Dynamic Transfer for Multi-Source Domain Adaptation

TL;DR

This paper introduces a dynamic transfer approach for multi-source domain adaptation, where model parameters adapt to individual samples, effectively handling domain conflicts and improving performance without relying on domain labels.

Contribution

The proposed method dynamically adapts model parameters at the sample level, transforming multi-source domains into a single-source domain and simplifying domain alignment.

Findings

Outperforms state-of-the-art by over 3% on DomainNet

Does not require domain labels for adaptation

Effectively handles conflicts across multiple domains

Abstract

Recent works of multi-source domain adaptation focus on learning a domain-agnostic model, of which the parameters are static. However, such a static model is difficult to handle conflicts across multiple domains, and suffers from a performance degradation in both source domains and target domain. In this paper, we present dynamic transfer to address domain conflicts, where the model parameters are adapted to samples. The key insight is that adapting model across domains is achieved via adapting model across samples. Thus, it breaks down source domain barriers and turns multi-source domains into a single-source domain. This also simplifies the alignment between source and target domains, as it only requires the target domain to be aligned with any part of the union of source domains. Furthermore, we find dynamic transfer can be simply modeled by aggregating residual matrices and a static convolution matrix. Experimental results show that, without using domain labels, our dynamic transfer outperforms the state-of-the-art method by more than 3% on the large multi-source domain adaptation datasets -- DomainNet. Source code is at https://github.com/liyunsheng13/DRT.