Interaction Feature-Guided Explainable Particle Chebyshev Networks (E-PCN) for Jet Tagging

Note: This project is ongoing and subject to continuous advancements and modifications.

---

Overview

This repository focuses on jet tagging—classifying collimated sprays of particles (jets) from high-energy collisions and associating them with their originating particles. It is an ongoing project of the Center for Computational and Data Sciences (CCDS) in collaboration with the Department of Theoretical Physics, University of Dhaka and has strong ties to CERN.

We enhance the Particle Chebyshev Network (PCN) architecture by integrating physics-motivated interaction features derived from particle 4-momentum vectors and inspired by the Lund jet plane. This approach improves discrimination capability for jet classification tasks using the JetClass dataset.

---

Key Features

• Physics-motivated feature integration into PCN
• End-to-end training with CERN Open Data (JetClass)
• Evaluation with standard HEP metrics

---

Physics-Motivated Interaction Features

These features capture key kinematic properties of particle interactions that are relevant for jet substructure and tagging tasks. The logarithmic transformation addresses the long-tail distributions of these variables in high-energy physics, making them more suitable for machine-learning models.

Feature Definitions

Feature	Formula	Description
$\Delta$	$\Delta = \sqrt{(y_a - y_b)^2 + (\phi_a - \phi_b)^2}$	Angular separation in rapidity–azimuth plane
$k_T$	$k_T = \min(p_{T,a}, p_{T,b}) \cdot \Delta$	Transverse momentum scale (soft / collinear observable)
$z$	$z = \frac{\min(p_{T,a}, p_{T,b})}{p_{T,a} + p_{T,b}}$	Momentum fraction (energy-sharing parameter)
$m^2$	$m^2 = (E_a + E_b)^2 - |\mathbf{p}_a + \mathbf{p}_b|^2$	Squared invariant mass of the particle pair

Notation

Symbol	Definition
$y_i$	Rapidity of particle $i$
$\phi_i$	Azimuthal angle of particle $i$
$p_{T,i}$	Transverse momentum of particle $i$: $p_{T,i} = \sqrt{p_{x,i}^2 + p_{y,i}^2}$
$p_{i}$	Momentum 3-vector of particle $i$: $p_{i} = (p_{x,i}, p_{y,i}, p_{z,i})$
$E_i$	Energy of particle $i$
$|\cdot|$	Euclidean norm

Since these variables typically have a long-tail distribution, we take the logarithm and use $(\ln \Delta, \ln k_T, \ln z, \ln m^2)$ as the interaction features for each particle pair.

Physical Motivation

These features bias the model toward fine-grained, QCD-informed inter-particle dependencies. This choice of features follows the work of Frédéric A. Dreyer & Huilin Qu (2021). For details see Jet tagging in the Lund plane with graph networks: Link.

---

Getting Started

Installation

git clone https://github.com/Adrita-Khan/Jet-Tagging.git
cd Jet-Tagging
pip install -r requirements.txt

Repository Structure

raqib-pcn-experiments/
├── raqib-pcn-experiments/   # Main experiment folder
├── .gitattributes           # Git configuration file
├── LICENSE                  # License file
├── README.md                # Project overview and instructions
├── pythia-data-gen.md       # Data generation tutorial for Pythia
├── pythia-installation.md   # Installation guide for Pythia
├── pythia-jet-tagging-data-generation-tutorial.md  # Jet tagging data generation tutorial
├── pythia-python-guide.md   # Python guide for Pythia
└── requirements.txt         # List of dependencies

---

References

#	Title	Link
1	JetClass: A Large-Scale Dataset for Deep Learning in Jet Physics	Springer
2	Particle Chebyshev Network (PCN)	PMLR
3	The Lund Jet Plane	Springer
4	Jet Substructure and Machine Learning	IOP Science
5	Jet Tagging via Particle Clouds	Physical Review D
6	PCN-Jet-Tagging GitHub Repository	GitHub

---

Acknowledgements

• CERN Open Data Portal for providing high-quality collision data
• Original PCN authors for the base architecture
• Center for Computational and Data Sciences (CCDS)

---

License

MIT License — see LICENSE for details.

---

Contact

For any inquiries or feedback, please contact:

Name	Email	LinkedIn	Twitter
Adrita Khan	[email protected]	LinkedIn	Twitter
Md Raqibul Islam	[email protected]	LinkedIn	Twitter

---

Maintained by the CCDS Team