Gradual Divergence for Seamless Adaptation: A Novel Domain Incremental Learning Method

TL;DR

This paper introduces DARE, a three-stage domain incremental learning method that reduces representation drift and catastrophic forgetting by gradually adapting representations and using a novel buffer sampling strategy.

Contribution

The paper presents DARE, a novel three-stage training process for DIL that effectively mitigates representation drift and catastrophic forgetting, with a new buffer sampling strategy.

Findings

DARE reduces catastrophic forgetting across multiple benchmarks.

The method prevents sudden representation drift at task boundaries.

DARE maintains high performance on previous tasks.

Abstract

Domain incremental learning (DIL) poses a significant challenge in real-world scenarios, as models need to be sequentially trained on diverse domains over time, all the while avoiding catastrophic forgetting. Mitigating representation drift, which refers to the phenomenon of learned representations undergoing changes as the model adapts to new tasks, can help alleviate catastrophic forgetting. In this study, we propose a novel DIL method named DARE, featuring a three-stage training process: Divergence, Adaptation, and REfinement. This process gradually adapts the representations associated with new tasks into the feature space spanned by samples from previous tasks, simultaneously integrating task-specific decision boundaries. Additionally, we introduce a novel strategy for buffer sampling and demonstrate the effectiveness of our proposed method, combined with this sampling strategy, in reducing representation drift within the feature encoder. This contribution effectively alleviates catastrophic forgetting across multiple DIL benchmarks. Furthermore, our approach prevents sudden representation drift at task boundaries, resulting in a well-calibrated DIL model that maintains the performance on previous tasks.