Soft large margin clustering for unsupervised domain adaptation

Abstract

Unsupervised domain adaptation (UDA) methods usually perform feature matching between domains by considering the domain shift. However, the cluster structure of data, which is one focus in traditional unsupervised learning, is not considered in those methods. In this paper, we attempt to explore such cluster structure in UDA. Specifically, a general transfer learning framework called Clustering for Domain Adaptation (DAC) has been proposed. DAC explores the cluster structure of target data with the help of source data. It seeks a domain-invariant classifier by simultaneously reducing the distribution shifts between domains and exploring the cluster structure for target instances. The optimization of DAC adopts the ADMM strategy, in which each iteration generates a closed-form solution. Empirical results demonstrate the effectiveness of DAC over several real datasets.