mlsanityData to 13 Oct 2026

Explaining machine-learned particle-flow reconstruction

arXiv 2111.12840 ↗First posted 24 Nov 2021physics.data-an cs.LG hep-excited by 12

Not on the map. Why?

Abstract (arXiv)

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.

Key facts. Cited by 12 papers in the map (data through 13 Oct 2026). Builds on 12 of its 61 references that are in the map. First posted on arXiv 24 Nov 2021.

Builds on

12 of 61 references.

References · 12

  1. 2019Fast Graph Representation Learning with PyTorch Geometriccited by 2,422
  2. 2016Convolutional Neural Networks on Graphs with Fast Localized Spectral Filteringcited by 2,086
  3. 2019GNNExplainer: Generating Explanations for Graph Neural Networkscited by 736
  4. 2020Higher-Order Explanations of Graph Neural Networks via Relevant Walkscited by 138
  5. 2019ParticleNet: Jet Tagging via Particle Cloudscited by 64
  6. 2021Particle Cloud Generation with Message Passing Generative Adversarial Networkscited by 45
  7. 2019Learning representations of irregular particle-detector geometry with distance-weighted graph networkscited by 43
  8. 2021MLPF: Efficient machine-learned particle-flow reconstruction using graph neural networkscited by 36
  9. 2020Object condensation: one-stage grid-free multi-object reconstruction in physics detectors, graph and image datacited by 23
  10. 2021Charged particle tracking via edge-classifying interaction networkscited by 19
  11. 2020Explainable AI for ML jet taggers using expert variables and layerwise relevance propagationcited by 14
  12. 2020Track Seeding and Labelling with Embedded-space Graph Neural Networkscited by 12

Built on by

Cited by year (2026 to 13 Oct)

Citing papers per year: 2022: 7; 2023: 2; 2024: 0; 2025: 2; 2026: 1.

12 papers.

  1. 2025Interpretable Machine Learning in Physics: A Reviewcited by 13
  2. 2022Graph Neural Networks in Particle Physics: Implementations, Innovations, and Challengescited by 12
  3. 2022Explainable AI for High Energy Physicscited by 7
  4. 2023Progress towards an improved particle flow algorithm at CMS with machine learningcited by 7
  5. 2022A Detailed Study of Interpretability of Deep Neural Network based Top Taggerscited by 6
  6. 2022Do graph neural networks learn traditional jet substructure?cited by 4
  7. 2022FAIR AI Models in High Energy Physicscited by 4
  8. 2022Data Science and Machine Learning in Educationcited by 3
  9. 2022Learning Tree Structures from Leaves For Particle Decay Reconstructioncited by 3
  10. 2025Mixture-of-Experts Graph Transformers for Interpretable Particle Collision Detectioncited by 3
  11. 2023Graph Structure from Point Clouds: Geometric Attention is All You Needcited by 2
  12. 2026Dissecting Jet-Tagger Through Mechanistic Interpretabilitycited by 0