compact graph representation of (a) PPC-TCs with coupling memory m = 1 from time instant t-1 to t+1, and (b) PPC-TCs of coupling memory m≥1 for time instant t. | 2012.13082 | graph_2.jpg |
0.491329 | b6ef6a94fd1a446ebaa16e87ba268bc4 | Compact graph representation of (a) PIC-TCs with coupling memory m = 1 from time instant t-1 to t+1, and (b) PIC-TCs of coupling memory m≥1 for time instant t. | 2012.13082 | graph_3.jpg |
0.400839 | 61062802b6914c15b432ac9e9a3b337e | Bit erasure rate of PIC-TCs and PPC-TCs for rate-0.2 under sliding window decoding. | 2012.13082 | window_decoding_BER1a.jpg |
0.388054 | 2f6099dc71ce4358b05ce99d312961f2 | Bit erasure rate of PIC-TCs and PPC-TCs for rate-0.3043 under sliding window decoding. | 2012.13082 | window_decoding_BER2a.jpg |
0.434619 | 228116f6683c44a5b0699aefda544c1e | Same as Figure <ref> but for the South Ecliptic Pole tile (0890m667). | 2012.13084 | 0890m667.jpg |
0.437802 | 385437b121ba4d49bf547e92c69d2e61 | CatWISE Preliminary Catalog (blue) and CatWISE2020 Catalog (red) astrometric performance with respect to Gaia DR2 in the COSMOS tile (1497p015). The top row shows the 1-σ dispersion between CatWISE and Gaia R.A. (specifically, Δα cos(δ)) and Dec. (left), and proper motion (right). The bottom row shows the median χ^2 computed taking into account CatWISE catalog uncertainties, Gaia catalog uncertainties, and the uncertainty introduced by the translation of Gaia's positions to the CatWISE epoch.) | 2012.13084 | 1497p015.jpg |
0.466679 | 3a26fdc66bf14fe7801e19e29387e133 | Comparison of CatWISE Preliminary Catalog (left) and CatWISE2020 Catalog (right) photometry to Spitzer photometry for COSMOS. Top: Difference between CatWISE W1 PSF and Spitzer S-COSMOS 29 radius aperture photometry at [3.6], for sources with -0.1 < [3.6] - [4.5] < 0 and < 10% flux increase from the 19 to 29 aperture. Median differences and standard deviations in 0.5 mag bins are shown by the red points and error bars. Bottom: Comparison for CatWISE W2 and Spitzer [4.5] photometry. No restriction on Spitzer source color needs to be applied in this case (see <ref>). | 2012.13084 | 1497p015_Spitzer.jpg |
0.456899 | a48c8866d09c43ea8019387684583077 | Same as Figure <ref> but for the Galactic Center tile (2657m288) | 2012.13084 | 2657m288.jpg |
0.435256 | 2fca2febb7924a218c99dc28c542d167 | Same as Figure <ref> but for the North Ecliptic Pole tile (2709p666) | 2012.13084 | 2709p666.jpg |
0.548606 | e244c2adf10f4721adf70cb3955fb22e | A comparison of CatWISE2020 total measured motion to values reported in the literature, for 224 ultra-cool dwarfs within 20 pc of the Sun. All have S/N ≥3 CatWISE2020 motion measurements. Vertical error bars are typically smaller than the symbols. | 2012.13084 | CW2020_vs_lit_PMtot.jpg |
0.460567 | d465d5e6101249c18c6a1fe2c26aeb1a | Comparison of CatWISE photometry to Spitzer photometry for the SSDF, using the same methodology as in Figure <ref>. The outer contour represents a source density of 10 sources per 0.05 × 0.05 mag bin, with each additional contour showing a factor of two increase in source density. | 2012.13084 | SSDF.jpg |
0.425496 | 2ed0051521bf4afba716a84f7a697b97 | CatWISE Preliminary and CatWISE2020 Catalog (top) and Reject Table (bottom) source density, plotted in Galactic Coordinates. See <ref> for details. | 2012.13084 | Source_density_map.jpg |
0.435223 | 80af32b577594183b04166db73fc323a | CatWISE Preliminary and CatWISE2020 astrometric performance with respect to Gaia DR2. CatWISE2020 positions and proper motions have been corrected for the systematic offsets, as discussed in <ref>. Top: the 1-σ dispersion between CatWISE and Gaia R.A. (specifically, Δα^*) and Dec. (left), and proper motion (right), for a subsample of ∼2.1 million sources in the 10<W1<17.5 mag range, uniformly distributed over the entire sky. Bottom: the median χ^2 computed taking into account CatWISE catalog uncertainties, Gaia catalog uncertainties, and the uncertainty introduced by the translation of Gaia's positions to the CatWISE epoch. | 2012.13084 | fullSky.jpg |
0.4239 | 69c0169da5db4aa4ab0250082a643ff7 | CatWISE2020 astrometric performance with respect to Gaia DR2. Top: the 1-σ dispersion between CatWISE2020 and Gaia R.A. and Dec. (left) and proper motion (right), as a function of Gaia G magnitude. Bottom: same as the top row, but as a function of Gaia measured distance, in bins of 20 pc. | 2012.13084 | fullSky_dist_gmag.jpg |
