Research map / Learning theory and algorithms / Fairness
FR-Train: A Mutual Information-Based Approach to Fair and Robust Training
Yuji Roh, Kangwook Lee, Steven Whang, Changho Suh
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 changed | Where | Measured | Effect |
|---|---|---|---|
| jointly train with fairness and robustness discriminators and example re-weighting | FR-Train · Synthetic, COMPAS, and AdultCensus poisoned test data | accuracy: Synthetic .814; COMPAS .670; AdultCensus .809 | Improved |
| jointly optimize fairness and robustness rather than fairness alone (ablation) | FR-Train · COMPAS, poisoned | accuracy: .580 -> .670 | Improved |
| add robust training (ablation) | FR-Train · COMPAS, poisoned | disparate impact: .802 -> .846 | Improved |
| add fair training (ablation) | FR-Train · COMPAS, poisoned | disparate impact: .420 -> .846 | Improved |
| add example re-weighting (ablation) | FR-Train · COMPAS, poisoned | accuracy: .624 -> .670 | Improved |
| use a 5% clean validation set | FR-Train · Synthetic poisoned data | accuracy-fairness tradeoff | No clear effect |
| use a 0.1% validation set | FR-Train · Synthetic poisoned data | accuracy-fairness tradeoff | Hurt |
| decrease λ₂ to 0.1 with a very small validation set | FR-Train · Synthetic poisoned data | accuracy-fairness tradeoff | Improved |
| apply RML sanitization before fairness training | FC · Synthetic, poisoned | accuracy: .760 -> .529 | Hurt |
| increase crowdsourced answers averaged per example from N = 1 to N = 11 | Crowdsourced validation labels · COMPAS | label accuracy: 0.609 -> 0.667 | Improved |
| use FR-Train with equalized-odds fairness instead of logistic regression | Synthetic data · Synthetic test data | equalized odds, Y = 0: 0.351 -> 0.888 | Improved |
Later papers that compared with this method
2 later accepted papers report FRTrain as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Simple and Effective Specialized Representations for Fair Classifiers (NeurIPS 2025 · better on 0 of 1 shared result)
- Sample Selection for Fair and Robust Training (NeurIPS 2021 · better on 5 of 6 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DAFA: Distance-Aware Fair Adversarial Training (ICLR 2024 · similarity 0.88)
- Can Private Machine Learning Be Fair? (AAAI 2025 · similarity 0.88)
- Towards Poisoning Fair Representations (ICLR 2024 · similarity 0.87)
- On the Alignment between Fairness and Accuracy: from the Perspective of Adversarial Robustness (ICML 2025 · similarity 0.87)
- Fair Representation Learning with Controllable High Confidence Guarantees via Adversarial Inference (NeurIPS 2025 · similarity 0.87)
- Achievable Fairness on Your Data With Utility Guarantees (NeurIPS 2024 · similarity 0.86)
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.
