The Grey Wolf Optimizer (GWO) algorithm is an interesting swarm-based optimization algorithm for global optimization. It was inspired by the hunting strategy and leadership hierarchy of Grey wolves. The GWO algorithm has been successfully tailored to solve various continuous and discrete optimization problems. However, the main drawback of GWO is that it may converge to sub-optimal solutions in early stages of its simulation process due to the loss of diversity in its population. This paper introduces a Distributed variation of GWO (DGWO) that attempts to enhance the diversity of GWO by organizing its population into small independent groups (islands) based on a well-known distributed model called the island model. DGWO applies the original GWO to each island and then allows selected solutions to be exchanged between the islands based on the random ring topology and the best-worst migration policy. The island model in DGWO provides a better environment for unfit candidate solutions in each island to evolve into better solutions which increases the likelihood of finding global optimal solutions. Another interesting feature about DGWO is that it can run in parallel devices which means that its computational complexity can be reduced compared to the computational complexity of existing variations of GWO. DGWO was evaluated and compared to well-known swarm-based optimization algorithms using 30 CEC 2014 functions. In addition, the sensitivity of DGWO to its parameters was evaluated using 15 standard test functions. The comparative study and the sensitivity analysis for DGWO indicate that it provides competitive performance compared to the other tested algorithms. The source code of DGWO is available at:
https://www.dropbox.com/s/2d16t46598u03y0/DistributedGreyWolfOptimizer.zip?dl=0
Key words: Grey Wolf Optimizer, Island-based Model , Distributed Optimization Algorithm, Optimization, Swarm-based Optimization
|
