Domain-Adversarial Graph Neural Networks for $\Lambda$ Hyperon Identification with CLAS12

TL;DR

This paper introduces a novel domain-adversarial graph neural network approach for identifying $ ext{Lambda}$ hyperon events in high energy physics, significantly improving purity and providing a benchmark for future experiments.

Contribution

It presents a new GNN-based method with domain-adversarial training for $ ext{Lambda}$ hyperon identification, enhancing purity and establishing a benchmark for future event tagging.

Findings

Purity of $ ext{Lambda}$ yield increased by a factor of 1.95

Domain-adversarial training improved purity by a factor of 1.82

Provides a benchmark for event tagging in high energy physics

Abstract

Machine learning methods and in particular Graph Neural Networks (GNNs) have revolutionized many tasks within the high energy physics community. We report on the novel use of GNNs and a domain-adversarial training method to identify $\Lambda$ hyperon events with the CLAS12 experiment at Jefferson Lab. The GNN method we have developed increases the purity of the $\Lambda$ yield by a factor of $1.95$ and by $1.82$ using the domain-adversarial training. This work also provides a good benchmark for developing event tagging machine learning methods for the $\Lambda$ and other channels at CLAS12 and other experiments, such as the planned Electron Ion Collider.