Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Exoplanet Transit Candidate Identification in TESS Full-Frame Images via a Transformer-Based Algorithm

T0 review · 4 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read A transformer network finds 214 new exoplanet transit candidates in TESS light curves without phase-folding or assuming a period.

desk verdict A genuine periodicity-free search method and a useful candidate list, but the 214-candidate claim is under-validated and the single-transit count has an internal inconsistency; worth a serious referee with major revision. read the letter →

arxiv 2502.07542 v2 pith:RJBGPETX submitted 2025-02-11 astro-ph.EP astro-ph.GAastro-ph.IMcs.AI

Helem Salinas , Rafael Brahm , Greg Olmschenk , Richard K. Barry , Karim Pichara , Stela Ishitani Silva , Vladimir Araujo This is my paper · ORCID
classification astro-ph.EPastro-ph.GAastro-ph.IMcs.AI
keywords exoplanettransitsTESSfull-frameimagestransformernetworkself-attentionsingle-transitcandidateslight-curveclassificationtransittimingvariationsdeeplearning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper sets out to find exoplanet transits in TESS full-frame image light curves without first assuming a period or phase-folding the data, so that single-transit events and systems with transit-timing variations are not missed. It reports that a convolutional-embedding plus transformer-encoder network, trained on confirmed planets, eclipsing binaries, and non-transit light curves, flags 214 new planetary-system candidates across TESS sectors 1–26 with radii above 0.27 Jupiter radii. If the candidate list holds up, the method would add a periodicity-free search channel to exoplanet discovery and provide a sample of warm-Jupiter-like single-transit candidates for follow-up.

What carries the argument

The key object is a transformer encoder with multi-head self-attention preceded by two 1-D convolutional embedding layers. The convolutions turn local windows of the flux, centroid, and background time series into tokens, and the self-attention lets the model weigh any two time steps against each other, so it can recognize a transit dip embedded in stellar variability without needing repeated transits. The encoder's output is average-pooled and passed through a multilayer perceptron to give a single probability that the light curve contains an exoplanet transit.

What would settle it

Inject synthetic transits of known depth and duration into genuine SPOC light curves from sectors 1–26, run the trained network on them, and measure the recovery rate; if the model misses injected transits that are visible by eye, or flags a comparable fraction of pure non-transit light curves, then the 214 candidates cannot be trusted as transits. A cheaper check is the false-positive rate on the non-transit and eclipsing-binary test sets, which the paper does not report for the blind search.

Watch

Extended reading notes

Core claim

The central claim is that the transformer's multi-head self-attention, applied directly to full light curves alongside centroid and background time series, learns the dip shape of a planetary transit well enough to distinguish it from eclipsing binaries, stellar variability, and instrumental systematics, without any transit parameters. The network identifies 122 multi-transit light curves, 88 single-transit events, and 4 multi-planet systems among TESS sectors 1–26, and the paper argues that the detection does not rely on periodicity, extending automated searches to regimes that phase-folding pipelines systematically under-sample.

Load-bearing premise

The load-bearing premise is that training on confirmed planets plus injected transit signals teaches the network a general transit shape, so that when it flags an unseen light curve as planet-bearing, the curve really contains a transit rather than stellar variability or instrumental noise.

Editorial extensions

If this is right

  • Transit searches can be run directly on full light curves, removing the need for de-trending and phase-folding that can erase or distort signals.
  • Single-transit events—candidates with orbital periods longer than a single TESS sector—become retrievable, yielding 88 new single-transit candidates.
  • Systems with transit-timing variations, where transits are not strictly periodic, are accessible to automated detection.
  • Adding background and centroid time series as inputs raises the AUC from 0.85 to 0.88, cutting false positives from blended or background sources.
  • The current model detects planets only above about 0.27 Jupiter radii, so Earth-sized and super-Earth candidates remain out of reach of this approach.

