A hybrid fast descent method for globally optimizing high dimensional functions

Yong Jun Wang; Jiang She Zhang

A hybrid fast descent method for globally optimizing high dimensional functions

Yong Jun Wang
, Jiang She Zhang

School of Mathematics and Statistics

Research output: Contribution to journal › Article › peer-review

1 Scopus citations

Abstract

A new algorithm built on a hybrid method (GRSA) for large scale global optimization problems is proposed. Unlike the previous proposed method that the original objective functions keep unchanged during the whole course of optimizing, a convexized auxiliary function on the obtained local minimizer so far is employed to improve the SA search ability. The experiments conducted show that the new method provides excellent results especially for large scale problems, compared to other state-of-the-art algorithm.

Original language	English
Pages (from-to)	87-98
Number of pages	12
Journal	Dynamics of Continuous, Discrete and Impulsive Systems Series B: Applications and Algorithms
Volume	15
Issue number	1
State	Published - Feb 2008

Keywords

Auxiliary function
Global optimization
Gradient algorithm
Simulated annealing method (SA)

Cite this

@article{80cd841edb204d5fab5916735172d175,

title = "A hybrid fast descent method for globally optimizing high dimensional functions",

abstract = "A new algorithm built on a hybrid method (GRSA) for large scale global optimization problems is proposed. Unlike the previous proposed method that the original objective functions keep unchanged during the whole course of optimizing, a convexized auxiliary function on the obtained local minimizer so far is employed to improve the SA search ability. The experiments conducted show that the new method provides excellent results especially for large scale problems, compared to other state-of-the-art algorithm.",

keywords = "Auxiliary function, Global optimization, Gradient algorithm, Simulated annealing method (SA)",

author = "Wang, \{Yong Jun\} and Zhang, \{Jiang She\}",

year = "2008",

month = feb,

language = "英语",

volume = "15",

pages = "87--98",

journal = "Dynamics of Continuous, Discrete and Impulsive Systems Series B: Applications and Algorithms",

issn = "1492-8760",

publisher = "Watam Press",

number = "1",

}

TY - JOUR

T1 - A hybrid fast descent method for globally optimizing high dimensional functions

AU - Wang, Yong Jun

AU - Zhang, Jiang She

PY - 2008/2

Y1 - 2008/2

N2 - A new algorithm built on a hybrid method (GRSA) for large scale global optimization problems is proposed. Unlike the previous proposed method that the original objective functions keep unchanged during the whole course of optimizing, a convexized auxiliary function on the obtained local minimizer so far is employed to improve the SA search ability. The experiments conducted show that the new method provides excellent results especially for large scale problems, compared to other state-of-the-art algorithm.

AB - A new algorithm built on a hybrid method (GRSA) for large scale global optimization problems is proposed. Unlike the previous proposed method that the original objective functions keep unchanged during the whole course of optimizing, a convexized auxiliary function on the obtained local minimizer so far is employed to improve the SA search ability. The experiments conducted show that the new method provides excellent results especially for large scale problems, compared to other state-of-the-art algorithm.

KW - Auxiliary function

KW - Global optimization

KW - Gradient algorithm

KW - Simulated annealing method (SA)

UR - https://www.scopus.com/pages/publications/39449116993

M3 - 文章

AN - SCOPUS:39449116993

SN - 1492-8760

VL - 15

SP - 87

EP - 98

JO - Dynamics of Continuous, Discrete and Impulsive Systems Series B: Applications and Algorithms

JF - Dynamics of Continuous, Discrete and Impulsive Systems Series B: Applications and Algorithms

IS - 1

ER -

A hybrid fast descent method for globally optimizing high dimensional functions

Abstract

Keywords

Other files and links

Fingerprint

Cite this