Invariance Principle Meets Vicinal Risk Minimization
This addresses generalization failures in computer vision models for domains with significant diversity shifts, though it appears incremental as it builds on existing methods like IRM and SDA.
The paper tackles out-of-distribution generalization in deep learning by proposing a domain-shared Semantic Data Augmentation module based on Variance Risk Minimization, achieving consistent performance improvements on benchmarks like PACS, VLCS, OfficeHome, and TerraIncognita.
Deep learning models excel in computer vision tasks but often fail to generalize to out-of-distribution (OOD) domains. Invariant Risk Minimization (IRM) aims to address OOD generalization by learning domain-invariant features. However, IRM struggles with datasets exhibiting significant diversity shifts. While data augmentation methods like Mixup and Semantic Data Augmentation (SDA) enhance diversity, they risk over-augmentation and label instability. To address these challenges, we propose a domain-shared Semantic Data Augmentation (SDA) module, a novel implementation of Variance Risk Minimization (VRM) designed to enhance dataset diversity while maintaining label consistency. We further provide a Rademacher complexity analysis, establishing a tighter generalization error bound compared to baseline methods. Extensive evaluations on OOD benchmarks, including PACS, VLCS, OfficeHome, and TerraIncognita, demonstrate consistent performance improvements over state-of-the-art domain generalization methods.