On the expected probability of correct classification

Abstract

SUMMARY The Bayesian approach is applied to examine how the number of features used in a classification problem affects the performance. Exact and asymptotic results are derived for two particular classification problems, demonstrating the positive effect of increasing the number of features.