Pith. sign in

REVIEW 4 major objections 6 minor 57 references

Application of Graph Networks to a wide-field Water-Cherenkov-based Gamma-Ray Observatory

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Modeling air-shower footprints as graphs, a GNN rejects hadronic cosmic-ray background with a quality factor two to three times higher than the LIC/PINCness observables and reconstructs gamma-ray energies more accurately than…

desk verdict A carefully done simulation study showing GNNs beat hand-designed gamma/hadron separators for SWGO; the main caveat is that the simulation omits cosmic-ray noise, so the size of the advantage in the real detector is unproven. read the letter →

arxiv 2411.16565 v2 pith:6TK5CO6D submitted 2024-11-25 astro-ph.IM

Jonas Glombitza , Martin Schneider , Franziska Leitl , Stefan Funk , Christopher van Eldik This is my paper · ORCID
classification astro-ph.IM
keywords graphneuralnetworksgamma/hadronseparationwaterCherenkovdetectorsSWGOenergyreconstructionairshowersdeeplearningqualityfactor
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to show that graph neural networks (GNNs) classify and reconstruct air showers recorded by a water-Cherenkov gamma-ray observatory better than the hand-designed observables now used by the HAWC experiment. Each event is modeled as a graph whose nodes are triggered tanks carrying position, charge, and arrival time. On simulated events for a candidate SWGO array layout, the authors find that the GNN rejects hadronic background with a quality factor two to three times higher than the optimized combination of LIC and PINCness across the studied energy range. They also report more accurate energy reconstruction than the template-based likelihood method above about 200 GeV, reaching about 16% resolution at 17 TeV for zenith angles below 30 degrees. If these simulation results transfer to the real instrument, the planned southern-hemisphere observatory would gain a large increase in sensitivity to very-high-energy gamma-ray sources.

What carries the argument

The central object is the graph representation of the shower footprint: each triggered tank is a node with features $(x, y, \text{charge}, \text{arrival time})$, and edges connect each node to itself and to its six nearest spatial neighbors, matching the triangular tank layout. The network uses EdgeConvolution layers, whose kernel function $h_\Theta$ combines the central node's features with edge differences $x_j - x_i$ and aggregates over neighbors, followed by one DynamicEdgeConvolution layer to capture global structure, global pooling over nodes, and fully connected output layers. This machinery lets a single network handle events with tens to thousands of triggered tanks and exploit spatial, timing, and charge patterns jointly, in contrast to fixed-size image filters.

What would settle it

Run the same trained GNN and the LIC/PINCness baseline on a simulation that includes cosmic-ray-induced noise hits, or on real SWGO data after commissioning; if the quality-factor advantage drops below the claimed factor of two to three, or the energy resolution no longer beats the template method above 200 GeV, the central claim is refuted.

Watch

Extended reading notes

Core claim

The central claim is that a GNN operating on the sparse, irregular footprint of an air shower extracts more discriminative information than the observables currently used by water-Cherenkov experiments. On roughly 370,000 simulated gamma-ray and 440,000 proton events, the GNN trained on tank positions, charge, and arrival time achieves a quality factor two to three times higher than the optimized combination of LIC and PINCness over the whole energy range, and its background contamination at a fixed gamma-ray efficiency of 80% is up to an order of magnitude lower at high energies. For energy reconstruction, the paper reports a stable bias from 600 GeV to about 200 TeV and a resolution that consistently surpasses the template-based likelihood method above 200 GeV, with roughly 16% resolution at 17 TeV for zenith angles below 30 degrees, exceeding the performance requirements set out in the SWGO science case.

Load-bearing premise

The load-bearing premise is that the simulated detector response faithfully represents the planned SWGO array, especially the current absence of cosmic-ray-induced noise hits in the simulation; if real noise changes the pattern of triggered tanks, the GNN's graph structure changes and the claimed performance gain may not transfer.

