Bayes Decisions in a Neural Network-PAC Setting
Abstract. In this paper, we investigate the problem of automatic object classification. We assume that each object is represented by a feature vector \bar {x} and belongs to one of finitely many possible object-types T0,...,Tr. Given an object with feature vector \bar {x}, we want to decide to which type it belongs. This decision is, in general, not error-free because objects of different types may occasionally have the same feature vector. Let P(\bar {x}