Deep Learning in di-tau final states
Lead Research Organisation:
Lancaster University
Department Name: Physics
Abstract
Higgs decays into two hadronic tau leptons offer a unique opportunity in high energy physics for the ATLAS experiment. They are the only directly accessible Higgs couplings to fermions and offer a relatively large cross section. This can be exploited in measuring the CP properties of the Higgs boson, accessing the Higgs self coupling in the bbtautau final state and using the Higgs as a portal into physics beyond the Standard Model. In this project we will look at the di-tau final state with deep learning methods in order to improve sensitivity for this Higgs decay mode.
People |
ORCID iD |
Harald Fox (Primary Supervisor) | |
Sebastiano Spinali (Student) |
Studentship Projects
Project Reference | Relationship | Related To | Start | End | Student Name |
---|---|---|---|---|---|
ST/P006795/1 | 30/09/2017 | 29/09/2024 | |||
2063461 | Studentship | ST/P006795/1 | 30/09/2017 | 31/12/2021 | Sebastiano Spinali |
Title | Merged Vertices Classifier model |
Description | The InDetVertexValidation package is able to retrieve the whole set of vertex and track features, that afterwords are divided, using a label, in two different classes: Main Scatter vertex and Pile Up vertex. For each vertex are than stored different variables, such as the number of tracks, the sum of all the momenta of each track belongs to that vertex and a set of variables, called impact parameter momenta, related to both perpendicular and longitudinal impact parameters. For each given track are stored the momentum, coordinates informations and the impact parameters. Three different models were developed, one with a single branch made of vertex features, the second one is made of track features, and the last one is taking into account both branches, concatenated in a single last layer to still get a binary output for the evaluation step. For all of them is than compute both training and evaluation step using Keras, and the result are also compared with certain scikit-learn Classifiers. |
Type Of Material | Database/Collection of data |
Year Produced | 2019 |
Provided To Others? | No |
Impact | It would help to have a cleaner Hard Scatter vertex, that means a more accurate vertex informations and therefore more accurate informations on tracks beloging to those vertices for all the other analysis within the ATLAS Collaboration. |
Title | Vertex Recovery model |
Description | The InDetVertexValidation package is able to retrieve the whole set of track features, that afterwords are divided, using a label, in two different classes: Main Scatter vertex and Pile Up vertex. For each given track are stored the coordinates informations, pseudo-rapidity, azimuthal angle, transverse Momentum, longitudinal impact parameter and transverse impact parameter. Those features are than used as a input for a model, based on KMeans, able to perform the clusterization for each given vertex. |
Type Of Material | Database/Collection of data |
Year Produced | 2019 |
Provided To Others? | Yes |
Impact | It would help to have a cleaner Hard Scatter vertex, that means a more accurate vertex informations and therefore more accurate informations on tracks beloging to those vertices for all the other analysis within the ATLAS Collaboration. |
Description | Outreach at CERN - School Visit |
Form Of Engagement Activity | A talk or presentation |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Schools |
Results and Impact | Several UK School every year want to visit CERN Facilities. As a member of ATLAS, I have been requested by Lancaster Uni to give talk about the CERN topics involving the CERN world, introduction to particle physics and to Higgs field and mechanism. |
Year(s) Of Engagement Activity | 2018,2019 |