Input feature selection by mutual information based on Parzen window

被引:472
作者
Kwak, N [1 ]
Choi, CH
机构
[1] Seoul Natl Univ, Sch Elect Engn & Comp Sci, Seoul, South Korea
[2] Seoul Natl Univ, Automat & Syst Res Inst, Seoul, South Korea
关键词
feature selection; mutual information; Parzen window;
D O I
10.1109/TPAMI.2002.1114861
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Mutual information is a good indicator of relevance between variables, and have been used as a measure in several feature selection algorithms: However, calculating the mutual information is difficult, and the performance of a feature selection algorithm depends on the accuracy of the mutual information. In this paper, we propose a new method of calculating mutual information between input and class variables based on the Parzen window, and we apply this to a feature selection algorithm for classification problems.
引用
收藏
页码:1667 / 1671
页数:5
相关论文
共 17 条
[1]   Weighted Parzen windows for pattern classification [J].
Babich, GA ;
Camps, OI .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 1996, 18 (05) :567-570
[2]   USING MUTUAL INFORMATION FOR SELECTING FEATURES IN SUPERVISED NEURAL-NET LEARNING [J].
BATTITI, R .
IEEE TRANSACTIONS ON NEURAL NETWORKS, 1994, 5 (04) :537-550
[3]  
Cover T. M., 2005, ELEM INF THEORY, DOI 10.1002/047174882X
[4]  
Draper N. R., 1966, APPL REGRESSION ANAL
[5]  
FUKUDA K, 1989, TRENDS PHARM SCI S, V11, P4
[6]   On the behavior of artificial neural network classifiers in high-dimensional spaces [J].
Hamamoto, Y ;
Uchimura, S ;
Tomita, S .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 1996, 18 (05) :571-574
[7]   Input feature selection for classification problems [J].
Kwak, N ;
Choi, CH .
IEEE TRANSACTIONS ON NEURAL NETWORKS, 2002, 13 (01) :143-159
[8]  
Meilhac C, 1999, IEEE INTERNATIONAL CONFERENCE ON MULTIMEDIA COMPUTING AND SYSTEMS, PROCEEDINGS VOL 1, P512, DOI 10.1109/MMCS.1999.779254
[9]  
Murphy P.M., 1994, UCI REPOSITORY MACHI
[10]  
Muto Y, 2000, INT C PATT RECOG, P67, DOI 10.1109/ICPR.2000.906019