Pith. sign in

REVIEW 4 major objections 4 minor 52 references

A deep neural network trained on augmented solar spectrograms can detect radio bursts with expert-level precision and higher recall, enough to automate the e-Callisto network.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 00:50 UTC pith:TES3AUVS

load-bearing objection A valuable open dataset and a plausible detector, but the headline recall gain over the human catalog is an in-sample artifact until the threshold is fixed prospectively. the 4 major comments →

arxiv 2607.26014 v1 pith:TES3AUVS submitted 2026-07-28 astro-ph.SR astro-ph.IM

Automated Solar Radio Burst Detection Using Deep Learning on Augmented e-Callisto Data

Vincenzo Timmel , Andr\'e Csillaghy , Christian Monstein This is my paper
classification astro-ph.SR astro-ph.IM
keywords solar radio burstse-Callistodeep learningResNetSpecAugmentTimeWarpspectrogram analysisexplainable AI
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper tries to establish that a single deep learning model, FlareSense, can replace the human expert's routine solar radio burst detection across the heterogeneous e-Callisto network. Using a ResNet trained on manually labeled spectrograms and augmented with SpecAugment and TimeWarp, the model reaches 93% precision and 73.15% recall on a clean test set. At the same precision as the routine expert catalog, it recovers more bursts (73.15% vs 63%), suggesting that day-to-day manual inspection misses a substantial fraction of events. If true, this makes near-real-time burst alerts and long-term statistical studies feasible without the labeling bottleneck.

Core claim

FlareSense, a Residual Network binary classifier, detects solar radio bursts in e-Callisto spectrograms with 93% precision and 73.15% recall on a clean, re-verified test set. When its decision threshold is calibrated to match the precision of the routine expert catalog (93%), it achieves higher recall than routine cataloging (73.15% vs 63%). The paper argues this shows the model recovers bursts missed under operational conditions, and that data augmentation is a key driver: disabling both augmentations drops median test F1 from 77.2% to 71.0%, and augmentation narrows the performance gap on instruments never seen in training.

What carries the argument

The central mechanism is a ResNet-34 binary classifier fed with 128×512 spectrogram crops, trained with weighted binary cross-entropy on a 304,750-sample dataset where each burst window is matched by ten non-burst windows. Two speech-processing augmentations adapted to solar spectrograms carry the generalization argument: SpecAugment masks random time/frequency bands, and TimeWarp elastically deforms the time axis, imitating burst drift-rate variation and instrument-specific interference. A temperature-scaling calibration step (temperature 0.4974) maps raw logits to probabilities for threshold comparison with the human catalog.

Load-bearing premise

The ground truth for the test set is a single expert's second-pass re-inspection of the same catalog the model was trained on, so the reported precision and recall measure agreement with that one expert's judgment; if that re-verification is biased, the headline numbers may not hold against other experts.

What would settle it

Have a second, independent solar radio astronomer label the same 30,549 test spectrograms without seeing the first labels or the model outputs; if the two experts' labels disagree on more than a few percent of cases, or if the model's recall against the second expert's labels drops materially below 73%, the claimed superiority over routine cataloging is an artifact of the re-verification procedure.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • A single model can monitor the whole e-Callisto network, issuing a burst alert roughly 30 seconds after each 15-minute observation is uploaded.
  • At equal precision, the automated detector recovers more bursts than routine day-to-day expert cataloging, meaning some events are currently being missed operationally.
  • Data augmentation with SpecAugment and TimeWarp improves detection on instruments not seen during training, supporting deployment to newly added stations.
  • The public release of dataset, code, and model (MIT license) lets other groups reproduce and extend the detector.
  • The model's precision likely underestimates its true performance, since some apparent false positives were re-identified as real bursts missed by the ground-truth annotation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • One implication the authors leave implicit: the detector could be run retrospectively on years of archived e-Callisto data to build a denser burst catalog, since its false positives often appear to be genuine bursts the routine catalog omitted.
  • If the clean-test labels reflect a single expert's judgment, the reported 73% recall may not be an objective upper bound; comparing against a second independent annotator would clarify how much of the gap over routine cataloging reflects detector skill vs label noise.
  • The same augmentation recipe could transfer to other spectrogram-classification problems, such as burst-type classification or detection in other radio arrays, since it targets generic failure modes of missing bands and drift-rate variation.
  • If deployed, the decision threshold can be tuned to trade precision for recall—for example, lowered to capture faint type III bursts at the cost of more false alarms.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper introduces FlareSense, a ResNet-34-based binary classifier for solar radio burst detection in e-Callisto spectrograms, trained on a large manually labeled dataset with SpecAugment and TimeWarp augmentation. The model is claimed to achieve 93% precision and 73.15% recall on a clean, re-verified test set, and to outperform the routine expert catalog (63% recall) at matched precision. The authors also evaluate generalization to unseen instruments and present Gradient SHAP attribution maps. The dataset, code, and model are publicly released.

