Orateur
Description
he reconstruction of particle trajectories is a key challenge of many high-energy and nuclear physics experiments. Indeed, its performance directly impacts particle identification and physics performances while also representing one of the primary CPU consumers of many experiments. As the luminosity of particle colliders increases, this reconstruction will become more challenging and resource-intensive. New algorithms are thus needed to address these challenges efficiently.
Since the TrackML challenge in 2018, the use of Deep Learning has emerged as a viable alternative to traditional tracking techniques, thanks to its ability to exploit global event context and its potential for fast inference on GPU systems. This period also saw the development of ACTS, the Open-Source tracking library that became a central testing environment for many new DL algorithms.
In this talk, after a short overview of the traditional tracking techniques and the ACTS library. We will present an overview of what has been developed so far in the DL world, including Simple Neural Network, Graph Neural Network and Transformers. And we will discuss the advantages and disadvantages compared to the classical solution.