A priori error estimate of deep mixed residual method for elliptic PDEs

DOI10.1007/S10915-023-02432-XMaRDI QIDQ6182315zbMATH OpenWikidataFDO

Authors Ling-Feng Li, Xue-Cheng Tai, Jiang Yang, Quanhui Zhu

Publication date 25 January 2024

Published in Journal of Scientific Computing (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2206.07474

zbMATH Keywords

deep learningmixed residual method

Mathematics Subject Classification ID

Error bounds for boundary value problems involving PDEs (65N15)

Abstract: In this work, we derive a priori error estimate of the mixed residual method when solving some elliptic PDEs. Our work is the first theoretical study of this method. We prove that the neural network solutions will converge if we increase the training samples and network size without any constraint on the ratio of training samples to the network size. Besides, our results suggest that the mixed residual method can recover high order derivatives better than the deep Ritz method, which has also been verified by our numerical experiments.

Review text

The paper presents a groundbreaking theoretical analysis of the deep mixed residual method (DMRM) in the context of solving elliptic partial differential equations (PDEs). DMRM, a novel deep learning-based approach, has demonstrated remarkable efficiency and accuracy in previous empirical studies, yet its theoretical foundations remained unexplored until now. The authors embark on a rigorous examination of DMRM's convergence properties, offering the first comprehensive a priori error estimate for this method. Their analysis unveils a significant finding: the convergence of neural network solutions is ensured with the augmentation of training samples and network size, without imposing any constraints on their ratio. This pivotal insight not only solidifies the reliability of DMRM but also highlights its scalability, a crucial attribute in practical applications.

Recommendations

• Error analysis of deep Ritz methods for elliptic equations
• Analysis of deep Ritz methods for semilinear elliptic equations
• Inf-sup neural networks for high-dimensional elliptic PDE problems
• On the convergence of physics informed neural networks for linear second-order elliptic and parabolic type PDEs

Cites work

• A Local Deep Learning Method for Solving High Order Partial Differential Equations
• Approximation rates for neural networks with general activation functions
• Deep least-squares methods: an unsupervised learning-based numerical method for solving elliptic PDEs
• DeepXDE: a deep learning library for solving differential equations
• DGM: a deep learning algorithm for solving partial differential equations
• Elliptic problems in nonsmooth domains
• Estimates on the generalization error of physics-informed neural networks for approximating a class of inverse problems for PDEs
• Estimates on the generalization error of physics-informed neural networks for approximating PDEs
• First-Order System Least Squares for Second-Order Partial Differential Equations: Part I
• scientific article; zbMATH DE number 49190 (Why is no real title available?)
• MIM: a deep mixed residual method for solving high-order partial differential equations
• PDE-Net 2.0: learning PDEs from data with a numeric-symbolic hybrid deep network
• Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
• Smoothed particle hydrodynamics: theory and application to non-spherical stars
• Solving high-dimensional partial differential equations using deep learning
• Some applications of concentration inequalities to statistics
• The deep Ritz method: a deep learning-based numerical algorithm for solving variational problems
• Universal approximation bounds for superpositions of a sigmoidal function
• Weak adversarial networks for high-dimensional partial differential equations

Cited in

(5)

• MIM: a deep mixed residual method for solving high-order partial differential equations
• Error analysis of deep Ritz methods for elliptic equations
• Generalization error in the deep Ritz method with smooth activation functions
• Deep mixed residual method for solving PDE-constrained optimization problems
• Error analysis of the deep mixed residual method for high-order elliptic equations

This page was built for publication: A priori error estimate of deep mixed residual method for elliptic PDEs

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6182315)