A hybrid learning-based genetic and grey-wolf optimizer for global optimization

The grey-wolf optimizer (GWO) is a comparatively recent and competent algorithm in Swarm Intelligence (SI) to solve numerical and real-world optimization problems. However, the biggest challenge is the quick stabilization of its search agents to the local optima. Therefore, to bring effectiveness in the global search, it is imperative to relocate the leading agents through the procreation of their positions in the search space. This paper proposes GL-GWO, a genetic learning (GL)-based GWO, which imitates the genetic offspring generation scheme to improve the intelligence of GWO's leading agents. The GL scheme expedites the global effectiveness of leading agents by constructing the exemplars for them through genetic operators using their historical information. The obtained exemplars are well diversified and highly intelligent; therefore, the rest of the population's global searchability and search efficiency are enhanced under their guidance. The GL-GWO is tested on widely adopted 20 benchmark functions from the IEEE-CEC-2005 dataset and 38 functions from the IEEE-CEC-2014 dataset. The efficacy of GL-GWO is tested on four real-world engineering problems, namely recommendation systems, face image super-resolution, tension/compression spring, and welded beam. The obtained results on benchmark functions and considered engineering problems conclude that the GL-GWO is an efficient, effective, and reliable algorithm for solving real-world optimization problems.