Relatore
Graziella Russo
(Istituto Nazionale di Fisica Nucleare)
Descrizione
Graph Neural Networks provide a promising technique for performing Anomaly Detection tasks, by expressing potentially heterogeneous detector information in graph form. In our approach, graphs can be used to represent large-radius jets with interconnected topocluster "nodes", leveraging graph information and message passing to identify unexpected signatures as anomalies. We will discuss our ongoing work based on the open data of LHC Olympics 2020 and its application for ATLAS Run 3 diboson searches in fully hadronic final states.
Autore principale
Graziella Russo
(Istituto Nazionale di Fisica Nucleare)
