第1題WhichcountrydidthefamousEnigmaciphersystem,crackedbyTuring,belongto?()AUnitedKingdomBUnitedStatesCGermanyDJapan第2題In1937,TuringfirstproposedtheconceptoftheTuringmachineinapaper.Whatwasthenameofthispaper?()A"EquivalenceofLeftandRightPeriodicity"B"OnComputableNumbers,withanApplicationtotheEntscheidungsproblem"C"ComputabilityandDefinability"D"OntheGaussianErrorFunction"第3題Whichofthefollowingstatementsisincorrect?()ATheTuringtestconfusesintelligencewiththehumancondition.BThemechanismofmachineintelligencemustbethesameashumanintelligence.CMachineintelligencecancompletelysurpasshumanintelligenceincertaindomains.DMachineintelligencecanhavethecreativityofhumanintelligence.第4題Whichcountry'sgovernmentreportusedtheterm"machineintelligence"?()AChinaBUnitedKingdomCGermanyDUnitedStates第5題Whichofthefollowingisacorebeliefofthesymbolistschoolofthought?()AIntelligenceisachievedbysimulatingconnectionsbetweenhumanbrainneurons.BArtificialintelligenceoriginatesfrommathematicallogic,simulatinghumanthoughtthroughsymbolicsystemsandrules.CIntelligencedependsonperceptionandaction,learningandadaptingthroughinteractionwiththeenvironment.DIntelligenceisachievedbyconstructingperception-actioncontrolsystems.第6題Whichofthefollowingisarepresentativeachievementofthesymbolistschoolofthought?()ADeeplearningBKnowledgegraphCAutonomousnavigationrobotsDExpertsystems第7題Whichofthefollowingtechnologiesbelongstotheconnectionistschool'sresearchmethods?()ADecisiontreesBDeeplearningCExpertsystemsDPerception-actioncontrolsystems第8題Whichofthefollowingtechnologiesisnotafrontiertechnologyintegratedintointelligentlogisticsrobotsforindoortransportationanddistribution?()AArtificialintelligenceBAutonomousdrivingCVirtualrealityD5G第9題Whichapplicationscenariobelongstoartificialintelligenceinthemedicalfield?()AIntelligentcustomerserviceBPulmonarydiseasediagnosisCSmarthomeDAutonomousdriving第10題WhichprinciplewasproposedbyChinaintheethicalgovernanceofartificialintelligence?()ATechnologyfirstBEconomyfirstCEthicsfirstDPolicyfirst第11題ThedifferencebetweenstrongAIandweakAIisthatstrongAIhasself-awareness.()第12題TheTuringtestistheonlystandardtodeterminewhetheramachinehasintelligence.()第13題AI'sintelligentrecommendationsystemsintheentertainmentfieldcanhelpusersdiscovernewmusicormovies.()第14題ThedevelopmentofAIwillcausehumanstolosecontroloftheworld.()第15題Theemergenceofself-drivingcarswillcompletelysolvetrafficcongestionproblems.()第16題AIcanautonomouslymakemoraljudgments.()第17題ThedevelopmentofAIwillleadtothelossofhumancreativity.()第18題AIcanfullyunderstandhumanemotions.()第19題ThedevelopmentofAIwillleadtothelossofsocialconnectionsforhumans.()第20題ThedevelopmentofAIwillcausehumanstolosefreewill.()第1題Whatisthedifferencebetweensupervisedlearningandunsupervisedlearning?()ADatasizeBPresenceofateachersignalCApplicationdomainDAlgorithmused第2題Thehistoryofmachinelearningcanbetracedbacktowhichcentury?()A17thcenturyB18thcenturyC19thcenturyD20thcentury第3題Whatisthepurposeofalinearregressionmodel?()ATofindoutliersinthedatasetBTominimizetheerrorbetweenthepredictedandactualvaluesCToincreasethecomplexityofthemodelDToreducethedimensionalityofthedata第4題WhatdoestheR-squaredcoefficientinregressionmeasure?()AThecomplexityofthemodelBThepredictiveabilityofthemodelCThegeneralizationabilityofthemodelDThedegreeofoverfittingofthemodel第5題Whatisthemainuseoftheleastsquaresmethod?()APredictfuturedataBFindthebestfitlineCClassifydataDClusterdata第6題WhichofthefollowingisNOTacharacteristicofthedecisiontreealgorithm?()AEasytounderstandandinterpretBCanhandlenumericalandcategoricaldataCRequiresalotofcomputationalresourcesDCangeneraterules第7題GiventheeventsAandB,theprobability

