带时滞的ADALINE神经网络在时变线性系统自适应建模中的应用研究

本文提出了带时滞的ADALINE神经网络对时变复杂线性系统的自适应建模方法,通过对性能函数是权系数二次型的神经网络的算法改进,以梯度变化的共轭向量为搜索方向,以解析式确定最优步长,理论分析和仿真实验证明了此算法的快速性和高精度逼近能力,也证明了基于这种快速算法的带时滞的ADALINE网络实现复杂时变线性系统高精度动态自适应建模的有效性。

ADALINE Neural Network With Time-delay Used in on-time Dynamic and Adaptive Modeling for Time-varying Linear System

This paper puts forward a adaptive modeling method for time-varying linear system adopting the ADALINE neural network with time-delay. Improving on the algorithm of the neural network which performance function is quadratic function for weight coefficient, its search direction is the conjugate direction for gradient changing and its optimal learning rate is obtained by analytical method. Theoretical analysis and simulation results show that the kind of algorithm is an fast and high precision method. They show that the neural network with time-delay based on the rapid algorithm can effectively and accurately carry out dynamic and adaptive modeling for time-varying and complicated linear system.