Computational Astro- and Particle Physics (CDL1)
Our work spans over several generic aspects of precision measurements in high-energy physics and morphological classification of radio galaxies in astrophysics. The interaction between scientists from the two domains has proved to be beneficial as we apply similar techniques.
First, generative models are used to improve and speed up the simulation of low- and high-level physics objects, like calorimeter showers (https://arxiv.org/abs/2112.09709) and jet resolution (Sven Harder's bachelor thesis at UHH & Shruthi Janardhan's master thesis at TUHH). In radio astronomy, we improve radio galaxy classifiers to face the challenge of limited labelled data (https://arxiv.org/abs/2206.15131, accepted to ml.astro).
Furthermore we develop generic frameworks for physics analysis (https://gitlab.cern.ch/DasAnalysisSystem, presently only for jet analysis in CMS, e.g. https://doi.org/10.1007/JHEP02%282022%29142, easily generalisable) and test the quality of the statistical description of differential distributions (https://arxiv.org/abs/2111.09968v3, resubmitted to SciPost Pysics Core).
Finally, we are involved in anomaly detection with classical and ML techniques in data taking with the CMS silicon tracker (https://cds.cern.ch/record/2812026).