非Matlab强化学习代码及资源说明.pdfVIP

ton/book/code/code.html

Codefor:

ReinforcementLearning:AnIntroduction

byRichardS.SuttonandAndrewG.Barto

Belowarelinkstoavarietyofsoftwarerelatedtoexampandexercises

inthebook,organizedbychapters(somefiappearinmultipleces).

SeeparticularlytheMountainCarcode.Mostoftherestofthecodeis

writteninCommonLispandrequiresutilityroutinesavailablehere.For

thegraphics,youwillneedthethepackagesforGandinsomecasesmy

graphingtool.Evenifyoucannotrunthiscode,itstillmayclarify

someofthedetailsoftheexperiments.However,thereisnoguarantee

thattheexampinthebookwererunusingexactlythesoftwaregiven.

Thiscodealsohasnotbeenextensivelytestedordocumentedandisbeing

madeavailableasis.Ifyouhavecorrections,extensions,additions

orimprovementsofanykind,pleasesendthemtomeatrich@

forinclusionhere.

Matlabcodefornearlyalltheexampandexcercisesinthebookhas

beencontributedbyJohnWeatherwax.ThanksJohn!

•Chapter1:Introduction

oTic-Tac-ToeExample(Lisp).InC.

•Chapter2:EvaluativeFeedback

o10-armedTestbedExample,Figure2.1(Lisp)

oTestbedwithSoftmaxActionSelection,Exercise2.2(Lisp)

oBanditsAandB,Figure2.3(Lisp)

oTestbedwithConstantAlpha,cf.Exercise2.7(Lisp)

oOptimisticInitialValuesExample,Figure2.4(Lisp)

oCodePertainingtoReinforcementComparison:File1,File2,

File3(Lisp)

oPursuitMethodsExample,Figure2.6(Lisp)

•Chapter3:TheReinforcementLearningProblem

oPole-BalancingExample,Figure3.2(C)

oGridworldExample3.8,Cod