(computing theory) A model-free reinforcement learning algorithm to learn a policy telling an agent what action to take under what circumstances.