
6 CHAPTER 1 Classifiers Based on Bayes Decision Theory
1.4 MINIMUM DISTANCE CLASSIFIERS
1.4.1 The Euclidean Distance Classifier
The optimal Bayesian classifier is significant ly simplified under the following assumptions:
• The classes are equiprobable.
• The data in all classes follow Gaussian distributions.
• The covariance matrix is the same for all classes.
• The covariancematrix i s diagonal and all elements across the diagonal are equal.Thatis,S = σ
2
I,
where I is the identity matrix.
Under these assumpt ions, it turns out t hat t he optimal Bayesian classifier is equivalent t o the minimum
Eucl i dean distance classifier. That i s, given an unknown x