版權(quán)說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請(qǐng)進(jìn)行舉報(bào)或認(rèn)領(lǐng)
文檔簡(jiǎn)介
SHANGHAIJIAOTONGUNIVERSITYProjectTitle:PlayingtheGameofFlappyBirdwithDeepReinforcementLearningGroupNumber:G-07GroupMembers:WangWenqingGaoXiaoningContents1 Introduction (10)endforEveryCstepsreset=:endforendforExperimentsThissectionwilldescribeouralgorithm’sparameterssettingandtheanalysisofexperimentresults.ParametersSettingsREF_Ref484591565\hFigure6illustratesourCNN’slayerssetting.Theneuralnetworkshas3CNNhiddenlayersfollowedby2fullyconnectedhiddenlayers.Table1showthedetailedparametersofeverylayer.HerewejustuseamaxpoolinginthefirstCNNhiddenlayer.Also,weusetheReLUactivationfunctiontoproducetheneuraloutput.FigureSEQFigure\*ARABIC6:ThelayersettingofCNN:thisCNNhas3convolutionallayersfollowedby2fullyconnectedlayers.Asfortraining,weuseAdamoptimizertoupdatetheCNN’sparameters.TableSEQTable\*ARABIC1:ThedetailedlayerssettingofCNNLayerInputFiltersizeStrideNumfiltersActivationOutputconv180×80×48×8432ReLU20×20×32max_pool20×20×322×2210×10×32conv210×10×324×4264ReLU5×5×64conv35×5×643×3164ReLU5×5×64fc45×5×64512ReLU512fc55122Linear2REF_Ref484591593\hTable1listsalltheparametersettingofDQN.Weuseadecayedrangingfrom0.1to0.001tobalanceexplorationandexploitation.What’smore,REF_Ref484591626\hTable2showsthatthebatchstochasticgradientdescentoptimizerisAdamwithbatchsizeof32.Finally,wealsoallocatealargereplaymemory.TableSEQTable\*ARABIC2:ThetrainingparametersofDQNParametersvalueObservesteps100000Exploresteps3000000Initial_epsilon0.1Final_epsilon0.001Replay_memory50000batchsize32learningrate0.000001FPS30optimizationalgorithmAdamResultsAnalysisWetrainourmodelabout4millionepochs.REF_Ref484591669\hFigure7showstheweightsandbiasesofCNN’sfirsthiddenlayer.Theweightsandbiasesfinallycentralizearound0,withlowvariance,whichdirectlystabilizeCNN’soutputQ-valueandreduceprobabilityofrandomaction.ThestabilityofCNN’sparametersleadstoobtainingoptimalpolicy.FigureSEQFigure\*ARABIC7:Left(right)figureisthehistogramofweights(biases)ofCNN’sfirsthiddenlayerREF_Ref484591680\hFigure8isthecostvalueofDQNduringtraining.Thecostfunctionhasaslowdowntrend,closeto0after3.5millionepochs.ItmeansthatDQNhaslearnedthemostcommonstatesubspaceandwillperformoptimalactionwhencomingacrossknownstate.Inaword,DQNhasobtaineditsbestactionpolicy.FigureSEQFigure\*ARABIC8:DQN’scostfunction:theplotshowsthetrainingprogressofDQN.Wetrainedourmodelabout4millionepochs.Whenplayingflappybird,ifthebirdgetsthroughthepipe,wegiveareward1,ifdead,give-1,otherwise0.1.REF_Ref484591694\hFigure9istheaveragereturnedrewardfromenvironment.Thestabiltiyinfinaltrainingstatemeansthattheagentcanautomaticallychoosethebestaction,andtheenvironmentgivesthebestrewardinturns.Weknowthattheagentandenvironmenthasenterintoafriendlyinteraction,guaranteeingthemaximaltotalreward.FigureSEQFigure\*ARABIC9:Theaveragereturnedrewardfromenvironment.Weaveragethereturnedrewardevery1000epochs.FromthisREF_Ref484591711\hFigure10,thepredictedmaxQ-valuefromCNNconvergesandstabilizesinavalueafterabout100000.ItmeansthatCNNcanaccuratelypredictthequalityofactionsinspecificstate,andwecansteadilyperformactionswithmaxQ-value.TheconvergenceofmaxQ-valuesstatesthatCNNhasexploredthestatespacewidelyandgreatlyapproximatedtheenvironmentwell.FigureSEQFigure\*ARABIC10:TheaveragemaxQ-valueobtainedfromCNN’soutput.WeaveragethemaxQ-valueevery1000epochs.REF_Ref484591726\hFigure11illustratestheDQN’sactionstrategy.IfthepredictedmaxQ-valueissohigh,weareconfidentthatwewillgetthroughthegapwhenperformtheactionwithmaxQ-valuelikeA,C.IfthemaxQ-valueisrelativelylow,andweperformtheaction,wemighthitthepipe,likeB.Inthefinalstateoftraining,themaxQ-valueisdramaticallyhigh,meaningthatweareconfidenttogetthroughthegapsifperformingtheactionswithmaxQ-value.FigureSEQFigure\*ARABIC11:TheleftmostplotshowstheCNN’spredictedmaxQ-valuefora100framessegmentofthegameflappybird.ThethreescreenshotscorrespondtotheframeslabeledbyA,B,andCrespectively.ConclusionWesuccessfullyuseDQNtoplayflappybird,whichcanoutperformhumanbeings.DQNcanautomaticallylearnknowledgefromenvironmentjustusingrawimagetoplaygameswithoutpriorknowledge.ThisfeaturegiveDQNthepowertoplayalmostsimplegames.Moreover,theuseofCNNasafunctionapproximationallowDQNtodealwithlargeenvironmentwhichhasalmostinfinitestatespace.Lastbutnotleast,CNNcanalsogreatlyrepresentfeaturespacewithouthandcraftedfeatureextractionreducingthemassivemanualwork.
