Yes, but did it work?: Evaluating variational inference

Yuling Yao; Aki Vehtari; Daniel Simpson; Andrew Gelman

Yes, but did it work? Evaluating variational inference

Yuling Yao, Aki Vehtari, Daniel Simpson, Andrew Gelman

Research output: Chapter in Book/Report/Conference proceeding › Conference Paper › Other › peer-review

Abstract

While it's always possible to compute a variational approximation to a posterior distribution, it can be difficult to discover problems with this approximation". We propose two diagnostic algorithms to alleviate this problem. The Paretosmoothed importance sampling (PSIS) diagnostic gives a goodness of fit measurement for joint distributions, while simultaneously improving the error in the estimate. The variational simulationbased calibration (VSBC) assesses the average performance of point estimates.

Original language	English
Title of host publication	Proceedings of the 35th International Conference on Machine Learning
Editors	Jennifer Dy, Andreas Krause
Publisher	International Machine Learning Society (IMLS)
Pages	8887-8895
Number of pages	9
ISBN (Electronic)	9781510867963
Publication status	Published - 2018
Externally published	Yes
Event	International Conference on Machine Learning 2018 - Stockholmsmässan, Stockholm, Sweden Duration: 10 Jul 2018 → 15 Jul 2018 Conference number: 35th

Conference

Conference	International Conference on Machine Learning 2018
Abbreviated title	ICML 2018
Country/Territory	Sweden
City	Stockholm
Period	10/07/18 → 15/07/18

Access to Document

http://proceedings.mlr.press/v80/yao18a/yao18a.pdf

Cite this

@inproceedings{6489ab04bed44053aa830181fc1c2446,

title = "Yes, but did it work?: Evaluating variational inference",

abstract = "While it's always possible to compute a variational approximation to a posterior distribution, it can be difficult to discover problems with this approximation{"}. We propose two diagnostic algorithms to alleviate this problem. The Paretosmoothed importance sampling (PSIS) diagnostic gives a goodness of fit measurement for joint distributions, while simultaneously improving the error in the estimate. The variational simulationbased calibration (VSBC) assesses the average performance of point estimates.",

author = "Yuling Yao and Aki Vehtari and Daniel Simpson and Andrew Gelman",

note = "Publisher Copyright: {\textcopyright} 35th International Conference on Machine Learning, ICML 2018.All Rights Reserved. Copyright: Copyright 2019 Elsevier B.V., All rights reserved.; International Conference on Machine Learning 2018, ICML 2018 ; Conference date: 10-07-2018 Through 15-07-2018",

year = "2018",

language = "English",

pages = "8887--8895",

editor = "Jennifer Dy and Andreas Krause",

booktitle = "Proceedings of the 35th International Conference on Machine Learning",

publisher = "International Machine Learning Society (IMLS)",

}

Yao, Y, Vehtari, A, Simpson, D & Gelman, A 2018, Yes, but did it work? Evaluating variational inference. in J Dy & A Krause (eds), Proceedings of the 35th International Conference on Machine Learning. International Machine Learning Society (IMLS), pp. 8887-8895, International Conference on Machine Learning 2018, Stockholm, Sweden, 10/07/18. <http://proceedings.mlr.press/v80/yao18a/yao18a.pdf>

Yes, but did it work? Evaluating variational inference. / Yao, Yuling; Vehtari, Aki; Simpson, Daniel et al.

Proceedings of the 35th International Conference on Machine Learning. ed. / Jennifer Dy; Andreas Krause. International Machine Learning Society (IMLS), 2018. p. 8887-8895.

Research output: Chapter in Book/Report/Conference proceeding › Conference Paper › Other › peer-review

TY - GEN

T1 - Yes, but did it work?

T2 - International Conference on Machine Learning 2018

AU - Yao, Yuling

AU - Vehtari, Aki

AU - Simpson, Daniel

AU - Gelman, Andrew

N1 - Conference code: 35th

PY - 2018

Y1 - 2018

N2 - While it's always possible to compute a variational approximation to a posterior distribution, it can be difficult to discover problems with this approximation". We propose two diagnostic algorithms to alleviate this problem. The Paretosmoothed importance sampling (PSIS) diagnostic gives a goodness of fit measurement for joint distributions, while simultaneously improving the error in the estimate. The variational simulationbased calibration (VSBC) assesses the average performance of point estimates.

AB - While it's always possible to compute a variational approximation to a posterior distribution, it can be difficult to discover problems with this approximation". We propose two diagnostic algorithms to alleviate this problem. The Paretosmoothed importance sampling (PSIS) diagnostic gives a goodness of fit measurement for joint distributions, while simultaneously improving the error in the estimate. The variational simulationbased calibration (VSBC) assesses the average performance of point estimates.

UR - http://www.scopus.com/inward/record.url?scp=85057318872&partnerID=8YFLogxK

M3 - Conference Paper

AN - SCOPUS:85057318872

SP - 8887

EP - 8895

BT - Proceedings of the 35th International Conference on Machine Learning

A2 - Dy, Jennifer

A2 - Krause, Andreas

PB - International Machine Learning Society (IMLS)

Y2 - 10 July 2018 through 15 July 2018

ER -

Yes, but did it work? Evaluating variational inference

Abstract

Conference

Access to Document

Other files and links

Cite this