Conference proceeding
WASP: Scalable Bayes via barycenters of subset posteriors
ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 38, Vol.38, pp.912-920
JMLR Workshop and Conference Proceedings
01/01/2015
Abstract
The promise of Bayesian methods for big data sets has not fully been realized due to the lack of scalable computational algorithms. For massive data, it is necessary to store and process subsets on different machines in a distributed manner. We propose a simple, general, and highly efficient approach, which first runs a posterior sampling algorithm in parallel on different machines for subsets of a large data set. To combine these subset posteriors, we calculate the Wasserstein barycenter via a highly efficient linear program. The resulting estimate for the Wasserstein posterior (WASP) has an atomic form, facilitating straightforward estimation of posterior summaries of functionals of interest. The WASP approach allows posterior sampling algorithms for smaller data sets to be trivially scaled to huge data. We provide theoretical justification in terms of posterior consistency and algorithm efficiency. Examples are provided in complex settings including Gaussian process regression and nonparametric Bayes mixture models.
Details
- Title: Subtitle
- WASP: Scalable Bayes via barycenters of subset posteriors
- Creators
- Sanvesh Srivastava - Duke UniversityVolkan Cevher - Federal ReserveQuoc Tran-Dinh - Federal ReserveDavid B. Dunson - Duke University
- Contributors
- G Lebanon (Editor)SVN Vishwanathan (Editor)
- Resource Type
- Conference proceeding
- Publication Details
- ARTIFICIAL INTELLIGENCE AND STATISTICS, VOL 38, Vol.38, pp.912-920
- Publisher
- Microtome Publishing
- Series
- JMLR Workshop and Conference Proceedings
- ISSN
- 1938-7288
- Number of pages
- 9
- Grant note
- ERC Future Proof DMS-1127914 / National Science Foundation; National Science Foundation (NSF) MIRG-268398 / European Commission; European Commission Joint Research Centre SNF 200021-132548; SNF 200021-146750; SNF CRSII2-147633 / Swiss Science Foundation; Swiss National Science Foundation (SNSF) R01-ES-017436 / National Institute of Environmental Health Sciences of the National Institutes of Health; United States Department of Health & Human Services; National Institutes of Health (NIH) - USA; NIH National Institute of Environmental Health Sciences (NIEHS)
- Language
- English
- Date published
- 01/01/2015
- Academic Unit
- Statistics and Actuarial Science
- Record Identifier
- 9984288737002771
Metrics
4 Record Views