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

Gaussian processes meet NeuralODEs: a Bayesian framework for learning the dynamics of partially observed systems from scarce and noisy data

Journal Article · · Philosophical Transactions of the Royal Society. A, Mathematical, Physical and Engineering Sciences
 [1];  [1]
  1. Department of Mechanical Engineering, and Applied Mechanics, University of Pennsylvania, Philadelphia, PA 19104, USA

We present a machine learning framework (GP-NODE) for Bayesian model discovery from partial, noisy and irregular observations of nonlinear dynamical systems. The proposed method takes advantage of differentiable programming to propagate gradient information through ordinary differential equation solvers and perform Bayesian inference with respect to unknown model parameters using Hamiltonian Monte Carlo sampling and Gaussian Process priors over the observed system states. This allows us to exploit temporal correlations in the observed data, and efficiently infer posterior distributions over plausible models with quantified uncertainty. The use of the Finnish Horseshoe as a sparsity-promoting prior for free model parameters also enables the discovery of parsimonious representations for the latent dynamics. A series of numerical studies is presented to demonstrate the effectiveness of the proposed GP-NODE method including predator–prey systems, systems biology and a 50-dimensional human motion dynamical system. This article is part of the theme issue ‘Data-driven prediction in dynamical systems’.

Research Organization:
Raytheon Technologies Corp., Waltham, MA (United States); Univ. of Pennsylvania, Philadelphia, PA (United States)
Sponsoring Organization:
USDOE Advanced Research Projects Agency - Energy (ARPA-E)
DOE Contract Number:
AR0001201; SC0019116
OSTI ID:
1979597
Journal Information:
Philosophical Transactions of the Royal Society. A, Mathematical, Physical and Engineering Sciences, Vol. 380, Issue 2229; ISSN 1364-503X
Publisher:
The Royal Society Publishing
Country of Publication:
United States
Language:
English

References (20)

Discovering governing equations from data by sparse identification of nonlinear dynamical systems journal March 2016
Data-driven discovery of partial differential equations journal April 2017
Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations journal February 2019
Adversarial uncertainty quantification in physics-informed neural networks journal October 2019
Physics-Informed Generative Adversarial Networks for Stochastic Differential Equations journal January 2020
Hidden physics models: Machine learning of nonlinear partial differential equations journal March 2018
Machine learning in cardiovascular flows modeling: Predicting arterial blood pressure from non-invasive 4D flow MRI data using physics-informed neural networks journal January 2020
Bayesian differential programming for robust systems identification under uncertainty
  • Yang, Yibo; Aziz Bhouri, Mohamed; Perdikaris, Paris
  • Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, Vol. 476, Issue 2243 https://doi.org/10.1098/rspa.2020.0290
journal November 2020
Kernel methods in system identification, machine learning and function estimation: A survey journal March 2014
Learning dynamical systems from data: A simple cross-validation perspective, part I: Parametric kernel flows journal July 2021
Bayesian Interpolation journal May 1992
Sparsity information and regularization in the horseshoe and other shrinkage priors journal January 2017
MCMC Using Hamiltonian Dynamics book May 2011
Mathematical Theory of Optimal Processes book May 2018
Simple chemical reaction systems with limit cycle behaviour journal December 1979
Systems biology informed deep learning for inferring parameters and hidden dynamics journal November 2020
Data-driven discovery of coordinates and governing equations journal October 2019
Data driven governing equations approximation using deep neural networks journal October 2019
Temperature dependency and temperature compensation in a model of yeast glycolytic oscillations journal November 2003
Gaussian Process Dynamical Models for Human Motion journal February 2008

Similar Records

Bayesian differential programming for robust systems identification under uncertainty
Journal Article · 2020 · Proceedings of the Royal Society. A. Mathematical, Physical and Engineering Sciences · OSTI ID:1852841

Sparsifying priors for Bayesian uncertainty quantification in model discovery
Journal Article · 2022 · Royal Society Open Science · OSTI ID:1861893

Physics makes the difference: Bayesian optimization and active learning via augmented Gaussian process
Journal Article · 2022 · Machine Learning: Science and Technology · OSTI ID:1840072