Synchronous and asynchronous Pareto-based multi-objective Artificial Bee Colony algorithms


Akay B.

JOURNAL OF GLOBAL OPTIMIZATION, cilt.57, ss.415-445, 2013 (SCI İndekslerine Giren Dergi) identifier identifier

  • Cilt numarası: 57 Konu: 2
  • Basım Tarihi: 2013
  • Doi Numarası: 10.1007/s10898-012-9993-1
  • Dergi Adı: JOURNAL OF GLOBAL OPTIMIZATION
  • Sayfa Sayıları: ss.415-445

Özet

Pareto-based multi-objective optimization algorithms prefer non-dominated solutions over dominated solutions and maintain as much as possible diversity in the Pareto optimal set to represent the whole Pareto-front. This paper proposes three multi-objective Artificial Bee Colony (ABC) algorithms based on synchronous and asynchronous models using Pareto-dominance and non-dominated sorting: asynchronous multi-objective ABC using only Pareto-dominance rule (A-MOABC/PD), asynchronous multi-objective ABC using non-dominated sorting procedure (A-MOABC/NS) and synchronous multi-objective ABC using non-dominated sorting procedure (S-MOABC/NS). These algorithms were investigated in terms of the inverted generational distance, hypervolume and spread performance metrics, running time, approximation to whole Pareto-front and Pareto-solutions spaces. It was shown that S-MOABC/NS is more scalable and efficient compared to its asynchronous counterpart and more efficient and robust than A-MOABC/PD. An investigation on parameter sensitivity of S-MOABC/NS was presented to relate the behavior of the algorithm to the values of the control parameters. The results of S-MOABC/NS were compared to some state-of-the art algorithms. Results show that S-MOABC/NS can provide good approximations to well distributed and high quality non-dominated fronts and can be used as a promising alternative tool to solve multi-objective problems with the advantage of being simple and employing a few control parameters.