Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Automatic detection of Ellerman bombs using Deep Learning

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read Trained neural networks segment Ellerman bombs from SST H-alpha spectra with high fidelity, while SDO/AIA four-passband models cannot reliably separate EBs from other bright patches.

desk verdict The SST detection result is solid and citable; the SDO/AIA negative conclusion overreaches because the labels are threshold-derived. read the letter →

arxiv 2505.03023 v1 pith:UI2VSTP3 submitted 2025-05-05 astro-ph.SR

I. J. Soler Poquet , C. J. Díaz Baso , L. H. M. Rouppe van der Voort , G. J. M. Vissers This is my paper · ORCID
classification astro-ph.SR
keywords EllermanbombssemanticsegmentationconvolutionalneuralnetworksdeeplearningH-alphalinewingsSDO/AIApassbandsfeatureimportancesmall-scalemagneticreconnection
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

This paper sets out to replace hand-tuned intensity thresholds with learned detectors that label every pixel as belonging to an Ellerman bomb or not, working both on high-resolution H-alpha spectra from the Swedish 1-m Solar Telescope and on SDO/AIA images. On SST data the networks are very effective, reaching a balanced precision-recall measure around 0.9, and the H-alpha line wings near ±1 Å carry most of the information, so spatial context is not needed when the spectrum is well resolved. When the spatial and spectral resolution is degraded, spatial context becomes important, and the convolutional model outperforms the pixel-wise model. On SDO/AIA, the same approach does not succeed: the best model reaches only about 0.77, barely better than a simple brightness threshold, and the four passbands together are not informative enough to tell EBs from network bright points and UV bursts. The paper concludes that adding the temporal evolution of bright patches is the most promising route toward a useful SDO-based detector, and a reader should care because a reliable automatic detector would turn a labor-intensive visual classification into a systematic survey of small-scale magnetic reconnection.

What carries the argument

The load-bearing mechanism is a semantic-segmentation pipeline built on two deliberately contrasted neural architectures: a fully convolutional network (a CNN that sees image patches and therefore spatial context) and a fully connected network (an FNN that classifies each pixel from its spectral vector alone). Training them on the same balanced patches, then applying a probabilistic calibration step and a threshold, lets the authors attribute any performance gap to the presence or absence of spatial context. The argument is carried further by spatial degradation experiments, which degrade SST data to SDO pixel size while preserving the masks, and by permutation feature importance, which shuffles each input channel and measures the drop in F1. Together these tools convert a practical detection task into a diagnostic about which information—spectral, spatial, or temporal—is genuinely needed.

What would settle it

Re-label the SDO test set without the UV-threshold step—for example, by directly thresholding co-aligned SST H-alpha wing brightness and then training the same four-passband CNN on those labels; if the resulting model exceeds an F1 of roughly 0.85 on held-out frames, the paper's claim that the four static passbands are not informative enough would be falsified. Conversely, if a classifier using only 1600 and 1700 Å light-curve statistics separates H-alpha-confirmed EBs from network bright points with high accuracy, the paper's proposed temporal route would receive direct support.

Watch

Extended reading notes

Core claim

The central discovery is a controlled demonstration of what information suffices to detect Ellerman bombs. In SST H-alpha observations at 0.057 arcsec per pixel, both a convolutional network and a per-pixel fully connected network classify EBs with nearly identical performance, showing that the spectral shape alone—especially the H-alpha wings around ±1 Å, identified by permutation feature importance—is enough when spectral resolution is high. Adding the center-to-limb parameter μ does not improve the results. After spatial degradation to SDO-like pixels, the convolutional model clearly beats the pixel-wise model, showing that spatial context substitutes for lost spectral detail. On SDO/AIA, the CNN reaches a maximum F1 of about 0.77 and the FNN about 0.66, and the 1600 Å channel is the dominant input, with 1700 Å, 171 Å, and 304 Å contributing less. A simple 2σ brightness threshold on 1700 Å achieves an F1 of 0.75, very close to the CNN, which the authors read as evidence that the four static intensity maps do not contain enough information to separate EBs from pseudo-EBs, network bright points, and UV bursts. Sample light curves of four bright patches show that H-alpha-confirmed EBs produce stronger short-timescale 1600 Å intensity variations than unrelated bright patches, pointing to temporal information as the key missing ingredient.

Load-bearing premise

The SDO conclusion depends on the assumption that the per-dataset UV brightness threshold, applied to the square root of the 1600×1700 Å product and overlaid on SST H-alpha masks, correctly labels SDO pixels as Ellerman bombs; if that label set is wrong, the models learn the wrong target and the claim that AIA passbands are insufficient is not established.

Editorial extensions

If this is right

  • SST H-alpha observations with adequate spectral sampling can be processed automatically to produce large, consistent EB catalogs without per-dataset threshold tuning, enabling statistical studies of reconnection occurrence and active-region evolution.
  • Since the H-alpha wings near ±1 Å dominate detection, future instruments can concentrate spectral sampling there without losing detection power.
  • At lower spatial resolution, detection depends on spatial context, so wide-field or satellite observations with degraded resolution should include contextual information rather than relying on single-pixel spectra.
  • SDO/AIA static four-passband maps are insufficient for reliable EB detection; any full-disk EB survey will need temporal information, such as short-timescale light-curve variability, or additional diagnostics.
  • The trained SDO CNN can serve as a first-stage candidate detector that flags bright patches whose 1600/1700 Å light curves can then be screened for EB-like variability, a pipeline the authors explicitly propose.

