内容简介
Geneticprogrammingmaybemorepowerfulthanneuralnetworksandothermachinelearningtechniques,abletosolveproblemsinawiderrangeofdisciplines.Inthisground-breakingbook,JohnKozashowshowthisremarkableparadigmworksandprovidessubstantialempiricalevidencethatsolutionstoagreatvarietyofproblemsfrommanydifferentfieldscanbefoundbygeneticallybreedingpopulationsofcomputerprograms.GeneticProgrammingcontainsagreatmanyworkedexamplesandincludesasamplecomputercodethatwillallowreaderstoruntheirownprograms.Ingettingcomputerstosolveproblemswithoutbeingexplicitlyprogrammed,Kozastressestwopoints:thatseeminglydifferentproblemsfromavarietyoffieldscanbereformulatedasproblemsofprograminduction,andthattherecentlydevelopedgeneticprogrammingparadigmprovidesawaytosearchthespaceofpossiblecomputerprogramsforahighlyfitindividualcomputerprogramtosolvetheproblemsofprograminduction.Goodprogramsarefoundbyevolvingtheminacomputeragainstafitnessmeasureinsteadofbysittingdownandwritingthem.JohnR.KozaisConsultingAssociateProfessorintheComputerScienceDepartmentatStanfordUniversity.