项目名称: 进化规划算法的计算时间难题研究
项目编号: No.61003066
项目类型: 青年科学基金项目
立项/批准年度: 2011
项目学科: 金属学与金属工艺
项目作者: 黄翰
作者单位: 华南理工大学
项目金额: 7万元
中文摘要: 进化算法的计算时间分析是进化计算领域的公开难题。现有研究大多数集中在(1+1)EA 等简单离散 优化进化算法的时间复杂性,较少涉及连续优化进化算法的计算时间。由于进化规划算法是一类重要 的连续优化进化算法,是许多新型进化算法设计的原型,因此本项目着重分析进化规划算法收敛于最 优邻域的计算时间。主要工作包括:分析Gauss 变异、Cauchy 变异、Lé 变异、差异变异等进化规 划算法的计算时间上下界;建立进化规划算法的多项式和指数式时间收敛的判定条件;分析算法参数 与计算时间的关系。本项目还计划将研究结论用于分析与改进进化规划、差异进化算法、粒子群优化 算法等连续型进化算法。
中文关键词: 进化规划算法;进化计算;收敛;计算时间;最优邻域
英文摘要: Running time analysis of evolutionary algorithm (EA) has been an open problem of evolutionary computation for several years. Although most of the researches focus on easy cases like (1+1)EA, there is less result on running time analysis of continuous evolutionary algorithm. Evolutionary programming (EP) is one of the most important continuous evolutionary algorithms, as a basic model of other EAs' improvement, so that our project mainly works on the running time of EP converging to the optimal neighborhood. We planned to analyze the running time bounds of EPs based on Gaussian mutation, Cauchy mutation, Lé mutation and individual difference. Moreover, we will propose the results of EP converging in polynomial or exponential running time to study how parameters impact the running time of EP. The results of our project will be also used to improve EPs and other continuous EAs like difference evolution, particle swarm optimization and so on
英文关键词: Evolutionary Programming Algorithm; Evolutionary Computation; Convergence; Running Time; Continuous Optimization