Deep learning for reconstructing and simulating particles in collider experiments

by Etienne Dreyer

Weizmann Institute

at Particles and Fields Seminar

Mon, 11 Nov 2024, 14:00
Sacta-Rashi Building for Physics (54), room 207

Abstract

Particle detectors like ATLAS and CMS are highly sophisticated systems designed to record the aftermath of high-energy particle collisions. To interpret the vast and complex data generated, physicists rely on pattern recognition algorithms to reconstruct the particles produced in these collisions. In this talk, I will present new results from an algorithm that leverages deep learning with hypergraphs to enhance particle reconstruction quality. Additionally, I will explore the potential of deep learning as a groundbreaking tool for modeling particle interactions in detectors. These “virtual detector” models can rapidly generate data that mimic the full particle reconstruction chain, bypassing the computationally expensive step of simulating particle interactions at the microscopic level. Both developments underscore how deep learning is shaping the future of reconstruction and simulation pipelines in collider experiments.

Created on 29-10-2024 by Kats, Yevgeny (katsye)
Updaded on 29-10-2024 by Kats, Yevgeny (katsye)