摘要
提出了一种用于多维函数逼近的进化策略修正泛函网络基函数系数的新算法,并给出了其算法学习过程.利用进化策略的自适应性来确定基函数前的系数,改进了泛函网络的参数通过解方程组来得到这一传统方法.仿真结果表明,这种新的逼近算法简单可行,能够逼近给定的函数到预先给定的精度,具有较快的收敛速度和良好的逼近性能.
An evolutionary functional networks approximation method for multi-dimensional functions is proposed, evolution strategy is used to study the coefficient of the basis functions and a learning algorithm for function approximation is given. Using the adaptive evolution strategy algorithm to define the coefficient before the basis functions, the new Mgorithm improved the traditional method that the functional network＇s parameters are obtained by solving linear equations. The simulation results demonstrate that the approximation method is simple and easy to realize, it can approximate a given continuous function satisfying given precision. The new evolutionary functional networks algorithm has rapid convergence and powerful performance.
出处
《数学的实践与认识》
CSCD
北大核心
2013年第11期231-238,共8页
Mathematics in Practice and Theory
基金
贵州省教育厅科研项目(2010093)
泰州市社会发展计划项目(2011044
2012114)
江苏省高等学校大学生实践创新训练计划项目(2012JSSPITP3029)
南京师范大学泰州学院资助项目(Q201231
Q201232)
关键词
泛函网络
进化策略
进化泛函网络
多维函数逼近
学习算法
functional networks
evolution strategy
evolutionary functional networks
multidimensional function approximation
learning algorithm
