Pith. sign in

REVIEW 1 major objections 7 minor 1 cited by

Machine Learning in Gamma Astronomy

T0 review · 1 major / 7 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read This review argues that deep learning has become a practical, standard toolkit for IACT data analysis, and maps the field's main methods and results.

desk verdict A useful but unambitious review that curates deep learning references for IACT analysis; it is honest about being a shortened version of an earlier review, and its main weakness is an unsubstantiated 'most popular' scope claim. read the letter →

arxiv 2501.19064 v1 pith:R7IOHNSK submitted 2025-01-31 astro-ph.IM astro-ph.HEcs.LG

A.P.Kryukov , A.P.Demichev , V.A.Ilyin This is my paper · ORCID
classification astro-ph.IMastro-ph.HEcs.LG
keywords gamma-rayastronomyimagingatmosphericCherenkovtelescopesdeeplearningconvolutionalneuralnetworkseventclassificationairshowerreconstructiongenerativeadversarialmulti-task
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

Imaging Atmospheric Cherenkov Telescopes record air showers produced by gamma rays and cosmic rays, and their central data-analysis problem is separating the rare gamma-ray events—swamped by a background roughly ten thousand times larger—while reconstructing each primary's energy and direction. This paper is a review of the deep learning methods that have been brought to bear on that problem, with the goal of mapping the most popular approaches and pointing readers to the original papers. Its conclusion is that convolutional neural networks applied directly to camera images have become a workable alternative to hand-crafted shower-parameter analysis, and that the field is moving toward richer architectures that combine convolutional backbones with recurrent layers, attention mechanisms, multi-task output heads, and generative models for fast simulation. A sympathetic reader would take the paper as a practical entry point: it collects the main architectures, the quality gains they report, and the software tools that make them usable.

What carries the argument

The load-bearing object is the convolutional neural network (CNN) applied to IACT camera images, together with the components that extend it: deep residual and multi-scale convolutional backbones, recurrent layers (RNNs and LSTM cells) for sequences of images from different telescopes, attention blocks that let the network focus on relevant parts of the feature map, and multi-task decoders that share representations across classification and regression objectives. A second machinery track is generative: GANs and conditional variational autoencoders trained on Monte Carlo images, used to synthesize new camera images faster than full simulation. The review also treats the geometric obstacle that hexagonal camera pixels pose for rectangular-grid convolutions, and it catalogs interpolation and indexed-convolution strategies for bridging that gap.

What would settle it

A systematic bibliometric or expert survey of IACT deep learning papers published up to 2025 that finds the most-used methods are, for example, transformer or graph-neural-network models rather than the CNN/RNN/GAN families emphasized here would falsify the review's implicit claim about what is most popular.

Watch

Extended reading notes

Core claim

The paper's central claim is that deep learning has matured into a standard, practical toolkit for IACT data analysis, and that the leading methods form a coherent family centered on convolutional neural networks. For classification, these networks separate gamma-induced from hadron-induced showers directly from pixel images, with reported accuracies that improve with primary energy (for one tested architecture, from about 81% at 0.1 TeV to about 92% at 100 TeV). For parameter reconstruction, multi-task networks augmented with attention mechanisms—exemplified by the γ-PhysNet system—jointly estimate energy, arrival direction, and event class, and in simulation they outperform the conventional moment-based shower-parameter plus random-forest baseline, especially below 100 GeV. For simulation speed, generative adversarial networks and conditional variational autoencoders produce camera images that are statistically indistinguishable from Monte Carlo output but orders of magnitude faster, which is proposed as a way to augment training data and correct class imbalance. The review also reports that recurrent-convolutional hybrids can process image sequences from multiple telescopes, and it describes two open software frameworks that supply the plumbing for these models. In sum, the review claims the field has shifted from feature engineering to end-to-end learned analysis, while noting that gains on simulated data carry the caveat of a simulation-to-real gap.

