Theoretical Performance Guarantees for Partial Domain Adaptation via Partial Optimal Transport

2025-06-03

AI Summary

This paper addresses partial domain adaptation (PDA), where the target label space is a proper subset of the source label space—a setting lacking theoretical foundations. Method: We establish, for the first time, a theoretically grounded generalization bound based on Partial Optimal Transport (POT), enabling an interpretable, theory-backed explicit weighting scheme for source samples. Building upon this, we propose WARMPOT, a unified algorithm that jointly guides domain alignment and weighted empirical risk minimization, integrating the partial Wasserstein distance with rigorous generalization error analysis. Contribution/Results: WARMPOT achieves state-of-the-art performance across multiple benchmark datasets. Its principled sample-weighting mechanism significantly outperforms heuristic alternatives, demonstrating both theoretical soundness and empirical effectiveness in mitigating negative transfer from irrelevant source classes.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Search and Optimization: Sampling/Simulation-based Search
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

In many scenarios of practical interest, labeled data from a target distribution are scarce while labeled data from a related source distribution are abundant. One particular setting of interest arises when the target label space is a subset of the source label space, leading to the framework of partial domain adaptation (PDA). Typical approaches to PDA involve minimizing a domain alignment term and a weighted empirical loss on the source data, with the aim of transferring knowledge between domains. However, a theoretical basis for this procedure is lacking, and in particular, most existing weighting schemes are heuristic. In this work, we derive generalization bounds for the PDA problem based on partial optimal transport. These bounds corroborate the use of the partial Wasserstein distance as a domain alignment term, and lead to theoretically motivated explicit expressions for the empirical source loss weights. Inspired by these bounds, we devise a practical algorithm for PDA, termed WARMPOT. Through extensive numerical experiments, we show that WARMPOT is competitive with recent approaches, and that our proposed weights improve on existing schemes.

Problem

• Lack of theoretical basis for partial domain adaptation (PDA)
• Heuristic weighting schemes in existing PDA approaches
• Need for generalization bounds and practical PDA algorithms

Innovation

• Uses partial optimal transport theory
• Proposes partial Wasserstein distance alignment
• Introduces WARMPOT algorithm for PDA