pith. machine review for the scientific record. sign in

arxiv: 2604.10016 · v1 · submitted 2026-04-11 · 🌌 astro-ph.SR · cs.LG

Recognition: unknown

Predicting Associations between Solar Flares and Coronal Mass Ejections Using SDO/HMI Magnetograms and a Hybrid Neural Network

Jialiang Li , Vasyl Yurchyshyn , Jason T. L. Wang , Haimin Wang , Manolis K. Georgoulis , Wen He , Yasser Abduallah , Hameedullah A. Farooki
show 1 more author
Yan Xu
Authors on Pith no claims yet

Pith reviewed 2026-05-10 16:30 UTC · model grok-4.3

classification 🌌 astro-ph.SR cs.LG
keywords solar flarescoronal mass ejectionshybrid neural networkline-of-sight magnetogramsactive regionspolarity inversion linesspace weatherSDO/HMI
0
0 comments X

The pith

Hybrid neural network using vision transformer and LSTM predicts whether solar flares will produce coronal mass ejections from active-region magnetogram sequences.

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

The paper introduces a hybrid neural network that ingests time series of line-of-sight magnetograms from solar active regions to forecast if an impending flare will launch a coronal mass ejection or remain confined. The model learns spatio-temporal patterns in the data collected by SDO/HMI and achieves good performance on this binary classification task. A secondary result is that the network's attention highlights magnetic flux cancellation along polarity inversion lines as a distinguishing feature, consistent with prior physical understanding. If the patterns are reliable, the approach could supply earlier and more specific alerts for space-weather hazards that affect satellites, power grids, and communications.

Core claim

The hybrid neural network (HNN) combines a vision transformer with long short-term memory layers to extract features from sequences of line-of-sight magnetograms of active regions; it then predicts whether a flare expected within the next 24 hours will be eruptive (CME-associated) or confined. Experiments demonstrate that the HNN performs well at this distinction, and inspection of its learned patterns indicates that magnetic flux cancellation in polarity inversion line regions is likely involved in triggering the eruptive cases.

What carries the argument

The hybrid neural network (HNN) that merges a vision transformer for spatial feature extraction with LSTM layers for temporal modeling, applied to sequences of line-of-sight magnetograms to classify future flares as eruptive or confined.

If this is right

  • The HNN method achieves good predictive performance for distinguishing eruptive from confined flares.
  • Magnetic flux cancellation along polarity inversion lines is identified as a candidate trigger for flare-associated CMEs.
  • The result is consistent with existing literature on the role of flux cancellation in solar eruptions.
  • Early classification of flare eruptiveness could support more targeted space-weather forecasting.

Where Pith is reading between the lines

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

  • If the learned patterns prove causal, routine monitoring of flux cancellation in polarity inversion lines could become a direct input to operational eruption forecasts.
  • The same magnetogram-sequence approach might be tested on other eruptive phenomena such as filament lift-off or coronal dimming.
  • Combining the HNN with simultaneous EUV or X-ray observations could test whether magnetogram-only inputs are sufficient or whether multi-wavelength context adds predictive power.

Load-bearing premise

Spatio-temporal patterns in line-of-sight magnetogram time series of active regions supply enough information to distinguish, 24 hours in advance, which flares will be accompanied by coronal mass ejections.

What would settle it

A held-out test set of active regions in which the model's accuracy on eruptive versus confined flares drops to random-chance levels, or in which measured flux cancellation rates show no statistical difference between the two classes.

Figures

Figures reproduced from arXiv: 2604.10016 by Haimin Wang, Hameedullah A. Farooki, Jason T. L. Wang, Jialiang Li, Manolis K. Georgoulis, Vasyl Yurchyshyn, Wen He, Yan Xu, Yasser Abduallah.

Figure 1
Figure 1. Figure 1: Construction of positive and negative magnetograms used in our prediction task. The magnetograms are collected at a frequency of 1 hour. Each rectangular box corresponds to 1 hour and contains one magnetogram. The red vertical line indicates the peak time of a γ-class flare, where γ implies ≥M5.0, ≥M, or ≥C. The yellow rectangular boxes shown in the left panel contain magnetograms that are within the 24 ho… view at source ↗
Figure 2
Figure 2. Figure 2: Overall architecture of our HNN model. Given is a magnetogram xt at time point t in an AR that will produce a γ-class flare within the next 24 hours of t. The HNN model accepts, as input, a time series or sequence of m magnetograms xt−m+1, xt−m+2, xt−m+3 . . . , xt−2, xt−1, xt where m is set to 10. The sequence is first processed image-by-image by a vision transformer (ViT) to extract spatial features in t… view at source ↗
Figure 3
Figure 3. Figure 3: Architectural design of the vision transformer (ViT). Each input magnetogram image is divided into non-overlapping patches, which are embedded via linear projection and positional encoding. These patch embeddings are fed into a stack of transformer encoder layers, each of which contains a multi-head self-attention (MHA) module, layer normalization (Norm), and a multilayer perceptron (MLP) module, to genera… view at source ↗
Figure 4
Figure 4. Figure 4: Performance comparison between ViT and HNN. 4.2. Performance Evaluation of the HNN Model [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Attention heat maps of the HNN model for two test magnetograms. The grayscale bar corresponds to the LOS magnetic field, in Gauss. The color bar shows the attention score. A larger attention score at a region indicates that more attention is paid to the region, where large attention scores are represented by dark red and small attention scores are represented by dark blue. (a) A positive prediction (true p… view at source ↗
Figure 6
Figure 6. Figure 6: Illustration of the changes of the unsigned magnetic flux in the PIL regions of the 10 magnetograms at time points t − 9, t − 8, . . . , t − 1, t used as the input of our HNN model in six ARs where t is as specified in each panel. Left panels: results for three positive predictions (true positives), where a positive prediction indicates that the corresponding AR will produce an eruptive flare within the ne… view at source ↗
Figure 7
Figure 7. Figure 7: Illustration of the changes of the total unsigned magnetic flux in the 10 magnetograms at time points t − 9, t − 8, . . . , t − 1, t used as the input of our HNN model in the six ARs shown in [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Bar graphs showing the comparison between ViT and HNN based on the five-fold cross-validation scheme. SDO/HMI data is provided by the Joint Science Operations Center (JSOC) Science Data Processing (SDP). DONKI is developed by the Community Coordinated Modeling Center (CCMC) at NASA. The proposed HNN model is implemented in PyTorch and Astropy packages. V.Y. acknowledges support from NSF AGS grants 2401229,… view at source ↗
read the original abstract

Solar eruptions, including flares and coronal mass ejections (CMEs), have a significant impact on Earth. Some flares are associated with CMEs, and some flares are not. The association between flares and CMEs is not always obvious. In this study, we propose a new deep learning method, specifically a hybrid neural network (HNN) that combines a vision transformer with long short-term memory, to predict associations between flares and CMEs. HNN finds spatio-temporal patterns in the time series of line-of-sight magnetograms of solar active regions (ARs) collected by the Helioseismic and Magnetic Imager on board the Solar Dynamics Observatory and uses the patterns to predict whether a flare projected to occur within the next 24 hours will be eruptive (i.e., CME-associated) or confined (i.e., not CME-associated). Our experimental results demonstrate the good performance of the HNN method. Furthermore, the results show that magnetic flux cancellation in polarity inversion line regions may well play a role in triggering flare-associated CMEs, a finding consistent with literature.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript proposes a hybrid neural network (HNN) combining a vision transformer and LSTM to analyze time series of SDO/HMI line-of-sight magnetograms from solar active regions. It aims to predict 24 hours in advance whether a flare will be eruptive (CME-associated) or confined, claiming good performance and interpreting model attention on polarity inversion line (PIL) regions as evidence that magnetic flux cancellation may trigger flare-associated CMEs, consistent with existing literature.

Significance. If the performance metrics and physical interpretation hold after proper validation, the work could support improved space-weather forecasting by identifying spatio-temporal precursors in magnetogram data. The hybrid architecture is a reasonable choice for capturing both spatial patterns and temporal evolution in active-region time series. However, the absence of quantitative results, baselines, and validation tests in the provided material limits assessment of its potential impact.

major comments (3)
  1. [Abstract] Abstract: The central claim of 'good performance' is unsupported by any reported quantitative metrics, baseline comparisons, cross-validation details, error estimates, or statistical significance tests. This is load-bearing for the primary contribution and prevents evaluation of whether the HNN outperforms existing methods or random baselines.
  2. [Abstract] Abstract and results discussion: The inference that magnetic flux cancellation in PIL regions plays a triggering role is presented as a post-hoc model insight from attention maps without controlled ablation, feature importance tests, or counterfactual experiments to distinguish causal drivers from dataset confounders (e.g., AR complexity or labeling criteria).
  3. [Methods] Methods and data sections: No explicit description of training/test split construction, handling of class imbalance, or out-of-distribution testing is provided. Given that the model is trained and evaluated on the same class of LOS magnetogram time series, this raises a risk of circularity or selection bias that could inflate apparent performance.
minor comments (2)
  1. [Abstract] Abstract: The 24-hour forecast horizon and active-region selection criteria should be stated more precisely to allow reproducibility.
  2. [Discussion] The manuscript would benefit from a dedicated limitations section addressing projection effects and missing transverse-field information inherent to LOS magnetograms.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for their constructive and detailed feedback. We have revised the manuscript to address the concerns about unsupported claims, interpretive overreach, and missing methodological details, while maintaining the core contributions of the hybrid neural network approach.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim of 'good performance' is unsupported by any reported quantitative metrics, baseline comparisons, cross-validation details, error estimates, or statistical significance tests. This is load-bearing for the primary contribution and prevents evaluation of whether the HNN outperforms existing methods or random baselines.

