Title Information
Title
Anomaly Detection for the CMS Pixel Detector Using Machine Learning
Type of Resource (primo)
dissertations
Name: Personal
Name Part
Gu, Tianshu
Role
Role Term: Text
creator
Name: Personal
Name Part
Gouskos, Loukas
Role
Role Term: Text
Advisor
Name: Corporate
Name Part
Brown University. Department of Physics
Role
Role Term: Text
sponsor
Origin Information
Copyright Date
2026
Physical Description
Extent
vii, 33 p.
digitalOrigin
born digital
Note: thesis
Thesis (Sc. M.)--Brown University, 2026
Genre (aat)
theses
Abstract
As data volumes in high-energy physics continue to increase, data quality monitoring becomes increasingly important, while traditional monitoring procedures are no longer sufficient on their own. This thesis studies anomaly detection for the CMS pixel detector using machine-learning methods based on lumisection-level Data Quality Monitoring outputs. Two approaches are investigated: a variational autoencoder (VAE)-based method and a Contrastive Language-Image Pre-training (CLIP)-based method. The VAE-based method converts reconstruction loss into anomaly decisions and identifies more anomalous samples while maintaining good interpretability, although its generalization remains limited. The CLIP-based method is first tested in a few-shot setting, where its performance is weak. Its performance improves after anomaly injection is used to enlarge the training set, but it still remains below that of the VAE-based method.
Subject
Topic
Machine Learning
Subject
Topic
CMS Experiment
Subject
Topic
Anomaly Detection
Subject
Topic
Data quality monitoring
Language
Language Term (ISO639-2B)
English
Record Information
Record Content Source (marcorg)
RPB
Record Creation Date (encoding="iso8601")
20260516