Abstract
In this paper, we tackle the challenges of multilabel classi?cation by developing a general conditional dependency network model. The proposed model is a cyclic directed graphical model, which provides an intuitive representation for the dependencies among multiple label variables, and a well integrated framework for ef?cient model training using binary classi?ers and label predictions using Gibbs sampling inference. Our experiments show the proposed conditional model can effectively exploit the label dependency to improve multi-label classi?cation performance.