You are here

Exploration of Deep Learning Methods for Vector Boson Fusion Event Discrimination

Title: Exploration of Deep Learning Methods for Vector Boson Fusion Event Discrimination.
41 views
13 downloads
Name(s): Sanchez, Alejandro Mario, author
Type of Resource: text
Genre: Text
Bachelor Thesis
Date Issued: 2017-04-26
Physical Form: computer
online resource
Extent: 1 online resource
Language(s): English
Abstract/Description: In the past, the discrimination of Higgs boson events from background events has been done with the introduction of clever variables to make discrimina- tion easier for machine learning methods. By using a complex neural network with two hidden layers and a stochastic optimizer algorithm called Adam, I can discriminate Higgs Boson events from non-Higgs events with a receiver oper- ating characteristic (ROC) score of 0.961 and a validation accuracy score of 94.4% using the raw data alone. I also attempted to discriminate those Higgs events that were produced by the process of Vector Boson Fusion (VBF), and achieved a ROC score of 0.858 and a validation score of 79.7%. Although not as high a score as the Higgs/non-Higgs discrimination, the correct choice of raw data variables outperformed the historical clever variables used for this discrimination task. These results confirm something that it intuitively ob- vious, namely, that all the information about events is contained in the raw data and can be extracted directly with a sufficiently smart machine learning algorithm. A lot of time currently goes into designing clever variables for these purposes, but, in the future, computers will be able to discriminate with no need of these clever variables, allowing researchers to spend less time de- signing complicated, clever, variables and more time inventing better ways to analyze data for new physics.
Identifier: FSU_libsubv1_scholarship_submission_1493403077 (IID)
Keywords: High Energy Physics, HEP, Machine Learning, Neural Network, Multilayer perceptron, MLP, Adam, VBF, Higgs
Persistent Link to This Record: http://purl.flvc.org/fsu/fd/FSU_libsubv1_scholarship_submission_1493403077
Owner Institution: FSU

Choose the citation style.
Sanchez, A. M. (2017). Exploration of Deep Learning Methods for Vector Boson Fusion Event Discrimination. Retrieved from http://purl.flvc.org/fsu/fd/FSU_libsubv1_scholarship_submission_1493403077