Searches for new physics using unsupervised machine learning for anomaly detection at the ATLAS detector and the development of particle identification algorithms for the HL-LHC
Crosby, Jacob E.
Citations
Abstract
This document contains discussions on completed and ongoing projects that have developed over the past few years while working on the ATLAS detector located at CERN in Geneva, Switzerland. The first discussion will be on my qualification task for the ATLAS collaboration which is on the development and future plans of the DL1d tagger that is currently used as a base- line tagger for run 4 of the ATLAS detector. After, the discussion will transition to an analysis that applies a novel anomaly detection technique which uses a neural network architecture called the autoencoder. This neural network is then trained on 1% randomly selected events of run 2 data from the ATLAS detector. Once the model and anomalous regions are defined, the model is used to find phase spaces where events that contain physics beyond the standard model may occur. Statistical analysis is then applied to these phase spaces in order to find signatures of new physics. No significant signatures are found. Lastly, I will discuss an ongoing search for a new massive scalar X decaying into a new light scalar Y and the standard model Higgs boson H through the process X→YH→bbbb in the boosted topology.