Reading between the lines

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

  • The 88 single-transit candidates form a ready list for radial-velocity and photometric follow-up; if even a fraction are confirmed, they would populate the warm Jupiter gap that short-period surveys under-sample.
  • The same architecture should transfer to future surveys like PLATO or Roman, where single-transit and TTV cases will be proportionally more common; a direct test would be fine-tuning on simulated Roman light curves.
  • A fallback that combines this detector with a shallow periodic search (e.g., BLS) might recover smaller planets while keeping the single-transit channel, since the paper itself finds its method misses shallow transits.
  • The reported 0.1% positive rate in the blind search is only an upper bound on true candidacy until an injection-recovery experiment measures the completeness and false-positive rate on the actual survey light curves.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes a Transformer-based binary classifier that takes TESS SPOC FFI light curves as input, using concatenated flux, centroid, and background time series, and predicts whether a light curve contains a planetary transit without requiring phase folding or prior transit parameters. The model is trained on confirmed ExoFOP planets, eclipsing binaries, non-transit signals, and injected transit signals, and is evaluated with 10-fold stratified cross-validation using a target-aware split, reporting an average AUC of 0.88 and F1 of 0.82. The trained model is then applied to 4.1 million SPOC light curves from TESS sectors 1-26. After BLS, DAVE, human vetting, and a Gaia radial-velocity error filter, the authors report 214 new planetary system candidates: 122 single-planet multi-transit light curves, 88 single-transit candidates, and 4 multi-planet systems.

Significance. If the central claim is supported, the paper would demonstrate a useful periodicity-free transit search method that can find single-transit and TTV-bearing systems that phase-folding pipelines may miss. The paper has several concrete strengths: it uses target-aware train/test splitting, includes auxiliary centroid and background time series, reports an explicit false-positive example, and makes the code publicly available. The candidate list, with transit parameters and radii, could provide useful follow-up targets. However, the validation of the blind survey search is incomplete: the paper reports no injection-recovery experiments in the actual survey light curves, no false-positive rate for the end-to-end pipeline, and the single-transit vetting rests on a Gaia RV-error cutoff that does not discriminate true long-period planets from blended eclipsing binaries or instrumental artifacts. The significance of the 214-candidate claim is therefore currently limited by missing quantitative validation rather than by the core model design.

major comments (4)
  1. [Section 6.2, Table 4, Abstract, Section 7] The number of single-transit candidates is internally inconsistent. Section 6.2 states that 'leaving 82 validated as single transiters' and Table 4 lists 82 entries, but the Abstract and Section 7 report 88 single-transit candidates and a total of 214 candidates. This discrepancy affects the headline result and must be resolved; the final counts must be reproducible from the tables.
  2. [Sections 6.1-6.2] The paper does not provide a measured false-positive rate or completeness for the blind survey search. The curated test-set AUC/F1 (Section 5.2, Table 2) measures performance on labeled confirmed planets, EBs, and non-transits, but the 4.1-million-light-curve search is unlabeled and the funnel from model positives to final candidates is only given as '~20%' after BLS and '~25%' after DAVE, with no cumulative counts and no injection-recovery experiment into actual survey light curves. Without an end-to-end false-positive rate, the claim that the 214 candidates are 'new planetary system candidates' is not quantitatively supported; a large unknown fraction could be false positives.
  3. [Section 6.2] The single-transit vetting uses a Gaia RV-error cutoff of 2 km/s and visual inspection, but this does not discriminate a true long-period transiting planet from a blended eclipsing binary or an instrumental dip. The paper itself shows in Section 6.5 that a false positive (TIC 296945443) passed the network with high probability because the aperture did not capture the centroid shift. The single-transit candidates are not subjected to the same DAVE centroid and secondary-eclipse checks as the multi-transit candidates, so their status as planetary candidates is weakly supported. The manuscript should either apply quantitative false-positive diagnostics to these candidates or explicitly label them as transit-like single events requiring confirmation.
  4. [Section 6.5] The false-positive example TIC 296945443 demonstrates an aperture-dependent failure mode in which the network assigns high probability despite background-star contamination. Since this failure mode is identified only through external examination in a later sector, and since the final candidate list is not screened in this way for all entries, the rate at which such false positives contaminate the 214-candidate list is unknown. Reporting the number of model positives that were rejected at each vetting stage, and ideally a small cross-match against known false-positive catalogs, would help bound this contamination.
minor comments (5)
  1. [Eq. (5), Section 4.3.1] The positional encoding formula uses sin for both even and odd indices; the standard Vaswani et al. (2017) formulation uses sine for even indices and cosine for odd indices. If the implementation follows the standard formulation, the formula should be corrected; otherwise the deviation should be explained.
  2. [Section 6.1] The text contains an unresolved 'Figure??' reference in the paragraph discussing radius versus orbital period. This must be fixed before publication.
  3. [Tables 3 and 4] The column labeled 'Transit Epoch' lists values such as 1386.206, but the unit (presumably BTJD) and the precision convention are not stated. Please define the epoch unit and explain why four significant figures are used.
  4. [Table 4] The table header contains a typo: 'lighr curve' should be 'light curve'.
  5. [Section 5.2] The sentence 'our model identified that approximately 0.1% of the light curves contain planetary transit signals' should be phrased as 'predicted to contain', since these are model-positive predictions, not confirmed transit detections.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: trained labels are external and the candidate search is a genuine model application; remaining issues are validation gaps, not circular reasoning.

