• Corpus ID: 219708206

Learning continuous-time PDEs from sparse data with graph neural networks

  title={Learning continuous-time PDEs from sparse data with graph neural networks},
  author={Valerii Iakovlev and Markus Heinonen and Harri L{\"a}hdesm{\"a}ki},
The behavior of many dynamical systems follow complex, yet still unknown partial differential equations (PDEs). While several machine learning methods have been proposed to learn PDEs directly from data, previous methods are limited to discrete-time approximations or make the limiting assumption of the observations arriving at regular grids. We propose a general continuous-time differential model for dynamical systems whose governing equations are parameterized by message passing graph neural… 

Figures and Tables from this paper

Learning latent field dynamics of PDEs
A new approach to learning surrogate models for simulation of complex physical systems described by nonlinear partial differential equations that achieves top performance and competitive inference speed compared to baseline methods while operating with a 4-times more compact representation.
Learning time-dependent PDE solver using Message Passing Graph Neural Networks
A discretization and domain invariant neural network time-dependent PDE solver based on message passing graph neural nets (MPGNN) which is trained on a sample domain with different discretizations, although the underlying physics remains to be the same.
Learning the Dynamics of Physical Systems from Sparse Observations with Finite Element Networks
This work derives a transport variant of their model from the convection equation and shows that it improves the transfer performance to higher-resolution meshes on sea surface temperature and gas flow forecasting against baseline models representing a selection of spatio-temporal forecasting methods.
Reconstructing Nonlinear Dynamical Systems from Multi-Modal Time Series
A general framework for multi- modal data integration for the purpose of nonlinear DS reconstruction and the analysis of cross-modal relations is proposed, based on dynamically interpretable recurrent neural networks as general approximators of non linear DSs, coupled to sets of modality-specific decoder models from the class of generalized linear models.
Transformer for Partial Differential Equations' Operator Learning
This work presents an attention-based framework for data-driven operator learning, which is built upon self-attention, cross-att attention, and a set of point-wise multilayer perceptrons (MLPs), and thus it makes few assumptions on the sampling pattern of the input function or query locations.
Scalable algorithms for physics-informed neural and graph networks
Some of the prevailing trends in embedding physics into machine learning are reviewed, using physics-informed neural networks (PINNs) based primarily on feed-forward neural networks and automatic differentiation and graph neural networks based on graph exterior calculus to construct differential operators.
Improved Training of Physics-Informed Neural Networks with Model Ensembles
This paper proposes to expand the solution interval gradually to make the PINN converge to the correct solution and uses the ensemble agreement as the criterion for including new points for computing the loss derived from PDEs.
When Physics Meets Machine Learning: A Survey of Physics-Informed Machine Learning
An abundant number of recent works in physics-informed machine learning are surveyed and summarized from three aspects: motivations of PIML, physics knowledge in PIMl, and methods of physics knowledge integration in P IML.
Unravelled multilevel transformation networks for predicting sparsely observed spatio-temporal dynamics
This paper proposes a deep learning model that learns to predict unknown spatio-temporal dynamics using data from sparsely-distributed data sites using the radial basis function (RBF) collocation method which is often used for meshfree solution of partial differential equations.
Application of Lie Group-based Neural Network Method to Nonlinear Dynamical Systems
In this paper, a Lie group-based neural network method is proposed for solving initial value problems of non linear dynamics. Due to its single-layer structure (MLP), the approach is substantially


Physics-aware Difference Graph Networks for Sparsely-Observed Dynamics
A novel architecture named Physics-aware Difference Graph Networks (PA-DGN) is proposed that exploits neighboring information to learn finite differences inspired by physics equations and further leverages data-driven end-to-end learning to discover underlying dynamical relations between the spatial and temporal differences in given observations.
Modeling the Dynamics of PDE Systems with Physics-Constrained Deep Auto-Regressive Networks
DL-PDE: Deep-learning based data-driven discovery of partial differential equations from discrete and noisy data
The proposed DL-PDE method combines deep learning via neural networks and data-driven discovery of PDE via sparse regressions to discover the governing PDEs of underlying physical processes.
PDE-Net 2.0: Learning PDEs from Data with A Numeric-Symbolic Hybrid Deep Network
PDE-Net: Learning PDEs from Data
Numerical experiments show that the PDE-Net has the potential to uncover the hidden PDE of the observed dynamics, and predict the dynamical behavior for a relatively long time, even in a noisy environment.
Physics Informed Deep Learning (Part II): Data-driven Discovery of Nonlinear Partial Differential Equations
We introduce physics informed neural networks -- neural networks that are trained to solve supervised learning tasks while respecting any given law of physics described by general nonlinear partial
Physics Informed Deep Learning (Part I): Data-driven Solutions of Nonlinear Partial Differential Equations
This two part treatise introduces physics informed neural networks – neural networks that are trained to solve supervised learning tasks while respecting any given law of physics described by general nonlinear partial differential equations and demonstrates how these networks can be used to infer solutions topartial differential equations, and obtain physics-informed surrogate models that are fully differentiable with respect to all input coordinates and free parameters.
Solving parametric PDE problems with artificial neural networks
This work proposes using neural network to parameterise the physical quantity of interest as a function of input coefficients and demonstrates the simplicity and accuracy of the approach through notable examples of PDEs in engineering and physics.