Actor-Critic

A reinforcement learning architecture with two components: an actor (policy) that selects actions, and a critic (value function) that evaluates those actions.