EUVIMEDEUVIMED entdecken
Crossref

On Elliptic and Parabolic Interface Problems with a Nonlinear Jump Condition: A Deep Learning Approach

Yu-Hau Tseng, Ting-Syu Chou

Taiwanese Journal of Mathematics · 01.01.2026 · -1 (-1) · 1-22

DOI: 10.11650/tjm/261002 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Taiwanese Journal of Mathematics, erschienen am 01.01.2026. Verfasst von Yu-Hau Tseng und 1 weiteren Autor:innen. Die Arbeit nennt 25 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

We apply a discontinuity-capturing physics-informed neural network for parabolic interface problems with a nonlinear jump condition. An augmented neural network representation, via an indicator variable that distinguishes subdomains, expresses the discontinuous solution as the restriction of a smooth extension defined on a higher-dimensional space. This construction enables the use of standard smooth neural networks while explicitly preserving discontinuities across the interface. We incorporate the governing partial differential equations, boundary and initial conditions, nonlinear jump condition, and flux continuity constraint into a unified loss functional. The framework remains fully mesh-free and requires no body-fitted discretization. A Levenberg–Marquardt optimization strategy is employed for training. Numerical experiments in one- and two-dimensional spaces demonstrate that the proposed approach accurately captures nonlinear jump behavior and achieves relative $L^{\infty}$ errors down to $O(10^{-7})$, which is comparable to fitted finite difference schemes in accuracy while providing greater flexibility for PDEs defined on complex geometries.

Zitieren

APA

Tseng, Y., & Chou, T. (2026). On Elliptic and Parabolic Interface Problems with a Nonlinear Jump Condition: A Deep Learning Approach. Taiwanese Journal of Mathematics, -1(-1), 1-22. https://doi.org/10.11650/tjm/261002

Vancouver

Tseng Y, Chou T. On Elliptic and Parabolic Interface Problems with a Nonlinear Jump Condition: A Deep Learning Approach. Taiwanese Journal of Mathematics. 2026;-1(-1):1-22. doi:10.11650/tjm/261002

BibTeX
@article{Tseng2026,
  author = {Tseng, Yu-Hau and Chou, Ting-Syu},
  title = {On Elliptic and Parabolic Interface Problems with a Nonlinear Jump Condition: A Deep Learning Approach},
  journal = {Taiwanese Journal of Mathematics},
  year = {2026},
  volume = {-1},
  number = {-1},
  pages = {1--22},
  doi = {10.11650/tjm/261002}
}
Kostenlos starten →