Scalable algorithms for physics-informed neural and graph networks
Abstract
Physics-informed machine learning (PIML) has emerged as a promising new approach for simulating complex physical and biological systems that are governed by complex multiscale processes for which some data are also available. In some instances, the objective is to discover part of the hidden physics from the available data, and PIML has been shown to be particularly effective for such problems for which conventional methods may fail. Unlike commercial machine learning where training of deep neural networks requires big data, in PIML big data are not available. Instead, we can train such networks from additional information obtained by employing the physical laws and evaluating them at random points in the space–time domain. Such PIML integrates multimodality and multifidelity data with mathematical models, and implements them using neural networks or graph networks. Here, we review some of the prevailing trends in embedding physics into machine learning, using physics-informed neural networks (PINNs) based primarily on feed-forward neural networks and automatic differentiation. For more complex systems or systems of systems and unstructured data, graph neural networks (GNNs) present some distinct advantages, and here we review how physics-informed learning can be accomplished with GNNs based on graph exterior calculus to construct differential operators; wemore »
- Authors:
-
- Brown University, Providence, RI (United States)
- Brown University, Providence, RI (United States); Massachusetts Institute of Technology (MIT), Cambridge, MA (United States)
- Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
- Publication Date:
- Research Org.:
- Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
- Sponsoring Org.:
- USDOE Office of Science (SC), Advanced Scientific Computing Research (ASCR)
- OSTI Identifier:
- 1872036
- Report Number(s):
- SAND2022-6955J
Journal ID: ISSN 2632-6736; 706667
- Grant/Contract Number:
- NA0003525; SC0019453; FA9550-20-1-0358
- Resource Type:
- Accepted Manuscript
- Journal Name:
- Data-Centric Engineering
- Additional Journal Information:
- Journal Volume: 3; Journal ID: ISSN 2632-6736
- Publisher:
- Cambridge University Press
- Country of Publication:
- United States
- Language:
- English
- Subject:
- 42 ENGINEERING; graph calculus; graph neural networks; PINNS; scalability; scientific machine learning
Citation Formats
Shukla, Khemraj, Xu, Mengjia, Trask, Nathaniel, and Karniadakis, George E. Scalable algorithms for physics-informed neural and graph networks. United States: N. p., 2022.
Web. doi:10.1017/dce.2022.24.
Shukla, Khemraj, Xu, Mengjia, Trask, Nathaniel, & Karniadakis, George E. Scalable algorithms for physics-informed neural and graph networks. United States. https://doi.org/10.1017/dce.2022.24
Shukla, Khemraj, Xu, Mengjia, Trask, Nathaniel, and Karniadakis, George E. Wed .
"Scalable algorithms for physics-informed neural and graph networks". United States. https://doi.org/10.1017/dce.2022.24. https://www.osti.gov/servlets/purl/1872036.
@article{osti_1872036,
title = {Scalable algorithms for physics-informed neural and graph networks},
author = {Shukla, Khemraj and Xu, Mengjia and Trask, Nathaniel and Karniadakis, George E.},
abstractNote = {Physics-informed machine learning (PIML) has emerged as a promising new approach for simulating complex physical and biological systems that are governed by complex multiscale processes for which some data are also available. In some instances, the objective is to discover part of the hidden physics from the available data, and PIML has been shown to be particularly effective for such problems for which conventional methods may fail. Unlike commercial machine learning where training of deep neural networks requires big data, in PIML big data are not available. Instead, we can train such networks from additional information obtained by employing the physical laws and evaluating them at random points in the space–time domain. Such PIML integrates multimodality and multifidelity data with mathematical models, and implements them using neural networks or graph networks. Here, we review some of the prevailing trends in embedding physics into machine learning, using physics-informed neural networks (PINNs) based primarily on feed-forward neural networks and automatic differentiation. For more complex systems or systems of systems and unstructured data, graph neural networks (GNNs) present some distinct advantages, and here we review how physics-informed learning can be accomplished with GNNs based on graph exterior calculus to construct differential operators; we refer to these architectures as physics-informed graph networks (PIGNs). We present representative examples for both forward and inverse problems and discuss what advances are needed to scale up PINNs, PIGNs and more broadly GNNs for large-scale engineering problems.},
