Policy Iteration
A dynamic programming algorithm for solving the optimal policy of an MDP. It alternates between policy evaluation and policy improvement, and is guaranteed to converge to the optimal policy in finite MDPs.
Timeline (last 90 days)
Policy iteration alternates between policy evaluation and policy improvement, and in a finite MDP it necessarily converges to the optimal policy within a finite number of steps.
Policy iteration and value function estimation are two fundamental solver frameworks in reinforcement learning.
All Facts (2)
Policy iteration alternates between policy evaluation and policy improvement, and in a finite MDP it necessarily converges to the optimal policy within a finite number of steps.
50%UnverifiedPolicy iteration and value function estimation are two fundamental solver frameworks in reinforcement learning.
50%