Using Self-Supervised Learning Can Improve Model Robustness and Uncertainty

TL;DR

This paper demonstrates that self-supervised learning significantly enhances model robustness against adversarial attacks, label noise, and input corruptions, and improves out-of-distribution detection, surpassing fully supervised methods.

Contribution

It reveals that self-supervised learning can improve robustness and uncertainty estimation, establishing these as key evaluation axes for future research.

Findings

Self-supervision improves robustness to adversarial examples.

Self-supervision enhances detection of out-of-distribution samples.

Self-supervised models outperform fully supervised ones in robustness metrics.

Abstract

Self-supervision provides effective representations for downstream tasks without requiring labels. However, existing approaches lag behind fully supervised training and are often not thought beneficial beyond obviating or reducing the need for annotations. We find that self-supervision can benefit robustness in a variety of ways, including robustness to adversarial examples, label corruption, and common input corruptions. Additionally, self-supervision greatly benefits out-of-distribution detection on difficult, near-distribution outliers, so much so that it exceeds the performance of fully supervised methods. These results demonstrate the promise of self-supervision for improving robustness and uncertainty estimation and establish these tasks as new axes of evaluation for future self-supervised learning research.