Load-bearing premise

The review's map of the field is only as reliable as its selection of papers, and the paper gives no systematic search or inclusion criteria; because it is a shortened version of the authors' own longer review and draws many examples from one experimental collaboration, the 'most popular' label may reflect the authors' vantage point rather than the whole field.

Editorial extensions

If this is right

  • Gamma/proton classification can be done with no prior parameterization of shower images; for one tested architecture accuracy rises from about 81% at 0.1 TeV to about 92% at 100 TeV.
  • Attention-augmented multi-task networks can beat moment-based parameter baselines on angular and energy resolution, most clearly below 100 GeV, which would allow fainter gamma sources to be detected.
  • Generative models can replace Monte Carlo for some simulation purposes, producing statistically matched images orders of magnitude faster and enabling synthetic resampling to fight class imbalance.
  • Recurrent-convolutional hybrids make stereoscopic data usable as image sequences, although the tested telescope-orderings do not yet show a quality gain.
  • Dedicated software platforms make the deep learning pipeline reproducible and reduce the barrier for applying these methods to new and future IACT arrays.

Reading between the lines

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

  • A testable extension: the simulation-to-real gap the paper flags suggests that domain adaptation—training on simulated data augmented with realistic night-sky noise, then fine-tuning on real events—will become a central research direction as large arrays come online; this is an editorial inference, not a claim of the review.
  • The review's selection implicitly defines 'most popular' by its reference list rather than by citation counts; one could seed a bibliometric study from that list to measure how much of the IACT deep learning literature it captures.
  • The attention-augmented architectures highlighted here point toward transformer-style and graph-based models that operate natively on hexagonal pixel grids, a direction the review mentions only indirectly through attention.
  • If end-to-end learned reconstruction continues to close the gap with conventional methods on real data, the traditional shower-parameterization stage in IACT analysis software may eventually be removed entirely.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

1 major / 7 minor

Summary. This manuscript is a review of deep learning methods for imaging atmospheric Cherenkov telescope (IACT) data analysis. It covers particle-type classification, reconstruction of extensive-air-shower parameters, and dedicated software frameworks such as CTLearn and GammaLearn. The abstract states that the paper's purpose is to review 'the most popular deep learning methods' used for IACT data and to provide references to original papers. The text is an explicit abridged version of the authors' earlier review [13], with a number of newer references added. The paper is narrative and does not present new experimental results or systematic search methodology.

Significance. If its map of the field is accurate, the review would be a useful entry point for practitioners, particularly because it collects practical topics such as hexagonal pixel-grid handling, multi-task and attention architectures, and GAN-based fast simulation. The organization is clear and the references point to original sources. The main weakness is that the 'most popular' scope claim is not backed by an explicit selection methodology, and the sample is skewed toward the authors' own TAIGA-related work (refs. 32, 33, 36, 51-53, 57-59). This limits but does not invalidate the review as a partial map of the field. The paper ships no code or machine-checked proofs, which is expected for a literature review.

major comments (1)
  1. [Abstract and Section 1] The abstract's central claim—that the paper reviews 'the most popular deep learning methods'—is an empirical statement about the field, but the manuscript provides no selection methodology. Section 1 only states that the paper is a shortened version of the authors' own review [13], and the acknowledgments thank the TAIGA collaboration. A substantial subset of the reviewed applications (refs. 32, 33, 36, 51-53, 57-59) come from TAIGA or from the authors' own publications. Without an explicit search strategy, inclusion/exclusion criteria, or a usage-based measure of popularity, the representativeness of the sample is unsupported. Please add a short methods paragraph describing how the papers were identified and selected, or reframe the abstract and Section 1 to describe 'selected deep learning applications' rather than 'the most popular' ones.
minor comments (7)
  1. [Section 2] Typo: 'tranformation' should be 'transformation' in the sentence about oblique coordinates.
  2. [Section 3] Typo: 'CNNss' should be 'CNNs'; additionally, 'Let us focused' should be 'Let us focus' and 'charge cosmic rays' should be 'charged cosmic rays'.
  3. [Section 3] The sentence 'It was shown that GAN, cGAN and cV AE simulate proton and gamma events for the TAIGA-IACT experiment with a high degree of accuracy and reliability' is repeated verbatim two sentences later with slightly different wording; the duplicate should be removed.
  4. [Section 3] The phrase 'implements a method of upload data [24]' is unclear and likely incorrect; it should refer to indexed operations on hexagonal lattices and should be rephrased accordingly.
  5. [Section 3] Grammar issues in the attention-mechanism paragraph: 'The paper [38] explore' should be 'explores' and 'which improve of the accuracy' should be 'which improves the accuracy'.
  6. [Section 4] Missing prepositions in software descriptions: 'event reconstruction IACT' should be 'event reconstruction for IACT' and 'metrics IACT' should be 'metrics for IACT'.
  7. [References] Reference [4] lists the year as '201' (incomplete); reference [10] is missing an article number in the journal citation; reference [14] from 1993 predates the modern deep-learning era and describing it as deep learning is anachronistic, though it may be historically relevant.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper is a review with no derivation or prediction loop; the only self-citation is an honest provenance statement, and the TAIGA-heavy reference set is a coverage concern, not a circularity.

