Neural network and fuzzy system for the tuning of Gravitational Search Algorithm parameters. (15th July 2018)

• Record Type: Journal Article
• Authors: Pelusi, Danilo; Mascella, Raffaele; Tallini, Luca; Nayak, Janmenjoy; Naik, Bighnaraj; Abraham, Ajith
• Is Part Of: Expert systems with applications. Volume 102(2018)
• Page Start: 234
• Page End: 244
• Publication Date: 2018-07-15
• Subjects: Computational intelligence -- Fuzzy systems -- Gravitational Search Algorithm -- Neural networks -- Search algorithms.; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2018.02.026
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: Neural network and Fuzzy Inference system for the tuning of GSA parameters are designed. A revised version of GSA is proposed with a new parameters definition. The proposed algorithm improves GSA with the same computational complexity. Abstract: A good trade-off between exploration and exploitation to find optimal values in search algorithms is very hard to achieve. On the other hand, the combination of search methods may cause computational complexity increase problems. The Gravitational Search Algorithm (GSA) is a swarm optimization algorithm based on the law of gravity, where the solution search process depends on the velocity of particles. The application of intelligent techniques can improve the search performances of GSA. This paper proposes the design of a Neuro and Fuzzy Gravitational Search Algorithm (NFGSA) to achieve better results than GSA in terms of global optimum search capability and convergence speed, without increasing the computational complexity. Both the algorithms have the same computational complexity O ( nd ), where n is the number of agents and d is the search space dimension. The main task of the designed intelligent system is to adjust a GSA parameter on a revised version of GSA. NFGSA is compared with GSA, a Plane Surface Gravitational Search Algorithm (PSGSA) and a Modified Gravitational Search Algorithm (MGSA). The results show that NFGSA improves the optimization performances of GSA and PSGSA, without adding computational costs. Moreover, the proposed algorithm is better than MGSA for a benchmark function and achieves similar results for two test functions. The analysis on the computational complexity shows that NFGSA has a better computational complexity than MGSA, because NFGSA has complexity O ( nd ), whereas MGSA has complexity O (( nd ) 2 ).