POEM: Polarization of Embeddings for Domain-Invariant Representations
This addresses the problem of handling out-of-distribution samples for deep visual models, offering a novel approach to domain generalization that is incremental but shows strong gains when combined with existing methods.
The paper tackles domain generalization in deep visual models by proposing POEM, which learns both domain-invariant and domain-specific representations and polarizes them to be orthogonal, resulting in consistent performance gains across multiple benchmarks including PACS, VLCS, OfficeHome, TerraIncognita, and DomainNet.
Handling out-of-distribution samples is a long-lasting challenge for deep visual models. In particular, domain generalization (DG) is one of the most relevant tasks that aims to train a model with a generalization capability on novel domains. Most existing DG approaches share the same philosophy to minimize the discrepancy between domains by finding the domain-invariant representations. On the contrary, our proposed method called POEM acquires a strong DG capability by learning domain-invariant and domain-specific representations and polarizing them. Specifically, POEM cotrains category-classifying and domain-classifying embeddings while regularizing them to be orthogonal via minimizing the cosine-similarity between their features, i.e., the polarization of embeddings. The clear separation of embeddings suppresses domain-specific features in the domain-invariant embeddings. The concept of POEM shows a unique direction to enhance the domain robustness of representations that brings considerable and consistent performance gains when combined with existing DG methods. Extensive simulation results in popular DG benchmarks with the PACS, VLCS, OfficeHome, TerraIncognita, and DomainNet datasets show that POEM indeed facilitates the category-classifying embedding to be more domain-invariant.