AI Paper Atlas

Research map / Learning theory and algorithms / Fairness

FIFA: Making Fairness More Generalizable in Classifiers Trained on Imbalanced Data

Zhun Deng, Jiayao Zhang, Linjun Zhang, Ting Ye, Yates Coley, Weijie J Su, James Zou

ICLR 2023 · Learning theory and algorithms › Fairness · approach: Post-processing and transport for fairness · 6 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
Add subgroup- and class-dependent FIFA logit marginsGrid Search with ResNet-18 · CelebA, ε=0.01Test fairness violation: 20.29% -> 2.75%Improved
Add subgroup- and class-dependent FIFA logit marginsGrid Search with ResNet-18 · CelebA, ε=0.05Test combined loss: 13.05% -> 6.34%Improved
Use FIFA margins instead of vanilla trainingLogistic regression with Exponentiated Gradient · AdultIncome, ε=0.01Test fairness violation: 6.7475% -> 2.7776%Improved
Use FIFA margins instead of vanilla trainingLogistic regression with Exponentiated Gradient · DutchConsensus, ε=0.05Test fairness violation: 4.4323% -> 4.4323%No clear effect
Add logits-based LDAM margins (α=0 and all δ=0) (ablation)Grid Search with ResNet-18 · CelebA, ε=0.01Test combined loss: 14.01% -> 7.29%Improved
Tune subgroup fairness margins and α beyond LDAM (ablation)LDAM / FIFA loss with Grid Search · CelebA, ε=0.01Test fairness violation: 5.39% -> 2.75%Improved
Add FIFA margins instead of vanilla lossGrid Search with ResNet-18 · CelebA, ε=0.10Test combined loss: 16.71% -> 6.54%Improved

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