Multi-guide particle swarm optimization for many-objective optimization problems
Date
2021-03
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT: The scalability of the multi-guide particle swarm optimization (MGPSO) algorithm,
with respect to the number of objectives for a problem, is investigated.
Two MGPSO algorithm adaptations are proposed; that is, the partialdominance
multi-guide particle swarm optimization (PMGPSO) algorithm and
the knee-point driven multi-guide particle swarm optimization (KnMGPSO)
algorithm. As a sub-objective, the effect of different archive balance coefficient
update strategies for the MGPSO, the PMGPSO, and the KnMGPSO
algorithms are investigated. The proposed algorithms attempt to address the
scalability limitations associated with a certain component of the MGPSO algorithm.
This study does not consider scalability with respect to the number
of decision variables. This study assumes a static search space; that is, where
the number of objectives remains fixed throughout the optimization. This
study also assumes that each objective remains static throughout the search
process. This study also considers only problems with boundary constraints.
The results indicate that the MGPSO algorithm scaled to many-objectives
competitively compared to other state-of-the-art many-objective optimization
algorithms. The results were unexpected because the MGPSO algorithm uses
the Pareto-dominance relation, which is known to degrade as the number
of objectives increases. The proposed PMGPSO and KnMGPSO algorithms
also scaled competitively, however, these algorithms were not superior to the
MGPSO algorithm. The investigated dynamic archive balance coefficient update
strategies did not improve the performance of the MGPSO, the PMGPSO,
or the KnMGPSO algorithms. Keywords: Multi-guide particle swarm optimization, particle swarm
optimization, many-objective optimization, scalability
AFRIKAANSE OPSOMMING: Geen Afrikaanse opsomming beskikbaar.
AFRIKAANSE OPSOMMING: Geen Afrikaanse opsomming beskikbaar.
Description
Thesis (MSc)--Stellenbosch University, 2021.
Keywords
Optimization problems, Swarm intelligence, Optimization algorithms, Mathematical optimization, Computer networks -- Scalability, UCTD