Solving hyperbolic conservation laws with characteristic based neural network

Chuyu Liao, Ye Li, Zhongyi HuangOpenReview 2025 · Submission to ICLR 2025 · Withdrawn · OpenReview

Its authors withdrew this paper, so there is no full text to read or summarize.

Abstract

Neural network PDE solvers have recently gained popularity.However, it faces difficulty to deal with sharp discontinuity like shock waves in hyperbolic conservation laws.In this paper we propose a characteristic-based neural network to solve one dimension hyperbolic laws.The smooth solution can be derived by equation of characteristic lines ,and shock waves are decided by simple ODE solver.This method achieves a high accuracy with high efficiency. In the future it is hopeful to apply this method to higher dimension problems.
BibTeX
@misc{liao2025solving,
  author = {Chuyu Liao and Ye Li and Zhongyi Huang},
  title = {{Solving hyperbolic conservation laws with characteristic based neural network}},
  year = {2025},
  url = {https://openreview.net/forum?id=HDmmwwTIlf},
  howpublished = {OpenReview},
  note = {Submitted to ICLR 2025},
}

Loading related content...

Figure