Clustered intrinsic label correlations for multi-label classification

作者：

Highlights：

• The classifier for each label consists of a label-specific part and a shared one.

• The label-specific part characterizes the corresponding label.

• The shared part represents the information shared by all labels.

• Intrinsic label correlations are represented by label-specific parts.

• The proposed method extends SVM to the multi-label setting.

摘要

•The classifier for each label consists of a label-specific part and a shared one.•The label-specific part characterizes the corresponding label.•The shared part represents the information shared by all labels.•Intrinsic label correlations are represented by label-specific parts.•The proposed method extends SVM to the multi-label setting.

论文关键词：Multi-label classification,Label correlation,Clustered structure,Label-specific component,Shared component,Block coordinate descent

论文评审过程：Received 1 February 2016, Revised 23 March 2017, Accepted 24 March 2017, Available online 27 March 2017, Version of Record 30 March 2017.

论文官网地址：https://doi.org/10.1016/j.eswa.2017.03.054