full rationale

This paper is a narrative review of deep-learning applications in IACT gamma astronomy. It contains no equations, no fitted parameters named as predictions, no uniqueness theorem invoked to force a choice, and no derivation whose output equals its input by construction. The closest thing to a self-referential element is the statement in Section 1: "This paper is a shortened version of the extensive review [13]," where [13] is the authors' own prior review. This is a provenance disclosure, not a load-bearing argument; it does not substitute for evidence or justify a technical conclusion. The abstract's scope claim to cover "the most popular deep learning methods" is not supported by an explicit search strategy or inclusion criteria, and a substantial subset of the cited applications concerns TAIGA, with which the authors are affiliated. That is a legitimate concern about representativeness and coverage, but it is not circular reasoning: the review does not claim to derive its own conclusions from those references, and it explicitly points readers to original papers for details. No step in the paper reduces to its own input, either statistically or definitionally. The paper is self-contained as a review, and the appropriate finding is no significant circularity.

Assumptions & free parameters 0 free parameters · 1 assumptions · 0 invented entities

No free parameters, fitted values, or invented entities appear because the paper is a literature review. The only background assumption is accuracy of the reported literature, which is unverified here.

assumptions (1)
  • domain assumption The review's summaries of the cited deep learning papers are faithful to the originals.
    A review cannot independently confirm every quoted accuracy figure or architectural detail; correctness of the survey depends on this trust.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Machine Learning in Gamma Astronomy." pith.science (2026). https://pith.science/paper/R7IOHNSK

@misc{pith2026250119064,
  author       = {Pith},
  title        = {Pith review of: Machine Learning in Gamma Astronomy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R7IOHNSK}},
  note         = {Machine review of arXiv:2501.19064}
}
read the original abstract

The purpose of this paper is to review the most popular deep learning methods used to analyze astroparticle data obtained with Imaging Atmospheric Cherenkov Telescopes and provide references to the original papers.

Figures

Figures reproduced from arXiv: 2501.19064 by the authors.

Figure 1
Figure 1. Examples of simulated EAS images in an IACT camera for the [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. The key components of the deep learning based IACT data pro [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. It consists of two parts. The first one is a very deep CNN (ResNet [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 3
Figure 3. Figure 3: The simplified γ-PhysNet architecture. the energy, extract gamma rays from the background, and reconstruct the direction of the EAS axis and the arrival point. Multi-task learning helps to improve the generalization properties of the model and, therefore, the quality o…
Figure 4
Figure 4. Figure 4: The general GAN architecture. ods. Additionally, the specialized program OPTICA-TAIGA [54] is used to simulate the response IACT and perform a full tracing of Cherenkov photons through the telescope optics. The difficulty lies in the fact that the computational models …
Figure 5
Figure 5. Figure 5: The simplified general architecture of GammaLearn. [PITH_FULL_IMAGE:figures/full_fig_p012_5.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. HAWC Performance Enhanced by Machine Learning in Gamma-Hadron Separation

    astro-ph.IM 2025-06 conditional novelty 5.0 of 10

    Using 20 event features and a unified training set, an MLP improves HAWC gamma-hadron separation, increasing Crab Nebula significance by 19% and differential sensitivity by 23-40%.

