Analysis of Lower Bounds for Simple Policy Iteration
share
Policy iteration is a family of algorithms that are used to find an optimal policy for a given Markov Decision Problem (MDP). Simple Policy iteration (SPI) is a type of policy iteration where the strategy is to change the policy at exactly one improvable state at every step. Melekopoglou and Condon [1990] showed an exponential lower bound on the number of iterations taken by SPI for a 2 action MDP. The results have not been generalized to k-action MDP since. In this paper, we revisit the algorithm and the analysis done by Melekopoglou and Condon. We generalize the previous result and prove a novel exponential lower bound on the number of iterations taken by policy iteration for N-state, k-action MDPs. We construct a family of MDPs and give an index-based switching rule that yields a strong lower bound of O((3+k)2^N/2-3).
READ FULL TEXT