Research map / Learning theory and algorithms / Fairness
Fast and Efficient MMD-Based Fair PCA via Optimization over Stiefel Manifold
Junghyun Lee, Gwangsu Kim, Mahbod Olfat, Mark Hasegawa-Johnson, Chang D. Yoo
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 |
|---|---|---|---|
| Use MbF-PCA (10^-3) instead of FPCA (0.1, 0.01) | Fair PCA · German Credit, d=2 | %Var: 7.43 -> 10.17 | Improved |
| Use MbF-PCA (10^-3) instead of FPCA (0.1, 0.01) | Fair PCA · Adult Income, d=2 | %Var: 4.05 -> 6.03 | Improved |
| Use MbF-PCA (10^-6) instead of FPCA (0, 0.01) | Fair PCA · COMPAS, d=2 | MMD^2: 0.011 -> 0.002 | Improved |
| Use MbF-PCA (10^-6) instead of FPCA (0, 0.01) | Fair PCA · German Credit, d=10 | MMD^2: 0.020 -> 0.013 | Improved |
| Use MbF-PCA (10^-3) instead of FPCA (0.1, 0.01) | Fair PCA · COMPAS, d=2 | %Var: 35.06 -> 33.95 | Hurt |
| Use stricter fairness tolerance, MbF-PCA (10^-6) rather than MbF-PCA (10^-3) (ablation) | MbF-PCA · COMPAS, d=2 | %Var: 33.95 -> 11.83 | Hurt |
| Use stricter fairness tolerance, MbF-PCA (10^-6) rather than MbF-PCA (10^-3) (ablation) | MbF-PCA · German Credit, d=10 | %Var: 34.10 -> 16.95 | Hurt |
| Use stricter fairness tolerance, MbF-PCA (10^-6) rather than MbF-PCA (10^-3) (ablation) | MbF-PCA · German Credit, d=10 | MMD^2: 0.020 -> 0.013 | Improved |
| Use stricter fairness tolerance, MbF-PCA (10^-6) rather than MbF-PCA (10^-3) (ablation) | MbF-PCA · COMPAS, d=10 | %Var: 87.75 -> 87.75 | No clear effect |
| Use stricter fairness tolerance, MbF-PCA (10^-6) rather than MbF-PCA (10^-3) (ablation) | MbF-PCA · COMPAS, d=10 | MMD^2: 0.014 -> 0.014 | No clear effect |
| Use MbF-PCA rather than FPCA | Fair PCA · Synthetic data #2 | runtime | Improved |
Closest papers by meaning
- Distributionally Robust Fair Principal Components via Geodesic Descents (ICLR 2022 · similarity 0.90)
- Convex Formulations for Fair Principal Component Analysis (AAAI 2019 · similarity 0.90)
- Fair Streaming Principal Component Analysis: Statistical and Algorithmic Viewpoint (NeurIPS 2023 · similarity 0.89)
- An Iterative Algorithm for Differentially Private $k$-PCA with Adaptive Noise (NeurIPS 2025 · similarity 0.85)
- Fun with Flags: Robust Principal Directions via Flag Manifolds (CVPR 2024 · similarity 0.84)
- Rethinking PCA Through Duality (NeurIPS 2025 · similarity 0.84)
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.