P(B|A)

canbederivedusingBayes'formulaas(P(B|A)=P(A|B)

*

M/P(),whatisM=()AP(A)BP(AB)CP(A')DP(B')第8題Intheartificialneuronmodel(M-Pmodel),whatarethedendritesandtheirsynapsessimplifiedandabstractedas?()AAsetofinputvariablesBAnoutputvariableCAnactivationfunctionDAbias第9題Whichofthefollowingstatementsaboutensemblelearningisincorrect?()AEnsemblelearningalwaysimprovestheperformanceofindividuallearnersBInBagging,individuallearnersareindependentofeachotherCBoostingisamethodthatemphasizeslearningfromincorrectsamplesDInBoosting,individuallearnershavestrongdependencies第10題IntheK-meansclusteringalgorithm,theinitialchoiceofclustercentersaffectsthefinalresult.Whichmethodcanhelpchoosebetterinitialclustercenters?()ARandomselectionBK-means++CSelectionbasedondataorderDManualspecification第11題Supervisedlearningalgorithmsinmachinelearningcanonlybeusedforclassificationtasks.()第12題Thebasicprocessofmachinelearningincludesdatacollectionandpreparation,featureengineering,modelselectionandtraining,modelevaluation,optimization,anddeploymentmonitoring.()第13題Alinearregressionmodelcanonlybeusedtomodelrelationshipsbetweentwovariables.()第14題Linearregressionmodelsaretypicallyusedtosolveclassificationproblems.()第15題Boththeleastsquaresmethodandthegradientdescentmethodareusedtosolveunconstrainedoptimizationproblems.()第16題Inadecisiontree,leafnodesaredecisionnodes.()第17題Bayesianriskreflectsthebestperformanceaclassifiercanachieve,representingthetheoreticalupperlimitofmodelaccuracyachievedthroughmachinelearning.()第18題Theoutputvariableofanartificialneuronisdeterminedbytheactivationfunction.()第19題TheoutputoftheBoostingalgorithmistheweightedsumofallbaselearners.()第20題ThevalueofkintheK-meansalgorithmisdeterminedautomaticallybythealgorithm.()Chapter3:DeepLearning-ChapterExecises(25points)第1題Whatisthemostcommonlyusedprogramminglanguagefordeeplearning?()APythonBJavaCC++DJavaScript第2題Whatisthemainfunctionofaconvolutionallayer?()AReducedatadimensionalityBExtractimagefeaturesCConnectinputsandoutputsDClassifydata第3題WhatisthemainfeatureofanRNN?()ACannotprocesssequentialdataBOnlyhandlesfixed-lengthdataCCanprocesstimeseriesinformationDLacksmemorycapability第4題InaGAN,whatistheprimarytaskofthediscriminator?()AGeneraterealdataBDistinguishbetweenrealandfakedataCOptimizethegeneratorDPredictdatacategories第5題OneofthecorecomponentsoftheTransformerismulti-headattention.Whatisitsfunction?()ACapturerelationshipsbetweendifferentpartsofasequenceBIncreasethemodel'strainingspeedCEnhancethedepthofthemodelDGeneratesequentialdata第6題WhatarethetwomaincomponentsofaGenerativeAdversarialNetwork(GAN)?AInputlayerandoutputlayerBConvolutionallayerandpoolinglayerCGeneratoranddiscriminatorDEncoderanddecoder第7題Whatistheprimarytaskofdiffusionmodels?()AImageclassificationBImagegenerationCObjectdetectionDDatacompression第8題Whichoptimizationalgorithmismostcommonlyusedindeeplearning?()AGradientDescentBRandomForestCK-meansClusteringDLinearRegression第9題Whatisthemainpurposeofincreasingthenumberoflayersinadeepneuralnetwork(DNN)?

