基于并行PSO算法的RBF建模

RBF Modeling Based on Parallel Particle Swarm Optimization

针对RBF网络的建模问题,提出了基于并行PSO算法的RBF网络建模方法。其中,隐层单元数由一系列随机产生的整数训练得到;中心向量从输入样本空间内随机选择。随后,通过误差适应度来评价全局最优粒子,进而实现网络性能。从对非线性系统的仿真效果看,该方法隐层单元数比较少,与相同隐层的RBF网络相比,显示出了一定的优越性。

To sovle the modeling problem of RBF network,the parallel PSO algorithm based on RBF network is proposed.Among them,the number of hidden units by a series of randomly generated integers training will be received;center vectors will be selected randomly.Then through the error fitness to evaluate the global optimum particles,thus network performance will be achieved.From the simulation results of the nonlinear system to see that the method ultimately determined the number of hidden units is relatively small,compared to the same hidden layer of RBF network,showing a certain superiority.