0
Research Papers

Computationally Efficient Variational Approximations for Bayesian Inverse Problems

[+] Author and Article Information
Panagiotis Tsilifis

Department of Mathematics,
University of Southern California,
Los Angeles, CA 90089-2532
e-mail: tsilifis@usc.edu

Ilias Bilionis

Assistant Professor
School of Mechanical Engineering,
Purdue University,
West Lafayette, IN 47906-2088
e-mail: ibilion@purdue.edu

Ioannis Katsounaros

Leiden Institute of Chemistry,
Leiden University,
Einsteinweg 55, P.O. Box 9502,
Leiden 2300 RA, The Netherlands
e-mail: katsounaros@anl.gov

Nicholas Zabaras

Professor
Warwick Centre for Predictive Modeling,
University of Warwick,
Coventry CV4 7AL, UK
e-mail: nzabaras@gmail.com

Manuscript received September 15, 2015; final manuscript received July 5, 2016; published online July 26, 2016. Editor: Ashley F. Emery.

J. Verif. Valid. Uncert 1(3), 031004 (Jul 26, 2016) (13 pages) Paper No: VVUQ-15-1041; doi: 10.1115/1.4034102 History: Received September 15, 2015; Revised July 05, 2016

The major drawback of the Bayesian approach to model calibration is the computational burden involved in describing the posterior distribution of the unknown model parameters arising from the fact that typical Markov chain Monte Carlo (MCMC) samplers require thousands of forward model evaluations. In this work, we develop a variational Bayesian approach to model calibration which uses an information theoretic criterion to recast the posterior problem as an optimization problem. Specifically, we parameterize the posterior using the family of Gaussian mixtures and seek to minimize the information loss incurred by replacing the true posterior with an approximate one. Our approach is of particular importance in underdetermined problems with expensive forward models in which both the classical approach of minimizing a potentially regularized misfit function and MCMC are not viable options. We test our methodology on two surrogate-free examples and show that it dramatically outperforms MCMC methods.

FIGURES IN THIS ARTICLE
<>
Copyright © 2016 by ASME
Your Session has timed out. Please sign back in to continue.

References

Figures

Grahic Jump Location
Fig. 1

Reaction kinetic model: Simplified reaction scheme for the reduction of nitrate

Grahic Jump Location
Fig. 2

Reaction kinetic model: Comparison of the variational posterior (VAR (L = 1)) of thescaled kinetic rate constants κi as well as of the likelihood noise σ to the MCMC (MALA) histograms of the same quantity. The prior probability density of each quantity is shown as a dashed line.

Grahic Jump Location
Fig. 3

Reaction kinetic model: The “•,” “ ∇,” “,” “◇,” and “⬠” signs indicate the scaled experimental measurements for NO3−, NO2−, N2, N2O, and NH3, respectively. The lines and the shaded areas around them correspond to the median and 95% credible intervals of the scaled concentration, ui, as a function of the scaled time τ. The left plot shows the results obtained by approximating the posterior of the parameters via the variational approach. The right plot shows the results obtained via MCMC (MALA).

Grahic Jump Location
Fig. 4

Concentration map of the contaminant at the four time instances when the measurements are taken: (a) t = 0.075, (b) t = 0.15, (c) t = 0.225, and (d) t = 0.3

Grahic Jump Location
Fig. 5

Contaminant source identification, first case: Comparison of the variational posterior (VAR (L = 1)) of the source location ξ=(ξ1,ξ2) as well as of the likelihood noise σ to the MCMC (MALA) histograms of the same quantities. The true value of each quantity is marked by a vertical, dashed line.

Grahic Jump Location
Fig. 6

Contaminant source identification, second case: Comparison of the variational posterior (VAR (L = 1)) of the source location ξ=(ξ1,ξ2) as well as of the likelihood noise σ to the MCMC (MALA) histograms of the same quantities. The true value of each quantity is marked by a vertical, dashed line.

Tables

Errata

Discussions

Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In