Objectives
The joint project Artificial Intelligence for the Fast Simulation of Scientific Data (KISS) conducts interdisciplinary research to simplify, accelerate, and improve simulations through artificial intelligence (AI).
Both for the fundamental research, and beyond, simulations are critical to the design of experiments and the analysis of rapidly growing data sets. In this project, the goal is to create highly accurate and effective simulation algorithms with minimal resources for a broad application in particle physics, hadron- and nuclear physics, astro-particle physics, and astronomy. Such efficient simulation algorithms are scientifically important and crucial in terms of sustainability.
In our project, we develop simulations using machine learning methods. We use deep learning to accelerate classical simulation approaches or to create fast surrogate models. Starting from immediate scientific research on large-scale facilities, we work on interdisciplinary approaches and produce software packages for comprehensive use. The commonality of problems and methods in particle, astroparticle and astrophysics allows interdisciplinary cooperation. The project partners each have large individual methodological experience and can rely on the precision tools/simulation programs of these communities. Together with partners from computer science and mathematics, we also work on fundamentally relevant questions in dealing with simulations, their statistical properties and the quantification of uncertainties.