摘要
人工神经网络理论是解决非线性问题的有力工具.文章研究了离散型反馈神经网络模型的建立,基于带有输入的非线性离散系统的输出可以被一类反馈神经网络输出神经元的状态向量任意逼近,笔者将已有的反馈神经网络近似非线性离散系统的能力从定常系统推广到时变系统,对反馈神经网络近似非线性离散系统的能力进行了扩展研究,针对于更普遍的非自治的非线性离散系统,证明了它们在有限时间段内的输出轨迹可以被反馈神经网络输出神经元的状态向量近似到任何程度.
The neural networks theory is a more e ffective way to solve nonlinear prob lems,since the output of nonlinear d is-crete-time system with input can be a pproximated by the state vector of th e output units of recurrent neural networks to any degree.This paper studies the model of discrete-time recurrent neural n etworks,then exploits the approxim ation ability of re-current neural networks to nonlinear discrete-time systems.For nonaut onomous nonlinear discrete-time sy stems,which are more common,this paper extends the appro ximation ability of RNN from instant systems to time-variant systems,pr oves that their finite time trajectory can be approximated by the state vector of the output units of recurrent neural networks to any degree.
出处
《宁夏工程技术》
CAS
2003年第1期24-26,共3页
Ningxia Engineering Technology
基金
宁夏大学科学研究基金资助项目(012502)
