Abstract

In this paper, we specifically examine the training of amulti-label classifier from data with incompletely assignedlabels. This problem is fundamentally important in manymulti-label applications because it is almost impossible for human annotators to assign a complete set of labels, al-though their judgments are reliable. In other words, a multilabel dataset usually has properties by which (1) assignedlabels are definitely positive and (2) some labels are ab-sent but are still considered positive. Such a setting hasbeen studied as a positive and unlabeled (PU) classifica-tion problem in a binary setting. We treat incomplete label assignment problems as a multi-label PU ranking, which is an extension of classical binary PU problems to the wellstudied rank-based multi-label classification. We derive theconditions that should be satisfied to cancel the negative effects of label incompleteness. Our experimentally obtained results demonstrate the effectiveness of these conditions.