Domain Generalization via Nuclear Norm Regularization

Abstract

The ability to generalize to unseen domains is crucial for machine learning systems deployed in the realworld, especially whenwe only have data from limited training domains. In this paper, we propose a simple and effective regularization method based on the nuclear norm of the learned features for domain generalization. Intuitively, the proposed regularizer mitigates the impacts of environmental features and encourages learning domain-invariant features. Theoretically, we provide insights intowhy nuclear normregularization is more effective compared toERMand alternative regularization methods. Empirically,we conduct extensive experiments on both synthetic and real datasets. We shownuclear normregularization achieves strong performance compared to baselines in a wide range of domain generalization tasks. Moreover, our regularizer is broadly applicable with various methods such as ERM and SWAD with consistently improved performance, e.g., 1.7% and 0.9% test accuracy improvements respectively on the DomainBed benchmark.