Title: Recent developments in data subsampling for large-scale Bayesian inference
Authors: Mattias Villani - Stockholm University (Sweden) [presenting]
Matias Quiroz - University of Technology Sydney (Australia)
Robert Kohn - University of New South Wales (Australia)
Minh-Ngoc Tran - University of Sydney (Australia)
Doan Khue Dung Dang - University of New South Wales (Australia)
Abstract: Hamiltonian Monte Carlo (HMC) is an increasingly popular simulation algorithm for Bayesian inference which has proven to be especially suitable in high-dimensional problems. A drawback of HMC is that it requires a large number of evaluations of the posterior and its gradient, which can be computationally costly, particularly in problems with large datasets. Results on accelerating HMC by data subsampling and how to optimally tune the algorithm are presented. Extensions to dependent data are also discussed.