AReducecomputationalcomplexityBReducetheriskofoverfittingCIncreasethemodel'srepresentationabilityDDecreasetrainingtime第10題Inwhichneuralnetworkdoesthevanishinggradientproblemtypicallyoccur?

AConvolutionalNeuralNetworkBFullyConnectedNeuralNetworkCGenerativeAdversarialNetworkDDeepNeuralNetwork第11題Deeplearningisasubfieldofmachinelearning.()第12題TheconvolutionallayerisacorecomponentofConvolutionalNeuralNetworks().()第13題StandardRNNshandlelongsequencedataverywell.()第14題InaGAN,thetaskofthegeneratoristodistinguishbetweenrealandfakedata.()第15題TheTransformernetworkmainlyreliesonself-attentionmechanismstoprocesssequentialdata.()第16題Diffusionmodelsgeneratedatabygraduallyaddingnoiseandlearningthedenoisingprocess.()第17題ReLUisacommonlyusedactivationfunction.()第18題Dropoutisatechniqueusedtopreventoverfittingindeeplearning.()第19題Alearningratethatistoolargemaycausethemodeltofailtoconverge.()第20題Onceadeeplearningmodelistrained,itnolongerrequiresvalidation.()Chapter4:AIGenerationContent-ChapterExecises(25points)第1題WhatisthemainadvantageoftheTransformerarchitectureinAIGC?()AHandlingshorttextBHandlinglong-rangedependencies,globalinformationcapture,andcomputationalefficiencyCOnlyprocessingimagedataDRequiringlargeamountsofmanuallylabeleddata第2題IntowhatthreecategoriescanAIGCtextgenerationtasksbedivided?()AUnimodal,bimodal,multimodalBUnimodal,cross-modal,hybrid-modalCUnimodal,multimodal,cross-modalDText,image,video第3題Whatisthecoreofpromptlearning?()ASupervisedlearningBDesigning"prompts"CAutomaticfeatureextractionDPretrainingandfine-tuning第4題Inwhichlearningscenariodoesautomaticdiscretepromptconstructionperformexceptionallywell?()ALarge-samplelearningBFew-shotlearningCUnsupervisedlearningDSupervisedlearning第5題OnwhichtwostagesdoestheICLperformanceoflargelanguagemodelsmainlydepend?()ATrainingstageandinferencestageBPretrainingstageandfine-tuningstageCInferencestageandpost-processingstageDTrainingstageandpost-processingstage第6題InICL,whatdoes"context"referto?()AThemodel’sparametersBExamplesofspecifictasksorquestionsprovidedtothemodelCThemodel’strainingdataDTheinput-outputpairofthemodel第7題Whatmechanismdoesreinforcementlearningusetotrainanagenttomakeoptimaldecisions?()APunishmentmechanismBRewardandpunishmentmechanismCRandomselectionmechanismDGeneticalgorithm第8題Whatisoneofthemainchallengesfacedbyreinforcementlearningfromhumanfeedbackinreal-worldapplications?()AScalabilityBTechnicalimplementationdifficultyCLackofsufficientdataDComputationalresourcelimitations第9題Whichofthefollowingtasksbelongstoamultimodallargemodel'sgenerationtask?()AAudio-to-imagegenerationBText-to-imagegenerationCImageclassificationDSpeechrecognition第10題WhatisthedefinitionofanAIAgent?()AAsoftwarethatcanonlyperformspecifictasksBAnintelligentsystemthatcanautonomouslyperceivetheenvironment,understandtaskrequirements,makedecisions,andperformactionsCAsimpleautomationtoolDAsystemthatrequirescontinuoushumaninterventiontofunction第11題AIGCtechnologycangeneratecon