Error Analysis and Optimization of Neural PDE Solvers
Date: Mon. December 14, 2026
Host: FAU DCN-AvH Chair for Dynamics, Control, Machine Learning and Numerics -Alexander von Humboldt Professorship at Friedrich-Alexander-Universität Erlangen-Nürnberg (Germany)
Talk: Error Analysis and Optimization of Neural PDE Solvers
Speaker: Dr. Marius Zeinhofer
Affiliation: ETH Zürich
Abstract. I will give an overview of the current state of error analysis for neural network-based PDE solvers. In particular, I will discuss a fundamental issue with the popular physics-informed neural network (PINN) formulation, which minimizes the PDE residual in strong form and penalizes boundary values in the L^2 norm. For elliptic equations, for instance, this formulation only controls the error in the H^{1/2}(\Omega) norm but requires approximation in H^2(\Omega). Exact enforcement of boundary conditions removes this issue. I will then present a geometric approach to the optimization of neural PDE solvers that produces highly accurate solutions, rivaling classical methods. The approach will be illustrated on high-dimensional eigenvalue problems in quantum physics.
WHEN
Mon. December 14, 2026 at 14:00H
WHERE
On-site / Online
Room H13. Johann-Radon-Hörsaal
Cauerstraße 11, 91058 Erlangen. Bavaria, Germany
GPS-Koord. Raum: 49.573764N, 11.030028E [Online] FAUtv live-streaming: https://www.fau.tv/clip/id/64145
You might like:
• FAU MoD Lectures
• Upcoming events
• FAU MoD Courses & Workshops
• FAU/Jilin Workshop: Workshop on Recent Trends in Applied Mathematics and Machine Learning 2026
• FAU MoD Course: Discovery of singularities with physics-informed neural networks by Prof. Dr. Gonzalo Cao-Labora
• FAU MoD Lecture: Moving your robotic hand is a matter of time and practice by Prof. Dr. Claudio Castellini
• FAU MoD Lecture: Trustworthy AI: Reliable, Explainable, Revisable by Prof. Dr. Ute Schmid
• MLPDES26 workshop: MLPDES26, Machine Learning and PDEs Workshop (2026)
• FAU MoD Lecture: Reverse typography and the theory of shape: Can old books be brought back to life? by Prof. Dr. Jean-Michel Morel
• FAU MoD Lecture: A data-driven approach to closed-loop control of wound state progression to drive healing outcomes by Prof. Dr. Marcella M. Gomez
• FAU MoD Lecture: Data Driven Modeling for Scientific Discovery and Digital Twins by Prof. Dr. Dongbin Xiu
• FAU MoD Lecture: A long life: How desirable is it, evolutionarily speaking? by Prof. Dr. Hanna Kokko
• FAU MoD Lecture: Bridging numerics and scientific machine learning for industrial applications by Dr. Christopher Straub
• FAU MoD Workshop: FAU MoD Workshop (Dec. 2025) by Prof. Giovanni Fantuzzi | Prof. Denisa Martonova
• FAU MoD Lecture: Quantum firmware: optimal control for quantum processors by Prof. Dr. Tommaso Calarco
• FAU MoD Lecture: AI Components in PDE Solvers by Prof. Dr. Nils Thürey
• FAU MoD Lecture: Disruption in science and engineering happens at scale by Prof. Dr. Johannes Brandstetter
• FAU MoD Workshop: FAU MoD Workshop (Sep. 2025) by Prof. Lorenzo Liverani | Prof. Hagen Holthusen
• FAU MoD Lecture: Exemplary applications of machine learning and optimization in quantum chemistry by Prof. Dr. Andreas Görling
• FAU MoD Lecture & workshop: AI for maths and maths for AI by Dr. François Charton
• FAU MoD Lecture: Optimization-based control for large-scale and complex systems: When and why does it work? by Prof. Dr. Lars Grüne
• FAU MoD Lecture: Mathematics of neural stem cells: Linking data and processes by Prof. Dr. Ana Martin-Villalba
• FAU MoD Lecture: FAU MoD Lecture S. Jin / N. Liu (double session) by Prof. Dr. Shi Jin and Prof. Dr. Nana Liu
• FAU MoD Lecture: Do you think you understand sex and death? Why predictions about biological processes require more than just intuition by Prof. Dr. Hanna Kokko
• FAU MoD Lecture: FAU MoD Lecture. Special December 2024 by Prof. Dr. Holger Rauhut and Prof. Dr. Christian Bär
• FAU MoD Lecture: Measuring productivity and fixedness in lexico-syntactic constructions by Prof. Dr. Stephanie Evert
• FAU MoD Lecture: New avenues for the interaction of computational mechanics and machine learning by Prof. Dr. Paolo Zunino
• FAU MoD Lecture: Discovering and Communicating Excellence by Prof. Dr. Ute Klammer
• FAU MoD Lecture: Thoughts on Machine Learning by Prof. Dr. Rupert Klein
• FAU MoD Lecture: Using system knowledge for improved sample efficiency in data-driven modeling and control of complex technical systems by Prof. Dr. Sebastian Peitz
• FAU MoD Lecture: Image Reconstruction – The Dialectic of Modelling and Learning by Prof. Dr. Martin Burger
• FAU MoD Lecture: The role of Artificial Intelligence in the future of mathematics by Prof. Dr. Amaury Hayat
• FAU MoD Lecture: FAU MoD Lecture. Special November 2023 by Prof. Dr. Michael Kohlhase and Prof. Dr. Edriss S. Titi
• FAU MoD Lecture: Free boundary regularity for the obstacle problem by Prof. Dr. Alessio Figalli
• FAU MoD Lecture: Physics-Based and Data-Driven-Based Algorithms for the Simulation of the Heart Function by Prof. Dr. Alfio Quarteroni
• FAU MoD Lecture: From Physics-Informed Machine Learning to Physics-Informed Machine Intelligence: Quo Vadimus? by Prof. Dr. George Karniadakis
• FAU MoD Lecture: From Alan Turing to contact geometry: Towards a “Fluid computer” by Prof. Dr. Eva Miranda
• FAU MoD Lecture: Applications of AAA Rational Approximation by Prof. Dr. Nick Trefethen
• FAU MoD Lecture: Learning-Based Optimization and PDE Control in User-Assignable Finite Time by Prof. Dr. Miroslav Krstic
_
Don’t miss out our last news and connect with us!