GroupSearchOptimizer:AnOptimizationAlgorithmInspiredbyAnimalSearchingBehavior
S.He,Member,IEEE,Q.H.Wu,IEEE,SeniorMember,andJ.R.Saunders
Abstract—Nature-inspiredoptimizationalgorithms,notablyevolutionaryalgorithms(EAs),havebeenwidelyusedtosolvevariousscienti candengineeringproblemsbecauseoftotheirsimplicityand exibility.Herewereportanoveloptimizationalgorithm,groupsearchoptimizer(GSO),whichisinspiredbyanimalbehavior,especiallyanimalsearchingbehavior.Theframeworkismainlybasedontheproducer–scroungermodel,whichassumesthatgroupmemberssearcheitherfor“ nding”(producer)orfor“joining”(scrounger)opportunities.Basedonthisframework,conceptsfromanimalsearchingbehavior,e.g.,animalscanningmechanisms,areemployedmetaphoricallytodesignoptimumsearchingstrategiesforsolvingcontinuousopti-mizationproblems.Whentestedagainstbenchmarkfunctions,inlowandhighdimensions,theGSOalgorithmhascompetitiveperformancetootherEAsintermsofaccuracyandconvergencespeed,especiallyonhigh-dimensionalmultimodalproblems.TheGSOalgorithmisalsoappliedtotrainarti cialneuralnetworks.Thepromisingresultsonthreereal-worldbenchmarkproblemsshowtheapplicabilityofGSOforproblemsolving.
IndexTerms—Animalbehavior,behavioralecology,evolution-aryalgorithm,optimization,swarmintelligence.
I.INTRODUCTION
NTHEPASTfewdecades,nature-inspiredcomputationhasattractedmoreandmoreattention[1].Natureservesasafertilesourceofconcepts,principles,andmechanismsfordesigningarti cialcomputationsystemstotacklecomplexcomputationalproblems.Amongthem,themostsuccessfulareevolutionaryalgorithms(EAs)whichdrawinspirationfromevolutionbynaturalselection.Currently,thereareseveraldif-ferenttypesofEAswhichincludegeneticalgorithms(GAs),geneticprogramming(GP),evolutionaryprogramming(EP),andevolutionarystrategies(ES)[2].
Inrecentyears,anewkindofcomputationalintelligenceknownasswarmintelligence(SI),whichwasinspiredbycollectiveanimalbehavior,hasbeendeveloped.Currently,SIincludestwodifferentalgorithms.The rstoneisantcolonyoptimization(ACO),whichwasdevelopedbasedonants’foragingbehavior[3].AnotherSIalgorithmisparticleswarmoptimizer(PSO),whichgleanedideasfromswarmbehavior
ManuscriptreceivedSeptember23,2007;revisedJune20,2008,September11,2008,andOctober16,2008;acceptedDecember14,2008.CurrentversionpublishedSeptember30,2009.
S.HeandQ.H.WuarewiththeDepartmentofElectricalEngineeringandElectronics,UniversityofLiverpool,Liverpool,L693GJ,U.K.(e-mail:heshan@ieee.org;q.h.wu@liv.ac.uk).
J.R.SaundersiswiththeSchoolofBiologicalSciences,UniversityofLiverpool,Liverpool,L693BX,U.K.(e-mail:jrs@liv.ac.uk).
Colorversionsofoneormoreofthe guresinthispaperareavailableonlineatdoc.guandang.net.
DigitalObjectIdenti er10.1109/TEVC.2009.2011992
I
ofbird ockingor shschooling[4].InterestedreadersaredirectedtoanexcellentsurveyofrecentdevelopmentofPSO[5].
Broadlyspeaking,thetwoSIalgorithmsmentionedabovewereinspiredbysomeaspectsofanimalbehavior,whichisascienti cstudyabouteverythinganimalsdo[6].Fromthisaspect,wecanaddmorealgorithmstothecategoryoftheso-calledanimalbehavior-inspiredalgorithms.In[7],area-restrictedsearchingbehaviorhasinspiredsyntheticpredatorsearch(SPS)algorithmforsolvingcombinatorialoptimizationproblems.
Intra-andintersocietyinteractionsofanimalsocieties,e.g.,humanandsocialinsectsocieties,havebeenusedtodesignastochasticoptimizationalgorithm,calledsocietyandcivi-lizationalgorithm[8].Thisalgorithmwasproposedtosolvesingleobjective-constrainedoptimizationproblemsbasedonaformalsocietyandthecivilizationmodel.
Bacteria,whicharesimplesingle-celledorganisms,havebeenstudiedfordecades.Recently,bacterialforagingbehav-ior,knownasbacterialchemotaxis,hasservedastheinspi-rationoftwodifferentstochasticoptimizationalgorithms.The rstoneisthebacterialchemotaxis(BC)algorithm,whichwasbasedonabacterialchemotaxismodel[9].Thewayinwhichbacteriareacttochemoattractantsinconcentra-tiongradientsareemployedtotackleoptimizationproblems.Anotherbacteria-inspiredoptimizationalgorithmisbacterialforagingalgorithm[10].Itwasinspiredbythechemotactic(foraging)behaviorofE.colibacteria.Onlysmall-scale(upto vedimensions)optimizationproblemsweretackledbythesetwoalgorithms.Thepotentialofbacteria-inspiredalgorithmsremainstobeexploited.
Idealsfromanimalbehaviorhavealsobeenincorporatedtomultiobjectiveevolutionaryalgorithm.In[11],predator–preymodelfromanimalbehaviorhasbeenusedtoapproximatetheshapeofthePareto-optimalsetofmultiobjectiveoptimizationproblems.
Inthispaper,inspiredbyanimalbehavior,especiallyanimalsearching(foraging)behavior,weproposeanoveloptimiza-tionalgorithm,calledgroupsearchoptimizer(GSO),primar-ilyforcontinuousoptimizationproblems.Animalsearchingbehaviormaybedescribedasanactivemovementbywhichananimal ndsorattemptsto ndresourcessuchasfood,mates,oviposition,ornestingsites[12],anditisperhapsthemostimportantkindofbehaviorinwhichananimalengages.Theultimatesuccessofananimal’ssearchingdependson[12]1)thestrategiesitusesinrelationshiptotheavailabilityofresourcesandtheirspatialandtemporaldistributionsin
1051-8215/$26.00©2009IEEE
热门文档