How Does Distribution Matching Help Domain Generalization: An Information-theoretic Analysis
This work addresses the domain generalization problem for machine learning models to improve robustness against out-of-distribution data, providing theoretical insights and a new method, but it is incremental as it builds on existing distribution matching approaches.
The paper tackles the problem of domain generalization by analyzing how distribution matching methods like gradient and representation alignment work, revealing their complementary roles and showing that existing methods focusing on only one are insufficient. It introduces IDM, which simultaneously aligns inter-domain gradients and representations, achieving superior performance over various baselines.
Domain generalization aims to learn invariance across multiple training domains, thereby enhancing generalization against out-of-distribution data. While gradient or representation matching algorithms have achieved remarkable success, these methods generally lack generalization guarantees or depend on strong assumptions, leaving a gap in understanding the underlying mechanism of distribution matching. In this work, we formulate domain generalization from a novel probabilistic perspective, ensuring robustness while avoiding overly conservative solutions. Through comprehensive information-theoretic analysis, we provide key insights into the roles of gradient and representation matching in promoting generalization. Our results reveal the complementary relationship between these two components, indicating that existing works focusing solely on either gradient or representation alignment are insufficient to solve the domain generalization problem. In light of these theoretical findings, we introduce IDM to simultaneously align the inter-domain gradients and representations. Integrated with the proposed PDM method for complex distribution matching, IDM achieves superior performance over various baseline methods.