Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Constrained or unconstrained? Neural-network-based equation discovery from data

Journal Article · · Computer Methods in Applied Mechanics and Engineering

Throughout many fields, practitioners often rely on differential equations to model systems. Yet, for many applications, the theoretical derivation of such equations and/or the accurate resolution of their solutions may be intractable. Instead, recently developed methods, including those based on parameter estimation, operator subset selection, and neural networks, allow for the data-driven discovery of both ordinary and partial differential equations (PDEs), on a spectrum of interpretability. The success of these strategies is often contingent upon the correct identification of representative equations from noisy observations of state variables and, as importantly and intertwined with that, the mathematical strategies utilized to enforce those equations. Specifically, the latter has been commonly addressed via unconstrained optimization strategies. Representing the PDE as a neural network, we propose to discover the PDE (or the associated operator) by solving a constrained optimization problem and using an intermediate state representation similar to a physics-informed neural network (PINN). The objective function of this constrained optimization problem promotes matching the data, while the constraints require that the discovered PDE is satisfied at a number of spatial collocation points. We present a penalty method and a widely used trust-region barrier method to solve this constrained optimization problem, and we compare these methods on numerical examples. Our results on several example problems demonstrate that the latter constrained method outperforms the penalty method, particularly for higher noise levels or fewer collocation points. This work motivates further exploration into using sophisticated constrained optimization methods in scientific machine learning, as opposed to their commonly used, penalty-method or unconstrained counterparts. For both of these methods, we solve these discovered neural network PDEs with classical methods, such as finite difference methods, as opposed to PINNs-type methods relying on automatic differentiation. Here, we briefly highlight how simultaneously fitting the data while discovering the PDE improves the robustness to noise and other small, yet crucial, implementation details.

Research Organization:
University of Colorado, Boulder, CO (United States)
Sponsoring Organization:
USDOE National Nuclear Security Administration (NNSA); National Science Foundation (NSF)
Grant/Contract Number:
NA0003962; NA0003525
OSTI ID:
2589832
Journal Information:
Computer Methods in Applied Mechanics and Engineering, Journal Name: Computer Methods in Applied Mechanics and Engineering Vol. 436; ISSN 0045-7825
Publisher:
Elsevier BVCopyright Statement
Country of Publication:
United States
Language:
English

References (43)

Scientific Machine Learning Through Physics–Informed Neural Networks: Where we are and What’s Next journal July 2022
Evaluating time series forecasting models: an empirical study on performance estimation methods journal October 2020
Conservative physics-informed neural networks on discrete domains for conservation laws: Applications to forward and inverse problems journal June 2020
Gradient-enhanced physics-informed neural networks for forward and inverse PDE problems journal April 2022
Derivative-based SINDy (DSINDy): Addressing the challenge of discovering governing equations from noisy data journal August 2023
Discovering governing equations in discrete systems using PINNs journal November 2023
Data-driven modeling and learning in science and engineering journal November 2019
Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations journal February 2019
Deep learning of dynamics and signal-noise decomposition with time-stepping constraints journal November 2019
PDE-Net 2.0: Learning PDEs from data with a numeric-symbolic hybrid deep network journal December 2019
Data-driven deep learning of partial differential equations in modal space journal May 2020
Weak SINDy for partial differential equations journal October 2021
SPINN: Sparse, Physics-based, and partially Interpretable Neural Networks for PDEs journal November 2021
Parallel physics-informed neural networks via domain decomposition journal December 2021
When and why PINNs fail to train: A neural tangent kernel perspective journal January 2022
Deep neural network modeling of unknown partial differential equations in nodal space journal January 2022
Physics constrained learning for data-driven inverse modeling from sparse observations journal March 2022
Physics and equality constrained artificial neural networks: Application to forward and inverse problems with multi-fidelity data fusion journal August 2022
Self-adaptive physics-informed neural networks journal February 2023
A practical PINN framework for multi-scale problems with multi-magnitude loss terms journal August 2024
Enhanced physics-informed neural networks with Augmented Lagrangian relaxation method (AL-PINNs) journal September 2023
Physics-informed learning of governing equations from scarce data journal October 2021
SciPy 1.0: fundamental algorithms for scientific computing in Python journal February 2020
Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators journal March 2021
Automated reverse engineering of nonlinear dynamical systems journal June 2007
Discovering governing equations from data by sparse identification of nonlinear dynamical systems journal March 2016
Structural and practical identifiability analysis of partially observed dynamical models by exploiting the profile likelihood journal June 2009
Learning partial differential equations via data discovery and sparse optimization journal January 2017
Learning partial differential equations for biological transport models from noisy spatio-temporal data
  • Lagergren, John H.; Nardini, John T.; Michael Lavigne, G.
  • Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, Vol. 476, Issue 2234 https://doi.org/10.1098/rspa.2019.0800
journal February 2020
SINDy-PI: a robust algorithm for parallel implicit sparse identification of nonlinear dynamics
  • Kaheman, Kadierdan; Kutz, J. Nathan; Brunton, Steven L.
  • Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, Vol. 476, Issue 2242 https://doi.org/10.1098/rspa.2020.0279
journal October 2020
Method for Solving the Sine-Gordon Equation journal June 1973
Order of Nonlinearity as a Complexity Measure for Models Generated by Symbolic Regression via Pareto Genetic Programming journal April 2009
Efficient training of physics‐informed neural networks via importance sampling journal April 2021
Distilling Free-Form Natural Laws from Experimental Data journal April 2009
Hidden fluid mechanics: Learning velocity and pressure fields from flow visualizations journal January 2020
Understanding and Mitigating Gradient Flow Pathologies in Physics-Informed Neural Networks journal January 2021
Physics-Informed Neural Networks with Hard Constraints for Inverse Design journal January 2021
On the Implementation of an Algorithm for Large-Scale Equality Constrained Optimization journal August 1998
An Interior Point Algorithm for Large-Scale Nonlinear Programming journal January 1999
Training with Noise is Equivalent to Tikhonov Regularization journal January 1995
PySINDy: A comprehensive Python package for robust sparse system identification journal January 2022
Investigating and Mitigating Failure Modes in Physics-Informed Neural Networks (PINNs) journal June 2023
Numerical Differentiation of Noisy, Nonsmooth Data journal January 2011

Similar Records

Gradient-enhanced physics-informed neural networks for forward and inverse PDE problems
Journal Article · Fri Mar 18 00:00:00 EDT 2022 · Computer Methods in Applied Mechanics and Engineering · OSTI ID:1976976

Exact enforcement of temporal continuity in sequential physics-informed neural networks
Journal Article · Mon Jul 15 00:00:00 EDT 2024 · Computer Methods in Applied Mechanics and Engineering · OSTI ID:2429369

Quantifying local and global mass balance errors in physics-informed neural networks
Journal Article · Fri Jul 05 00:00:00 EDT 2024 · Scientific Reports · OSTI ID:2426324