Optimization of classifier chains via conditional likelihood maximization

作者：

Highlights：

• A general framework is proposed for multi-label classification from the viewpoint of conditional likelihood maximization.

• Based on the proposed framework, the popular classifier chains method is optimized in terms of label correlation modeling and multi-label feature selection.

• The contribution of the proposed method is demonstrated theoretically and experimentally.

摘要

•A general framework is proposed for multi-label classification from the viewpoint of conditional likelihood maximization.•Based on the proposed framework, the popular classifier chains method is optimized in terms of label correlation modeling and multi-label feature selection.•The contribution of the proposed method is demonstrated theoretically and experimentally.

论文关键词：Multi-label classification,Classifier chains,Conditional likelihood maximization,k-dependence Bayesian network,Multi-label feature selection

论文评审过程：Received 12 June 2016, Revised 20 September 2017, Accepted 23 September 2017, Available online 25 September 2017, Version of Record 9 October 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.09.034