粒子群算法在神经网络非线性函数拟合中的应用

The Application of SPSO-BP Neural Network for Nonlinear System Identification

针对传统神经网络学习算法中存在的收敛速度慢、容易陷入局部最优等缺点,设计了基于标准粒子群算法(SPSO)的神经网络非线性函数拟合系统。将神经网络中的权值看作一个粒子,通过粒子之间的竞争与合作以完成网络的学习过程。仿真结果表明,基于SPSO的神经网络学习算法在收敛速度、辨识精度等方面要优于传统的BP神经网络。

The traditional neural network learning algorithms exist in slow convergence and is easy to fall into local optimum and other shortcomings,the design of Standard PSO（SPSO） neural network nonlinear system identification.Neural network weights as a particle,through competition and cooperation between particles in order to complete the network learning process.Simulation results show that,based on SPSO neural network learning algorithm in convergence speed,recognition accuracy is superior to traditional BP neural network..