Significance. If the evaluation were prospective, this would be a valuable contribution to automated space-weather monitoring. The paper's strengths are its public release of data, code, and model, the large multi-instrument training corpus, and the per-instrument diagnostics. However, the central performance claim is weakened by an evaluation protocol in which the decision threshold is selected on the test set, and by a test set whose labels are re-verified by the same expert who created the training labels. These issues are fixable and do not invalidate the underlying engineering, but the reported numbers should not be taken at face value.

major comments (4)
  1. [Section 6, Table 3] The matched-precision comparison uses a calibrated probability threshold of 0.426 that is selected on the clean test set ("at which FlareSense has the same precision as the routine catalog baseline"). This makes the reported 73.15% recall an in-sample operating point rather than a prospective result. The headline claim that FlareSense outperforms routine cataloging at equal precision is therefore not established by the reported protocol. The threshold should be fixed on the validation set (or via a pre-registered rule) and then applied to the test set, or the full PR curve with confidence bands should be presented with recall reported at a threshold chosen independently of the test data.
  2. [Section 5.2, Section 6, Table 3] The metric aggregation is ambiguous. Section 5.2 states that validation F1 is computed as an unweighted mean across instruments (macro), while Section 6 says "we report both micro and macro metrics" but Table 3 does not specify whether the 93% precision / 73.15% recall are micro or macro. Per-instrument Table 5 implies a macro recall of approximately 71.7%, noticeably different from 73.15%; the distinction matters. Please label Table 3 as micro or macro (or report both) and provide confidence intervals, e.g., via bootstrap.
  3. [Section 4.4] The "clean test labels" were re-verified by the same PI who created the training catalog; no independent second annotator was used. The test set is therefore not an external gold standard, and the reported precision/recall measure agreement with one expert's second-pass judgment. This limits the external validity of the absolute performance numbers. If feasible, obtain a second independent annotation of a subset and report inter-annotator agreement; otherwise temper the claim that the test set is "clean" and "double-checked" — or explicitly state that it is double-checked by the same observer.
  4. [Section 5.3, Figure 3] The ablation comparing augmentation conditions reports "subsequent performance on the held-out test data," and the conclusion that both SpecAugment and TimeWarp together are best is drawn from these test-set comparisons. Since the final model uses both augmentations, the test set has influenced the choice of model configuration, meaning the test set is not fully held out. This is a separate test-set usage from the threshold issue and also compromises the confirmatory evaluation. Please clearly separate exploratory/ablation analysis from final evaluation, e.g., by using a nested holdout or a separate test set for the chosen configuration.
minor comments (4)
  1. [Section 7, Figure 7] The caption and text suggest the shown false negative "might not be a true burst" and "might represent a mislabeled data point," yet the evaluation treats the clean test labels as ground truth. Please clarify whether this concern applies to the test label or to the model's prediction, and whether the evaluation would change if the label were corrected. As written, this creates confusion about the ground-truth status.
  2. [Section 6] The statement "most of the false negatives are either very faint, Type III bursts or missed during annotation" is qualitative. Please provide a quantitative breakdown of the false-negative categories if possible, since it bears on the practical significance of the missing detections.
  3. [Abstract and Section 8] The headline numbers "93% precision and 73.15% recall" should be qualified as an operating point selected on the test set, or the protocol should be changed so they represent a prospective result. If the latter, update the abstract accordingly.
  4. [Figure 2 caption] Minor grammatical issue: "applied to a e-Callisto spectrogram" should be "applied to an e-Callisto spectrogram."

Circularity Check

0 steps flagged

No significant circularity: empirical pipeline, no derivation reduces to its inputs; caveats are evaluation-protocol issues, not circular steps.

full rationale

This paper is an empirical machine-learning study, not a derivation chain. The central claims are (i) a ResNet trained on catalog-derived spectrograms reaches 93% precision and 73.15% recall on a re-verified test set, and (ii) at matched precision it outperforms routine catalog recall. Neither claim is produced by an equation whose inputs already contain the output. Training labels, validation-based hyperparameter selection, temperature scaling fit on training predictions, and test evaluation are distinct stages. The main concerns flagged in the manuscript and by reviewers—same-PI label re-verification (Sec. 4.4) and the test-set-selected threshold 0.426 used for the matched-precision comparison (Sec. 6)—are real threats to the generalizability and optimism of the reported operating point, but they are not circular reductions: the recall at the chosen threshold is an empirical measurement of the model's score ranking, not a quantity forced to equal an input by construction. The self-citations (e.g., ecallisto_ng [29], Callisto quicklook [4]) are tooling/data references and do not carry the argument. Therefore no circular step meeting the quoted-evidence threshold can be identified.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The central claim is an empirical ML benchmark. It rests on the fidelity of the burst catalog as ground truth and on the PI's second-pass re-verification of the test set. Hyperparameters and temperature are fitted to validation/training predictions. No new physical entities are introduced.

free parameters (4)
  • Temperature scaling T = 0.4974
    Fitted on training-set predictions to calibrate probabilities; used to set the 0.426 threshold for the matched-precision comparison.
  • Hyperparameters (learning rate, weight decay, label smoothing, epochs, warm-up, model type, augmentation parameters) = 2.4e-4 LR, 5.2e-4 WD, 0.1174 LS, 25 epochs, 12 warm-up, ResNet34, freq mask 25, time mask 70, warp width 389
    Chosen via Bayesian hyperparameter sweep maximizing validation F1 (Section 5.2, Table 2).
  • Class weight w1 = ~10
    Derived from the 1:10 positive-to-negative sampling ratio; affects the loss and operating point.
  • Data-generation thresholds (offset range, fmin, duration) = offset [0,10] min, fmin=150, duration ≥14 min
    Ad hoc quality criteria in Algorithms 2 and 3; not tuned to model performance but shape the dataset.
axioms (4)
  • domain assumption The e-Callisto burst catalog (Monstein catalog) provides sufficiently accurate labels for training.
    Section 4.2 treats the human-labeled catalog as ground truth; the paper acknowledges label noise but relies on it.
  • domain assumption The PI's second-pass re-verification of test labels is a correct and unbiased ground truth.
    Section 4.4: test labels re-inspected by the PI without model access; no independent annotator.
  • standard math ResNet training with weighted BCE and the described preprocessing preserves burst-relevant information.
    Standard ML practice; not justified beyond citing prior work.
  • domain assumption Gradient SHAP attributions are at least qualitatively indicative of the features the model uses.
    Used as evidence in Section 7; the paper itself notes these maps 'do not by themselves establish causal model behavior.'

pith-pipeline@v1.3.0-alltime-deepseek · 16879 in / 12361 out tokens · 104593 ms · 2026-08-01T00:50:26.487722+00:00 · methodology

0 comments
read the original abstract

Solar radio bursts are signatures of energetic events associated with solar flares and coronal mass ejections and can interfere with terrestrial and space-based communication systems. Real-time automatic burst monitoring enables early warnings tens of minutes to hours before associated particles reach Earth and provides the basis for long-term statistical studies. The e-Callisto network is a worldwide system of solar radio spectrometers providing continuous observations, with its instruments collectively covering frequencies from approximately 20 MHz to 1 GHz. Burst detection and labeling currently rely largely on human experts, limiting scalability and real-time applicability due to hardware heterogeneity and low signal-to-noise ratios.

Figures

Figures reproduced from arXiv: 2607.26014 by Andr\'e Csillaghy, Christian Monstein, Vincenzo Timmel.

Figure 1
Figure 1. Figure 1: A solar radio burst seen by one of the instruments of the e-Callisto network in Australia. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Examples of spectrogram data augmentation methods being applied to a e-Callisto spectrogram. Top: Unaugmented spectrogram, showing a solar radio burst seen by the instrument AUSTRALIA-ASSA_01 on the 7th of May 2021. This example is shown for illustration only and is not part of the dataset described in Section 4.3. Bottom: TimeWarp smoothly deforms the spectrogram along the time axis, imitating variations … view at source ↗
Figure 3
Figure 3. Figure 3: Comparison of FlareSense test-set performance under different data augmentation strategies. Each box represents the distribution of test F1-scores across models produced by hyperparameter sweeps with identical search ranges (see Section 5.2); model optimization within each sweep used the validation split. The results show that both SpecAugment and TimeWarp independently improve detection accuracy, while th… view at source ↗
Figure 4
Figure 4. Figure 4: Left: Validation precision across the 17 instruments withheld from training for models trained with and without data augmentation. Right: Corresponding validation recall for the same comparison. Both panels show results only for instruments unseen during training, illustrating how data augmentation improves cross-instrument generalization [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Precision-recall (PR) curves for the train and test datasets. Each point corresponds to a different probability threshold. Macro PR values are computed by averaging precision and recall across instruments (equal weight per instrument), while micro PR values are calculated globally over all samples (weighting instruments by the number of examples). The higher test-set curve reflects the cleaner, double-chec… view at source ↗
Figure 6
Figure 6. Figure 6: A true positive detection by FlareSense, with the corresponding Gradient SHAP attribution map below. The [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: False negative by FlareSense. Before temperature scaling, the model assigns a burst probability of 45.9%, [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Apparent false positive originating from a ground-truth annotation miss. FlareSense assigns a burst probability [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: True positive - BIR_01. Attribution concentrates on the burst ridge and suppresses background structure. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: True positive - Australia-ASSA_62. Attribution follows the salient emission trace, showing consistent behavior across instruments [PITH_FULL_IMAGE:figures/full_fig_p021_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: True positive - Germany-DLR_63. Attribution localizes along the drift-like burst pattern. 21 [PITH_FULL_IMAGE:figures/full_fig_p021_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: True positive - Glasgow_01 [PITH_FULL_IMAGE:figures/full_fig_p022_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: True positive - Humain_59. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: False positive - India-Gauri_01. This is another example for an annotation miss (see Section 7), so in effect, this case is another true positive [PITH_FULL_IMAGE:figures/full_fig_p023_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: True positive - SSRT_59. Strong attribution along the emission band indicates sensitivity to real events. E Probability Calibration Neural-network confidence scores are not necessarily calibrated probabilities (46). Additionally, label smoothing (47) makes the distribution of the output probabilities less sharp. We therefore fitted a single temperature parameter by minimizing the negative log likelihood o… view at source ↗
Figure 16
Figure 16. Figure 16: Distributions of model probabilities before and after temperature scaling. The fitted temperature of 0.4974 [PITH_FULL_IMAGE:figures/full_fig_p024_16.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

52 extracted references · 17 canonical work pages

  1. [1]

    Callisto radio spectrometer

    ASSA. Callisto radio spectrometer. https://www.assa.org.au/resources/radio-astronomy/ callisto-radio-spectrometer/, 2023. Most recent update 2 Nov 2023; accessed 8 Oct 2025

    2023

  2. [2]

    Solar radio bursts and space weather.arXiv:2405.00959, 2024

    Stephen M White. Solar radio bursts and space weather.arXiv:2405.00959, 2024

    Pith/arXiv arXiv 2024

  3. [3]

    A world-wide net of solar radio spectrometers: e-callisto.Earth, Moon, and Planets, 104(1–4):277–285, 2009

    AO Benz, C Monstein, H Meyer, PK Manoharan, R Ramesh, A Altyntsev, A Lara, J Paez, and K-S Cho. A world-wide net of solar radio spectrometers: e-callisto.Earth, Moon, and Planets, 104(1–4):277–285, 2009. doi: 10.1007/s11038-008-9267-6. URLhttps://doi.org/10.1007/s11038-008-9267-6

    doi:10.1007/s11038-008-9267-6 2009

  4. [4]

    Christian Monstein, André Csillaghy, and Arnold O. Benz. Callisto quicklook solar spectrogram plots, 2023. URL https://spase-metadata.org/ISWI/DisplayData/Callisto/FAS/PT15M

    2023

  5. [5]

    Deep residual learning for image recognition, 2015

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition, 2015. URLhttps://arxiv.org/abs/1512.03385

    Pith/arXiv arXiv 2015

  6. [6]

    Nur Zulaikha Mohd Afandi, N. H. Sabri, Roslan Umar, and Christian Monstein. Burst-finder: burst recognition for e-callisto spectra.Indian Journal of Physics, 94:947–957, 2020. doi: 10.1007/s12648-019-01551-2. URL https://doi.org/10.1007/s12648-019-01551-2

    work page doi:10.1007/s12648-019-01551-2 2020

  7. [7]

