FUZZY-CONNECTIVES BASED CROSSOVER OPERATORS TO MODEL GENETIC ALGORITHMS POPULATION DIVERSITY

Genetic algorithms are adaptive methods which may be used to solve sea rch and optimization problems. Genetic algorithms process a population of search space solutions with three operations: selection, crossover and mutation. An important problem in the use of genetic algorithms i s the premature convergence in a local optimum. Their main causes are the lack of diversity in the population and the disproportionate relat ionship between exploitation and exploration. The crossover operator i s considered one of the most determinant elements for solving this pro blem. In this paper, we present new crossover operators based on fuzzy connectives for real-coded genetic algorithms. These operators are de signed to avoid the premature convergence problem. To do so, they shou ld keep the right exploitation/exploration balance to suitably model t he diversity of the population. (C) 1997 Elsevier Science B.V.