摘要
1 引言在许多动态系统中,系统的输入输出数据容易得到,而其模型难以确定,使得系统辨识和控制变得非常困难。多层前向神经刚络可以通过输入输出数据点对逼近任意连续函数,为这个问题的解决提供了一种方法。但传统前向神经网络的BP学习算法存在以下问题:
This paper presents a dynamic modeling method based on orthogonal neural network,it fully uses the characteristics of the nonlinear processing ability of neural networks and the efficient disposal of the large scaling sparse problems that Givens transform can process. It can not only train the network quickly,but also can optimize the structure of the networks. Simulating experiments show that the new modeling method is a simple universal modeling method for the nonlinear systems.
出处
《计算机科学》
CSCD
北大核心
2000年第1期62-64,61,共4页
Computer Science
基金
浙江大学CAD8&CG国家重点实验室
