Parallel random injection differential evolution

  • Authors:

  • Matthieu Weber;Ferrante Neri;Ville Tirronen

  • Affiliations:

  • Department of Mathematical Information Technology, University of Jyväskylä, (Agora), Finland;Department of Mathematical Information Technology, University of Jyväskylä, (Agora), Finland;Department of Mathematical Information Technology, University of Jyväskylä, (Agora), Finland

  • Venue:
  • EvoApplicatons'10 Proceedings of the 2010 international conference on Applications of Evolutionary Computation - Volume Part I
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper proposes the introduction of a generator of random individuals within the ring topology of a Parallel Differential Evolution algorithm. The generated random individuals are then injected within a sub-population. A crucial point in the proposed study is that a proper balance between migration and random injection can determine the success of a distributed Differential Evolution scheme. An experimental study of this balance is carried out in this paper. Numerical results show that the proposed Parallel Random Injection Differential Evolution seems to be a simple, robust, and efficient algorithm which can be used for various applications. An important finding of this paper is that premature convergence problems due to an excessively frequent migration can be overcome by the injection of random individuals. In this way, the resulting algorithm maintains the high convergence speed properties of a parallel algorithm with high migration but differs in that it is able to continue improving upon the available genotypes and detect high quality solutions.