Policy Iteration

An algorithm for solving sequential decision problems (typically Markov decision processes) that alternates between evaluating the value of a current policy and improving the policy greedily with respect to that value until convergence to an optimal policy.