Presentation 2005/6/17
Identification of time series signal using DNN with GA-based training method
Daisaku YONEDA, Kunihiko NAKAZONO, Hiroshi KINJO, Tetsuhiko YAMAMOTO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Recently, recurrent neural network (RNN) has attracted more research interest than layered neural network having static mapping capability. However, the structure in RNN is complex compared to that in layered neural network with a training algorithm. In this paper, we propose a dynamical neural network (DNN) that have the properties of inertia, viscosity, and stiffness and its GA-based training method. In a previous research, we propose the DNN which has the dynamical neurons with hidden layer and output layer (DNN1). Also, we propose the simplified DNN which has the dynamical neuron only with hidden layer (DNN2). The GA-based training method is designed to train not only the connecting weights but also the property parameters of the DNN. Simulation results show that the DNN obtains good training performance compared with RNN for time series signals of periodic function.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Dynamical neural network / Recurrent neural network / GA-based training method / Time series signal
Paper # NC2005-26
Date of Issue

Conference Information
Committee NC
Conference Date 2005/6/17(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Identification of time series signal using DNN with GA-based training method
Sub Title (in English)
Keyword(1) Dynamical neural network
Keyword(2) Recurrent neural network
Keyword(3) GA-based training method
Keyword(4) Time series signal
1st Author's Name Daisaku YONEDA
1st Author's Affiliation Graduate school of Engineering and Science, University of the Ryukyus()
2nd Author's Name Kunihiko NAKAZONO
2nd Author's Affiliation Faculty of Engineering, University of the Ryukyus
3rd Author's Name Hiroshi KINJO
3rd Author's Affiliation Faculty of Engineering, University of the Ryukyus
4th Author's Name Tetsuhiko YAMAMOTO
4th Author's Affiliation Faculty of Engineering, University of the Ryukyus
Date 2005/6/17
Paper # NC2005-26
Volume (vol) vol.105
Number (no) 131
Page pp.pp.-
#Pages 6
Date of Issue