PyTorch 1.x Reinforcement Learning Cookbook
Reinforcementlearning(RL)isabranchofmachinelearningthathasgainedpopularityinrecenttimes.ItallowsyoutotrainAImodelsthatlearnfromtheirownactionsandoptimizetheirbehavior.PyTorchhasalsoemergedasthepreferredtoolfortrainingRLmodelsbecauseofitsefficiencyandeaseofuse.Withthisbook,you'llexploretheimportantRLconceptsandtheimplementationofalgorithmsinPyTorch1.x.Therecipesinthebook,alongwithreal-worldexamples,willhelpyoumastervariousRLtechniques,suchasdynamicprogramming,MonteCarlosimulations,temporaldifference,andQ-learning.You'llalsogaininsightsintoindustry-specificapplicationsofthesetechniques.Laterchapterswillguideyouthroughsolvingproblemssuchasthemulti-armedbanditproblemandthecartpoleproblemusingthemulti-armedbanditalgorithmandfunctionapproximation.You'llalsolearnhowtouseDeepQ-NetworkstocompleteAtarigames,alongwithhowtoeffectivelyimplementpolicygradients.Finally,you'lldiscoverhowRLtechniquesareappliedtoBlackjack,Gridworldenvironments,internetadvertising,andtheFlappyBirdgame.Bytheendofthisbook,you'llhavedevelopedtheskillsyouneedtoimplementpopularRLalgorithmsanduseRLtechniquestosolvereal-worldproblems.
·6.6萬字