On the Design of Optimal Zoning for Pattern Classification

Recently, a new technique for zoning design has been proposed in which zoning is considered as the result of an optimization problem. In this case the optimal zoning can be found as the one which minimizes the value of the cost function associated to the classification. In this paper, on the basis of this technique, the cardinality of the set of zones versus the cardinality of the set of classes is experimentally analyzed. The experimental tests, carried out in the field of hand-written numeral and character recognition, show that the best results are obtained when the number of zones is similar to the number of classes, whatever feature set is considered.