Título: Evolução Lamarckiana Baseada em Programação Evolutiva e Método de Hooke-Jeeves Aplicada à Otimização de um Sistema Nebuloso
Autores: Coelho, Leandro dos Santos; Mariani, Viviana Cocco
Resumo: An approach to design of Takagi-Sugeno-Kang fuzzy systems via Lamarckian evolution is presented based on antecedent and consequent parts optimization. The new conception of Lamarckian evolution is a combination of fast evolutionary programming and Hooke-Jeeves optimizaton method. The proposed methodology involves two levels of optimization, namely (fast evolutionary programming) and individual learning (Hooke-Jeeves method), which cooperate in a global procedure of optimization. The fuzzy system is evaluated in two case studies: (i) prediction of chaotic system called Lozi map, and (ii) prediction of a physiological control system (Mackey-Glass system). The simulation results indicate that the fuzzy system optimization based on Lamarckian evolution presents enhancement in solution quality, reliability, and accuracy properties for nonlinear prediction of chaotic time series.
Código DOI: 10.21528/CBRN2003-055
Artigo em PDF: 6CBRN_055.PDF
Arquivo BibTex: 6CBRN_055.bib