Skip to Main content Skip to Navigation
Theses

Machine Learning Anomaly Detection Applications to Compact Muon Solenoid Data Quality Monitoring

Abstract : The Data Quality Monitoring of High Energy Physics experiments is a crucial and demanding task to deliver high-quality data used for physics analysis. At the Compact Muon Solenoid experiment operating at the CERN Large Hadron Collider, the current quality assessment paradigm, is based on the scrutiny of a large number of statistical tests. However, the ever increasing detector complexity and the volume of monitoring data call for a growing paradigm shift. Here, Machine Learning techniques promise a breakthrough. This dissertation deals with the problem of automating Data Quality Monitoring scrutiny with Machine Learning Anomaly Detection methods. The high-dimensionality of the data precludes the usage of classic detection methods, pointing to novel ones, based on deep learning. Anomalies caused by detector malfunctioning are difficult to enumerate a priori and rare, limiting the amount of labeled data. This thesis explores the landscape of existing algorithms with particular attention to semi-supervised problems and demonstrates their validity and usefulness on real test cases using the experiment data. As part of this project, the monitoring infrastructure was further optimized and extended, delivering methods with higher sensitivity to various failure modes.
Document type :
Theses
Complete list of metadatas

Cited literature [228 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02924477
Contributor : Abes Star :  Contact
Submitted on : Friday, August 28, 2020 - 10:09:09 AM
Last modification on : Friday, October 23, 2020 - 5:03:54 PM

File

83652_POL_2020_archivage.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02924477, version 1

Citation

Adrian Alan Pol. Machine Learning Anomaly Detection Applications to Compact Muon Solenoid Data Quality Monitoring. Artificial Intelligence [cs.AI]. Université Paris-Saclay, 2020. English. ⟨NNT : 2020UPASS083⟩. ⟨tel-02924477⟩

Share

Metrics

Record views

95

Files downloads

98