The Scientific Computing group at CWI organizes a research semester programme on the emerging field of Scientific Machine Learning (SciML). SciML lies at the intersection of Scientific Computing and Machine Learning, combining state-of-the-art development in mathematics and computer science. SciML concerns the application of artificial intelligence / machine learning techniques to complex physics problems (e.g. climate simulation). Such complex applications require expensive computer simulation, are often highly sensitive to perturbations and feature ‘rare events’ that have low probability but high impact. Classical machine learning techniques, relying on an abundance of data, do not perform well on such problems. In SciML, algorithms are proposed that combine physics-driven models with data-driven models in a hybrid manner to make robust and accurate predictions. Scientific computing can and should provide the mathematical underpinning of ML methods, helping to achieve interpretability, robustness and generalisation.
The semester programme aims to bring together researchers in The Netherlands that work or aim to work at the interface of physics-based modeling and machine learning. In particular, we naturally welcome mathematicians and computer scientists involved in physics problems, but are also inviting physicists and engineers to come and see the value that scientific machine learning has in many practical applications.
- Physics-informed machine learning
- Neural differential equations
- Reduced-order models and deep learning
- Group-equivariant learning
- Multi-scale problems and closure models
- Mathematical theory for deep learning
Lectures and hands-on tutorials for PhD students
SOCIETY & INDUSTRY
Bridging SciML theory and practice
Connecting the international and national research community
Throughout the program
International speakers with open problems and brainstorming