Towards Self-Similarity Consistency And Feature Discrimination For Unsupervised Domain Adaptation

Recent advances in unsupervised domain adaptation mainly focus on learning shared representations by global statistics alignment, such as the Maximum Mean Discrepancy (MMD) which matches the Mean statistics across domains. The lack of class information, however, may lead to partial alignment (or even misalignment) and poor generalization performance. For robust domain alignment, we argue that the similarities across different features in the source domain should be consistent with that in the target domain. Based on this assumption, we propose a new domain discrepancy metric, i.e., Self-similarity Consistency (SSC), to enforce the pairwise relationship between different features being consistent across domains. The Gram matrix matching and Correlation Alignment is proven to be a special case, and a sub-optimal measure of our proposed SSC. Furthermore, we also propose to mitigate the side effect of the partial alignment and misalignment by incorporating the discriminative information of the deep representations. Specifically, a simple yet effective feature norm constraint is exploited to enlarge the discrepancy of inter-class samples. It relieves the requirements of strict alignment when performing adaptation, therefore improving the adaptation performance significantly. Extensive experiments on visual domain adaptation tasks demonstrate the effectiveness of our proposed SSC metric and feature discrimination approach.