×

A one-pass sequential Monte Carlo method for Bayesian analysis of massive datasets. (English) Zbl 1333.62007

Summary: For Bayesian analysis of massive data, Markov chain Monte Carlo (MCMC) techniques often prove infeasible due to computational resource constraints. Standard MCMC methods generally require a complete scan of the dataset for each iteration. G. Ridgeway and D. Madigan [“A sequential Monte Carlo method for Bayesian analysis of massive datasets”, J. Knowl. Discov. Data Min. 7, No. 3, 301–319 (2002; doi:10.1023/A:1024084221803)] and N. Chopin [Biometrika 89, No. 3, 539–551 (2002; Zbl 1036.62062)] recently presented importance sampling algorithms that combined simulations from a posterior distribution conditioned on a small portion of the dataset with a reweighting of those simulations to condition on the remainder of the dataset. While these algorithms drastically reduce the number of data accesses as compared to traditional MCMC, they still require substantially more than a single pass over the dataset. In this paper, we present “1PFS”, an efficient, one-pass algorithm. The algorithm employs a simple modification of the Ridgeway and Madigan [loc. cit.] particle filtering algorithm that replaces the MCMC based “rejuvenation” step with a more efficient “shrinkage” kernel smoothing based step. To show proof-of-concept and to enable a direct comparison, we demonstrate 1PFS on the same examples presented in Ridgeway and Madigan [loc. cit.], namely a mixture model for Markov chains and Bayesian logistic regression. Our results indicate the proposed scheme delivers accurate parameter estimates while employing only a single pass through the data.

MSC:

62-07 Data analysis (statistics) (MSC2010)
62L05 Sequential statistical design
62F15 Bayesian inference
62F40 Bootstrap, jackknife and other resampling methods
65C05 Monte Carlo methods

Citations:

Zbl 1036.62062
PDFBibTeX XMLCite
Full Text: DOI Euclid