This website link is outdated.

My up-to-date website is

Here are quick links to up-to-date content you might have been looking for:

The outdated site is below...

Convolutional Neural Networks and Particle Physics

Last modified: 27 September 2022

This project explores the use of convolutional neural networks for classifying the products of high-energy collisions produced in particle physics experiments like the Large Hadron Collider at CERN and was written in the scope of Seminar course requirements at the Faculty of Math and Physics at the University of Ljubljana.

You might be interested in…

About the project

Particle classification means identifying the results of a collision, often as simply as with a binary yes/no answer, e.g. “this collision produced a Higgs boson” or “this collision did not produce a Higgs boson”. Performing classification with a high degree of certainty is vital if, say, you are a research group interested in announcing the discovery of a new elementary particle.

Particle classification

This project is distinguished by its focus on performing classification using as raw, low-level data as possible (i.e. the data produced directly by a particle detector’s trackers and calorimeters, without further processing). Classification techniques that produce results directly from low-level data are called “end-to-end” classifiers. End-to-end classifiers are interesting because:

And what about convolutional neural networks? Convolutional neural networks (CNNs) are a class of machine learning systems well-suited to processing image-like data. Raw detector data—the type used in end-to-end classification—takes the form of energy and position information distributed on a two- or three-dimensional spatial grid across the detector’s volume. This data is essentially a series of image-like snapshots of the energy and position left behind by particles as they fly through the particle detector. Since CNNs are well-suited to image-like data, they find an application in end-to-end particle classification.


The paper and presentations address the following questions in the progression given below:

About the Seminar course at FMF

Seminar is a required course for students in the final semester of the undergraduate physics program at the Faculty of Mathematics and Physics at the University of Ljubljana. In the scope of the course, students, under the guidance of a faculty mentor, write a short undergraduate thesis on a currently relevant physics topic and present the topic to their classmates and professors.

The project encompasses two parts:

The project is intended primarily as an exercise in clear scientific writing and presentation, a training, in some sense, for giving presentations at scientific conferences. However, undergraduate students are neither expected nor encouraged to produce original research in the scope of the Seminar course, simply to clearly present their chosen topic at a level suitable (i.e. not too advanced) for a general final-year undergraduate audience.