Research map / Learning theory and algorithms / Fairness
A Fair Bayesian Inference through Matched Gibbs Posterior
Jihu Lee, Kunwoong Kim, Sehyun Park, Insung Kong, Dongyoon Yang, Yongdai Kim
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 |
|---|---|---|---|
| add matched deviation penalty with learnable matching T (matched Gibbs posterior) vs deterministic fair baselines | fair ReLU MLP · CRIME consistency score at Delta_W^1/2 ~ 0.09 | Con: 0.767 -> 0.837 | Improved |
| use matched Gibbs posterior instead of MMD-based Gibbs/variational posteriors | fair Bayesian inference · CRIME | accuracy: 0.736 vs 0.696 (gibbs mmd), 0.729 (variational mmd) | Improved |
| use matched Gibbs posterior instead of adv, reduction, gapreg | fair classifier · ADULT consistency score | Con: 0.927 -> 0.955 | Improved |
| use matched Gibbs posterior instead of baselines | fair classifier · DUTCH consistency score | Con: 0.939 -> 0.965 | Improved |
| reduce prior temperature tau of matching prior from 1.0 to 0.1 and 0.01 (ablation) | matched Gibbs posterior · ADULT | Acc (x10^2): 82.888 -> 82.892 | No clear effect |
| vary number of pretrain epochs (10, 20, 30, 40) for MCMC initialization (ablation) | matched Gibbs posterior · ADULT | Pareto trade-off | No clear effect |
| use matched Gibbs posterior with strict fairness instead of constrained random-walk MH | posterior sampling under fairness constraint · ADULT | acceptance ratio: 0.02 vs 0.897 | Improved |
| tighten constraint threshold in constrained random-walk MH | random-walk Metropolis-Hastings · ADULT | acceptance ratio: 0.58 -> 0.02 | Hurt |
| tighten DP level delta for samples from standard mean-field Gaussian VI | mean-field Gaussian VI · CRIME logistic regression | proportion of fair samples: 0.014 -> 0.000 | Hurt |
| use gibbs matched for Ece under strict fairness | matched Gibbs posterior · CRIME and DUTCH | Ece | Mixed |
Closest papers by meaning
- On the Burden of Achieving Fairness in Conformal Prediction (NeurIPS 2026 · similarity 0.87)
- Multigroup Fairness and Omniprediction: Separations and Equivalences (NeurIPS 2026 · similarity 0.86)
- Fairness without Harm: An Influence-Guided Active Sampling Approach (NeurIPS 2024 · similarity 0.85)
- Fairness-Aware Estimation of Graphical Models (NeurIPS 2024 · similarity 0.85)
- Is Your Model Fairly Certain? Uncertainty-Aware Fairness Evaluation for LLMs (ICML 2025 · similarity 0.85)
- Fair GLASSO: Estimating Fair Graphical Models with Unbiased Statistical Behavior (NeurIPS 2024 · similarity 0.85)
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.
