Bayesian optimal experimental design (BOED) is a principled framework for
making efficient use of limited experimental resources. Unfortunately, its
applicability is hampered by the difficulty of obtaining accurate estimates of
the expected information gain (EIG) of an experiment. To address this, we
introduce several classes of fast EIG estimators suited to the experiment
design context by building on ideas from variational inference and mutual
information estimation. We show theoretically and empirically that these
estimators can provide significant gains in speed and accuracy over previous
approaches. We demonstrate the practicality of our approach via a number of
experiments, including an adaptive experiment with human participants.
Description
[1903.05480] Variational Estimators for Bayesian Optimal Experimental Design
%0 Journal Article
%1 foster2019variational
%A Foster, Adam
%A Jankowiak, Martin
%A Bingham, Eli
%A Horsfall, Paul
%A Teh, Yee Whye
%A Rainforth, Tom
%A Goodman, Noah
%D 2019
%K bayesian
%T Variational Estimators for Bayesian Optimal Experimental Design
%U http://arxiv.org/abs/1903.05480
%X Bayesian optimal experimental design (BOED) is a principled framework for
making efficient use of limited experimental resources. Unfortunately, its
applicability is hampered by the difficulty of obtaining accurate estimates of
the expected information gain (EIG) of an experiment. To address this, we
introduce several classes of fast EIG estimators suited to the experiment
design context by building on ideas from variational inference and mutual
information estimation. We show theoretically and empirically that these
estimators can provide significant gains in speed and accuracy over previous
approaches. We demonstrate the practicality of our approach via a number of
experiments, including an adaptive experiment with human participants.
@article{foster2019variational,
abstract = {Bayesian optimal experimental design (BOED) is a principled framework for
making efficient use of limited experimental resources. Unfortunately, its
applicability is hampered by the difficulty of obtaining accurate estimates of
the expected information gain (EIG) of an experiment. To address this, we
introduce several classes of fast EIG estimators suited to the experiment
design context by building on ideas from variational inference and mutual
information estimation. We show theoretically and empirically that these
estimators can provide significant gains in speed and accuracy over previous
approaches. We demonstrate the practicality of our approach via a number of
experiments, including an adaptive experiment with human participants.},
added-at = {2019-03-15T00:30:20.000+0100},
author = {Foster, Adam and Jankowiak, Martin and Bingham, Eli and Horsfall, Paul and Teh, Yee Whye and Rainforth, Tom and Goodman, Noah},
biburl = {https://www.bibsonomy.org/bibtex/21e1095d36c172d599a01c07c649bdb08/kirk86},
description = {[1903.05480] Variational Estimators for Bayesian Optimal Experimental Design},
interhash = {16e281aa8b73eeba07c237345f6cee32},
intrahash = {1e1095d36c172d599a01c07c649bdb08},
keywords = {bayesian},
note = {cite arxiv:1903.05480},
timestamp = {2019-03-15T00:30:20.000+0100},
title = {Variational Estimators for Bayesian Optimal Experimental Design},
url = {http://arxiv.org/abs/1903.05480},
year = 2019
}