Reading between the lines

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

  • Because the SDO labels are derived from a UV-intensity threshold that is fine-tuned per dataset, the negative SDO result should be read as 'static AIA maps cannot reproduce these threshold-defined labels'; it does not rule out the possibility that a differently labeled training set, built for example from spectral classification of H-alpha profiles, would be learnable from the same passbands.
  • The light-curve analysis is based on only four bright patches; a natural extension would be to compute 1600/1700 Å rise rates or fluctuation amplitudes for all SDO CNN candidates and compare their distributions for H-alpha-confirmed EBs versus network bright points, testing the paper's temporal conjecture at scale.
  • The same CNN-versus-FNN comparison could be applied to other chromospheric diagnostics such as Ca II 8542 Å or H-beta, and to upcoming high-resolution facilities, where the question of whether spectra alone suffice is equally practical.
  • If the sub-0.1 arcsec² H-alpha-wing enhancements found by the models are genuine EBs, the event-size distribution extends below the previous 0.035 arcsec² cutoff, which would raise the estimated contribution of small reconnection events to the Sun's energy balance; this is a consequence the paper only hints at.
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

3 major / 5 minor

Summary. The paper presents a deep-learning pipeline for semantic segmentation of Ellerman bombs (EBs) in two data modalities: high-resolution SST Hα spectral scans and SDO/AIA four-passband images (1600, 1700, 171, 304 Å). For SST, the authors train FNN and CNN models on labels inherited from Vissers et al. (2019b) and report test-set F1 scores up to about 0.9, with the FNN matching the CNN at native resolution, and the CNN clearly outperforming the FNN when the spatial resolution is degraded. For SDO/AIA, the ground truth is constructed by applying a per-dataset fine-tuned intensity threshold to the square root of the 1600×1700 Å product and retaining contours that overlap the SST Hα EB masks; on this target, the CNN reaches F1≈0.77, only slightly above a 2σ intensity threshold on the 1700 Å channel (F1≈0.75). The paper concludes that the four AIA passbands are not informative enough to classify EBs and suggests that temporal information may help. The work also includes permutation feature importance, an application to a new 2024 SST dataset, and a calibration study.

Significance. If the SST result is taken at face value, it is a useful demonstration that a pixel-wise classifier can reproduce expert-label EB segmentation with high fidelity from high-resolution Hα spectra alone, and that spatial context becomes important only as spectral and spatial resolution degrade. The careful ablation design, the balancing patch-selection algorithm, and the evaluation on an unseen dataset are genuine strengths. However, the SST conclusion is tied to the Vissers et al. (2019b) threshold-based EB definition, and the SDO/AIA negative conclusion is weakened by the fact that its ground-truth labels are themselves generated by an intensity threshold on the same passbands used as input. The significance of the SDO claim is therefore conditional on that label definition; the paper partially acknowledges this in Sect. 4.3 but does not carry the caveat into the abstract and conclusions.

major comments (3)
  1. [Sect. 2.2.2 and Sect. 4.2/Abstract] The SDO ground-truth labels are constructed in Sect. 2.2.2 by applying a per-dataset fine-tuned intensity threshold to sqrt(1600 Å × 1700 Å) and keeping only the resulting bright contours that overlap SST Hα EB masks. This makes the positive class "UV brightening that overlaps an Hα-defined EB," not an independent definition of an EB. The models are then trained and evaluated on the same four AIA passbands to reproduce this target, so the abstract and Sect. 4.2 conclusion that the four AIA passbands are "not informative enough" to classify EBs is established only with respect to this threshold-generated label set. In particular, the near-tie between the CNN (F1=0.76) and the 1700 Å intensity threshold at 2σ (F1=0.75) in Sect. 3.4 and Fig. 11 is not an independent comparison: the labels were themselves produced by an intensity threshold on sqrt(1600×1700). The dependence is acknowledged in Sect. 4.3, but the abstract and Sect. 5 still state the insufficiency result without this caveat. The authors should rephrase the conclusion to state that the four passbands cannot reproduce the threshold-derived overlap labels, and ideally validate on manually labeled or independently defined SDO EB candidates.
  2. [Sect. 2.2.1 and Abstract] The SST ground-truth labels are inherited from Vissers et al. (2019b), which defined EBs via a brightness threshold on Hα wings plus area and lifetime criteria. Because the FNN, which has no spatial context, achieves nearly the same F1 as the CNN at native resolution, the conclusion that "spatial context is not required to detect EBs" may largely reflect the fact that the target labels are themselves constructed from per-pixel intensity thresholds on the Hα wings. This is not a flaw in the experimental design, but the abstract and Sect. 4.1 should qualify the claim as being relative to the Vissers et al. definition of EBs, not to a physically independent definition.
  3. [Sect. 3.4 and Fig. 11] The intensity-threshold benchmark is evaluated by scanning thresholds from 1σ to 9σ and reporting the best F1 (0.75 at 2σ) on the test set, while the NN performance is reported at its best probability threshold. Selecting the threshold on the test set gives the threshold method an optimistic bias in the comparison, and the near-tie with the CNN is therefore not on an equal footing. The authors should specify how the threshold was chosen (e.g., on the validation set) or account for the selection bias, because this comparison is used to support the conclusion that the AIA passbands are insufficient.
minor comments (5)
  1. [General] There are several typographical issues, e.g., "Futhermore" in Sect. 2.2.2, "di fferent" in multiple places, "classifiy" in Sect. 3.1, and inconsistent spacing in "V oort" in the author names and text.
  2. [Fig. 10] The tick labels in the left panel appear disordered and contain what looks like a duplicated "+1.2" offset; please check that the wavelength offsets are correctly mapped to the spectral sampling described in Sect. 2.2.1.
  3. [Sect. 2.2.2] The per-dataset intensity thresholds used to create the SDO masks are described as "fine-tuned" but the numerical values are not reported. Please provide them or specify the tuning criterion so that the ground-truth construction is reproducible.
  4. [Sect. 3.1 and Fig. 6] The text states that detections were selected with "a minimum linear extent of about 2′′," which seems inconsistent with the subarcsecond EB sizes quoted elsewhere and with the simultaneous 0.035 arcsec² area cut; please clarify this criterion.
  5. [Sect. 4.2] The interpretation that the SDO model "applies an intensity threshold over a non-linear combination of the passbands" is phrased as a speculation; it should be explicitly marked as an interpretation rather than a demonstrated property of the model.

