Non-Euclidean data structures are present everywhere in the physical and digital world. In recent years, an increasing number of scientific fields have started to leverage the information contained within such data structures with the advent of Geometric Deep Learning. High Energy Physics is no exception, as nowadays modern methods using Graph Neural Networks are being developed and validated for various tasks across different reconstruction steps.
In this lecture we will first demonstrate the inherent expressive power of graphs as a data structure and introduce the key concepts of graph theory. Then we will discuss Graph Neural Networks and lay the mathematical foundation of important neural mechanisms such as Neural Message Passing or Graph Convolution. Lastly we will examine practical applications of Graph Neural Networks in High Energy Physics highlighting how these technologies can be effectively utilized.
Designed for students seeking practical knowledge of Graph Neural Networks, this lecture has two primary objectives. Firstly to illustrate the reasons that Graph Neural Networks are powerful deep learning tools and secondly to present the minimal background required to engage with computer science literature and successfully apply established technologies to High Energy Physics research.
University of Bonn
I am a PhD student at the Belle II experiment, actively engaged in conducting lepton flavor universality tests. Prior to this, I completed my undergraduate degree at the Aristotle University of Thessaloniki, followed by an M.Sc. program at the University of Strasbourg.
Throughout my studies I have contributed to the development of deep learning methods for particle identification and decay reconstruction. Recognizing the immense potential of machine learning in high energy physics, I firmly believe in its capacity to revolutionize the field, despite the challenges posed by its rapid evolution.
In my upcoming lecture, I endeavor to bridge the nomenclature gaps between the high energy physics and machine learning communities. I consider this effort essential for fostering the creation of high-quality deep learning algorithms, thereby unlocking new frontiers in both fields.