A novel optimization algorithm EMEL: Exploration of moving and ever-shrinking layers

This study presents a novel optimization algorithm EMEL: Exploration of Moving and Ever-shrinking Layers. EMEL's layer building strategy embodies a balance between exploration and exploitation. The search space is partitioned into layers that are bounded by concentric rectangles whose dimensions increase exponentially. Contrary to its popular counterparts, EMEL does not produce candidate solutions influenced by others. Instead, the candidate solutions are produced anywhere inside a randomly selected layer. The probability of selecting a layer in the close vicinity of the best solution is higher than elsewhere. Therefore, most of the solutions exploit the inner layers, where a few of them still explore the distal layers. To demonstrate its search abilities, EMEL was tested on a total of 120 benchmarks which consist of sixty standard benchmarks, fourteen CEC 2005 benchmarks, twenty-nine CEC 2017 benchmarks, thirteen constrained benchmarks and four real-life constrained engineering design problems. The benchmark set consisting various characteristics such as shifted/rotated functions, hybrid functions, composition functions and also constrained problems include equality and inequality constraints to be satisfied. Four algorithms; Grey Wolf Optimizer, Artificial Bee Colony, Particle Swarm Optimization and Vortex Search Algorithm were included in the experiments to carry out a comparative work. Rather than using fixed parameters, design parameters of all algorithms were adjusted for each benchmark. The results were compared and analysed by utilizing Mann Whitney U test. The results showed that EMEL is superior to its counterparts by achieving high scores in the tests.