DocumentCode :
1115248
Title :
Estimation of Mutual Information in Two-Class Pattern Recognition
Author :
Butler, G.A. ; Ritea, H. Barry
Author_Institution :
Judson B. Branch Research Center, Allstate Insurance Company
Issue :
4
fYear :
1974
fDate :
4/1/1974 12:00:00 AM
Firstpage :
410
Lastpage :
420
Abstract :
Although mutual information (MI) has been proposed for some time as a measure of the dependence between the class variable and pattern recognition features, it is only recently that the practical problems of designing computer programs to use MI have been raised. Within the two-class context, this paper compares two traditional approaches to the requisite entropy estimation (using the maximum likelihood and expected value estimators of class probabilities) with a new estimator: the expected value of binomial entropy (E). The latter is shown to be superior where one class has a priori dominance. E is also related to expected probability of error and, in a surprising result, it is shown that E is a better estimator of class probabilities than the maximum likelihood and expected value estimators over a wide range.
Keywords :
Binomial distribution, entropy, feature selection information, mutual information, nonparametric classifier design, pattern recognition, two-class sampling.; Atomic measurements; Entropy; Insurance; Maximum likelihood estimation; Multidimensional systems; Mutual information; Pattern recognition; Random variables; Sampling methods; Time measurement; Binomial distribution, entropy, feature selection information, mutual information, nonparametric classifier design, pattern recognition, two-class sampling.;
fLanguage :
English
Journal_Title :
Computers, IEEE Transactions on
Publisher :
ieee
ISSN :
0018-9340
Type :
jour
DOI :
10.1109/T-C.1974.223956
Filename :
1672549
Link To Document :
بازگشت