doi = {10.1017/dce.2022.24},
journal = {Data-Centric Engineering},
number = ,
volume = 3,
place = {United States},
year = {Wed Jun 29 00:00:00 EDT 2022},
month = {Wed Jun 29 00:00:00 EDT 2022}
}
Works referenced in this record:
Protein protein interaction network analysis of differentially expressed genes to understand involved biological processes in coronary artery disease and its different severity
journal, September 2018
- Kashyap, Shiridhar; Kumar, Sudeep; Agarwal, Vikas
- Gene Reports, Vol. 12
EvolveGCN: Evolving Graph Convolutional Networks for Dynamic Graphs
journal, April 2020
- Pareja, Aldo; Domeniconi, Giacomo; Chen, Jie
- Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 34, Issue 04
Mapping higher-order relations between brain structure and function with embedded vector representations of connectomes
journal, June 2018
- Rosenthal, Gideon; Váša, František; Griffa, Alessandra
- Nature Communications, Vol. 9, Issue 1
Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations
journal, January 2020
- Raissi, Maziar; Yazdani, Alireza; Karniadakis, George Em
- Science, Vol. 367, Issue 6481
The general inefficiency of batch training for gradient descent learning
journal, December 2003
- Wilson, D. Randall; Martinez, Tony R.
- Neural Networks, Vol. 16, Issue 10
Quantifying total uncertainty in physics-informed neural networks for solving forward and inverse stochastic problems
journal, November 2019
- Zhang, Dongkun; Lu, Lu; Guo, Ling
- Journal of Computational Physics, Vol. 397
Physics-informed machine learning
journal, May 2021
- Karniadakis, George Em; Kevrekidis, Ioannis G.; Lu, Lu
- Nature Reviews Physics, Vol. 3, Issue 6
GINNs: Graph-Informed Neural Networks for multiscale physics
journal, May 2021
- Hall, Eric J.; Taverniers, Søren; Katsoulakis, Markos A.
- Journal of Computational Physics, Vol. 433
A Physics-Informed Neural Network for Quantifying the Microstructural Properties of Polycrystalline Nickel Using Ultrasound Data: A promising approach for solving inverse problems
journal, January 2022
- Shukla, Khemraj; Jagtap, Ameya D.; Blackshire, James L.
- IEEE Signal Processing Magazine, Vol. 39, Issue 1
DeepXDE: A Deep Learning Library for Solving Differential Equations
journal, January 2021
- Lu, Lu; Meng, Xuhui; Mao, Zhiping
- SIAM Review, Vol. 63, Issue 1
Systems biology informed deep learning for inferring parameters and hidden dynamics
journal, November 2020
- Yazdani, Alireza; Lu, Lu; Raissi, Maziar
- PLOS Computational Biology, Vol. 16, Issue 11
Chemi-Net: A Molecular Graph Convolutional Network for Accurate Drug Property Prediction
journal, July 2019
- Liu, Ke; Sun, Xiangyan; Jia, Lei
- International Journal of Molecular Sciences, Vol. 20, Issue 14
Conservative physics-informed neural networks on discrete domains for conservation laws: Applications to forward and inverse problems
journal, June 2020
- Jagtap, Ameya D.; Kharazmi, Ehsan; Karniadakis, George Em
- Computer Methods in Applied Mechanics and Engineering, Vol. 365
Physics-constrained deep learning of multi-zone building thermal dynamics
journal, July 2021
- Drgoňa, Ján; Tuor, Aaron R.; Chandan, Vikas
- Energy and Buildings, Vol. 243
fPINNs: Fractional Physics-Informed Neural Networks
journal, January 2019
- Pang, Guofei; Lu, Lu; Karniadakis, George Em
- SIAM Journal on Scientific Computing, Vol. 41, Issue 4
Physics-Informed Neural Networks with Hard Constraints for Inverse Design
journal, January 2021
- Lu, Lu; Pestourie, Raphaël; Yao, Wenjie
- SIAM Journal on Scientific Computing, Vol. 43, Issue 6
hp-VPINNs: Variational physics-informed neural networks with domain decomposition
journal, February 2021
- Kharazmi, Ehsan; Zhang, Zhongqiang; Karniadakis, George E. M.
