Lecture and hands-on tutorial at the NFDI4ING Caden Summer School 2026

I gave a lecture and a hands-on tutorial on machine learning for data analysis at the NFDI4ING Caden Summer School 2026, Practical Research Data Management in Materials Science, held from October 13 to 15, 2026 at Forschungszentrum Jülich (hosted by IMD-2, hybrid). I presented online on October 14.

The lecture, “Machine learning for materials data analysis: from measurements to trustworthy, FAIR results”, was about one idea: in machine learning, the data and the labels are the program. Good research data management (metadata, provenance, documentation) therefore decides whether an ML-based analysis can be trusted and reproduced.

The hands-on tutorial trains a U-Net to segment martensite-austenite (M-A) islands in SEM images of a bainitic steel, tests how much the result depends on how the data are split, and saves the analysis as a FAIR run folder with its configuration, provenance, a model card and an RO-Crate. A bonus notebook applies the same ideas to XRD patterns. The notebooks run on Google Colab and are openly available.

• Summer school page (NFDI4ING)
• Hands-on notebooks on GitHub

Thank you to the organizers, Hartmut Schlenz, Torsten Bronger and Michael Selzer, for the invitation.