Physics-informed Machine Learning

We’re offering the Seminar Physics-informed Machine Learning in the Winter Term 2024/2025.

In this seminar, we will explore influential papers in the field of physics-informed machine learning. This includes well-established concepts such as Gaussian process-based PDE solvers, Neural ODEs, and Neural Operators, as well as more recent advancements like hybrid models and foundational models for PDE solving. The goal is for you to prepare a self-contained tutorial based on a selected paper, which you will present in a block seminar at the end of the semester. Through these presentations and tutorials, we will discuss how physical knowledge can be encoded into machine learning models and examine the current limitations of these methods.

Organization and Evaluation

Credits: 3 ECTS

Language: English

Mandatory Dates & Room:

• Friday, 18.10.2024, 2–4 pm: Lecture Hall TTR2, AI Research Building (Maria-von-Linden Str. 6)
• Thursday & Friday, 20.02.2025 & 21.02.2025, 9 am–6 pm: Lecture Hall TTR2, AI Research Building (Maria-von-Linden Str. 6)

In the first meeting, we will introduce a list of potential topics/papers and provide guidance on what makes a good tutorial. Once each participant has chosen a topic, we will schedule an individual meeting in December or January to address any questions and refine the content of your presentation. You will then submit your tutorial by the end of January or the beginning of February and deliver a 45-minute presentation during the final seminar days. After the seminar, the submitted notebooks will be made available in a public GitHub repository, creating a valuable resource for you and others interested in scientific machine learning.

Registration

If you wish to participate in the seminar, please register via Moodle and attend the first session on Friday, October 18th, at 2 pm.