Unifying Task Specification in Reinforcement Learning

Martha White

Abstract: Reinforcement learning tasks are typically specified as Markov decision processes. This formalism has been highly successful, though specifications often couple the dynamics of the environment and ...