    Authors: We agree that the abstract was insufficiently specific and that quantitative support is essential. We have revised the abstract to explicitly report the performance metrics from our experiments (accuracy, F1-score, AUC), added baseline comparisons to standard CNN-LSTM and random models, and included cross-validation details with error estimates and statistical tests in the results section. These changes directly substantiate the performance claim. revision: yes

  2. Referee: [Abstract] Abstract and results discussion: The inference that magnetic flux cancellation in PIL regions plays a triggering role is presented as a post-hoc model insight from attention maps without controlled ablation, feature importance tests, or counterfactual experiments to distinguish causal drivers from dataset confounders (e.g., AR complexity or labeling criteria).

    Authors: We acknowledge the post-hoc and correlational nature of the attention-based interpretation. We have revised the abstract and discussion to present the finding as suggestive evidence consistent with existing literature on flux cancellation, rather than a definitive causal claim. We added explicit discussion of potential confounders such as active-region complexity and labeling criteria, along with a limitations paragraph noting the absence of ablation or counterfactual tests. revision: partial

  3. Referee: [Methods] Methods and data sections: No explicit description of training/test split construction, handling of class imbalance, or out-of-distribution testing is provided. Given that the model is trained and evaluated on the same class of LOS magnetogram time series, this raises a risk of circularity or selection bias that could inflate apparent performance.

    Authors: We agree that the original methods description lacked necessary detail on these points. We have substantially expanded the Methods and Data sections to describe the training/test split strategy (temporal and active-region partitioning to reduce leakage), class-imbalance handling via weighted loss, and out-of-distribution validation using data from separate time periods. These additions aim to demonstrate robustness against the noted risks. revision: yes

standing simulated objections not resolved
  • Performing new controlled ablation studies, feature importance tests, or counterfactual experiments to establish causality for flux cancellation would require additional experiments and resources beyond the scope of the current study.

Circularity Check

0 steps flagged

No circularity: standard supervised ML classification on held-out magnetogram sequences

full rationale

The paper trains a hybrid vision-transformer + LSTM network on time-series LOS magnetograms to classify whether an active-region flare will be eruptive or confined. Performance metrics are reported on (presumably) held-out test data, and the flux-cancellation interpretation is presented as a post-hoc observation consistent with existing literature rather than a derived first-principles result. No equation or claim reduces by construction to a fitted parameter renamed as a prediction, no self-citation is invoked as a uniqueness theorem, and the central result remains an empirical classifier output rather than a tautological re-expression of its inputs.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that magnetogram time series contain predictive signals for eruption potential and that the trained network generalizes beyond the training distribution. No new physical entities are introduced; the model parameters themselves constitute the fitted content.

free parameters (1)
  • Neural network weights and hyperparameters
    All model parameters are fitted to the magnetogram training data; no count or specific values are given in the abstract.
axioms (1)
  • domain assumption Line-of-sight magnetograms of active regions contain sufficient spatio-temporal information to distinguish eruptive from confined flares 24 hours ahead.
    This premise justifies the choice of input data and is required for the prediction task to be feasible.

pith-pipeline@v0.9.0 · 5541 in / 1546 out tokens · 43987 ms · 2026-05-10T16:30:16.608949+00:00 · methodology

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

69 extracted references · 61 canonical work pages · 1 internal anchor

  1. [1]

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

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

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION format.url url empty "" new.block "" url * "" * if FUNCTION format.eprint eprint empty "" archivePrefix empty "" archivePrefix "arXiv" = new.block " " eprint * " " * new.block " " eprint * " " * if if if FUNCTION format.doi doi empty "" " " doi * " " * if FUNCTION format.pid doi empty eprint empty ur...

  3. [3]

    -T 2@ | l wu 20 p 1s2@ Fg㸫 ے 2P @Uz 2g㸫8 = 3g㸫 S

    thebibliography [1] 20pt to REFERENCES 6pt =0pt \@twocolumntrue 12pt -12pt 10pt plus 3pt =0pt =0pt =1pt plus 1pt =0pt =0pt -12pt =13pt plus 1pt =20pt =13pt plus 1pt \@M =10000 =-1.0em =0pt =0pt 0pt =0pt =1.0em @enumiv\@empty 10000 10000 `\.\@m \@noitemerr \@latex@warning Empty `thebibliography' environment \@ifnextchar \@reference \@latexerr Missing key o...

