Machine Learning Enhances Jet Classification in High-Energy Physics

Recent advancements in machine learning have been applied to two significant challenges in high-energy physics, specifically at the Large Hadron Collider (LHC). The paper titled "From strange-quark tagging to fragmentation tagging with machine learning" by Yevgeny Kats and Edo Ofir explores these challenges: distinguishing strange-quark jets from down-quark jets, and identifying the fragmentation channel of a quark.

The authors highlight that traditional methods struggle with these tasks due to the lack of discriminating tools provided by quark lifetimes and masses, as well as parton showering. Instead, the study relies on analyzing the distributions of hadron types and their kinematics within the jets.

To tackle these classification problems, the researchers employed advanced machine learning architectures, including Graph Attention Networks and the Particle Transformer. These models utilize the properties of jets and their constituents as inputs. The performance of these sophisticated architectures was compared to a simpler Multilayer Perceptron that used basic variables.

The findings indicate that while the advanced models did not significantly enhance the differentiation between strange-quark and down-quark jets, they did achieve notable improvements in distinguishing between bottom baryon and bottom meson jets. This advancement could have implications for future research and experiments at the LHC, particularly in enhancing the precision of particle identification and classification.

This research was submitted on August 22, 2024, and revised on August 30, 2024. The full paper can be accessed via arXiv under the identifier arXiv:2408.12377.