full rationale

The derivation chain is not circular. The network is trained on externally labelled ground truth (ExoFOP confirmed planets, EBs from Prša et al. 2022, ExoFOP, and Yu et al. 2019, plus non-transit light curves), with injected transit signals used as data augmentation. This is standard supervised learning: the labels come from independent catalogues and follow-up, not from the model's own candidate list. The blind search is a genuine application of the trained network to 4.1 million SPOC light curves from sectors 1-26, and the 214 candidates are those surviving BLS, DAVE, Gaia RV-error filtering, and visual vetting; no fitted parameter is renamed as a prediction. The radius estimate r_p = r_* sqrt(d(c+1)) uses BLS transit depth plus TIC/Gaia stellar radii and contamination values, all external to the training labels. Self-citations to Salinas et al. (2023) and Olmschenk et al. (2021) describe the Transformer architecture and the injection technique, but they are not load-bearing: the architecture and training are specified in Sections 4.3-4.4, and the central claim rests on the reported test metrics rather than on those citations. Consequently, the paper's weaknesses are validation risks rather than circularity: the blind search has no measured false-positive rate or injection-recovery experiment in the actual survey light curves, the single-transit count is internally inconsistent (82 in Section 6.2 versus 88 in the abstract), and Section 6.5 openly documents a false positive that passed the network. These issues affect the strength of the 214-candidate claim, but they do not make the derivation equivalent to its inputs.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical entities. The central claim rests on the quality of external catalogs (ExoFOP, TESS-EBs, Yu et al. 2019), on standard transit-parameter assumptions, and on a set of hand-chosen vetting thresholds. The free parameters are all vetting cutoffs, not fitted constants in a derivation.

free parameters (3)
  • candidate radius threshold (1.8 RJup) = 1.8 RJup
    Hand-chosen cutoff in Section 6.1 to discard objects likely to be low-mass stars or brown dwarfs; directly affects the final candidate count.
  • radial velocity error cutoff (2 km/s) = 2 km/s
    Hand-chosen filter in Section 6.2 for single-transit candidates; directly affects the 88 single-transit sample.
  • detection probability threshold = not stated (default 0.5)
    The model output threshold used to decide which light curves contain transits is not specified anywhere in the paper; it determines the 0.1% detection rate and the input to the vetting pipeline.
assumptions (5)
  • domain assumption Ground truth labels from ExoFOP are accurate and complete enough for training positive examples.
    These labels are the only positive signal used in supervised training (Section 4.2). If confirmed planets in ExoFOP are mislabeled or if the sample is not representative, the model will learn an incorrect notion of 'transit'.
  • standard math The transit-depth to radius conversion r_p = r_* sqrt(d(c+1)) is valid for these candidates.
    Used in Section 6.1 to compute all candidate radii. It assumes no limb darkening, no significant impact parameter, a circular orbit, and accurate contamination c from the TIC. None of these are verified for individual candidates.
  • domain assumption SPOC PDCSAP flux, centroid, and background time series are sufficient to distinguish true transits from false positives.
    The model's discrimination relies on these three input channels (Section 3.1). If the centroid or background data contain systematics that mimic transit-like signals, the model can be fooled, as the paper itself notes for TIC 296945443.
  • domain assumption Injected light curves preserve the physical transit shape and noise properties of real transits.
    Positive training examples include transits artificially injected into non-transit light curves (Section 4.2). If injection creates artifacts or atypical noise correlations, the model may learn to detect injection artifacts rather than real transits.
  • domain assumption The BLS periodogram and DAVE vetting correctly recover transit parameters and false-positive flags for these FFI light curves.
    The candidate list is filtered through these tools (Section 6.1). No independent validation of BLS and DAVE behavior on SPOC FFI light curves is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Exoplanet Transit Candidate Identification in TESS Full-Frame Images via a Transformer-Based Algorithm." pith.science (2026). https://pith.science/paper/RJBGPETX

