Eventos Anais de eventos
MecSol 2017
6th International Symposium on Solid Mechanics
Investigation of the diversity index trend of the Search Group Algorithm in comparison with others meta-heuristic algorithms
Submission Author:
Rafael Souza , SC
Co-Authors:
Leandro Fadel Miguel, Matheus Silva Gonçalves, Rafael Holdor Lopez
Presenter: Rafael Souza
doi://10.26678/ABCM.MecSol2017.MSL17-0009
Abstract
Several meta-heuristic optimization algorithms have been developed and tested over the last few decades. The development of these algorithms continues to be an active field of research, mainly because of their ability to solve real life complex problems. However, one of the issues that meta-heuristic algorithms present is that their efficiency is problem dependent, i.e. the performance of a given algorithm solving a certain class of problems does not guarantee its good performance for other problems. Thus, the performance of a meta-heuristic algorithm may only be assessed a-posteriori, e.g. analyzing the best result, mean value, standard deviation and convergence curve of several independent runs. In the context of meta-heuristic algorithms, it is well-recognized that they should have two capabilities, exploration and exploitation, in order to be able to find reasonable solutions. Exploration may be described as the ability of the algorithm to find promising regions on the design domain, while exploitation is the ability of the algorithm to refine the solution on these promising regions. It is important for meta-heuristic algorithms to maintain an adequate balance between the exploration and exploitation tendencies in order to be competitive in terms of robustness and performance. Kaveh and Zolghadr (2014) presents the so called diversity index, which is a practical approach to analyze these features. One of the conclusions of the mentioned study is that the diversity index plays a major role in understanding the behavior of meta-heuristic algorithms. Indeed, it is indicated that satisfactory results are related to a specific behavior of the diversity index trend. Thus, the main goal of this paper is to show that using the Search Group Algorithm (SGA) (developed by Gonçalves et. al (2015)), the designer is able to choose a-priori the diversity index trend and consequently a good balance between exploration and exploitation. In contrast to other meta-heuristic algorithms, where the correlation between the parameters and the diversity index trend is not clear. To achieve this goal, several benchmark optimization problems are solved, employing the SGA, the FA (firefly algorithm) and the BSA (Backtrack Search Algorithm).
Keywords
Optimization algorithm, Search Group Algorithm, Diversity index, meta-heuristic algorithms
