BAYESC

Bayes classifier based on given density estimates

     W = BAYESC(WA,WB, ... ,P,LABLIST)
    D = X*W

Input
WA,	WB, ... Trained mappings for supplying class density estimates
P	Vector with class prior probabilities  Default: equal priors
LABLIST	List of class names (labels)
X	Testset

Output
W	Bayes classifier.
D	Classification matrix

Description

The trained mappings WA,WB, ... should supply proper densities estimates  D for a dataset X by D = X*WA, etcetera. E.g. they should be trained by  commands like GAUSSM(A), PARZENM(A), KNNM(A). Consequently, they should  have a size of K x 1 (assuming that X and A are K-dimensional). Also  sizes of K x N are supported, assuming a combined density estimate for N classes simultaneously. BAYESC weighs the class densitites by the class  priors in P and names the classes by LABLIST. If LABLIST is not supplied,  the labels stored in the mappings are used.

Reference(s)

1. R.O. Duda, P.E. Hart, and D.G. Stork, Pattern classification, 2nd edition, John Wiley and Sons, New York, 2001.
2. A. Webb, Statistical Pattern Recognition, John Wiley && Sons, New York, 2002.

See also

datasets, mappings, gaussm, parzenm, knnm,