Title: | Machine learning methods for the raw data analysis of cryogenic dark matter experiments | Language: | English | Authors: | Wagner, Felix | Qualification level: | Diploma | Keywords: | dunkle Materie Maschine Learning kryogenik dark matter machine learning |
Advisor: | Schieck, Jochen | Assisting Advisor: | Reindl, Florian | Issue Date: | 2020 | Number of Pages: | 123 | Qualification level: | Diploma | Abstract: | Kryogene Dunkle Materie Experimente erfordern eine sensible Datenanalyse, um für Sub-GeV / c ^ 2-Dunkle Materie empfindlich zu sein. In dieser Arbeit werden gelabelte Datensätze, die auf Daten der direkten Suche nach dunkler Materie mit CRESST-II basieren, für die Verwendung mit modernen Methoden des maschinellen Lernens simuliert. Deep Learning-Modelle werden trainiert, um die Rückstoßenergie von Ereignissen zu bestimmen und sie von nicht-physischen Artefakten und Rauschen zu unterscheiden. Unsupervised Learning wird angewendet, um verschiedene Ereignistypen in den CRESST-II-Daten zu untersuchen. Die Ergebnisse werden mit der aktuellen Analyse verglichen, die auf angepassten Filtern und Fit-Techniken basiert. Cryogenic dark matter experiments require a sensible data analysis to be sensitive to Sub-GeV/c^2 dark matter. In this thesis, supervised datasets, based on data from the CRESST-II direct dark matter search, are simulated for the use with modern machine learning methods. Deep Learning models are trained to determine the recoil energy of events and distinguish them from non-physical artifacts and noise. Unsupervised methods are applied to explore different event-types in the CRESST-II data. Results are compared with the current analysis, which is based on matched filters and fitting techniques. |
URI: | https://doi.org/10.34726/hss.2020.77322 http://hdl.handle.net/20.500.12708/15023 |
DOI: | 10.34726/hss.2020.77322 | Library ID: | AC15675604 | Organisation: | E141 - Atominstitut | Publication Type: | Thesis Hochschulschrift |
Appears in Collections: | Thesis |
Files in this item:
File | Description | Size | Format | |
---|---|---|---|---|
Machine Learning Methods for the Raw Data Analysis of Cryogenic Dark Matter Experiments.pdf | 39.69 MB | Adobe PDF | ![]() View/Open |
Page view(s)
38
checked on Feb 21, 2021
Download(s)
28
checked on Feb 21, 2021

Google ScholarTM
Check
Items in reposiTUm are protected by copyright, with all rights reserved, unless otherwise indicated.