Pii: S0031-3203(00)00059-5

To this end, the so-called Bayes decision rule assigns each pattern x to the class for which the a posteriori probability P(u i Dx) is maximum. An error probability lower than the one provided by the above Bayes rule can be obtained using the so-called `rejecta option. Namely, the patterns that are the most likely to be misclassi"ed are rejected (i.e., they are not classi"ed); they are then handled by more sophisticated procedures (e.g., a manual classi"cation is performed). However, handling high reject rates is usually too time-consuming for application purposes. Therefore, a trade-o! between error and reject is mandatory. The formulation of the best error-reject trade-o! and the related optimal reject rule was given by Chow [1]. According to Chow's rule, a pattern x is rejected if