Circularity Check

2 steps flagged · score 6.0 of 10

The SDO/AIA negative result is partially circular: its labels are generated by a threshold on the same 1600/1700 inputs used for training and benchmarking, so the 'not informative enough' conclusion is partly fixed by construction.

  1. self definitional [Section 2.2.2 (SDO datasets) and Section 4.2/5 (SDO insufficiency conclusion)]
    "To identify these UV brightenings, we applied an intensity threshold to the SDO datasets using the square root of the product of the 1600 Å and 1700 Å channels. ... The value for the intensity threshold is different for each dataset and has been fine-tuned to isolate the brighter regions ... Finally, we selected the UV brightenings contours that overlap with EB masks from the SST as final SDO mask."

    The SDO ground-truth label is defined as a thresholded function of the 1600/1700 input channels (combined with overlap against external SST Hα masks). Training and evaluating the SDO models against this target therefore measures how well the network can reproduce a threshold-plus-overlap rule, not whether the AIA intensity maps contain physical information about Ellerman bombs as independent objects. The abstract conclusion that 'the combination of the four different SDO/AIA passbands is not informative enough to accurately classify EBs' is consequently a statement about this self-generated target: the overlap component is not in the AIA inputs by construction, while the threshold component is merely a function of those same inputs.

  2. fitted input called prediction [Section 3.4 (Fig. 11) and Section 4.3]
    "For the SDO models, this comparison is possible because we have re-defined our own ground-truth data as explained in Sect. 2.2.2. ... We applied a minimum brightness threshold over the 1700 Å channel. ... The best F1 score is 0.75 for a threshold of 2σ over the quiet-Sun average intensity. ... For comparison, the FNN and the CNN achieved a maximum F1 0.66 and 0.76."

    The benchmark threshold on 1700 Å is the single-channel analogue of the label-generation threshold on sqrt(1600×1700), and the label-generation threshold itself was fine-tuned per dataset. Scoring a 1700 Å threshold against labels produced by a threshold on the same two UV channels is therefore not an independent baseline: the near-tie between the CNN (F1 0.76) and the threshold (F1 0.75) is partly a match between a label generator and a close variant of itself. The paper partially concedes this in Sect. 4.3, calling the threshold dependence 'a consequence of the method followed to build the ground-truth data for the SDO train, validation and test sets,' but the abstract and conclusions still state the insufficiency result without that caveat.

full rationale

The SST branch is not circular: the ground truth is explicitly taken from the published Vissers et al. (2019b) threshold criteria, and the paper openly states in Sect. 3.4 that comparing the SST models with an intensity threshold on the same test set would be 'redundant by definition.' That branch is a supervised-learning performance measurement against an external standard, even though one of the cited authors overlaps with the present paper. The SDO branch, however, builds its own labels by fine-tuning a threshold on sqrt(1600 Å × 1700 Å) and keeping contours that overlap SST EB masks (Sect. 2.2.2). The central SDO claim — that the four AIA passbands are not informative enough to classify EBs — is then evaluated against those threshold-generated labels, and the comparison threshold in Sect. 3.4 is a close variant of the same label generator. This makes the near-tie between CNN and threshold partly built into the label definition, and the 1600 Å feature-importance result partly an echo of the label construction. The paper acknowledges part of this dependence in Sect. 4.3, but the abstract and conclusions present the insufficiency result without that caveat. Overall, the circularity is partial and affects the SDO negative result, not the SST methodology or the new-dataset demonstration.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claims rest on the validity of the ground-truth labels (themselves derived from intensity thresholds), the accuracy of SST-SDO co-alignment, and standard machine learning assumptions about calibration and feature importance. The contribution is the application of standard deep learning tools to a new detection problem; no new physical entities are introduced.

free parameters (3)
  • SDO UV brightness threshold for ground-truth masks = Not reported; fine-tuned per dataset
    In Section 2.2.2, the SDO labels are created by thresholding sqrt(1600 Å x 1700 Å) and selecting contours that overlap SST EB masks; the threshold is adjusted for each dataset to isolate brighter regions. This threshold directly defines the target labels for training and evaluation.
  • Decision probability threshold = 0.5
    Section 3.1: probability of 0.5 is chosen as the best trade-off between precision and recall; it is used for area histograms and feature importance analysis.
  • Neural network hyperparameters (learning rate, batch size, architecture sizes) = Learning rates between 1e-2 and 1e-3; batch sizes 126 (SST) and 500 (SDO)
    Section 2.5: learning rates and batch sizes were selected as the ones presenting the best performance; different architectures were tested and the final ones are reported. These choices affect all reported metrics.
assumptions (4)
  • domain assumption Ground-truth EBs in SST data from Vissers et al. (2019b) are correct by construction.
    Section 2.2.1: labels are based on an intensity threshold of 145% for the EB core and 140% for the halo, plus size and lifetime criteria. The neural network is trained to reproduce these labels, so its performance is bounded by that definition.
  • domain assumption The co-alignment and downsampling between SST and SDO preserve EB correspondences.
    Section 2.2.2: SST masks are transferred to SDO via coalignment of the Hα blue wing and the SDO/AIA 1700 Å passband. Any misalignment or resampling error propagates into the SDO labels and the evaluation of SDO models.
  • domain assumption Data reductions (MOMFBD/SSTRED for SST, AIAPY for SDO) produce intensity maps in which EB signatures are comparable.
    Sections 2.2.1 and 2.2.2: SST data are normalized to the average quiet-Sun intensity, and SDO data are normalized by exposure and corrected for time-dependent degradation. The models rely on these calibration steps for consistent features.
  • standard math Permutation feature importance is a valid indicator of model reliance.
    Section 2.10: the method follows Fisher et al. (2018) and is a standard machine learning technique; it assumes that shuffling a feature breaks its relationship to the target without introducing confounding effects.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Automatic detection of Ellerman bombs using Deep Learning." pith.science (2026). https://pith.science/paper/UI2VSTP3

@misc{pith2026250503023,
  author       = {Pith},
  title        = {Pith review of: Automatic detection of Ellerman bombs using Deep Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UI2VSTP3}},
  note         = {Machine review of arXiv:2505.03023}
}
abstract

Ellerman bombs (EBs) are observable signatures of photospheric small-scale magnetic reconnection events. The reliable automatic detection of EBs would enable the study of the impact of magnetic reconnection on the Sun's dynamics. We aim to develop a method to automatically detect EBs in H$\alpha$ observations from the Swedish 1-m Solar Telescope (SST) and in SDO/AIA observations using the 1600\r{A}, 1700\r{A}, 171\r{A} and 304\r{A} passbands. We trained models based on neural networks (NNs) to perform automatic detection of EBs. Additionally, we used different types of NNs to study how different properties contribute to the detection of EBs. We find that for SST observations, the NN-based models are proficient at detecting EBs. With sufficiently high spectral resolution, the spatial context is not required to detect EBs. However, as we degrade the spectral and spatial resolution, the spatial information becomes more important. Models that include both dimensions perform best. For SDO/AIA, the models struggle to reliably distinguish between EBs and bright patches of different origin. Permutation feature importance revealed that the H$\alpha$ line wings (around $\pm$1 \r{A} from line center) are the most informative features for EB detection. For the SDO/AIA case, the 1600\r{A} channel is the most relevant one when used in combination with 171\r{A} and 304\r{A}. The combination of the four different SDO/AIA passbands is not informative enough to accurately classify EBs. From our analysis of a few sample SDO/AIA 1600\r{A} and 1700\r{A} light curves, we conclude that inclusion of the temporal variation may be a significant step towards establishing an effective EB detection method that can be applied to the extensive SDO/AIA database of observations. Abstract modified for ArXiv purposes.

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. Automatic detection of Ellerman bombs in the H$\alpha$ line

    astro-ph.SR 2025-08 unverdicted novelty 5.0 of 10

    An automated star-finding-based pipeline detects and tracks 2,257 Ellerman bombs across ten solar datasets, reporting an average area of 0.44 arcsec^2, peak contrast of 1.4 times the quiet Sun, and a median lifetime o...

Reference graph

Works this paper leans on

74 extracted references · 19 canonical work pages · cited by 1 Pith paper

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archiveprefix author booktitle chapter edition editor howpublished institution eprint doi url journal key month note number organization pages publisher school series title type volume year adsurl label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'be...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in " " * FUNCTION format....

    work page

  3. [3]

    Armstrong , J. A. & Fletcher , L. 2019, http://dx.doi.org/10.1007/s11207-019-1473-z blue , 294, 80 https://ui.adsabs.harvard.edu/abs/2019SoPh..294...80A

    doi:10.1007/s11207-019-1473-z 2019

  4. [4]

    Asensio Ramos , A., Cheung , M. C. M., Chifu , I., & Gafeira , R. 2023, http://dx.doi.org/10.1007/s41116-023-00038-x blue Living Reviews in Solar Physics , 20, 4 https://ui.adsabs.harvard.edu/abs/2023LRSP...20....4A

    doi:10.1007/s41116-023-00038-x 2023

  5. [5]

    T., Cheung, M

    Barnes, W. T., Cheung, M. C. M., Bobra, M. G., et al. 2020, http://dx.doi.org/10.21105/joss.02801 blue Journal of Open Source Software , 5, 5

    doi:10.21105/joss.02801 2020

  6. [6]

    Transition region response to Quiet Sun Ellerman Bombs

    Bhatnagar , A., Rouppe van der Voort , L., & Joshi , J. 2024, https://ui.adsabs.harvard.edu/abs/2024arXiv240609585B http://dx.doi.org/10.48550/arXiv.2406.09585 blue arXiv e-prints , arXiv:2406.09585

    work page Pith review arXiv doi:10.48550/arxiv.2406.09585 2024

  7. [7]

    Bobra , M. G. & Couvidat , S. 2015, http://dx.doi.org/10.1088/0004-637X/798/2/135 blue , 798, 135 https://ui.adsabs.harvard.edu/abs/2015ApJ...798..135B

    doi:10.1088/0004-637x/798/2/135 2015

  8. [8]

