A2063
Title: A general framework for fair and robust regression
Authors: Wenhai Cui - The Hong Kong Polytechnic University (China) [presenting]
Xiaoting Ji - The Hong Kong Polytechnic University (China)
Wen Su - City University of Hong Kong (Hong Kong)
Xingqiu Zhao - The Hong Kong Polytechnic University (Hong Kong)
Abstract: Fair regression methods typically rely on squared error loss, making them fragile under heavy-tailed noise. A general framework for robust regression under demographic parity (DP) is proposed that applies to a wide class of M-estimators, including Cauchy, Huber, least absolute deviation, quantile, and Tukey losses. An optimal fair transformation is introduced that guarantees DP while achieving the minimum population risk among all rank-preserving fair predictors, with convergence rates established for the resulting estimators. To balance fairness and predictive accuracy, an interpolation scheme is developed whose risk decreases while unfairness grows linearly with the interpolation parameter. The proposed framework can be further extended to conditional DP to account for legitimate covariates. Extensive simulation studies and real data applications demonstrate clear improvements over existing fair regression approaches in both robustness and predictive performance.