0.40938 | 9ba13f868b3a49bf9807d3bade6bb9ae | CatWISE2020 astrometric performance with respect to GPS1+. Top: the 1-σ dispersion between CatWISE2020 and GPS1+ R.A. and Dec. (left), and proper motion (right). Bottom: the median χ^2 computed taking into account CatWISE2020 Catalog uncertainties, GPS1+ catalog uncertainties, and the uncertainty introduced by the translation of GPS1+ positions to the CatWISE2020 epoch. | 2012.13084 | gps1p.jpg |
0.432868 | 8254d9d4648d49fb9481f10a0e8d298d | GCD with momentum (capped, θ = 0.5, τ_1 = 1, τ_2 = m): comparison among momentum variants on Gaussian and Matrix market data, residual error and relative error vs CPU time and No. of iterations. | 2012.13087 | gcd_capped.jpg |
0.43579 | c768cdaf5f45416e953d03e1bd66f801 | GCD: sampling rules comparison on random and real data, left 2 panels: residual error vs No. of iterations and time, right 2 panels: relative error vs No. of iterations and time. | 2012.13087 | gcd_sampling_rules.jpg |
0.416967 | 18bd881f0619439d984613d19998179d | GCD with momentum (sketch Sample size, τ = 1): comparison among momentum variants on Gaussian and Matrix market data, residual error and relative error vs CPU time and No. of iterations. | 2012.13087 | gcd_tau_1.jpg |
0.44248 | 776ffd0565974351bc2a7be0486c038d | GCD with momentum (sketch Sample size, τ = 10): comparison among momentum variants on Gaussian and Matrix market data, residual error and relative error vs CPU time and No. of iterations. | 2012.13087 | gcd_tau_10.jpg |
0.433018 | ff029328678f4d618b6d0f1810dad621 | GCD with momentum (sketch Sample size, τ = 20): comparison among momentum variants on Gaussian and Matrix market data, residual error and relative error vs CPU time and No. of iterations. | 2012.13087 | gcd_tau_20.jpg |
0.44297 | 4775668e90814b04b7f7a7c6a03072ba | GCD with momentum (sketch Sample size, τ = 30): comparison among momentum variants on Gaussian and Matrix market data, residual error and relative error vs CPU time and No. of iterations. | 2012.13087 | gcd_tau_30.jpg |
0.446602 | 28d585c77e3240a48b153309ee4a597a | GCD with momentum (sketch Sample size, τ = 5): comparison among momentum variants on Gaussian and Matrix market data, residual error and relative error vs CPU time and No. of iterations. | 2012.13087 | gcd_tau_5.jpg |
0.437335 | 6d67b9c55fa045d6a40f8ed2a98058f7 | GCD with momentum (sketch Sample size, τ = m): comparison among momentum variants on Gaussian and Matrix market data, residual error and relative error vs CPU time and No. of iterations. | 2012.13087 | gcd_tau_max.jpg |
0.572996 | 97fa57ea6b7843339d8b0677f7cc02d1 | GCD momentum (effect of sketch sample size τ and capped parameter θ). | 2012.13087 | gcd_tau_theta_.jpg |
0.455258 | dff0986dd4984cbaaa4e7237b95f7dc4 | GK with momentum (capped, θ = 0.5, τ_1 = 1, τ_2 = m): comparison among momentum variants on LIBSVM data, residual error and relative error vs CPU time and No. of iterations. | 2012.13087 | gk_capped.jpg |
0.433314 | f0e8862bfc594fc29ed4d02bb90a3c50 | GK: sampling rules comparison on random data, left 2 panels: residual error vs No. of iterations and time, right 2 panels: relative error vs No. of iterations and time. | 2012.13087 | gk_random_sampling_rules.jpg |
0.433899 | 4f92ed0a403945d3b637ccd9c7240891 | GK with momentum (sketch Sample size, τ = 5): comparison among momentum variants on Gaussian data, residual error and relative error vs CPU time and No. of iterations. | 2012.13087 | gk_random_tau_5.jpg |
0.475595 | c1a5d7541f7c4dc29c65975e0e5cce3c | GK: sampling rules comparison on LIBSVM data, left 2 panels: residual error vs No. of iterations and time, right 2 panels: relative error vs No. of iterations and time. | 2012.13087 | gk_sampling_rules.jpg |
0.442013 | 20e2e439aa5c4561ac5fbf14e57db6a7 | GK with momentum (sketch Sample size, τ = 1): comparison among momentum variants on LIBSVM data, residual error and relative error vs CPU time and No. of iterations. | 2012.13087 | gk_tau_1.jpg |
0.444924 | afd634af2c13445d94073eb08188e823 | GK with momentum (sketch Sample size, τ = 20): comparison among momentum variants on LIBSVM data, residual error and relative error vs CPU time and No. of iterations. | 2012.13087 | gk_tau_20.jpg |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.