    2017

  4. [4]

    Abduallah , Y., Wang , J. T. L., Wang , H., & Xu , Y. 2023, title Operational prediction of solar flares using a transformer-based framework , Scientific Reports, 13, 13665, 10.1038/s41598-023-40884-1

    work page doi:10.1038/s41598-023-40884-1 2023

  5. [5]

    N., Fan , Y., Kazachenko , M

    Afanasyev , A. N., Fan , Y., Kazachenko , M. D., & Cheung , M. C. M. 2023, title Hybrid Data-driven Magnetofrictional and Magnetohydrodynamic Simulations of an Eruptive Solar Active Region , , 952, 136, 10.3847/1538-4357/acd7e9

    work page doi:10.3847/1538-4357/acd7e9 2023

  6. [6]

    A., Abduallah , Y., Wang , J

    Alobaid , K. A., Abduallah , Y., Wang , J. T. L., et al. 2023, title Estimating Coronal Mass Ejection Mass and Kinetic Energy by Fusion of Multiple Deep-learning Models , , 958, L34, 10.3847/2041-8213/ad0c4a

    work page doi:10.3847/2041-8213/ad0c4a 2023

  7. [7]

    2016, Machine Learning: The New AI (MIT Press)

    Alpaydin, E. 2016, Machine Learning: The New AI (MIT Press)

    2016

  8. [8]

    Aslam , O. P. M., MacTaggart , D., Williams , T., Fletcher , L., & Romano , P. 2024, title Photospheric signatures of CME onset , , 534, 444, 10.1093/mnras/stae2110

    work page doi:10.1093/mnras/stae2110 2024

  9. [9]

    N., Daly , E., Daglis , I., Kappenman , J

    Baker , D. N., Daly , E., Daglis , I., Kappenman , J. G., & Panasyuk , M. 2004, title Effects of Space Weather on Technology Infrastructure , Space Weather, 2, S02004, 10.1029/2003SW000044

    work page doi:10.1029/2003sw000044 2004

  10. [10]

    K., & Veronig , A

    Baumgartner , C., Thalmann , J. K., & Veronig , A. M. 2018, title On the Factors Determining the Eruptive Character of Solar Flares , , 853, 105, 10.3847/1538-4357/aaa243

    work page doi:10.3847/1538-4357/aaa243 2018

  11. [11]

    M., Bein , B

    Berkebile-Stoiser , S., Veronig , A. M., Bein , B. M., & Temmer , M. 2012, title Relation between the Coronal Mass Ejection Acceleration and the Non-thermal Flare Characteristics , , 753, 88, 10.1088/0004-637X/753/1/88

    work page doi:10.1088/0004-637x/753/1/88 2012

  12. [12]

    S., Higgins , P

    Bloomfield , D. S., Higgins , P. A., McAteer , R. T. J., & Gallagher , P. T. 2012, title Toward Reliable Benchmarking of Solar Flare Forecasting Methods , , 747, L41, 10.1088/2041-8205/747/2/L41

    work page doi:10.1088/2041-8205/747/2/l41 2012

  13. [13]

    G., & Couvidat , S

    Bobra , M. G., & Couvidat , S. 2015, title Solar Flare Prediction Using SDO/HMI Vector Magnetic Field Data with a Machine-learning Algorithm , , 798, 135, 10.1088/0004-637X/798/2/135

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

  14. [14]

    G., & Ilonidis , S

    Bobra , M. G., & Ilonidis , S. 2016, title Predicting Coronal Mass Ejections Using Machine Learning Methods , , 821, 127, 10.3847/0004-637X/821/2/127

    work page doi:10.3847/0004-637x/821/2/127 2016

  15. [15]

    G., Sun , X., Hoeksema , J

    Bobra, M. G., Sun, X., Hoeksema, J. T., et al. 2014, title The Helioseismic and Magnetic Imager ( HMI ) Vector Magnetic Field Pipeline: SHARPs Space-Weather HMI Active Region Patches, , 289, 3549, 10.1007/s11207-014-0529-3

    work page doi:10.1007/s11207-014-0529-3 2014

  16. [16]

    Chen , P. F. 2011, title Coronal Mass Ejections: Models and Their Observational Basis , Living Reviews in Solar Physics, 8, 1, 10.12942/lrsp-2011-1

    work page doi:10.12942/lrsp-2011-1 2011

  17. [17]

    R., & Antiochos , S

    DeVore , C. R., & Antiochos , S. K. 2008, title Homologous Confined Filament Eruptions via Magnetic Breakout , , 680, 740, 10.1086/588011

    work page doi:10.1086/588011 2008

  18. [18]

    Djilali, Y. A. D., McGuinness, K., & O'Connor, N. E. 2023, in 34th British Machine Vision Conference, 771--774. http://proceedings.bmvc2023.org/771/

    2023

  19. [19]

    2021, in 9th International Conference on Learning Representations

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., et al. 2021, in 9th International Conference on Learning Representations. https://openreview.net/forum?id=YicbFdNTTy

    2021

  20. [20]

    2018, title Forecasting Solar Flares Using Magnetogram-based Predictors and Machine Learning , , 293, 28, 10.1007/s11207-018-1250-4

    Florios , K., Kontogiannis , I., Park , S.-H., et al. 2018, title Forecasting Solar Flares Using Magnetogram-based Predictors and Machine Learning , , 293, 28, 10.1007/s11207-018-1250-4

    work page doi:10.1007/s11207-018-1250-4 2018

  21. [21]

    K., Yardley, S

    Georgoulis, M. K., Yardley, S. L., Guerra, J. A., et al. 2024, title Prediction of solar energetic events impacting space weather conditions, AdSpR, 10.1016/j.asr.2024.02.030

    work page doi:10.1016/j.asr.2024.02.030 2024

  22. [22]

    Harrison , R. A. 1995, title The nature of solar flares associated with coronal mass ejection. , , 304, 585

    1995

  23. [23]

    Deep Residual Learning for Image Recognition , isbn =

    He, K., Zhang, X., Ren, S., & Sun, J. 2016, in IEEE Conference on Computer Vision and Pattern Recognition, 770--778, 10.1109/CVPR.2016.90

    work page doi:10.1109/cvpr.2016.90 2016

  24. [24]

    Neural Computation 9(8), 1735–1780 (1997)

    Hochreiter, S., & Schmidhuber, J. 1997, title Long Short-Term Memory, Neural Computation, 9, 1735, 10.1162/neco.1997.9.8.1735

    work page doi:10.1162/neco.1997.9.8.1735 1997

  25. [25]

    Hopfield , J. J. 1982, title Neural Networks and Physical Systems with Emergent Collective Computational Abilities , Proceedings of the National Academy of Sciences, 79, 2554, 10.1073/pnas.79.8.2554

    work page doi:10.1073/pnas.79.8.2554 1982

  26. [26]

    2018, The Astrophysical Journal, 856, 7, doi: 10.3847/1538-4357/aaae00

    Huang , X., Wang , H., Xu , L., et al. 2018, title Deep Learning Based Solar Flare Forecasting Model. I. Results for Line-of-sight Magnetograms , , 856, 7, 10.3847/1538-4357/aaae00

    work page doi:10.3847/1538-4357/aaae00 2018

  27. [27]

    H., Kongstad , P., et al

    Inceoglu , F., Jeppesen , J. H., Kongstad , P., et al. 2018, title Using Machine Learning Methods to Forecast if Solar Flares Will Be Associated with CMEs and SEPs , , 861, 128, 10.3847/1538-4357/aac81e

    work page doi:10.3847/1538-4357/aac81e 2018

  28. [28]

    2023, title A Systematic Magnetic Polarity Inversion Line Data Set from SDO/HMI Magnetograms , , 265, 28, 10.3847/1538-4365/acb43a

    Ji , A., Cai , X., Khasayeva , N., et al. 2023, title A Systematic Magnetic Polarity Inversion Line Data Set from SDO/HMI Magnetograms , , 265, 28, 10.3847/1538-4365/acb43a

    work page doi:10.3847/1538-4365/acb43a 2023

  29. [29]

    K., & Aydin , B

    Ji , A., Georgoulis , M. K., & Aydin , B. 2025, title Integration of solar flare and coronal mass ejection event data , Data in Brief, 60, 111539, 10.1016/j.dib.2025.111539

    work page doi:10.1016/j.dib.2025.111539 2025

  30. [30]

    T., & Stephenson, D

    Jolliffe, I. T., & Stephenson, D. B., eds. 2011, Forecast Verification: A Practitioner's Guide in Atmospheric Science (Wiley), 10.1002/9781119960003

    work page doi:10.1002/9781119960003 2011

  31. [31]

    2018, title Flare Prediction Using Photospheric and Coronal Image Data , , 293, 48, 10.1007/s11207-018-1258-9

    Jonas , E., Bobra , M., Shankar , V., Todd Hoeksema , J., & Recht , B. 2018, title Flare Prediction Using Photospheric and Coronal Image Data , , 293, 48, 10.1007/s11207-018-1258-9

    work page doi:10.1007/s11207-018-1258-9 2018

  32. [32]

