Open AccessHybrid particle swarm optimization with spiral-shaped mechanism for solving high-dimension problems
Author(s) -
Humberto Martins Mendonça Duarte,
Rafael Lima de Carvalho
Publication year - 2020
Publication title -
academic journal on computing, engineering and applied mathematics
Language(s) - English
Resource type - Journals
ISSN - 2675-3588
DOI - 10.20873/uft.2675-3588.2020v1n1p1
Subject(s) - metaheuristic , local optimum , mathematical optimization , particle swarm optimization , multi swarm optimization , benchmark (surveying) , inertia , convergence (economics) , dimension (graph theory) , local search (optimization) , position (finance) , mathematics , computer science , optimization problem , physics , geodesy , finance , classical mechanics , economic growth , pure mathematics , economics , geography
Particle swarm optimization (PSO) is a well-known metaheuristic, whose performance for solving global optimization problems has been thoroughly explored. It has been established that without proper manipulation of the inertia weight parameter, the search for a global optima may fail. In order to handle this problem, we investigate the experimental performance of a PSO-based metaheuristic known as HPSO-SSM, which uses a logistic map sequence to control the inertia weight to enhance the diversity in the search process, and a spiral-shaped mechanism as a local search operator, as well as two dynamic correction factors to the position formula. Thus, we present an application of this variant for solving high-dimensional optimization problems, and evaluate its effectiveness against 24 benchmark functions. A comparison between both methods showed that the proposed variant can escape from local optima, and demonstrates faster convergence for almost every evaluated function.