Improving generalization ability of universal learning networks with superfluous parameters

Min Han, Kotaro Hirasawa, Takayuki Furuzuki, Junichi Murata, Chun zhi Jin

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

The parameters in large scale neural networks can be divided into two classes. One class is necessary for a certain purpose while another class is not directly needed. The parameters in the latter are defined as superfluous parameters. How to use these superfluous parameters effectively is an interesting subject. In this paper, it is studied how the generalization ability of dynamic systems can be improved by use of network's superfluous parameters. And a calculation technique is proposed which use second order derivatives of the criterion function with respect to superfluous parameters. So as to investigate the effectiveness of the proposed method, simulations of modeling a nonlinear robot dynamics system is studied. Simulation results show that the proposed method is useful for improving the generalization ability of neural networks, which may model nonlinear dynamic systems.

Original languageEnglish
Title of host publicationProceedings of the IEEE International Conference on Systems, Man and Cybernetics
PublisherIEEE
Volume5
Publication statusPublished - 1999
Externally publishedYes
Event1999 IEEE International Conference on Systems, Man, and Cybernetics 'Human Communication and Cybernetics' - Tokyo, Jpn
Duration: 1999 Oct 121999 Oct 15

Other

Other1999 IEEE International Conference on Systems, Man, and Cybernetics 'Human Communication and Cybernetics'
CityTokyo, Jpn
Period99/10/1299/10/15

Fingerprint

Dynamical systems
Neural networks
Robots
Derivatives

ASJC Scopus subject areas

  • Hardware and Architecture
  • Control and Systems Engineering

Cite this

Han, M., Hirasawa, K., Furuzuki, T., Murata, J., & Jin, C. Z. (1999). Improving generalization ability of universal learning networks with superfluous parameters. In Proceedings of the IEEE International Conference on Systems, Man and Cybernetics (Vol. 5). IEEE.

Improving generalization ability of universal learning networks with superfluous parameters. / Han, Min; Hirasawa, Kotaro; Furuzuki, Takayuki; Murata, Junichi; Jin, Chun zhi.

Proceedings of the IEEE International Conference on Systems, Man and Cybernetics. Vol. 5 IEEE, 1999.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Han, M, Hirasawa, K, Furuzuki, T, Murata, J & Jin, CZ 1999, Improving generalization ability of universal learning networks with superfluous parameters. in Proceedings of the IEEE International Conference on Systems, Man and Cybernetics. vol. 5, IEEE, 1999 IEEE International Conference on Systems, Man, and Cybernetics 'Human Communication and Cybernetics', Tokyo, Jpn, 99/10/12.
Han M, Hirasawa K, Furuzuki T, Murata J, Jin CZ. Improving generalization ability of universal learning networks with superfluous parameters. In Proceedings of the IEEE International Conference on Systems, Man and Cybernetics. Vol. 5. IEEE. 1999
Han, Min ; Hirasawa, Kotaro ; Furuzuki, Takayuki ; Murata, Junichi ; Jin, Chun zhi. / Improving generalization ability of universal learning networks with superfluous parameters. Proceedings of the IEEE International Conference on Systems, Man and Cybernetics. Vol. 5 IEEE, 1999.
@inproceedings{67597e0c182c44cca0b875d60abeb695,
title = "Improving generalization ability of universal learning networks with superfluous parameters",
abstract = "The parameters in large scale neural networks can be divided into two classes. One class is necessary for a certain purpose while another class is not directly needed. The parameters in the latter are defined as superfluous parameters. How to use these superfluous parameters effectively is an interesting subject. In this paper, it is studied how the generalization ability of dynamic systems can be improved by use of network's superfluous parameters. And a calculation technique is proposed which use second order derivatives of the criterion function with respect to superfluous parameters. So as to investigate the effectiveness of the proposed method, simulations of modeling a nonlinear robot dynamics system is studied. Simulation results show that the proposed method is useful for improving the generalization ability of neural networks, which may model nonlinear dynamic systems.",
author = "Min Han and Kotaro Hirasawa and Takayuki Furuzuki and Junichi Murata and Jin, {Chun zhi}",
year = "1999",
language = "English",
volume = "5",
booktitle = "Proceedings of the IEEE International Conference on Systems, Man and Cybernetics",
publisher = "IEEE",

}

TY - GEN

T1 - Improving generalization ability of universal learning networks with superfluous parameters

AU - Han, Min

AU - Hirasawa, Kotaro

AU - Furuzuki, Takayuki

AU - Murata, Junichi

AU - Jin, Chun zhi

PY - 1999

Y1 - 1999

N2 - The parameters in large scale neural networks can be divided into two classes. One class is necessary for a certain purpose while another class is not directly needed. The parameters in the latter are defined as superfluous parameters. How to use these superfluous parameters effectively is an interesting subject. In this paper, it is studied how the generalization ability of dynamic systems can be improved by use of network's superfluous parameters. And a calculation technique is proposed which use second order derivatives of the criterion function with respect to superfluous parameters. So as to investigate the effectiveness of the proposed method, simulations of modeling a nonlinear robot dynamics system is studied. Simulation results show that the proposed method is useful for improving the generalization ability of neural networks, which may model nonlinear dynamic systems.

AB - The parameters in large scale neural networks can be divided into two classes. One class is necessary for a certain purpose while another class is not directly needed. The parameters in the latter are defined as superfluous parameters. How to use these superfluous parameters effectively is an interesting subject. In this paper, it is studied how the generalization ability of dynamic systems can be improved by use of network's superfluous parameters. And a calculation technique is proposed which use second order derivatives of the criterion function with respect to superfluous parameters. So as to investigate the effectiveness of the proposed method, simulations of modeling a nonlinear robot dynamics system is studied. Simulation results show that the proposed method is useful for improving the generalization ability of neural networks, which may model nonlinear dynamic systems.

UR - http://www.scopus.com/inward/record.url?scp=0033329404&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0033329404&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:0033329404

VL - 5

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

PB - IEEE

ER -