Editorial extensions

If this is right

  • Source-detection significance scales with the quality factor $Q = \epsilon_\gamma / \sqrt{\epsilon_p}$, so a two-to-three-fold higher $Q$ translates directly into a two-to-three-fold gain in discovery sensitivity for gamma-ray sources.
  • Arrival-time information, previously little used for gamma/hadron separation in water-Cherenkov arrays, adds measurable separation power on top of charge and position.
  • The GNN delivers its separation performance without any quality cuts on the classifier output, simplifying the analysis chain relative to methods that require tuned cuts on LIC and PINCness.
  • The same graph architecture can be adapted to changes in detector layout or tank design, such as adding muon-tagging capabilities, without redesigning the algorithm.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Because real night-sky and cosmic-ray noise will alter which tanks trigger, the simulation's missing cosmic-ray noise is the most likely place for the claimed advantage to shrink; adding such noise to the simulation and retraining would quantify this.
  • The graph formulation is layout-agnostic and could transfer to other sparse detector arrays with strongly varying hit multiplicities, such as HAWC, LHAASO, or neutrino detectors.
  • The paper's own data thin out above 50 TeV, so the high-energy quality-factor gain is the least certain part of the claim and could be pinned down with a larger simulation set.
  • Attention mechanisms or separate graph clustering in the dense and sparse detector zones, which the authors list as future work, may push performance further, but those gains are not demonstrated here.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper presents a simulation study of graph neural networks (GNNs) for gamma/hadron separation and energy reconstruction in a candidate SWGO water-Cherenkov detector layout. Events are represented as graphs whose nodes are triggered tanks carrying position, charge, and arrival-time features, connected via k-nearest neighbors. The authors train EdgeConvolution-based GNNs with hyperparameter searches and compare them against the classical LIC/PINCness observables (optimized per energy bin) and a template-based likelihood method. They report a quality factor two to three times higher than the classical observables across the studied energy range, and an energy resolution of about 16% at 17 TeV for zenith angles below 30 degrees, outperforming the template method above about 200 GeV.

Significance. If the results transfer to the real detector, the work provides a strong quantitative case for using GNNs in SWGO event reconstruction. The study is internally well controlled: a 70/10/20 train/validation/test split, 70-trial hyperparameter searches for each task, validation-based model selection, multiple performance metrics, statistical uncertainties following Ref. [50], and per-bin optimization of the LIC/PINCness baselines. The main caveat is that the simulation explicitly omits cosmic-ray-induced noise hits, which will alter the graph topology that the GNN consumes; the relative robustness of the GNN versus the baselines under such noise is not quantified. This limitation is clearly stated in the text but is load-bearing for the title-level claim about the observatory, so a quantitative robustness estimate or a more restricted claim is needed.

major comments (4)
  1. [Sec. 2.1 / Sec. 6] The simulation explicitly lacks cosmic-ray-induced noise hits (Sec. 2.1: “still lacks cosmic ray-induced noise hits”), and Sec. 6 defers this to future work. Because the GNN graph has one node per triggered tank and uses k-nearest-neighbor connectivity (Sec. 2.2.1), random noise hits will add nodes and edges and change the local neighborhoods seen by the EdgeConvolution layers. The paper provides no estimate of how much of the reported quality-factor and energy-resolution advantage survives such noise relative to the LIC/PINCness and template baselines, which are also affected but potentially to a different degree. Since the title and framing address the SWGO observatory, this external-validity gap is load-bearing. Please add a quantitative robustness study (for example, injecting cosmic-ray noise hits at expected rates into the simulated events and recomputing all metrics) or explicitly and prominently restrict the claims to the noise-free simulation.
  2. [Sec. 3.1 / Tables 2 and 3] The graph aggregation function is specified inconsistently: Sec. 3.1 states “the sum (used in this work)”, while the architecture tables in Appendix A.2 specify “□j: mean” for every graph layer. Because the aggregation choice determines the network's behavior and thus the reported performance, this contradiction must be resolved. Please state which aggregation was actually used in the final models and correct the text accordingly.
  3. [Sec. 5 (Data selection)] For the energy-reconstruction comparison, the authors apply a dedicated selection that includes a cut on the GNN gamma-score (“The cut value was taken from the previous γ/hadron separation”). It is not stated whether the template-based method is evaluated on the same selected event sample or on its own default selection. If the GNN-score cut is applied before the comparison, it may preferentially retain events that the GNN reconstructs well, biasing the energy-resolution comparison in the GNN's favor. Please clarify whether both methods use the identical event sample and, if the GNN-score cut is applied, discuss its effect on the template method.
  4. [Sec. 4.3.1] The LIC and PINCness baseline cuts are said to be “optimized” for each energy bin “in our dataset”, but the paper does not specify whether this optimization is performed on the training/validation split or on the test set later used for the reported comparisons. To ensure a fair comparison with the GNN (whose hyperparameters are selected on the validation set), please state explicitly which data were used for the baseline cut optimization. If the cuts were tuned on the test set, the comparison is conservative but should still be reported as such.
minor comments (6)
  1. [Sec. 2.2.1] The abbreviation “knn” should be written as “kNN” or “k-nearest neighbors” when first introduced, and the figure caption should match.
  2. [Sec. 2.1] The sentence “This is slightly lower than the trigger threshold of 30 hits investigated in the following” is ambiguous; consider “used in the following analysis” instead of “investigated in the following”.
  3. [Eq. (4.2)] The definition of CxPE40 appears after the equation; placing it before the equation would improve readability.
  4. [Sec. 4.3.2] The text reports that background rejection improves by “around a factor of two” at low energies and “by one order of magnitude” at high energies, while the quality-factor comparison mentions a factor of three at low energies and five to eight at medium and high energies. These are different metrics, but the juxtaposition may confuse readers; consider adding an explicit statement that the two quantities are not directly comparable.
  5. [Sec. 5.2] The yellow marked region in Fig. 11 may be difficult to see in grayscale; please also indicate the ±10% region with dashed lines or similar.
  6. [Sec. 5] The exact value of the GNN score cut used in the energy-reconstruction data selection is not stated; please provide it for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the GNN benchmark is self-contained and measured against external baselines on held-out simulated data.

full rationale

The paper's central claims are empirical simulation benchmarks, not analytical derivations: a GNN is trained on simulated SWGO events (Sec. 2.1, with a 70/10/20 train/validation/test split) to predict MC truth labels and energies, and evaluated on the held-out test set (Secs. 4.1 and 5.2). The LIC and PINCness baselines are defined by their standard formulas in Sec. 4.3.1 and compared at matched gamma efficiency or optimized quality factor; the template energy-reconstruction baseline is an external method from Refs. [13,51]. No fitted parameter is renamed as a prediction: the GNN output score and reconstructed energy are functions of tank positions, charges, and times, and their quality is assessed by contamination, efficiency, bias, and resolution on test events. Self-references such as Ref. [44] (architectural inspiration) and Ref. [51] (template comparator) are contextual and non-load-bearing; they do not impose the result. The explicit limitation that the simulation lacks cosmic-ray-induced noise hits (Sec. 2.1 and Sec. 6), and the possible use of the same events for baseline cut optimization, are external-validity and statistical-fairness concerns rather than circularity. Hence no circular step is present.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical entities or fitted physical constants. The central result rests on the fidelity of the Monte Carlo detector simulation, the graph construction choices, and the assumption that excluding cosmic-ray-induced noise hits does not alter the relative performance ranking.

assumptions (5)
  • domain assumption The HAWC-based simulation package modified for SWGO provides a faithful model of detector response (CORSIKA showers, GEANT4 detector interaction).
    All performance numbers derive from this simulation; Section 2.1 describes the setup but no comparison to real data is provided.
  • domain assumption Omitting cosmic-ray-induced noise hits does not change the relative ranking of GNN and baseline methods.
    Section 2.1 explicitly states the simulation lacks cosmic-ray-induced noise hits; the impact on graph-based GNNs is not quantified.
  • domain assumption The knn graph construction with k=6 and self-connections is a sufficient representation of local shower structure.
    Section 2.2.1 motivates k=6 by the triangular tank grid; no ablation on k is shown.
  • domain assumption The template-based likelihood method [13, 51] and the LIC/PINCness observables are appropriate state-of-the-art baselines.
    Section 4.3 and 5.2 compare against these baselines; the correctness of their implementation is taken as given.
  • domain assumption Training with identical spectral index -2 for protons and gamma rays produces a classifier whose performance transfers to physics analyses.
    Section 2.1 states this choice is intended to make the classifier spectrum-independent; real source spectra differ.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Application of Graph Networks to a wide-field Water-Cherenkov-based Gamma-Ray Observatory." pith.science (2026). https://pith.science/paper/6TK5CO6D

@misc{pith2026241116565,
  author       = {Pith},
  title        = {Pith review of: Application of Graph Networks to a wide-field Water-Cherenkov-based Gamma-Ray Observatory},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6TK5CO6D}},
  note         = {Machine review of arXiv:2411.16565}
}
read the original abstract

With their wide field of view and high duty cycle, water-Cherenkov-based observatories are integral to studying the very high-energy gamma-ray sky. For gamma-ray observations, precise event reconstruction and highly effective background rejection are crucial and have been continuously improving in recent years. In this work, we investigate the application of graph neural networks (GNNs) to background rejection and energy reconstruction and benchmark their performance against state-of-the-art methods. In our simulation study, we find that GNNs outperform hand-designed classification algorithms and observables in background rejection and find an improved energy resolution compared to template-based methods.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 24 canonical work pages

  1. [50]

    Ullrich and Z

    T. Ullrich and Z. Xu. Treatment of errors in efficiency calculations, 2012. arXiv:0701199v1

    work page 2012

  2. [1]

    Kildea (Whipple Collaboration) et al

    J. Kildea (Whipple Collaboration) et al. The Whipple Observatory 10m gamma-ray telescope, 1997–2006. Astropart. Phys., 28(2):182–195, 2007. doi: 10.1016/j.astropartphys.2007.05.004

    doi:10.1016/j.astropartphys.2007.05.004 1997

  3. [2]

    Aharonian (H.E.S.S

    F. Aharonian (H.E.S.S. Collaboration) et al. Observations of the Crab nebula with HESS. A&A, 457(3):899–915, 2006. doi: 10.1051/0004-6361:20065351

    doi:10.1051/0004-6361:20065351 2006

  4. [3]

    Aleksić (MAGIC Collaboration) et al

    J. Aleksić (MAGIC Collaboration) et al. Performance of the MAGIC stereo system obtained with Crab Nebula data.Astropart. Phys., 35(7):435–448, 2012. doi: 10.1016/j.astropartphys.2011.11.007

    doi:10.1016/j.astropartphys.2011.11.007 2012

  5. [4]

    Holder (VERITAS Collaboration) et al

    J. Holder (VERITAS Collaboration) et al. The first VERITAS telescope.Astropart. Phys., 25 (6):391–401, 2006. doi: 10.1016/j.astropartphys.2006.04.002

    doi:10.1016/j.astropartphys.2006.04.002 2006

  6. [5]

    B. S. Acharya (CTA Collaboration) et al. Introducing the CTA concept.Astropart. Phys., 43: 3–18, 2013. doi: 10.1016/j.astropartphys.2013.01.007. Seeing the High-Energy Universe with the Cherenkov Telescope Array - The Science Explored with the CTA

    doi:10.1016/j.astropartphys.2013.01.007 2013

  7. [6]

    Abreu (SWGO Collaboration) et al

    P. Abreu (SWGO Collaboration) et al. The Southern Wide-Field Gamma-Ray Observatory (SWGO): A Next-Generation Ground-Based Survey Instrument for VHE Gamma-Ray Astronomy. 2019. arXiv:1907.07737

    arXiv 2019

  8. [7]

    Atkins (Milagro Collaboration) et al

    R. Atkins (Milagro Collaboration) et al. Observation of TeV Gamma Rays from the Crab Nebula with Milagro Using a New Background Rejection Technique.Astrophys. J., 595(2):803,

    work page

Show all 57 references
  1. [8]

    Atkins (Milagro Collaboration) et al

    R. Atkins (Milagro Collaboration) et al. Milagrito, a TeV air-shower array.Nucl. Instrum. Methods Phys. Res. A , 449(3):478–499, 2000. doi: 10.1016/S0168-9002(00)00146-7

    2000 doi

  2. [9]

    A. U. Abeysekara (HAWC Collaboration) et al. The High-Altitude Water Cherenkov (HAWC) observatory in México: The primary detector.Nucl. Instrum. Methods Phys. Res. A , 1052: 168253, 2023. doi: 10.1016/j.nima.2023.168253

    2023

  3. [10]

    Aharonian (LHAASO collaboration) et al

    F. Aharonian (LHAASO collaboration) et al. Performance of LHAASO-WCDA and Observation of Crab Nebula as a Standard Candle, 2021. arXiv:2101.03508. – 17 –