- Computer Methods in Applied Mechanics and Engineering, Vol. 374
Physics-informed neural networks for high-speed flows
journal, March 2020
- Mao, Zhiping; Jagtap, Ameya D.; Karniadakis, George Em
- Computer Methods in Applied Mechanics and Engineering, Vol. 360
On generalized moving least squares and diffuse derivatives
journal, September 2011
- Mirzaei, D.; Schaback, R.; Dehghan, M.
- IMA Journal of Numerical Analysis, Vol. 32, Issue 3
Highly accurate protein structure prediction with AlphaFold
journal, July 2021
- Jumper, John; Evans, Richard; Pritzel, Alexander
- Nature
Efficient BackProp
book, January 2012
- LeCun, Yann A.; Bottou, Léon; Orr, Genevieve B.
- Lecture Notes in Computer Science
Gradient-enhanced physics-informed neural networks for forward and inverse PDE problems
journal, April 2022
- Yu, Jeremy; Lu, Lu; Meng, Xuhui
- Computer Methods in Applied Mechanics and Engineering, Vol. 393
Accurate, Efficient and Scalable Graph Embedding
conference, May 2019
- Zeng, Hanqing; Zhou, Hongkuan; Srivastava, Ajitesh
- 2019 IEEE International Parallel and Distributed Processing Symposium (IPDPS)
Understanding Graph Embedding Methods and Their Applications
journal, January 2021
- Xu, Mengjia
- SIAM Review, Vol. 63, Issue 4
Enforcing exact physics in scientific machine learning: A data-driven exterior calculus on graphs
journal, May 2022
- Trask, Nathaniel; Huang, Andy; Hu, Xiaozhe
- Journal of Computational Physics, Vol. 456
Non-invasive inference of thrombus material properties with physics-informed neural networks
journal, March 2021
- Yin, Minglang; Zheng, Xiaoning; Humphrey, Jay D.
- Computer Methods in Applied Mechanics and Engineering, Vol. 375
Advancing mathematics by guiding human intuition with AI
journal, December 2021
- Davies, Alex; Veličković, Petar; Buesing, Lars
- Nature, Vol. 600, Issue 7887
Numerical solution of initial boundary value problems involving maxwell's equations in isotropic media
journal, May 1966
- Kane Yee,
- IEEE Transactions on Antennas and Propagation, Vol. 14, Issue 3
A Recommendation System Based on Hierarchical Clustering of an Article-Level Citation Network
journal, June 2016
- West, Jevin D.; Wesley-Smith, Ian; Bergstrom, Carl T.
- IEEE Transactions on Big Data, Vol. 2, Issue 2
EXAGRAPH: Graph and combinatorial methods for enabling exascale applications
journal, September 2021
- Acer, Seher; Azad, Ariful; Boman, Erik G.
- The International Journal of High Performance Computing Applications, Vol. 35, Issue 6
Topological Data Analysis
journal, March 2018
- Wasserman, Larry
- Annual Review of Statistics and Its Application, Vol. 5, Issue 1
User identity linkage across social networks via linked heterogeneous network embedding
journal, April 2018
- Wang, Yaqing; Feng, Chunyan; Chen, Ling
- World Wide Web, Vol. 22, Issue 6
Physics-Informed Graph Neural Network for Circuit Compact Model Development
conference, September 2020
- Gao, Xujiao; Huang, Andy; Trask, Nathaniel
- 2020 International Conference on Simulation of Semiconductor Processes and Devices (SISPAD)
Spectral/hp Element Methods for Computational Fluid Dynamics
book, January 2005
- Karniadakis, George; Sherwin, Spencer
- Oxford University Press
DynG2G: An Efficient Stochastic Graph Embedding Method for Temporal Graphs
journal, January 2022
- Xu, Mengjia; Singh, Apoorva Vikram; Karniadakis, George Em
- IEEE Transactions on Neural Networks and Learning Systems
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
Topology and data
journal, January 2009
- Carlsson, Gunnar
- Bulletin of the American Mathematical Society, Vol. 46, Issue 2
Statistical ranking and combinatorial Hodge theory
journal, November 2010
- Jiang, Xiaoye; Lim, Lek-Heng; Yao, Yuan
- Mathematical Programming, Vol. 127, Issue 1
