Close Search Advanced
Journals
Resources
About Us
Open Access

Learning to Discretize: Solving 1D Scalar Conservation Laws via Deep Reinforcement Learning

Learning to Discretize: Solving 1D Scalar Conservation Laws via Deep Reinforcement Learning
Preview
Add to basket

Year:    2020

Author:    Yufei Wang, Ziju Shen, Zichao Long, Bin Dong

Communications in Computational Physics, Vol. 28 (2020), Iss. 5 : pp. 2158–2179

Abstract

Conservation laws are considered to be fundamental laws of nature. It has broad applications in many fields, including physics, chemistry, biology, geology, and engineering. Solving the differential equations associated with conservation laws is a major branch in computational mathematics. The recent success of machine learning, especially deep learning in areas such as computer vision and natural language processing, has attracted a lot of attention from the community of computational mathematics and inspired many intriguing works in combining machine learning with traditional methods. In this paper, we are the first to view numerical PDE solvers as an MDP and to use (deep) RL to learn new solvers. As proof of concept, we focus on 1-dimensional scalar conservation laws. We deploy the machinery of deep reinforcement learning to train a policy network that can decide on how the numerical solutions should be approximated in a sequential and spatial-temporal adaptive manner. We will show that the problem of solving conservation laws can be naturally viewed as a sequential decision-making process, and the numerical schemes learned in such a way can easily enforce long-term accuracy. Furthermore, the learned policy network is carefully designed to determine a good local discrete approximation based on the current state of the solution, which essentially makes the proposed method a meta-learning approach. In other words, the proposed method is capable of learning how to discretize for a given situation mimicking human experts. Finally, we will provide details on how the policy network is trained, how well it performs compared with some state-of-the-art numerical solvers such as WENO schemes, and supervised learning based approach L3D and PINN, and how well it generalizes.

Submit Article

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/cicp.OA-2020-0194

Communications in Computational Physics, Vol. 28 (2020), Iss. 5 : pp. 2158–2179

Published online:    2020-01

AMS Subject Headings:    Global Science Press

Copyright:    COPYRIGHT: © Global Science Press

Pages:    22

Keywords:    Conservation laws deep reinforcement learning finite difference approximation WENO.

Author Details

Yufei Wang

Ziju Shen

Zichao Long

Bin Dong

  1. Deep smoothness weighted essentially non-oscillatory method for two-dimensional hyperbolic conservation laws: A deep learning approach for learning smoothness indicators

    Kossaczká, Tatiana | Jagtap, Ameya D. | Ehrhardt, Matthias

    Physics of Fluids, Vol. 36 (2024), Iss. 3

    https://doi.org/10.1063/5.0199322 [Citations: 1]

  2. Error analysis of deep Ritz methods for elliptic equations

    Jiao, Yuling | Lai, Yanming | Lo, Yisu | Wang, Yang | Yang, Yunfei

    Analysis and Applications, Vol. 22 (2024), Iss. 01 P.57

    https://doi.org/10.1142/S021953052350015X [Citations: 4]

  3. Mastering the Cahn–Hilliard equation and Camassa–Holm equation with cell-average-based neural network method

    Zhou, Xiaofang | Qiu, Changxin | Yan, Wenjing | Li, Biao

    Nonlinear Dynamics, Vol. 111 (2023), Iss. 5 P.4823

    https://doi.org/10.1007/s11071-022-08088-w [Citations: 4]

  4. ConvStabNet: a CNN-based approach for the prediction of local stabilization parameter for SUPG scheme

    Yadav, Sangeeta | Ganesan, Sashikumaar

    Calcolo, Vol. 61 (2024), Iss. 3

    https://doi.org/10.1007/s10092-024-00597-x [Citations: 0]
  5. Advances in Practical Applications of Agents, Multi-Agent Systems, and Complex Systems Simulation. The PAAMS Collection

    Multi-agent Learning of Numerical Methods for Hyperbolic PDEs with Factored Dec-MDP

