Accelerating MCMC via Parallel Predictive Prefetching

Angelino, E., Kohler, E., Waterland, A., Seltzer, M., & Adams, R. P. (2014). Accelerating MCMC via Parallel Predictive Prefetching. Proceedings of the 30th Conference on Uncertainty in Artificial Intelligence (UAI).

Parallel predictive prefetching is a new frame- work for accelerating a large class of widely-used Markov chain Monte Carlo (MCMC) algorithms. It speculatively evaluates many potential steps of an MCMC chain in parallel while exploiting fast, iterative approximations to the tar- get density. This can accelerate sampling from target distributions in Bayesian inference problems. Our approach takes advantage of whatever parallel resources are available, but produces results exactly equivalent to standard serial execution. In the initial burn-in phase of chain evaluation, we achieve speedup close to linear in the number of available cores.

  @conference{angelino2014accelerating,
  year = {2014},
  author = {Angelino, Elaine and Kohler, Eddie and Waterland, Amos and Seltzer, Margo and Adams, Ryan P.},
  title = {Accelerating {MCMC} via Parallel Predictive Prefetching},
  booktitle = {Proceedings of the 30th Conference on Uncertainty in Artificial Intelligence (UAI)},
  keywords = {UAI, Markov chain Monte Carlo, parallel computing, Bayesian methods},
  note = {arXiv:1403.7265 [stat.ML]}
}