Evolutionary algorithm with ensemble strategies based on maximum a posteriori for continuous optimization

A. Ghoumari A. Nakib 1 P. Siarry 1
1 SIMO
LISSI - Laboratoire Images, Signaux et Systèmes Intelligents
Abstract : To be efficient and reach the best solution, evolutionary algorithms (EAs) are composed of three evolutionary operators: selection, crossover and mutation. Nevertheless, there is no established rule to choose them from a large panel of operators available in the literature. In this paper, a Maximum A Posteriori (MAP) principle based rule is proposed to palliate the operators’ selection problem for solving different kinds of continuous optimization problems. This learning based approach allows switching between different strategies during the optimization. This algorithm is called Maximum a posteriori based Evolutionary Algorithm (MEA). Experimental analysis was performed and proved its efficiency and robustness on a large set of 34 problems.
Type de document :
Article dans une revue
Information Sciences, Elsevier, 2018, 〈10.1016/j.ins.2018.05.041〉
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01823466
Contributeur : Lab Lissi <>
Soumis le : mardi 26 juin 2018 - 09:40:29
Dernière modification le : jeudi 14 mars 2019 - 17:23:06

Identifiants

Collections

Citation

A. Ghoumari, A. Nakib, P. Siarry. Evolutionary algorithm with ensemble strategies based on maximum a posteriori for continuous optimization. Information Sciences, Elsevier, 2018, 〈10.1016/j.ins.2018.05.041〉. 〈hal-01823466〉

Partager

Métriques

Consultations de la notice

46