Reconstruction of muon bundles in KM3NeT detectors using machine learning methods
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: High Energy Physics: Experiment. The day in high-energy experiment: detectors, searches and limits.
Kai: We're looking at the paper: "Reconstruction of muon bundles in KM3NeT detectors using machine learning methods".
Mira: Here's the gist of the paper. The work explores the potential of machine learning models for the reconstruction of muon bundles in KM3NeT detectors, which is important because it aims to pinpoint shortcomings in current theoretical models used to compute muon flux.
Kai: Reconstruction Targets The machine learning models are trained to predict several important observables related to muon bundles, including the total number of muons in a bundle, their total energy, and even the energy of the primary cosmic ray.
Mira: The reconstruction process is performed separately for two detector configurations for both ARCA and ORCA.
Kai: Data preprocessing involved extracting 46 features, event weights, and target values before the division into training, validation, and test datasets.
Title and authors: Kai: Let's talk about the title and who wrote this. "Reconstruction of muon bundles in KM3NeT detectors using machine learning methods". It’s very direct about what they did.
Mira: The authors are Piotr Kalaczyński on behalf of the KM3NeT Collaboration, which is a big international effort building those neutrino telescopes at the bottom of the Mediterranean Sea.
Lev: So it’s a collaboration effort, which means they're dealing with data coming from two different detector setups, ARCA and ORCA. That adds complexity to the whole reconstruction process.
Kai: Right. And what this paper suggests is that machine learning is a tool we can use to study these muon bundles better than maybe the existing theoretical predictions alone allow us to do right now.
Mira: The implication here is that by reconstructing things like the total number of muons and the primary cosmic ray energy, we can actually start seeing where those current theoretical models for cosmic ray interactions need adjustment.
Lev: I wonder how much overhead this ML system would really land on real hardware. We're talking about training these models on data from two different configurations and then deploying them for real-time reconstruction, right?
The paper's summary: Kai: Now, let’s go over the summary of what they actually did in this paper. They took data collected with intermediate detector configurations of ARCA and ORCA, along with simulated data from the final setups, and used it to train a machine learning model.
Mira: The goal of that training was to predict several key observables: the total number of muons in a bundle, their total energy, and even the energy of that primary cosmic ray.
Lev: So they’re not just looking at one thing; they’re trying to get a whole picture of the event by predicting these three different physical properties simultaneously.
Kai: That’s right. They also present how this reconstruction helps in understanding the muon puzzle, which is that persistent deficit in muon events observed at high energies and high multiplicities compared to what current theories predict.
Mira: The paper suggests that being able to reconstruct those quantities—the number, the energy of the bundle, and the primary cosmic ray energy—can help pinpoint exactly where those current theoretical models are lacking.
Lev: So they’re trying to use this ML system as a diagnostic tool for theory rather than just a data processing pipeline.
The paper's improvements: Kai: They talk about some specific improvements they found in their approach, and those are quite interesting from an engineering standpoint. One big thing is that they identified LightGBM as the best performer specifically for reconstructing the bundle energy.
Mira: LightGBM was better than other models they tried, like XGBoost or maybe others, for predicting the total energy of a muon bundle in ARCA115 data, achieving a median weighted coefficient of determination R squared score of zero point six nine eight plus or minus zero point ten.
Lev: A coefficient of determination around zero point seven means it’s not perfect, but it’s definitely doing the job better than the baseline methods they compared it against, right?
Kai: It is, and they also found a way to select features that makes the model more robust. They said requiring positive importance for all features provided the best results when choosing which inputs to feed into LightGBM.
Mira: So not just picking a good algorithm, but also finding a smart way to pick the right input data—the features—to make it work better, which points toward some deeper understanding of what matters in these muon bundle events.
Lev: That feature selection idea is interesting because it tries to prune the noise before the main learning happens, which should reduce computational overhead if it works well.
Conclusion: Kai: So to wrap up, this paper on "Reconstruction of muon bundles in KM3NeT detectors using machine learning methods" shows that machine learning is a viable path for improving our understanding of cosmic ray interactions at the KM3NeT detectors.
Mira: The main implication is that by training models like LightGBM to reconstruct things like the total number of muons, their total energy, and even the primary cosmic ray energy, we can start figuring out where current theoretical models for muon flux are missing pieces.
Lev: From an engineering side, it suggests that these ML systems could potentially be used to probe physics beyond what we can get from pure simulation alone.
Kai: They also point out some areas where they think the system could get better, like using even lower-level data for feature extraction to minimize information loss when reconstructing the primary cosmic ray energy.
Mira: And they mention that looking into other models, like graph neural networks, might give them valuable results moving forward for this kind of reconstruction.
Lev: I think if you were to run this on real hardware, you'd have to worry about how much overhead those complex models add compared to simpler methods we’ve used before.
Kai: That’s the practical hurdle there, but the potential for gaining insight into the muon puzzle is certainly worth looking into.
Mira: Anyway, this paper gives us a concrete example of how data-driven techniques can help bridge gaps between what we observe and what our current theories predict.
KM3NeT Collaboration · Nicolaus Copernicus Astronomical Center · Polish Academy of Sciences · Center of Excellence in Artificial Intelligence, AGH University of Krakow
hep-ex, cs.LG
Submitted: 2025-03-03
Updated: 2025-03-03
Comments: 16 pages, 12 figures, proceedings of the 2nd International Workshop on Machine Learning and Quantum Computing Applications in Medicine and Physics, accepted for publication in Computer Science Journal
Journal ref: Comput. Sci. 26 (SI), 7062 (2025)
DOI: 10.7494/csci.2025.26.SI.7062
Code: https://github.com/carmeloevoli/The
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 82/100
The gist: The work explores the potential of machine learning models for the reconstruction of muon bundles in KM3NeT detectors…
Key concepts
- Muon Bundle Reconstruction
- This involves using machine learning to analyze the signals from muons clustered together in a detector. The goal is to accurately determine key physical characteristics like the total number of muons, their combined energy, and even the energy of the original cosmic ray that created them. This is crucial for understanding how cosmic rays interact with matter.
- LightGBM
- LightGBM was identified as the most effective machine learning model for reconstructing bundle energy. It uses gradient boosting decision trees to make predictions. Its high performance, particularly in the ARCA115 configuration, suggests it is a powerful tool for accurately estimating complex physical quantities from detector data.
- Feature Importance
- Feature importance measures how much each input variable contributes to the model's accuracy. In this study, it was used to select the most relevant data points. By prioritizing features with positive importance, researchers improved the model's performance and robustness across different energy scales.
Terminology
Summary
The work explores the potential of machine learning models for the reconstruction of muon bundles in KM3NeT detectors, which is important because it aims to pinpoint shortcomings in current theoretical models used to compute muon flux. p. 1 With a dedicated muon simulation, it was possible
to train a machine learning model to reconstruct several important observables, such
as the total number of muons in a bundle, their total energy and the energy of the
primary cosmic ray. Measuring those quantities can help to pinpoint the shortcomings
of the current theoretical models used to compute the muon flux.Where the paper says it
Reconstruction Targets
The machine learning models are trained to predict several important observables related to muon bundles, including the total number of muons in a bundle, their total energy, and even the energy of the primary cosmic ray. The reconstruction process is performed separately for two detector configurations for both ARCA and ORCA. p. 4 The data used to produce
the results is summarized in Tab. The reconstruction was performed separately
for two detector configurations for both ARCA and ORCA. The data preprocessing included extracting 46 features (described in detail in
[13]), event weights and target (observable) values. Each feature had its mean removed
and was scaled to unit variance (standard scaling).Where the paper says it
Where the paper says it
Model Selection and Feature Importance
The best performer for bundle energy reconstruction was identified as LightGBM. Feature importance was defined as the mean decrease of the R2-score upon removal of a given feature. p. 6 Not all were
equally important and several were strongly inter-correlated, which could indicate
redundancy. The feature importance was defined as the mean decrease of the R2
-
score upon removal of a given feature. The comparison between those four categories for ARCA115 is presented in Fig. The feature selection method requiring the positive importance of each feature
provided the best results.Where the paper says it
Where the paper says it
Performance Across Energy Scales
The reconstructed bundle energy was most accurate for ARCA115—this configuration instruments the largest volume and hence can observe the largest energy deposits. There is a threshold energy of about 1 TeV below which the model cannot make reasonable predictions, and towards the highest energies, there is a growing underestimation of the true bundle energy. p. 7 There is a threshold energy of about 1 TeV below which the model cannot make reasonable predictions. Towards the highest energies, there is a growing underestimation
of the true bundle energy. The comparison of the experimental data against the MC simulations for the
reconstructed bundle energy is shown in Fig. The simulations reproduce the data
within the few-TeV energy scale, however, at higher energies discrepancies are visible,
similarly as in Fig.Where the paper says it
Where the paper says it
Primary Cosmic Ray Energy and Multiplicity
The reconstruction of primary CR energy showed promising results above 100 PeV, suggesting KM3NeT may contribute to the primary CR measurements at the highest energies, notably to the study of the Greisen-Zatsepin-Kuzmin (GZK) cutoff. Multiplicity reconstruction is more accurate for ARCA, thanks to the typical muon signature in a detector (elongated track). p. 0 This work explores the potential of machine
learning models for the reconstruction of muon bundles, which are multi-muon
events. For this, data collected with intermediate detector configurations of
ARCA and ORCA was used in addition to simulated data from the envisaged
final configurations of those detectors. Before
reaching the final result of reconstruction, several loops of training and validation
were performed to select the best machine learning (ML) regression model for the
task, pick the features that improve the performance and tune the hyperparameters
(internal parameters of the model). The regressor selection was performed only for
bundle energy; the result was used for primary CR energy and muon multiplicity
reconstruction.Where the paper says it
Where the paper says it
Conclusion and Future Work
The potential for further improvement lies in achieving a better understanding of the discrepancies between the data and MC. The model training could potentially benefit from extracting features from even lower-level data to ensure minimal information loss. p. 5 The data preprocessing included extracting 46 features (described in detail in
[13]), event weights and target (observable) values. Each feature had its mean removed
and was scaled to unit variance (standard scaling). Both
detectors are already operational in their intermediate states and collect valuable data, including the measurements of the muons produced by cosmic ray
interactions in the atmosphere. This work explores the potential of machine
learning models for the reconstruction of muon bundles, which are multi-muon
events. The reconstruction was performed separately
for two detector configurations for both ARCA and ORCA. In each case, it was a
full building block (ARCA115, ORCA115) and an intermediate state with 6 detection
units installed (ARCA6, ORCA6).Where the paper says it
Where the paper says it
Where the paper says it
References
[1] Aiello S., et al.: Atmospheric muons measured with the KM3NeT detectors in comparison with updated numeric predictions. p. 2 Its more dense instrumentation allows for precise studies of atmospheric neutrino
oscillations in the GeV-TeV energy range. The design of the KM3NeT detectors is
summarised in Fig.Where the paper says it
[2] Akiba T., Sano S., Yanase T., Ohta T., Koyama M.: Optuna. In: Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. ACM, 2019.
[3] Albrecht J., et al.: The Muon Puzzle in cosmic-ray induced air showers and its connection to the Large Hadron Collider. p. 2 Energy spectra of various CR primary particles measured by several experiments. The "LHC" arrow indicates the kinetic energy achievable with the Large Hadron Collider
(as of 2023). Since then, a wide range of
experiments have embarked on a quest for better understanding of the nature and
origin of cosmic rays (CRs) and pushed to improve the precision of CR measurements
[18, 24, 19, 23, 17]. A CR is a high-energy particle or atomic nucleus arriving at Earth from outer
space.Where the paper says it
Where the paper says it
[4] Carminati G., Bazzotti M., Biagi S., Cecchini S., Chiarusi T., Margiotta A., Sioli M., Spurio M.: MUPAGE: a fast atmospheric MUon GEnerator for neutrino telescopes based on PArametric formulas. p. 3 Both were used for this work and are briefly described in what follows. MUPAGE (atmospheric MUons from PArametric formulas: a fast GEnerator for
neutrino telescopes) is a code inherited by KM3NeT from the ANTARES experiment.Where the paper says it
[5] Chen T., Guestrin C.: XGBoost: A Scalable Tree Boosting System. p. 5 As demonstrated in Fig. 4, the best performer was
LightGBM, a histogram-based gradient-boosting model.Where the paper says it
[6] Evoli C.: The Cosmic-Ray Energy Spectrum, 2020. p. 0 For this, data collected with intermediate detector configurations of
ARCA and ORCA was used in addition to simulated data from the envisaged
final configurations of those detectors. Prediction of the total number of muons
in a bundle as well as their total energy and even the energy of the primary
cosmic ray is presented.Where the paper says it
[7] Filippini F., Androutsou E., Domi A., Spisso B., Drakopoulou E.: Data reconstruction and classification with Graph neural networks in KM3NeT/ARCA. p. 0 This work explores the potential of machine
learning models for the reconstruction of muon bundles, which are multi-muon
events. For this, data collected with intermediate detector configurations of
ARCA and ORCA was used in addition to simulated data from the envisaged
final configurations of those detectors. Energy spectra of various CR primary particles measured by several experiments. The "LHC" arrow indicates the kinetic energy achievable with the Large Hadron Collider
(as of 2023).Where the paper says it
Where the paper says it
[8] Giacomelli G., Margiotta A.: The MACRO Experiment at Gran Sasso. p. 3 It extensively makes use of parametrised functions, which can be tuned to experimental data and simulations. The original tuning was performed based on the HEMAS
[16] Monte Carlo (MC) simulation code and data from the MACRO experiment at
Gran Sasso in Italy [8].Where the paper says it
[9] Greisen K. p. 9 9 look promising above 100 PeV, given how much of the original
information is lost when the EAS has reached the detector. It suggests that KM3NeT
may contribute to the primary CR measurements at the highest energies, notably
to the study of the Greisen-Zatsepin-Kuzmin (GZK) cutoff [9, 25]. The focus of ARCA is neutrino
astronomy, while ORCA is optimised for neutrino oscillation studies. Both
detectors are already operational in their intermediate states and collect valuable data, including the measurements of the muons produced by cosmic ray
interactions in the atmosphere.Where the paper says it
Where the paper says it
[10] Heck D., Knapp J., Capdevielle J.N., Schatz G., Thouw T.: CORSIKA: A Monte Carlo code to simulate extensive air showers. p. 3 CORSIKA (COsmic Ray SImulations for KAscade) originated as an MC simulation code for the KASKADE experiment [19]. With time, it evolved into a stateof-the-art solution for simulating cosmic ray air showers.Where the paper says it
[11] Hess V. p. 1 Measuring those quantities can help to pinpoint the shortcomings
of the current theoretical models used to compute the muon flux. Cosmic rays
The existence of cosmic radiation was discovered by Victor Hess in 1912 following
a series of balloon-borne radiation measurements [11]. Measuring those quantities can help to pinpoint the shortcomings
of the current theoretical models used to compute the muon flux. Cosmic rays
The existence of cosmic radiation was discovered by Victor Hess in 1912 following
a series of balloon-borne radiation measurements [11].Where the paper says it
Where the paper says it
[12] de Jong M.: Multi-dimensional interpolations in C++. p. 0 Both
detectors are already operational in their intermediate states and collect valuable data, including the measurements of the muons produced by cosmic ray
interactions in the atmosphere. This work explores the potential of machine
learning models for the reconstruction of muon bundles, which are multi-muon
events.Where the paper says it
[13] Kalaczyński P. p. 0 Piotr Kalaczyński
on behalf of the KM3NeT Collaboration
RECONSTRUCTION OF MUON BUNDLES
IN KM3NET DETECTORS USING
MACHINE LEARNING METHODS
Abstract The KM3NeT Collaboration is installing the ARCA and ORCA neutrino detectors at the bottom of the Mediterranean Sea. The focus of ARCA is neutrino
astronomy, while ORCA is optimised for neutrino oscillation studies. The focus of ARCA is neutrino
astronomy, while ORCA is optimised for neutrino oscillation studies. Both
detectors are already operational in their intermediate states and collect valuable data, including the measurements of the muons produced by cosmic ray
interactions in the atmosphere.Where the paper says it
Where the paper says it
[14] Ke G., Meng Q., Finley T., Wang T., Chen W., Ma W., Ye Q., Liu T.Y.: LightGBM: A Highly Efficient Gradient Boosting Decision Tree. p. 5 As demonstrated in Fig. 4, the best performer was
LightGBM, a histogram-based gradient-boosting model. reported in Fig. 9 look promising above 100 PeV, given how much of the original
information is lost when the EAS has reached the detector.Where the paper says it
Where the paper says it
[15] Pedregosa F., Varoquaux G., Gramfort A., Michel V., Thirion B., Grisel O., Blondel M., Prettenhofer P., Weiss R., Dubourg V., Vanderplas J., Passos A.: Scikit-learn: Machine Learning in Python. p. 5 Muon bundle energy
The first step in determining the muon bundle energy was the selection of the best ML
model. A range of regression models available in the Scikit-learn library [15], as well
as LightGBM [14] and XGBoost [5] were considered. Piotr Kalaczyński
on behalf of the KM3NeT Collaboration
RECONSTRUCTION OF MUON BUNDLES
IN KM3NET DETECTORS USING
MACHINE LEARNING METHODS
Abstract The KM3NeT Collaboration is installing the ARCA and ORCA neutrino detectors at the bottom of the Mediterranean Sea. The focus of ARCA is neutrino
astronomy, while ORCA is optimised for neutrino oscillation studies.Where the paper says it
Where the paper says it
[16] Scapparone E.: HEMAS: A Monte Carlo code for hadronic, electromagnetic and TeV muon components in air shower. p. 3 The original tuning was performed based on the HEMAS
[16] Monte Carlo (MC) simulation code and data from the MACRO experiment at
Gran Sasso in Italy [8]. The output of MUPAGE consists of muon bundle events
sampled on a cylindrical volume surrounding the detector, called the can.Where the paper says it
[17] Telescope Array Collaboration: R. p. 2 1 103 106 109 1012
Energy [GeV]
10−7
10−5
10−3
10−1
101
103
E
2 Intensity [GeV m
−
2
s
−
1 sr
−
1
]
AMS-02
AUGER
BESS
CALET
CREAM
DAMPE
FERMI
HAWC
HESS
ICECUBE
ICETOP+ICECUBE
KASCADE
KASCADE-Grande
NUCLEON
PAMELA
Telescope Array
Tibet-III
TUNKA
VERITAS
LHC
Knee
Ankle
1/m2/s
1/m2/yr
1/km2/yr
e−+e+
e+
p¯
p
ν + ν¯
γ
γ IGRB
github. com/carmeloevoli/The CR Spectrum
10−10 10−8 10−6 10−4 10−2 1 102
Energy [J]
Figure 1. CORSIKA (COsmic Ray SImulations for KAscade) originated as an MC simulation code for the KASKADE experiment [19]. With time, it evolved into a stateof-the-art solution for simulating cosmic ray air showers.Where the paper says it
Where the paper says it
[18] The Alpha Magnetic Spectrometer Collaboration: M. p. 0 Piotr Kalaczyński
on behalf of the KM3NeT Collaboration
RECONSTRUCTION OF MUON BUNDLES
IN KM3NET DETECTORS USING
MACHINE LEARNING METHODS
Abstract The KM3NeT Collaboration is installing the ARCA and ORCA neutrino detectors at the bottom of the Mediterranean Sea. The focus of ARCA is neutrino
astronomy, while ORCA is optimised for neutrino oscillation studies. The KM3NeT experiment
KM3NeT [20] is a network of two neutrino telescopes at the bottom of the Mediterranean Sea, currently under construction: ARCA near Portopalo di Capo Passero in
Italy and ORCA offshore Toulon in France. They share the same hardware, however,
their physics focus and thus geometry differ. The
simulated data was split between training, validation and test datasets in proportion 64:16:20
and was extracted from the official KM3NeT CORSIKA simulation produced by the author
of this work. The experimental data comes from the measurements taken by ARCA6 and
ORCA6 detectors in 2020 and 2021.Where the paper says it
Where the paper says it
Where the paper says it
[19] The KASCADE Collaboration: W. p. 0 Piotr Kalaczyński
on behalf of the KM3NeT Collaboration
RECONSTRUCTION OF MUON BUNDLES
IN KM3NET DETECTORS USING
MACHINE LEARNING METHODS
Abstract The KM3NeT Collaboration is installing the ARCA and ORCA neutrino detectors at the bottom of the Mediterranean Sea. The focus of ARCA is neutrino
astronomy, while ORCA is optimised for neutrino oscillation studies. 1 103 106 109 1012
Energy [GeV]
10−7
10−5
10−3
10−1
101
103
E
2 Intensity [GeV m
−
2
s
−
1 sr
−
1
]
AMS-02
AUGER
BESS
CALET
CREAM
DAMPE
FERMI
HAWC
HESS
ICECUBE
ICETOP+ICECUBE
KASCADE
KASCADE-Grande
NUCLEON
PAMELA
Telescope Array
Tibet-III
TUNKA
VERITAS
LHC
Knee
Ankle
1/m2/s
1/m2/yr
1/km2/yr
e−+e+
e+
p¯
p
ν + ν¯
γ
γ IGRB
github. com/carmeloevoli/The CR Spectrum
10−10 10−8 10−6 10−4 10−2 1 102
Energy [J]
Figure 1. The KM3NeT experiment
KM3NeT [20] is a network of two neutrino telescopes at the bottom of the Mediterranean Sea, currently under construction: ARCA near Portopalo di Capo Passero in
Italy and ORCA offshore Toulon in France. They share the same hardware, however,
their physics focus and thus geometry differ.Where the paper says it
Where the paper says it
Where the paper says it
[20] The KM3NeT Collaboration: S. p. 0 Piotr Kalaczyński
on behalf of the KM3NeT Collaboration
RECONSTRUCTION OF MUON BUNDLES
IN KM3NET DETECTORS USING
MACHINE LEARNING METHODS
Abstract The KM3NeT Collaboration is installing the ARCA and ORCA neutrino detectors at the bottom of the Mediterranean Sea. The focus of ARCA is neutrino
astronomy, while ORCA is optimised for neutrino oscillation studies. Piotr Kalaczyński
on behalf of the KM3NeT Collaboration
RECONSTRUCTION OF MUON BUNDLES
IN KM3NET DETECTORS USING
MACHINE LEARNING METHODS
Abstract The KM3NeT Collaboration is installing the ARCA and ORCA neutrino detectors at the bottom of the Mediterranean Sea. The focus of ARCA is neutrino
astronomy, while ORCA is optimised for neutrino oscillation studies. The KM3NeT experiment
KM3NeT [20] is a network of two neutrino telescopes at the bottom of the Mediterranean Sea, currently under construction: ARCA near Portopalo di Capo Passero in
Italy and ORCA offshore Toulon in France. They share the same hardware, however,
their physics focus and thus geometry differ.Where the paper says it
Where the paper says it
Where the paper says it
[21] The KM3NeT Collaboration: S. p. 0 Piotr Kalaczyński
on behalf of the KM3NeT Collaboration
RECONSTRUCTION OF MUON BUNDLES
IN KM3NET DETECTORS USING
MACHINE LEARNING METHODS
Abstract The KM3NeT Collaboration is installing the ARCA and ORCA neutrino detectors at the bottom of the Mediterranean Sea. The focus of ARCA is neutrino
astronomy, while ORCA is optimised for neutrino oscillation studies. Both were used for this work and are briefly described in what follows. MUPAGE (atmospheric MUons from PArametric formulas: a fast GEnerator for
neutrino telescopes) is a code inherited by KM3NeT from the ANTARES experiment. The KM3NeT experiment
KM3NeT [20] is a network of two neutrino telescopes at the bottom of the Mediterranean Sea, currently under construction: ARCA near Portopalo di Capo Passero in
Italy and ORCA offshore Toulon in France. They share the same hardware, however,
their physics focus and thus geometry differ. The
simulated data was split between training, validation and test datasets in proportion 64:16:20
and was extracted from the official KM3NeT CORSIKA simulation produced by the author
of this work. The experimental data comes from the measurements taken by ARCA6 and
ORCA6 detectors in 2020 and 2021.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
[22] The KM3NeT Collaboration: S. p. 0 Piotr Kalaczyński
on behalf of the KM3NeT Collaboration
RECONSTRUCTION OF MUON BUNDLES
IN KM3NET DETECTORS USING
MACHINE LEARNING METHODS
Abstract The KM3NeT Collaboration is installing the ARCA and ORCA neutrino detectors at the bottom of the Mediterranean Sea. The focus of ARCA is neutrino
astronomy, while ORCA is optimised for neutrino oscillation studies. Pictorial summary of the design of KM3NeT neutrino telescopes. From left to
right: the components of an optical module, the module integrated into a detection unit, an
artist’s impression of a detection unit and a building block. ORCA is placed shallower, at 2500 m underwater. Its more dense instrumentation allows for precise studies of atmospheric neutrino
oscillations in the GeV-TeV energy range.Where the paper says it
Where the paper says it
Where the paper says it
[23] The PAMELA Collaboration: V. p. 0 Piotr Kalaczyński
on behalf of the KM3NeT Collaboration
RECONSTRUCTION OF MUON BUNDLES
IN KM3NET DETECTORS USING
MACHINE LEARNING METHODS
Abstract The KM3NeT Collaboration is installing the ARCA and ORCA neutrino detectors at the bottom of the Mediterranean Sea. The focus of ARCA is neutrino
astronomy, while ORCA is optimised for neutrino oscillation studies. 1, the vast majority of primary CRs are protons, followed by
electrons, positrons and photons. A tiny contribution from the cosmic neutrinos has
been observed as well. The KM3NeT experiment
KM3NeT [20] is a network of two neutrino telescopes at the bottom of the Mediterranean Sea, currently under construction: ARCA near Portopalo di Capo Passero in
Italy and ORCA offshore Toulon in France. They share the same hardware, however,
their physics focus and thus geometry differ. The experimental data comes from the measurements taken by ARCA6 and
ORCA6 detectors in 2020 and 2021. ARCA115 ARCA6 ORCA115 ORCA6
Training data 14 374 415 3 894 293 15 439 946 6 154 895
Validation data 3 598 133 974 371 3 857 715 1 539 988
Test data 4 489 177 1 215 552 4 825 073 1 923 475
Experimental data × 3 013 982 × 4 574 887
JSirene [12] is an application developed to simulate photon emission and detection
by the PMTs, including the signal from environmental background.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
[24] The Pierre Auger collaboration: A. p. 0 Piotr Kalaczyński
on behalf of the KM3NeT Collaboration
RECONSTRUCTION OF MUON BUNDLES
IN KM3NET DETECTORS USING
MACHINE LEARNING METHODS
Abstract The KM3NeT Collaboration is installing the ARCA and ORCA neutrino detectors at the bottom of the Mediterranean Sea. The focus of ARCA is neutrino
astronomy, while ORCA is optimised for neutrino oscillation studies. Since then, a wide range of
experiments have embarked on a quest for better understanding of the nature and
origin of cosmic rays (CRs) and pushed to improve the precision of CR measurements
[18, 24, 19, 23, 17]. A CR is a high-energy particle or atomic nucleus arriving at Earth from outer
space. The KM3NeT experiment
KM3NeT [20] is a network of two neutrino telescopes at the bottom of the Mediterranean Sea, currently under construction: ARCA near Portopalo di Capo Passero in
Italy and ORCA offshore Toulon in France. They share the same hardware, however,
their physics focus and thus geometry differ.Where the paper says it
Where the paper says it
Where the paper says it
[25] Zatsepin G. p. 9 9 look promising above 100 PeV, given how much of the original
information is lost when the EAS has reached the detector. It suggests that KM3NeT
may contribute to the primary CR measurements at the highest energies, notably
to the study of the Greisen-Zatsepin-Kuzmin (GZK) cutoff [9, 25]. Measuring those quantities can help to pinpoint the shortcomings
of the current theoretical models used to compute the muon flux. Cosmic rays
The existence of cosmic radiation was discovered by Victor Hess in 1912 following
a series of balloon-borne radiation measurements [11]. The KM3NeT experiment
KM3NeT [20] is a network of two neutrino telescopes at the bottom of the Mediterranean Sea, currently under construction: ARCA near Portopalo di Capo Passero in
Italy and ORCA offshore Toulon in France. They share the same hardware, however,
their physics focus and thus geometry differ.Where the paper says it
Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Precision reconstruction of muon bundle energy can be improved by utilizing LightGBM as it was
the best performer
in terms of weighted coefficient of determination R2-score for ARCA115 data, achieving a median c=0.698±0.010. -
Feature selection methods should prioritize the
feature selection method requiring the positive importance of each feature
to improve model robustness, as this approachprovided the best results
in terms of mean R2 decrease for ARCA115. -
The ML system can be enhanced by integrating Graph Neural Networks to potentially yield valuable results, as suggested by
Considering other machine learning models, like graph neural networks [7], might yield valuable results.
-
The AI system can be made more accurate for high-energy predictions by addressing the underestimation of true bundle energy above 1 TeV through further work on CR shower simulations, as the paper notes that
at higher energies discrepancies are visible.
-
Primary cosmic ray energy reconstruction can be refined by attempting
to extract features from even lower-level data to ensure minimal information loss
in the model training process.
Abstract
The KM3NeT Collaboration is installing the ARCA and ORCA neutrino detectors at the bottom of the Mediterranean Sea. The focus of ARCA is neutrino astronomy, while ORCA is optimised for neutrino oscillation studies. Both detectors are already operational in their intermediate states and collect valuable data, including the measurements of the muons produced by cosmic ray interactions in the atmosphere. This work explores the potential of machine learning models for the reconstruction of muon bundles, which are multi-muon events. For this, data collected with intermediate detector configurations of ARCA and ORCA was used in addition to simulated data from the envisaged final configurations of those detectors. Prediction of the total number of muons in a bundle as well as their total energy and even the energy of the primary cosmic ray is presented.
Sources
- MUPAGE: a fast atmospheric MUon GEnerator for neutrino telescopes based on PArametric formulas
- The MACRO Experiment at Gran Sasso
- Multi-dimensional interpolations in C++
- HEMAS: a Monte Carlo code for hadronic, electromagnetic and TeV muon components in air shower
Related papers
- DrSai: an agentic AI system toward autonomous physics analysis workflows at BESIII
- CANTON- mu Proposal: A Next-Generation Muon g-2 Measurement at Sub-0.1 ppm Precision
- Search for dark photon dark matter near 1,meV using a superconducting mixer
- ANUBIS detector and its sensitivity to neutral long-lived particles
- How to scale your HEP ML models: A recipe for robust architecture comparisons at scale
- Search for the 16 O(ppp) to 13 C pi+ pi+ e+ Decay Mode in Super-Kamiokande Using Machine Learning Techniques