English

Variational Bayesian Optimal Experimental Design

Machine Learning 2020-01-15 v3 Machine Learning Computation Methodology

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 by building on ideas from amortized variational inference. We show theoretically and empirically that these estimators can provide significant gains in speed and accuracy over previous approaches. We further demonstrate the practicality of our approach on a number of end-to-end experiments.

Keywords

Cite

@article{arxiv.1903.05480,
  title  = {Variational Bayesian Optimal Experimental Design},
  author = {Adam Foster and Martin Jankowiak and Eli Bingham and Paul Horsfall and Yee Whye Teh and Tom Rainforth and Noah Goodman},
  journal= {arXiv preprint arXiv:1903.05480},
  year   = {2020}
}

Comments

Published as a conference paper at the Thirty-third Conference on Neural Information Processing Systems, Vancouver 2019. https://papers.nips.cc/paper/9553-variational-bayesian-optimal-experimental-design.pdf