AI Paper Atlas

Research map / Image generation / GANs

Generative Adversarial Nets for Robust Scatter Estimation: A Proper Scoring Rule Perspective

Chao Gao, Yuan Yao, Weizhi Zhu

JMLR 2020 · Image generation › GANs · approach: Generative adversarial network theory · 14 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use JS-GAN instead of Tyler's M-estimatorJS-GAN G1 · t-distribution v=1spectral norm error: 372.9637 -> 0.2808Improved
use JS-GAN instead of dimension halvingJS-GAN G1 · t-distribution v=4spectral norm error: 1.2834 -> 0.2751Improved
increase contamination distance sJS-GAN G1 · Dirac contaminationspectral norm error: 0.2000 -> 0.1600Improved
increase contamination distance sKendall's tau · Dirac contaminationspectral norm error: 3.8098 -> 92.1079Hurt
increase contamination proportion epsilonTyler's M-estimator, Kendall's tau, MVE · Gaussian contamination, p=100spectral norm error: 4.1114 -> 39.3082Hurt
increase contamination proportion epsilonJS-GAN · p=100spectral norm error: 0.1438 -> 0.1620No clear effect
increase sample size n from 5,000 to 50,000JS-GAN G1 · N(5,5I) contaminationspectral norm error: 1.0186 -> 0.1547Improved
increase sample size nTyler's M-estimator · N(5,5I) contaminationspectral norm error: 38.6494 -> 39.2105No clear effect
increase dimension p from 10 to 200JS-GAN G1 · N(5,5I) contaminationspectral norm error: 0.0498 -> 0.2154Hurt
learn the radial distribution with extra network (G2 vs G1) (ablation)JS-GAN · Simultaneous estimationspectral norm error: 0.2808 -> 0.3350Hurt
add location estimation (G3/G4 vs G1/G2) (ablation)JS-GAN · Simultaneous estimationscatter error: 0.1615 -> 0.1537No clear effect
use asymmetric Beta-GAN(1,0.5) instead of JS-GAN (ablation)Beta-GAN · N(0,I) with N(5,5I), p=100spectral norm error: 0.1188 -> 0.1040Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy