Efficient Bias Mitigation Without Privileged Information
This addresses bias mitigation in machine learning for practitioners, offering a more efficient solution without privileged information, though it is incremental as it builds on existing bias mitigation approaches.
The paper tackles performance disparities in deep neural networks due to spurious correlations by proposing TAB, a hyperparameter-free framework that improves worst-group performance without group information, outperforming existing methods while maintaining overall accuracy.
Deep neural networks trained via empirical risk minimisation often exhibit significant performance disparities across groups, particularly when group and task labels are spuriously correlated (e.g., "grassy background" and "cows"). Existing bias mitigation methods that aim to address this issue often either rely on group labels for training or validation, or require an extensive hyperparameter search. Such data and computational requirements hinder the practical deployment of these methods, especially when datasets are too large to be group-annotated, computational resources are limited, and models are trained through already complex pipelines. In this paper, we propose Targeted Augmentations for Bias Mitigation (TAB), a simple hyperparameter-free framework that leverages the entire training history of a helper model to identify spurious samples, and generate a group-balanced training set from which a robust model can be trained. We show that TAB improves worst-group performance without any group information or model selection, outperforming existing methods while maintaining overall accuracy.