Open AccessComparing the Robustness of Evolutionary Algorithms on the Basis of Benchmark Functions
Author(s) -
Ezgi Deniz Ülker,
Ali Haydar
Publication year - 2013
Publication title -
advances in electrical and computer engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.254
H-Index - 23
eISSN - 1844-7600
pISSN - 1582-7445
DOI - 10.4316/aece.2013.02010
Subject(s) - robustness (evolution) , benchmark (surveying) , computer science , algorithm , evolutionary algorithm , basis (linear algebra) , evolutionary computation , mathematical optimization , artificial intelligence , mathematics , biochemistry , geodesy , gene , geography , chemistry , geometry
In real-world optimization problems, even though the solution quality is of great importance, the robustness of the solution is also an important aspect. This paper investigates how the optimization algorithms are sensitive to the variations of control parameters and to the random initialization of the solution set for fixed control parameters. The comparison is performed of three well-known evolutionary algorithms which are Particle Swarm Optimization (PSO) algorithm, Differential Evolution (DE) algorithm and the Harmony Search (HS) algorithm. Various benchmark functions with different characteristics are used for the evaluation of these algorithms. The experimental results show that the solution quality of the algorithms is not directly related to their robustness. In particular, the algorithm that is highly robust can have a low solution quality, or the algorithm that has a high quality of solution can be quite sensitive to the parameter variations