We revisit expectation propagation (EP) as a prototype for scalable
algorithms that partition big datasets into many parts and analyze each part in
parallel to perform inference of shared parameters. The algorithm should be
particularly efficient for hierarchical models, for which the EP algorithm
works on the shared parameters (hyperparameters) of the model.
The central idea of EP is to work at each step with a "tilted distribution"
that combines the likelihood for a part of the data with the "cavity
distribution," which is the approximate model for the prior and all other parts
of the data. EP iteratively approximates the moments of the tilted
distributions and incorporates those approximations into a global posterior
approximation. As such, EP can be used to divide the computation for large
models into manageable sizes. The computation for each partition can be made
parallel with occasional exchanging of information between processes through
the global posterior approximation. Moments of multivariate tilted
distributions can be approximated in various ways, including, MCMC, Laplace
approximations, and importance sampling.
%0 Journal Article
%1 gelman2014expectation
%A Gelman, Andrew
%A Vehtari, Aki
%A Jylänki, Pasi
%A Robert, Christian
%A Chopin, Nicolas
%A Cunningham, John P.
%D 2014
%K Bayes
%T Expectation propagation as a way of life
%U http://arxiv.org/abs/1412.4869
%X We revisit expectation propagation (EP) as a prototype for scalable
algorithms that partition big datasets into many parts and analyze each part in
parallel to perform inference of shared parameters. The algorithm should be
particularly efficient for hierarchical models, for which the EP algorithm
works on the shared parameters (hyperparameters) of the model.
The central idea of EP is to work at each step with a "tilted distribution"
that combines the likelihood for a part of the data with the "cavity
distribution," which is the approximate model for the prior and all other parts
of the data. EP iteratively approximates the moments of the tilted
distributions and incorporates those approximations into a global posterior
approximation. As such, EP can be used to divide the computation for large
models into manageable sizes. The computation for each partition can be made
parallel with occasional exchanging of information between processes through
the global posterior approximation. Moments of multivariate tilted
distributions can be approximated in various ways, including, MCMC, Laplace
approximations, and importance sampling.
@article{gelman2014expectation,
abstract = {We revisit expectation propagation (EP) as a prototype for scalable
algorithms that partition big datasets into many parts and analyze each part in
parallel to perform inference of shared parameters. The algorithm should be
particularly efficient for hierarchical models, for which the EP algorithm
works on the shared parameters (hyperparameters) of the model.
The central idea of EP is to work at each step with a "tilted distribution"
that combines the likelihood for a part of the data with the "cavity
distribution," which is the approximate model for the prior and all other parts
of the data. EP iteratively approximates the moments of the tilted
distributions and incorporates those approximations into a global posterior
approximation. As such, EP can be used to divide the computation for large
models into manageable sizes. The computation for each partition can be made
parallel with occasional exchanging of information between processes through
the global posterior approximation. Moments of multivariate tilted
distributions can be approximated in various ways, including, MCMC, Laplace
approximations, and importance sampling.},
added-at = {2015-07-12T15:00:23.000+0200},
author = {Gelman, Andrew and Vehtari, Aki and Jylänki, Pasi and Robert, Christian and Chopin, Nicolas and Cunningham, John P.},
biburl = {https://www.bibsonomy.org/bibtex/2840b635a1d8ee017f929a3680ff919dd/murraystat},
interhash = {51164ad6ddbe8222c69cb9007b359484},
intrahash = {840b635a1d8ee017f929a3680ff919dd},
keywords = {Bayes},
note = {cite arxiv:1412.4869v1.pdfComment: 19 pages (+ appendix), 5 figures},
timestamp = {2015-07-12T15:00:23.000+0200},
title = {Expectation propagation as a way of life},
url = {http://arxiv.org/abs/1412.4869},
year = 2014
}