Feature selection for label distribution learning using dual-similarity based neighborhood fuzzy entropy

that is used to measure sample similarity in both the feature and the label spaces. Second, we invent a novel