Skip to main navigation Skip to search Skip to main content

On the implementation of a class of stochastic search algorithms

  • Georgia Institute of Technology

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

We propose a stochastic approximation approach for implementing a class of random search-based optimization algorithms called the model-based methods. The approach makes efficient use of the past sampling information as the search progresses and can significantly reduce the number of function evaluations needed to obtain high quality solutions. We illustrate our approach through a specific algorithm called Model-based Annealing Random Search with Stochastic Averaging (MARS-SA), which maintains the per-iteration sample size at a small constant value. We present the global convergence property of MARS-SA and report on numerical results.

Original languageEnglish
Title of host publicationAdvances in Global Optimization
EditorsWenxun Xing, David Gao, Ning Ruan
PublisherSpringer New York LLC
Pages427-435
Number of pages9
ISBN (Electronic)9783319083766
DOIs
StatePublished - 2015
Event3rd World Congress on Global Optimization in Engineering and Science, WCGO 2013 - Anhui, China
Duration: Jul 8 2013Jul 12 2013

Publication series

NameSpringer Proceedings in Mathematics and Statistics
Volume95
ISSN (Print)2194-1009
ISSN (Electronic)2194-1017

Conference

Conference3rd World Congress on Global Optimization in Engineering and Science, WCGO 2013
Country/TerritoryChina
CityAnhui
Period07/8/1307/12/13

Keywords

  • Global optimization
  • Model-based annealing random search
  • Stochastic approximation

Fingerprint

Dive into the research topics of 'On the implementation of a class of stochastic search algorithms'. Together they form a unique fingerprint.

Cite this