Adversarial uncertainty quantification in physics-informed neural networks
Abstract
Here, we introduce a deep learning framework for quantifying and propagating uncertainty in systems governed by non-linear differential equations using physics-informed neural networks. Specifically, we employ latent variable models to construct probabilistic representations for the system states, and put forth an adversarial inference procedure for training them on data, while constraining their predictions to satisfy given physical laws expressed by partial differential equations. Such physics-informed constraints provide a regularization mechanism for effectively training deep generative models as surrogates of physical systems in which the cost of data acquisition is high, and training data-sets are typically small. This provides a flexible framework for characterizing uncertainty in the outputs of physical systems due to randomness in their inputs or noise in their observations that entirely bypasses the need for repeatedly sampling expensive experiments or numerical simulators. Moreover, we demonstrate the effectiveness of our approach through a series of examples involving uncertainty propagation in non-linear conservation laws, and the discovery of constitutive laws for flow through porous media directly from noisy data.
- Authors:
-
- Univ. of Pennsylvania, Philadelphia, PA (United States)
- Publication Date:
- Research Org.:
- Univ. of Pennsylvania, Philadelphia, PA (United States)
- Sponsoring Org.:
- USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR)
- OSTI Identifier:
- 1595802
- Alternate Identifier(s):
- OSTI ID: 1691928
- Grant/Contract Number:
- SC0019116; HR00111890034
- Resource Type:
- Accepted Manuscript
- Journal Name:
- Journal of Computational Physics
- Additional Journal Information:
- Journal Volume: 394; Journal Issue: C; Related Information: https://github.com/PredictiveIntelligenceLab/UQPINNs; Journal ID: ISSN 0021-9991
- Publisher:
- Elsevier
- Country of Publication:
- United States
- Language:
- English
- Subject:
- 97 MATHEMATICS AND COMPUTING; Variational inference; Generative adversarial networks; Probabilistic deep learning; Probabilistic scientific computing; Data-driven modeling
Citation Formats
Yang, Yibo, and Perdikaris, Paris. Adversarial uncertainty quantification in physics-informed neural networks. United States: N. p., 2019.
Web. doi:10.1016/j.jcp.2019.05.027.
Yang, Yibo, & Perdikaris, Paris. Adversarial uncertainty quantification in physics-informed neural networks. United States. https://doi.org/10.1016/j.jcp.2019.05.027
Yang, Yibo, and Perdikaris, Paris. Fri .
"Adversarial uncertainty quantification in physics-informed neural networks". United States. https://doi.org/10.1016/j.jcp.2019.05.027. https://www.osti.gov/servlets/purl/1595802.
@article{osti_1595802,
title = {Adversarial uncertainty quantification in physics-informed neural networks},
author = {Yang, Yibo and Perdikaris, Paris},
abstractNote = {Here, we introduce a deep learning framework for quantifying and propagating uncertainty in systems governed by non-linear differential equations using physics-informed neural networks. Specifically, we employ latent variable models to construct probabilistic representations for the system states, and put forth an adversarial inference procedure for training them on data, while constraining their predictions to satisfy given physical laws expressed by partial differential equations. Such physics-informed constraints provide a regularization mechanism for effectively training deep generative models as surrogates of physical systems in which the cost of data acquisition is high, and training data-sets are typically small. This provides a flexible framework for characterizing uncertainty in the outputs of physical systems due to randomness in their inputs or noise in their observations that entirely bypasses the need for repeatedly sampling expensive experiments or numerical simulators. Moreover, we demonstrate the effectiveness of our approach through a series of examples involving uncertainty propagation in non-linear conservation laws, and the discovery of constitutive laws for flow through porous media directly from noisy data.},
doi = {10.1016/j.jcp.2019.05.027},
journal = {Journal of Computational Physics},
number = C,
volume = 394,
place = {United States},
year = {Fri May 24 00:00:00 EDT 2019},
month = {Fri May 24 00:00:00 EDT 2019}
}
Web of Science
Works referenced in this record:
Deep learning
journal, May 2015
- LeCun, Yann; Bengio, Yoshua; Hinton, Geoffrey
- Nature, Vol. 521, Issue 7553
Human-level concept learning through probabilistic program induction
journal, December 2015
- Lake, B. M.; Salakhutdinov, R.; Tenenbaum, J. B.
- Science, Vol. 350, Issue 6266
Predicting the sequence specificities of DNA- and RNA-binding proteins by deep learning
journal, July 2015
- Alipanahi, Babak; Delong, Andrew; Weirauch, Matthew T.
- Nature Biotechnology, Vol. 33, Issue 8
Hidden physics models: Machine learning of nonlinear partial differential equations
journal, March 2018
- Raissi, Maziar; Karniadakis, George Em
- Journal of Computational Physics, Vol. 357
A hybrid neural network-first principles approach to process modeling
journal, October 1992
- Psichogios, Dimitris C.; Ungar, Lyle H.
- AIChE Journal, Vol. 38, Issue 10
Artificial neural networks for solving ordinary and partial differential equations
journal, January 1998
- Lagaris, I. E.; Likas, A.; Fotiadis, D. I.
- IEEE Transactions on Neural Networks, Vol. 9, Issue 5
Multiscale modeling and simulation of brain blood flow
journal, February 2016
- Perdikaris, Paris; Grinberg, Leopold; Karniadakis, George Em
- Physics of Fluids, Vol. 28, Issue 2
Polynomial Chaos in Stochastic Finite Elements
journal, March 1990
- Ghanem, Roger; Spanos, P. D.
