About PhysML

PhysML Project

PhysML: Structure-based Machine Learning for Physical Systems is a research project dedicated to developing machine learning methods that are firmly grounded in physical principles.

The project is led by the Analytics and Artificial Intelligence group in the Mathematics and Cybernetics department at SINTEF, Oslo, with key academic collaborators Prof. Elena Celledoni (NTNU) and Prof. George Em Karniadakis (Brown University).

The PhysML project has two complementary goals:

• use machine learning to gain knowledge about physical systems, and

• use physical knowledge to obtain machine learning models that are open, trustable, robust, and flexible.

The project aims to build on theory developed within numerical analysis and specifically the field called geometric numerical integration, which concerns numerical integrators that preserve some structure of the system they model, to create structure-preserving machine learning models.

SINTEF is one of Europe’s largest independent research institutes, with broad expertise spanning technology, natural sciences, and social sciences. Founded in 1950, SINTEF has more than 2,000 employees and works with partners in both the private and public sectors in Norway and internationally.

PhysML Workshop

The PhysML Workshop is the main scientific meeting of the PhysML project and a forum for exchange between researchers in machine learning, applied mathematics, and physics working on physically grounded learning methods. The first edition was held in 2024 and established the workshop as a focused venue for discussing structure-aware and physics-informed approaches to scientific machine learning, with future editions planned as part of the project.