Manipulating subpopulations of feasible and infeasible solutions in genetic algorithms

1 January 1993

conference paper
Published by Association for Computing Machinery (ACM)

p. 118-125
https://doi.org/10.1145/162754.162831

Abstract

This paper explores the partitioning of the population pool in genetic algorithms into separatesubpopulationsof feasible and infeasible solutions, and the interactionon a regular basis of crossover operations among and within the subPopulations. The set covering problem was chosen as a representative optimization problem to apply our subpopuIation strategies. We designed a class of nine algorithms for manipulating the two population pools and compared this against the traditional GA. The traditional GA uses a single population pool where infeasible solutions are generalIy considered infrequently or ignored. All of our algorithms significantly and consistently outperformed the traditional GA in all of the test problems illustrating the importance of infeasible solutions as a source of good genetic material. Furthermore, results show that the random select consistently outperformed the bias select from the infeasible pool suggesting all infeasible solutions should be considered equal.

Keywords

This publication has 0 references indexed in Scilit: