一种混沌神经网络的混沌时间序列预测  

Chaotic Neural Network Approach to Chaotic Time Series Prediction

作　　者：王燚[1] 郭伟[1] 

机构地区：[1]电子科技大学通信抗干扰国家重点实验室

出　　处：《电子测量与仪器学报》2007年第6期5-9,共5页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金资助项目(编号:10577007)

摘　　要：混沌是一种普遍存在的非线性动力学行为,针对混沌时间序列预测问题,提出了一种新颖的混沌对角递归神经网络模型,然后,给出了易实现的动量梯度学习算法。为了验证网络的预测性能,采用该神经网络预测模型对Mackey-Glass混沌时间序列进行了仿真。结果表明,所提出的混沌神经神经网络模型能快速、精确地预测混沌时间序列,并能在一定精度上满足多步预测需要,是研究复杂非线性动力系统辨识和控制的一种有效方法。Chaos is a nonlinear dynamic behavior existed generally. Aiming at chaotic time series prediction, the paper presents a novel chaotic diagonally recurrent neural network approach and also presents a momentum gradient back-propagation training algorithm, which can be implemented effectively. In order to evaluate the prediction performance of the network, Mackey-Glass chaotic series simulation is carried out using this model. Simulation results indicate that the presented network model can make a rapid and accurate prediction of the chaotic series, and reconstruct the original feature space, which shows that the model can resume the series trajectory. Further more the network model can realize multiple step prediction in certain accurate degree, which is an effective method for studying the recognition and control of the complicated nonlinear dynamic system.

关 键 词：混沌对角递归神经网络 动量梯度学习算法 混沌时间序列 预测 

分 类 号：TN911[电子电信—通信与信息系统]