Bachelor thesis defense by Rasmus S. Thomsen

Title: Revealing hidden particles with neutral networks


The Standard model of particle physics is an extremely successful theory, whose major predictions have been verified by numerous experiments. Nevertheless there are solid indications that there exist more elementary particles. Some of them can even be accessible with the modern-day experiments, but have never been detected because their presence should be deduced by analyzing incomplete and noisy data.

This project explores the possibility of doing such an analysis using a neural networks. To this end, we generate computer-simulated events describing W-bosons, decaying to three leptons and a neutrino via an intermediate heavy neutral lepton – a hypothetical particle able to explain neutrino oscillations. One of the three final leptons is a τ-lepton – a short lived particle, whose decay chain always contains at least one neutrino. Some of the kinematic information about this process is lost at colliders (due to two escaping neutrinos) which prohibits an analytic reconstruction of the mass of the intermediate particle.

We train a neural network-based regressor to reconstruct the particle’s mass, given the incomplete kinematic information. We demonstrate that the neural network is able to predict the mass of the particle with a precision of above 10% for the intermediate mass particles that the neural network has not been trained upon. Furthermore, it is also shown that the neural network is generally able to differentiate between different signals from these intermediate particles, if their mass is at least 4 − 5 GeV apart.

Censor: Else Lytken 

Join Zoom Meeting

Meeting ID: 698 9165 5716
Passcode: 658745