Identification of nonlinear dynamical systems using recurrent neural networks

Laxmidhar Behera, Swagat Kumar, Subhas Chandra Das

Research output: Contribution to conferencePaper

2 Citations (Scopus)

Abstract

This paper discusses three learning algorithms to train Recurrent Neural Networks for identification of non-linear dynamical systems. We select Memory Neural Networks(MNN) topology for the recurrent network in our work. MNNs are themselves dynamical systems that have internal memory obtained by adding trainable temporal elements to feed-forward networks. Three learning procedures namely Back-Propagation Through Time(BPTT), Real Time Recurrent Learning(RTRL) and Extended Kalman Filtering(EKF) are used for adjusting the weights in MNN to train such networks to identify the plant. The relative effectiveness of different learning algorithms have been discussed by comparing the mean square error associated with them and corresponding computational requirements. The simulation results show that RTRL algorithm is efficient for training MNNs to model nonlinear dynamical systems by considering both computational complexity and modelling accuracy. Eventhough, the accuracy of system identification is best with EKF, but it has the drawback of being computationally intensive.

Original languageEnglish
Pages1120-1124
Number of pages5
Publication statusPublished - 1 Dec 2003
EventIEEE TENCON 2003: Conference on Convergent Technologies for the Asia-Pacific Region - Bangalore, India
Duration: 15 Oct 200317 Oct 2003

Conference

ConferenceIEEE TENCON 2003: Conference on Convergent Technologies for the Asia-Pacific Region
CountryIndia
CityBangalore
Period15/10/0317/10/03

Fingerprint

Nonlinear dynamical systems
Recurrent neural networks
Learning algorithms
Data storage equipment
Neural networks
Backpropagation
Mean square error
Computational complexity
Identification (control systems)
Dynamical systems
Topology

Cite this

Behera, L., Kumar, S., & Das, S. C. (2003). Identification of nonlinear dynamical systems using recurrent neural networks. 1120-1124. Paper presented at IEEE TENCON 2003: Conference on Convergent Technologies for the Asia-Pacific Region, Bangalore, India.
Behera, Laxmidhar ; Kumar, Swagat ; Das, Subhas Chandra. / Identification of nonlinear dynamical systems using recurrent neural networks. Paper presented at IEEE TENCON 2003: Conference on Convergent Technologies for the Asia-Pacific Region, Bangalore, India.5 p.
@conference{1816bfd1cdc94641bd281aa9f73f0ad3,
title = "Identification of nonlinear dynamical systems using recurrent neural networks",
abstract = "This paper discusses three learning algorithms to train Recurrent Neural Networks for identification of non-linear dynamical systems. We select Memory Neural Networks(MNN) topology for the recurrent network in our work. MNNs are themselves dynamical systems that have internal memory obtained by adding trainable temporal elements to feed-forward networks. Three learning procedures namely Back-Propagation Through Time(BPTT), Real Time Recurrent Learning(RTRL) and Extended Kalman Filtering(EKF) are used for adjusting the weights in MNN to train such networks to identify the plant. The relative effectiveness of different learning algorithms have been discussed by comparing the mean square error associated with them and corresponding computational requirements. The simulation results show that RTRL algorithm is efficient for training MNNs to model nonlinear dynamical systems by considering both computational complexity and modelling accuracy. Eventhough, the accuracy of system identification is best with EKF, but it has the drawback of being computationally intensive.",
author = "Laxmidhar Behera and Swagat Kumar and Das, {Subhas Chandra}",
year = "2003",
month = "12",
day = "1",
language = "English",
pages = "1120--1124",
note = "IEEE TENCON 2003: Conference on Convergent Technologies for the Asia-Pacific Region ; Conference date: 15-10-2003 Through 17-10-2003",

}

Behera, L, Kumar, S & Das, SC 2003, 'Identification of nonlinear dynamical systems using recurrent neural networks' Paper presented at IEEE TENCON 2003: Conference on Convergent Technologies for the Asia-Pacific Region, Bangalore, India, 15/10/03 - 17/10/03, pp. 1120-1124.

Identification of nonlinear dynamical systems using recurrent neural networks. / Behera, Laxmidhar; Kumar, Swagat; Das, Subhas Chandra.

2003. 1120-1124 Paper presented at IEEE TENCON 2003: Conference on Convergent Technologies for the Asia-Pacific Region, Bangalore, India.

Research output: Contribution to conferencePaper

TY - CONF

T1 - Identification of nonlinear dynamical systems using recurrent neural networks

AU - Behera, Laxmidhar

AU - Kumar, Swagat

AU - Das, Subhas Chandra

PY - 2003/12/1

Y1 - 2003/12/1

N2 - This paper discusses three learning algorithms to train Recurrent Neural Networks for identification of non-linear dynamical systems. We select Memory Neural Networks(MNN) topology for the recurrent network in our work. MNNs are themselves dynamical systems that have internal memory obtained by adding trainable temporal elements to feed-forward networks. Three learning procedures namely Back-Propagation Through Time(BPTT), Real Time Recurrent Learning(RTRL) and Extended Kalman Filtering(EKF) are used for adjusting the weights in MNN to train such networks to identify the plant. The relative effectiveness of different learning algorithms have been discussed by comparing the mean square error associated with them and corresponding computational requirements. The simulation results show that RTRL algorithm is efficient for training MNNs to model nonlinear dynamical systems by considering both computational complexity and modelling accuracy. Eventhough, the accuracy of system identification is best with EKF, but it has the drawback of being computationally intensive.

AB - This paper discusses three learning algorithms to train Recurrent Neural Networks for identification of non-linear dynamical systems. We select Memory Neural Networks(MNN) topology for the recurrent network in our work. MNNs are themselves dynamical systems that have internal memory obtained by adding trainable temporal elements to feed-forward networks. Three learning procedures namely Back-Propagation Through Time(BPTT), Real Time Recurrent Learning(RTRL) and Extended Kalman Filtering(EKF) are used for adjusting the weights in MNN to train such networks to identify the plant. The relative effectiveness of different learning algorithms have been discussed by comparing the mean square error associated with them and corresponding computational requirements. The simulation results show that RTRL algorithm is efficient for training MNNs to model nonlinear dynamical systems by considering both computational complexity and modelling accuracy. Eventhough, the accuracy of system identification is best with EKF, but it has the drawback of being computationally intensive.

UR - http://www.scopus.com/inward/record.url?scp=2342519964&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=2342519964&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:2342519964

SP - 1120

EP - 1124

ER -

Behera L, Kumar S, Das SC. Identification of nonlinear dynamical systems using recurrent neural networks. 2003. Paper presented at IEEE TENCON 2003: Conference on Convergent Technologies for the Asia-Pacific Region, Bangalore, India.