MCDM'08 - paper no. 4


 

Back to MCDM'08 contents
 

A study of distributed evolutionary algorithms for multi-objective optimisation

Abdelbasset Essabri, Mariem Gzara, Taicir Loukil

Abstract:

Most popular Evolutionary Algorithms for single multi-objective optimization are motivated by the reduction of the computation time and the resolution larger problems. A promising alternative is to create new distributed schemes that improve the behavior of the search process of such algorithms. In the multi-objective optimization problems, more exploration of the search space is required to obtain the whole or the best approximation of the Pareto front. Almost all proposed Parallel Multi-Objective Evolutionary Algorithms (PMOEAs) are based on the specialization concept which means dividing the objective and/or the search space then assigning each part to a processor. One processor called the organizer or the coordinator is usually charged to direct the whole algorithm. In this paper, we present a new parallel scheme of multi-objective evolutionary algorithms which is based on a clustering technique. This new parallel algorithm is implemented and compared to three PMOEAs which are cone-separation [1], Divided Range Multi-Objective Genetic Algorithm (DRMOGA) [8] and a Parallel Strength Pareto Evolutionary Algorithm (PSPEA) based on the island model without migration.

Keywords:

Parallel computing, multi-objective optimisation, evolutionary algorithms, parallel genetic algorithms, clustering algorithms

Reference index:

Abdelbasset Essabri, Mariem Gzara, Taicir Loukil, (2009), A study of distributed evolutionary algorithms for multi-objective optimisation, Multiple Criteria Decision Making (4), pp. 89-106

Full text:

download