A new gradient-based search method: Grey-gradient search method

Chin-Ming Hong, Chih Ming Chen, Heng Kang Fan

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

1 Citation (Scopus)

Abstract

Optimization theory and methods play very important role for engineering design and applications. In many domains of engineering applications, it is usually the most important process to find near optimal solution. The gradient-descent method is widely used to solve many engineering optimization problems. But the gradient-descent method has some disadvantages for searching optimal solution. Firstly, its convergent speed is very slowly and is easy to trap into local minimum in the applications of many actual problems. Secondly, the learning rate of gradient-descent method must been determined adequately for different engineering problem. If the learning rate set very small, the convergent speed will be very slowly. If the learning rate is set very large, the searching of solution is very easy to generate trashing or divergence. The main goal of this research is to propose a new method that is based on grey prediction theory to improve the gradient-descent method. We use the idea of grey prediction to speed up effectively the searching speed of gradient-descent method, and improve the drawback that gradient-descent method is very easy trap into local minimum. From the experimental results, we can show the workings of the proposed method that can speed up effectively the searching speed of gradient-descent method, and improve the drawback that gradientdescent method is easy to trapped into local minimum.

Original languageEnglish
Title of host publicationMultiple Approaches to Intelligent Systems - 12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems IEA/AIE 1999, Proceedings
EditorsAyman El-Dessouki, Ibrahim Imam, Yves Kodratoff, Moonis Ali
PublisherSpringer Verlag
Pages185-194
Number of pages10
ISBN (Print)3540660763, 9783540660767
Publication statusPublished - 1999 Jan 1
Event12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, IEA/AIE 1999 - Cairo, Egypt
Duration: 1999 May 311999 Jun 3

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume1611
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, IEA/AIE 1999
CountryEgypt
CityCairo
Period99/5/3199/6/3

Fingerprint

Gradient Descent Method
Gradient methods
Gradient Method
Search Methods
Gradient
Learning Rate
Local Minima
Engineering Application
Trap
Signal filtering and prediction
Speedup
Optimal Solution
Grey Theory
Prediction Theory
Engineering
Optimization Theory
Engineering Design
Optimization Methods
Divergence
Optimization Problem

Keywords

  • Gradient-descent method
  • Grey prediction
  • Grey-gradient method
  • Optimization method

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Hong, C-M., Chen, C. M., & Fan, H. K. (1999). A new gradient-based search method: Grey-gradient search method. In A. El-Dessouki, I. Imam, Y. Kodratoff, & M. Ali (Eds.), Multiple Approaches to Intelligent Systems - 12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems IEA/AIE 1999, Proceedings (pp. 185-194). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1611). Springer Verlag.

A new gradient-based search method : Grey-gradient search method. / Hong, Chin-Ming; Chen, Chih Ming; Fan, Heng Kang.

Multiple Approaches to Intelligent Systems - 12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems IEA/AIE 1999, Proceedings. ed. / Ayman El-Dessouki; Ibrahim Imam; Yves Kodratoff; Moonis Ali. Springer Verlag, 1999. p. 185-194 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1611).

