RBFN输出层学习率的一种稳定算法  

A Stable Arithmetic for the Output-layer Learning Rate of RBFN

作　　者：郦小敏[1] 王卫红[1] 王宗学[1] 

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,北京100083

出　　处：《计算机仿真》2004年第2期95-97,共3页Computer Simulation

摘　　要：该文运用Lyapunov稳定性原理,针对RBF网络作为辨识器和控制器,分别推导出其输出权值学习率更新的稳定算法。然后以某转台位置跟踪系统为对象进行辨识器的仿真,结果显示辨识效果大大改善;对于作为控制器的情况,选用[4]中的仿真实例进行仿真验证,结果表明,该方法对控制性能有一定改进。Lyapunov stability theory is used in this paper to reason a kind of stable arithmetic for the output-layer weight of RBFN when it's used as identifier and controller respectively.Then,according to the result,a certain kind of turntable was choosed as the object for simulation.The result shows that the method applied has promoted the effect of identification a lot.As for the controller,the example in paper^([4]) is used in the simulation,and the result shows this arithmetic can promote the control effect in a certain degree.

关 键 词：神经网络 智能特性 RBF网络 输出层学习率 稳定算法 权值调整 仿真 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.9[自动化与计算机技术—控制科学与工程]