Bertsekas, Dimitri P.; Yu, Huizhen
(2010)
Report LIDS - 2831
We consider the classical nite-state discounted Markovian decision problem, and we introduce a new policy iteration-like algorithm for fi nding the optimal Q-factors. Instead of policy evaluation by solving
a linear system of equations, our algorithm requires (possibly inexact) solution of a nonlinear system of equations, involving estimates of state costs as well as Q-factors. This is Bellman's equation for an optimal
stopping problem that can be solved with simple Q-learning iterations, in the case where a lookup table representation is used; it can also be solved with the Q-learning algorithm of Tsitsiklis and Van Roy [TsV99], in the case where feature-based Q-factor approximations are used. In exact/lookup table representation form, our algorithm admits asynchronous and stochastic iterative implementations, in the spirit of
asynchronous/modi ed policy iteration, with lower overhead and more reliable convergence advantages over existing Q-learning schemes. Furthermore, for large-scale problems, where linear basis function approximations and simulation-based temporal di erence implementations are used, our algorithm resolves e ffectively the inherent difficulties of existing schemes due to inadequate exploration.