遗传算法参数和操作的序优化

来源期刊：中南大学学报(自然科学版)2003年第4期

论文作者：王凌 张亮 唐芳

文章页码：350 - 354

关键词：遗传算法;序优化;最优计算量分配;随机优化

Key words：genetic algorithm; ordinal optimization; optimal computing budget allocation; stochastic optimization

摘    要：遗传算法的优化质量和效率很大程度上依赖于算法参数和操作,包括种群大小、交叉和变异概率、选择、交叉和变异操作等,然而确定合适的参数和操作通常需要大量的试验和比较,需确定最佳参数和操作.为此,基于序优化和最优计算量分配技术,通过将问题描述为随机优化问题,提出了一种确定有限计算量下最佳GA参数和操作的系统性方法.仿真研究表明,该方法能够合理地确定最佳参数和操作,并同时给出GA的合理性能评价.

Abstract: The optimization quality and efficiency of genetic algorithm highly depend on its parameters and operators, such as population size, crossover and mutation probabilities, selection, crossover and mutation operators.However, it often needs trial and error to determine suitable parameters and operators, so it is still an important open problem on GA research. Based on ordinal optimization and optimal computing budget allocation, the considered issue was regards as a stochastic optimization problem and a systematical method to determine optimal parameters and operators for GA with limited computation effort was proposed. Simulation results show that the proposed approach is able to reasonably determine optimal parameters and operators and provide suitable evaluation for GA simultaneously.