Positive and unlabeled (PU) learning addresses binary classification when only confirmed positive instances and a pool of unlabeled data are available. This paradigm combines elements of supervised ...