    2020, in Computational Science -- ICCS 2020, ed

    Brabec, J., Kom \'a rek, T., Franc, V., & Machlica, L. 2020, in Computational Science -- ICCS 2020, ed. V. V. Krzhizhanovskaya, G. Z \'a vodszky, M. H. Lees, J. J. Dongarra, P. M. A. Sloot, S. Brissos, & J. Teixeira (Cham: Springer International Publishing), 74--87 https://link.springer.com/chapter/10.1007/978-3-030-50423-6_6#citeas

    doi:10.1007/978-3-030-50423-6_6 2020

Show all 74 references
  1. [9]

    G., Asensio Ramos , A., & Felipe , T

    Broock , E. G., Asensio Ramos , A., & Felipe , T. 2022, http://dx.doi.org/10.1051/0004-6361/202244206 blue , 667, A132 https://ui.adsabs.harvard.edu/abs/2022A&A...667A.132B

    2022 doi

  2. [10]

    2015, https://ui.adsabs.harvard.edu/abs/2015arXiv151107289C http://dx.doi.org/10.48550/arXiv.1511.07289 blue arXiv e-prints , arXiv:1511.07289

    Clevert , D.-A., Unterthiner , T., & Hochreiter , S. 2015, https://ui.adsabs.harvard.edu/abs/2015arXiv151107289C http://dx.doi.org/10.48550/arXiv.1511.07289 blue arXiv e-prints , arXiv:1511.07289

    2015 doi arXiv

  3. [11]

    2023, https://ui.adsabs.harvard.edu/abs/2023arXiv230206378C http://dx.doi.org/10.48550/arXiv.2302.06378 blue arXiv e-prints , arXiv:2302.06378

    Csurka , G., Volpi , R., & Chidlovskii , B. 2023, https://ui.adsabs.harvard.edu/abs/2023arXiv230206378C http://dx.doi.org/10.48550/arXiv.2302.06378 blue arXiv e-prints , arXiv:2302.06378

    2023 doi arXiv

  4. [12]

    o fdahl , M. G., S \

    de la Cruz Rodr \' guez , J., L \"o fdahl , M. G., S \"u tterlin , P., Hillberg , T., & Rouppe van der Voort , L. 2015, http://dx.doi.org/10.1051/0004-6361/201424319 blue , 573, A40 https://ui.adsabs.harvard.edu/abs/2015A&A...573A..40D

    2015 doi

  5. [13]

    R., & Cheung , C

    De Pontieu , B., Lemen , J. R., & Cheung , C. M. M. 2019, in AGU Fall Meeting Abstracts, Vol. 2019, SH33A--07 https://ui.adsabs.harvard.edu/abs/2019AGUFMSH33A..07D

    2019

  6. [14]

    J., Asensio Ramos , A., & de la Cruz Rodr \' guez , J

    D \' az Baso , C. J., Asensio Ramos , A., & de la Cruz Rodr \' guez , J. 2022, http://dx.doi.org/10.1051/0004-6361/202142018 blue , 659, A165 https://ui.adsabs.harvard.edu/abs/2022A&A...659A.165D

    2022 doi

  7. [15]

    J., Asensio Ramos , A., de la Cruz Rodr \' guez , J., da Silva Santos , J

    D \' az Baso , C. J., Asensio Ramos , A., de la Cruz Rodr \' guez , J., da Silva Santos , J. M., & Rouppe van der Voort , L. 2025, http://dx.doi.org/10.1051/0004-6361/202452172 blue , 693, A170 https://ui.adsabs.harvard.edu/abs/2025A&A...693A.170D

    2025 doi

  8. [16]

    J., de la Cruz Rodr \' guez , J., & Danilovic , S

    D \' az Baso , C. J., de la Cruz Rodr \' guez , J., & Danilovic , S. 2019, http://dx.doi.org/10.1051/0004-6361/201936069 blue , 629, A99 https://ui.adsabs.harvard.edu/abs/2019A&A...629A..99D

    2019 doi

  9. [17]

    J., de la Cruz Rodr \' guez , J., & Leenaarts , J

    D \' az Baso , C. J., de la Cruz Rodr \' guez , J., & Leenaarts , J. 2021, http://dx.doi.org/10.1051/0004-6361/202040111 blue , 647, A188 https://ui.adsabs.harvard.edu/abs/2021A&A...647A.188D

    2021 doi

  10. [18]

    J., Rouppe van der Voort , L., de la Cruz Rodr \' guez , J., & Leenaarts , J

    D \' az Baso , C. J., Rouppe van der Voort , L., de la Cruz Rodr \' guez , J., & Leenaarts , J. 2023, http://dx.doi.org/10.1051/0004-6361/202346230 blue , 673, A35 https://ui.adsabs.harvard.edu/abs/2023A&A...673A..35D

    2023 doi

  11. [19]

    M., Asensio Ramos , A., Fischer , C

    D \' az Castillo , S. M., Asensio Ramos , A., Fischer , C. E., & Berdyugina , S. V. 2022, http://dx.doi.org/10.3389/fspas.2022.896632 blue Frontiers in Astronomy and Space Sciences , 9, 896632 https://ui.adsabs.harvard.edu/abs/2022FrASS...9.6632D

    2022

  12. [20]

    1917, http://dx.doi.org/10.1086/142366 blue , 46, 46

    Ellerman, F. 1917, http://dx.doi.org/10.1086/142366 blue , 46, 46

    1917 doi

  13. [21]

    2006, http://dx.doi.org/https://doi.org/10.1016/j.patrec.2005.10.010 blue Pattern Recognition Letters , 27, 27

    Fawcett, T. 2006, http://dx.doi.org/https://doi.org/10.1016/j.patrec.2005.10.010 blue Pattern Recognition Letters , 27, 27

    2006 doi

  14. [22]

    2018, https://ui.adsabs.harvard.edu/abs/2018arXiv180101489F http://dx.doi.org/10.48550/arXiv.1801.01489 blue arXiv e-prints , arXiv:1801.01489

    Fisher , A., Rudin , C., & Dominici , F. 2018, https://ui.adsabs.harvard.edu/abs/2018arXiv180101489F http://dx.doi.org/10.48550/arXiv.1801.01489 blue arXiv e-prints , arXiv:1801.01489

    2018 doi arXiv

  15. [23]

    K., Rust , D

    Georgoulis , M. K., Rust , D. M., Bernasconi , P. N., & Schmieder , B. 2002, http://dx.doi.org/10.1086/341195 blue , 575, 506 https://ui.adsabs.harvard.edu/abs/2002ApJ...575..506G

    2002 doi

  16. [24]

    2016, Deep Learning (MIT Press), http://www.deeplearningbook.org

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

    2016

  17. [25]

    Guo, C., Pleiss, G., Sun, Y., & Weinberger, K. Q. 2017, CoRR, abs/1706.04599, abs/1706.04599

    2017 arXiv

  18. [26]

    2022, http://dx.doi.org/10.1007/s11207-022-02019-z blue , 297, 104 https://ui.adsabs.harvard.edu/abs/2022SoPh..297..104G

    Guo , X., Yang , Y., Feng , S., et al. 2022, http://dx.doi.org/10.1007/s11207-022-02019-z blue , 297, 104 https://ui.adsabs.harvard.edu/abs/2022SoPh..297..104G

    2022 doi

  19. [27]

    2019, http://dx.doi.org/10.1051/0004-6361/201935376 blue , 626, A33 https://ui.adsabs.harvard.edu/abs/2019A&A...626A..33H

    Hansteen , V., Ortiz , A., Archontis , V., et al. 2019, http://dx.doi.org/10.1051/0004-6361/201935376 blue , 626, A33 https://ui.adsabs.harvard.edu/abs/2019A&A...626A..33H

    2019 doi

  20. [28]

    & Berlicki , A

    Herlender , M. & Berlicki , A. 2011, Central European Astrophysical Bulletin, 35, 181 https://ui.adsabs.harvard.edu/abs/2011CEAB...35..181H

    2011

  21. [29]

    D., & Hao , Q

    Hong , J., Li , Y., Ding , M. D., & Hao , Q. 2021, http://dx.doi.org/10.3847/1538-4357/ac1ba0 blue , 921, 50 https://ui.adsabs.harvard.edu/abs/2021ApJ...921...50H

    2021 doi

  22. [30]

    Illarionov , E. A. & Tlatov , A. G. 2018, http://dx.doi.org/10.1093/mnras/sty2628 blue , 481, 5014 https://ui.adsabs.harvard.edu/abs/2018MNRAS.481.5014I

    2018 doi

  23. [31]

    M., Hofmeister , S., et al

    Jarolim , R., Veronig , A. M., Hofmeister , S., et al. 2021, http://dx.doi.org/10.1051/0004-6361/202140640 blue , 652, A13 https://ui.adsabs.harvard.edu/abs/2021A&A...652A..13J

    2021 doi

  24. [32]

    & Rouppe van der Voort , L

    Joshi , J. & Rouppe van der Voort , L. H. M. 2022, http://dx.doi.org/10.1051/0004-6361/202243051 blue , 664, A72 https://ui.adsabs.harvard.edu/abs/2022A&A...664A..72J

    2022 doi

  25. [33]

    Joshi , J., Rouppe van der Voort , L. H. M., & de la Cruz Rodr \' guez , J. 2020, http://dx.doi.org/10.1051/0004-6361/202038769 blue , 641, L5 https://ui.adsabs.harvard.edu/abs/2020A&A...641L...5J

    2020 doi

  26. [34]

    Kingma, D. P. & Ba, J. 2017, Adam: A Method for Stochastic Optimization

    2017

  27. [35]

    & Panos, B

    Kleint, L. & Panos, B. 2022, Astronomy & Astrophysics, 657, 657

    2022

  28. [36]

    Krikova , K., Pereira , T. M. D., & Rouppe van der Voort , L. H. M. 2023, http://dx.doi.org/10.1051/0004-6361/202346796 blue , 677, A52 https://ui.adsabs.harvard.edu/abs/2023A&A...677A..52K

    2023 doi

  29. [37]

    Kruskal, J. B. 1964, Psychometrika, 29, 29

    1964

  30. [38]

    R., Title , A

    Lemen , J. R., Title , A. M., Akin , D. J., et al. 2012, http://dx.doi.org/10.1007/s11207-011-9776-8 blue , 275, 17 https://ui.adsabs.harvard.edu/abs/2012SoPh..275...17L

    2012 doi

  31. [39]

    1982, http://dx.doi.org/10.1109/TIT.1982.1056489 blue IEEE Transactions on Information Theory , 28, 28

    Lloyd, S. 1982, http://dx.doi.org/10.1109/TIT.1982.1056489 blue IEEE Transactions on Information Theory , 28, 28

    1982

  32. [40]

    G., Hillberg , T., de la Cruz Rodr \' guez , J., et al

    L \"o fdahl , M. G., Hillberg , T., de la Cruz Rodr \' guez , J., et al. 2021, http://dx.doi.org/10.1051/0004-6361/202141326 blue , 653, A68 https://ui.adsabs.harvard.edu/abs/2021A&A...653A..68L

    2021 doi

  33. [41]

    2008, http://dx.doi.org/10.1093/pasj/60.3.577 blue Publications of the Astronomical Society of Japan , 60, 60

