A workshop on utilizing numerical analysis and physics knowledge to obtain accurate, explainable and robust machine learning models

Oslo, Norway, May 14th – 16th 2024

This three-day physical workshop will consist of morning and afternoon talks every day, with sessions for discussions and debates, and ample room for sharing ideas and discussing potential collaborations between the participants.

The overarching topic is machine learning models that incorporate structures based on the underlying physics. This may include but is not limited to:

• Geometric structures in machine learning architectures

• Numerical analysis techniques for improved neural network models

• Analysis of properties of neural networks

• Hard-constrained models, including Hamiltonian neural networks and extensions

• Soft-constrained models, including physics-informed neural networks

Invited speakers:

• J. Nathan Kutz (University of Washington)

• George Em Karniadakis (Brown University) (virtual speaker)

• Nat Trask (University of Pennsylvania)

• Daniel Worrall (DeepMind)

• Takaharu Yaguchi (Kobe University)

• Rima Alaifari (ETH Zurich)

• Wil Schilders (TU Eindhoven, TU Munich - IAS)

• Felix Dietrich (TU Munich)

• Elena Celledoni (NTNU)

• Morten Hjorth-Jensen (University of Oslo, Michigan State University)

The workhop is hosted by the Analytics and AI group at SINTEF Digital in Oslo, Norway. It is part of the project “PhysML: Structure-based machine learning for physical systems”, with academic partners from the Norwegian University of Science and Technology (NTNU) and Brown University.

If you are interested in attending and want more information, do not hesitate to contact us.