Uniform low-rank representation for unsupervised visual domain adaptation

摘要: Visual domain adaptation aims to adapt a model learned in source target domain, which has received much attention recent years. In this paper, we propose uniform low-rank representation based unsupervised method captures the intrinsic relationship among and samples meanwhile eliminates disturbance from noises outliers. particular, first align into common subspace using alignment technique. Then learn domain-invariant dictionary with respect transformed samples. Finally, all are represented on dictionary. Extensive experimental results show that our is beneficial reducing difference, achieve state-of-the-art performance widely used visual benchmark.