Unknown-Aware Domain Adversarial Learning for Open-Set Domain Adaptation

Open-Set Domain Adaptation (OSDA) assumes that a target domain contains unknown classes, which are not discovered in a source domain. Existing domain adversarial learning methods are not suitable for OSDA because distribution matching with textitunknown classes leads to negative transfer. Previous OSDA methods have focused on matching the source and the target distribution by only utilizing textitknown classes. However, this textitknown-only matching may fail to learn the target-textitunknown feature space. Therefore, we propose Unknown-Aware Domain Adversarial Learning (UADAL), which textitaligns the source and the target-textitknown distribution while simultaneously textitsegregating the target-textitunknown distribution in the feature alignment procedure. We provide theoretical analyses on the optimized state of the proposed textitunknown-aware feature alignment, so we can guarantee both textitalignment and textitsegregation theoretically. Empirically, we evaluate UADAL on the benchmark datasets, which shows that UADAL outperforms other methods with better feature alignments by reporting state-of-the-art performances.