    2021 arXiv

  4. [11]

    A. J. Smith on behalf of the HAWC Collaboration. HAWC: Design, Operation, Reconstruction and Analysis, 2015. arXiv:1508.05826

    2015 arXiv

  5. [12]

    R. D. Parsons and J. A. Hinton. A Monte Carlo template based analysis for air-Cherenkov arrays. Astropart. Phys., 56:26–34, 2014. doi: 10.1016/j.astropartphys.2014.03.002

    2014 doi

  6. [13]

    Joshi et al

    V. Joshi et al. A template-basedγ-ray reconstruction method for air shower arrays.JCAP, 2019(01):012–012, 2019. doi: 10.1088/1475-7516/2019/01/012

    2019 doi

  7. [14]

    S. Ohm, C. van Eldik, and K. Egberts.γ/hadron separation in very-high-energyγ-ray astronomy using a multivariate analysis method.Astropart. Phys., 31(5):383–391, 2009. doi: 10.1016/j.astropartphys.2009.04.001

    2009 doi

  8. [15]

    Krause, E

    M. Krause, E. Pueschel, and G. Maier. Improvedγ/hadron separation for the detection of faint γ-ray sources using boosted decision trees.Astropart. Phys., 89:1–9, 2017. doi: 10.1016/j.astropartphys.2017.01.004

    2017 doi

  9. [16]

    Albert (MAGIC Collaboration) et al

    J. Albert (MAGIC Collaboration) et al. Implementation of the Random Forest method for the Imaging Atmospheric Cherenkov Telescope MAGIC.Nucl. Instrum. Methods Phys. Res. A , 588 (3):424–432, 2008. doi: 10.1016/j.nima.2007.11.068

    2008 doi

  10. [17]

    LeCun, Y

    Y. LeCun, Y. Bengio, and G. Hinton. Deep Learning.Nature, 521(7553):436–44, 2015. doi: 10.1038/nature14539

    2015 doi

  11. [18]

    Erdmann et al.Deep Learning for Physics Research

    M. Erdmann et al.Deep Learning for Physics Research . WORLD SCIENTIFIC, 2021. ISBN 978-981-12-3745-4. doi: 10.1142/12294

    2021 doi

  12. [19]

    Westerhoff et al

    S. Westerhoff et al. Separatingγ- and hadron-induced cosmic ray air showers with feed-forward neural networks using the charged particle information.Astropart. Phys., 4(2):119–132, 1995. doi: 10.1016/0927-6505(95)00028-4

    1995 doi

  13. [20]

    Boinee et al

    P. Boinee et al. Neural Networks for gamma-hadron separation in MAGIC. InFFP, pages 297–302. Springer Netherlands, 2006

    2006

  14. [21]

    Erdmann, J

    M. Erdmann, J. Glombitza, and D. Walz. A deep learning-based reconstruction of cosmic ray-induced air showers.Astropart. Phys., 97:46–53, 2018. doi: 10.1016/j.astropartphys.2017.10.006

    2018 doi

  15. [22]

    Aab (Pierre Auger Collaboration) et al

    A. Aab (Pierre Auger Collaboration) et al. Deep-learning based reconstruction of the shower maximum Xmax using the water-Cherenkov detectors of the Pierre Auger Observatory. JINST, 16(07):P07019, 2021. doi: 10.1088/1748-0221/16/07/p07019

    2021 doi

  16. [23]

    Aab (Pierre Auger Collaboration) et al

    A. Aab (Pierre Auger Collaboration) et al. Extraction of the muon signals recorded with the surface detector of the Pierre Auger Observatory using recurrent neural networks.JINST, 16 (07):P07016, 2021. doi: 10.1088/1748-0221/16/07/p07016

    2021 doi

  17. [24]

    Abbasi (IceCube Collaboration) et al

    R. Abbasi (IceCube Collaboration) et al. A Convolutional Neural Network Based Cascade Reconstruction for the IceCube Neutrino Observatory.JINST, 16(07):P07041, 2021. doi: 10.1088/1748-0221/16/07/p07041

    2021 doi

  18. [25]

