Continual Learning of Domain-Invariant Representations

2026-05-15

AI Summary

This work addresses the limited generalization of existing continual learning methods to unseen domains, which often stems from their reliance on domain-specific cues. To overcome this limitation, the study introduces domain-invariant representation learning into the continual learning framework for the first time, proposing a deployment-oriented sequential invariance alignment strategy. This approach integrates experience replay with causal structure modeling to preserve the underlying invariant causal relationships across tasks. Evaluated on six benchmark and real-world datasets spanning computer vision, medical imaging, manufacturing, and ecology, the proposed method significantly outperforms current state-of-the-art approaches and demonstrates markedly stronger generalization performance on unseen target domains.

Abstract

Continual learning (CL) aims to train models sequentially over multiple domains without forgetting previously learned knowledge. However, existing CL methods optimize for in-domain performance and are therefore prone to learning spurious, domain-specific cues (``shortcut learning''), which limits generalization to unseen domains after deployment. In this paper, we address this limitation through continual learning of domain-invariant representation. We introduce a broad class of CL methods that sequentially learn representations capturing invariant structures across domains. Our methods are motivated by the observation that such invariant structures often preserve the underlying causal mechanisms, which can reduce the risk of overfitting to domain-specific cues and thus offer better out-of-domain generalization. Our proposed CL methods combine replay-based training with a tailored sequential invariance alignment to learn -- and preserve -- invariant structures over time. We evaluate our methods under a deployment-oriented protocol that measures performance on unseen target domains. Across six benchmark and real-world datasets spanning vision, medicine, manufacturing, and ecology, our methods consistently outperform existing CL baselines in terms of generalization to unseen target domains. As an ablation, we further show that naïve extensions of sequential training with existing domain-invariant representation learning (DIRL) methods provide only limited benefits. To the best of our knowledge, this is the first work to develop domain-invariant representation methods for CL.

Problem

• continual learning
• domain-invariant representation
• out-of-domain generalization
• shortcut learning
• catastrophic forgetting

Innovation

• Continual Learning
• Domain-Invariant Representations
• Out-of-Domain Generalization
• Sequential Invariance Alignment
• Causal Mechanisms