WIT Press


Designing Optimized Pattern Recognition Systems By Learning Voronoi Vectors Using Genetic Algorithms

Price

Free (open access)

Paper DOI

10.2495/DATA000401

Volume

25

Pages

10

Published

2000

Size

874 kb

Author(s)

Claudio M.N.A. Pereira and Roberto Schirru

Abstract

Designing optimized pattern recognition systems by learning Voronoi vectors using genetic algorithms Claudio M. N. A. Pereira ^ and Roberto Schirru * ' Institute de Engenharia Nuclear, CREA, Comissdo National de Energia Nuclear, Brazil ^ Programa de Engenharia Nuclear, Universidade Federal do Rio de Janeiro, Brazil. * Computer Science Department, Universidade Igua$u, Brazil. Abstract In this work is described a methodology for developing optimized pattern recognition systems by means of genetic machine learning. The idea is to redefine the set of classes that must be learned by the classification system, mapping them into another set which the Voronoi vectors can successfully classify a sample into one of the original classes. The main objective of this approach is to find the minimum number of classification rules (by minimizing the number of classes in the new set) that maximizes the number of correct classifications. To accomplish that, a genetic algorithm was design. In

Keywords