Dynamic Reinforcement Learning for Actors

Katsunari Shibata

Dynamic Reinforcement Learning (Dynamic RL), proposed in this paper, directly controls system dynamics, instead of the actor (action-generating neural network) outputs at each mome…