Generalization ability of universal learning network by using second order derivatives

M. Han, K. Hirasawa, J. Hu, Junichi Murata

Research output: Contribution to journalConference article

4 Citations (Scopus)

Abstract

In this paper, it is studied how the generalization ability of modeling of the dynamic systems can be improved by taking advantages of the second order derivatives of the criterion function with respect to the external inputs. The proposed method is based on the regularization theory proposed by Poggio, but a main distinctive point in this paper is that extension to dynamic systems from static systems has been taken into account and actual second order derivatives of the Universal Learning Network have been used to train the parameters of the networks. The second order derivatives term of the criterion function may minimize the deviation caused by the external input changes. Simulation results show that the method is useful for improving the generalization ability of identifying nonlinear dynamic systems using neural networks.

Original languageEnglish
Pages (from-to)1818-1823
Number of pages6
JournalProceedings of the IEEE International Conference on Systems, Man and Cybernetics
Volume2
Publication statusPublished - Dec 1 1998
EventProceedings of the 1998 IEEE International Conference on Systems, Man, and Cybernetics. Part 2 (of 5) - San Diego, CA, USA
Duration: Oct 11 1998Oct 14 1998

Fingerprint

Dynamical systems
Derivatives
Neural networks

All Science Journal Classification (ASJC) codes

  • Control and Systems Engineering
  • Hardware and Architecture

Cite this

Generalization ability of universal learning network by using second order derivatives. / Han, M.; Hirasawa, K.; Hu, J.; Murata, Junichi.

In: Proceedings of the IEEE International Conference on Systems, Man and Cybernetics, Vol. 2, 01.12.1998, p. 1818-1823.

Research output: Contribution to journalConference article

@article{e24393f4060843c3bf0314a31d93e84f,
title = "Generalization ability of universal learning network by using second order derivatives",
abstract = "In this paper, it is studied how the generalization ability of modeling of the dynamic systems can be improved by taking advantages of the second order derivatives of the criterion function with respect to the external inputs. The proposed method is based on the regularization theory proposed by Poggio, but a main distinctive point in this paper is that extension to dynamic systems from static systems has been taken into account and actual second order derivatives of the Universal Learning Network have been used to train the parameters of the networks. The second order derivatives term of the criterion function may minimize the deviation caused by the external input changes. Simulation results show that the method is useful for improving the generalization ability of identifying nonlinear dynamic systems using neural networks.",
author = "M. Han and K. Hirasawa and J. Hu and Junichi Murata",
year = "1998",
month = "12",
day = "1",
language = "English",
volume = "2",
pages = "1818--1823",
journal = "Proceedings of the IEEE International Conference on Systems, Man and Cybernetics",
issn = "0884-3627",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - JOUR

T1 - Generalization ability of universal learning network by using second order derivatives

AU - Han, M.

AU - Hirasawa, K.

AU - Hu, J.

AU - Murata, Junichi

PY - 1998/12/1

Y1 - 1998/12/1

N2 - In this paper, it is studied how the generalization ability of modeling of the dynamic systems can be improved by taking advantages of the second order derivatives of the criterion function with respect to the external inputs. The proposed method is based on the regularization theory proposed by Poggio, but a main distinctive point in this paper is that extension to dynamic systems from static systems has been taken into account and actual second order derivatives of the Universal Learning Network have been used to train the parameters of the networks. The second order derivatives term of the criterion function may minimize the deviation caused by the external input changes. Simulation results show that the method is useful for improving the generalization ability of identifying nonlinear dynamic systems using neural networks.

AB - In this paper, it is studied how the generalization ability of modeling of the dynamic systems can be improved by taking advantages of the second order derivatives of the criterion function with respect to the external inputs. The proposed method is based on the regularization theory proposed by Poggio, but a main distinctive point in this paper is that extension to dynamic systems from static systems has been taken into account and actual second order derivatives of the Universal Learning Network have been used to train the parameters of the networks. The second order derivatives term of the criterion function may minimize the deviation caused by the external input changes. Simulation results show that the method is useful for improving the generalization ability of identifying nonlinear dynamic systems using neural networks.

UR - http://www.scopus.com/inward/record.url?scp=0032311985&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0032311985&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:0032311985

VL - 2

SP - 1818

EP - 1823

JO - Proceedings of the IEEE International Conference on Systems, Man and Cybernetics

JF - Proceedings of the IEEE International Conference on Systems, Man and Cybernetics

SN - 0884-3627

ER -