Reinforcement Learning with Non-Cumulative Objective

Wei Cui, Wei Yu

In reinforcement learning, the objective is almost always defined as a \emph{cumulative} function over the rewards along the process. However, there are many optimal control and re…