In Actor-Critic, we have both policy and value functions (or a combination of value-based and policy-based). This method is the best of both worlds:
