A diversity preservation scheme for DSMGA-II to conquer the hierarchical difficulty
Journal
GECCO 2017 - Proceedings of the 2017 Genetic and Evolutionary Computation Conference
Pages
841-848
Date Issued
2017
Author(s)
Abstract
Hierarchical problems represent an important class of nearly decomposable problems and come from hierarchical complex systems. Complex systems are important since they appear in a variety of different areas. The dependency structure matrix genetic algorithm II, performing exploration and exploitation properly, requires fewer number of function evaluations on several problems than some well-known evolutionary algorithms such as the linkage tree genetic algorithm and the hierarchical bayesian optimization algorithm. However, DSMGA-II does not preserve enough promising subsolutions to the upper levels in hierarchical problems due to the back mixing operator of DSMGA-II, so it fails to solve the hierarchical trap problem. This paper proposes a diversity preservation scheme for DSMGA-II to conquer the hierarchical difficulty by calculating the entropies of subsolutions and determining whether to perform the back mixing. The empirical results show that our algorithm works well on hierarchical problems and does not compromise the performance on other problems.
SDGs
Type
conference paper