Department of Mathematics
 Search | Help | Login | pdf version | printable version

Math @ Duke



Publications [#322553] of David B. Dunson


Papers Published

  1. Srivastava, S; Cevher, V; Tran-Dinh, Q; Dunson, DB, WASP: Scalable Bayes via barycenters of subset posteriors, Journal of Machine Learning Research, vol. 38 (January, 2015), pp. 912-920
    (last updated on 2019/05/22)

    Copyright 2015 by the authors. 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.
ph: 919.660.2800
fax: 919.660.2821

Mathematics Department
Duke University, Box 90320
Durham, NC 27708-0320