Research map / Learning theory and algorithms / Fairness
Fair Representation: Guaranteeing Approximate Multiple Group Fairness for Unknown Tasks
Xudong Shen, Yongkang Wong, Mohan Kankanhalli
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 |
|---|---|---|---|
| learn fair representation with kernel-based fairness regularizer instead of reconstruction only | reconstruction autoencoder · Adult (sensitive: gender, target: income) | sensitive attribute balanced accuracy: .846 -> .573 | Improved |
| same change | reconstruction autoencoder · Adult (sensitive: gender, target: income) | statistical parity difference: .177 -> .029 | Improved |
| same change | reconstruction autoencoder · Adult (sensitive: gender, target: income) | target attribute balanced accuracy: .811 -> .681 | Hurt |
| same change | reconstruction autoencoder · Adult (sensitive: gender, target: income) | DPC: .103 -> .100 | No clear effect |
| use ours instead of FFVAE | FFVAE · Adult (sensitive: gender, target: income) | target attribute balanced accuracy: .610 -> .733 | Improved |
| increase length scale from 2 to 2 sqrt2 (ablation) | Ours · Adult (sensitive: gender, target: income) | sensitive attribute balanced accuracy: .611 -> .573 | Improved |
| increase length scale from 2 to 2 sqrt2 (ablation) | Ours · Adult (sensitive: gender, target: income) | target attribute balanced accuracy: .733 -> .681 | Hurt |
| apply fairness representation learning on MPI3D | reconstruction autoencoder · MPI3D uncorrelated (sensitive: object color) | sensitive attribute balanced accuracy: .935 -> .507 | Improved |
| apply fairness representation learning on MPI3D | reconstruction autoencoder · MPI3D uncorrelated (sensitive: object color), target shape | balanced accuracy: .804 -> .846 | Improved |
| FFVAE on MPI3D background color | reconstruction autoencoder · MPI3D uncorrelated (sensitive: object color), target bkgd color | DOpp: .047 -> .232 | Hurt |
| apply gender-blind learning to ArcFace | ArcFace · LFW face verification | TAR@1e-3 (%): 94.4 -> 86.8 | Hurt |
| apply gender-blind learning to ArcFace | ArcFace · VGGFace2 gender prediction (logistic regression) | gender prediction (%): 96.9 -> 52.6 | Improved |
Closest papers by meaning
- Position: Fair Representations Cannot Hold What They Promise (NeurIPS 2026 · similarity 0.91)
- Kernel-based Equalized Odds: A Quantification of Accuracy-Fairness Trade-off in Fair Representation Learning (NeurIPS 2025 · similarity 0.90)
- Fair Representation Learning with Controllable High Confidence Guarantees via Adversarial Inference (NeurIPS 2025 · similarity 0.89)
- Rethinking Fair Representation Learning for Performance-Sensitive Tasks (ICLR 2025 · similarity 0.89)
- Understanding Fairness and Prediction Error through Subspace Decomposition and Influence Analysis (NeurIPS 2025 · similarity 0.88)
- Efficient Fairness-Performance Pareto Front Computation (NeurIPS 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.