    2018, The Astrophysical Journal, 869, 99, doi: 10.3847/1538-4357/aaebfc

    Kawabata , Y., Iida , Y., Doi , T., et al. 2018, title Statistical Relation between Solar Flares and Coronal Mass Ejections with Respect to Sigmoidal Structures in Active Regions , , 869, 99, 10.3847/1538-4357/aaebfc

    work page doi:10.3847/1538-4357/aaebfc 2018

  33. [33]

    Kazachenko , M. D. 2023, title A Database of Magnetic and Thermodynamic Properties of Confined and Eruptive Solar Flares , , 958, 104, 10.3847/1538-4357/ad004e

    work page doi:10.3847/1538-4357/ad004e 2023

  34. [34]

    Adam: A Method for Stochastic Optimization

    Kingma, D. P., & Ba, J. 2015, in 3rd International Conference on Learning Representations. http://arxiv.org/abs/1412.6980

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  35. [35]

    LeCun, Y

    LeCun, Y., Bengio, Y., & Hinton, G. 2015, title Deep learning, Nature, 521, 436, 10.1038/nature14539

    work page doi:10.1038/nature14539 2015

  36. [36]

    2020, title Predicting Solar Flares Using a Novel Deep Convolutional Neural Network , , 891, 10, 10.3847/1538-4357/ab6d04

    Li , X., Zheng , Y., Wang , X., & Wang , L. 2020, title Predicting Solar Flares Using a Novel Deep Convolutional Neural Network , , 891, 10, 10.3847/1538-4357/ab6d04

    work page doi:10.3847/1538-4357/ab6d04 2020

  37. [37]

    Lin , J., & Forbes , T. G. 2000, title Effects of reconnection on the coronal mass ejection process , , 105, 2375, 10.1029/1999JA900477

    work page doi:10.1029/1999ja900477 2000

  38. [38]

    Focal Loss for Dense Object Detection , booktitle =

    Lin, T., Goyal, P., Girshick, R. B., He, K., & Doll \' a r, P. 2017, in IEEE International Conference on Computer Vision, 2999--3007, 10.1109/ICCV.2017.324

    work page doi:10.1109/iccv.2017.324 2017

  39. [39]

    2019, title New data-driven method of simulating coronal mass ejections , , 626, A91, 10.1051/0004-6361/201935225

    Liu , C., Chen , T., & Zhao , X. 2019, title New data-driven method of simulating coronal mass ejections , , 626, A91, 10.1051/0004-6361/201935225

    work page doi:10.1051/0004-6361/201935225 2019

  40. [40]

    Liu , C., Deng , N., Wang , J. T. L., & Wang , H. 2017, title Predicting Solar Flares Using SDO/HMI Vector Magnetic Data Products and the Random Forest Algorithm , , 843, 104, 10.3847/1538-4357/aa789b

    work page doi:10.3847/1538-4357/aa789b 2017

  41. [41]

    Liu , H., Liu , C., Wang , J. T. L., & Wang , H. 2019, title Predicting Solar Flares Using a Long Short-term Memory Network , , 877, 121, 10.3847/1538-4357/ab1b3c

    work page doi:10.3847/1538-4357/ab1b3c 2019

  42. [42]

    Liu , H., Liu , C., Wang , J. T. L., & Wang , H. 2020, title Predicting Coronal Mass Ejections Using SDO/HMI Vector Magnetic Data Products and Recurrent Neural Networks , , 890, 12, 10.3847/1538-4357/ab6850

    work page doi:10.3847/1538-4357/ab6850 2020

  43. [43]

    2008, title Magnetic Field Overlying Solar Eruption Regions and Kink and Torus Instabilities , , 679, L151, 10.1086/589282

    Liu , Y. 2008, title Magnetic Field Overlying Solar Eruption Regions and Kink and Torus Instabilities , , 679, L151, 10.1086/589282

    work page doi:10.1086/589282 2008

  44. [44]

    T., Valori, G., et al

    Liu , Y., Welsch , B. T., Valori , G., et al. 2023, title Changes of Magnetic Energy and Helicity in Solar Active Regions from Major Flares , , 942, 27, 10.3847/1538-4357/aca3a6

    work page doi:10.3847/1538-4357/aca3a6 2023

  45. [45]

    Low , B. C. 1994, title Magnetohydrodynamic processes in the solar corona: Flares, coronal mass ejections, and magnetic helicity , Physics of Plasmas, 1, 1684, 10.1063/1.870671

    work page doi:10.1063/1.870671 1994

  46. [46]

