Domain Generalization via Invariant Feature Representation

TL;DR

This paper introduces DICA, a kernel-based method for domain generalization that learns invariant features across multiple domains, enhancing the ability of classifiers to perform well on unseen domains.

Contribution

The paper proposes a novel invariant feature learning algorithm, DICA, with theoretical analysis and empirical validation for improved domain generalization.

Findings

DICA effectively learns invariant features across domains.

Reducing domain dissimilarity improves generalization performance.

Experimental results show improved classifier accuracy on unseen domains.

Abstract

This paper investigates domain generalization: How to take knowledge acquired from an arbitrary number of related domains and apply it to previously unseen domains? We propose Domain-Invariant Component Analysis (DICA), a kernel-based optimization algorithm that learns an invariant transformation by minimizing the dissimilarity across domains, whilst preserving the functional relationship between input and output variables. A learning-theoretic analysis shows that reducing dissimilarity improves the expected generalization ability of classifiers on new domains, motivating the proposed algorithm. Experimental results on synthetic and real-world datasets demonstrate that DICA successfully learns invariant features and improves classifier performance in practice.