Approximate MCMC in Theory and Practice

Friday, December 1, 2017 - 3:30pm

James Johndrow, Stanford University


Rapid growth in the number of samples in typical datasets and the number of parameters in statistical models poses significant computational challenges. A popular strategy for reducing the computational cost of Markov chain Monte Carlo (MCMC) is to replace the exact Markov kernel with an approximation that is less costly to simulate. We give a number of results on the convergence properties of these approximating Markov chains and the performance of time-averages in approximating expectations of functions with respect to the target measure. The talk is structured around several canonical examples that both motivate the results and illustrate the power and limitations of this approach to scaling up MCMC.

Seminars generally take place in 116 Old Chemistry Building on Fridays from 3:30 - 4:30 pm. For additional information contact: or phone 919-684-8029. Sorry, but we do not have reprints available. Please feel free to contact the authors by email for follow-up information, articles, etc. Reception following seminar in 211 Old Chemistry

Old Chemistry 116

Location Info