Reference graph

Works this paper leans on

62 extracted references · 41 canonical work pages · cited by 1 Pith paper

  1. [13]

    Demichev, A

    A. Demichev, A. Kryukov, Using deep learning methods for iact data analysis in gamma-ray astronomy: A review. Astronomy and Computing 46, 100793 (2024). https://doi.org/10.1016/j.ascom.2024.100793

    work page arXiv 2024

  2. [1]

    Aggarwal, Neural Networks and Deep Learning - A Textbook (Springer, Berlin, 2018)

    C.C. Aggarwal, Neural Networks and Deep Learning - A Textbook (Springer, Berlin, 2018). https://doi.org/10.1007/978-3-319-94463-0

    doi:10.1007/978-3-319-94463-0 2018

  3. [2]

    Goodfellow, Y

    I.J. Goodfellow, Y. Bengio, A. Courville, Deep Learning. (MIT Press, Cambridge, MA USA, 2016). http://www.deeplearningbook.org

    work page 2016

  4. [3]

    Shrestha, A

    A. Shrestha, A. Mahmood, Review of deep learning algo- rithms and architectures. IEEE access 7, 53040–53065 (2019). https://doi.org/10.1109/access.2019.2912200

    work page arXiv 2019

  5. [4]

    Hatcher, W

    W.G. Hatcher, W. Yu, A survey of deep learning: Platforms, applica- tions and emerging research trends. IEEE Access 6, 24411–24432 (201). https://doi.org/10.1109/access.2018.2830661

    work page arXiv 2018

  6. [5]

    S. Dong, P. Wang, K. Abbas, A survey on deep learning and its applications. Computer Science Review 40, 100379 (2021). https://doi.org/10.1016/j.cosrev.2021.100379

    work page arXiv 2021

  7. [6]

    Albert, E

    J. Albert, E. Aliu, H. Anderhub, et al., Implementation of the ran- dom forest method for the imaging atmospheric Cherenkov telescope magic. Nuclear Instruments and Methods in Physics Research Section 13 A: Accelerators, Spectrometers, Detectors and Associated Equipment 588, 424–432 (2008). https://doi.org/10.1016/j.nima.2007.11.068

    doi:10.1016/j.nima.2007.11.068 2008

  8. [7]

    S. Ohm, C. van Eldik, K. Egberts, γ/hadron separation in very-high-energy γ/-ray astronomy using a multivariate analysis method. Astroparticle Physics 31, 383–391 (2009). https://doi.org/10.1016/j.astropartphys.2009.04.001

    doi:10.1016/j.astropartphys.2009.04.001 2009

Show all 62 references
  1. [8]

    Becherini, A

    Y. Becherini, A. Djannati-Atai, V. Marandon, et al., A new analy- sis strategy for detection of faint γ/-ray sources with imaging atmo- spheric Cherenkov telescopes. Astroparticle Physics 34, 858–870 (2011). https://doi.org/10.1016/j.astropartphys.2011.03.005

    2011 doi

  2. [9]

    Sharma, J

    M. Sharma, J. Nayak, M.K. Koul, et al., Gamma/hadron segregation for a ground based imaging atmospheric Cherenkov telescope using machine learning methods: Random forest leads. Research in Astron- omy and Astrophysics 14, 1491 (2014). https://doi.org/10.1088/1674- 4527/14/11/012

    2014 doi

  3. [10]

    D’Amico, Statistical tools for imaging atmo- spheric Cherenkov telescopes

    G. D’Amico, Statistical tools for imaging atmo- spheric Cherenkov telescopes. Universe 8, (2022). https://doi.org/10.20944/preprints202201.0347.v1

