Self-supervised Learning is More Robust to Dataset Imbalance
This work addresses the problem of dataset imbalance in self-supervised learning for computer vision, offering insights and a method to enhance robustness, though it is incremental in improving existing SSL approaches.
The paper investigates self-supervised learning (SSL) under dataset imbalance and finds that SSL representations are more robust to class imbalance than supervised ones, with a smaller performance gap between balanced and imbalanced pre-training, especially for out-of-domain evaluation. It proposes a re-weighted regularization technique that improves SSL representation quality on imbalanced datasets, closing the gap with balanced datasets.
Self-supervised learning (SSL) is a scalable way to learn general visual representations since it learns without labels. However, large-scale unlabeled datasets in the wild often have long-tailed label distributions, where we know little about the behavior of SSL. In this work, we systematically investigate self-supervised learning under dataset imbalance. First, we find out via extensive experiments that off-the-shelf self-supervised representations are already more robust to class imbalance than supervised representations. The performance gap between balanced and imbalanced pre-training with SSL is significantly smaller than the gap with supervised learning, across sample sizes, for both in-domain and, especially, out-of-domain evaluation. Second, towards understanding the robustness of SSL, we hypothesize that SSL learns richer features from frequent data: it may learn label-irrelevant-but-transferable features that help classify the rare classes and downstream tasks. In contrast, supervised learning has no incentive to learn features irrelevant to the labels from frequent examples. We validate this hypothesis with semi-synthetic experiments and theoretical analyses on a simplified setting. Third, inspired by the theoretical insights, we devise a re-weighted regularization technique that consistently improves the SSL representation quality on imbalanced datasets with several evaluation criteria, closing the small gap between balanced and imbalanced datasets with the same number of examples.