Solution recombination in an indicator-based many-objective ant colony optimizer for continuous search spaces
Ashraf M. Abdelbar, Khalid M. Salama · 2017
In this paper, we present a recombination-based extension of the recently-introduced iMOACORalgorithm. iMOACORitself is based on ACOR, an Ant Colony Optimization (ACO) optimizer for continuous search spaces, and extends ACORto Multi-Objective Optimization (MOO) problems. Our proposal uses an externally supplied probability to decide whether to apply a recombination operator or to invoke iMOACOR's usual solution construction mechanism. In the former case, one parent is selected from the iMOACORsolution archive by rank-proportionate selection (as is the case in ACOR), and the other parent is selected by uniform selection. Our proposal is compared to iMOACORusing 64 standard problems from the MOO literature with number of objectives ranging from 3 to 10, and found to perform better than iMOACORto a statistically significant extent, as assessed by the hypervolume indicator.