    Automatic detection of e-callisto solar radio bursts by deep neural networks

    Mario Fernández Ruiz, Javier Bussons Gordo, Manuel Prieto Mateo, and Christian Monstein. Automatic detection of e-callisto solar radio bursts by deep neural networks. In2022 3rd URSI Atlantic and Asia Pacific Radio Science Meeting (AT-AP-RASC), pages 1–4, 2022. doi: 10.23919/AT-AP-RASC54737.2022.9814298

    arXiv 2022

  8. [8]

    Temmer, F

    Lukas Höfig, M. Temmer, F. Koller, L. Drescher, Marcin Latocha, and C. Monstein. Robust – a radio burst identification algorithm for space weather warning using the e-callisto station at university of graz. 2023. 19th European Space-Weather Week ; Conference date: 20-11-2023 Through 24-11-2023

    2023

  9. [9]

    Mohd Afandi, R

    N.Z. Mohd Afandi, R. Umar, N.H. Sabri, S. Safei, C. Monstein, C.C. Lau, and S.N.A. Syed Zafar. Burst- classifier: Automated classification of solar radio burst type ii, iii and iv for callisto spectra using physical properties during maximum of solar cycle 24.Advances in Space Research, 74(11):6104–6123, 2024. ISSN 0273-1177. doi: 10.1016/j.asr.2024.09.00...

    work page doi:10.1016/j.asr.2024.09.001 2024

  10. [10]

    A study of real-time detection methods for solar radio burst identification.The Astrophysical Journal Supplement Series, 276 (1):18, jan 2025

    YongLin Yu, ChengLong Qiao, Shuo Xu, QinZheng Li, YanRui Su, HongQiang Song, and FaBao Yan. A study of real-time detection methods for solar radio burst identification.The Astrophysical Journal Supplement Series, 276 (1):18, jan 2025. doi: 10.3847/1538-4365/ad91a2. URL https://dx.doi.org/10.3847/1538-4365/ad91a2

    work page doi:10.3847/1538-4365/ad91a2 2025

  11. [11]

    The analysis of type ii and type iii solar radio bursts: Gui for the e-callisto data.New Astronomy, 109:102194,

    Yashan Hettiarachchi, Janaka Adassuriya, Chandana Jayaratne, Sasani Jayawardhana, and Christian Monstein. The analysis of type ii and type iii solar radio bursts: Gui for the e-callisto data.New Astronomy, 109:102194,

  12. [12]

    Automated recognition of type iii solar radio bursts using mathematical morphology

    James Jones. Automated recognition of type iii solar radio bursts using mathematical morphology. InProceedings of the Advanced Maui Optical and Space Surveillance Technologies Conference. Maui Economic Development Board, September 2014. Atmospherics/Space Weather track

    2014

  13. [13]

    A method for the automated detection of solar radio bursts in dynamic spectra.J

    Houssam Salmane, Rodolphe Weber, Karim Abed-Meraim, Karl-Ludwig Klein, and Xavier Bonnin. A method for the automated detection of solar radio bursts in dynamic spectra.J. Space Weather Space Clim., 8:A43, 2018. doi: 10.1051/swsc/2018028. URLhttps://doi.org/10.1051/swsc/2018028

    arXiv 2018

  14. [14]

    V . V . Lobzin, I. H. Cairns, P. A. Robinson, G. Steward, and G. Patterson. Automatic recognition of type III solar radio bursts: Automated radio burst identification system method and first observations.Space Weather, 7:S04002,

  15. [15]

    Random forests.Machine Learning, 45(1):5–32, 2001

    Leo Breiman. Random forests.Machine Learning, 45(1):5–32, 2001. ISSN 1573-0565. doi: 10.1023/A: 1010933404324. URLhttps://doi.org/10.1023/A:1010933404324

    doi:10.1023/a: 2001

  16. [16]

    Carley, Peter T

    Eoin P. Carley, Peter T. Gallagher, Joe McCauley, and Pearse Murphy. Using supervised machine learning to automatically detect type ii and iii solar radio bursts. InEGU General Assembly 2020, EGU2020-5109, 2020. doi: 10.5194/egusphere-egu2020-5109. Online, 4-8 May 2020

    work page doi:10.5194/egusphere-egu2020-5109 2020

  17. [17]

