Continual Learning of Domain-Invariant Representations

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.

Machine learning models often learn correlations, and not causation. This causes them to fail when they are used on new data: the correlation does not hold here. We propose methods that help the machine learning systems learn causal relationships between the input data (such as images) and the output (such as "this is a dog"). We do this in a continual manner. In the beginning, for example, only outdoor images are available, then indoor images, then images with rain. After training on these different domains, we test the model on a completely now domain. For example, this could be images with snow. Our results show that methods that only learn correlations can fail on these new domains. In contrast, our methods are able to make predictions on these new domains.