    Aiello (KM3NeT Collaboration) et al

    S. Aiello (KM3NeT Collaboration) et al. Event reconstruction for KM3NeT/ORCA using convolutional neural networks.JINST, 15(10):P10005, 2020. doi: 10.1088/1748-0221/15/10/P10005

    2020 doi

  19. [26]

    Bom et al

    C. Bom et al. Bayesian Deep Learning for Shower Parameter Reconstruction in Water Cherenkov Detectors. PoS, ICRC2021:739, 2021. doi: 10.22323/1.395.0739

    2021 doi

  20. [27]

    Assunçã et al

    F. Assunçã et al. Automatic design of artificial neural networks for gamma-ray detection. IEEE Access, 7:110531–110540, 2019. doi: 10.1109/access.2019.2933947

    2019

  21. [28]

    B. S. González et al. Tackling the muon identification in water cherenkov detectors problem for the future southern wide-field gamma-ray observatory by means of machine learning.Neural Comput. Appl., pages 1–14, 2022. doi: https://doi.org/10.1007/s00521-021-06730-z. – 18 –

    2022 doi

  22. [29]

    Conceição et al

    R. Conceição et al. Muon identification in a compact single-layered water Cherenkov detector and gamma/hadron discrimination using machine learning techniques.EPJ C, 81(6):542, 2021. doi: 10.1140/epjc/s10052-021-09312-4

    2021 doi

  23. [30]

    Watson on behalf of the HAWC Collaboration

    I. Watson on behalf of the HAWC Collaboration. Deep Learning for the HAWC Gamma-Ray Observatory. PoS, ICRC2023:927, 2023. doi: 10.22323/1.444.0927

    2023 doi

  24. [31]

    Abbasi (IceCube Collaboration) et al

    R. Abbasi (IceCube Collaboration) et al. Observation of high-energy neutrinos from the Galactic plane. Science, 380(6652):1338–1343, 2023. doi: 10.1126/science.adc9818

    2023 doi

  25. [32]

    Abdul Halim (Pierre Auger Collaboration) et al

    A. Abdul Halim (Pierre Auger Collaboration) et al. Inference of the Mass Composition of Cosmic Rays with energies from1018.5 to 1020 eV using the Pierre Auger Observatory and Deep Learning. Phys. Rev. Lett., 134(2), 2025. doi: 10.1103/physrevlett.134.021001

    2025 doi

  26. [33]

    Abdul Halim (Pierre Auger Collaboration) et al

    A. Abdul Halim (Pierre Auger Collaboration) et al. Measurement of the Depth of Maximum of Air-Shower Profiles with energies between1018.5 and 1020 eV using the Surface Detector of the Pierre Auger Observatory and Deep Learning.Phys. Rev. D , 111:022003, 2025. doi: 10.1103/Phys...

    2025 doi

  27. [34]

    M. M. Bronstein et al. Geometric Deep Learning: Going beyond Euclidean data.IEEE Signal Process. Mag., 34(4):18–42, 2017. doi: 10.1109/MSP.2017.2693418

    2017

  28. [35]

    Heck et al

    D. Heck et al. CORSIKA: A Monte Carlo Code to Simulate Extensive Air Showers. Technical report, 1998. URL https://hal.in2p3.fr/in2p3-00005094

    1998

  29. [36]

    Amako et al

    K. Amako et al. S. Agostinelli, J. Allison. Geant4—a simulation toolkit.Nucl. Instrum. Methods Phys. Res. A , 506(3):250–303, 2003. doi: 10.1016/S0168-9002(03)01368-8

    2003 doi

  30. [37]

    A. U. Abeysekara (HAWC collaboration) et al. Observation of the Crab Nebula with the HAWC Gamma-Ray Observatory.Astrophys. J., 843(1):39, 2017. doi: 10.3847/1538-4357/aa7555

    2017 doi

  31. [38]

    Schoorlemmer, R

    H. Schoorlemmer, R. Conceiçao, and A. J. Smith. Simulating the performance of the Southern Wide-view Gamma-ray Observatory.PoS, ICRC2021:903, 2021. PoS(ICRC2021)903

    2021

  32. [39]

    Conceiçao on behalf of the SWGO Collaboration

    R. Conceiçao on behalf of the SWGO Collaboration. The Southern Wide-field Gamma-ray Observatory. PoS, ICRC2023:963, 2023. doi: 10.22323/1.444.0963

