Convolutional Neural Operators for robust and accurate learning of PDEs (2023-02-02T00:00:00.000000Z)

TL;DR

Novel adaptations for convolutional neural networks are presented to demonstrate that they are indeed able to process functions as inputs and outputs, and it is proved a universality theorem to show that CNOs can approximate operators arising in PDEs to desired accuracy.

Abstract

Although very successfully used in conventional machine learning, convolution based neural network architectures -- believed to be inconsistent in function space -- have been largely ignored in the context of learning solution operators of PDEs. Here, we present novel adaptations for convolutional neural networks to demonstrate that they are indeed able to process functions as inputs and outputs. The resulting architecture, termed as convolutional neural operators (CNOs), is designed specifically to preserve its underlying continuous nature, even when implemented in a discretized form on a computer. We prove a universality theorem to show that CNOs can approximate operators arising in PDEs to desired accuracy. CNOs are tested on a novel suite of benchmarks, encompassing a diverse set of PDEs with possibly multi-scale solutions and are observed to significantly outperform baselines, paving the way for an alternative framework for robust and accurate operator learning. Our code is publicly available at https://github.com/bogdanraonic3/ConvolutionalNeuralOperator

Authors

Bogdan Raoni'c

1 papers

R. Molinaro

1 papers

T. D. Ryck

1 papers

TL;DR

Abstract

Authors

References68 items

Neural Inverse Operators for Solving PDE Inverse Problems

Nonlinear Reconstruction for Operator Learning of PDEs with Discontinuities

Towards Multi-spatiotemporal-scale Generalized PDE Modeling

Fourier Neural Operator with Learned Deformations for PDEs on General Geometries

NOMAD: Nonlinear Manifold Decoders for Operator Learning

Generic bounds on the approximation error for physics-informed (and) operator learning

Variable-Input Deep Operator Networks

Spectral Neural Operators

FourCastNet: A Global Data-driven High-resolution Weather Model using Adaptive Fourier Neural Operators

A ConvNet for the 2020s

Learning Operators with Coupled Attention

Factorized Fourier Neural Operators

Physics-Informed Neural Operator for Learning Partial Differential Equations

ResNet strikes back: An improved training procedure in timm

Neural Operator: Learning Maps Between Function Spaces

On universal approximation and error bounds for Fourier Neural Operators

Alias-Free Generative Adversarial Networks

Choose a Transformer: Fourier or Galerkin

Physics-informed machine learning

On the approximation of functions by tanh neural networks

Learning the solution operator of parametric partial differential equations with physics-informed DeepONets

Adaptive Rational Activations to Boost Deep Reinforcement Learning

Error estimates for DeepOnets: A deep learning framework in infinite dimensions

DeepM&Mnet for hypersonics: Predicting the coupled flow and finite-rate chemistry behind a normal shock using neural-network approximation of operators

Fourier Neural Operator for Parametric Partial Differential Equations

DeepM&Mnet: Inferring the electroconvection multiphysics fields based on operator approximation by neural networks

Iterative Surrogate Model Optimization (ISMO): An active learning algorithm for PDE constrained optimization with deep neural networks

Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains

Multipole Graph Neural Operator for Parametric Partial Differential Equations

Model Reduction and Neural Networks for Parametric PDEs

Rational neural networks

Neural Operator: Graph Kernel Network for Partial Differential Equations

Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators

Statistical solutions of the incompressible Euler equations

Padé Activation Units: End-to-end Learning of Flexible Activation Functions in Deep Networks

Learning Dynamical Systems from Partial Observations

Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations

Equivalence of approximation by convolutional neural networks and fully-connected networks

Deep Parametric Continuous Convolutional Neural Networks

Bayesian Deep Convolutional Encoder-Decoder Networks for Surrogate Modeling and Uncertainty Quantification

PDE-Net: Learning PDEs from Data

Neural Networks and Rational Functions

Stable architectures for deep neural networks

Image-to-Image Translation with Conditional Adversarial Networks

A Geometric Comparison of Aerofoil Shape Parameterisation Methods

Deep Residual Learning for Image Recognition

U-Net: Convolutional Networks for Biomedical Image Segmentation

Foundations of Signal Processing

Numerical Approximation of Partial Differential Equations

SPECTRAL METHODS FOR TIME-DEPENDENT PROBLEMS.

Burgers' Equation with Vanishing Hyper-Viscosity

Sampling-50 years after Shannon

Universal approximation to nonlinear operators by neural networks with arbitrary activation functions and its application to dynamical systems

Sinc interpolation of discrete periodic signals

Original Contribution: Multilayer feedforward networks with a nonpolynomial activation function can approximate any function

Spectral methods for time dependent problems

A second-order projection method for the incompressible navier-stokes equations

Convergence of spectral methods for nonlinear conservation laws. Final report

Are Neural Operators Really Neural Operators? Frame Theory Meets Operator Learning

Aliasing-Free Nonlinear Signal Processing Using Implicitly Defined Functions

UNIFORM L p -BOUND OF THE ALLEN–CAHN EQUATION AND ITS NUMERICAL DISCRETIZATION

Deep Learning

Partial Differential Equations

Introduction to Inverse Problems in Imaging

Gradient-based learning applied to document recognition

Numerical Approximation of Partial Differential Equations.

and as an in

multi-spatiotemporal-scale

Field of Study

Venue Information

Name

Type

URL

Alternate Names