From May 6 to 9, 2025, the fifth edition of the Deep Learning School Basic Concepts took place – this time at the Erholungsgesellschaft in Aachen. Around 50 dedicated participants – including students, PhD candidates, and postdocs – took the opportunity to dive deep into the fundamentals and current developments in the field of deep learning.

The program offered a well-balanced mix of theoretical input and hands-on practical sessions. Following a welcome by the ErUM-Data team, the school officially began on Tuesday afternoon with the first lecture by Dr. Ethan Simpson (University of Manchester), who introduced the topic Neural Network Building Blocks. In the subsequent practical session, participants had the chance to immediately apply their newly gained knowledge. The evening concluded with a shared dinner, providing a relaxed atmosphere for initial conversations and networking.

On Wednesday morning, Dr. Maximilian Horzela (University of Göttingen) took over with the second topic block Mastering Model Building, followed by another hands-on session. After lunch, the program continued with a session on Convolutional Neural Networks, led by Dr. Theo Heimel (UC Louvain). Once again, theoretical input was directly combined with practical application.

In addition to the academic exchange, the school also placed great emphasis on community-building activities. On Wednesday evening, a social event took place, starting with a short city tour followed by a pub quiz. In a relaxed atmosphere, participants competed in mixed teams with creative names like We Come With Pretrained Knowledge, Backprop to the Future, and Maximum Likelihood Minimizer – sparking laughter, friendly rivalry, and lots of fun.

The final thematic block on Thursday morning was led by Dr. Knut Zoch (Harvard University), who gave an engaging introduction to Transformers & Prompt Engineering. Highly motivated, the participants then set to work on solving the task assigned to them in the final hands-on session.

To conclude the school, participants faced a special challenge: the Solar Orbiter Images Challenge. Working in randomly assigned teams, they tackled a hands-on deep learning task, demonstrating great dedication and teamwork—sometimes working late into the night. On Friday morning, the finalized results were presented in short presentations. The jury, composed of the speakers and tutors, was visibly impressed by the diversity and quality of the solutions. The winning team was Dr. Sun and his Friends—Marina, Rico, Birgit, Baodong, and David. Congratulations to the champions!

The Deep Learning School Basic Concepts was once again a great success in its fifth edition—thanks to the tremendous dedication of everyone involved. Our heartfelt thanks go to the lecturers Dr. Ethan Simpson, Dr. Maximilian Horzela, Dr. Theo Heimel, and Dr. Knut Zoch for their outstanding contributions, as well as to the tutors Mikhail Romodin (University of Tübingen), Markus Pirke (FAU), Max Fusté Costa (DESY), and Kees Benkendorfer (Harvard University) for their invaluable support. A special thanks also goes to the BMBF for the financial support. BMBF für die finanzielle Förderung dieser Veranstaltung.