Algoritmos evolutivos hídridos com detecção de regiões promissoras em espaços de busca contínuos e discretos / Hybrid evolutionary algorithms with detection of promising areas in continuous discrete search spaces

AUTOR(ES)
DATA DE PUBLICAÇÃO

2004

RESUMO

This work presents three strategies for exploitation in hybrid evolutionary algorithms. These strategies are the base of three approaches: Population Training Heuristic (called TPH), Evolutionary Clustering Search (ECS) and the Parallel Adaptive Hierarchical Fair Competition Genetic Algorithm (called APHAC). The TPH employs problem-specific heuristics for fitness evaluation, guiding the population to settle down in promising search areas where such heuristic would not yield further improvement. In the ECS, an iterative clustering scheme works concurrently with an evolutionary algorithm generating a set of clusters that are used as references to assumed promising search areas. An alternative exploitation mechanism, called assimilation, makes the search strategy more aggressive in the areas framed by cluster. Finally, the APHAC implements a fair competition scheme by segregating individuals with di erent fitness ranges in di erent evolving demes. A heterogeneous evolutionary environment allows to employ a di erent search strategy in each deme. For instance, a local search operator is applied to the individuals of the elite deme. These three approaches are applied to standard test problems, associated to continuous or discrete search spaces. The results obtained with these approaches are similar, or even better, than those found in literature.

ASSUNTO(S)

métodos heurísticos algoritmos genéticos otimização processamento paralelo inteligência artificial heuristic methods genetic algorithms optimization parallel processing artificial intelligence

ACESSO AO ARTIGO

Documentos Relacionados