    L., Sterling, A

    Moore , R. L., Sterling , A. C., Hudson , H. S., & Lemen , J. R. 2001, title Onset of the Magnetic Explosion in Solar Flares and Coronal Mass Ejections , , 552, 833, 10.1086/320559

    work page doi:10.1086/320559 2001

  47. [47]

    2018, The Astrophysical Journal, 858, 113, doi: 10.3847/1538-4357/aab9a7

    Nishizuka , N., Sugiura , K., Kubo , Y., Den , M., & Ishii , M. 2018, title Deep Flare Net (DeFN) Model for Solar Flare Prediction , , 858, 113, 10.3847/1538-4357/aab9a7

    work page doi:10.3847/1538-4357/aab9a7 2018

  48. [48]

    2025, title Magnetic flux cancellation in a flux-emergence magnetohydrodynamics simulation of coronal hole eruptions and jets , , 699, A87, 10.1051/0004-6361/202554580

    Patsourakos , S., & Archontis , V. 2025, title Magnetic flux cancellation in a flux-emergence magnetohydrodynamics simulation of coronal hole eruptions and jets , , 699, A87, 10.1051/0004-6361/202554580

    work page doi:10.1051/0004-6361/202554580 2025

  49. [49]

    2011, title Scikit-learn: machine learning in Python, J

    Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, title Scikit-learn: machine learning in Python, J. Mach. Learn. Res., 12, 2825. http://scikit-learn.sourceforge.net

    2011

  50. [50]

    D., Thompson , B

    Pesnell , W. D., Thompson , B. J., & Chamberlin , P. C. 2012, title The Solar Dynamics Observatory (SDO) , , 275, 3, 10.1007/s11207-011-9841-3

    work page doi:10.1007/s11207-011-9841-3 2012

  51. [51]

    2008, title Automated Prediction of CMEs Using Machine Learning of CME - Flare Associations , , 248, 471, 10.1007/s11207-007-9108-1

    Qahwaji , R., Colak , T., Al-Omari , M., & Ipson , S. 2008, title Automated Prediction of CMEs Using Machine Learning of CME - Flare Associations , , 248, 471, 10.1007/s11207-007-9108-1

    work page doi:10.1007/s11207-007-9108-1 2008

  52. [52]

    M., & Kumar , A

    Rust , D. M., & Kumar , A. 1996, title Evidence for Helically Kinked Magnetic Flux Ropes in Solar Eruptions , , 464, L199, 10.1086/310118

    work page doi:10.1086/310118 1996

  53. [53]

    H., Bush, R

    Schou , J., Scherrer , P. H., Bush , R. I., et al. 2012, title Design and Ground Calibration of the Helioseismic and Magnetic Imager (HMI) Instrument on the Solar Dynamics Observatory (SDO) , , 275, 229, 10.1007/s11207-011-9842-2

    work page doi:10.1007/s11207-011-9842-2 2012

  54. [54]

    C., Moore , R

    Sterling , A. C., Moore , R. L., & Panesar , N. K. 2018, title Magnetic Flux Cancelation as the Buildup and Trigger Mechanism for CME-producing Eruptions in Two Small Active Regions , , 864, 68, 10.3847/1538-4357/aad550

    work page doi:10.3847/1538-4357/aad550 2018

  55. [55]

    2022, title Solar Flare Forecast Using 3D Convolutional Neural Networks , , 941, 1, 10.3847/1538-4357/ac9e53

    Sun , P., Dai , W., Ding , W., et al. 2022, title Solar Flare Forecast Using 3D Convolutional Neural Networks , , 941, 1, 10.3847/1538-4357/ac9e53

    work page doi:10.3847/1538-4357/ac9e53 2022

  56. [56]

    Teraoka , K., Yamasaki , D., Kawabata , Y., Imada , S., & Shimizu , T. 2025, title Observational Comparison between Confined and Eruptive Flares: Magnetohydrodynamics Instability Parameters in a Similar Magnetic Configuration , , 983, 126, 10.3847/1538-4357/adc12d

    work page doi:10.3847/1538-4357/adc12d 2025

  57. [57]

    K., Su , Y., Temmer , M., & Veronig , A

    Thalmann , J. K., Su , Y., Temmer , M., & Veronig , A. M. 2015, title The Confined X-class Flares of Solar Active Region 2192 , , 801, L23, 10.1088/2041-8205/801/2/L23

    work page doi:10.1088/2041-8205/801/2/l23 2015

  58. [58]

    2005, title Confined and Ejective Eruptions of Kink-unstable Flux Ropes , , 630, L97, 10.1086/462412

