Group Distributionally Robust Machine Learning under Group Level Distributional Uncertainty
This addresses fairness and robustness issues in ML for applications with heterogeneous data sources, but is incremental as it builds on prior worst-group optimization methods.
The paper tackles the problem of machine learning models degrading performance for underrepresented groups due to spurious correlations, by proposing a distributionally robust optimization framework that accounts for group-level distributional uncertainty to improve worst-group performance, and validates it on real-world data.
The performance of machine learning (ML) models critically depends on the quality and representativeness of the training data. In applications with multiple heterogeneous data generating sources, standard ML methods often learn spurious correlations that perform well on average but degrade performance for atypical or underrepresented groups. Prior work addresses this issue by optimizing the worst-group performance. However, these approaches typically assume that the underlying data distributions for each group can be accurately estimated using the training data, a condition that is frequently violated in noisy, non-stationary, and evolving environments. In this work, we propose a novel framework that relies on Wasserstein-based distributionally robust optimization (DRO) to account for the distributional uncertainty within each group, while simultaneously preserving the objective of improving the worst-group performance. We develop a gradient descent-ascent algorithm to solve the proposed DRO problem and provide convergence results. Finally, we validate the effectiveness of our method on real-world data.