PoS - Proceedings of Science
Volume 450 - The Eleventh Annual Conference on Large Hadron Collider Physics (LHCP2023) - session Poster session
AutoEncoders for per-lumisection data quality monitoring at CMS
A. Papanastassiou*, V. Gori  on behalf of the CMS Collaboration
Full text: pdf
Pre-published on: January 16, 2024
Published on:
The monitoring of data quality is crucial both online, during the data taking, to promptly spot issues and act on them, and offline, to provide analysts with datasets that are cleaned against the occasional failures that may have crept in. Typically, data quality monitoring (DQM) is performed by shifters who look at a set of integrated quantities, compare them with reference histograms, and, based on their experience and training, assign quality flags. Recently CMS has developed the possibility of producing DQM plots per-lumisection, where a lumisection is a time unit corresponding to about 23 𝑠 of data taking. To analyze per-lumisection data, a manual approach would be prohibitive due to the high number of lumisections, therefore an automated one would be preferable. In this work, the first use in CMS of AutoEncoders to perform anomaly detection on per-lumisection data, specifically for quantities associated with jets and missing transverse energy, is presented. The technique developed allows the detection of anomalies at the level of individual lumisections, which might be overlooked when examining integrated quantities, and serves as a proof of concept regarding the efficacy of this and similar approaches.
DOI: https://doi.org/10.22323/1.450.0284
How to cite

Metadata are provided both in "article" format (very similar to INSPIRE) as this helps creating very compact bibliographies which can be beneficial to authors and readers, and in "proceeding" format which is more detailed and complete.

Open Access
Creative Commons LicenseCopyright owned by the author(s) under the term of the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.