Dual evolutionary optimization

Abstract : The most general strategy for handling constraints in evolutionary optimization is through penalty functions. The choice of the penalty function is critical to both success and efficiency of the optimization. Many strategies have been proposed for formulating penalty functions, most of which rely on arbitrary tuning of parameters. An new insight on function penalization is proposed in this paper that relies on the dual optimization problem. An evolutionary algorithm for approximately solving dual optimization problems is first presented. Next, an efficient and exact penalty function without penalization parameter to be tuned is proposed. Numerical tests are provided for continuous variables and inequality constraints.
Type de document :
Chapitre d'ouvrage
Collet, Pierre : Fonlupt, Cyril : Hao, Jin-Kao : Lutton, Evelyne : Schoenauer, Marc. Artificial Evolution, Springer Berlin / Heidelberg, p 139 - 148, 2002, Lecture Notes in Computer Science, 978-3-540-43544-0. 〈10.1007/3-540-46033-0_23〉
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-00298026
Contributeur : Sms Stagiaire <>
Soumis le : mercredi 16 juillet 2008 - 13:39:05
Dernière modification le : mardi 23 octobre 2018 - 14:36:09

Lien texte intégral

Identifiants

Citation

Rodolphe Le Riche, F. Guyon. Dual evolutionary optimization. Collet, Pierre : Fonlupt, Cyril : Hao, Jin-Kao : Lutton, Evelyne : Schoenauer, Marc. Artificial Evolution, Springer Berlin / Heidelberg, p 139 - 148, 2002, Lecture Notes in Computer Science, 978-3-540-43544-0. 〈10.1007/3-540-46033-0_23〉. 〈hal-00298026〉

Partager

Métriques

Consultations de la notice

77