Group-robust Sample Reweighting for Subpopulation Shifts via Influence Functions
This addresses the challenge of costly group labels for enhancing model robustness to subpopulation shifts, offering a more label-efficient solution.
The paper tackles the problem of uneven model performance across subpopulations when group proportions shift during deployment, proposing Group-robust Sample Reweighting (GSR) to improve robustness with fewer group labels, achieving state-of-the-art results compared to methods requiring similar or more labels.
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.