Trainable Spline Representations for Physics-Informed Learning

DOI10.48550/arXiv.2607.15751MaRDI QIDQ7284400arXivFDO

Authors Giovanni Canali, Nicola Demo, Gianluigi Rozza

Publication date 17 July 2026

Tldr The paper introduces Physics-Informed Splines (PI-Splines), a spline-based architecture for physics-informed learning, which represents the solution of differential equations through tensor-product B-spline expansions with trainable control coefficients while preserving residual-based training paradigm, compact support, explicit smoothness control, and analytical derivatives.

Full work available at URL https://doi.org/10.48550/arXiv.2607.15751

arXiv classification

cs.LG math.NA

Cites work

An Expert's Guide to Training Physics-informed Neural Networks
Can physics-informed neural networks beat the finite element method?
Complex physics-informed neural network
Constraint-driven Optimization and Parametrization of Industrial NURBS Geometries via Neural Deformation Field
Deep NURBS -- Admissible Physics-informed Neural Networks
Extended Physics-Informed Neural Networks (XPINNs): A Generalized Space-Time Domain Decomposition Based Deep Learning Framework for Nonlinear Partial Differential Equations
On the eigenvector bias of Fourier feature networks: from regression to solving multi-scale PDEs with physics-informed neural networks
Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
PirateNets: physics-informed deep learning with residual adaptive networks
Residual-based attention in physics-informed neural networks
Understanding and Mitigating Gradient Flow Pathologies in Physics-Informed Neural Networks
When and why PINNs fail to train: A neural tangent kernel perspective

This page was built for publication: Trainable Spline Representations for Physics-Informed Learning

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7284400)