Adaptive and Balanced Re-initialization for Long-timescale Continual Test-time Domain Adaptation

TL;DR

This paper introduces ABR, a re-initialization strategy for long-term continual test-time domain adaptation, which adaptively resets model weights based on label flip changes to improve long-term performance.

Contribution

It proposes a novel adaptive re-initialization policy that dynamically resets model weights to enhance long-term CTTA performance, addressing a key challenge in non-stationary environments.

Findings

ABR outperforms existing CTTA methods on multiple benchmarks.

Adaptive re-initialization effectively maintains model performance over long periods.

The method is simple, effective, and generalizable across different scenarios.

Abstract

Continual test-time domain adaptation (CTTA) aims to adjust models so that they can perform well over time across non-stationary environments. While previous methods have made considerable efforts to optimize the adaptation process, a crucial question remains: Can the model adapt to continually changing environments over a long time? In this work, we explore facilitating better CTTA in the long run using a re-initialization (or reset) based method. First, we observe that the long-term performance is associated with the trajectory pattern in label flip. Based on this observed correlation, we propose a simple yet effective policy, Adaptive-and-Balanced Re-initialization (ABR), towards preserving the model's long-term performance. In particular, ABR performs weight re-initialization using adaptive intervals. The adaptive interval is determined based on the change in label flip. The proposed method is validated on extensive CTTA benchmarks, achieving superior performance.