    T \"o r \"o k , T., & Kliem , B. 2005, title Confined and Ejective Eruptions of Kink-unstable Flux Ropes , , 630, L97, 10.1086/462412

    work page doi:10.1086/462412 2005

  59. [59]

    2020, title Predicting Solar Flares with Machine Learning: Investigating Solar Cycle Dependence , , 895, 3, 10.3847/1538-4357/ab89ac

    Wang , X., Chen , Y., Toth , G., et al. 2020, title Predicting Solar Flares with Machine Learning: Investigating Solar Cycle Dependence , , 895, 3, 10.3847/1538-4357/ab89ac

    work page doi:10.3847/1538-4357/ab89ac 2020

  60. [60]

    2007, title A Comparative Study between Eruptive X-Class Flares Associated with Coronal Mass Ejections and Confined X-Class Flares , , 665, 1428, 10.1086/519765

    Wang , Y., & Zhang , J. 2007, title A Comparative Study between Eruptive X-Class Flares Associated with Coronal Mass Ejections and Confined X-Class Flares , , 665, 1428, 10.1086/519765

    work page doi:10.1086/519765 2007

  61. [61]

    F., & Howard, T

    Webb , D. F., & Howard , T. A. 2012, title Coronal Mass Ejections: Observations , Living Reviews in Solar Physics, 9, 3, 10.12942/lrsp-2012-3

    work page doi:10.12942/lrsp-2012-3 2012

  62. [62]

    M., & Angryk, R

    Wen , J., Ahmadzadeh , A., Georgoulis , M. K., Sadykov , V. M., & Angryk , R. A. 2025, title Outlier Detection and Removal in Multivariate Time Series for a More Robust Machine Learning based Solar Flare Prediction , , 277, 60, 10.3847/1538-4365/adb9e3

    work page doi:10.3847/1538-4365/adb9e3 2025

  63. [63]

    2025, title Solar Flare Forecasting Using Hybrid Neural Networks , , 276, 68, 10.3847/1538-4365/ada281

    Xu , D., Sun , P., Feng , S., Liang , B., & Dai , W. 2025, title Solar Flare Forecasting Using Hybrid Neural Networks , , 276, 68, 10.3847/1538-4365/ada281

    work page doi:10.3847/1538-4365/ada281 2025

  64. [64]

    2023, title Attribution rollout: a new way to interpret visual transformer, J

    Xu, L., Yan, X., Ding, W., & Liu, Z. 2023, title Attribution rollout: a new way to interpret visual transformer, J. Ambient Intell. Humaniz. Comput., 14, 163, 10.1007/S12652-022-04354-2

    work page doi:10.1007/s12652-022-04354-2 2023

  65. [65]

    Yashiro , S., Akiyama , S., Gopalswamy , N., & Howard , R. A. 2006, title Different Power-Law Indices in the Frequency Distributions of Flares with and without Coronal Mass Ejections , , 650, L143, 10.1086/508876

    work page doi:10.1086/508876 2006

  66. [66]

    2008, title Statistical relationship between solar flares and coronal mass ejections, Proceedings of the International Astronomical Union, 4, 233–243, 10.1017/S1743921309029342

    Yashiro, S., & Gopalswamy, N. 2008, title Statistical relationship between solar flares and coronal mass ejections, Proceedings of the International Astronomical Union, 4, 233–243, 10.1017/S1743921309029342

    work page doi:10.1017/s1743921309029342 2008

  67. [67]

    2022, title Solar Flare Index Prediction Using SDO/HMI Vector Magnetic Data Products with Statistical and Machine-learning Methods , , 263, 28, 10.3847/1538-4365/ac9b17

    Zhang , H., Li , Q., Yang , Y., et al. 2022, title Solar Flare Index Prediction Using SDO/HMI Vector Magnetic Data Products with Statistical and Machine-learning Methods , , 263, 28, 10.3847/1538-4365/ac9b17

    work page doi:10.3847/1538-4365/ac9b17 2022

  68. [68]

    Zhang , H., Jing , J., Wang , J. T. L., et al. 2025, title Prediction of Halo Coronal Mass Ejections Using SDO/HMI Vector Magnetic Data Products and a Transformer Model , , 981, 37, 10.3847/1538-4357/adafa0

    work page doi:10.3847/1538-4357/adafa0 2025

  69. [69]

    2021, title Hybrid deep convolutional neural network with one-versus-one approach for solar flare prediction , , 507, 3519, 10.1093/mnras/stab2132

    Zheng , Y., Li , X., Si , Y., Qin , W., & Tian , H. 2021, title Hybrid deep convolutional neural network with one-versus-one approach for solar flare prediction , , 507, 3519, 10.1093/mnras/stab2132

    work page doi:10.1093/mnras/stab2132 2021