Felix Streichert, Holger Ulmer, and Andreas Zell

Parallelization of Multi-Objective Evolutionary Algorithms using Clustering Algorithms

Proceedings of the Conference on Evolutionary Multi-Criterion Optimization (EMO 2005)

Abstract

While single-objective Evolutionary Algorithms (EAs) parallelization schemes are both well established and easy to implement, this is not the case for Multi-Objective Evolutionary Algorithms (MOEAs). Nevertheless, the need for parallelizing MOEAs arises in many real-world applications, where fitness evaluations and the optimization process can be very time consuming. In this paper, we test the `divide and conquer' approach to parallelize MOEAs, aimed at improving the speed of convergence beyond a parallel island MOEA with migration. We also suggest a clustering based parallelization scheme for MOEAs and compare it to several alternative MOEA parallelization schemes on multiple standard multi-objective test functions.

Download

[pdf]

BibTeX 

    
@inproceedings{Streichert05Parallelization,
  author    = {F. Streichert and H. Ulmer and A. Zell},
  title     = {Parallelization of Multi-Objective Evolutionary Algorithms using Clustering Algorithms},
  booktitle = {Conference on Evolutionary Multi-Criterion Optimization  {(EMO 2005)}},
  year 	    = {2005},
  month     = {9-11 March},
  editor    = {Carlos A. {Coello Coello} and Arturo Hernandez Aguirre and Eckart Ziztler},
  series    = {LNCS},
  volume    = {3410},
  address   = {Guanajuato, Mexico},
  publisher = SPRINGER,
  pages     = {92-107},
  keywords  = {evolutionary computation},
  ISBN 	    = {3-540-24983-4}  
}