Bounds for Markov Decision Processes
نویسندگان
چکیده
We consider the problem of producing lower bounds on the optimal cost-to-go function of a Markov decision problem. We present two approaches to this problem: one based on the methodology of approximate linear programming (ALP) and another based on the so-called martingale duality approach. We show that these two approaches are intimately connected. Exploring this connection leads us to the problem of finding ‘optimal’ martingale penalties within the martingale duality approach which we dub the pathwise optimization (PO) problem. We show interesting cases where the PO problem admits a tractable solution and establish that these solutions produce tighter approximations than the ALP approach.
منابع مشابه
Accelerated decomposition techniques for large discounted Markov decision processes
Many hierarchical techniques to solve large Markov decision processes (MDPs) are based on the partition of the state space into strongly connected components (SCCs) that can be classified into some levels. In each level, smaller problems named restricted MDPs are solved, and then these partial solutions are combined to obtain the global solution. In this paper, we first propose a novel algorith...
متن کاملOnline Regret Bounds for Markov Decision Processes with Deterministic Transitions
We consider an upper confidence bound algorithm for Markov decision processes (MDPs) with deterministic transitions. For this algorithm we derive upper bounds on the online regret (with respect to an (ε-)optimal policy) that are logarithmic in the number of steps taken. These bounds also match known asymptotic bounds for the general MDP setting. We also present corresponding lower bounds. As an...
متن کاملExponential Lower Bounds for Policy Iteration
We study policy iteration for infinite-horizon Markov decision processes. It has recently been shown policy iteration style algorithms have exponential lower bounds in a two player game setting. We extend these lower bounds to Markov decision processes with the total reward and average-reward optimality criteria.
متن کاملPseudometrics for State Aggregation in Average Reward Markov Decision Processes
We consider how state similarity in average reward Markov decision processes (MDPs) may be described by pseudometrics. Introducing the notion of adequate pseudometrics which are well adapted to the structure of the MDP, we show how these may be used for state aggregation. Upper bounds on the loss that may be caused by working on the aggregated instead of the original MDP are given and compared ...
متن کاملLoss Bounds for Uncertain Transition Probabilities in Markov Decision Processes
We analyze losses resulting from uncertain transition probabilities in Markov decision processes with bounded nonnegative rewards. We assume that policies are pre-computed using exact dynamic programming with the estimated transition probabilities, but the system evolves according to different, true transition probabilities. Our approach analyzes the growth of errors incurred by stepping backwa...
متن کاملAdaptive aggregation for reinforcement learning in average reward Markov decision processes
We present an algorithm which aggregates online when learning to behave optimally in an average reward Markov decision process. The algorithm is based on the reinforcement learning algorithm UCRL and uses confidence intervals for aggregating the state space. We derive bounds on the regret our algorithm suffers with respect to an optimal policy. These bounds are only slightly worse than the orig...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2011