First edition of the “ETP Blockpraktikum: Data Science”
Design a detector and use Artificial Intelligence (AI) to analyse its signals - that is the challenge students rose to in this first-ever ETP Blockpraktikum organised and led by Dr. Jan Kieseler as part of the research-oriented teaching effort.
Integral to the team were PhD students Isabel Haide, Lars Sowa, and Artur Monsch from ETP, who helped prepare and supervise the hands-on session.
After two weeks of preparatory exercises, students convened at the physics building. Equipped with their laptops, they embarked on simulating their first calorimeter with Geant4, the same state-of-the art software the CERN experiments use. Their primary tool? A simple browser, as the exercises were facilitated on Jupyter notebooks, running on machines provided by the SCC.
The initial phase revolved around electromagnetic particle showers, stemming from photons or electrons. The students delved into different calorimeter concepts and explored, for example, the optimal depth to contain even high energetic particles. The shower energies were determined by classic methods as well as deep neural networks to improve the resolution. Now ready for an advanced challenge, their detectors were subjected to hadronic showers, prompting a re-evaluation of design approaches. Furthermore, the neural networks transitioned to more sophisticated convolutional frameworks, also used for complex computer vision tasks.
Finally, everything absorbed from the lectures, group dialogues, and hands-on experiences was put to the ultimate test: a competition between groups, to build the best detector and the best algorithm aimed to measure the energy of hadrons and photons at the same time. To add more realism, restrictions were placed on the detector's size, its material cost, and the total number of read-out channels. It was a tough competition, but the team “T für Technik” emerged victorious and won a card deck with famous physicists from CERN. This prize was awarded during the course's concluding segment, a discussion forum where each group showcased their designs, offering a platform for comparative analysis.
All in all, it was an intense, fun, and highly rewarding week - we're looking forward to next time!
Contact: Dr. Jan Kieseler, Prof. Markus Klute, Prof. Torben Ferber