Smooth Piecewise Cutting for Neural Operator to Handle Discontinuities and Sharp Transitions
Abstract
Neural operators have achieved strong performance in learning solution operators of partial differential equations (PDEs), but their inherently continuous representations struggle to capture discontinuities and sharp transitions. Existing approaches typically approximate such features within continuous function spaces, often requiring increased model capacity and high-resolution data. In this work, we propose Cut-DeepONet, a two-stage training framework that explicitly models discontinuities while reducing learning complexity. Our approach reformulates the problem via a lifting strategy, partitioning the domain into smooth subregions while representing discontinuities as boundaries in a higher-dimensional space. This separation aligns the operator learning task with the inductive bias of neural networks and avoids directly approximating discontinuities. An additional network predicts input-dependent discontinuity locations for unseen inputs, which are then used to guide the neural operator in generating smooth components within each region. Experiments on benchmark PDEs show that Cut-DeepONet outperforms state-of-the-art methods, even when trained on low-resolution datasets. The method excels on problems with discontinuities and sharp transitions, while using fewer trainable parameters. Our results highlight the benefits of changing the representation of operator learning rather than increasing model complexity.
Related
- Finite volume quotients of bounded domains with piecewise smooth boundary
- Accelerated Prox-Level Methods for Unknown Piecewise-Smooth Optimization II: Function-constrained Optimization
- Finite element approximation on piecewise smooth domains
- Differentiability of Shape Function for Piecewise Smooth Lagrangians in Poissonian Random Environments
- Normal form method of center-focus problem in piecewise-smooth systems and algorithm design
BibTeX
@misc{dang2026smooth,
title = {{Smooth Piecewise Cutting for Neural Operator to Handle Discontinuities and Sharp Transitions}},
author = {Ha Dang and Sebastian Schmidt and Juergen Hesser},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2605.19823},
eprint = {2605.19823},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.19823},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Smooth Piecewise Cutting for Neural Operator to Handle Discontinuities and Sharp Transitions
AU - Ha Dang
AU - Sebastian Schmidt
AU - Juergen Hesser
PY - 2026
DA - 2026/09/25
PB - arXiv
DO - 10.48550/arxiv.2605.19823
UR - https://arxiv.org/abs/2605.19823
AB - Neural operators have achieved strong performance in learning solution operators of partial differential equations (PDEs), but their inherently continuous representations struggle to capture discontinuities and sharp transitions. Existing approaches typically approximate such features within continuous function spaces, often requiring increased model capacity and high-resolution data. In this work, we propose Cut-DeepONet, a two-stage training framework that explicitly models discontinuities while reducing learning complexity. Our approach reformulates the problem via a lifting strategy, partitioning the domain into smooth subregions while representing discontinuities as boundaries in a higher-dimensional space. This separation aligns the operator learning task with the inductive bias of neural networks and avoids directly approximating discontinuities. An additional network predicts input-dependent discontinuity locations for unseen inputs, which are then used to guide the neural operator in generating smooth components within each region. Experiments on benchmark PDEs show that Cut-DeepONet outperforms state-of-the-art methods, even when trained on low-resolution datasets. The method excels on problems with discontinuities and sharp transitions, while using fewer trainable parameters. Our results highlight the benefits of changing the representation of operator learning rather than increasing model complexity.
ER -