IGA-ODIL: Optimizing DIscretre robust Loss with Isogeometric Analysis to solve forward and inverse problems faster using machine learning tools

Physics-informed neural networks (PINNs) formulate the solution of partial differential equations as residual minimization problems over neural network parameterizations. Although highly flexible, optimization of PINNs using modern variants of Stochastic Gradient Descent algorithms is expensive. On the other hand, iterative computation of PINN parameterization using the Gauss-Newton method suffers from convergence difficulties, dense Jacobian structures, and poor conditioning that limit the effectiveness of second-order optimization methods. In this work, we introduce IGA-ODIL [1], a spline-based residual minimization framework combining ideas from Optimizing DIscrete Loss (ODIL), robust variational residual minimization, and Isogeometric Analysis (IGA). Instead of neural-network parameterizations of PINNs, the unknown solution is represented by smooth B-spline basis functions, leading to sparse structured Jacobians and efficient Gauss--Newton optimization. We also derive robust residual formulations based on weighted Gram operators, making the loss function related with the true error. The resulting systems inherit locality, sparsity, and approximation-theoretic properties of classical finite element and isogeometric methods while preserving the residual-learning philosophy of scientific machine learning. The proposed methodology is evaluated on several benchmark problems, including Poisson equations, convection-dominated advection--diffusion equations, Helmholtz problems with highly oscillatory solutions, nonlinear Allen--Cahn equations, and inverse Helmholtz parameter identification. Numerical experiments demonstrate orders-of-magnitude speedups compared with PINNs and CRVPINNs while maintaining high accuracy and robustness.

[1] Maciej Paszynski, Tomasz Sluzalec, https://arxiv.org/abs/2605.30272

Maciej Paszynski is a Professor of Computer Science at AGH University of Krakow. He received his PhD in Mathematics with applications to Computer Science from Jagiellonian University in Krakow, Poland. He was a postdoctoral researcher under Leszek Demkowicz at the Oden Institute for Computational Engineering and Sciences and has remained a frequent visiting researcher there, completing more than 25 research visits since 2003. He serves as Workshops Chair of the International Conference on Computational Science and has recently been nominated as an Associate Editor of the Journal of Computational Science. In 2025, he was included in the Stanford University/Elsevier ranking of the world's top 2% of scientists (Main Field: Information & Communication Technologies; Subfield 1: Artificial Intelligence & Image Processing; Subfield 2: Applied Mathematics). Together with his research group, Adaptive Algorithms and Systems (A2S), he conducts research on adaptive finite element methods, isogeometric analysis, physics-informed machine learning, and scalable algorithms for high-performance computing. He has authored more than 150 peer-reviewed publications, including over 70 papers in Q1 and Q2 journals, and has delivered more than 100 presentations at international conferences. He collaborates with the Basque Center for Applied Mathematics (Spain), the Pontifical Catholic University of Valparaíso (Chile), SIMULA Research Laboratory (Norway), and the University Institute for Intelligent Systems and Numerical Applications in Engineering (SIANI), Universidad de Las Palmas de Gran Canaria (Spain).