计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (35): 46-48.

• 学术探讨 • 上一篇    下一篇

RBF网络的微分进化正交最小二乘算法

苏美娟,邓 伟   

  1. 苏州大学 计算机科学与技术学院,江苏 苏州 215006
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2007-12-11 发布日期:2007-12-11
  • 通讯作者: 苏美娟

Differential Evolution Orthogonal Least Square(DEOLS) algorithm for RBF network

SU Mei-juan,DENG Wei   

  1. School of Computer Science & Technology,Soochow University,Suzhou,Jiangsu 215006,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2007-12-11 Published:2007-12-11
  • Contact: SU Mei-juan

摘要: 研究用于径向基函数(RBF)网络训练的一种微分进化正交最小二乘(DEOLS)算法。把微分进化(DE)算法的种群作为正交最小二乘(OLS)算法的候选径向基函数集合,利用OLS对DE的种群个体进行评断,以确定RBF网络的隐结点的数目、中心和宽度。该算法融合了DE的强大搜索能力和OLS的高效评断能力,隐结点的选择比OLS要合理,同时避免DE的复杂性。最后使用实验验证了该算法的优越性。

关键词: RBF神经网络, 正交最小二乘方法, 微分进化算法

Abstract: Differential Evolution Orthogonal Least Square(DEOLS) algorithm for RBF network is presented,a hybrid algorithm blending Orthogonal Least Squares(OLS) method with Differential Evolution(DE) algorithm.The population of the DE algorithm,encoding the center and width of RBF hidden node,corresponds to the set of candidate RBFs of OLS,while DE exploits OLS to evaluate each individual of the population.This algorithm compromises the strong search power of DE and the efficient evaluation ability of OLS,selecting hidden nodes more rationally than OLS,without incurring the computational cost of DE.At last,experiments are used to demonstrate the effectiveness of this algorithm.

Key words: Radial Basis Function(RBF) network, Orthogonal Least Square(OLS) method, Differential Evolution(DE) algorithm