    Matsumoto, T., Kitai, R., Shibata, K., et al. 2008, http://dx.doi.org/10.1093/pasj/60.3.577 blue Publications of the Astronomical Society of Japan , 60, 60

    2008 doi

  34. [42]

    M \"u ller , D., St. Cyr , O. C., Zouganelis , I., et al. 2020, http://dx.doi.org/10.1051/0004-6361/202038467 blue , 642, A1 https://ui.adsabs.harvard.edu/abs/2020A&A...642A...1M

    2020 doi

  35. [43]

    & Hinton, G

    Nair, V. & Hinton, G. E. 2010, in Proceedings of the 27th International Conference on International Conference on Machine Learning, ICML'10 (Madison, WI, USA: Omnipress), 807–814 https://dl.acm.org/doi/10.5555/3104322.3104425

    2010

  36. [44]

    J., Doyle , J

    Nelson , C. J., Doyle , J. G., Erd \'e lyi , R., et al. 2013, http://dx.doi.org/10.1007/s11207-012-0222-3 blue , 283, 307 https://ui.adsabs.harvard.edu/abs/2013SoPh..283..307N

    2013 doi

  37. [45]

    2007, http://dx.doi.org/10.1051/0004-6361:20067011 blue , 473, 279 https://ui.adsabs.harvard.edu/abs/2007A&A...473..279P

    Pariat , E., Schmieder , B., Berlicki , A., et al. 2007, http://dx.doi.org/10.1051/0004-6361:20067011 blue , 473, 279 https://ui.adsabs.harvard.edu/abs/2007A&A...473..279P

    2007 doi

  38. [46]

    2019, in Advances in Neural Information Processing Systems 32 (Curran Associates, Inc.), 8024--8035

    Paszke, A., Gross, S., Massa, F., et al. 2019, in Advances in Neural Information Processing Systems 32 (Curran Associates, Inc.), 8024--8035

    2019

  39. [47]

    2011, Journal of machine learning research, 12, 12

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of machine learning research, 12, 12

    2011

  40. [48]

    1990, http://dx.doi.org/10.1029/GM030p0063 blue Mitteilungen der Astronomischen Gesellschaft Hamburg , 65, 65

    Priest, E. 1990, http://dx.doi.org/10.1029/GM030p0063 blue Mitteilungen der Astronomischen Gesellschaft Hamburg , 65, 65

    1990 doi

  41. [49]

    Prince, S. J. 2023, Understanding Deep Learning (The MIT Press)

    2023

  42. [50]

    D., Wang , H., Denker , C., & Goode , P

    Qiu , J., Ding , M. D., Wang , H., Denker , C., & Goode , P. R. 2000, http://dx.doi.org/10.1086/317310 blue , 544, L157 https://ui.adsabs.harvard.edu/abs/2000ApJ...544L.157Q

    2000 doi

  43. [51]

    R., et al

    Quintero Noda , C., Schlichenmaier , R., Bellot Rubio , L. R., et al. 2022, http://dx.doi.org/10.1051/0004-6361/202243867 blue , 666, A21 https://ui.adsabs.harvard.edu/abs/2022A&A...666A..21Q

    2022 doi

  44. [52]

    G., et al

    Reid , A., Mathioudakis , M., Doyle , J. G., et al. 2016, http://dx.doi.org/10.3847/0004-637X/823/2/110 blue , 823, 110 https://ui.adsabs.harvard.edu/abs/2016ApJ...823..110R

    2016 doi

  45. [53]

    R., Warner , M., Keil , S

    Rimmele , T. R., Warner , M., Keil , S. L., et al. 2020, http://dx.doi.org/10.1007/s11207-020-01736-7 blue , 295, 172 https://ui.adsabs.harvard.edu/abs/2020SoPh..295..172R

    2020 doi

  46. [54]

    2015, https://ui.adsabs.harvard.edu/abs/2015arXiv150504597R http://dx.doi.org/10.48550/arXiv.1505.04597 blue arXiv e-prints , arXiv:1505.04597

    Ronneberger , O., Fischer , P., & Brox , T. 2015, https://ui.adsabs.harvard.edu/abs/2015arXiv150504597R http://dx.doi.org/10.48550/arXiv.1505.04597 blue arXiv e-prints , arXiv:1505.04597

    2015 doi arXiv

  47. [55]

    Rouppe van der Voort , L. H. M., Joshi , J., Henriques , V. M. J., & Bose , S. 2021, http://dx.doi.org/10.1051/0004-6361/202040171 blue , 648, A54 https://ui.adsabs.harvard.edu/abs/2021A&A...648A..54R

    2021 doi

  48. [56]

    Rouppe van der Voort , L. H. M., Joshi , J., & Krikova , K. 2024, http://dx.doi.org/10.1051/0004-6361/202348976 blue , 683, A190 https://ui.adsabs.harvard.edu/abs/2024A&A...683A.190R

    2024 doi

  49. [57]

    Rouppe van der Voort , L. H. M., van Noort , M., & de la Cruz Rodr \' guez , J. 2023, http://dx.doi.org/10.1051/0004-6361/202345933 blue , 673, A11 https://ui.adsabs.harvard.edu/abs/2023A&A...673A..11R

    2023 doi

  50. [58]

    J., Vissers , G

    Rutten , R. J., Vissers , G. J. M., Rouppe van der Voort , L. H. M., S \"u tterlin , P., & Vitas , N. 2013, in Journal of Physics Conference Series, Vol. 440, Journal of Physics Conference Series, 012007 https://ui.adsabs.harvard.edu/abs/2013JPhCS.440a2007R

    2013

  51. [59]

    B., Bjelksjo , K., Korhonen , T

    Scharmer , G. B., Bjelksjo , K., Korhonen , T. K., Lindberg , B., & Petterson , B. 2003, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 4853, Innovative Telescopes and Instrumentation for Solar Astrophysics, ed. S. L. Keil & S. V. Avakyan ...

