Group-robust Sample Reweighting for Subpopulation Shifts via Influence Functions

2025-03-10

AI Summary

Existing group-robust learning methods address imbalanced generalization under subgroup distribution shifts but require extensive group-label annotations—rendering them costly and impractical. Method: We propose GSR, a two-stage sample reweighting framework that leverages only a minimal set of group-labeled samples as a target set. GSR quantifies the influence of unlabeled samples on the worst-group loss via influence functions and iteratively optimizes both sample weights and the model’s final layer. Contribution/Results: This work is the first to incorporate influence functions into group-robust optimization, eliminating the need for full group labeling. GSR significantly improves worst-group performance under low-label budgets. Empirical evaluation across multiple benchmarks demonstrates that GSR surpasses state-of-the-art methods using substantially fewer group labels, while offering theoretical interpretability, computational efficiency, and strong robustness.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Intelligent Robots: Learning & Optimization for ROB
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Machine learning models often have uneven performance among subpopulations (a.k.a., groups) in the data distributions. This poses a significant challenge for the models to generalize when the proportions of the groups shift during deployment. To improve robustness to such shifts, existing approaches have developed strategies that train models or perform hyperparameter tuning using the group-labeled data to minimize the worst-case loss over groups. However, a non-trivial amount of high-quality labels is often required to obtain noticeable improvements. Given the costliness of the labels, we propose to adopt a different paradigm to enhance group label efficiency: utilizing the group-labeled data as a target set to optimize the weights of other group-unlabeled data. We introduce Group-robust Sample Reweighting (GSR), a two-stage approach that first learns the representations from group-unlabeled data, and then tinkers the model by iteratively retraining its last layer on the reweighted data using influence functions. Our GSR is theoretically sound, practically lightweight, and effective in improving the robustness to subpopulation shifts. In particular, GSR outperforms the previous state-of-the-art approaches that require the same amount or even more group labels.

Problem

• Improves robustness to subpopulation shifts in data distributions.
• Enhances group label efficiency using group-labeled data as target.
• Proposes Group-robust Sample Reweighting for better model generalization.

Innovation

• Utilizes group-labeled data to optimize unlabeled data weights
• Employs influence functions for iterative model retraining
• Enhances robustness to subpopulation shifts efficiently