作　　者： ; ; ; ; ;

机构地区： 惠州学院

出　　处： 《计算机科学》 2011年第9期197-200,共4页

摘　　要： 连续型进化算法的计算时间分析是目前国内外研究的难题,对此研究了Lévy变异进化规划(evolutionary programming based on Lévy mutation,LEP)算法的计算时间分析理论。具体的分析步骤如下:首先在将LEP算法建模为吸收态Markov过程的基础上,证明了LEP算法的收敛性;然后,结合LEP算法选择算子的特点,以首达最优解的期望时间作为计算时间分析的主要指标;最后,利用Lévy分布的近似变形给出LEP算法计算时间的估计式。研究结果表明,最优解空间的Lebesgue测度、算法的种群规模和搜索范围对计算时间有直接影响。 Running-time analysis of the continuous evolutionary algorithm is a difficult problem now existing in the field at home and abroad.To deal with this issue,the paper gave an in-depth studies about the running time of evolutionary programming based on Lévy mutation（LEP）.The procedure is as follows：First,LEP algorithm was modeled on the basis of an absorbing Markov process,which proved the optimal solution of the LEP convergence.Second,the expected first hitting time was used to evaluate the running time of LEP algorithm by taking its computational properties into consideration.Finally,based on the similar transformation of Lévy distribution,an estimation equation of LEP running time was proposed.The research results indicate that the upper bounds for the running time are directly influenced by the Lebesgue measurement of the optimal space,its population scale and the searching range.

关 键 词： 人工智能 进化计算 进化规划 计算时间 变异

领　　域： [自动化与计算机技术] [自动化与计算机技术]