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

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.