    Ignacio Hidalgo, and Christian Monstein

    Javier Bussons Gordo, Mario Fernández Ruiz, Manuel Prieto Mateo, Jorge Alvarado Díaz, Francisco Chavez de la O, J. Ignacio Hidalgo, and Christian Monstein. Automatic burst detection in solar radio spectrograms 15 APREPRINT- JULY29, 2026 using deep learning: dearce method.Solar Physics, 298(6):82, 2023. doi: 10.1007/s11207-023-02171-0. URL https://doi.org/...

    work page doi:10.1007/s11207-023-02171-0 2026

  18. [18]

    Solar radio burst detection based on the mobilevit-ssdlite lightweight model.The Astrophysical Journal Supplement Series, 269(2):51, November 2023

    Hailan He, Guowu Yuan, Hao Zhou, Chengming Tan, and Shaojie Guo. Solar radio burst detection based on the mobilevit-ssdlite lightweight model.The Astrophysical Journal Supplement Series, 269(2):51, November 2023. doi: 10.3847/1538-4365/ad036c

    work page doi:10.3847/1538-4365/ad036c 2023

  19. [19]

    Stalder, Luis Salgueiro, and Jorge Molina

    Iván González Orué, Diego H. Stalder, Luis Salgueiro, and Jorge Molina. Automatic solar radio burst detection using deep learning. In2023 IEEE CHILEAN Conference on Electrical, Electronics Engineering, Information and Communication Technologies (CHILECON), pages 1–6, 2023. doi: 10.1109/CHILECON60335.2023.10418691. URLhttps://doi.org/10.1109/CHILECON60335....

    arXiv 2023

  20. [20]

    Deep active learning–based classification of solar radio spectrogram data.The Astrophysical Journal Supplement Series, 279(1):25, 2025

    Yan Liu, HongQiang Song, Fabao Yan, and Yan-Rui Su. Deep active learning–based classification of solar radio spectrogram data.The Astrophysical Journal Supplement Series, 279(1):25, 2025. doi: 10.3847/1538-4365/adda30. URLhttps://doi.org/10.3847/1538-4365/adda30

    work page doi:10.3847/1538-4365/adda30 2025

  21. [21]

    Y . H. Wang, S. W. Feng, Q. F. Du, Y . Q. Zhong, J. Wang, J. Y . Chen, X. Yang, and Y . Zhou. Solar radio burst prediction based on a multimodal model.Solar Physics, 299(4):49, 2024. doi: 10.1007/s11207-024-02296-w. URLhttps://doi.org/10.1007/s11207-024-02296-w

    work page doi:10.1007/s11207-024-02296-w 2024

  22. [22]

    Gallagher, Jeremiah Scully, Shane A

    Herman le Roux, Ruhann Steyn, Du Toit Strauss, Mark Daly, Peter T. Gallagher, Jeremiah Scully, Shane A. Maloney, Christian Monstein, and Günther Drevin. Type ii and type iii solar radio burst classification using transfer learning.Solar Physics, 300(12):179, 2025. doi: 10.1007/s11207-025-02595-w. URL https://doi. org/10.1007/s11207-025-02595-w

    work page doi:10.1007/s11207-025-02595-w 2025

  23. [23]

    Elisa Tassan-Din, Akhil Gunessee, Philippe V ong, Christophe Marqué, Laurent Dolla, Antonio Martínez Picar, and Christian Monstein. Automated detection and classification of solar radio bursts in callisto spectrograms using deep-learning yolov5 model and ensemble methods.Journal of Space Weather and Space Climate, 16:16,

  24. [24]

    Real-time automated detection of multi- category solar radio bursts.Astrophysics and Space Science, 369:99, 2024

    Jingyu Deng, Guowu Yuan, Hao Zhou, Hao Wu, and Chengming Tan. Real-time automated detection of multi- category solar radio bursts.Astrophysics and Space Science, 369:99, 2024. doi: 10.1007/s10509-024-04364-w

    work page doi:10.1007/s10509-024-04364-w 2024

  25. [25]

    Multi-category solar radio burst detection based on task-aligned one-stage object detection model.Astrophysics and Space Sci- ence, 370:23, 2025

    Mingming Wang, Guowu Yuan, Hailan He, Chengming Tan, Hao Wu, and Hao Zhou. Multi-category solar radio burst detection based on task-aligned one-stage object detection model.Astrophysics and Space Sci- ence, 370:23, 2025. doi: 10.1007/s10509-025-04414-x. URL https://github.com/onewangqianqian/ MobileNetVitv2-TOOD

    work page doi:10.1007/s10509-025-04414-x 2025

  26. [26]