Physics-Informed Neural Network for Ultrasound Nondestructive Quantification of Surface Breaking Cracks
journal, August 2020
- Shukla, Khemraj; Di Leoni, Patricio Clark; Blackshire, James
- Journal of Nondestructive Evaluation, Vol. 39, Issue 3
Asset diversification and systemic risk in the financial system
journal, November 2017
- Zhou, Yichen; Li, Honggang
- Journal of Economic Interaction and Coordination, Vol. 14, Issue 2
Systematic Construction of Neural Forms for Solving Partial Differential Equations Inside Rectangular Domains, Subject to Initial, Boundary and Interface Conditions
journal, August 2020
- Lagari, Pola Lydia; Tsoukalas, Lefteri H.; Safarkhani, Salar
- International Journal on Artificial Intelligence Tools, Vol. 29, Issue 05
Multisymplectic Geometry, Variational Integrators, and Nonlinear PDEs
journal, December 1998
- Marsden, Jerrold E.; Patrick, George W.; Shkoller, Steve
- Communications in Mathematical Physics, Vol. 199, Issue 2
Learning in Modal Space: Solving Time-Dependent Stochastic PDEs Using Physics-Informed Neural Networks
journal, January 2020
- Zhang, Dongkun; Guo, Ling; Karniadakis, George Em
- SIAM Journal on Scientific Computing, Vol. 42, Issue 2
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
Extended Physics-Informed Neural Networks (XPINNs): A Generalized Space-Time Domain Decomposition Based Deep Learning Framework for Nonlinear Partial Differential Equations
journal, June 2020
- Karniadakis, Ameya D. Jagtap & George Em
- Communications in Computational Physics, Vol. 28, Issue 5
A new Graph Gaussian embedding method for analyzing the effects of cognitive training
journal, September 2020
- Xu, Mengjia; Wang, Zhijiang; Zhang, Haifeng
- PLOS Computational Biology, Vol. 16, Issue 9
Learning data-driven discretizations for partial differential equations
journal, July 2019
- Bar-Sinai, Yohai; Hoyer, Stephan; Hickey, Jason
- Proceedings of the National Academy of Sciences, Vol. 116, Issue 31
Physics-Informed Neural Networks for Cardiac Activation Mapping
journal, February 2020
- Sahli Costabal, Francisco; Yang, Yibo; Perdikaris, Paris
- Frontiers in Physics, Vol. 8
DeepSpeed
conference, August 2020
- Rasley, Jeff; Rajbhandari, Samyam; Ruwase, Olatunji
- Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining
A Graph Gaussian Embedding Method for Predicting Alzheimer's Disease Progression With MEG Brain Networks
journal, May 2021
- Xu, Mengjia; Sanz, David Lopez; Garces, Pilar
- IEEE Transactions on Biomedical Engineering, Vol. 68, Issue 5
Principles of Mimetic Discretizations of Differential Operators
book,
- Bochev, Pavel B.; Hyman, James M.
- Compatible Spatial Discretizations
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
Numerical Calculation of Time-Dependent Viscous Incompressible Flow of Fluid with Free Surface
journal, January 1965
- Harlow, Francis H.; Welch, J. Eddie
- Physics of Fluids, Vol. 8, Issue 12
Physics-informed graph neural Galerkin networks: A unified framework for solving PDE-governed forward and inverse problems
journal, February 2022
- Gao, Han; Zahr, Matthew J.; Wang, Jian-Xun
- Computer Methods in Applied Mechanics and Engineering, Vol. 390
Thermodynamics-informed Graph Neural Networks
journal, January 2022
- Hernandez, Quercus; Badias, Alberto; Chinesta, Francisco
- IEEE Transactions on Artificial Intelligence
Parallel physics-informed neural networks via domain decomposition
journal, December 2021
- Shukla, Khemraj; Jagtap, Ameya D.; Karniadakis, George Em
- Journal of Computational Physics, Vol. 447
Works referencing / citing this record:
Multimodal learning with graphs
journal, April 2023
- Ektefaie, Yasha; Dasoulas, George; Noori, Ayush
- Nature Machine Intelligence, Vol. 5, Issue 4