Nonlinear sparse Bayesian learning for physics-based models
Abstract
This paper addresses the issue of overfitting while calibrating unknown parameters of over-parameterized physics-based models with noisy and incomplete observations. Here, a semi-analytical Bayesian framework of nonlinear sparse Bayesian learning (NSBL) is proposed to identify sparsity among model parameters during Bayesian inversion. NSBL offers significant advantages over machine learning algorithm of sparse Bayesian learning (SBL) for physics-based models, such as 1) the likelihood function or the posterior parameter distribution is not required to be Gaussian, and 2) prior parameter knowledge is incorporated into sparse learning (i.e. not all parameters are treated as questionable). NSBL employs the concept of automatic relevance determination (ARD) to facilitate sparsity among questionable parameters through parameterized prior distributions. The analytical tractability of NSBL is enabled by employing Gaussian ARD priors and by building a Gaussian mixture-model approximation of the posterior parameter distribution that excludes the contribution of ARD priors. Subsequently, type-II maximum likelihood is executed using Newton's method whereby the evidence and its gradient and Hessian information are computed in a semi-analytical fashion. We show numerically and analytically that SBL is a special case of NSBL for linear regression models. Subsequently, a linear regression example involving multimodality in both parameter posterior pdf and model evidence ismore »
- Authors:
-
- Carleton Univ., Ottawa, ON (Canada). Dept. of Civil & Environmental Engineering
- Sandia National Lab. (SNL-CA), Livermore, CA (United States). Quantitative Modeling & Analysis Dept.
- United States Naval Academy, Annapolis, MD (United States). Dept. of Aerospace Engineering
- Royal Military College of Canada, Kingston, ON (Canada). Dept. of Mechanical & Aerospace Engineering
- Publication Date:
- Research Org.:
- Sandia National Lab. (SNL-CA), Livermore, CA (United States)
- Sponsoring Org.:
- USDOE National Nuclear Security Administration (NNSA)
- OSTI Identifier:
- 1670717
- Report Number(s):
- SAND-2019-14718J
Journal ID: ISSN 0021-9991; 682468; TRN: US2203877
- Grant/Contract Number:
- AC04-94AL85000; NA0003525
- Resource Type:
- Accepted Manuscript
- Journal Name:
- Journal of Computational Physics
- Additional Journal Information:
- Journal Volume: 426; Journal ID: ISSN 0021-9991
- Publisher:
- Elsevier
- Country of Publication:
- United States
- Language:
- English
- Subject:
- 71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSICS; automatic relevance determination; Bayesian inference; Bayesian model selection; Gaussian mixture-model; inverse problems; physics-based modelling; sparse learning
Citation Formats
Sandhu, Rimple, Khalil, Mohammad, Pettit, Chris, Poirel, Dominique, and Sarkar, Abhijit. Nonlinear sparse Bayesian learning for physics-based models. United States: N. p., 2020.
Web. doi:10.1016/j.jcp.2020.109728.
Sandhu, Rimple, Khalil, Mohammad, Pettit, Chris, Poirel, Dominique, & Sarkar, Abhijit. Nonlinear sparse Bayesian learning for physics-based models. United States. https://doi.org/10.1016/j.jcp.2020.109728
Sandhu, Rimple, Khalil, Mohammad, Pettit, Chris, Poirel, Dominique, and Sarkar, Abhijit. Wed .
"Nonlinear sparse Bayesian learning for physics-based models". United States. https://doi.org/10.1016/j.jcp.2020.109728. https://www.osti.gov/servlets/purl/1670717.
@article{osti_1670717,
title = {Nonlinear sparse Bayesian learning for physics-based models},
author = {Sandhu, Rimple and Khalil, Mohammad and Pettit, Chris and Poirel, Dominique and Sarkar, Abhijit},
abstractNote = {This paper addresses the issue of overfitting while calibrating unknown parameters of over-parameterized physics-based models with noisy and incomplete observations. Here, a semi-analytical Bayesian framework of nonlinear sparse Bayesian learning (NSBL) is proposed to identify sparsity among model parameters during Bayesian inversion. NSBL offers significant advantages over machine learning algorithm of sparse Bayesian learning (SBL) for physics-based models, such as 1) the likelihood function or the posterior parameter distribution is not required to be Gaussian, and 2) prior parameter knowledge is incorporated into sparse learning (i.e. not all parameters are treated as questionable). NSBL employs the concept of automatic relevance determination (ARD) to facilitate sparsity among questionable parameters through parameterized prior distributions. The analytical tractability of NSBL is enabled by employing Gaussian ARD priors and by building a Gaussian mixture-model approximation of the posterior parameter distribution that excludes the contribution of ARD priors. Subsequently, type-II maximum likelihood is executed using Newton's method whereby the evidence and its gradient and Hessian information are computed in a semi-analytical fashion. We show numerically and analytically that SBL is a special case of NSBL for linear regression models. Subsequently, a linear regression example involving multimodality in both parameter posterior pdf and model evidence is considered to demonstrate the performance of NSBL in cases where SBL is inapplicable. Next, NSBL is applied to identify sparsity among the damping coefficients of a mass-spring-damper model of a shear building frame. These numerical studies demonstrate the robustness and efficiency of NSBL in alleviating overfitting during Bayesian inversion of nonlinear physics-based models.},
doi = {10.1016/j.jcp.2020.109728},
journal = {Journal of Computational Physics},
number = ,
volume = 426,
place = {United States},
year = {Wed Jul 29 00:00:00 EDT 2020},
month = {Wed Jul 29 00:00:00 EDT 2020}
}
Works referenced in this record:
Updating Models and Their Uncertainties. I: Bayesian Statistical Framework
journal, April 1998
- Beck, J. L.; Katafygiotis, L. S.
