Title: Two algorithmic enhancements for the parallel differential evolution

Authors: Matthieu Weber, Ferrante Neri, Ville Tirronen

Addresses: Department of Mathematical Information Technology, University of Jyvaskyla, FIN-40014 Jyvaskyla, Finland. ' Department of Mathematical Information Technology, University of Jyvaskyla, FIN-40014 Jyvaskyla, Finland. ' Department of Mathematical Information Technology, University of Jyvaskyla, FIN-40014 Jyvaskyla, Finland

Abstract: This paper proposes the use of two algorithms based on the parallel differential evolution. The first algorithm proposes the use of endemic control parameters within a parallel differential evolution algorithm; the differential evolution running at each subpopulation is associated with randomly initialised scale factor and crossover rate, which are then repeatedly updated during the optimisation process. The second algorithm proposes decomposing the search space of large-scale problems into lower-dimensionality subspaces, and associating each of these to one subpopulation of a parallel differential evolution algorithm. Each subpopulation is running a modified differential evolution algorithm, where the crossover function is limited to components of the subpopulation|s associated subspace. According to numerical results, both algorithms seem to be clear improvements over the original parallel distributed evolution; they are simple, robust, and efficient algorithms suited for various applications.

Keywords: evolutionary algorithms; structured population; parallel differential evolution; PDE; component decomposition; endemic control parameters.

DOI: 10.1504/IJICA.2011.037948

International Journal of Innovative Computing and Applications, 2011 Vol.3 No.1, pp.20 - 30

Received: 22 Jul 2010
Accepted: 26 Jul 2010
Published online: 21 Mar 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article