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

Hong, C-M, Chen, CM & Fan, HK 1999, A new gradient-based search method: Grey-gradient search method. in A El-Dessouki, I Imam, Y Kodratoff & M Ali (eds), Multiple Approaches to Intelligent Systems - 12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems IEA/AIE 1999, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 1611, Springer Verlag, pp. 185-194, 12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, IEA/AIE 1999, Cairo, Egypt, 99/5/31.
Hong C-M, Chen CM, Fan HK. A new gradient-based search method: Grey-gradient search method. In El-Dessouki A, Imam I, Kodratoff Y, Ali M, editors, Multiple Approaches to Intelligent Systems - 12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems IEA/AIE 1999, Proceedings. Springer Verlag. 1999. p. 185-194. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
Hong, Chin-Ming ; Chen, Chih Ming ; Fan, Heng Kang. / A new gradient-based search method : Grey-gradient search method. Multiple Approaches to Intelligent Systems - 12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems IEA/AIE 1999, Proceedings. editor / Ayman El-Dessouki ; Ibrahim Imam ; Yves Kodratoff ; Moonis Ali. Springer Verlag, 1999. pp. 185-194 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{c54bd42aeb60457aa3b24f13f3e9947c,
title = "A new gradient-based search method: Grey-gradient search method",
abstract = "Optimization theory and methods play very important role for engineering design and applications. In many domains of engineering applications, it is usually the most important process to find near optimal solution. The gradient-descent method is widely used to solve many engineering optimization problems. But the gradient-descent method has some disadvantages for searching optimal solution. Firstly, its convergent speed is very slowly and is easy to trap into local minimum in the applications of many actual problems. Secondly, the learning rate of gradient-descent method must been determined adequately for different engineering problem. If the learning rate set very small, the convergent speed will be very slowly. If the learning rate is set very large, the searching of solution is very easy to generate trashing or divergence. The main goal of this research is to propose a new method that is based on grey prediction theory to improve the gradient-descent method. We use the idea of grey prediction to speed up effectively the searching speed of gradient-descent method, and improve the drawback that gradient-descent method is very easy trap into local minimum. From the experimental results, we can show the workings of the proposed method that can speed up effectively the searching speed of gradient-descent method, and improve the drawback that gradientdescent method is easy to trapped into local minimum.",
keywords = "Gradient-descent method, Grey prediction, Grey-gradient method, Optimization method",
author = "Chin-Ming Hong and Chen, {Chih Ming} and Fan, {Heng Kang}",
year = "1999",
month = "1",
day = "1",
language = "English",
isbn = "3540660763",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "185--194",
editor = "Ayman El-Dessouki and Ibrahim Imam and Yves Kodratoff and Moonis Ali",
booktitle = "Multiple Approaches to Intelligent Systems - 12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems IEA/AIE 1999, Proceedings",

}

TY - GEN

T1 - A new gradient-based search method

T2 - Grey-gradient search method

AU - Hong, Chin-Ming

AU - Chen, Chih Ming

AU - Fan, Heng Kang

PY - 1999/1/1

Y1 - 1999/1/1

N2 - Optimization theory and methods play very important role for engineering design and applications. In many domains of engineering applications, it is usually the most important process to find near optimal solution. The gradient-descent method is widely used to solve many engineering optimization problems. But the gradient-descent method has some disadvantages for searching optimal solution. Firstly, its convergent speed is very slowly and is easy to trap into local minimum in the applications of many actual problems. Secondly, the learning rate of gradient-descent method must been determined adequately for different engineering problem. If the learning rate set very small, the convergent speed will be very slowly. If the learning rate is set very large, the searching of solution is very easy to generate trashing or divergence. The main goal of this research is to propose a new method that is based on grey prediction theory to improve the gradient-descent method. We use the idea of grey prediction to speed up effectively the searching speed of gradient-descent method, and improve the drawback that gradient-descent method is very easy trap into local minimum. From the experimental results, we can show the workings of the proposed method that can speed up effectively the searching speed of gradient-descent method, and improve the drawback that gradientdescent method is easy to trapped into local minimum.

AB - Optimization theory and methods play very important role for engineering design and applications. In many domains of engineering applications, it is usually the most important process to find near optimal solution. The gradient-descent method is widely used to solve many engineering optimization problems. But the gradient-descent method has some disadvantages for searching optimal solution. Firstly, its convergent speed is very slowly and is easy to trap into local minimum in the applications of many actual problems. Secondly, the learning rate of gradient-descent method must been determined adequately for different engineering problem. If the learning rate set very small, the convergent speed will be very slowly. If the learning rate is set very large, the searching of solution is very easy to generate trashing or divergence. The main goal of this research is to propose a new method that is based on grey prediction theory to improve the gradient-descent method. We use the idea of grey prediction to speed up effectively the searching speed of gradient-descent method, and improve the drawback that gradient-descent method is very easy trap into local minimum. From the experimental results, we can show the workings of the proposed method that can speed up effectively the searching speed of gradient-descent method, and improve the drawback that gradientdescent method is easy to trapped into local minimum.

KW - Gradient-descent method

KW - Grey prediction

KW - Grey-gradient method

KW - Optimization method

UR - http://www.scopus.com/inward/record.url?scp=84947777721&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84947777721&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84947777721

SN - 3540660763

SN - 9783540660767

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

SP - 185

EP - 194

BT - Multiple Approaches to Intelligent Systems - 12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems IEA/AIE 1999, Proceedings

A2 - El-Dessouki, Ayman

A2 - Imam, Ibrahim

A2 - Kodratoff, Yves

A2 - Ali, Moonis

PB - Springer Verlag

ER -