English

Deep least-squares methods: an unsupervised learning-based numerical method for solving elliptic PDEs

Machine Learning 2020-08-26 v3 Numerical Analysis Numerical Analysis Computational Physics Machine Learning

Abstract

This paper studies an unsupervised deep learning-based numerical approach for solving partial differential equations (PDEs). The approach makes use of the deep neural network to approximate solutions of PDEs through the compositional construction and employs least-squares functionals as loss functions to determine parameters of the deep neural network. There are various least-squares functionals for a partial differential equation. This paper focuses on the so-called first-order system least-squares (FOSLS) functional studied in [3], which is based on a first-order system of scalar second-order elliptic PDEs. Numerical results for second-order elliptic PDEs in one dimension are presented.

Keywords

Cite

@article{arxiv.1911.02109,
  title  = {Deep least-squares methods: an unsupervised learning-based numerical method for solving elliptic PDEs},
  author = {Zhiqiang Cai and Jingshuang Chen and Min Liu and Xinyu Liu},
  journal= {arXiv preprint arXiv:1911.02109},
  year   = {2020}
}

Comments

15 pages, 6 figures, 5 tables, accepted by Journal of Computational Physics

R2 v1 2026-06-23T12:06:49.055Z