TY - GEN

T1 - Selecting computations

T2 - 28th Conference on Uncertainty in Artificial Intelligence, UAI 2012

AU - Hay, Nicholas

AU - Russell, Stuart

AU - Tolpin, David

AU - Shimony, Solomon Eyal

PY - 2012/12/1

Y1 - 2012/12/1

N2 - Sequential decision problems are often approximately solvable by simulating possible future action sequences. Metalevel decision procedures have been developed for selecting which action sequences to simulate, based on estimating the expected improvement in decision quality that would result from any particular simulation; an example is the recent work on using bandit algorithms to control Monte Carlo tree search in the game of Go. In this paper we develop a theoretical basis for metalevel decisions in the statistical framework of Bayesian selection problems, arguing (as others have done) that this is more appropriate than the bandit framework. We derive a number of basic results applicable to Monte Carlo selection problems, including the first finite sampling bounds for optimal policies in certain cases; we also provide a simple counterexample to the intuitive conjecture that an optimal policy will necessarily reach a decision in all cases. We then derive heuristic approximations in both Bayesian and distribution-free settings and demonstrate their superiority to bandit-based heuristics in one-shot decision problems and in Go.

AB - Sequential decision problems are often approximately solvable by simulating possible future action sequences. Metalevel decision procedures have been developed for selecting which action sequences to simulate, based on estimating the expected improvement in decision quality that would result from any particular simulation; an example is the recent work on using bandit algorithms to control Monte Carlo tree search in the game of Go. In this paper we develop a theoretical basis for metalevel decisions in the statistical framework of Bayesian selection problems, arguing (as others have done) that this is more appropriate than the bandit framework. We derive a number of basic results applicable to Monte Carlo selection problems, including the first finite sampling bounds for optimal policies in certain cases; we also provide a simple counterexample to the intuitive conjecture that an optimal policy will necessarily reach a decision in all cases. We then derive heuristic approximations in both Bayesian and distribution-free settings and demonstrate their superiority to bandit-based heuristics in one-shot decision problems and in Go.

UR - http://www.scopus.com/inward/record.url?scp=84886054445&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84886054445

SN - 9780974903989

T3 - Uncertainty in Artificial Intelligence - Proceedings of the 28th Conference, UAI 2012

SP - 346

EP - 355

BT - Uncertainty in Artificial Intelligence - Proceedings of the 28th Conference, UAI 2012

Y2 - 15 August 2012 through 17 August 2012

ER -