Intelligent Search Strategies of Global Optima in Nature Inspired Metaheuristics

Bouthina Saib¹*, Mohamed-Rida Abdessemed², Riadh Hocin³ and Kamel Khoualdi^4
1,2,3Computer Science Department, Batna 2 University, Batna, Algeria
⁴Management Information Systems Department, King Abdulaziz University, Jeddah, Saudi Arabia
*Corresponding Author: Bouthina Saib, Computer Science Department, Batna 2 University, Batna, Algeria.

Published: March 29, 2023

View Pdf

Abstract  

Global optimization sets itself the mission of identifying the most interesting solutions in the overall search space. But actually, it is only the best of all discovered candidate solutions in the explored search space, which depends, in turn, on initial positions of searching points. It is practically impossible to cover the entire search space of NP-hard problems, in a reasonable time, whatever the method used, since the size of this space exceed the capabilities of all sophisticated algorithms implemented on any powerful computer even parallel ones. That is what justify the increased interest to nature-inspired metaheuristics, currently, based on a strategy of balance between exploitation and exploration, which make their approaches looking like smart. Exploration guarantees that the used algorithm will reach the widest possible extent of the undiscovered areas, whereas exploitation guarantees that such algorithm will search for the best solutions inside the most promising areas, already discovered. Authors study, in this work, different implementations of exploration and exploitation mechanisms utilized in many well-known nature-inspired metaheuristics applied to both uni-model and multi-model benchmarks. Uni-model benchmarks aim to test exploitation while multi-model benchmarks aim to test exploration. Obtained results show the superiority of TSO metaheuristic to find the adequate balance between exploration and exploitation leading it to discover in all tested cases, at least, one of the global optima or one of the best near global optima.

Keywords: Smart search; Load balancing; Exploitation; Exploration; Relative best solution; Nature inspired metaheuristics