Jet-Images: Computer Vision Inspired Techniques for Jet Tagging

摘要: We introduce a novel approach to jet tagging and classification through the use of techniques inspired by computer vision. Drawing parallels problem facial recognition in images, we define jet-image using calorimeter towers as elements image establish preprocessing methods. For processing step, develop discriminant for classifying jet-images derived Fisher analysis. The effectiveness technique is shown within context identifying boosted hadronic W boson decays with respect background quark- gluon- initiated jets. Using Monte Carlo simulation, demonstrate that performance this introduces additional discriminating power over other substructure approaches, gives significant insight into internal structure