Inferring solutions of differential equations using noisy multi-fidelity data

Abstract

For more than two centuries, solutions of differential equations have been obtained either analytically or numerically based on typically well-behaved forcing and boundary conditions for well-posed problems. We are changing this paradigm in a fundamental way by establishing an interface between probabilistic machine learning and differential equations. We develop data-driven algorithms for general linear equations using Gaussian process priors tailored to the corresponding integro-differential operators. The only observables are scarce noisy multi-fidelity data for the forcing and solution that are not required to reside on the domain boundary. The resulting predictive posterior distributions quantify uncertainty and naturally lead to adaptive solution refinement via active learning. This general framework circumvents the tyranny of numerical discretization as well as the consistency and stability issues of time-integration, and is scalable to high-dimensions.

arXiv comment19 pages, 3 figures

References in corpus (6)

• Gaussian Processes for Big Data
• Active Learning with Statistical Models
• Probabilistic Numerics and Uncertainty in Computations
• Generalized Fokker-Planck equation: Derivation and exact solutions
• Active Uncertainty Calibration in Bayesian ODE Solvers
• Probabilistic solvers for partial differential equations

Cited by in corpus (60)

• Hidden Physics Models: Machine Learning of Nonlinear Partial Differential Equations
• B-PINNs: Bayesian Physics-Informed Neural Networks for Forward and Inverse PDE Problems with Noisy Data
• Adaptive activation functions accelerate convergence in deep and physics-informed neural networks
• A composite neural network that learns from multi-fidelity data: Application to function approximation and inverse PDE problems
• Integrating Machine Learning and Multiscale Modeling: Perspectives, Challenges, and Opportunities in the Biological, Biomedical, and Behavioral Sciences
• Machine Learning of Linear Differential Equations using Gaussian Processes
• Adversarial Uncertainty Quantification in Physics-Informed Neural Networks
• Deep Learning of Vortex Induced Vibrations
• Model-Free Data-Driven Inelasticity
• A Physics Informed Neural Network for Time-Dependent Nonlinear and Higher Order Partial Differential Equations
• Multi-fidelity Bayesian Neural Networks: Algorithms and Applications
• Data-driven rogue waves and parameter discovery in the defocusing NLS equation with a potential using the PINN deep learning
• Physics-informed Neural Networks for Solving Nonlinear Diffusivity and Biot's equations
• Multi-fidelity regression using artificial neural networks: efficient approximation of parameter-dependent output quantities
• Distributed physics informed neural network for data-efficient solution to partial differential equations
• Hamiltonian neural networks for solving equations of motion
• On Learning Hamiltonian Systems from Data
• Neural-net-induced Gaussian process regression for function approximation and PDE solution
• Gamblets for opening the complexity-bottleneck of implicit schemes for hyperbolic and parabolic ODEs/PDEs with rough coefficients
• Physics-informed neural network methods based on Miura transformations and discovery of new localized wave solutions
• Active Training of Physics-Informed Neural Networks to Aggregate and Interpolate Parametric Solutions to the Navier-Stokes Equations
• Physics-informed neural networks method in high-dimensional integrable systems
• Bayesian Optimization with Output-Weighted Optimal Sampling
• A Deep Learning Approach for Predicting Two-dimensional Soil Consolidation Using Physics-Informed Neural Networks (PINN)
• Physical Symmetries Embedded in Neural Networks
• Symplectic Neural Networks in Taylor Series Form for Hamiltonian Systems
• Constructing coarse-scale bifurcation diagrams from spatio-temporal observations of microscopic simulations: A parsimonious machine learning approach
• Extraction of Material Properties through Multi-fidelity Deep Learning from Molecular Dynamics Simulation
• A physics-aware, probabilistic machine learning framework for coarse-graining high-dimensional systems in the Small Data regime
• A DeepONet multi-fidelity approach for residual learning in reduced order modeling
• A probabilistic generative model for semi-supervised training of coarse-grained surrogates and enforcing physical constraints through virtual observables
• Learning the constitutive relation of polymeric flows with memory
• Numerical modeling and simulation on nanosecond laser-target interactions
• A hybrid data driven-physics constrained Gaussian process regression framework with deep kernel for uncertainty quantification
• Conditional Karhunen-Loève expansion for uncertainty quantification and active learning in partial differential equation models
• Symplectic Gaussian Process Regression of Hamiltonian Flow Maps
• A probabilistic finite element method based on random meshes: Error estimators and Bayesian inverse problems
• Deep Hamiltonian networks based on symplectic integrators
• Modelling of physical systems with a Hopf bifurcation using mechanistic models and machine learning
• Constrained Hamiltonian Systems and Physics-Informed Neural Networks: Hamilton-Dirac Neural Networks
• Enforcing Boundary Conditions on Physical Fields in Bayesian Inversion
• Multi-fidelity data fusion through parameter space reduction with applications to automotive engineering
• Physically Consistent Learning of Conservative Lagrangian Systems with Gaussian Processes
• Physics-informed Neural Networks for Solving Inverse Problems of Nonlinear Biot's Equations: Batch Training
• Temporal Normalizing Flows
• Learning the Tangent Space of Dynamical Instabilities from Data
• Multi-fidelity data fusion for the approximation of scalar functions with low intrinsic dimensionality through active subspaces
• Towards information optimal simulation of partial differential equations
• Quantum Extremal Learning
• A Generalized Probabilistic Learning Approach for Multi-Fidelity Uncertainty Propagation in Complex Physical Simulations
• Decomposing Temperature Time Series with Non-Negative Matrix Factorization
• Physics-informed Gaussian process model for Euler-Bernoulli beam elements
• Neutrino Interactions with perturbed Rastall Gravity: A Novel Approach to Reducing the Hubble Tension
• Bayesian Reasoning for Physics Informed Neural Networks
• Towards a Machine Learning Solution for Hubble Tension: Physics-Informed Neural Network (PINN) Analysis of Tsallis Holographic Dark Energy in Presence of Neutrinos
• Active learning with physics-informed neural networks for optimal sensor placement in deep tunneling through transversely isotropic elastic rocks
• Sampling Methods for Bayesian Inference Involving Convergent Noisy Approximations of Forward Maps
• On the performance of multi-fidelity and reduced-dimensional neural emulators for inference of physiological boundary conditions
• Efficient Characterization of Dynamic Response Variation Using Multi-Fidelity Data Fusion through Composite Neural Network
• Physics-Coupled Spatio-Temporal Active Learning for Dynamical Systems