Abstract
We discuss a problem of formation of prototypes of classes of patterns. An approach studied here takes advantage of techniques of fuzzy sets in forming a variable perception perspective. This in turn implies that the prototypes of the classes are determined in a logical fashion by solving an optimization problem in the unit hypercube. Then the confidence intervals associated with the prototypes are computed in such a way that they reflect the geometry of the classes of the patterns and are directly implied by the value of the performance index achieved during the optimization process. Numerical studies applying synthetic and real-world data enhance capabilities of the approach studied.