    Solar radio burst detection based on deformable detr.The Astrophysical Journal Supplement Series, 278(1):4, 2025

    Mingming Wang, Guowu Yuan, Hao Zhou, Chengming Tan, and Hao Wu. Solar radio burst detection based on deformable detr.The Astrophysical Journal Supplement Series, 278(1):4, 2025. doi: 10.3847/1538-4365/adbf18. URLhttps://github.com/onewangqianqian/SSA-Co-Deformable-DETR

    doi:10.3847/1538-4365/adbf18 2025

  27. [27]

    Scully, R

    J. Scully, R. Flynn, and P. T. Gallagher. Improved type iii solar radio burst detection using congruent deep learning models.Astronomy & Astrophysics, 674:A218, 2023. doi: 10.1051/0004-6361/202346404

    work page doi:10.1051/0004-6361/202346404 2023

  28. [28]

    Park, William Chan, Yu Zhang, Chung-Cheng Chiu, Barret Zoph, Ekin D

    Daniel S. Park, William Chan, Yu Zhang, Chung-Cheng Chiu, Barret Zoph, Ekin D. Cubuk, and Quoc V . Le. Specaugment: A simple data augmentation method for automatic speech recognition. InInterspeech 2019, interspeech_2019. ISCA, September 2019. doi: 10.21437/interspeech.2019-2680. URL http://dx.doi.org/ 10.21437/Interspeech.2019-2680

    doi:10.21437/interspeech.2019-2680 2019

  29. [29]

    i4ds/ecallisto_ng: Zenodo release, December 2024

    Vincenzo Timmel and Gabriel Torres Gamez. i4ds/ecallisto_ng: Zenodo release, December 2024. URL https: //doi.org/10.5281/zenodo.14505852

    work page doi:10.5281/zenodo.14505852 2024

  30. [30]

    Delving deep into rectifiers: Surpassing human-level performance on imagenet classification

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. InProc. IEEE Int. Conf. on Computer Vision (ICCV), pages 1026–1034,

  31. [31]

    Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017

    Pith/arXiv arXiv 2017

  32. [32]

    Kaichao You, Mingsheng Long, Jianmin Wang, and Michael I. Jordan. How does learning rate decay help modern neural networks?arXiv preprint arXiv:1908.01878, 2019

    Pith/arXiv arXiv 1908

  33. [33]

    Jasper Snoek, Hugo Larochelle, and Ryan P. Adams. Practical bayesian optimization of machine learning algorithms, 2012. URLhttps://arxiv.org/abs/1206.2944

    Pith/arXiv arXiv 2012

  34. [34]

    Experiment tracking with weights and biases, 2020

    Lukas Biewald. Experiment tracking with weights and biases, 2020. URL https://www.wandb.com/. Software available from wandb.com. 16 APREPRINT- JULY29, 2026

    2020

  35. [35]

    A unified approach to interpreting model predictions

    Scott M Lundberg and Su-In Lee. A unified approach to interpreting model predictions. In I. Guyon, U. V on Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, editors,Advances in Neural Information Processing Systems, volume 30. Curran Associates, Inc., 2017. URL https://proceedings. neurips.cc/paper_files/paper/2017/file/8a20a8621...

    2017

  36. [36]

    Captum: A unified and generic model interpretability library for pytorch.arXiv:2009.07896, 2020

    Narine Kokhlikyan, Vivek Miglani, Miguel Martin, Edward Wang, Bilal Alsallakh, Jonathan Reynolds, Alexander Melnikov, Natalia Kliushkina, Carlos Araya, Siqi Yan, et al. Captum: A unified and generic model interpretability library for pytorch.arXiv:2009.07896, 2020

    Pith/arXiv arXiv 2009

  37. [37]

    Mixmatch: A holistic approach to semi-supervised learning, 2019

    David Berthelot, Nicholas Carlini, Ian Goodfellow, Nicolas Papernot, Avital Oliver, and Colin Raffel. Mixmatch: A holistic approach to semi-supervised learning, 2019. URLhttps://arxiv.org/abs/1905.02249

    Pith/arXiv arXiv 2019

  38. [38]

    Cubuk, Alex Kurakin, Han Zhang, and Colin Raffel

    Kihyuk Sohn, David Berthelot, Chun-Liang Li, Zizhao Zhang, Nicholas Carlini, Ekin D. Cubuk, Alex Kurakin, Han Zhang, and Colin Raffel. Fixmatch: Simplifying semi-supervised learning with consistency and confidence,

