Determining the minimum cover based on the grouping of class objects

Abstract

The method of dividing the training sample into non-intersecting groups of objects

based on the property of their connection according to the defined subset of boundary

objects of classes is considered. Grouping is used to find the coverage of the sample with

reference objects. The formation of a new feature space for representing objects is

described by nonlinear mapping of non-intersecting set of features onto the number axis.