Learning Label Correlations For Multi-label Classification

Multi-label classification deals with the problem where each training example is associated with a set of labels simultaneously, with the set of labels corresponding to multiple concepts or semantic meanings. It is evident that multi-label classification can be regarded as a generalized version of traditional classification task where each example is confined to have only one single label. However, the generality of multi-label classification inevitably makes the corresponding classification task much more difficult to solve. Intuitively, the multiple labels are usually correlated in some semantic space while sharing the same input space. As a consequence, the multi-label classification process can be augmented significantly by exploiting the label correlations effectively. Most of the existing approaches share the limitations in that the label correlations are typically taken as prior knowledge, which may not depict the true dependencies among labels correctly; or they do not adequately address the issue of missing labels.In this dissertation, we propose an integrated framework that learns the correlations among labels while training the multi-label model simultaneously. Specifically, a low rank structure is adopted to capture the complex correlations among labels. In the meantime, to address the issue of incomplete labels, we incorporate a supplementary label matrix which augments the incomplete original label matrix by exploiting the label correlations. An alternating algorithm is then developed to solve the optimization problem. Extensive experiments are conducted on a number of image and text data sets to demonstrate the advantages of the proposed approach from aspects of multi-label classification, capturing inherent semantic relations, as well as the robustness to missing labels.