An hybrid training method for B-spline neural networks

Cristiano Cabrita, János Botzheim, Antonio E B Ruano, L. Kóczy

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

1 Citation (Scopus)

Abstract

Current and past research has brought up new views related to the optimization of neural networks. For a fixed structure, second order methods are seen as the most promising. From previous works we have shown how second order methods are of easy applicability to a neural network. Namely, we have proved how the Levenberg-Marquard possesses not only better convergence but how it can assure the convergence to a local minima. However, as any gradient-based method, the results obtained depend on the startup point. In this work, a reformulated Evolutionary algorithm - the Bacterial Programming for Levenberg-Marquardt is proposed, as an heuristic which can be used to determine the most suitable starting points, therefore achieving, in most cases, the global optimum.

Original languageEnglish
Title of host publication2005 IEEE International Workshop on Intelligent Signal Processing - Proceedings
Pages165-170
Number of pages6
Publication statusPublished - 2005
Event2005 IEEE International Workshop on Intelligent Signal Processing - Faro, Portugal
Duration: Sep 1 2005Sep 3 2005

Other

Other2005 IEEE International Workshop on Intelligent Signal Processing
CountryPortugal
CityFaro
Period9/1/059/3/05

Fingerprint

Splines
Neural networks
Evolutionary algorithms

Keywords

  • B-Splines
  • Bacterial Algorithm
  • Genetic Programming
  • Levenberg-Marquard algorithm
  • Local and global minima

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Cabrita, C., Botzheim, J., Ruano, A. E. B., & Kóczy, L. (2005). An hybrid training method for B-spline neural networks. In 2005 IEEE International Workshop on Intelligent Signal Processing - Proceedings (pp. 165-170). [1531652]

An hybrid training method for B-spline neural networks. / Cabrita, Cristiano; Botzheim, János; Ruano, Antonio E B; Kóczy, L.

2005 IEEE International Workshop on Intelligent Signal Processing - Proceedings. 2005. p. 165-170 1531652.

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

Cabrita, C, Botzheim, J, Ruano, AEB & Kóczy, L 2005, An hybrid training method for B-spline neural networks. in 2005 IEEE International Workshop on Intelligent Signal Processing - Proceedings., 1531652, pp. 165-170, 2005 IEEE International Workshop on Intelligent Signal Processing, Faro, Portugal, 9/1/05.
Cabrita C, Botzheim J, Ruano AEB, Kóczy L. An hybrid training method for B-spline neural networks. In 2005 IEEE International Workshop on Intelligent Signal Processing - Proceedings. 2005. p. 165-170. 1531652
Cabrita, Cristiano ; Botzheim, János ; Ruano, Antonio E B ; Kóczy, L. / An hybrid training method for B-spline neural networks. 2005 IEEE International Workshop on Intelligent Signal Processing - Proceedings. 2005. pp. 165-170
@inproceedings{f5fdb6112d614f8b904f8264cb20e8f0,
title = "An hybrid training method for B-spline neural networks",
abstract = "Current and past research has brought up new views related to the optimization of neural networks. For a fixed structure, second order methods are seen as the most promising. From previous works we have shown how second order methods are of easy applicability to a neural network. Namely, we have proved how the Levenberg-Marquard possesses not only better convergence but how it can assure the convergence to a local minima. However, as any gradient-based method, the results obtained depend on the startup point. In this work, a reformulated Evolutionary algorithm - the Bacterial Programming for Levenberg-Marquardt is proposed, as an heuristic which can be used to determine the most suitable starting points, therefore achieving, in most cases, the global optimum.",
keywords = "B-Splines, Bacterial Algorithm, Genetic Programming, Levenberg-Marquard algorithm, Local and global minima",
author = "Cristiano Cabrita and J{\'a}nos Botzheim and Ruano, {Antonio E B} and L. K{\'o}czy",
year = "2005",
language = "English",
isbn = "078039030X",
pages = "165--170",
booktitle = "2005 IEEE International Workshop on Intelligent Signal Processing - Proceedings",

}

TY - GEN

T1 - An hybrid training method for B-spline neural networks

AU - Cabrita, Cristiano

AU - Botzheim, János

AU - Ruano, Antonio E B

AU - Kóczy, L.

PY - 2005

Y1 - 2005

N2 - Current and past research has brought up new views related to the optimization of neural networks. For a fixed structure, second order methods are seen as the most promising. From previous works we have shown how second order methods are of easy applicability to a neural network. Namely, we have proved how the Levenberg-Marquard possesses not only better convergence but how it can assure the convergence to a local minima. However, as any gradient-based method, the results obtained depend on the startup point. In this work, a reformulated Evolutionary algorithm - the Bacterial Programming for Levenberg-Marquardt is proposed, as an heuristic which can be used to determine the most suitable starting points, therefore achieving, in most cases, the global optimum.

AB - Current and past research has brought up new views related to the optimization of neural networks. For a fixed structure, second order methods are seen as the most promising. From previous works we have shown how second order methods are of easy applicability to a neural network. Namely, we have proved how the Levenberg-Marquard possesses not only better convergence but how it can assure the convergence to a local minima. However, as any gradient-based method, the results obtained depend on the startup point. In this work, a reformulated Evolutionary algorithm - the Bacterial Programming for Levenberg-Marquardt is proposed, as an heuristic which can be used to determine the most suitable starting points, therefore achieving, in most cases, the global optimum.

KW - B-Splines

KW - Bacterial Algorithm

KW - Genetic Programming

KW - Levenberg-Marquard algorithm

KW - Local and global minima

UR - http://www.scopus.com/inward/record.url?scp=33749078606&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33749078606&partnerID=8YFLogxK

M3 - Conference contribution

SN - 078039030X

SN - 9780780390300

SP - 165

EP - 170

BT - 2005 IEEE International Workshop on Intelligent Signal Processing - Proceedings

ER -