BEGIN:VCALENDAR VERSION:2.0 PRODID:-//linuxsoftware.nz//NONSGML Joyous v1.4//EN BEGIN:VEVENT SUMMARY:Optimization of a Transformer-Based Model for Flavour-Tagging in t he ATLAS Experiment DTSTART:20251125T160000Z DTEND:20251125T180000Z DTSTAMP:20260811T084821Z UID:1628c2d7-f57e-4557-8509-fad8c2cdb844 SEQUENCE:4 CREATED:20251121T092505Z DESCRIPTION: The Standard Model is the most widely accepted theory in the field of particle physics. In order to test and find some of the missing p ieces of the theory\, several experiments are being made around the world. One of the main drivers of experimental particle physics research are col lider experiments\, where particles are accelerated and carefully made to collide at specific regions with detectors ready to collect the products o f the interaction. This data is later treated and the events reconstructed so that it is possible to learn about the phenomena occurring near the co llision point. This complex task has several stages\, one of which is call ed jet flavour-tagging\, that aims to identify the flavour of the particle s that originated jets after the collision. Machine learning models are em ployed with this task due to its complexity derived from the large amounts of data collected from each collision. One such model\, created by the AT LAS Collaboration\, called GN2\, is studied in this work with a focus on t he input variables it receives. In this thesis\, different versions of the model were created with different input variables to test their importanc e to the model’s performance. Additionally\, a study was conducted to di rectly test the impact of these input variables on the model’s decision- making\, and the results suggest that a group of these variables related t o the number of particle interactions with the detector may have a limited effect on the model. LAST-MODIFIED:20251124T114741Z LOCATION:Online URL:http://df.vps.tecnico.ulisboa.pt/pt/eventos/optimization-of-a-transfor mer-based-model-for-flavour-tagging-in-the-atlas-experiment/ X-ALT-DESC;FMTTYPE=text/html:

The Standard Model is the most widely accepted theory in the field of particle physics. In o rder to test and find some of the missing pieces of the theory\, several e xperiments are being made around the world. One of the main drivers of exp erimental particle physics research are collider experiments\, where parti cles are accelerated and carefully made to collide at specific regions wit h detectors ready to collect the products of the interaction.

Th is data is later treated and the events reconstructed so that it is possib le to learn about the phenomena occurring near the collision point. This c omplex task has several stages\, one of which is called jet flavour-taggin g\, that aims to identify the flavour of the particles that originated jet s after the collision. Machine learning models are employed with this task due to its complexity derived from the large amounts of data collected fr om each collision.


One such model\, created by the ATLAS Co llaboration\, called GN2\, is studied in this work with a focus on the inp ut variables it receives. In this thesis\, different versions of the model were created with different input variables to test their importance to t he model’s performance. Additionally\, a study was conducted to directly test the impact of these input variables on the model’s decision-making \, and the results suggest that a group of these variables related to the number of particle interactions with the detector may have a limited effec t on the model.

END:VEVENT END:VCALENDAR