Exploring a Two-Population Genetic Algorithm

Steven Orla Kimbrough¹, Ming Lu¹, David Harlan Wood², and D.J. Wu³

¹University of Pennsylvania,
3730 Walnut Street,
Philadelphia, PA 19104-6340
{kimbrough,milu}@wharton.upenn.edu

²University of Delaware,
CIS Dept.,
Newark, DE 19716
wood@cis.udel.edu

³DuPree College of Management,
Georgia Institute of Technology,
Atlanta, GA30332
wudj@drexel.edu

Abstract. In a two-market genetic algorithm applied to a constrained optimization problem, two `markets' are maintained. One market establishes fitness in terms of the objective function only; the other market measures fitness in terms of the problem constraints only. Previous work on knapsack problems has shown promise for the two-market approach. In this paper we: (1) extend the investigation of two-market GAs to nonlinear optimization, (2) introduce a new, two-population variant on the two-market idea, and (3) report on experiments with the two-population, two-market GA that help explain how and why it works.

LNCS 2723, p. 1148 ff.

Full article in PDF

lncs@springer.de
© Springer-Verlag Berlin Heidelberg 2003