P. Poupart, C. Boutilier, R. Patrascu, and D. Schuurmans. Piecewise linear value function approximation for factored MDPs. In Proceedings of the Eighteenth National Conference on AI, pages 292--299, 2002.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....By searching within a restricted set of structured compressions and by exploiting DBN structure it is possible to efficiently solve the optimization program in Table 1. The question of factor selection remains: on what factors should F be defined A version of this question has been tackled in [12, 14] in the context of selecting a basis to approximately solve MDPs. The techniques proposed in those papers could be adapted to our optimization program. An alternative method for structuring the computation of F involves additive separability. Let X j (j m) be subsets of variables, and j (X j ....

P. Poupart, C. Boutilier, R. Patrascu, and D. Schuurmans. Piecewise linear value function approximation for factored MDPs. AAAI-02, pp.292--299, Edmonton, 2002.

....be thought of as relying on domain specific properties. We then describe a general framework for the incremental construction of a suitable basis for linear approximation of a factored MDP. This approach relies on no special domain properties, and can be instantiated in a number of concrete ways [14] . We focus in this paper on a particular instantiation of our framework that allows for the construction of a piecewise linear (PWL) combination of basis functions. We argue that this model is especially suited to the solution of WCMDPs, a fact supported by our empirical results. We begin in ....

....L 1 error, so it cannot strictly be viewed as minimizing L 1 error. L# error can be tackled directly using algorithms like policy and value iteration [8] but at higher computational cost. The difficulties associated with minimizing different error metrics in the LP context are discussed in [14] . 3 Basis Function Selection While linear approximations scale well, determining apriori the solution quality one can obtain using a given basis set is difficult. Ideally, V # would be an element of the subspace spanned by , in which case an exact solution could be found. If this is not ....

[Article contains additional citation context not shown here]

P. Poupart, C. Boutilier, R. Patrascu, and D. Schuurmans. Piecewise linear value function approximation for factored MDPs. In Proc. Eighteenth National Conf. on AI, Edmonton, 2002. to appear.

No context found.

P. Poupart, C. Boutilier, R. Patrascu, and D. Schuurmans. Piecewise linear value function approximation for factored MDPs. In Proceedings of the Eighteenth National Conference on AI, pages 292--299, 2002.

No context found.

Poupart, P., Boutilier, C., Patrascu, R., & Schuurmans, D. (2002). Piecewise linear value function approximation for factored mdps. In Eighteenth national conference on Artificial intelligence, pp. 292--299. American Association for Artificial Intelligence.

No context found.

Poupart, P.; Boutilier, C.; Patrascu, R.; and Schuurmans, D. 2002. Piecewise linear value function approximation for factored MDPs. In Proceedings of the Eighteenth National Conference on AI, 292--299.

No context found.

P. Poupart, C. Boutilier, R. Patrascu, and D. Schuurmans. Piecewise linear value function approximation for factored MDPs. In Proceedings of the Eighteenth National Conference on AI, pages 292--299, 2002.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC