Explaining machine-learned particle-flow reconstruction

Abstract

The particle-flow (PF) algorithm is used in general-purpose particle detectors to reconstruct a comprehensive particle-level view of the collision by combining information from different subdetectors. A graph neural network (GNN) model, known as the machine-learned particle-flow (MLPF) algorithm, has been developed to substitute the rule-based PF algorithm. However, understanding the model's decision making is not straightforward, especially given the complexity of the set-to-set prediction task, dynamic graph building, and message-passing steps. In this paper, we adapt the layerwise-relevance propagation technique for GNNs and apply it to the MLPF algorithm to gauge the relevant nodes and features for its predictions. Through this process, we gain insight into the model's decision-making.

arXiv comment5 pages, 3 figures. Accepted to Machine Learning for Physical Sciences NeurIPS 2021 workshop

References in corpus (25)

• PYTHIA 6.4 Physics and Manual
• PYTHIA 6.2 Physics and Manual
• A Brief Introduction to PYTHIA 8.1
• DELPHES 3, A modular framework for fast simulation of a generic collider experiment
• Fast Graph Representation Learning with PyTorch Geometric
• Particle-flow reconstruction and global event description with the CMS detector
• Description and performance of track and primary-vertex reconstruction with the CMS tracker
• GNNExplainer: Generating Explanations for Graph Neural Networks
• ParticleNet: Jet Tagging via Particle Clouds
• Jet reconstruction and performance using particle flow with the ATLAS Detector
• Higher-Order Explanations of Graph Neural Networks via Relevant Walks
• Performance of the ATLAS Track Reconstruction Algorithms in Dense Environments in LHC Run 2
• Learning representations of irregular particle-detector geometry with distance-weighted graph networks
• ABCNet: An attention-based method for particle tagging
• Graph Neural Networks for Particle Reconstruction in High Energy Physics detectors
• Interaction networks for the identification of boosted $H\to b\overline{b}$ decays
• The Tracking Machine Learning challenge : Accuracy phase
• Measurement of the Diffractive Structure Function F_2^D(4) at HERA
• Charged particle tracking via edge-classifying interaction networks
• Novel deep learning methods for track reconstruction
• HepSim: a repository with predictions for high-energy physics experiments
• Measurement of sigma(ppbar -> Z + X) Br(Z -> tau+tau-) at sqrt(s)=1.96 TeV
• Explainable machine learning of the underlying physics of high-energy particle collisions
• Track Seeding and Labelling with Embedded-space Graph Neural Networks
• A Search for Supersymmetric Higgs Bosons in the Di-Tau Decay Mode in Proton Anti-Proton Collisions at sqrt(s)=1.8 TeV