HybridParticleSwarmOptimiserwithBreedingandSubpopulationsMortenLøvbjergThomasKielRasmussenThiemoKrinkEvALifeprojectgroupEvALifeprojectgroupEvALifeprojectgroupDept.ofComputerscienceDept.ofComputerscienceDept.ofComputerscienceUniversityofAarhusUniversityofAarhusUniversityofAarhusDK-8000AarhusCDK-8000AarhusCDK-8000AarhusCDenmarkDenmarkDenmarklovbjerg@daimi.au.dkkiel@daimi.au.dkkrink@daimi.au.dkphone:+4589423357phone:+4589423357phone:+4589423358AbstractofthestandardGAandthePSO,couldleadtofurtherad-vances.InthispaperwepresenttwohybridParticleWepresentsuchahybridmodel.ThemodelincorporatesSwarmOptimiserscombiningtheideaofthepar-onemajoraspectofthestandardGAintothePSO,there-ticleswarmwithconceptsfromEvolutionaryAl-production.Inthefollowingwewillrefertotheusedre-gorithms.ThehybridPSOscombinethetradi-productionandrecombinationofgenesonlyas“breeding”.tionalvelocityandpositionupdateruleswiththeBreedingisoneofthecoreelementsthatmakesthestan-ideasofbreedingandsubpopulations.Bothhy-dardGAapowerfulalgorithm.HenceourhypothesiswasbridmodelsweretestedandcomparedwiththethataPSOhybridwithbreedinghasthepotentialtoreachstandardPSOandstandardGAmodels.ThisisabetteroptimumthanthestandardPSO.donetoillustratethatPSOswithbreedingstrate-Inadditiontobreedingweintroduceahybridwithbothgieshavethepotentialtoachievefasterconver-breedingandsubpopulations.Subpopulationshavepre-genceandthepotentialtofindabettersolution.viouslybeenintroducedtostandardGAmodelsmainlyTheobjectiveofthispaperistodescribehowtotopreventprematureconvergencetosuboptimalpointsmakethehybridsbenefitfromgeneticmethods([Spears94]).OurmotivationforthisextensionwasthattheandtotesttheirpotentialandcompetetivenessonPSOmodels,includingthehybridPSOwithbreeding,alsofunctionoptimisation.reachsuboptimalsolutions.Breedingbetweenparticlesindifferentsubpopulationswasalsoaddedasaninteractionmechanismbetweensubpopulations.1IntroductionTheintroducedhybridsweretestedagainstbothstandardTheParticleSwarmOptimisation