Modelling parameter configuration spaces with local optima networks

German Treimun-Costa, Elizabeth Montero, Gabriela Ochoa, Nicolás Rojas-Morales

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

Abstract

Most algorithms proposed for solving complex problems require the definition of some parameter values. The process of finding suitable parameter values is an optimization problem by itself. Understanding the global structure of search spaces of complex optimization problems remains a challenge. Moreover, understanding the relationship between parameter values and the performance of metaheuristics is a key issue on their development. Local optima networks propose a scheme to model search spaces as networks whose nodes represent local optima and edges represent transitions between them. In this work, we adapt the local optima network model to analyze and visualize the global structure of parameter configuration spaces. Our main objectives are to understand the structure of these networks and explore the difficulty of different tuning scenarios using common indicators previously proposed in local optima networks studies (e.g. number of local optima, number of global optima and presence of local and global funnels). For this, we use the well-known tuning method ParamILS to analyze configuration search spaces of a standard genetic algorithm that solves continuous optimization problems.

Original languageEnglish
Title of host publicationGECCO 2020 - Proceedings of the 2020 Genetic and Evolutionary Computation Conference
PublisherAssociation for Computing Machinery
Pages751-759
Number of pages9
ISBN (Electronic)9781450371285
DOIs
Publication statusPublished - 25 Jun 2020
Event2020 Genetic and Evolutionary Computation Conference, GECCO 2020 - Cancun, Mexico
Duration: 8 Jul 202012 Jul 2020

Publication series

NameGECCO 2020 - Proceedings of the 2020 Genetic and Evolutionary Computation Conference

Conference

Conference2020 Genetic and Evolutionary Computation Conference, GECCO 2020
Country/TerritoryMexico
CityCancun
Period8/07/2012/07/20

Keywords

  • Local optima networks
  • Parameter configuration space
  • Parameter setting problem
  • Tuning methods

ASJC Scopus subject areas

  • Artificial Intelligence
  • Software
  • Theoretical Computer Science

Fingerprint

Dive into the research topics of 'Modelling parameter configuration spaces with local optima networks'. Together they form a unique fingerprint.

Cite this