- Journal of Engineering Mechanics, Vol. 124, Issue 4
Probabilistic parameter estimation of a fluttering aeroelastic system in the transitional Reynolds number regime
journal, July 2013
- Khalil, Mohammad; Poirel, Dominique; Sarkar, Abhijit
- Journal of Sound and Vibration, Vol. 332, Issue 15
Global sensitivity analysis using polynomial chaos expansions
journal, July 2008
- Sudret, Bruno
- Reliability Engineering & System Safety, Vol. 93, Issue 7
Tracking noisy limit cycle oscillation with nonlinear filters
journal, January 2010
- Khalil, Mohammad; Sarkar, Abhijit; Adhikari, Sondipon
- Journal of Sound and Vibration, Vol. 329, Issue 2
Stochastic spectral methods for efficient Bayesian solution of inverse problems
journal, June 2007
- Marzouk, Youssef M.; Najm, Habib N.; Rahn, Larry A.
- Journal of Computational Physics, Vol. 224, Issue 2
Model Selection Using Response Measurements: Bayesian Probabilistic Approach
journal, February 2004
- Beck, James L.; Yuen, Ka-Veng
- Journal of Engineering Mechanics, Vol. 130, Issue 2
Bayesian model selection using automatic relevance determination for nonlinear dynamical systems
journal, June 2017
- Sandhu, Rimple; Pettit, Chris; Khalil, Mohammad
- Computer Methods in Applied Mechanics and Engineering, Vol. 320
Weak convergence and optimal scaling of random walk Metropolis algorithms
journal, February 1997
- Roberts, G. O.; Gelman, A.; Gilks, W. R.
- The Annals of Applied Probability, Vol. 7, Issue 1
Bayes factors: Prior sensitivity and model generalizability
journal, December 2008
- Liu, Charles C.; Aitkin, Murray
- Journal of Mathematical Psychology, Vol. 52, Issue 6
Transitional Markov Chain Monte Carlo Method for Bayesian Model Updating, Model Class Selection, and Model Averaging
journal, July 2007
- Ching, Jianye; Chen, Yi-Chu
- Journal of Engineering Mechanics, Vol. 133, Issue 7
Bayesian calibration of computer models
journal, August 2001
- Kennedy, Marc C.; O'Hagan, Anthony
- Journal of the Royal Statistical Society: Series B (Statistical Methodology), Vol. 63, Issue 3
Dimensionality reduction and polynomial chaos acceleration of Bayesian inference in inverse problems
journal, April 2009
- Marzouk, Youssef M.; Najm, Habib N.
- Journal of Computational Physics, Vol. 228, Issue 6
Data-driven model reduction for the Bayesian solution of inverse problems: DATA-DRIVEN MODEL REDUCTION FOR INVERSE PROBLEMS
journal, August 2014
- Cui, Tiangang; Marzouk, Youssef M.; Willcox, Karen E.
- International Journal for Numerical Methods in Engineering, Vol. 102, Issue 5
Variational inference of cluster-weighted models for local and global sensitivity analysis
journal, January 2014
- Pettit, Chris L.; Wilson, D. Keith
- International Journal of Reliability and Safety, Vol. 8, Issue 2/3/4
Calculation of Posterior Probabilities for Bayesian Model Class Assessment and Averaging from Posterior Samples Based on Dynamic System Data: Calculation of posterior probabilities
journal, February 2010
- Cheung, Sai Hung; Beck, James L.
- Computer-Aided Civil and Infrastructure Engineering, Vol. 25, Issue 5
The Wiener--Askey Polynomial Chaos for Stochastic Differential Equations
journal, January 2002
- Xiu, Dongbin; Karniadakis, George Em
- SIAM Journal on Scientific Computing, Vol. 24, Issue 2
The Homogeneous Chaos
journal, October 1938
- Wiener, Norbert
- American Journal of Mathematics, Vol. 60, Issue 4
Bayesian inference of nonlinear unsteady aerodynamics from aeroelastic limit cycle oscillations
journal, July 2016
- Sandhu, Rimple; Poirel, Dominique; Pettit, Chris
- Journal of Computational Physics, Vol. 316
Bayesian Compressive Sensing Using Laplace Priors
journal, January 2010
- Babacan, S. D.; Molina, R.; Katsaggelos, A. K.
- IEEE Transactions on Image Processing, Vol. 19, Issue 1
Bayesian system identification based on probability logic
journal, October 2010
- Beck, James L.
- Structural Control and Health Monitoring, Vol. 17, Issue 7