Lorentz Boost Networks: Autonomous Physics-Inspired Feature Engineering

Abstract

We present a two-stage neural network architecture that enables a fully autonomous and comprehensive characterization of collision events by exclusively exploiting the four momenta of final-state particles. We refer to the first stage of the architecture as Lorentz Boost Network (LBN). The LBN allows the creation of particle combinations representing rest frames. The LBN also enables the formation of further composite particles, which are then transformed into said rest frames by Lorentz transformation. The properties of the composite, transformed particles are compiled in the form of characteristic variables that serve as input for a subsequent network. This second network has to be configured for a specific analysis task such as the separation of signal and background events. Using the example of the classification of ttH and ttbb events, we compare the separation power of the LBN approach with that of domain-unspecific deep neural networks (DNN). We observe leading performance with the LBN, even though we provide the DNNs with extensive additional input information beyond the particle four momenta. Furthermore, we demonstrate that the LBN forms physically meaningful particle combinations and autonomously generates suitable characteristic variables.

References in corpus (33)

• Adam: A Method for Stochastic Optimization
• Observation of a new boson at a mass of 125 GeV with the CMS experiment at the LHC
• TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
• An Introduction to PYTHIA 8.2
• FastJet user manual
• Deep Residual Learning for Image Recognition
• DELPHES 3, A modular framework for fast simulation of a generic collider experiment
• Searching for Exotic Particles in High-Energy Physics with Deep Learning
• Learning Particle Physics by Example: Location-Aware Generative Adversarial Networks for Physics Synthesis
• Deep learning in color: towards automated quark/gluon jet discrimination
• Jet-Images -- Deep Learning Edition
• Deep-learning Top Taggers or The End of QCD?
• Jet Substructure Classification in High-Energy Physics with Deep Neural Networks
• A Convolutional Neural Network Neutrino Event Classifier
• Deep-learned Top Tagging with a Lorentz Layer
• Jet Flavor Classification in High-Energy Physics with Deep Neural Networks
• Top pair invariant mass distribution: a window on new physics
• Decorrelated Jet Substructure Tagging using Adversarial Neural Networks
• QCD-Aware Recursive Neural Networks for Jet Physics
• Enhanced Higgs to ${\tau }^{+}{\tau }^{-}$ Searches with Deep Learning
• Application of Deep Learning methods to analysis of Imaging Atmospheric Cherenkov Telescopes data
• How Much Information is in a Jet?
• A Deep Neural Network for Pixel-Level Electromagnetic Particle Identification in the MicroBooNE Liquid Argon Time Projection Chamber
• A Deep Learning-based Reconstruction of Cosmic Ray-induced Air Showers
• Search for $\mathrm{t}\overline{\mathrm{t}}$H production in the H $\to$ $\mathrm{b}\overline{\mathrm{b}}$ decay channel with leptonic $\mathrm{t}\overline{\mathrm{t}}$ decays in proton-proton collisions at $\sqrt{s}=$ 13 TeV
• Lorentz Boost Networks: Autonomous Physics-Inspired Feature Engineering
• Fox-Wolfram Moments in Higgs Physics
• Event shape criteria for single-lepton top signals
• N=2* Super-Yang-Mills Theory at Strong Coupling
• Deep Neural Networks for Energy and Position Reconstruction in EXO-200
• Measurement of the ${\rm Y}$ polarizations in $pp$ collisions at $\sqrt{s}$=7 and 8TeV
• Jet-Parton Assignment in ttH Events using Deep Learning
• Search for top quark partners with charge 5/3 in proton-proton collisions at sqrt(s) = 13 TeV

Cited by in corpus (34)

• ParticleNet: Jet Tagging via Particle Clouds
• The Machine Learning Landscape of Top Taggers
• An Efficient Lorentz Equivariant Graph Neural Network for Jet Tagging
• DisCo Fever: Robust Networks Through Distance Correlation
• Lorentz Group Equivariant Neural Network for Particle Physics
• A robust anomaly finder based on autoencoders
• When the Machine Chimes the Bell: Entanglement and Bell Inequalities with Boosted $t\bar{t}$
• A guide for deploying Deep Learning in LHC searches: How to achieve optimality and account for uncertainty
• Casting a graph net to catch dark showers
• Lorentz Boost Networks: Autonomous Physics-Inspired Feature Engineering
• Symmetries, Safety, and Self-Supervision
• Automating the Construction of Jet Observables with Machine Learning
• Measuring QCD Splittings with Invertible Networks
• Energy-weighted Message Passing: an infra-red and collinear safe graph neural network algorithm
• Jet Flavour Tagging for Future Colliders with Fast Simulation
• Explainable Equivariant Neural Networks for Particle Physics: PELICAN
• Shared Data and Algorithms for Deep Learning in Fundamental Physics
• Machine Learning in High Energy Physics: A review of heavy-flavor jet tagging at the LHC
• CapsNets Continuing the Convolutional Quest
• A Detailed Study of Interpretability of Deep Neural Network based Top Taggers
• Invisible Higgs search through Vector Boson Fusion: A deep learning approach
• Searches for Higgs boson production through decays of heavy resonances
• Does Lorentz-symmetric design boost network performance in jet physics?
• BIP: Boost Invariant Polynomials for Efficient Jet Tagging
• Comparing Point Cloud Strategies for Collider Event Classification
• Machine learning classification: case of Higgs boson CP state in H to tau tau decay at LHC
• Search for Higgs boson pair production in the $\mathrm{b}\overline{\mathrm{b}}$W${}^{+}$W${}^{-}$ decay mode in proton-proton collisions at $\sqrt{s}$ = 13 TeV
• Resolving Combinatorial Ambiguities in Dilepton $t\bar{t}$ Event Topologies with Neural Networks
• Attention to the strengths of physical interactions: Transformer and graph-based event classification for particle physics experiments
• A rotation-equivariant graph neural network for learning hadronic SMEFT effects
• Strange Jet Tagging
• Top-philic Machine Learning
• Solving Combinatorial Problems at Particle Colliders Using Machine Learning
• LHC Higgs CP Sensitive Observables in H to tau^+ tau^-; tau^pm to (3pi)^pm nu and Machine Learning Benefits