    Fu, Yiwei | Kapilavai, Dheeraj S. K. | Way, Elliot

    2022

    https://doi.org/10.1007/978-3-031-18192-4_15 [Citations: 0]

  6. Enhanced fifth order WENO shock-capturing schemes with deep learning

    Kossaczká, Tatiana | Ehrhardt, Matthias | Günther, Michael

    Results in Applied Mathematics, Vol. 12 (2021), Iss. P.100201

    https://doi.org/10.1016/j.rinam.2021.100201 [Citations: 13]

  7. Machine learning-based WENO5 scheme

    Nogueira, Xesús | Fernández-Fidalgo, Javier | Ramos, Lucía | Couceiro, Iván | Ramírez, Luis

    Computers & Mathematics with Applications, Vol. 168 (2024), Iss. P.84

    https://doi.org/10.1016/j.camwa.2024.05.031 [Citations: 0]

  8. Cell-average based neural network method for third order and fifth order KdV type equations

    Chen, Yongsheng | Yan, Jue | Zhong, Xinghui

    Frontiers in Applied Mathematics and Statistics, Vol. 8 (2022), Iss.

    https://doi.org/10.3389/fams.2022.1021069 [Citations: 2]

  9. Learning the flux and diffusion function for degenerate convection-diffusion equations using different types of observations

    Li, Qing | Evje, Steinar

    BIT Numerical Mathematics, Vol. 64 (2024), Iss. 2

    https://doi.org/10.1007/s10543-024-01018-9 [Citations: 0]

  10. Eno classification and regression neural networks for numerical approximation of discontinuous flow problems

    Jayswal, Vikas Kumar | Pandey, Prashant Kumar | Dubey, Ritesh Kumar

    Soft Computing, Vol. 28 (2024), Iss. 20 P.11787

    https://doi.org/10.1007/s00500-024-09944-0 [Citations: 0]
  11. Traffic Congestion Control by PDE Backstepping

    Comparison of Backstepping with Reinforcement Learning

    Yu, Huan | Krstic, Miroslav

    2022

    https://doi.org/10.1007/978-3-031-19346-0_7 [Citations: 0]

  12. On the learning of high order polynomial reconstructions for essentially non-oscillatory schemes

    Kumar Jayswal, Vikas | Kumar Dubey, Ritesh

    Physica Scripta, Vol. 99 (2024), Iss. 11 P.116009

    https://doi.org/10.1088/1402-4896/ad7f97 [Citations: 0]

  13. Learning based numerical methods for acoustic frequency-domain simulation with high frequency

    Li, Tingyue | Chen, Yu | Miao, Yun | Ma, Dingjiong

    Engineering Analysis with Boundary Elements, Vol. 163 (2024), Iss. P.200

    https://doi.org/10.1016/j.enganabound.2024.03.009 [Citations: 1]

  14. Data-Driven Modeling of Partially Observed Biological Systems

    Su, Wei-Hung | Chou, Ching-Shan | Xiu, Dongbin

    Communications on Applied Mathematics and Computation, Vol. 6 (2024), Iss. 1 P.739

    https://doi.org/10.1007/s42967-023-00317-2 [Citations: 0]

  15. A learned conservative semi-Lagrangian finite volume scheme for transport simulations

    Chen, Yongsheng | Guo, Wei | Zhong, Xinghui

    Journal of Computational Physics, Vol. 490 (2023), Iss. P.112329

    https://doi.org/10.1016/j.jcp.2023.112329 [Citations: 0]

  16. Neural control of discrete weak formulations: Galerkin, least squares & minimal-residual methods with quasi-optimal weights

    Brevis, Ignacio | Muga, Ignacio | van der Zee, Kristoffer G.

    Computer Methods in Applied Mechanics and Engineering, Vol. 402 (2022), Iss. P.115716

    https://doi.org/10.1016/j.cma.2022.115716 [Citations: 6]