A Cooperative Tree-based Hybrid GA-B&B Approach for Solving Challenging Permutation-based Problems

Abstract : The issue addressed in this paper is how to build low-level hybrid cooperative optimization methods that combine a Genetic Algorithm (GA) with a Branch-and-Bound algorithm (B&B). The key challenge is to provide a common solution and search space coding and associated transformation operators enabling an efficient cooperation between the two algorithms. The tree-based coding is traditionally used in exact optimization methods such as B&B. In this paper, we explore the idea of using such coding in Genetic Algorithms. Following this idea, we propose a pioneering approach hybridizing a GA with a B&B algorithm. The information (solutions and search subspaces) exchange between the two algorithms is performed at low-level and during the exploration process. From the implementation point of view, the common coding has facilitated the low-level coupling of two software frameworks: ParadisEO and BOB++ used to implement respectively the GA and the B&B algorithms. The proposed approach has been experimented on the 3D Quadratic Assignment Problem. In order to support the CPU cost of the hybridization mechanism, hierarchical parallel computing is used together with grid computing.
Type de document :
Communication dans un congrès
GECCO 2011, 13th Conf. on Genetic and Evolutionary Computation Conference, 2011, Dublin, Ireland. pp.513--520, 2011
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01222472
Contributeur : Jean-Michel Caricand <>
Soumis le : vendredi 30 octobre 2015 - 08:39:04
Dernière modification le : vendredi 22 mars 2019 - 01:35:47

Identifiants

  • HAL Id : hal-01222472, version 1

Citation

Malika Mehdi, Jean Charr, Nouredine Melab, El-Ghazali Talbi, Pascal Bouvry. A Cooperative Tree-based Hybrid GA-B&B Approach for Solving Challenging Permutation-based Problems. GECCO 2011, 13th Conf. on Genetic and Evolutionary Computation Conference, 2011, Dublin, Ireland. pp.513--520, 2011. 〈hal-01222472〉

Partager

Métriques

Consultations de la notice

573