QCD-Aware Recursive Neural Networks for Jet Physics
This work addresses the challenge of improving jet physics analysis for high-energy physics researchers by providing a more accurate and efficient machine learning approach, though it is incremental in building upon existing analogies in the field.
The authors tackled the problem of analyzing particle jets in high-energy physics by introducing a novel class of recursive neural networks that analogize QCD processes to natural language processing, where four-momenta are treated as words and jet clustering as sentence parsing. Their method achieved significantly higher accuracy and data efficiency compared to previous image-based networks, with experiments demonstrating its flexibility for task-specific jet embeddings and the first event-level classifier for LHC events.
Recent progress in applying machine learning for jet physics has been built upon an analogy between calorimeters and images. In this work, we present a novel class of recursive neural networks built instead upon an analogy between QCD and natural languages. In the analogy, four-momenta are like words and the clustering history of sequential recombination jet algorithms is like the parsing of a sentence. Our approach works directly with the four-momenta of a variable-length set of particles, and the jet-based tree structure varies on an event-by-event basis. Our experiments highlight the flexibility of our method for building task-specific jet embeddings and show that recursive architectures are significantly more accurate and data efficient than previous image-based networks. We extend the analogy from individual jets (sentences) to full events (paragraphs), and show for the first time an event-level classifier operating on all the stable particles produced in an LHC event.