Moment Matching for Multi-Source Domain Adaptation

TL;DR

This paper introduces a large multi-source domain adaptation dataset, proposes a moment matching-based deep learning method for knowledge transfer across multiple domains, and provides theoretical insights into moment matching techniques.

Contribution

It presents the largest UDA dataset called DomainNet, a novel moment matching approach for multi-source domain adaptation, and new theoretical understanding of moment matching methods.

Findings

The DomainNet dataset significantly advances multi-source UDA research.

The M3SDA model outperforms existing methods on benchmark tasks.

Theoretical analysis clarifies the effectiveness of moment matching in domain adaptation.

Abstract

Conventional unsupervised domain adaptation (UDA) assumes that training data are sampled from a single domain. This neglects the more practical scenario where training data are collected from multiple sources, requiring multi-source domain adaptation. We make three major contributions towards addressing this problem. First, we collect and annotate by far the largest UDA dataset, called DomainNet, which contains six domains and about 0.6 million images distributed among 345 categories, addressing the gap in data availability for multi-source UDA research. Second, we propose a new deep learning approach, Moment Matching for Multi-Source Domain Adaptation M3SDA, which aims to transfer knowledge learned from multiple labeled source domains to an unlabeled target domain by dynamically aligning moments of their feature distributions. Third, we provide new theoretical insights specifically for moment matching approaches in both single and multiple source domain adaptation. Extensive experiments are conducted to demonstrate the power of our new dataset in benchmarking state-of-the-art multi-source domain adaptation methods, as well as the advantage of our proposed model. Dataset and Code are available at \url{http://ai.bu.edu/M3SDA/}.