    2022

  4. [11]

    Fanelli, Machine learning for imaging Cherenkov detectors

    C. Fanelli, Machine learning for imaging Cherenkov detectors. Journal of Instrumentation 15, C02012 (2020). https://doi.org/10.1088/1748- 0221/15/02/c02012

    2020 doi

  5. [12]

    Malyshev, L

    D. Malyshev, L. Mohrmann, Analysis methods for gamma-ray astron- omy. arXiv preprint arXiv:2309.02966 (2023)

    2023 arXiv

  6. [14]

    Reynolds, Neural networks to vhe gamma-ray atmospheric Cherenkov crab nebula imaging data

    P.T. Reynolds, Neural networks to vhe gamma-ray atmospheric Cherenkov crab nebula imaging data. Irish Astronomical Journal, 21(2), 118 (1993)

    1993

  7. [15]

    Nieto Casta˜ no, A

    D. Nieto Casta˜ no, A. Brill, B. Kim, T.B. Humensky, Exploring deep learning as an event classification method for the Cherenkov Telescope Array. PoS ICRC2017, 809 (2017). https://doi.org/10.22323/1.301.0809

    2017 doi

  8. [16]

    Acharya, M

    B. Acharya, M. Actis, T. Aghajani, et al., Introducing the CTA concept. Astroparticle physics 43, 3–18 (2013). https://doi.org/10.1016/j.astropartphys.2013.01.007. 14

    2013 doi

  9. [17]

    D. Heck, J. Knapp, J. Capdevielle, et al., Corsika: A Monte Carlo code to simulate extensive air showers. Report FZKA-6019 (1998)

    1998

  10. [18]

    K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, in: Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770–778 (2016). https://doi.org/10.1109/cvpr.2016.90

    2016 doi

  11. [19]

    Szegedy, V

    C. Szegedy, V. Vanhoucke, S. Ioffe, et al., Rethinkingthe inception ar- chitecture for computer vision, in: Proceedings of the IEEE confer- ence on computer vision and pattern recognition, pp. 2818–2826 (2016). https://doi.org/10.1109/cvpr.2016.308

    2016 doi

  12. [20]

    Shilon, M

    I. Shilon, M. Kraus, M. B¨ uchele, et al., Application of deep learning methods to analysis of imaging atmospheric Cherenkov telescopes data. Astroparticle Physics 105, 44–53 (2019). https://doi.org/10.1016/j.astropartphys.2018.10.003

    2019 doi

  13. [21]

    Ashton, M

    T. Ashton, M. Backes, A. Balzer, et al., A nectar- based upgrade for the Cherenkov cameras of the hess 12- meter telescopes. Astroparticle Physics 118, 102425 (2020). https://doi.org/10.1016/j.astropartphys.2019.102425

    2020

  14. [22]

    Holch, I

    T.L. Holch, I. Shilon, M. B¨ uchele, et al., Probing convolu- tional neural networks for event reconstruction in γ-ray astronomy with Cherenkov telescopes. arXiv preprint arXiv:1711.06298 (2017). https://doi.org/10.22323/1.301.0795

    2017 arXiv

  15. [23]

    Nieto Casta˜ no, A

    D. Nieto Casta˜ no, A. Brill, Q. Feng, et al., Studying Deep Convolu- tional Neural Networks With Hexagonal Lattices for Imaging Atmo- spheric Cherenkov Telescope Event Reconstruction. PoS ICRC2019, 753 (2019). https://doi.org/10.22323/1.358.0753

    2019 doi

  16. [24]

    Jacquemont, L

    M. Jacquemont, L. Antiga, T. Vuillaume, et al., Indexed op- erations for non-rectangular lattices applied to convolutional neu- ral networks, in: VISAPP, 14th International Conference on Computer Vision Theory and Applications, pp. 362–371 (2019). https://doi.org/10.5220/00073...

