基于二维函数地貌的遗传算法控制参数优化研究

Research on Optimization of Control Parameters for Genetic Algorithm Based on Fitness Landscape

针对遗传算法求解函数优化问题时控制参数难于确定的问题,提出先将函数按地貌信息聚成不同类,按类给出最佳的控制参数指导知识的解决方案。本文以复杂二维函数优化为例,提取表达函数的地貌信息特征参数,采用模糊C均值法将其聚类,得到适于不同类函数的最佳控制参数的知识。该知识可以指导遗传算法以最佳的控制参数进行函数优化。本研究为获取遗传优化最佳控制参数提供了一种新方法。

With the aim of the problem difficult to determine suitable control parameters for using genetic algorithm to solve the function optimization,it is suggested that the function be first clustered into different types in accordance with the landscape information.So that the solution schemes of optimal control parameter guidance knowledge should be given according to the types.With the optimization of complex binary function as an example,this paper extracts the feature parameters of landscape information to express the function： the fuzzy C-means is adopted to cluster them so as to obtain the optimal control parameter knowledge suitable to different types of functions.This knowledge can guide the genetic algorithm to do the function optimization with the optimal control parameters and provide a new kind of method to obtain the optimal control parameters for genetic optimization.