- Journal of Applied Mechanics, Vol. 57, Issue 1
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
Uncertainty Quantification and Polynomial Chaos Techniques in Computational Fluid Dynamics
journal, January 2009
- Najm, Habib N.
- Annual Review of Fluid Mechanics, Vol. 41, Issue 1
Multi-level Monte Carlo Finite Element method for elliptic PDEs with stochastic coefficients
journal, April 2011
- Barth, Andrea; Schwab, Christoph; Zollinger, Nathaniel
- Numerische Mathematik, Vol. 119, Issue 1
Optimal Model Management for Multifidelity Monte Carlo Estimation
journal, January 2016
- Peherstorfer, Benjamin; Willcox, Karen; Gunzburger, Max
- SIAM Journal on Scientific Computing, Vol. 38, Issue 5
The Proper Orthogonal Decomposition in the Analysis of Turbulent Flows
journal, January 1993
- Berkooz, G.; Holmes, P.; Lumley, J. L.
- Annual Review of Fluid Mechanics, Vol. 25, Issue 1
Multi-output local Gaussian process regression: Applications to uncertainty quantification
journal, July 2012
- Bilionis, Ilias; Zabaras, Nicholas
- Journal of Computational Physics, Vol. 231, Issue 17
Multi-output separable Gaussian process: Towards an efficient, fully Bayesian paradigm for uncertainty quantification
journal, May 2013
- Bilionis, Ilias; Zabaras, Nicholas; Konomi, Bledar A.
- Journal of Computational Physics, Vol. 241
Multifidelity Information Fusion Algorithms for High-Dimensional Systems and Massive Data sets
journal, January 2016
- Perdikaris, Paris; Venturi, Daniele; Karniadakis, George Em
- SIAM Journal on Scientific Computing, Vol. 38, Issue 4
Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
journal, February 2019
- Raissi, M.; Perdikaris, P.; Karniadakis, G. E.
- Journal of Computational Physics, Vol. 378
Variational Inference: A Review for Statisticians
journal, July 2016
- Blei, David M.; Kucukelbir, Alp; McAuliffe, Jon D.
- Journal of the American Statistical Association, Vol. 112, Issue 518
Inferring solutions of differential equations using noisy multi-fidelity data
journal, April 2017
- Raissi, Maziar; Perdikaris, Paris; Karniadakis, George Em
- Journal of Computational Physics, Vol. 335
Numerical Gaussian Processes for Time-Dependent and Nonlinear Partial Differential Equations
journal, January 2018
- Raissi, Maziar; Perdikaris, Paris; Karniadakis, George Em
- SIAM Journal on Scientific Computing, Vol. 40, Issue 1
The partial differential equation ut + uux = μxx
journal, September 1950
- Hopf, Eberhard
- Communications on Pure and Applied Mathematics, Vol. 3, Issue 3
On the limited memory BFGS method for large scale optimization
journal, August 1989
- Liu, Dong C.; Nocedal, Jorge
- Mathematical Programming, Vol. 45, Issue 1-3
Modeling fluid flow and transport in variably saturated porous media with the STOMP simulator. 1. Nonvolatile three-phase model description
journal, January 1995
- White, M. D.; Oostrom, M.; Lenhard, R. J.
- Advances in Water Resources, Vol. 18, Issue 6
A Closed-form Equation for Predicting the Hydraulic Conductivity of Unsaturated Soils1
journal, January 1980
- van Genuchten, M. Th.
- Soil Science Society of America Journal, Vol. 44, Issue 5
The partial differential equation ut + uux = μxx
journal, September 1950
- Hopf, Eberhard
- Communications on Pure and Applied Mathematics, Vol. 3, Issue 3
Multiscale modeling and simulation of brain blood flow
journal, February 2016
- Perdikaris, Paris; Grinberg, Leopold; Karniadakis, George Em
- Physics of Fluids, Vol. 28, Issue 2
Variational Inference: A Review for Statisticians
journal, July 2016
- Blei, David M.; Kucukelbir, Alp; McAuliffe, Jon D.
- Journal of the American Statistical Association, Vol. 112, Issue 518
Automatic differentiation in machine learning: a survey
text, January 2015
- Baydin, Atilim Gunes; Pearlmutter, Barak A.; Radul, Alexey Andreyevich
- arXiv
Probabilistic Numerical Methods for PDE-constrained Bayesian Inverse Problems
text, January 2017
- Cockayne, Jon; Oates, Chris; Sullivan, Tim
- arXiv
Adversarial Variational Bayes: Unifying Variational Autoencoders and Generative Adversarial Networks
preprint, January 2017
- Mescheder, Lars; Nowozin, Sebastian; Geiger, Andreas
- arXiv
Optimal fidelity multi-level Monte Carlo for quantification of uncertainty in simulations of cloud cavitation collapse
preprint, January 2017
- Šukys, Jonas; Rasthofer, Ursula; Wermelinger, Fabian
- arXiv
Works referencing / citing this record:
Multiscale Modeling Meets Machine Learning: What Can We Learn?
journal, February 2020
- Peng, Grace C. Y.; Alber, Mark; Buganza Tepole, Adrian
- Archives of Computational Methods in Engineering
Integrating machine learning and multiscale modeling—perspectives, challenges, and opportunities in the biological, biomedical, and behavioral sciences
journal, November 2019
- Alber, Mark; Buganza Tepole, Adrian; Cannon, William R.
- npj Digital Medicine, Vol. 2, Issue 1
Multiscale modeling meets machine learning: What can we learn?
preprint, January 2019
- Peng, Grace C. Y.; Alber, Mark; Tepole, Adrian Buganza
- arXiv