Research map / Learning theory and algorithms / Fairness
Learning Adversarially Fair and Transferable Representations
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Learn representations with LAFTR's equalized-odds adversary observing the representation | LAFTR transfer learning · Heritage Health, 10 PCG transfer tasks | ΔEO | Helped |
| Learn representations with LAFTR instead of Transfer-Unfair | Transfer prediction · Heritage Health transfer: MSC2a3 | ΔEO: 0.370 -> 0.281 | Helped |
| Learn representations with LAFTR instead of Transfer-Unfair | Transfer prediction · Heritage Health transfer: SKNAUT | ΔEO: 0.125 -> 0.155 | Hurt |
| Use an output-only adversary rather than an adversary with access to the learned representation | Fair transfer representations · Heritage Health, PCG transfer tasks | ΔEO | Hurt |
| Add reconstruction loss (β = 1) to Transfer-Y-Adv instead of using β = 0 (ablation) | Transfer-Y-Adv · Heritage Health transfer representations | MMD: $1.1 \times 10^{-3}$ -> $3.4 \times 10^{-5}$ | Helped |
| Add reconstruction loss (β = 1) to Transfer-Y-Adv instead of using β = 0 (ablation) | Transfer-Y-Adv · Heritage Health transfer representations | AdvAcc: 0.786 -> 0.787 | No clear effect |
| Use group-normalized ℓ1 adversarial loss instead of cross-entropy | LAFTR demographic-parity adversary · Imbalanced synthetic example | test discrepancy: 0 -> approximately 0.368 | Helped |
| Match the adversarial objective to the target fairness criterion | LAFTR · Adult fair classification | accuracy at low unfairness | Helped |
| Add reconstruction loss to preserve information beyond the original task label | LAFTR transfer learning · Heritage Health PCG tasks | transfer utility | Helped |
| Use a fair-regularized MLP representation rather than a task-unfair MLP representation | Transfer representation learning · Heritage Health, PCG transfer tasks | ΔEO | No clear effect |
Closest papers by meaning
- Fair Representation Learning with Controllable High Confidence Guarantees via Adversarial Inference (NeurIPS 2025 · similarity 0.91)
- On the Fairness ROAD: Robust Optimization for Adversarial Debiasing (ICLR 2024 · similarity 0.90)
- On the Alignment between Fairness and Accuracy: from the Perspective of Adversarial Robustness (ICML 2025 · similarity 0.89)
- Adversarial Fairness Network (AAAI 2024 · similarity 0.89)
- FairICP: Encouraging Equalized Odds via Inverse Conditional Permutation (ICML 2025 · similarity 0.88)
- Can Private Machine Learning Be Fair? (AAAI 2025 · similarity 0.88)
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.