Σεμινάριο: "Variance reduction for MCMC estimators"
ΚΥΚΛΟΣ ΣΕΜΙΝΑΡΙΩΝ ΣΤΑΤΙΣΤΙΚΗΣ ΔΕΚΕΜΒΡΙΟΣ 2021
Ομιλητής: Angelos Alexopoulos (Research Associate, MRC Biostatistics Unit, University of Cambridge, UK)
Variance reduction for MCMC estimators
We provide a general methodology to construct control variates for any discrete time random walk Metropolis and Metropolis-adjusted Langevin algorithm. By adopting the proposed methods we achieve, in a post-processing manner and with a negligible additional computational cost, impressive variance reduction when compared to the standard MCMC ergodic averages. Our proposed estimators are based on an approximate solution of the Poisson equation for multivariate Gaussian target densities of any dimension.
This is joint work with P. Dellaportas (UCL and AUEB) and M. Titsias (DeepMind)