摘要
构造了一种用于线性时变系统辨识的神经网络 ,研究了它对线性时变控制系统的逼近能力 .在以 L2 ([0 ,t1 ];Rm)的任意一个有界子集为控制函数集上 ,神经网络具有一致逼近线性时变系统的状态的能力 .提出了采用标准正交系作为样本的新的训练方法 .按照这种方法训练后 ,在由这个标准正交系所生成的 L2 ([0 ,t1 ];Rm)的子空间上 。
This paper presents a neural network which can be used to identify linear time varying systems.The capatility of approximating linear time varying control systems on this neural network is studied.On any control functions set which is the bounded subset of L 2([0,t 1];R m),the neural network has the capatility of approximating uniform state of the linear time varying systems.We develep a new training method in which a normal orthogonal system is used as sample.After training according to this method,the output of the neural network can approximate uniform state of the linear time varying system on the subspace of L 2([0,t 1];R m) spun by this normal orthogonal system.
出处
《天津大学学报》
EI
CAS
CSCD
2000年第2期247-251,共5页
Journal of Tianjin University(Science and Technology)
基金
国家自然科学基金资助项目! (697740 12 )
关键词
线性时变系统
系统辨识
神经网络
标准正交系
linear time varying systems
system identification
neural network
normal orthogonal system