Closed state model for understanding the dynamics of MOEAs

Abstract : This work proposes the use of simple closed state models to capture, analyze and compare the dynamics of multi- and many-objective evolutionary algorithms. Two- and three-state models representing the composition of the instantaneous population are described and learned for representatives of the major approaches to multi-objective optimization, i.e. dominance, extensions of dominance, decomposition, and indicator algorithms. The model parameters are trained from data obtained running the algorithms with various population sizes on enumerable MNK-landscapes with 3, 4, 5 and 6 objectives. We show ways to interpret and use the model parameter values in order to analyze the population dynamics according to selected features. For example, we are interested in knowing how parameter values change for a given population size with the increase of the number of objectives. We also show a graphical representation capturing in one graph how the parameters magnitude and sign relate to the connections between states.
Type de document :
Communication dans un congrès
Genetic and Evolutionary Computation Conference (GECCO 2017), Jul 2017, Berlin, Germany. ACM, 2017, Genetic and Evolutionary Computation Conference (GECCO 2017). 〈http://gecco-2017.sigevo.org〉
Liste complète des métadonnées

Littérature citée [11 références]  Voir  Masquer  Télécharger

https://hal.archives-ouvertes.fr/hal-01496329
Contributeur : Sébastien Verel <>
Soumis le : mercredi 10 mai 2017 - 02:45:57
Dernière modification le : vendredi 22 mars 2019 - 01:35:46
Document(s) archivé(s) le : vendredi 11 août 2017 - 12:17:48

Fichier

monzon.gecco2017.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01496329, version 1

Collections

Citation

Hugo Monzón, Hernan Aguirre, Sébastien Verel, Arnaud Liefooghe, Bilel Derbel, et al.. Closed state model for understanding the dynamics of MOEAs. Genetic and Evolutionary Computation Conference (GECCO 2017), Jul 2017, Berlin, Germany. ACM, 2017, Genetic and Evolutionary Computation Conference (GECCO 2017). 〈http://gecco-2017.sigevo.org〉. 〈hal-01496329〉

Partager

Métriques

Consultations de la notice

442

Téléchargements de fichiers

123