TD-Regularized Actor-Critic Methods

Simone Parisi, Voot Tangkaratt, Jan Peters +1

Actor-critic methods can achieve incredible performance on difficult reinforcement learning problems, but they are also prone to instability. This is partly due to the interaction…