Unsupervised Heterogeneous Domain Adaptation with Fuzzy Rule Learning

Abstract

Heterogeneous domain adaptation is a technique that uses the knowledge of source domain to model the target domain. The source domain and the target domain are semantically related, but their feature spaces are different. Among existing heterogeneous domain adaptive methods, most of them belong to semi-supervised methods, which require some labeled samples in the target domain. However, this kind of dataset is rare in many heter-ogeneous adaptive tasks. In order to solve the above problem, this paper proposes a new unsupervised heter-ogeneous domain adaptive algorithm based on fuzzy rule learning. On the one hand, by introducing the TSK fuzzy system, the proposed method learns two feature transformation matrices corresponding to the source domain and the target domain respectively. By learning two feature transformation matrices, the source domain and the target domain are projected into a common feature subspace. On the other, in order to reduce the information loss caused by feature transformation, the proposed algorithm adopts a variety of information preservation strategies and maximizes the correlation between the transformed source domain data and target domain data. Through experi-ments on domain adaptive datasets, the results show that the proposed algorithm has certain advantages over the existing heterogeneous domain adaptive methods.

Keywords

• |fuzzy rule learning|tsk fuzzy system|information preserving|heterogeneous domain adaptation