摘要
本文从最优化原理出发,提出了一种适合于非线性方程组求解的神经元网络方法,把它应用到AR和MA模型参数估计中,取得了满意的结果.由于该方法采用了自适应变权连续Hopfield神经元网络,具有速度快、精度高等特点.
Based on the optimization theory, a neural network with variable weights is proposed for solving a class cf nonlinear equations in the form of F(X)X Y.The advantage of this method over traditional methods is that the higher accuracy and the higher speed of convergence can be achieved.It is applied to the estimation of parameters of AR and MA models, the satisfactory results are obtained.
出处
《北方交通大学学报》
CSCD
北大核心
1992年第1期16-21,共6页
Journal of Northern Jiaotong University
关键词
非线性
方程组
神经元网络
解法
nonlinear equations
time series analysis
autoregressive moving average model/artificial neural network
