Domain Adaptation via Maximizing Surrogate Mutual Information [article]

Haiteng Zhao, Chang Ma, Qinyu Chen, Zhi-Hong Deng
2022 arXiv   pre-print
Unsupervised domain adaptation (UDA) aims to predict unlabeled data from target domain with access to labeled data from the source domain. In this work, we propose a novel framework called SIDA (Surrogate Mutual Information Maximization Domain Adaptation) with strong theoretical guarantees. To be specific, SIDA implements adaptation by maximizing mutual information (MI) between features. In the framework, a surrogate joint distribution models the underlying joint distribution of the unlabeled
more » ... rget domain. Our theoretical analysis validates SIDA by bounding the expected risk on target domain with MI and surrogate distribution bias. Experiments show that our approach is comparable with state-of-the-art unsupervised adaptation methods on standard UDA tasks.
arXiv:2110.12184v2 fatcat:3tgdw34uyfa2zoep55w2evyjhu