Four crucial issues are considered by the proposed HoAL: 1) unlike binary cases, the selection granularity for multilabel active learning need to be fined from example to examplelabel pair 2) different labels are seldom independent, and label correla