A parallelized sequential random search global optimization algorithm

 P.M. Ortigosa, J. Balogh, I. García 1

  Acta Cybernetica 14 (1999) 403-418.

Abstract:

 This work deals with a stochastic global optimization algorithm, called CRS (Controlled Random Search), which originally was devised as a sequential algorithm. Our work is intended to analyze the degree of parallelism that can be introduced into CRS and to propose a new refined parallel CRS algorithm (RPCRS). As a first stage, evaluations of RPCRS were carried out by simulating parallel implementations. The degree of parallelism of RPCRS is controlled by a user given parameter whose value must be tuned to the size of the parallel computer system. It will be shown that the greater the degree of parallelism is the better the performance of the sequential and parallel executions are.

  Keywords: Parallel algorithm, Distributed Processing, Random Search, Global Optimization.

Footnotes

  1  This work was supported by the Tempus Program, by the Consejeria de Educación de la Junta de Andalucía (07/FSC/MDM) and by the Ministry of Education of Spain (CICYTTIC96-1125-C03-03).