Group-preserving label-specific feature selection for multi-label learning

作者：

Highlights：

• A group-preserving framework is designed for label-specific feature selection.

• Label-group and instance-group correlations are used to enhance the generalization.

• An alternating minimization algorithm is presented to seek the optimal solution.

• Empirical studies are conducted to validate the advantages of the proposed method.

摘要

•A group-preserving framework is designed for label-specific feature selection.•Label-group and instance-group correlations are used to enhance the generalization.•An alternating minimization algorithm is presented to seek the optimal solution.•Empirical studies are conducted to validate the advantages of the proposed method.

论文关键词：Feature selection,Multi-label learning,Label-specific,Group-preserving,Local label correlations

论文评审过程：Received 14 March 2022, Revised 13 September 2022, Accepted 15 September 2022, Available online 21 September 2022, Version of Record 27 September 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.118861