| 基于Kalman滤波的储备池多元时间序列在线预报器 |
| |
| 引用本文： | 韩敏,王亚楠.基于Kalman滤波的储备池多元时间序列在线预报器[J].自动化学报,2010,36(1):169-173. |
| |
| 作者姓名： | 韩敏 王亚楠 |
| |
| 作者单位： | 1.大连理工大学电信学院 大连 116024 |
| |
| 基金项目： | 国家高技术研究发展计划(863计划)(2007AA04Z158);;国家自然科学基金(60674073);;大连理工大学2007年数学+X交叉学科建设专项项目资助~~ |
| |
| 摘 要： | 针对多元非线性时间序列, 结合回声状态网络和Kalman滤波提出一种新的在线自适应预报方法. 该方法将Kalman滤波应用于回声状态网络储备池高维状态空间中, 直接对网络的输出权值进行在线更新, 省去了传统递归网络扩展Kalman滤波中Jacobian矩阵的计算, 在提高预测精度的同时令算法的适用范围得到扩展. 在回声状态网络稳定时给出所提算法的收敛性证明. 仿真实例验证了所提方法的有效性.
|
| 关 键 词： | 卡尔曼滤波 递归网络 回声状态网络 多变量序列 预测 |
| 收稿时间： | 2008-12-26 |
| 修稿时间： | 2009-3-24 |
Multivariate Time Series Online Predictor with Kalman Filter Trained Reservoir |
| |
| HAN Min WANG Ya-Nan.School of Electronic , Information Engineering,Dalian University of Technology,Dalian.Multivariate Time Series Online Predictor with Kalman Filter Trained Reservoir[J].Acta Automatica Sinica,2010,36(1):169-173. |
| |
| Authors: | HAN Min WANG Ya-NanSchool of Electronic Information Engineering Dalian University of Technology Dalian |
| |
| Affiliation: | 1.School of Electronic and Information Engineering, Dalian University of Technology, Dalian 116024 |
| |
| Abstract: | A novel online adaptive prediction method is proposed for multivariable nonlinear time series, which is based on echo state network (ESN) and Kalman filtering (KF) algorithm. The KF is adopted in the high-dimension ``reservoir' state space to directly update the output weights of the ESN online. It is implemented without the computation of Jacobian matrices which is in the expanded KF (EKF) algorithm of traditional recurrent neural network (RNN), so as to improve the prediction accuracy and extend the applications. The convergence of the proposed method is proved when the ESN is steady. Simulation examples demonstrate the validity of the proposed method. |
| |
| Keywords: | Kalman filter (KF) recurrent neural network (RNN) echo state network (ESN) multivariate series prediction |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《自动化学报》浏览原始摘要信息 |
|
点击此处可从《自动化学报》下载全文 |
