基于改进粒子群优化RBF神经网络的算法

摘要

In view of the defect of particle swarm optimization which easily gets into partial extremum, an improved particle swarm optimization algorithm is put out,and the algorithm is applied to the parameter selecting of RBF neural network kernel function. The best parameter vector is searched in the whole space, according to coding means, iterative formula, fitness function which are mentionedin the paper. The proves that RBF neural network based on improved PSO has faster convergent speed, and higher error precision.%针对粒子群算法易陷入局部极小的缺陷,提出了一种改进的粒子群优化算法,并将改进后的算法应用到RBF神经网络核函数参数的选取中.依照文中提出的编码方式、迭代公式和适应度函数,在全局空间中搜索具有最优适应值的参效向量.实例仿真表明,基于改进粒子群算法优化的RBF神经网络不仅收敛速度快,且误差精度高.