Active partial label learning based on adaptive sample selection

Partial label learning is a type of weak supervised learning which uses samples with candidate label sets to train a classifier. Most of the related researches assume that there are a lot of available training samples with partial labels in advance, that is to assume that the candidate label set is easy to obtain. In many practical problems, however, there are still a large number of unlabeled samples, and obtaining their partial labels is costly. In this paper, we consider using a small number of partially labeled samples and a large number of unlabeled samples to form the training set, and propose a partial label learning method based on active learning mechanism to construct an effective classifier. Firstly, the weak supervised information in candidate label set is used to determine the possible labels of the partially labeled samples by using iterative label transfer process; then an adaptive sample selection strategy in active learning framework is proposed to comprehensively measure the labeling value of each unlabeled sample based on its uncertainty, graph density and label transfer ability, and the most valuable samples are selected from unlabeled sample set for manual labeling. Finally, the labeled samples are used to re-optimize the existing partially labeled samples, and the final classifier is trained. The experimental results on some benchmark datasets show that the proposed active partial label learning method has higher classification accuracy than the representative similar methods, and only needs to label a small number of samples to achieve stable performance.