    2019 doi

  17. [25]

    Ketkar, Deep Learning with Python: A Hands-on Introduction(Apress, Berkeley CA, 2017) pp

    N. Ketkar, Deep Learning with Python: A Hands-on Introduction(Apress, Berkeley CA, 2017) pp. 195–208. https://doi.org/10.1007/978-1-4842-2766-4 12. 15

    2017 doi

  18. [26]

    Parsons, S

    R.D. Parsons, S. Ohm, Background rejection in atmospheric Cherenkov telescopes using recurrent convolutional neural net- works. The European Physical Journal C 80, 1–11 (2020). https://doi.org/10.1140/epjc/s10052-020-7953-3

    2020 doi

  19. [27]

    Spencer, T

    S. Spencer, T. Armstrong, J. Watson, et al., Deep learning with pho- tosensor timing information as a background rejection method for the Cherenkov telescope array. Astroparticle Physics 129, 102579 (2021). https://doi.org/10.1016/j.astropartphys.2021.102579

    2021

  20. [28]

    Feng, T.T

    Q. Feng, T.T. Lin, The analysis of veritas muon images using convolu- tional neural networks. Proceedings of the International Astronomical Union 12, 173–179 (2016). https://doi.org/10.1017/s1743921316012734

    2016 doi

  21. [29]

    Riquelme, M

    D. Riquelme, M. Araya, S. Borquez, et al.,Deep learning semi- supervised strategy for gamma/hadron classification of imag- ing atmospheric Cherenkov telescope events, in: Proceedings of the 12th International Conference on Pattern Recognition Applications and Methods (ICPRAM 2...

    2023 doi

  22. [30]

    S. De, W. Maitra, V. Rentala, A.M. Thalapillil, Deep learning techniques for imaging air Cherenkov telescopes. arXiv preprint arXiv:2206.05296 (2022). https://doi.org/10.1103/PhysRevD.107.083026

    2022 arXiv

  23. [31]

    Mangano, C

    S. Mangano, C. Delgado, M.I. Bernardos, et al., Extracting gamma-ray information from images with convolutional neural network methods on simulated Cherenkov telescope array data, in: IAPR Workshop on Ar- tificial Neural Networks in Pattern Recognition (Springer, Berlin, 2018)...

    2018 doi

  24. [32]

    Postnikov, A

    E. Postnikov, A. Kryukov, S. Polyakov, et al., Gamma/hadron sep- aration in imaging air Cherenkov telescopes using deep learning li- braries TensorFlow and PyTorch, in: Journal of Physics: Con- ference Series, 1181, 012048 (2019). https://doi.org/10.1088/1742- 6596/1181/1/012048

    2019 doi

  25. [33]

    Polyakov, A

    S. Polyakov, A. Demichev, A. Kryukov, E. Postnikov, The use of convolutional neural networks for processing images from mul- tiple IACTs in the TAIGA experiment. PoS 395, 753 (2021). https://doi.org/10.22323/1.395.0753. 16

    2021 doi

  26. [34]

    Budnev, I

    N. Budnev, I. Astapov, P. Bezyazeekov, et al., Taiga—an advanced hy- brid detector complex for astroparticle physics and high energy gamma- ray astronomy in the tunka valley. Journal of Instrumentation 15, C09031 (2020). https://doi.org/10.1088/1748-0221/15/09/C09031

    2020 doi

  27. [35]

    Abadi, A

    M. Abadi, A. Agarwal, P. Barham, et al., Tensorflow: Large-scale ma- chine learning on heterogeneous distributed systems. arXiv preprint arXiv:1603.04467 (2016) https://doi.org/10.48550/arXiv.1603.04467

    2016 doi arXiv

  28. [36]

    E. Gres, A. Kryukov, Energy reconstruction in analysis of Cherenkov telescopes images in TAIGA experiment using deep learning methods. PoS DLCP2022, 002 (2022). https://doi.org/10.22323/1.429.0002

    2022 doi

  29. [37]

    Jacquemont, T

    M. Jacquemont, T. Vuillaume, A. Benoit, et al., Single imaging at- mospheric Cherenkov telescope full-event reconstruction with a deep multi-task learning architecture, in: Astronomical Data Analysis Soft- ware and Systems ADASS XXX, pp. 1–5 (2020). https://hal.archives- ouver...

