Partners

Technische Universität Dortmund

Technische Universität Dresden

The Dresden group pursues a multifaceted research programme for the simulation of collision experiments in particle physics. On the theoretical side, we are significantly involved in the development of the event generator Sherpa, a programme for the realistic simulation of collisions up to hadronic final states as they can be measured by detectors. Furthermore, the group is responsible for the physics modelling with the help of event generators within the ATLAS experiment in leading roles. The challenges of computational complexity that become apparent at this interface and practicable solutions to accelerate event generation will be the main focus of the Dresden activities in KISS. Preliminary work on novel ML algorithms to optimise the efficiency of event generation already promises first significant improvements and will be realised and extended in the framework of KISS.

Frankfurt Institute for Advanced Studies (FIAS)

Georg-August-Universität Göttingen

Universität Hamburg

The group of Prof. Gregor Kasieczka is pursuing an extensive work program for the development of AI-based methods in particle physics. Important topics are the development of network architectures for physical data, the unsupervised detection of anomalies as well as the description of uncertainties and the improvement of the robustness of decisions. New simulation approaches play a central role in these efforts. This includes improved simulations of high-resolution detectors with various architectures, posthoc adjustment of simulation results, and the investigation of statistical properties of generative models. The group of Prof. Peter Schleper brings experience in the development of fast detector simulation methods and currently coordinates these activities in the CMS experiment. Prof. Marcus Brüggen’s group has been working on Deep Learning for source detection in astronomy, including in the context of the Data Challenges for the Square Kilometer Array (SKA). Other research topics are 3D Convolutional Networks for parameter estimation in cosmology and super-resolution methods in radio astronomy.

Universität Heidelberg

Ludwig-Maximilians-Universität München

Two LMU groups with expertise in AI are contributing to the KISS project. Based on preparatory work in the ErUM-Data pilot project, the particle physics group focuses on surrogate models and adaptive sampling to speed up simulations. Generative models are developed for background patterns in high-resolution silicon detectors. Adaptive sampling methods are studied for sophisticated selections of B meson decay channels. The cosmology group addresses the problem of high-dimensional surrogate models for cosmological simulations on fine-grained lattices.

Deutsches Elektronen-Synchrotron (DESY)

The group of Frank Gaede works on the development of generic algorithms and software solutions for high-energy physics with a focus on the simulation of current and future particle detectors. One key aspect here is the development of generative ML methods for the precise and fast simulation of highly granular calorimeters, such as those in preparation for CMS. The group of Jakob van Santen is part of the Neutrino Astronomy Group at the DESY Zeuthen site and works on the development and optimisation of simulation software for the IceCube neutrino observatory. The focus is on the treatment of light propagation in natural, inhomogeneous media and their systematic errors.

partners in KISS