Energy-weighted Message Passing: an infra-red and collinear safe graph neural network algorithm

Abstract

Hadronic signals of new-physics origin at the Large Hadron Collider can remain hidden within the copiously produced hadronic jets. Unveiling such signatures require highly performant deep-learning algorithms. We construct a class of Graph Neural Networks (GNN) in the message-passing formalism that makes the network output infra-red and collinear (IRC) safe, an important criterion satisfied within perturbative QCD calculations. Including IRC safety of the network output as a requirement in the construction of the GNN improves its explainability and robustness against theoretical uncertainties in the data. We generalise Energy Flow Networks (EFN), an IRC safe deep-learning algorithm on a point cloud, defining energy weighted local and global readouts on GNNs. Applying the simplest of such networks to identify top quarks, W bosons and quark/gluon jets, we find that it outperforms state-of-the-art EFNs. Additionally, we obtain a general class of graph construction algorithms that give structurally invariant graphs in the IRC limit, a necessary criterion for the IRC safety of the GNN output.

Added discussion of output stability for deeper network. Updated references and corrected typos. Accepted version in JHEP

References in corpus (15)

• The automated computation of tree-level and next-to-leading order differential cross sections, and their matching to parton shower simulations
• An Introduction to PYTHIA 8.2
• Fast Graph Representation Learning with PyTorch Geometric
• Jet substructure as a new Higgs search channel at the LHC
• Jet substructure as a new Higgs search channel at the LHC
• Power Counting to Better Jet Observables
• Sudakov Safety in Perturbative QCD
• Parton Shower Uncertainties in Jet Substructure Analyses with Deep Neural Networks
• Anomaly detection with Convolutional Graph Neural Networks
• Uncertainty Aware Learning for High Energy Physics
• Lorentz Boost Networks: Autonomous Physics-Inspired Feature Engineering
• Towards Machine Learning Analytics for Jet Substructure
• Higgs boson tagging with the Lund jet plane
• Particle Convolution for High Energy Physics
• Investigating top tagging with Y${}_{\text{m}}$-Splitter and N-subjettiness

Cited by in corpus (20)

• An Efficient Lorentz Equivariant Graph Neural Network for Jet Tagging
• Exploring QCD matter in extreme conditions with Machine Learning
• A Fragmentation Approach to Jet Flavor
• Explainable Equivariant Neural Networks for Particle Physics: PELICAN
• Lorentz group equivariant autoencoders
• Unbinned Deep Learning Jet Substructure Measurement in High $Q^2$ ep collisions at HERA
• Improved Constraints on Effective Top Quark Interactions using Edge Convolution Networks
• Does Lorentz-symmetric design boost network performance in jet physics?
• Equivariant, Safe and Sensitive -- Graph Networks for New Physics
• Is infrared-collinear safe information all you need for jet classification?
• Comparing Point Cloud Strategies for Collider Event Classification
• Hypergraphs in LHC Phenomenology -- The Next Frontier of IRC-Safe Feature Extraction
• Interpretable deep learning models for the inference and classification of LHC data
• Safe but Incalculable: Energy-weighting is not all you need
• Unveiling the Secrets of New Physics Through Top Quark Tagging
• A rotation-equivariant graph neural network for learning hadronic SMEFT effects
• Foundations of automatic feature extraction at LHC--point clouds and graphs
• Probing sub-TeV Higgsinos aided by a ML-based top tagger in the context of Trilinear RPV SUSY
• Graph theory inspired anomaly detection at the LHC
• Stable and Interpretable Jet Physics with IRC-Safe Equivariant Feature Extraction