ReferencesC.ClarkandA.Storkey.Teachingdeepconvolutionalneuralnetworkstoplaygo.arXivpreprintarXiv:1412.3409,2014.1.AlexKrizhevsky,IlyaSutskever,andGeoffHinton.Imagenetclassificationwithdeepconvolutionalneuralnetworks.InAdvancesinNeuralInformationProcessingSystems25,pages1106–1114,2012.GeorgeE.Dahl,DongYu,LiDeng,andAlexAcero.Context-dependentpre-traineddeepneuralnetworksforlarge-vocabularyspeechrecognition.Audio,Speech,andLanguageProcessing,IEEETransactionson,20(1):30–42,2012,1.RichardSuttonandAndrewBarto.ReinforcementLearning:AnIntroduction.MITPress,1998.BrianSallansandGeoffreyE.Hinton.Reinforcementlearningwithfactoredstatesandactions.JournalofMachineLearningResearch,5:1063–1088,2004.ChristopherJCHWatkinsandPeterDayan.Q-learning.Machinelearning,8(3-4):279–292,1992.HamidMaei,CsabaSzepesv′ari,ShalabhBhatnagar,andRichardS.Sutton.Towardoff-policylearningcontrolwithfunctionapproximation.InProceedingsofthe27thI
溫馨提示
- 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請(qǐng)下載最新的WinRAR軟件解壓。
- 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請(qǐng)聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
- 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會(huì)有圖紙預(yù)覽,若沒有圖紙預(yù)覽就沒有圖紙。
- 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
- 5. 人人文庫網(wǎng)僅提供信息存儲(chǔ)空間,僅對(duì)用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對(duì)用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對(duì)任何下載內(nèi)容負(fù)責(zé)。
- 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請(qǐng)與我們聯(lián)系,我們立即糾正。
- 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶因使用這些下載資源對(duì)自己和他人造成任何形式的傷害或損失。
最新文檔
- 2024年大數(shù)據(jù)中心機(jī)房建設(shè)工程合同
- 2024年廣告媒體采購發(fā)布合同
- 2024城市公園環(huán)境衛(wèi)生承包協(xié)議
- 2024年工程貸款協(xié)議模板助力項(xiàng)目發(fā)展
- 2024年危險(xiǎn)品運(yùn)輸合同范本
- 2024年應(yīng)急通信系統(tǒng)設(shè)備采購及安裝合同
- 2024年工程質(zhì)量檢測(cè)居間合同
- 2024年雙方共同投資建立教育咨詢公司的合同
- 挑食偏食課件教學(xué)課件
- 2024婚姻關(guān)系解除后債務(wù)清償合同
- 河北省石家莊市長(zhǎng)安區(qū)2023-2024學(xué)年五年級(jí)上學(xué)期期中英語試卷
- 品牌經(jīng)理招聘筆試題及解答(某大型國(guó)企)2025年
- 多能互補(bǔ)規(guī)劃
- 珍愛生命主題班會(huì)
- 《網(wǎng)絡(luò)數(shù)據(jù)安全管理?xiàng)l例》課件
- 消除“艾梅乙”醫(yī)療歧視-從我做起
- 2024年時(shí)事政治試題(帶答案)
- 第7課《回憶我的母親》課件-2024-2025學(xué)年統(tǒng)編版語文八年級(jí)上冊(cè)
- 八年級(jí)歷史上冊(cè)(部編版)第六單元中華民族的抗日戰(zhàn)爭(zhēng)(大單元教學(xué)設(shè)計(jì))
- 公司研發(fā)項(xiàng)目審核管理制度
- 《詩意的色彩》課件 2024-2025學(xué)年人美版(2024)初中美術(shù)七年級(jí)上冊(cè)
評(píng)論
0/150
提交評(píng)論