Deep Neural Networks for Electron-Positron discrimination in the JUNO experiment

Neutrino physics has always been an important area of research in particle physics, especially since the discovery of neutrino oscillations. The Jiangmen Underground Neutrino Observatory (JUNO) is a new large liquid scintillator detector that aims to solve the neutrino mass hierarchy measuring reactor neutrino interaction in the detector via inverse beta decay. One of the most relevant background eects is given by the presence of electrons which, even if they don't take part in the decay reaction, leave a trace very similar to that of positrons in the liquid scintillator. High energies experimental physics has always had to deal with the management and observation of large amounts of data; for this reason, nowadays, the development of algorithms and the use of computer techniques are some of the most important skills that form the background preparation of a physicist. With the advent of the so-called big data and the rapid development of hardware components, the eld of articial intelligence has made numerous progresses in recent years. In this thesis the electrons-positrons discrimination in JUNO experiment are investigated through the use of articial neural networks. The work is organized by rst introducing the main features of JUNO and the inverse beta decay; then a brief illustration of the modern deep learning techniques is given. Finally, the data set is presented, together with the development of the techniques, the data analysis and the obtained results.