摘要
基于全局搜索的进化算法和一种局部搜索算法——结构化的非线性参数优化方法(SNPOM),提出两种混合的优化算法来估计RBF神经网络中的参数:1)初始化一定数目的种群作为SNPOM的初始值得到其适应值,通过选择、交叉和替换策略来更新种群;2)采用进化算法运行一定的代数,从最终群体中选取一些个体进一步用SNPOM来优化.这两种混合优化算法的本质是用进化算法为SNPOM搜寻最优初始值,以得到全局最优解.仿真实验结果表明,该混合算法比单独使用进化算法或SNPOM更优,且优于其他一些算法.
Based on an evolutionary algorithm (EA) and a local search strategy - the structured nonlinear parameter optimization method (SNPOM), two hybrid parameter optimization algorithms for RBF neural networks are proposed. The first approach starts with a population of some random initial parameter values, and updates the population by selection, crossover and replacement according to the fitness values obtained by using SNPOM. The second method runs the EA for a reasonable amount of generations, after which the SNPOM is used to locate the refined local optimum. The basic idea of the two hybrid algorithms is to find the optimal initial values for SNPOM using EA. The simulation tests show that the combination method provides better results than either the single method (EA and SNPOM) or some other existing algorithms.
出处
《控制与决策》
EI
CSCD
北大核心
2009年第8期1172-1176,共5页
Control and Decision
基金
国家自然科学基金项目(60443008)
湖南省自然科学基金项目(07JJ3126)
关键词
RBF神经网络
参数估计
混合优化方法
RBF neural network
Parameter estimation
Hybrid optimization approach