Distributionally Robust Classification for Multi-source Unsupervised Domain Adaptation

2026-01-29

AI Summary

This work addresses the degradation of generalization performance in unsupervised domain adaptation caused by distributional shifts between source and target domains, scarcity of target-domain data, and spurious correlations in the source domain. To this end, the authors propose a distributionally robust learning framework that, for the first time in this setting, jointly models uncertainty in both the covariate distribution and the conditional label distribution. The approach is applicable to both single-source and multi-source scenarios and can be seamlessly integrated with existing unsupervised domain adaptation algorithms. By leveraging distributionally robust optimization, the method enhances model robustness and generalization under extreme target-data scarcity and the presence of spurious correlations. Extensive experiments demonstrate that the proposed framework consistently outperforms strong baselines across various distribution shift settings, maintaining superior performance even when only a few target samples are available.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Adversarial Attacks & Robustness
• Intelligent Robots: Learning & Optimization for ROB

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Unsupervised domain adaptation (UDA) is a statistical learning problem when the distribution of training (source) data is different from that of test (target) data. In this setting, one has access to labeled data only from the source domain and unlabeled data from the target domain. The central objective is to leverage the source data and the unlabeled target data to build models that generalize to the target domain. Despite its potential, existing UDA approaches often struggle in practice, particularly in scenarios where the target domain offers only limited unlabeled data or spurious correlations dominate the source domain. To address these challenges, we propose a novel distributionally robust learning framework that models uncertainty in both the covariate distribution and the conditional label distribution. Our approach is motivated by the multi-source domain adaptation setting but is also directly applicable to the single-source scenario, making it versatile in practice. We develop an efficient learning algorithm that can be seamlessly integrated with existing UDA methods. Extensive experiments under various distribution shift scenarios show that our method consistently outperforms strong baselines, especially when target data are extremely scarce.

Problem

• unsupervised domain adaptation
• distribution shift
• multi-source
• distributional robustness
• generalization

Innovation

• Distributionally Robust Learning
• Unsupervised Domain Adaptation
• Multi-source Adaptation
• Covariate Distribution
• Conditional Label Distribution