Post-processing for Fair Regression via Explainable SVD

2025-04-04

AI Summary

This work addresses the challenge of achieving statistical parity fairness in regression models without access to sensitive attributes during inference. We propose a post-processing method based on interpretable singular value decomposition (SVD) of the model’s weight matrix. Our approach establishes, for the first time, an analytical mapping between singular values and the first- and second-order moment disparities of output distributions across demographic groups. This enables explicit translation of statistical parity constraints into a constrained optimization over singular values, yielding a closed-form optimal solution with theoretical guarantees. Fairness is enforced via a linear weight transformation, requiring no sensitive attributes at inference time. Experiments across multiple benchmark datasets demonstrate that our method matches or surpasses state-of-the-art baselines in the fairness–accuracy trade-off, while offering strong interpretability, computational efficiency, and deployment compatibility.

Technology Category

• Computer Vision: Bias, Fairness & Privacy
• Machine Learning: Ethics, Bias, and Fairness
• Game Theory and Economic Paradigms: Fair Division

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

This paper presents a post-processing algorithm for training fair neural network regression models that satisfy statistical parity, utilizing an explainable singular value decomposition (SVD) of the weight matrix. We propose a linear transformation of the weight matrix, whereby the singular values derived from the SVD of the transformed matrix directly correspond to the differences in the first and second moments of the output distributions across two groups. Consequently, we can convert the fairness constraints into constraints on the singular values. We analytically solve the problem of finding the optimal weights under these constraints. Experimental validation on various datasets demonstrates that our method achieves a similar or superior fairness-accuracy trade-off compared to the baselines without using the sensitive attribute at the inference time.

Problem

• Ensures fair neural network regression via SVD
• Converts fairness constraints into singular value constraints
• Achieves optimal fairness-accuracy trade-off without sensitive attributes

Innovation

• Post-processing algorithm for fair regression
• Explainable SVD for weight matrix transformation
• Converts fairness constraints into singular values