A greedy policy is a reinforcement learning strategy that always selects the action with the highest estimated expected return. It favors exploitation of current knowledge and does not deliberately explore alternative actions.
In reinforcement learning, a policy that always chooses the action with the highest expected return.