Residual-based Adaptive Refinement of hp-Variational PINNs for elliptic PDEs with geometric singularities and localized features
Abstract
Physics-Informed Neural Networks approximate the solution of boundary value problems by minimizing the strong-form residual, their accuracy is often limited near corner and crack singularities, sharp layers and localized features and the enforcement of Neumann boundary conditions remains challenging. In hp-Variational Physics-Informed Neural Networks, the problem is formulated in a Petrov-Galerkin weak form, reducing the differentiation order applied to the neural network, while boundary constraints are imposed through penalty terms. The variational residual is evaluated using fixed Gauss quadrature points and the strong-form residual on a fixed collocation set, which may fail to capture regions of rapid variation. To address this issue, we integrate hp-Variational Physics-Informed Neural Networks with Residual-based Adaptive Refinement. During training, the strong-form residual is computed on a dense candidate pool and points with the largest residuals are added adaptively to the collocation set. Since the weak formulation is defined in H1(Ω), it avoids the H2(Ω) regularity requirement that fails on cracked and non-convex domains. The proposed method is validated on four two-dimensional elliptic problems and consistently achieves substantial reductions in relative L2, absolute L∞ and relative H1 semi-norm errors through adaptive residual-driven point placement.
Related
- Residual-Based Adaptive Refinement in Hybrid Variational-Collocation PINNs for Singular Elliptic PDEs with Mixed Dirichlet-Neumann Boundary Conditions
- INI-VPINN: A Variational Physics-Informed Neural Network with Implicit Neumann and Interface Handling for Multi-Material Domains with Geometric Singularities
- Residual Calibration via Local Feature-Space Refinement
- The Yau Geometric Conjecture for Isolated Hypersurface Singularities
- Limit theory for Lipschitz-localized statistics in random geometric models
BibTeX
@misc{giri2026residualbased,
title = {{Residual-based Adaptive Refinement of hp-Variational PINNs for elliptic PDEs with geometric singularities and localized features}},
author = {Somnath Giri and Muslim Malik and Rishikesh Yadav},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.37670},
eprint = {2610.37670},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.37670},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - Residual-based Adaptive Refinement of hp-Variational PINNs for elliptic PDEs with geometric singularities and localized features
AU - Somnath Giri
AU - Muslim Malik
AU - Rishikesh Yadav
PY - 2026
DA - 2026/10/01
PB - arXiv
DO - 10.48550/arxiv.2610.37670
UR - https://arxiv.org/abs/2610.37670
AB - Physics-Informed Neural Networks approximate the solution of boundary value problems by minimizing the strong-form residual, their accuracy is often limited near corner and crack singularities, sharp layers and localized features and the enforcement of Neumann boundary conditions remains challenging. In hp-Variational Physics-Informed Neural Networks, the problem is formulated in a Petrov-Galerkin weak form, reducing the differentiation order applied to the neural network, while boundary constraints are imposed through penalty terms. The variational residual is evaluated using fixed Gauss quadrature points and the strong-form residual on a fixed collocation set, which may fail to capture regions of rapid variation. To address this issue, we integrate hp-Variational Physics-Informed Neural Networks with Residual-based Adaptive Refinement. During training, the strong-form residual is computed on a dense candidate pool and points with the largest residuals are added adaptively to the collocation set. Since the weak formulation is defined in H^1(Ω), it avoids the H^2(Ω) regularity requirement that fails on cracked and non-convex domains. The proposed method is validated on four two-dimensional elliptic problems and consistently achieves substantial reductions in relative L^2, absolute L^∞ and relative H^1 semi-norm errors through adaptive residual-driven point placement.
ER -