Surrogate-Assisted NSGA-II Algorithm for Expensive Multiobjective Optimization - IRT SystemX Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Surrogate-Assisted NSGA-II Algorithm for Expensive Multiobjective Optimization

Mouadh Yagoubi
  • Fonction : Auteur
  • PersonId : 1054942
Hiba Bederina

Résumé

Most real world problems are multiobjective by nature and expensive to evaluate. We propose in this work a surrogate assisted approach for multiobjective evolutionary algorithms by building a surrogate model on each objective. However, integrating a surrogate model within an optimization process generates complexity with additional hyper-parameters to tune. Empirical validation on standards MOO benchmark problems of a use case based on NSGA-II and surrogates using SVM regression shows a significant improvement of the optimization cost in terms of true objectives evaluations, especially for low budget. We also discuss the behavior of the proposed algorithm using different values of the parameter calibrating the use of the surrogate model.
Fichier principal
Vignette du fichier
SurrogateMultiObjective-6.pdf (1.93 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04096875 , version 1 (13-05-2023)

Identifiants

Citer

Mouadh Yagoubi, Hiba Bederina. Surrogate-Assisted NSGA-II Algorithm for Expensive Multiobjective Optimization. Genetic and Evolutionary Computation Conference (GECCO 2023), Jul 2023, Lisbon, Portugal. ⟨10.1145/3583133.3590746⟩. ⟨hal-04096875⟩

Collections

IRT-SYSTEMX
69 Consultations
90 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More