Enhancing multi-label classification by modeling dependencies among labels

作者：

Highlights：

• Enhance multi-label classification by modeling dependencies among labels using BN.

• Our approach can deal with both complete and incomplete labels.

• Relations modeled by our approach are more flexible than current direct methods.

摘要

Highlights•Enhance multi-label classification by modeling dependencies among labels using BN.•Our approach can deal with both complete and incomplete labels.•Relations modeled by our approach are more flexible than current direct methods.

论文关键词：Multi-label classification,Bayesian network,Structure learning,Incomplete label assignments,Maximum likelihood estimation

论文评审过程：Received 2 July 2013, Revised 4 April 2014, Accepted 8 April 2014, Available online 16 April 2014.

论文官网地址：https://doi.org/10.1016/j.patcog.2014.04.009