Multilevel Monte Carlo estimation of expected information gains
share
In this paper we develop an efficient Monte Carlo algorithm for estimating the expected information gain that measures how much the information entropy about uncertain quantity of interest θ is reduced on average by collecting relevant data Y. The expected information gain is expressed as a nested expectation, with an outer expectation with respect to Y and an inner expectation with respect to θ. The standard, nested Monte Carlo method requires a total computational cost of O(ε^-3) to achieve a root-mean-square accuracy of ε. In this paper we reduce this to optimal O(ε^-2) by applying a multilevel Monte Carlo (MLMC) method. More precisely, we introduce an antithetic MLMC estimator for the expected information gain and provide a sufficient condition on the data model under which the antithetic property of the MLMC estimator is well exploited such that optimal complexity of O(ε^-2) is achieved. Furthermore, we discuss how to incorporate importance sampling techniques within the MLMC estimator to avoid so-called arithmetic underflow. Numerical experiments show the considerable computational savings compared to the nested Monte Carlo method for a simple test case and a more realistic pharmacokinetic model.
READ FULL TEXT
