AI Paper Atlas

Research map / Label-efficient and robust learning / Noisy labels

R-divergence for Estimating Model-oriented Distribution Discrepancy

Zhilin Zhao, Longbing Cao

NeurIPS 2023 · Label-efficient and robust learning › Noisy labels · approach: Contrastive learning on noisy data · 0 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
R-Div instead of H-Div (hypothesis learned on mixed data, risk on each dataset)distribution discrepancy test · HIGGS averagetest power: 0.847 -> 1.000Improved
R-Div instead of H-Div at N=1000distribution discrepancy test · HIGGStest power: 0.240 -> 1.000Improved
R-Div instead of H-Divdistribution discrepancy test · MNIST vs Colored-MNISTtest power: 0.951 -> 1.000Improved
R-Div instead of H-Divdistribution discrepancy test · PACS six domain pairs averagetest power: 0.714 -> 0.849Improved
R-Div instead of H-Div on hardest domain pair C+Pdistribution discrepancy test · PACS C+Ptest power: 0.448 -> 0.527Improved
R-Div instead of H-Div on SENet at N=5000distribution discrepancy test · CIFAR10 vs corrupted CIFAR10test power: 0.478 -> 0.790Improved
R-Div instead of H-Div on ViT at N=1000distribution discrepancy test · CIFAR10 vs corrupted CIFAR10test power: 0.119 -> 0.211Improved
Increase sample size to N=10000R-Div and H-Div · CIFAR10 vs corrupted CIFAR10test power: 1.000 / 1.000No clear effect
R-Div instead of H-Div on MNISTdistribution discrepancy test · MNISTtest power: 1.000 -> 1.000No clear effect
R-Div instead of H-Div at 1% corruptiondistribution discrepancy test · CIFAR10 Snow corruptiontest power: 0.096 -> 0.189Improved
R-Div instead of H-Div at 0.5% corruption (mild)distribution discrepancy test · CIFAR10 and ImageNet corruptionstest power: 0.000 / 0.000No clear effect
Mixed-data minimum hypothesis (R-Div) vs per-dataset hypotheses (H-Div)discrepancy estimator · MNIST N=200risk gap: 7.1813 -> 44.4703Improved

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