参会心得

Xi’anUniversityofTechnology参会心得学生：徐庆征2009-6-19Xi’anUniversityofTechnology提纲会议基本信息精彩报告几点体会Xi’anUniversityofTechnology会议基本信息会议名称：2009WorldSummitonGeneticandEvolutionaryComputation会议地点：主办方：共同组织者：Xi’anUniversityofTechnology论文数量Submissions:372Fullpaper:153Postpaper:126Xi’anUniversityofTechnology会议规模Keynote:1Tutorial:14Intro:6Special:3Advanced:5Session:28ScheduledPaperSession:27SpecialPosterSession:1Xi’anUniversityofTechnology投稿过程提交日期：2008-12-5录用日期：2009-2-24提交正式稿日期：2009-3-3注册日期：2009-3-11会议日期：2009-6-12至2009-6-14Xi’anUniversityofTechnology会议经历开幕式KeynoteTutorial:6PaperPresentation:16做报告Xi’anUniversityofTechnology精彩报告1Keynote:Practical&PhilosophicalReflectionsonaLifeinGeneticAlgorithmsTutorial:Fast,EffectiveGA’sforLarge,HardProblemsAuthor:DavidE.GoldbergXi’anUniversityofTechnologyLessonsLearningtoAskLearningtoLabelLearningtoDecomposeLearningtoModelXi’anUniversityofTechnologyLesson1:LearningtoAskIn1984hadmanyquestionsabouthowGAswork,whentheyfail?Wasn’texperiencedinaskinggoodframingquestions.Keyproblem:UsingGAstosolveengineeringproblems,butGAsweren’tengineeredwell.Socrates(470-399BCE)Xi’anUniversityofTechnologyWhat’saGoodQuestion?Socratesaskedvarietyofquestions.Whatistruth?Whatiscourage?Moreoftenthecritic.Rarelygaveanswers.Increativeenterprises,manygoodquestionsareframingquestions:–Getatheartoftheissue.–Helpdefinetheproblemorelicitdefinition.–Sometimescauseproblemtoberepresentedinnovelwayorfromunusualorcreativeperspective.Fundamentalimportanceofdialectic.Creativeprocessofaskingandansweringquestions.Xi’anUniversityofTechnologyTermsgatherthoughtsunderconsistentrubrics.Canbepartoflargertaxonomy.Definesattentionareas.Canhaveinfluenceonhowothersthink.Catchyorstickytermspropagatevirally.Lesson2:LearningtoLabelTermsReallyDoMatterXi’anUniversityofTechnologyLesson3:LearningtoDecomposeDecember17,1903:TheMostFamousMomentinAviationHistoryXi’anUniversityofTechnologyTheWrightBrothers’SecretFunctionaldecomposition.Threesubproblems:Stability:wing-warpingpluselevatorin1899glidermodel.1902gliderhadthree-axisactivecontrol.LiftandDrag:wingshapeimprovedonLilenthal’sthroughairtunnelexperiments.Propulsion:rotarywingwithforwardliftisapropeller.Xi’anUniversityofTechnologyEffectiveTheoryinGADesignManyGAsdon’tscale&muchGAtheoryinapplicable.Needdesigntheorythatworks:Understandbuildingblocks(BBs),notionsorsubideas.EnsureBBsupply.EnsureBBgrowth.ControlBBspeed.EnsuregoodBBdecisions.EnsuregoodBBmixing(exchange).KnowBBchallengers.Canusetheorytodesignscalable&efficientGAs.Xi’anUniversityofTechnologyLesson4:LearningtoModelError,εCostofModeling,CEngineer/InventorScientist/MathematicianAModelofModelsXi’anUniversityofTechnologyALifeinGeneticAlgorithmsEventsBumpedintoGAsbyaccident.Joinedfieldattimeofgrowth.Fluidstrainingasdisciplinarygroundingincomplexity.WroteabookIwastoldnottowrite.Becamephilosophicalinaaction-orientedfield.Tookonreformeffortnotadmiredbypeers.Lessons?Importantthingscanberandom.Opportunityisknocking?Willyouanswerthedoor?Beingappropriatelydifferentcanbebeneficial.Authorityfiguresarenotnecessarilyrightorwise.Exploringtheunexploredcanyieldinterestinginsights.Sometimesimportantjobsarenotvaluedbyothers.Xi’anUniversityofTechnology精彩报告2Tutorial:IntroductiontoGeneticAlgorithmsTutorial:IntroductiontoGeneticalgorithmTheoryandPracticeErikGoodmanDarrellWhitleyXi’anUniversityofTechnology精彩报告3Tutorial:AUnifiedFramworkforEvolutionaryComputationAuthor:KenDeJongXi’anUniversityofTechnologyHistoricalRootsEvolutionStrategies(ESs)－developedbyRechenberg,Schwefel,etc.in1960sEvolutionaryProgramming(EP)－developedbyFogelin1960sGeneticAlgorithms(GAs)－developedbyHollandin1960sXi’anUniversityofTechnologyPresentStatuswidevarietyofevolutionaryalgorithms(EAs)widevarietyofapplications–optimization–search–learning,adaptationwell-developedanalysis–theoretical–experimentalXi’anUniversityofTechnologyViewpointDevelopageneralframeworkthat:–Helpsonecompareandcontrastapproaches.–Encouragescrossbreeding.–Facilitatesintelligentdesignchoices.Xi’anUniversityofTechnologyAnEATemplate1.Randomlygenerateaninitialpopulation.2.Dountilsomestoppingcriteriaismet:Selectindividualstobeparents(biasedbyfitness).Produceoffspring.Selectindividualstodie(biasedbyfitness).EndDo.3.Returnaresult.Basicelements:–apopulationof“individuals”–anotionof“fitness”–abirth/deathcyclebiasedbyfitness–anotionof“inheritance”Xi’anUniversityofTechnologyNewDevelopmentsandDirectionsExploitingparallelism:–coarselygrainednetworkmodels–finelygraineddiffusionmodelsCo-evolutionarymodels:–competitiveco-evolutionExploitingMorphogenesis:–sophisticatedgenotype--phenotypemappings–evolveplansforbuildingcomplexobjectsratherthantheobjectsthemselves.Xi’anUniversityofTechnologyNewDevelopmentsandDirectionsSelf-adaptiveEAs:–dynamicallyadapttoproblemcharacteristics:–goal:robust“blackbox”optimizerHybridSystems:–combineEAswithothertechniquesTime-varyingenvironments:–fitnesslandscapechangesduringevolution–goal:adaptation,tracking–standardoptimization-orien