AN EFFICIENT PRIMARY POPULATION INITIALIZATION METHOD FOR METAHEURISTIC ALGORITHMS
A.V. VARDUMYAN · 2022
It is widely recognized that convergence capabilities of metaheuristic optimization algorithms can be enhanced by properly chosen initial population. Most of the state-of-theart initialization techniques are suffering from numerous shortcomings, that ultimately make them non-viable or not efficient enough. To overcome this, this paper proposes a new algorithm for population initialization, which adopts the approach of dividing the search space into nested cubes and picking edge-points for sampling. Based on the data obtained after testing the method on 8 complex benchmark functions against other popular initialization strategies in the scope of WOA algorithm, the proposed approach outperformed all of the candidates in finding the global optimum.