无人驾驶DeepReforcementLearning.pdf

PhilosophicalMotivationforReinforcementLearningTakeawayfromSupervisedLearning:Neuralnetworksaregreatatmemorizationandnot(yet)greatatreasoning.HopeforReinforcementLearning:Brute-forcepropagationofoutcomestoknowledgeaboutstatesandactions.Thisisakindofbrute-force“reasoning”.