How Robust is Your Fairness? Evaluating and Sustaining Fairness under Unseen Distribution Shifts
How Robust is Your Fairness? Evaluating and Sustaining Fairness under Unseen Distribution Shifts
复制标题
DOI：
10.48550/arxiv.2207.01168
复制
发表时间：
2022-07
期刊：
影响因子：
--
通讯作者：
Haotao Wang;Junyuan Hong;Jiayu Zhou;Zhangyang Wang
中科院分区：
文献类型：
--
作者：
Haotao Wang;Junyuan Hong;Jiayu Zhou;Zhangyang Wang
近年来，人们越来越关注深度学习的公平性。现有的公平性感知机器学习方法主要关注分布数据的公平性。然而，在实际应用中，训练数据和测试数据之间的分布偏移是常见的。在本文中，我们首先表明，现有的方法实现的公平性可以很容易地打破轻微的分布变化。为了解决这个问题，我们提出了一种新的公平性学习方法，称为CUrvature匹配（CUrvature MAtching），它可以实现鲁棒的公平性推广到未知分布变化的未知域。具体来说，通过匹配两个群体的损失曲率分布，CNOAA强制模型在多数群体和少数群体上具有相似的泛化能力。我们在三个流行的公平数据集上评估我们的方法。与现有的方法相比，在不牺牲整体精度或分布内公平性的情况下，在不可见的分布变化下，CRAMA实现了上级公平性。
Increasing concerns have been raised on deep learning fairness in recent years. Existing fairness-aware machine learning methods mainly focus on the fairness of in-distribution data. However, in real-world applications, it is common to have distribution shift between the training and test data. In this paper, we first show that the fairness achieved by existing methods can be easily broken by slight distribution shifts. To solve this problem, we propose a novel fairness learning method termed CUrvature MAtching (CUMA), which can achieve robust fairness generalizable to unseen domains with unknown distributional shifts. Specifically, CUMA enforces the model to have similar generalization ability on the majority and minority groups, by matching the loss curvature distributions of the two groups. We evaluate our method on three popular fairness datasets. Compared with existing methods, CUMA achieves superior fairness under unseen distribution shifts, without sacrificing either the overall accuracy or the in-distribution fairness.