On the Expressivity of Neural Networks for Deep Reinforcement Learning

ICML 2020