Jet tagging via particle clouds

Abstract

How to represent a jet is at the core of machine learning on jet physics. Inspired by the notion of point cloud, we propose a new approach that considers a jet as an unordered set of its constituent particles, effectively a "particle cloud". Such particle cloud representation of jets is efficient in incorporating raw information of jets and also explicitly respects the permutation symmetry. Based on the particle cloud representation, we propose ParticleNet, a customized neural network architecture using Dynamic Graph CNN for jet tagging problems. The ParticleNet architecture achieves state-of-the-art performance on two representative jet tagging benchmarks and improves significantly over existing methods.ArXiv ePrint: 1902.08570

FPGA-Accelerated Machine Learning Inference as a Service for Particle Physics Computing

“…In this training, we treat the jets as 2D grayscale images in the η-φ plane and send them as input to the ResNet-50 algorithm. Jet images are created by summing jet constitutent p T in a 2D grid of 224 × 224 in η and φ units from −1.2 to 1.2 centered on the jet axis [32]. In order to apply the standard ResNet-50 architecture, the images are normalized such that each image has a range between 0 and 225 and duplicated 3 times, once for each RGB channel.…”

“…The typical particle physics models used for top tagging are often several orders of magnitude smaller than ResNet-50 in terms of the numbers of parameters and operations. However, it should be noted that the best-performing models to date (ResNeXt50 and a directed graph CNN) [32,24] are within a factor of a few in size with respect to the ResNet-50 model. We emphasize here that this study is a proof-of-concept for the physics performance and that there are many other very challenging, computationally intensive algo- rithms where machine learning is being explored.…”

Jets with electrons from boosted top quarks

“…Tagging these top jets, which contains all the decay products of hadronically decaying top quarks is quite a mature field. A plethora of tagging algorithms have been proposed which range from the substructure analyses [2][3][4][5][6][7][8][9][10][11][12][13] to methods taking full advantage of recent advances in the machine learning [14][15][16][17][18][19][20][21][22][23][24][25][26][27].…”

binary junipr: An Interpretable Probabilistic Model for Discrimination

“…A key question that is beginning to be addressed is: what is the optimal representation of the information in an event? Is it through analogy with images [1,2], natural-language processing [8,11], or set theory [19,20]? In many of these approaches, there is a competition between effectiveness in some task (e.g.…”

“…We found that untuned Binary Junipr comes close to state-of-the-art top discrimination. Specifically, Junipr achieves an AUC of 0.9810 ± 0.0002 as compared to 0.9819 ± 0.0001 attained using Particle Flow Networks [19], and 0.9848 reported for ParticleNet [20]; all significantly outperform traditional boosted top-tagging methods [29]. For a recent overview of machine learning in top tagging, see [30].…”