Abstract
Multi-label learning deals with the problem where each training example is represented by a single instance while associated with a set of class labels. For an unseen example, existing approaches choose to determine the membership of each possible class label to it based on identical feature set, i.e. the very instance representation of the unseen example is employed in the discrimination processes of all labels. However, this commonly-used strategy might be suboptimal as different class labels usually carry speci?c characteristics of their own, and it could be bene?cial to exploit different feature sets for the discrimination of different labels. Based on the above re?ection, we propose a new strategy to multi-label learning by leveraging labelspeci?c features, where a simple yet effective algorithm named L IFT is presented. Brie?y, L IFT constructs features speci?c to each label by conducting clustering analysis on its positive and negative instances, and then performs training and testing by querying the clustering results. Extensive experiments across sixteen diversi?ed data sets clearly validate the superiority of L IFT against other wellestablished multi-label learning algorithms.