Domain adaptation of image classification exploiting target adaptive collaborative local-neighbor representation

Domain adaptation aims to deal with a kind of problem, in which the distribution of training scenarios and testing scenarios are different. Traditional solutions consider this problem in the point of the distribution matching. For the problem of domain adaptation of image classification, this paper proposes a new collaborative representation from the view of image representation. First, all source samples are supposed to form the dictionary. Second, we encode the target sample by combining this dictionary and exploiting adaptively the local-neighbor geometrical construction. Third, based on this proposed representation, named target adaptive collaborative local-neighbor representation, and the nearest classifier, the classification is implemented. Our core contribution is that the adaptive local-neighbor information of the target sample is technically absorbed to form more robust representation. The experimental results confirm the effectiveness of the proposed method.