Published Articles >> Table of Contents >> Abstract

International Parallel and Distributed Processing Symposium (IPDPS'03)   p. 144a
ParadisEO: A Framework for Parallel and Distributed Metaheuristics
Sebastien Cahon, Université des Sciences et Technologies de Lille1
El-Ghazali Talbi, Université des Sciences et Technologies de Lille1
Nordine Melab, Université des Sciences et Technologies de Lille1

Full Article Text:

DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/IPDPS.2003.1213274
Send link to a friend

Abstract
In this paper we present PARADISEO , an open source framework for flexible parallel and distributed design of hybrid metaheuristics. Flexibility means that the parameters such as data representation and variation operators can be evolved. It is inherited from the EO object-oriented library for evolutionary computation. PARADISEO provides different parallel and/or distributed models and allows a tranparent multi-threaded implementation. Moreover, it supplies different natural hybridization mechanisms mainly for metaheuristics including evolutioanry algorithms and local search methods. The framework is experimented here in the spectroscopic data mining field. The flexibilty property allowed an easy and straightforward development of a GA-based attribute selection for models discovery in NIR spectroscopic data. Experiments on a cluster of SMPs (IBM SP3) show that a good speed-up is achieved by using the provided parallel distributed models and multi-threading. Furthermore, the hybridization of the GA with the efficient PLS method allows to discover high-quality models. Indeed, their accuracy and understandability are improved respectively by 37% and 88%.
Back to Top Additional Information
Index Terms- Parallelism, Distributed Systems, Hybrid Metaheuristics, Flexible Library, NIR Spectroscopic Data Mining Citation:  Sebastien Cahon, El-Ghazali Talbi, Nordine Melab, "ParadisEO: A Framework for Parallel and Distributed Metaheuristics," ipdps, p. 144a,  International Parallel and Distributed Processing Symposium (IPDPS'03),  2003