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

Chin Ming Hong, Chih Ming Chen, Heng Kang Fan

研究成果: 書貢獻/報告類型會議論文篇章

3 引文 斯高帕斯(Scopus)

摘要

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.

原文英語
主出版物標題Multiple Approaches to Intelligent Systems - 12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems IEA/AIE 1999, Proceedings
編輯Ibrahim Imam, Yves Kodratoff, Ayman El-Dessouki, Moonis Ali
發行者Springer Verlag
頁面185-194
頁數10
ISBN(列印)3540660763, 9783540660767
DOIs
出版狀態已發佈 - 1999
事件12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, IEA/AIE 1999 - Cairo, 埃及
持續時間: 1999 5月 311999 6月 3

出版系列

名字Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
1611
ISSN(列印)0302-9743
ISSN(電子)1611-3349

會議

會議12th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, IEA/AIE 1999
國家/地區埃及
城市Cairo
期間1999/05/311999/06/03

ASJC Scopus subject areas

  • 理論電腦科學
  • 電腦科學(全部)

指紋

深入研究「A new gradient-based search method: Grey-gradient search method」主題。共同形成了獨特的指紋。

引用此