    2020

  30. [38]

    Jacquemont, T

    M. Jacquemont, T. Vuillaume, A. Benoit, et al., Multi-task ar- chitecture with attention for imaging atmospheric Cherenkov tele- scope data analysis, in: 16th International Conference on Comput- erVision Theory and Applications (VISAPP 2021), pp. 1–12 (2021). https://doi.org/1...

    2021 doi

  31. [39]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, et al., Attention is all you need. arXiv preprint arXiv:1706.03762 (2017). https://doi.org/10.48550/arXiv.1706.03762

    2017 doi arXiv

  32. [40]

    J. Sun, F. Darbehani, M. Zaidi, B. Wang, Saunet: Shape attentive u-net for interpretable medical image segmentation, in: International Con- ference on Medical Image Computing and Computer-Assisted Interven- tion, (Springer, Berlin, 2020) pp. 797–806. https://doi.org/10.1007/97...

    2020 doi

  33. [41]

    Zhang, Q

    Y. Zhang, Q. Yang, A survey on multi-task learning. arXiv preprint arXiv:1707.08114 (2021). https://doi.org/10.48550/arXiv.1707.08114

    2021 doi arXiv

  34. [42]

    Parmar, A

    N. Parmar, A. Vaswani, J. Uszkoreit, et al., Im- age transformer, arXiv preprint arXiv:1802.05751 (2018). https://doi.org/10.48550/arXiv.1802.05751. 17

    2018 doi arXiv

  35. [43]

    Miener, D

    T. Miener, D. Nieto, A. Brill, et al., Reconstruction of stereoscopic CTA events using deep learning with CTLearn. PoS ICRC2021, 730 (2021). https://doi.org/10.22323/1.395.0730

    2021 doi

  36. [44]

    Miener, D

    T. Miener, D. Nieto, R. L´ opez-Coto, et al., The perfor- mance of the magic telescopes using deep convolutional neu- ral networks with ctlearn. arXiv preprint arXiv:2211.16009 (2022). https://doi.org/10.48550/arXiv.2211.16009

    2022 doi arXiv

  37. [45]

    Grespan, M

    P. Grespan, M. Jacquemont, R. Lopez-Coto, et al., Deep-learning- driven event reconstruction applied to simulated data from a sin- gle Large-Sized Telescope of CTA. PoS ICRC2021, 771 (2022). https://doi.org/10.22323/1.395.0771

    2022 doi

  38. [46]

    Jacquemont, T

    M. Jacquemont, T. Vuillaume, A. Benoit, et al., First full-event reconstruction from imaging atmospheric Cherenkov telescope real data with deep learning, in: 2021 International Conference on Content-Based Multimedia Indexing (IEEE, New York, 2021) pp. 1–6. https://doi.org/10....

    2021

  39. [47]

    H. Abe, A. Aguasca, I. Agudo, et al., Analysis of the Cherenkov Telescope Array first Large Size Telescope real data using convolutional neural networks. PoS ICRC2021, 703 (2021). https://doi.org/10.22323/1.395.0703

    2021 doi

  40. [48]

    Juryˇ sek, E

    J. Juryˇ sek, E. Lyard, R. Walter, Full lst-1 data reconstruction with the use of convolutional neural networks. arXiv preprint arXiv:2111.14478 (2021). https://doi.org/10.48550/arXiv.2111.14478

    2021 doi arXiv

  41. [49]

    Bylund, G

    T. Bylund, G. Kukec Mezek, M. Senniappan, et al., Studies of Gamma Ray Shower Reconstruction Using Deep Learning. PoS ICRC2021, 758 (2021). https://doi.org/10.22323/1.395.0758

    2021 doi

  42. [50]

