Jet-Images: Computer Vision Inspired Techniques for Jet Tagging

TL;DR

The paper presents a computer vision-inspired framework for jet tagging by representing jets as fixed-size images derived from calorimeter towers. It preprocesses these jet-images (noise reduction, alignment, equalization) and applies a regularized Fisher Linear Discriminant to produce a Fisher-jet discriminant, $D[A] = \sum_{k=1}^{N^{2}} \bar{F}_{k} \cdot \bar{A}_{k}$, enabling interpretable, linear classification of jet classes. Through Monte Carlo studies of boosted $W$ jets versus QCD jets, the method demonstrates competitive or superior discrimination compared to N-subjettiness across $p_T$ and subjet separation bins, with visualizable discriminant features that illuminate jet structure. The approach offers a transparent, fast, and extensible tool for jet substructure analysis and can be adapted to other hadronic final states, providing a bridge between jet physics and computer vision.

Abstract

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

Figures (6)

• Figure 1: The preprocessing of jet-images and the impact on the average jet-image for W jets in which the leading jet with $p_{T}$ between $200$ and $250$ GeV. Note that the grid in figure \ref{['subfig:proj']} appears shifted down to represent the jet-image before translation, which is subsequently translated such that the leading subjet lies in the location ($Q_{1}\sim1.5, Q_{2}\sim1.25$) as see in the final average jet-image of figure \ref{['subfig:proj_avg']}.
• Figure 2: A Fisher's linear discriminant presented as an image (left) and the distributions of the discriminant output when applied to W-jets and Light-jets (right), when the FLD is trained on jets with $p_{T} \in [250, 300]$ GeV, mass $M \in [65, 95]$ GeV, and separation between subjets of $\Delta R \in [0.6, 0.8]$.
• Figure 3: Background rejection vs signal efficiency curves obtained by training discriminants on samples with $p_{T} \in [250, 300]$ GeV (left), and the W-jet efficiency at fixed $\times10$ QCD jet rejection versus jet $p_{T}$ (right).
• Figure 4: The Fisher-jet discriminant output for the bin $p_{T} \in [250, 300]$ GeV and $\Delta R \in [0.6, 0.8]$ plotted against: (\ref{['subfig:nsubjetiness_w']}) the output of N-subjettiness for W-jets, (\ref{['subfig:nsubjetiness_l']}) the output of N-subjettiness for QCD jets, (\ref{['subfig:mass_w']}) the invariant mass of the jet for W-jets, and (\ref{['subfig:mass_l']}) the invariant mass of the jet for QCD jets.
• Figure 5: The Fisher-jet discriminant output, trained on a sample of Pythia8$W$ and QCD jets applied to Pythia8 and Herwig++ samples.