High-dimensional supervised feature selection via optimized kernel mutual information

作者：

Highlights：

• This would be the first publication of work that integrates kernel learning and MI.

• The OKMI method avoids the problem by finding the optimal features at low computational cost.

• The experiment results show that the OKMI method is effective and robust over a wide range.

摘要

•This would be the first publication of work that integrates kernel learning and MI.•The OKMI method avoids the problem by finding the optimal features at low computational cost.•The experiment results show that the OKMI method is effective and robust over a wide range.

论文关键词：Feature selection,Kernel method,Mutual information,Classification,Optimize function,Machine learning

论文评审过程：Received 16 July 2017, Revised 29 January 2018, Accepted 27 April 2018, Available online 2 May 2018, Version of Record 7 May 2018.

论文官网地址：https://doi.org/10.1016/j.eswa.2018.04.037