Jet-Images -- Deep Learning Edition

TL;DR

The paper investigates identifying highly boosted W bosons by treating jets as images and applying deep learning architectures. It demonstrates that convolutional and MaxOut networks can outperform traditional physics-driven features like jet mass and τ21, while also introducing visualization tools to interpret what the networks learn about jet substructure and color-flow. Through systematic phase-space manipulations (general, uniform, highly restricted) and comprehensive analysis of filters, activations, and per-pixel correlations, it shows that networks capture information beyond conventional variables. The work provides a methodological framework for interpreting deep networks in jet physics and suggests practical pathways to enhance LHC analyses with image-based tagging.

Abstract

Building on the notion of a particle physics detector as a camera and the collimated streams of high energy particles, or jets, it measures as an image, we investigate the potential of machine learning techniques based on deep learning architectures to identify highly boosted W bosons. Modern deep learning algorithms trained on jet images can out-perform standard physically-motivated feature driven approaches to jet tagging. We develop techniques for visualizing how these features are learned by the network and what additional information is used to improve performance. This interplay between physically-motivated feature driven tools and supervised learning algorithms is general and can be used to significantly increase the sensitivity to discover new particles and new forces, and gain a deeper understanding of the physics within jets.

Figures (24)

• Figure 1: The distributions of the jet mass (top left), $\tau_{21}$ (top right) and the $\Delta R$ between subjets (bottom) for signal (blue) and background (red) jets.
• Figure 2: The average jet image for signal $W$ jets (top) and background QCD jets (bottom) before (left) and after (right) applying the rotation, re-pixelation, and inversion steps of the pre-processing. The average is taken over images of jets with $240$ GeV $<p_T<$ 260 GeV and 65 GeV $<$ mass $<$ 95 GeV.
• Figure 3: The distribution of the image mass after various states of pre-processing for signal jets (left) and background jets (right). The No pixelation line is the jet mass without any detector granularity and without any pre-processing. Only pixelation has only detector granularity but no pre-processing and all subsequent lines have this pixelation applied as well as translation to center the image at the origin. The translation is called naive when the energy is used as the pixel intensity instead of the pixel transverse momentum. Flip denotes the parity inversion operation and the $p_T^2$ norm is a $L^2$ normalization scheme. The naive translation and the $I^2$ normalization image masses are both multiplied by constants so that the centers of the distribution are roughly in the same location as for the other distributions.
• Figure 4: The tradeoff between $W$ boson (signal) jet efficiency and inverse QCD (background) efficiency for various pre-processing algorithms applied to the jet (images). The No pixelation line is the jet mass without any detector granularity and without any pre-processing. Only pixelation has only detector granularity but no pre-processing and all subsequent lines have this pixelation applied as well as translation to center the image at the origin. The translation is called naive when the energy is used as the pixel intensity instead of the pixel transverse momentum. Flip denotes the parity inversion operation and the $p_T^2$ norm is a $L^2$ normalization scheme.
• Figure 5: The convolution neural network concept as applied to jet-images.