遗传算法参数自适应控制的新方法

New Adaptive Control Strategies for Parameters of Genetic Algorithms

根据遗传算法参数自适应控制方法的不同分类,采用基于启发式规则的参数控制方法对遗传算法的种群数进行了宏观调控和微观调控。并采用不同特点的模糊控制器分别控制交叉率和变异率,使种群数、交叉率和变异率都能够随进化的实际情况发生自动调整,形成了一种新的种群数变化的模糊自适应遗传算法。实验数据表明这种算法能够有效防止遗传算法早收敛,同时也说明对参数进行自适应控制能够使遗传算法性能大大提高。

Adaptive control strategies for parameters of genetic algorithms can be built according to the heuristic rules or artificial intelligent techniques. A fuzzy adaptive genetic algorithm with variable population size （FAGAVPS） is proposed to overcome premature convergence and slow convergence speed at later evolution process of simple genetic algorithm. FAGAVPS uses both macroscopical and microscopical control based on heuristic rules to realize population size adaptation. The crossover rate and mutation rate of the algorithm are also tuned automatically in the evolutionary process by two fuzzy controllers with different characteristics. The experiments show that FAGAVPS can efficiently avoid premature convergence and the performance of genetic algorithm can be greatly raised by applying adaptive control strategies.