Development, Application and Representation of Algorithms for Discoveries with the ATLAS Forward Proton (AFP) Detector at CERN

The goal of this thesis is to extend selection criteria for getting signal/background samples from data collected during the AFP experiment at CERN in years 2016 and 2017. This experiment studies interaction of two near-missing pro-tons in two crossing beams where both of them go in opposite direction. For extending the selection criteria machine learning is used. The thesis first introduces the AFP experiment and explains signal/background processes and the current critera for retrieving them from data. Then follows an introduction to machine learning with explanation of a method which was requested for realization - neural networks. The practical part explains the dataset I worked with and which adjustments I made in dataset. Then I explain which machine learning method I have used and why. Also two approaches for getting signal/background samples are presented and how I combined both of them for obtaining the final results. At the end I describe the output of the thesis - webpage and script for getting desired samples.