A ROBUST BP ALGORITHM AND ITS APPLICATION ON THE IDENTIFICATION OF NONLINEAR DYNAMIC SYSTEM
-
摘要: 利用多层前馈神经网络的非线性建模特性，基于动态BP网络的串并联和并联模型，提出一种高鲁棒性BP算法．与传统的BP算法相比，鲁棒BP算法有5个优点：（1）适合于非线性动态系统辨识；（2）辨识精度高；（3）不必内插所有训练样本；（4）具有高鲁棒性，能抵制过失误差和量测误差；（5）收敛速度得到了改进，因为错误样本的影响得到了适度的抑制．把该算法用于非线性动态系统辨识，仿真结果表明此方法是有效的．Abstract: The paper presents a high robust BP algorithm using the nonlinear of multilayer feedforward neural networks and based on the series parallel model of the dynamic BP network. In contrast to the conventional BP algorithm,five advantages of the robust BP algorithm are:(1) fitting to the dynamic identification of nonlinear system;(2) the identifical accuracy is very high;(3) not interpolating all the training points;(4) it is robust against gross errors and measuring errors;(5) its rate of convergence is improved since the influence of incorrect sample is gracefully suppressed.The algorithm is applied to the dynamic identification of nonlinear system and the simulation result shows the new method is efficient.
-
-
1 Funahashi K I.On the Approximate Realization of Continuous Mapping by Neural Networks.Neural Networks,1989,(2):183～192
2 Narendra K S,Parthasarat hy K.Identification and Control of Dynamic Systems Using Neural Net works.IEEE Trans Neural Network,1990,1(1):4～27
3 田明等.基于动态BP神经网络的系统辨识方法.自动化学报,1993,19(4):450～453
4 Hampel F R,Rousseeuw P J,Rouchetti E,Stahel W A.Robust Statistics-The Approach Based on Influence Function.New York:Wiley,1986
5 Efrom B,Tibshirani R.Bootstrap Methods for Standard Errors Confidence Intervals and Other Measures of Statistical Accuracy.Statist Sci,1988,1:54～77
6 Cun Y Le.A Theoretical Framework for Back propagation.In:D Touretzky,G Hinton,T Sejnowski Eds,Proc 1988 Connectionist Models Summer School,San Mateo,CA:Morgan Kanfmann,1988:21～28
计量
- 文章访问数: 817
- HTML全文浏览量: 0
- PDF下载量: 279
下载:
