Classifying Waveform MAGIC Telescope Data Using Graph Neural Networks

Jarred Green

TeVPA, Valencia, 6 November 2025

Max Planck Institute for Physics

jgreen@mpp.mpg.de

  1. The Data
  2. The Challenge
  3. Neutrinos?
  4. ML Playground

Outline

Jarred Green - jgreen@mpp.mpg.de

1. The Data

0. Raw data

1. Calibration

2. Cleaning

3. Parameterization

4. ML

5. Science

}

Dimensionality Reduction

Jarred Green - jgreen@mpp.mpg.de

0. Raw data

1. Calibration

2. Cleaning

3. Parameterization

4. ML

5. Science

Jarred Green - jgreen@mpp.mpg.de

1. The Data

0. Raw data

1. Calibration

2. Cleaning

3. Parameterization

4. ML

5. Science

Jarred Green - jgreen@mpp.mpg.de

1. The Data

Random Forests*

1039 \ \text{pixels} \\ \times 50 \ \text{time slices} \\ \times 2 \ \text{telescopes} \\ = 103900 \ \text{params}
{
	'size': 546.88637,
	'xc': -3.78983,
	'yc': 7.29427,
	'length': 7.76769,
	'width': 1.76480,
	'delta': -0.56875,
	 ...
}
  • Particle Type
  • Energy
  • Arrival Direction

Dimensionality

reduction

tl;dr

Jarred Green - jgreen@mpp.mpg.de

1. The Data

*Source: MAGIC Collaboration 2008 [arXiv:0709.3719]

Proton

Gamma

Muon

103900 \ \text{params per event (raw)}
\sim 300 \frac{\text{triggers}}{\text{s}}

Jarred Green - jgreen@mpp.mpg.de

\text{S/N ratio} \sim \frac{1}{10000}

2. The Challenge

3. Neutrinos

  • Particle Type
  • Energy
  • Arrival Direction

👀

Jarred Green - jgreen@mpp.mpg.de

  • A GNN framework for Neutrino observatories
  • Easy to adapt to other detectors

With Graph Neural Networks, IceCube can improve certain reconstruction metrics by 13-20%

Source: IceCube 2022 [arXiv:2209.03042]

3. Neutrinos

4. ML Playground

Jarred Green - jgreen@mpp.mpg.de

Jarred Green - jgreen@mpp.mpg.de

Particle Classification with

Model:

  • Binary classification
  • DynEdge starting with 12 NN
  • Each telescope initially separate in graph
  • 5-10% of training data

DynEdge

Confidence of gamma classification

Gammas

Protons

PRELIMINARY

4. ML Playground

Source: MAGIC 2016 [1409.5594]

At ~100 GeV, 80% of gammas have confidence higher than:

0.92

Jarred Green - jgreen@mpp.mpg.de

Direction Reconstruction with

Model:

  • 2nd place winner of IceCube Kaggle Competition
  • Transformer + DynEdge
  • Very robust model for neutrino direction reconstruction

DeepIce

Radial Error [deg]
Cumulative Fraction

MC 68% angular offset

~0.07º at 1 TeV

PRELIMINARY

4. ML Playground

RF method:

~0.07º at 1 TeV

Source: MAGIC 2016 [1409.5594]

MAGIC Raw Data

  • Contains full time evolution of events

  • Saves on data processing

  • Are naturally represented as graphs

Graph Neural Networks

  • Apply domain knowledge from IceCube
  • Basic neutrino models work on gamma-ray data
  • Robust performance with minimal modifications

Evidence for the benefits
of cross-domain collaboration

Jarred Green - jgreen@mpp.mpg.de

Future Motivation

70+ Telescopes!!

Cherenkov Telescope Array

Jarred Green - jgreen@mpp.mpg.de

Thank you!

Jarred Green - jgreen@mpp.mpg.de

Credit: G. Ceribella