Parallel Hybrid Multi-Objective Island Model in Peer-to-Peer Environment

Author

Melab, N. ; Mezmaz, M. ; Talbi, E.-G.

Author_Institution

LIFL, CNRS, France

fYear

2005

fDate

04-08 April 2005

Abstract

Solving large size and time-intensive combinatorial optimization problems with parallel hybrid multi-objective evolutionary algorithms (MO-EAs) requires a large amount of computational resources. Peer-to-Peer (P2P) computing is recently revealed as a powerful way to harness these resources and efficiently deal with such problems. In this paper, we focus on the parallel hybrid multi-objective island model for P2P systems. We address its design, implementation, and fault-tolerant deployment in a P2P context. The proposed model have been experimented on the Bi-criterion Permutation Flow-Shop Problem (BPFSP) on a network of 120 heterogeneous PCs. The preliminary results demonstrate the effectiveness of this model and its capabilities to fully exploit the hybridization.

Keywords

evolutionary computation; fault tolerant computing; middleware; parallel algorithms; peer-to-peer computing; scheduling; search problems; bi-criterion permutation flow-shop problem; combinatorial optimization problem; fault-tolerance computing; local search; middleware; multiobjective evolutionary algorithm; parallel hybrid multiobjective island model; peer-to-peer environment; Concurrent computing; Dolphins; Evolutionary computation; Fault tolerance; Middleware; Peer to peer computing; Personal communication networks; Power system modeling; Robustness; Scheduling; Flow-Shop; Hybridization; Local Search; Multi-objective Evolutionary Algorithms; Parallel Island Model; Peer-to-Peer Computing;

fLanguage

English

Publisher

ieee

Conference_Titel

Parallel and Distributed Processing Symposium, 2005. Proceedings. 19th IEEE International

Print_ISBN

0-7695-2312-9

Type

conf

DOI

10.1109/IPDPS.2005.327

Filename

1420079