On a class of computationally efficient feature selection criteria
作者：
Highlights：
•
摘要
This paper considers the computation problems in feature selection. A recursive computation procedure is presented for feature selection and ordering by using the indirect measures such as the Bhattacharyya distance and mutual information. Both binary and quantized measurements are considered. Supporting computer results are provided.
论文关键词：Sequential feature selection,Feature ordering,Bhattacharyya distance,Mutual information,Markov dependence,Recursive algorithm,Computational complexity,Error probability
论文评审过程：Received 11 April 1974, Revised 7 October 1974, Available online 16 May 2003.
论文官网地址：https://doi.org/10.1016/0031-3203(75)90018-7
