Diversity Management in Evolutionary Dynamic Optimization

Bernhard Werth*, Johannes Karder, Stefan Wagner, Michael Affenzeller

*Corresponding author for this work

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

Abstract

The retention of diversity of genetic information is an important aspect of many population-based evolutionary optimizers. With the increasing relevance of dynamic optimization, where live data is streamed directly into a running optimization system, this algorithmic facet gains new importance. This study compares five different strategies for handling diversity in genetic algorithms in a dynamic open-ended optimization scenario. Using the traveling salesman problem as a benchmark, the algorithmic variations are compared and analyzed with respect to their performance and retained diversity. Results indicate that convergence patterns behave differently from static optimization and several algorithm features that are well understood for static optimization may have unintended consequences in dynamic scenarios.

Original languageEnglish
Title of host publicationComputer Aided Systems Theory - EUROCAST 2024 - 19th International Conference, 2024, Revised Selected Papers
EditorsAlexis Quesada-Arencibia, Michael Affenzeller, Roberto Moreno-Díaz
PublisherSpringer
Pages140-147
Number of pages8
ISBN (Print)9783031829512
DOIs
Publication statusPublished - 2025
Event19th International Conference on Computer Aided Systems Theory, EUROCAST 2024 - Las Palmas de Canaria, Spain
Duration: 25 Feb 20241 Mar 2024

Publication series

NameLecture Notes in Computer Science
Volume15172 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference19th International Conference on Computer Aided Systems Theory, EUROCAST 2024
Country/TerritorySpain
CityLas Palmas de Canaria
Period25.02.202401.03.2024

Keywords

  • dynamic optimization
  • evolutionary algorithms
  • genetic algorithms

Fingerprint

Dive into the research topics of 'Diversity Management in Evolutionary Dynamic Optimization'. Together they form a unique fingerprint.

Cite this