基于回声状态网络的混沌时间序列多步预测
首发时间:2008-12-26
摘要:针对回声状态网络在权值学习过程中存在的病态问题,提出了一种基于正则化方法的回声状态网络学习算法。在标准误差项基础上增加一个限制逼近函数复杂性的正则化项,从而对解的精度和平滑性进行折中,使得神经网络冗余的连接权在学习的过程中逐渐衰减到零值附近,在保持学习精度的前提下神经网络结构更加精简,泛化能力得到提高。仿真实验中通过Lorenz映射所产生的时间序列对所提出的算法进行验证,结果表明该算法在直接多步预测中比RBF神经网络预测模型有更高的预测精度。
For information in English, please click here
Multi-Steps Prediction of Chaotic Time Series Based on Echo State Network
Abstract:Considering of the ill-posed problem of in learning process of echo state network(ESN), a learning algorithm of ESN is proposed based on regularization method. The regularization term provides a stable solution to function approximation with a tradeoff between accuracy and smoothness of the solutions. So the redundant weights of neural network are damped and converge to the zero state. The structure of neural network become more compact in a particular accuracy. The neural network has a good generalization. The simulation results show that the proposed algorithm has higher accuracy than the prediction model based on RBF network in a multi-steps prediction by Lorenz map.
Keywords: chaos prediction;echo state network;regularization;phase space reconstruction
论文图表:
引用
No.2699237878012302****
同行评议
共计0人参与
勘误表
基于回声状态网络的混沌时间序列多步预测
评论
全部评论0/1000