Similarity-based constraint score for feature selection
作者：
Highlights：
• A new constraint score is proposed for feature selection.
• It can be used in the context of both supervised and semi-supervised learnings.
• It is based on similarity matrices to evaluate the relevance of a subset of features.
• It evaluates a subset of features at once and can identify redundant features.
• It outperforms the other state-of-the-art constraint scores.
摘要
•A new constraint score is proposed for feature selection.•It can be used in the context of both supervised and semi-supervised learnings.•It is based on similarity matrices to evaluate the relevance of a subset of features.•It evaluates a subset of features at once and can identify redundant features.•It outperforms the other state-of-the-art constraint scores.
论文关键词：Constraint score,Feature selection,Pairwise constraints,Similarity matrix
论文评审过程：Received 18 February 2020, Revised 22 July 2020, Accepted 9 September 2020, Available online 17 September 2020, Version of Record 23 September 2020.
论文官网地址：https://doi.org/10.1016/j.knosys.2020.106429
