Cluster analysis of genetic algorithms results

• Autores: Katercyna Adamska
• Localización: Inteligencia artificial: Revista Iberoamericana de Inteligencia Artificial, ISSN-e 1988-3064, ISSN 1137-3601, Vol. 9, Nº. 28, 2005 (Ejemplar dedicado a: Nuevas tendencias en Sistemas Multiagente y Soft Computing), págs. 25-32
• Idioma: español
• Enlaces
  • Texto completo
• Resumen

  • The work is concerned on the problem of approximation of central parts of basins of attraction of an objective in continuous global optimization problems. It presents the general strategy of Clustered Genetic Search (CGS), which consists in finding clusters in a genetic sample to get the approximations of basins of attraction of an objective. DR-CGS is an instance of CGS which utilizes a construction of a Finite Mixture Model of normal componenets as a clustering method. CR-CGS brings wide opportunities of asymptotic analysis. Due to features of a normal mixture, it also allows for very easy definition of approximations of basins of attraction Presented computational tests illustrate how the method works and are a practical evidence of its good results.