Performance analysis of localisation strategy for island model genetic algorithm in population diversity preservation

ABSTRACT

The genetic algorithm (GA) is one of the most common solutions to solve many optimisation problems. Its distributed version, Island Model GA (IMGA), was introduced to overcome more complex and scalable cases. However, there is a recurrent problem in IMGA called premature convergence as a consequence of selection in the migration. This process is a mechanism of migrating individuals from one into another island to keep population diversity. The primary cause is the structural similarity of a migrated individual because of the genetic operator configurations are identical. Localised IMGA (LIMGA) tries to implement different island characteristics to avoid premature convergence. The main motivation of this paper is to investigate the performance of LIMGA capability in maintaining population diversity. In detail, the contributions of this research are (1) to prove LIMGA concept in handling general optimisation problem, (2) to analyse the performance LIMGA in diversity preservation, and (3) compare LIMGA performance with the current solvers. By harmonising three different GA cores, LIMGA could overcome computationally expensive functions with a great result and acceptable execution time. Moreover, because of its success in maintaining the diversity, Localised Island Model Genetic Algorithm (LIMGA) could lead to the among other current solvers for this case.

KEYWORDS:

• Genetic algorithms
• Island model genetic algorithm
• localisation strategy
• computationally expensive optimisation

Acknowledgments

Indonesia Endowment Fund for Education (LPDP), a scholarship from the Ministry of Finance, Republic of Indonesia, supports this work. We conduct this research while at the Graduate School of Information, Production, and Systems, Waseda University, Japan.

Disclosure statement

No potential conflict of interest was reported by the authors.