Trade-off between reconstruction loss and feature alignment for domain generalization

TL;DR

This paper investigates the balance between reconstruction loss and feature alignment in domain generalization, proposing a joint optimization framework that improves transferability to unseen domains.

Contribution

It reveals the importance of constraining reconstruction loss alongside domain alignment, introducing a new framework for better domain generalization.

Findings

Joint optimization of reconstruction loss and domain discrepancy improves generalization.

Theoretical analysis supports the necessity of reconstruction constraints.

Numerical experiments validate the proposed approach.

Abstract

Domain generalization (DG) is a branch of transfer learning that aims to train the learning models on several seen domains and subsequently apply these pre-trained models to other unseen (unknown but related) domains. To deal with challenging settings in DG where both data and label of the unseen domain are not available at training time, the most common approach is to design the classifiers based on the domain-invariant representation features, i.e., the latent representations that are unchanged and transferable between domains. Contrary to popular belief, we show that designing classifiers based on invariant representation features alone is necessary but insufficient in DG. Our analysis indicates the necessity of imposing a constraint on the reconstruction loss induced by representation functions to preserve most of the relevant information about the label in the latent space. More importantly, we point out the trade-off between minimizing the reconstruction loss and achieving domain alignment in DG. Our theoretical results motivate a new DG framework that jointly optimizes the reconstruction loss and the domain discrepancy. Both theoretical and numerical results are provided to justify our approach.