An evolutionaryapproachwithdiversityguaranteeandwell-informed grouping recombinationforgraphcoloring

We presentadiversity-orientedhybridevolutionaryapproachforthegraphcoloringproblem.This approachisbasedonbothgenerallyapplicablestrategiesandspecificallytailoredtechniques.Particular attentionispaidtoensuringpopulationdiversitybycarefullycontrollingspacingamongindividuals. Using adistancemeasurebetweenpotentialsolutions,thegeneralpopulationmanagementstrategy decides whetheranoffspringshouldbeacceptedinthepopulation,whichindividualneedstobe replacedandwhenmutationisapplied.Furthermore,weintroduceaspecialgrouping-basedmulti- parent crossoveroperatorwhichreliesonseveralrelevantfeaturestoidentifymeaningfulbuilding blocks foroffspringconstruction.Theproposedapproachcanbegenerallycharacterizedas‘‘well- informed’’,inthesensethatthedesignofeachcomponentisbasedonthemostpertinentinformation which isidentifiedbybothexperimentalobservationandcarefulanalysisofthegivenproblem.The resultingalgorithmprovestobehighlycompetitivewhenitisappliedonthewholesetoftheDIMACS benchmarkgraphs.