    Brill, Q

    A. Brill, Q. Feng, T.B. Humensky, et al., Investigating a deep learning method to analyze images from multiple gamma-ray telescopes, in: 2019 New York Scientific Data Summit (IEEE, New York, 2019) pp. 1–4. https://doi.org/10.1109/nysds.2019.8909697

    2019

  43. [51]

    Dubenskaya, A

    J. Dubenskaya, A. Kryukov, A. Demichev, Fast simulation of gamma/proton event images for the TAIGA-IACT experiment us- ing generative adversarial networks. PoS ICRC2021, 874 (2021). https://doi.org/10.22323/1.395.0874. 18

    2021 doi

  44. [52]

    Dubenskaya, A

    J. Dubenskaya, A. Kryukov, A. Demichev, Modeling Images of Proton Events for the TAIGA Project Using a Generative Adversaria Network: Features of the Network Architecture and the Learning Process. PoS DLCP2021, 011 (2021). https://doi.org/10.22323/1.410.0011

    2021 doi

  45. [53]

    Dubenskaya, A

    J. Dubenskaya, A. Kryukov, A. Demichev, 2021a. Architecture of a gen- erative adversarial network and preparation of input data for modeling gamma event images for the taiga-iact experiment. CEUR Workshop Proceedings 3041, 270–274. doi:10.54546/MLIT.2021.11.22.001

    doi

  46. [54]

    Postnikov, A.A

    E.B. Postnikov, A.A. Grinyuk, L.A. Kuzmichev, L.G. Svesh- nikova, Hybrid method for identifying mass groups of primary cosmic rays in the joint operation of IACTs and wide angle Cherenkov timing arrays. J.Phys.: Conf. Series 798, 012030 (2017). https://doi.org/10.48550/arXiv.1...

    2017 doi arXiv

  47. [55]

    Goodfellow, J

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, et al., Generative adversar- ial networks. Communications of the ACM, 63 (11), 139 - 144 (2020). http://dx.doi.org/10.1145/3422622

    2020 doi

  48. [56]

    Johnson, T.M

    J.M. Johnson, T.M. Khoshgoftaar, Survey on deep learning with class imbalance. Journal of Big Data 6, 1–54 (2019). https://doi.org/10.1186/s40537-019-0192-5

    2019 doi

  49. [57]

    Dubenskaya, A

    J. Dubenskaya, A. Kryukov, A. Demichev, et al., Using a Conditional Generative Adversarial Network to Control the Statistical Characteris- tics of Generated Images for IACT Data Analysis. PoS 429, 004 (2022). https://doi.org/10.22323/1.429.0004

    2022 doi

  50. [58]

    Dubenskaya, A

    J. Dubenskaya, A. Kryukov, A. Demichev, et al., Image Data Augmen- tation for the TAIGA-IACT Experiment with Conditional Generative Adversarial Networks. Moscow University Physics Bulletin 79 (2024)

    2024

  51. [59]

    Polyakov, A

    S. Polyakov, A. Kryukov, E. Gres, et al., Using conditional varia- tional autoencoders to generate images from atmospheric Cherenkov telescopes. PoS 429, 003 (2022). https://doi.org/10.22323/1.429.0003

    2022 doi

  52. [60]

    Nieto Casta˜ no, A

    D. Nieto Casta˜ no, A. Brill, Q. Feng, et al., CTLearn: Deep Learning for Gamma-ray Astronomy. PoS 358, 752 (2019). https://doi.org/10.22323/1.358.0752

    2019 doi

  53. [61]

    Jacquemont, T

    M. Jacquemont, T. Vuillaume, A. Benoit, et al., GammaLearn: A Deep Learning Framework for IACT Data. PoS 358, 705 (2019). https://doi.org/10.22323/1.358.0705. 19

    2019 doi

  54. [62]

    Vuillaume, J

    T. Vuillaume, J. Mikael, L. Antiga, et al., Gammalearn-first steps to apply deep learning to the Cherenkov telescope array data, in: EPJ Web of Conferences, EDP Sciences. p. 06020 (2019). doi:https://doi.org/10.1051/epjconf/201921406020. 20

    2019

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.