Lyapunov design of a simple step-size adaptation strategy based on success

Claudia R. Correa, Elizabeth F. Wanner*, Carlos M. Fonseca

*Corresponding author for this work

Research output: Chapter in Book/Published conference outputConference publication

Abstract

A simple success-based step-size adaptation rule for singleparent Evolution Strategies is formulated, and the setting of the corresponding parameters is considered. Theoretical convergence on the class of strictly unimodal functions of one variable that are symmetric around the optimum is investigated using a stochastic Lyapunov function method developed by Semenov and Terkel [5] in the context of martingale theory. General expressions for the conditional expectations of the next values of step size and distance to the optimum under (1 +, λ)-selection are analytically derived, and an appropriate Lyapunov function is constructed. Convergence rate upper bounds, as well as adaptation parameter values, are obtained through numerical optimization for increasing values of λ. By selecting the number of offspring that minimizes the bound on the convergence rate with respect to the number of function evaluations, all strategy parameter values result from the analysis.

Original languageEnglish
Title of host publicationParallel Problem Solving from Nature – PPSN XIV
Subtitle of host publication14th International Conference, Edinburgh, UK, September 17-21, 2016, Proceedings
EditorsJulia Handl, Emma Hart, Peter R. Lewis, et al
Place of PublicationCham (CH)
PublisherSpringer
Pages101-110
Number of pages10
ISBN (Electronic)978-3-319-45823-6
ISBN (Print)978-3-319-45822-9
DOIs
Publication statusE-pub ahead of print - 31 Aug 2016
Event14th International Conference on Parallel Problem Solving from Nature - Edinburgh, United Kingdom
Duration: 17 Sept 201621 Sept 2016

Publication series

NameLecture Notes in Computer Science
PublisherSpringer
Volume9921
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference14th International Conference on Parallel Problem Solving from Nature
Abbreviated titlePPSN 2016
Country/TerritoryUnited Kingdom
CityEdinburgh
Period17/09/1621/09/16

Keywords

  • convergence rate
  • evolution strategy
  • Lyapunov function theory
  • step-size adaptation

Fingerprint

Dive into the research topics of 'Lyapunov design of a simple step-size adaptation strategy based on success'. Together they form a unique fingerprint.
  • Parallel Problem Solving from Nature – PPSN XIV: 14th International Conference, Edinburgh, UK, September 17-21, 2016, Proceedings

    Handl, J. (Editor), Hart, E. (Editor), Lewis, P. R. (Editor), López-Ibáñez, M. (Editor), Ochoa, G. (Editor) & Paechter, B. (Editor), 31 Aug 2016, Cham (CH): Springer. 1026 p. (Theoretical Computer Science and General Issues; vol. 9921)

    Research output: Book/ReportScholarly edition

  • Preface

    Handl, J., Hart, E., Lewis, P. R., López-Ibáñez, M., Ochoa, G. & Paechter, B., 31 Aug 2016, Parallel Problem Solving from Nature – PPSN XIV: 14th International Conference, Edinburgh, UK, September 17-21, 2016, Proceedings. Handl, J., Hart, E., Lewis, P. R. & et al (eds.). Cham (CH): Springer, p. V-VI 2 p. (Theoretical Computer Science and General Issues; vol. 9921).

    Research output: Chapter in Book/Published conference outputForeword/postscript

Cite this