Policy iteration
= Policy iteration
Policy iteration alternates evaluation of a fixed decision policy with improvement of its actions using the resulting <value function>. A stable policy satisfies the <Bellman equation> when each chosen action attains the required optimum.