Heterogeneous Domain Adaptation for Multiple Classes

In this paper, we present an efficient multi-class heterogeneous domain adaptation method, where data from source and target domains are repre-sented by heterogeneous features of different di-mensions. Specifically, we propose to recon-struct a sparse feature transformation matrix to map the weight vector of classifiers learned from the source domain to the target domain. We cast this learning task as a compressed sensing prob-lem, where each binary classifier induced from multiple classes can be deemed as a measure-ment sensor. Based on the compressive sensing theory, the estimation error of the transformation matrix decreases with the increasing number of classifiers. Therefore, to guarantee reconstruc-tion performance, we construct sufficiently many binary classifiers based on the error correcting output coding. Extensive experiments are con-ducted on both a toy dataset and three real-world datasets to verify the superiority of our proposed method over existing state-of-the-art HDA meth-ods in terms of prediction accuracy.