KIGNet: Physics-motivated graph learning for explainable jet tagging

Researchers have developed KIGNet, a new graph neural network designed for explainable jet tagging in high-energy physics. KIGNet integrates kinematic variables like angular separation, relative transverse momentum, momentum fraction, and invariant mass squared into its classification process. The model demonstrates that it learns physically interpretable representations, with angular separation and relative transverse momentum being the most dominant factors in classification, aligning with theoretical predictions of QCD radiation. KIGNet achieves state-of-the-art performance on benchmark datasets, showing significant improvements in accuracy and representation quality compared to existing methods. AI

IMPACT Enhances explainability in physics-based AI models, potentially improving scientific discovery and model trust.

RANK_REASON The cluster contains a research paper detailing a new model and its performance on scientific benchmarks. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Aspen Open Jets
• Grad-CAM++
• JetClass
• KIGNet
• Kinematic Interaction Graph Network
• Lund jet plane
• Md Raqibul Islam

• paper
• model release
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →