Deep Reinforcement Learning Hands-On
Recentdevelopmentsinreinforcementlearning(RL),combinedwithdeeplearning(DL),haveseenunprecedentedprogressmadetowardstrainingagentstosolvecomplexproblemsinahuman-likeway.Google’suseofalgorithmstoplayanddefeatthewell-knownAtariarcadegameshaspropelledthefieldtoprominence,andresearchersaregeneratingnewideasatarapidpace.DeepReinforcementLearningHands-OnisacomprehensiveguidetotheverylatestDLtoolsandtheirlimitations.YouwillevaluatemethodsincludingCross-entropyandpolicygradients,beforeapplyingthemtoreal-worldenvironments.TakeonboththeAtarisetofvirtualgamesandfamilyfavoritessuchasConnect4.ThebookprovidesanintroductiontothebasicsofRL,givingyoutheknow-howtocodeintelligentlearningagentstotakeonaformidablearrayofpracticaltasks.DiscoverhowtoimplementQ-learningon‘gridworld’environments,teachyouragenttobuyandtradestocks,andfindouthownaturallanguagemodelsaredrivingtheboominchatbots.
·13.4萬字