AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Unbiased Classification through Bias-Contrastive and Bias-Balanced Learning

Youngkyu Hong, Eunho Yang

NeurIPS 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Group-robust classification · 26 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
BiasCon loss vs vanilla trainingResNet/CNN · BiasedMNIST rho=0.999unbiased accuracy: 11.8 -> 94.5Improved
add BiasBal loss to BiasCon (BC+BB) (ablation)BiasCon · BiasedMNIST rho=0.995unbiased accuracy: 97.4 -> 97.7Improved
add BiasBal loss to BiasCon (BC+BB) (ablation)BiasCon · BiasedMNIST rho=0.999unbiased accuracy: 94.5 -> 94.0Hurt
add BiasBal loss to BiasCon (BC+BB) (ablation)BiasCon · CelebA Blondeunbiased accuracy: 90.0 -> 91.4Improved
add BiasBal loss to BiasCon (BC+BB) (ablation)BiasCon · CelebA Makeupunbiased accuracy: 77.5 -> 78.6Improved
add BiasBal loss to BiasCon (BC+BB) (ablation)BiasCon · UTKFace Ageunbiased accuracy: 75.7 -> 79.1Improved
BiasBal alone vs BiasCon alone (ablation)BiasCon · CelebA Blondeunbiased accuracy: 90.0 -> 91.1Mixed
BiasCon vs EnD (triplet-loss feature entangling)EnD · BiasedMNIST rho=0.999unbiased accuracy: 59.5 -> 94.5Improved
SoftCon vs vanilla without bias labelsResNet/CNN · BiasedMNIST rho=0.999unbiased accuracy: 11.8 -> 65.0Improved
SoftCon vs vanillaResNet18 · 9-Class ImageNetunbiased accuracy: 92.7 -> 94.1Improved
SoftCon vs vanillaResNet18 · ImageNet-Aaccuracy: 30.5 -> 34.1Improved
BiasBal alone vs vanilla (ablation)CNN · BiasedMNIST rho=0.999unbiased accuracy: 11.8 -> 76.8Improved

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