Generalizing to Unseen Domains with Wasserstein Distributional Robustness under Limited Source Knowledge

TL;DR

This paper introduces WDRDG, a novel framework for domain generalization that enhances robustness to unseen domains with limited source data by leveraging Wasserstein distributional robustness and test-time adaptation.

Contribution

The paper proposes a new domain generalization method combining Wasserstein distributional robustness with test-time adaptation, addressing limited source data and unseen domain shifts.

Findings

Effective on Rotated MNIST, PACS, VLCS datasets

Balances robustness and discriminability

Outperforms existing methods in challenging scenarios

Abstract

Domain generalization aims at learning a universal model that performs well on unseen target domains, incorporating knowledge from multiple source domains. In this research, we consider the scenario where different domain shifts occur among conditional distributions of different classes across domains. When labeled samples in the source domains are limited, existing approaches are not sufficiently robust. To address this problem, we propose a novel domain generalization framework called {Wasserstein Distributionally Robust Domain Generalization} (WDRDG), inspired by the concept of distributionally robust optimization. We encourage robustness over conditional distributions within class-specific Wasserstein uncertainty sets and optimize the worst-case performance of a classifier over these uncertainty sets. We further develop a test-time adaptation module leveraging optimal transport to quantify the relationship between the unseen target domain and source domains to make adaptive inference for target data. Experiments on the Rotated MNIST, PACS and the VLCS datasets demonstrate that our method could effectively balance the robustness and discriminability in challenging generalization scenarios.