A CUDA-Based Cooperative Evolutionary Multi-Swarm Optimization Applied to Engineering Problems
Laboratory of Natural Computing (LCN), Area of Exact and Natural Sciences (ACET), University Centre of Para (CESUPA), Belem, Brazil
CSBC, 2012
@article{souza2012cuda,
title={A CUDA-Based Cooperative Evolutionary Multi-Swarm Optimization Applied to Engineering Problems},
author={Souza, Daniel Leal and Roberto, Ot{‘a}vio Noura Teixeira1 Dionne Monteiro and de Oliveira, C{‘e}lio Lim{~a}o},
year={2012}
}
This paper presents a variation of Evolutionary Particle Swarm Optimization applied to the concept of master/slave swarm with mechanism of sharing data for the acceleration of convergence. The implementation called Cooperative Evolutionary MultiSwarm Optimization on Graphics Processing Units (CMEPSOGPU) consists in using thousands of threads in various slave swarms on the CUDA parallel architecture, where each one works in a parallel and cooperative way in order to improve the search for best result and reduce the number of iterations. The use of CMEPSO-GPU applied to engineering problems showed superior results when compared to other implementations found in the scientific literature.
March 20, 2013 by hgpu
