A heuristic to determine the initial gravitational constant of the GSA

• Main Authors: Barbosa, Alfredo J. P, Moreira, Edmilson M, Moraes, Carlos H. V, Carpinteiro, Otávio A. S
• Format: Journal Article
• Language: English
• Published: 21-04-2022
• Subjects: Computer Science - Artificial Intelligence; Computer Science - Neural and Evolutionary Computing
• DOI: 10.48550/arxiv.2205.06770

The Gravitational Search Algorithm (GSA) is an optimization algorithm based
on Newton's laws of gravity and dynamics. Introduced in 2009, the GSA already
has several versions and applications. However, its performance depends on the
values of its parameters, which are determined empirically. Hence, its
generality is compromised, because the parameters that are suitable for a
particular application are not necessarily suitable for another. This paper
proposes the Gravitational Search Algorithm with Normalized Gravitational
Constant (GSA-NGC), which defines a new heuristic to determine the initial
gravitational constant of the GSA. The new heuristic is grounded in the
Brans-Dicke theory of gravitation and takes into consideration the multiple
dimensions of the search space of the application. It aims to improve the final
solution and reduce the number of iterations and premature convergences of the
GSA. The GSA-NGC is validated experimentally, proving to be suitable for
various applications and improving significantly the generality, performance,
and efficiency of the GSA.