Regularised Zero-Variance Control Variates


Dr. Leah South


University of Lancaster


Mon, 15/07/2019 - 12:30pm


RC-4082, The Red Centre, UNSW


Zero-variance control variates (ZV-CV) are a post-processing method to reduce the variance of Monte Carlo estimators of expectations using the derivatives of the log target. Once the derivatives are available, the only additional computational effort is solving a linear regression problem. Significant variance reductions have been achieved with this method in low dimensional examples, but the number of covariates in the regression rapidly increases with the dimension of the target. We propose 
to exploit different types of regularisation to make the method more flexible and feasible, particularly in higher dimensions. Our novel methods retain the unbiasedness property of the estimators. The benefits of regularised ZV-CV for Bayesian inference will be illustrated using several examples, including a 61-dimensional example. This is work joint with Antonietta Mira, Chris Oates and Chris Drovandi.

School Seminar Series: