Multi-label feature selection via adaptive label correlation estimation

In multi-label learning, each instance is associated with multiple labels simultaneously. Multi-label data often has noisy, irrelevant, and redundant features of high dimensionality. feature selection received considerable attention as an effective means for dealing high-dimensional data. Many methods exploit label correlations to help select features. However, finding selecting in existing are two separate processes, the existence noises outliers training makes exploited from space less reliable. Therefore, learned may mislead process result informative This paper proposes a novel algorithm named ROAD, i.e., featuRe selectiOn via ADaptive correlation estimation. ROAD jointly performs adaptive exploration alternating optimization obtain reliable estimation correlations, which can more effectively reveal intrinsic manifold structure among lead proper subset. Comprehensive experiments on several frequently used sets validate superiority against state-of-the-art algorithms.