Year: 2021
Author: Liyao Lyu, Keke Wu, Rui Du, Jingrun Chen
CSIAM Transactions on Applied Mathematics, Vol. 2 (2021), Iss. 4 : pp. 748–775
Abstract
Boundary and initial conditions are important for the well-posedness of partial differential equations (PDEs). Numerically, these conditions can be enforced exactly in classical numerical methods, such as finite difference method and finite element method. Recent years, we have witnessed growing interests in solving PDEs by deep neural networks (DNNs), especially in the high-dimensional case. However, in the generic situation, a careful literature review shows that boundary conditions cannot be enforced exactly for DNNs, which inevitably leads to a modeling error. In this work, based on the recently developed deep mixed residual method (MIM), we demonstrate how to make DNNs satisfy boundary and initial conditions automatically by using distance functions and explicit constructions. As a consequence, the loss function in MIM is free of the penalty term and does not have any modeling error. Using numerous examples, including Dirichlet, Neumann, mixed, Robin, and periodic boundary conditions for elliptic equations, and initial conditions for parabolic and hyperbolic equations, we show that enforcing exact boundary and initial conditions not only provides a better approximate solution but also facilitates the training process.
You do not have full access to this article.
Already a Subscriber? Sign in as an individual or via your institution
Journal Article Details
Publisher Name: Global Science Press
Language: English
DOI: https://doi.org/10.4208/csiam-am.SO-2021-0011
CSIAM Transactions on Applied Mathematics, Vol. 2 (2021), Iss. 4 : pp. 748–775
Published online: 2021-01
AMS Subject Headings: Global Science Press
Copyright: COPYRIGHT: © Global Science Press
Pages: 28
Keywords: Machine learning deep neural networks enforcement of boundary/initial conditions penalty.
Author Details
-
Deep mixed residual method for solving PDE-constrained optimization problems
Yong, Jinjun | Luo, Xianbing | Sun, Shuyu | Ye, ChanglunComputers & Mathematics with Applications, Vol. 176 (2024), Iss. P.510
https://doi.org/10.1016/j.camwa.2024.11.009 [Citations: 0] -
Lagrangian dual framework for conservative neural network solutions of kinetic equations
Hwang, Hyung Ju | Son, HwijaeKinetic and Related Models, Vol. 15 (2022), Iss. 4 P.551
https://doi.org/10.3934/krm.2021046 [Citations: 1] -
Discovering physics-informed neural networks model for solving partial differential equations through evolutionary computation
Zhang, Bo | Yang, ChaoSwarm and Evolutionary Computation, Vol. 88 (2024), Iss. P.101589
https://doi.org/10.1016/j.swevo.2024.101589 [Citations: 0] -
Asymptotic-Preserving Neural Networks for Multiscale Time-Dependent Linear Transport Equations
Jin, Shi | Ma, Zheng | Wu, KekeJournal of Scientific Computing, Vol. 94 (2023), Iss. 3
https://doi.org/10.1007/s10915-023-02100-0 [Citations: 13] -
MIM: A deep mixed residual method for solving high-order partial differential equations
Lyu, Liyao | Zhang, Zhen | Chen, Minxin | Chen, JingrunJournal of Computational Physics, Vol. 452 (2022), Iss. P.110930
https://doi.org/10.1016/j.jcp.2021.110930 [Citations: 39] -
Neural network guided adjoint computations in dual weighted residual error estimation
Roth, Julian | Schröder, Max | Wick, ThomasSN Applied Sciences, Vol. 4 (2022), Iss. 2
https://doi.org/10.1007/s42452-022-04938-9 [Citations: 5] -
A Priori Error Estimate of Deep Mixed Residual Method for Elliptic PDEs
Li, Lingfeng | Tai, Xue-Cheng | Yang, Jiang | Zhu, QuanhuiJournal of Scientific Computing, Vol. 98 (2024), Iss. 2
https://doi.org/10.1007/s10915-023-02432-x [Citations: 2] -
A physics-informed neural network based simulation tool for reacting flow with multicomponent reactants
Sun, Zewei | Du, Honghan | Miao, Chunfu | Hou, QingzhiAdvances in Engineering Software, Vol. 185 (2023), Iss. P.103525
https://doi.org/10.1016/j.advengsoft.2023.103525 [Citations: 5]