RT info:eu-repo/semantics/article
T1 A novel framework for improving multi-population algorithms for dynamic optimization problems: a scheduling approach
A1 Novoa Hernández, Pavel
A1 Kordestani, Javidan Kazemi
A1 Ranginkaman, Amir Ehsan
A1 Meybodi, Mohammad Reza
A2 Ingeniería Informática y de Sistemas
K1 Dynamic optimization problems
K1 Differential evolution
K1 Moving peaks benchmark
K1 Evolutionary computation
K1 Scheduling
K1 Learning automata
AB This paper presents a novel framework for improving the performance of multi-population algorithms in solving dynamic optimization problems (DOPs). The fundamental idea of the proposed framework is to incorporate the concept of scheduling into multi-population methods with the aim to allocate more function evaluations to the best performing sub-populations. Two methods are developed based on the proposed framework, each of which uses a different approach for scheduling the sub-populations. The first method combines the quality of subpopulations and the degree of diversity among them into a single feedback parameter for detecting the best performing sub-population. The second method uses the learning automata as the central unit for performing the scheduling operation. In order to validate the applicability of the proposed methods, they are incorporated into three well-known algorithms for DOPs. The experimental results show the efficiency of the scheduling approach for improving the multi-population methods on the moving peaks benchmark (MPB) and generalized dynamic benchmark generator.
SN 2210-6502
YR 2019
FD 2019
LK http://riull.ull.es/xmlui/handle/915/41293
UL http://riull.ull.es/xmlui/handle/915/41293
LA en
DS Repositorio institucional de la Universidad de La Laguna
RD 25-abr-2025