A Two-Way alignment approach for unsupervised multi-Source domain adaptation

•We introduce a two-way alignment framework for unsupervised multisource domain adaptation task. We first align the target and multiple source domains on domain-level by an adversarial learning process, and then reduce the domain gap on category-level by minimizing the distance between the category prototypes and target instances with a minimax entropy loss.•We propose an instance weighting strategy to mitigate the impact of instance variations inside each domain.•Experimental results on several datasets demonstrate the superiority of the proposed algorithm to several state-of-the-art approaches. Domain adaptation aims at transferring knowledge from labeled source domain to unlabeled target domain. Current advances primarily concern single source domain and neglect the setting of multiple source domains. Previous unsupervised multi-source domain adaptation (MDA) algorithms only consider domain-level alignment, while neglecting the category-level information among multiple domains and the instance variations inside each domain. This paper introduces a Two-Way alignment framework for MDA (TWMDA), which considers both domain-level and category-level alignments, and addresses the instance variations. We first align the target and multiple sources on the domain-level by an adversarial learning process. To circumvent the drawbacks of adversarial learning, we further reduce the domain gap on the category-level by minimizing the distance between the category prototypes and unlabeled target instances. To address the instance variations, we design an instance weighting strategy for diverse source instances. The effectiveness of TWMDA is demonstrated on three benchmark datasets for image classification.