    2023 doi

  33. [40]

    Wang et al

    Y. Wang et al. Dynamic Graph CNN for Learning on Point Clouds, 2018. arXiv:1801.07829

    2018 arXiv

  34. [41]

    Qu and L

    H. Qu and L. Gouskos. Jet tagging via particle clouds.Phys. Rev. D , 101(5), 2020. doi: 10.1103/physrevd.101.056019

    2020 doi

  35. [42]

    Abbasi (IceCube Collaboration) et al

    R. Abbasi (IceCube Collaboration) et al. Graph Neural Networks for low-energy event classification and reconstruction in IceCube.JINST, 17(11):P11003, 2022. doi: 10.1088/1748-0221/17/11/p11003

    2022 doi

  36. [43]

    Bister et al

    T. Bister et al. Identification of patterns in cosmic-ray arrival directions using dynamic graph convolutional neural networks.Astropart. Phys., 126:102527, 2021. doi: 10.1016/j.astropartphys.2020.102527

    2021

  37. [44]

    Glombitza et al

    J. Glombitza et al. Application of graph networks to background rejection in Imaging Air Cherenkov Telescopes. JCAP, 2023(11):008, 2023. doi: 10.1088/1475-7516/2023/11/008

    2023 doi

  38. [45]

    Fey and J

    M. Fey and J. E. Lenssen. Fast Graph Representation Learning with PyTorch Geometric

  39. [46]

    D. P. Kingma and J. Ba. Adam: A Method for Stochastic Optimization. arXiv:1412.6980

    arXiv

  40. [47]

    Srivastava et al

    N. Srivastava et al. Dropout: A Simple Way to Prevent Neural Networks from Overfitting. JMLR, 15(56):1929–1958, 2014. URLhttp://jmlr.org/papers/v15/srivastava14a.html

    1929

  41. [48]

    Ioffe and C

    S. Ioffe and C. Szegedy. Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift, 2015. arXiv:1502.03167. – 19 –

    2015 arXiv

  42. [49]

    He et al

    K. He et al. Deep Residual Learning for Image Recognition, 2015. arXiv:1512.03385

    2015 arXiv

  43. [51]

    Leitl, V

    F. Leitl, V. Joshi, S. Funk on behalf of the SWGO Collaboration. Status of the SWGO air shower reconstruction using a template-based likelihood method, 2023. PoS(ICRC2023)593

    2023

  44. [52]

    Albert (HAWC Collaboration) et al

    A. Albert (HAWC Collaboration) et al. Performance of the HAWC Observatory and TeV Gamma-Ray Measurements of the Crab Nebula with Improved Extensive Air Shower Reconstruction Algorithms, 2024. arXiv:2405.06050

    2024 arXiv

  45. [53]

    Hampel-Arias and S

    Z. Hampel-Arias and S. on bahlf of the HAWC Collaboration Westerhoff. Gamma Hadron Separation using Pairwise Compactness Method with HAWC.PoS, ICRC2015:1001, 2016. doi: 10.22323/1.236.1001

    2016 doi

  46. [54]

    Conceição et al

    R. Conceição et al. Gamma/hadron discrimination at high energies through the azimuthal fluctuations of air shower particle distributions at the ground.JCAP, 2022(10):086, 2022. doi: 10.1088/1475-7516/2022/10/086

    2022 doi

  47. [55]

    Alfaro (HAWC Collaboration) et al

    R. Alfaro (HAWC Collaboration) et al. Gamma/hadron separation with the HAWC observatory. Nucl. Instrum. Methods Phys. Res. A , 1039:166984, 2022. doi: 10.1016/j.nima.2022.166984

    2022

  48. [56]

    Albert et al

    A. Albert et al. Science Case for a Wide Field-of-View Very-High-Energy Gamma-Ray Observatory in the Southern Hemisphere, 2019. arXiv:1902.08429

    2019 arXiv

  49. [57]

    Addition

    A.U. Abeysekara (HAWC Collanboration) et al. Data acquisition architecture and online processing system for the hawc gamma-ray observatory.Nucl. Instrum. Methods Phys. Res. A , 888:138–146, 2018. doi: 10.1016/j.nima.2018.01.051. – 20 – A Network architectures A.1 Parameters af...

    2018 doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.