Deep belief echo-state network and its application to time series prediction

作者：

Highlights：

• A novel deep neural network architecture called DBEN is presented for time series prediction.

• The deep model contains a pre-train network and a novel regression layer.

• The ESN methodology is introduced in the regression layer to perform a local fine-tuning.

• Experimental results show considerable improvements in the prediction performance, learning speed and STM.

摘要

•A novel deep neural network architecture called DBEN is presented for time series prediction.•The deep model contains a pre-train network and a novel regression layer.•The ESN methodology is introduced in the regression layer to perform a local fine-tuning.•Experimental results show considerable improvements in the prediction performance, learning speed and STM.

论文关键词：Deep belief network,Echo state network,Memory capacity,Time series prediction

论文评审过程：Received 19 November 2016, Revised 19 May 2017, Accepted 23 May 2017, Available online 25 May 2017, Version of Record 6 June 2017.

论文官网地址：https://doi.org/10.1016/j.knosys.2017.05.022