摘要
分析了一般神经网络控制系统中学习误差选择的问题,指出系统误差不等于理论上神经网络用于学习的误差,因而网络的性能会受到影响。进而针对局部逼近神经网,提出了一种改进的控制器结构,并讨论了其学习算法。仿真实验研究表明该方法收敛速度快,学习能力强,证明其在系统控制中的合理性和有效性。
There is a problem in neural networks control system. The error that is adopted by learning methods in neural networks in theory does not equal the system's error in reality, so the behavior of the neural networks will be affected. A new control structure that can settle the question to some extent is put forward in this paper, and the corresponding learning methods are discussed. Through the simulation experiments research, the reasonableness and effectiveness of the method is proved.
出处
《系统仿真学报》
EI
CAS
CSCD
2001年第6期730-731,共2页
Journal of System Simulation
基金
青年骨干教师资助计划资助