Abstract:We develop a new Bayesian framework based on deep neural networks to be able to extrapolate in space-time using historical data and to quantify uncertainties arising from both noisy and gappy data in physical problems. Specifically, the proposed approach has two stages: (1) prior learning and (2) posterior estimation. At the first stage, we employ the physics-informed Generative Adversarial Networks (PI-GAN) to learn a functional prior either from a prescribed function distribution, e.g., Gaussian process, or from historical data and physics. At the second stage, we employ the Hamiltonian Monte Carlo (HMC) method to estimate the posterior in the latent space of PI-GANs. In addition, we use two different approaches to encode the physics: (1) automatic differentiation, used in the physics-informed neural networks (PINNs) for scenarios with explicitly known partial differential equations (PDEs), and (2) operator regression using the deep operator network (DeepONet) for PDE-agnostic scenarios. We then test the proposed method for (1) meta-learning for one-dimensional regression, and forward/inverse PDE problems (combined with PINNs); (2) PDE-agnostic physical problems (combined with DeepONet), e.g., fractional diffusion as well as saturated stochastic (100-dimensional) flows in heterogeneous porous media; and (3) spatial-temporal regression problems, i.e., inference of a marine riser displacement field using experimental data from the Norwegian Deepwater Programme (NDP). The results demonstrate that the proposed approach can provide accurate predictions as well as uncertainty quantification given very limited scattered and noisy data, since historical data could be available to provide informative priors. In , the proposed method is capable of learning flexible functional priors, e.g., both Gaussian and non-Gaussian process, and can be readily extended to big data problems by enabling mini-batch training using stochastic HMC or normalizing flows since the latent space is generally characterized as low dimensional.

Predicting parametric spatiotemporal dynamics by multi-resolution PDE structure-preserved deep learning

Multi-resolution partial differential equations preserved learning framework for spatiotemporal dynamics

A Deep Learning Approach for Predicting Spatiotemporal Dynamics From Sparsely Observed Data

Discovering Nonlinear PDEs from Scarce Data with Physics-encoded Learning

Encoding physics to learn reaction–diffusion processes

Physics-Based Deep Learning for Flow Problems

PhyCRNet: Physics-informed convolutional-recurrent network for solving spatiotemporal PDEs

Learning functional priors and posteriors from data and physics

A Spatial-Temporally Adaptive PINN Framework for 3D Bi-Ventricular Electrophysiological Simulations and Parameter Inference

PARCv2: Physics-aware Recurrent Convolutional Neural Networks for Spatiotemporal Dynamics Modeling

Physics-guided machine learning from simulated data with different physical parameters

Physics informed deep learning for computational elastodynamics without labeled data

Advancing Generalization in PINNs through Latent-Space Representations

Spatio-Temporal Super-Resolution of Dynamical Systems using Physics-Informed Deep-Learning

PhyMPGN: Physics-encoded Message Passing Graph Network for spatiotemporal PDE systems

Physics-informed learning of governing equations from scarce data

Physics-embedded Fourier Neural Network for Partial Differential Equations

A Physics-Informed Recurrent Neural Network for Solving Time-Dependent Partial Differential Equations

Physics-Guided Discovery of Highly Nonlinear Parametric Partial Differential Equations

Enhanced Spatiotemporal Prediction Using Physical-guided And Frequency-enhanced Recurrent Neural Networks