摘要
为了提高回声状态网对时间序列的预测精度,将改进的小世界网络和泄露积分型回声状态网结合,提出了一种新型时间序列预测方法.泄露积分型回声状态网储备池神经元采用随机网络进行连接,首先利用改进的小世界网络替代随机网络,提高了储备池的适应性,从而改善回声状态网的泛化能力和稳定性.然后将利用改进的回声状态网预测典型的非线性时间序列.最后利用Matlab仿真软件进行验证,仿真结果表明,该方法较传统回声状态网预测模型具有更高的效率和预测精度.
In order to improve the prediction accuracy of echo state network for time series, this paper pro-poses an optimization method by combining modified small world network with the leaky echo state network.The neurons of reservoir of leaky echo state network adopt randomly connected network.Firstly,using the modified small world network instead of random network, A small world network is used to improve connection mode of reservoir processing unit and boost the adaptability of reservoir,thus generalization ability and stability of echo state network are improved.Next,the improved echo state network model is used to predict the typical nonlinear time series.Finally,Matlab software is used to verify in this paper.The simulation results show that the method proposed in this paper has faster convergence speed and higher precision than the traditional echo state network .
出处
《渤海大学学报（自然科学版）》
CAS
2015年第3期284-288,共5页
Journal of Bohai University:Natural Science Edition
基金
教育部"新世纪优秀人才支持计划"项目(No:NCET-11-1005)
2011年辽宁省第一批次科学计划项目(No:2011402001)
辽宁省自然科学基金项目(No:2014020143)
辽宁省百千万人才工程项目(No:2012921061)
关键词
时间序列
回声状态网
小世界网络
泄漏
time series
echo state network
small world network
leaky
