FEATURE-SELECTION FOR BEST MEAN-SQUARE APPROXIMATION OF CLASS DENSITIES

被引：5
作者：
PETERS, C
机构：
[1] Department of Mathematics, University of Houston, Houston, TX
来源：
PATTERN RECOGNITION | 1979年 / 11卷 / 5-6期
关键词：
Discriminant analysis; Feature selection; Pattern class separability; Pattern recognition;
D O I：
10.1016/0031-3203(79)90048-7
中图分类号：
TP18 [人工智能理论];
学科分类号：
081104 ; 0812 ; 0835 ; 1405 ;
摘要：
A criterion for feature selection is proposed which is based on mean square approximation of class density functions. It is shown that for the widest possible class of approximants, the criterion reduces to Devijer's Bayesian distance. For linear approximants the criterion is equivalent to well known generalized Fisher criteria. © 1979.
引用
收藏
页码：361 / 364
页数：4
相关论文
共 11 条
12