  39. [39]

    Barlow twins: Self-supervised learning via redundancy reduction, 2021

    Jure Zbontar, Li Jing, Ishan Misra, Yann LeCun, and Stéphane Deny. Barlow twins: Self-supervised learning via redundancy reduction, 2021. URLhttps://arxiv.org/abs/2103.03230

    Pith/arXiv arXiv 2021

  40. [40]

    Self-supervised learning for solar radio spectrum classification.Universe, 8(12):656, 2022

    Siqi Li, Guowu Yuan, Jian Chen, Chengming Tan, and Hao Zhou. Self-supervised learning for solar radio spectrum classification.Universe, 8(12):656, 2022. doi: 10.3390/universe8120656. URL https://github. com/filterbank/spectrumcls

    work page doi:10.3390/universe8120656 2022

  41. [41]

    Detection of various solar radio bursts based on stable diffusion and self-supervised pretraining.Solar Physics, 300:171, 2025

    Xinlei Zhao, Guowu Yuan, Hao Zhou, Chengming Tan, and Liang Dong. Detection of various solar radio bursts based on stable diffusion and self-supervised pretraining.Solar Physics, 300:171, 2025. doi: 10.1007/ s11207-025-02590-1

    2025

  42. [42]

    M. P. van Haarlem, M. W. Wise, A. W. Gunst, G. Heald, J. P. McKean, J. W. T. Hessels, A. G. de Bruyn, R. Nijboer, J. Swinbank, R. Fallows, M. Brentjens, A. Nelles, R. Beck, H. Falcke, R. Fender, J. Hörandel, L. V . E. Koopmans, G. Mann, G. Miley, H. Röttgering, B. W. Stappers, R. A. M. J. Wijers, S. Zaroubi, M. van den Akker, A. Alexov, J. Anderson, K. An...

    2013

  43. [43]

    S. J. Tingay, R. Goeke, J. D. Bowman, D. Emrich, S. M. Ord, D. A. Mitchell, M. F. Morales, et al. The murchison widefield array: The square kilometre array precursor at low radio frequencies.Publications of the Astronomical Society of Australia, 30:e007, 2013. doi: 10.1017/pasa.2012.007. URL https://doi.org/10.1017/pasa. 2012.007

    doi:10.1017/pasa.2012.007 2013

  44. [44]

    L. A. Cañizares, S. T. Badman, S. A. Maloney, M. J. Owens, D. M. Weigt, E. P. Carley, and P. T. Gallagher. Tracking solar radio bursts using bayesian multilateration.A&A, 684:A182, 2024. doi: 10.1051/0004-6361/202347747. URLhttps://doi.org/10.1051/0004-6361/202347747

    work page doi:10.1051/0004-6361/202347747 2024

  45. [45]

    Solar type iii radio burst identification using few-shot object detection.Universe, 12(5):139, 2026

    Haoxiang Jiang, Shoulin Wei, Linjie Chen, Bo Liang, Wei Dai, Zhijian Zhang, and Heng Zhang. Solar type iii radio burst identification using few-shot object detection.Universe, 12(5):139, 2026. doi: 10.3390/universe12050139. 17 APREPRINT- JULY29, 2026

    work page doi:10.3390/universe12050139 2026

  46. [46]

    Weinberger

    Chuan Guo, Geoff Pleiss, Yu Sun, and Kilian Q. Weinberger. On calibration of modern neural networks, 2017. URLhttps://arxiv.org/abs/1706.04599

    Pith/arXiv arXiv 2017

  47. [47]

    non-burst

    Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jon Shlens, and Zbigniew Wojna. Rethinking the inception architecture for computer vision. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 2818–2826, 2016. A Stations and Instruments Used Table 4: Stations and corresponding instruments used to generate and evaluate ...

    2016

  48. [2009]

    doi: 10.1029/2008SW000425

    work page doi:10.1029/2008sw000425

  49. [2015]

    doi: 10.1109/ICCV .2015.123

    doi:10.1109/iccv 2015

  50. [2020]

    URLhttps://arxiv.org/abs/2001.07685

    Pith/arXiv arXiv 2001

  51. [2024]

    URLhttps://doi.org/10.1016/j.newast.2024.102194

    doi: 10.1016/j.newast.2024.102194. URLhttps://doi.org/10.1016/j.newast.2024.102194

    arXiv 2024

  52. [2026]

    doi: 10.1051/swsc/2026014

    arXiv