Bayesian Hidden Physics Models: Uncertainty Quantification for Discovery of Nonlinear Partial Differential Operators from Data
arXiv:2006.04228
Abstract
What do data tell us about physics-and what don't they tell us? There has been a surge of interest in using machine learning models to discover governing physical laws such as differential equations from data, but current methods lack uncertainty quantification to communicate their credibility. This work addresses this shortcoming from a Bayesian perspective. We introduce a novel model comprising "leaf" modules that learn to represent distinct experiments' spatiotemporal functional data as neural networks and a single "root" module that expresses a nonparametric distribution over their governing nonlinear differential operator as a Gaussian process. Automatic differentiation is used to compute the required partial derivatives from the leaf functions as inputs to the root. Our approach quantifies the reliability of the learned physics in terms of a posterior distribution over operators and propagates this uncertainty to solutions of novel initial-boundary value problem instances. Numerical experiments demonstrate the method on several nonlinear PDEs.
16 pages, 8 figures
References in corpus (13)
- Adam: A Method for Stochastic Optimization
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- ADADELTA: An Adaptive Learning Rate Method
- NSFnets (Navier-Stokes Flow nets): Physics-informed neural networks for the incompressible Navier-Stokes 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
- Physics Informed Deep Learning (Part I): Data-driven Solutions of Nonlinear Partial Differential Equations
- Understanding and mitigating gradient pathologies in physics-informed neural networks
- Bayesian Inference and Learning in Gaussian Process State-Space Models with Particle MCMC
- Data-driven discovery of free-form governing differential equations
- Learning Symbolic Physics with Graph Networks
- Closed-form Inference and Prediction in Gaussian Process State-Space Models
- Non-Factorised Variational Inference in Dynamical Systems