Dortmund 2021 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 21: Data analysis, Information technology I

T 21.1: Vortrag

Montag, 15. März 2021, 16:00–16:15, Tu

Anomaly searches for new physics based on generative classifiers — •Sven Bollweg and Gregor Kasieczka — Universität Hamburg, Germany

There exist strong hints for the existence of physics beyond the standard model (BSM). Many models for BSM physics have been investigated but none of these could be observed in data so far. Another strategy are model-independent searches. The idea is that events originating from BSM processes differ from events originating from SM processes. Without applying any knowledge of possible BSM processes, it can be used to search for anomalous events.

To search for anomalies, we use a generative classifier (GC) based on invertible neural networks. A GC learns the likelihood of the input data. The likelihood can be used either for classification or anomaly detection. In the ideal case, anomalous events are less likely than all the other events if we train the GC on SM events. We show a first attempt to apply this method in the context of searching for new physics with the CMS experiment in the dijet final state. We investigate different input representations and anomaly scores based on the likelihood.