    2003

  52. [60]

    B., Narayan , G., Hillberg , T., et al

    Scharmer , G. B., Narayan , G., Hillberg , T., et al. 2008, http://dx.doi.org/10.1086/595744 blue , 689, L69 https://ui.adsabs.harvard.edu/abs/2008ApJ...689L..69S

    2008 doi

  53. [61]

    2017, http://dx.doi.org/10.1109/TPAMI.2016.2572683 blue IEEE Transactions on Pattern Analysis and Machine Intelligence , 39, 39

    Shelhamer, E., Long, J., & Darrell, T. 2017, http://dx.doi.org/10.1109/TPAMI.2016.2572683 blue IEEE Transactions on Pattern Analysis and Machine Intelligence , 39, 39

    2017

  54. [62]

    2015, https://ui.adsabs.harvard.edu/abs/2015arXiv150604214S http://dx.doi.org/10.48550/arXiv.1506.04214 blue arXiv e-prints , arXiv:1506.04214

    Shi , X., Chen , Z., Wang , H., et al. 2015, https://ui.adsabs.harvard.edu/abs/2015arXiv150604214S http://dx.doi.org/10.48550/arXiv.1506.04214 blue arXiv e-prints , arXiv:1506.04214

    2015 doi arXiv

  55. [63]

    Sim \ o es , P. J. A., Reid , H. A. S., Milligan , R. O., & Fletcher , L. 2019, http://dx.doi.org/10.3847/1538-4357/aaf28d blue , 870, 114 https://ui.adsabs.harvard.edu/abs/2019ApJ...870..114S

    2019 doi

  56. [64]

    2005, http://dx.doi.org/10.1007/s11207-006-0049-x blue Solar Physics , 235, 235

    Socas-Navarro, H., Martínez-Pillet, V., Elmore, D., et al. 2005, http://dx.doi.org/10.1007/s11207-006-0049-x blue Solar Physics , 235, 235

    2005 doi

  57. [65]

    I., Pevtsov , A

    T \"a htinen , I., Virtanen , I. I., Pevtsov , A. A., & Mursula , K. 2022, http://dx.doi.org/10.1051/0004-6361/202141164 blue , 664, A2 https://ui.adsabs.harvard.edu/abs/2022A&A...664A...2T

    2022 doi

  58. [66]

    2022, http://dx.doi.org/10.1109/ACCESS.2022.3216565 blue IEEE Access , 10, 10

    Tortora, M., Cordelli, E., & Soda, P. 2022, http://dx.doi.org/10.1109/ACCESS.2022.3216565 blue IEEE Access , 10, 10

    2022

  59. [67]

    Van Noort , M., Rouppe Van Der Voort , L., & L \"o fdahl , M. G. 2005, http://dx.doi.org/10.1007/s11207-005-5782-z blue , 228, 191 https://ui.adsabs.harvard.edu/abs/2005SoPh..228..191V

    2005 doi

  60. [68]

    Vissers , G. J. M., de la Cruz Rodr \' guez , J., Libbrecht , T., et al. 2019 a , http://dx.doi.org/10.1051/0004-6361/201833560 blue , 627, A101 https://ui.adsabs.harvard.edu/abs/2019A&A...627A.101V

    2019 doi

  61. [69]

    Vissers , G. J. M., Rouppe van der Voort , L. H. M., & Rutten , R. J. 2013, http://dx.doi.org/10.1088/0004-637X/774/1/32 blue , 774, 32 https://ui.adsabs.harvard.edu/abs/2013ApJ...774...32V

    2013 doi

  62. [70]

    Vissers , G. J. M., Rouppe van der Voort , L. H. M., & Rutten , R. J. 2019 b , http://dx.doi.org/10.1051/0004-6361/201834811 blue , 626, A4 https://ui.adsabs.harvard.edu/abs/2019A&A...626A...4V

    2019 doi

  63. [71]

    Vissers , G. J. M., Rouppe van der Voort , L. H. M., Rutten , R. J., Carlsson , M., & De Pontieu , B. 2015, http://dx.doi.org/10.1088/0004-637X/812/1/11 blue , 812, 11 https://ui.adsabs.harvard.edu/abs/2015ApJ...812...11V

    2015 doi

  64. [72]

    Watanabe , H., Vissers , G., Kitai , R., Rouppe van der Voort , L., & Rutten , R. J. 2011, http://dx.doi.org/10.1088/0004-637X/736/1/71 blue , 736, 71 https://ui.adsabs.harvard.edu/abs/2011ApJ...736...71W

    2011 doi

  65. [73]

    R., Tian , H., Peter , H., et al

    Young , P. R., Tian , H., Peter , H., et al. 2018, http://dx.doi.org/10.1007/s11214-018-0551-0 blue , 214, 120 https://ui.adsabs.harvard.edu/abs/2018SSRv..214..120Y

    2018 doi

  66. [74]

    2019, http://dx.doi.org/10.1007/s11207-019-1517-4 blue , 294, 117 https://ui.adsabs.harvard.edu/abs/2019SoPh..294..117Z

    Zhu , G., Lin , G., Wang , D., Liu , S., & Yang , X. 2019, http://dx.doi.org/10.1007/s11207-019-1517-4 blue , 294, 117 https://ui.adsabs.harvard.edu/abs/2019SoPh..294..117Z

    2019 doi

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.