10-15 March 2019
Steinmatte conference center
Europe/Zurich timezone

Monitoring of time evolution of the trigger rates exploiting deep representation learning at the CMS experiment

Not scheduled
Steinmatte conference center

Steinmatte conference center

Hotel Allalin, Saas Fee, Switzerland https://allalin.ch/conference/
Poster Track 1: Computing Technology for Physics Research Poster Session


Adrian Alan Pol (Université Paris-Saclay (FR))


Real time monitoring of Compact Muon Solenoid (CMS) trigger system is a vital task to ensure the quality of all physics results published by the collaboration. Today, the trigger monitoring software reports on potential problems given the time evolution of the reported rates. The anomalous rates are identified given the deviation from the prediction which is calculated using a regression model generated independently for each trigger path. The CMS experiment implements a two-level trigger system. The fast hardware-based Level 1 triggers filter the data before the software-based High Level Triggers which access the full detector information. In most cases real problems present in the detector manifest themselves in abnormal trigger rates for a number of trigger paths which share a common infrastructure; whereas a few unrelated triggers misbehaving can be a result of statistical fluctuation. As such, the alarms require a considerable human interpretation. This contribution presents steps undertaken towards extending the current framework taking into account interdependence of different trigger paths. Our prototype, based on a deep autoencoder, exploits a global configuration of the trigger system and in particular its hierarchical nature.

Primary authors

Adrian Alan Pol (Université Paris-Saclay (FR)) Maurizio Pierini (CERN) Gianluca Cerminara (CERN) Cecile Germain (Universite Paris Sud)

Presentation materials