Protected Test-Time Adaptation via Online Entropy Matching: A Betting Approach

TL;DR

This paper introduces a novel test-time adaptation method that detects distribution shifts using online entropy monitoring and dynamically updates the classifier, improving accuracy under shifts while maintaining performance in stable conditions.

Contribution

The paper proposes a new online self-training framework combining entropy-based shift detection with a betting approach for dynamic model adaptation, linked to optimal transport theory.

Findings

Improves test-time accuracy under distribution shifts.

Maintains calibration and accuracy in stable conditions.

Outperforms existing entropy minimization methods.

Abstract

We present a novel approach for test-time adaptation via online self-training, consisting of two components. First, we introduce a statistical framework that detects distribution shifts in the classifier's entropy values obtained on a stream of unlabeled samples. Second, we devise an online adaptation mechanism that utilizes the evidence of distribution shifts captured by the detection tool to dynamically update the classifier's parameters. The resulting adaptation process drives the distribution of test entropy values obtained from the self-trained classifier to match those of the source domain, building invariance to distribution shifts. This approach departs from the conventional self-training method, which focuses on minimizing the classifier's entropy. Our approach combines concepts in betting martingales and online learning to form a detection tool capable of quickly reacting to distribution shifts. We then reveal a tight relation between our adaptation scheme and optimal transport, which forms the basis of our novel self-supervised loss. Experimental results demonstrate that our approach improves test-time accuracy under distribution shifts while maintaining accuracy and calibration in their absence, outperforming leading entropy minimization methods across various scenarios.