Branch and bound algorithm for the Bayes classifier
Lui Sze, CH Leung · 1996
Given the feature vector from an unknown class, the branch and bound algorithm (BAB) is very efficient for finding the nearest neighbor among the set of reference vectors. The Euclidean distance measure is adopted. In this article, the BAB algorithm is extended so that it can be used with the Bayes classifier which uses the probability measure instead of the Euclidean distance for classification. Gaussian statistics is assumed in the derivations. Satisfactory results are obtained in recognition experiments.