A hierarchical method for training embedded sigmoidal neural networks

Takayuki Furuzuki, Kotaro Hirasawa

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

3 Citations (Scopus)

Abstract

This paper discusses the problem of applying sigmoidal neural networks to identification of nonlinear dynamical systems. When using sigmoidal neural networks directly as nonlinear models, one often meets problems such as model parameters lack of physical meaning, sensitivity to noise in model training. In this paper, we introduce an embedded sigmoidal neural network model, in which the neural network is not used directly as a model, but is embedded in a shield such that part of the model parameters become meaningful. Corresponding to the meaningful part and the meaningless part of model parameters, a hierarchical training algorithm consisting of two learning loops is then introduced to train the model. Simulation results show that such a dual loop learning algorithm can solve the noise sensitivity and local minimum problems to some extent.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages937-942
Number of pages6
Volume2130
ISBN (Print)3540424865, 9783540446682
DOIs
Publication statusPublished - 2001
Externally publishedYes
EventInternational Conference on Artificial Neural Networks, ICANN 2001 - Vienna, Austria
Duration: 2001 Aug 212001 Aug 25

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume2130
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

OtherInternational Conference on Artificial Neural Networks, ICANN 2001
CountryAustria
CityVienna
Period01/8/2101/8/25

Fingerprint

Neural Networks
Neural networks
Model
Training Algorithm
Nonlinear Dynamical Systems
Local Minima
Neural Network Model
Nonlinear dynamical systems
Nonlinear Model
Training
Learning Algorithm
Learning algorithms
Simulation

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Furuzuki, T., & Hirasawa, K. (2001). A hierarchical method for training embedded sigmoidal neural networks. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 2130, pp. 937-942). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 2130). Springer Verlag. https://doi.org/10.1007/3-540-44668-0_129

A hierarchical method for training embedded sigmoidal neural networks. / Furuzuki, Takayuki; Hirasawa, Kotaro.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 2130 Springer Verlag, 2001. p. 937-942 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 2130).

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

Furuzuki, T & Hirasawa, K 2001, A hierarchical method for training embedded sigmoidal neural networks. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 2130, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 2130, Springer Verlag, pp. 937-942, International Conference on Artificial Neural Networks, ICANN 2001, Vienna, Austria, 01/8/21. https://doi.org/10.1007/3-540-44668-0_129
Furuzuki T, Hirasawa K. A hierarchical method for training embedded sigmoidal neural networks. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 2130. Springer Verlag. 2001. p. 937-942. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/3-540-44668-0_129
Furuzuki, Takayuki ; Hirasawa, Kotaro. / A hierarchical method for training embedded sigmoidal neural networks. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 2130 Springer Verlag, 2001. pp. 937-942 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{03a5d5394f0c436985e3ace54a03e6f4,
title = "A hierarchical method for training embedded sigmoidal neural networks",
abstract = "This paper discusses the problem of applying sigmoidal neural networks to identification of nonlinear dynamical systems. When using sigmoidal neural networks directly as nonlinear models, one often meets problems such as model parameters lack of physical meaning, sensitivity to noise in model training. In this paper, we introduce an embedded sigmoidal neural network model, in which the neural network is not used directly as a model, but is embedded in a shield such that part of the model parameters become meaningful. Corresponding to the meaningful part and the meaningless part of model parameters, a hierarchical training algorithm consisting of two learning loops is then introduced to train the model. Simulation results show that such a dual loop learning algorithm can solve the noise sensitivity and local minimum problems to some extent.",
author = "Takayuki Furuzuki and Kotaro Hirasawa",
year = "2001",
doi = "10.1007/3-540-44668-0_129",
language = "English",
isbn = "3540424865",
volume = "2130",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "937--942",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - A hierarchical method for training embedded sigmoidal neural networks

AU - Furuzuki, Takayuki

AU - Hirasawa, Kotaro

PY - 2001

Y1 - 2001

N2 - This paper discusses the problem of applying sigmoidal neural networks to identification of nonlinear dynamical systems. When using sigmoidal neural networks directly as nonlinear models, one often meets problems such as model parameters lack of physical meaning, sensitivity to noise in model training. In this paper, we introduce an embedded sigmoidal neural network model, in which the neural network is not used directly as a model, but is embedded in a shield such that part of the model parameters become meaningful. Corresponding to the meaningful part and the meaningless part of model parameters, a hierarchical training algorithm consisting of two learning loops is then introduced to train the model. Simulation results show that such a dual loop learning algorithm can solve the noise sensitivity and local minimum problems to some extent.

AB - This paper discusses the problem of applying sigmoidal neural networks to identification of nonlinear dynamical systems. When using sigmoidal neural networks directly as nonlinear models, one often meets problems such as model parameters lack of physical meaning, sensitivity to noise in model training. In this paper, we introduce an embedded sigmoidal neural network model, in which the neural network is not used directly as a model, but is embedded in a shield such that part of the model parameters become meaningful. Corresponding to the meaningful part and the meaningless part of model parameters, a hierarchical training algorithm consisting of two learning loops is then introduced to train the model. Simulation results show that such a dual loop learning algorithm can solve the noise sensitivity and local minimum problems to some extent.

UR - http://www.scopus.com/inward/record.url?scp=33645276097&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33645276097&partnerID=8YFLogxK

U2 - 10.1007/3-540-44668-0_129

DO - 10.1007/3-540-44668-0_129

M3 - Conference contribution

SN - 3540424865

SN - 9783540446682

VL - 2130

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 937

EP - 942

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

PB - Springer Verlag

ER -