20.500.12556/DKUM-88983 Koevolucijski algoritem roja delcev z metodo rekurzivnega diferencialnega grupiranja za reševanje problemov velikih dimenzij magistrsko delo Coevolution particle swarm algorithm with recursive differential grouping method for solving large-scale optimization problems Kooperativna koevolucija je podzvrst evolucijskega računanja, ki se uporablja kot ogrodje za optimizacijo problemov z velikim številom dimenzij preko pristopa deli in vladaj. Glavni izzivi uporabe ogrodja kooperativne koevolucije ležijo v dekompoziciji problema ter v uporabi primernega optimizacijskega algoritma. Dekompozicija se v glavnem ukvarja z deljenjem problema v manjše podprobleme, kjer je glavni izziv, kako ugotoviti povezave med komponentami problema. V našem delu smo razvili kooperativni koevolucijski algoritem, ki uporablja rekurzivne strategije diferencialnega grupiranja za dekompozicijo problema, ter algoritem roja delcev, kot optimizacijski algoritem. V delu smo analizirali šest optimizacijskih algoritmov roja delcev na naboru testnih funkcij iz CEC2013, ki spadajo v probleme z velikim številom dimenzij, ter je njihova dimenzionalnost 1000. Na podlagi te analize smo v naš predlagan kooperativni koevolucijski algoritem vključili optimizacijski algoritem roja delcev, ki se je najbolje izkazal na naboru izbranih funkcij. Izvedli smo primerjalno analizo med najboljšim algoritmom roja delcev in predlaganimi kooperativnimi koevolucijskimi algoritmi, kjer smo uporabili pet različnih strategij rekurzivnega diferencialnega grupiranja. Ugotovili smo, da kooperativni koevolucijski algoritem deluje boljše od algoritmov roja delcev, ki smo jih uporabili v našem delu. Prav tako smo ugotovili, da izbira strategije dekompozicije problema igra pomembno vlogo. Cooperative coevolution is a subgenre of evolutionary computation used as a framework for optimizing high-dimensional problems through a divide-and-conquer approach. The main challenges of using the framework of cooperative coevolution lie in the decomposition of the problem and the use of a suitable optimization algorithm. Decomposition is mainly concerned with dividing the problem into smaller sub-problems, where the main challenge is determining the connections between the components of the problem. In our work, we developed a cooperative coevolutionary algorithm that uses a recursive differential grouping algorithm for problem decomposition, and a particle swarm algorithm as an optimization algorithm. In the work, we analyzed six particle swarm optimization algorithms on a set of test functions from CEC2013, which belong to large-scale optimization problems that have a large number of dimensions, and their dimensionality is 1000. Based on this analysis, we included the particle swarm optimization algorithm in our proposed cooperative coevolution algorithm, which performed best on the selected set of functions. We conducted a comparative analysis between the best particle swarm algorithm and the proposed cooperative coevolutionary algorithm, where we used five different recursive differential grouping strategies. We found that the cooperative coevolution algorithm performs better than the best particle swarm algorithm used in this work. We also found that the choice of problem decomposition strategy plays an important role. roj delcev rekurzivno diferencialno grupiranje kooperativna koevolucija optimizacija veliko število dimenzij particle swarm recursive differential grouping cooperative coevolution optimization large scale true true false [K. Berkovič] Slovenski jezik Angleški jezik Magistrsko delo/naloga 2024-06-03 16:52:11 2024-07-01 11:52:16 2024-08-31 03:50:39 0000-00-00 00:00:00 2024 0 Maribor 0 Maribor 1 spletni vir (1 datoteka PDF (XV, 84 f.)) 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.421(043.2) 205822467 MAG_Berkovic_Klemen_2024.pdf MAG_Berkovic_Klemen_2024.pdf 1 6CF08C8A2FCB98FA6D9DE0632AFFC8DC 627b3a958a661326aa3005ebf85ac0922973434798c600aa8e72323af5269555 3cdbed9d-21bb-11ef-bc37-00155d000105 https://dk.um.si/Dokument.php?lang=slv&id=180136 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0