In this paper, we propose a novel learning framework for the problem ofdomain transfer learning. We map the data of two domains to one single commonspace, and learn a classifier in this common space. Then we adapt the commonclassifier to the two domains by adding two adaptive functions to itrespectively. In the common space, the target domain data points are weightedand matched to the target domain in term of distributions. The weighting termsof source domain data points and the target domain classification responses arealso regularized by the local reconstruction coefficients. The novel transferlearning framework is evaluated over some benchmark cross-domain data sets, andit outperforms the existing state-of-the-art transfer learning methods.
展开▼
