Challenges and Opportunities in Improving Worst-Group Generalization in Presence of Spurious Features
It addresses a critical issue in machine learning fairness and robustness for models that exploit spurious correlations, though it is incremental in identifying new challenges rather than proposing a novel solution.
This paper tackles the problem of poor worst-group test accuracy in deep neural networks due to spurious features, by systematically benchmarking 8 state-of-the-art methods across 5 vision datasets and training over 5,000 models. It uncovers challenges in settings with slowly learned spurious features, more classes, and more groups, and proposes cost-efficient model selection strategies.
Deep neural networks often exploit *spurious* features that are present in the majority of examples within a class during training. This leads to *poor worst-group test accuracy*, i.e., poor accuracy for minority groups that lack these spurious features. Despite the growing body of recent efforts to address spurious correlations (SC), several challenging settings remain unexplored.In this work, we propose studying methods to mitigate SC in settings with: 1) spurious features that are learned more slowly, 2) a larger number of classes, and 3) a larger number of groups. We introduce two new datasets, Animals and SUN, to facilitate this study and conduct a systematic benchmarking of 8 state-of-the-art (SOTA) methods across a total of 5 vision datasets, training over 5,000 models. Through this, we highlight how existing group inference methods struggle in the presence of spurious features that are learned later in training. Additionally, we demonstrate how all existing methods struggle in settings with more groups and/or classes. Finally, we show the importance of careful model selection (hyperparameter tuning) in extracting optimal performance, especially in the more challenging settings we introduced, and propose more cost-efficient strategies for model selection. Overall, through extensive and systematic experiments, this work uncovers a suite of new challenges and opportunities for improving worst-group generalization in the presence of spurious features. Our datasets, methods and scripts available at https://github.com/BigML-CS-UCLA/SpuCo.