@misc{pith2026250207542,
  author       = {Pith},
  title        = {Pith review of: Exoplanet Transit Candidate Identification in TESS Full-Frame Images via a Transformer-Based Algorithm},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RJBGPETX}},
  note         = {Machine review of arXiv:2502.07542}
}
abstract

The Transiting Exoplanet Survey Satellite (TESS) is surveying a large fraction of the sky, generating a vast database of photometric time series data that requires thorough analysis to identify exoplanetary transit signals. Automated learning approaches have been successfully applied to identify transit signals. However, most existing methods focus on the classification and validation of candidates, while few efforts have explored new techniques for the search of candidates. To search for new exoplanet transit candidates, we propose an approach to identify exoplanet transit signals without the need for phase folding or assuming periodicity in the transit signals, such as those observed in multi-transit light curves. To achieve this, we implement a new neural network inspired by Transformers to directly process Full Frame Image (FFI) light curves to detect exoplanet transits. Transformers, originally developed for natural language processing, have recently demonstrated significant success in capturing long-range dependencies compared to previous approaches focused on sequential data. This ability allows us to employ multi-head self-attention to identify exoplanet transit signals directly from the complete light curves, combined with background and centroid time series, without requiring prior transit parameters. The network is trained to learn characteristics of the transit signal, like the dip shape, which helps distinguish planetary transits from other variability sources. Our model successfully identified 214 new planetary system candidates, including 122 multi-transit light curves, 88 single-transit and 4 multi-planet systems from TESS sectors 1-26 with a radius > 0.27 $R_{\mathrm{Jupiter}}$, demonstrating its ability to detect transits regardless of their periodicity.

Figures

Figures reproduced from arXiv: 2502.07542 by the authors.

Figure 1
Figure 1. (a) the transformer encoder, which processes time series inputs using positional encodings combined with input embeddings, and computes feature representations through a self-attention mechanism. (b) explains the self-attention mechanism, where the 𝑄, 𝐾 matrices are used to calculate attention scores. These scores are then applied to the 𝑉 to produce the self-attention feature map. mission. BTJD is defined as BTJD =… view at source ↗
Figure 2
Figure 2. Schematic of the proposed architecture. The input includes flux 𝑥𝑖 , centroid 𝑐𝑖 , and background 𝑏𝑖 time series, which are concatenated into input embeddings 𝑥 ′ 𝑖 which are processed using convolutional embeddings. The tokens, along with positional encodings, are passed through a MSA mechanism within a transformer encoder. The features embedding produced by the transformer encoder are then passed through to averag… view at source ↗
Figure 3
Figure 3. CNN embedding, where the kernels slide across the input time series, transforming each local window into an embedding vector. Xemb = CNN2 (CNN1 (𝑋)) (3) where, Xemb ∈ R 𝑇×𝑑 is the embedded representation, with 𝑇 being the number of tokens generated after the convolution process, 𝑋 represents the input features (flux, centroid, background) and 𝑑 is the dimension of the output embedding space. After the layer where we… view at source ↗
Figures from the paper (9 more)
Figure 4
Figure 4. Figure 4: General diagram for the identification of new exoplanet candidates based on model predictions. The blue boxes represent the inputs to the NN and the outputs after inference (candidates and transit vetting). The grey boxes are related to the NN, which includes the creat…
Figure 5
Figure 5. Figure 5: Distribution of single-planet multi-transit light curve candidate radii measured in Jupiter raidii. Moreover, these examples underscore the challenges in analyzing single transit events. Our model demonstrates the ability to detect the presence of a transit signal, but…
Figure 7
Figure 7. Figure 7: Relationship between TESS magnitude and transit depth. Each point represents a confirmed planet from ExoFOP-TESS or our candidates. observed due to long orbital periods. However, through our analysis, we found that our model did not identify any Earth-sized or super￾Ea…
Figure 8
Figure 8. Figure 8: Relationship between radius and log orbital period. Each point represents a confirmed planet from ExoFOP-TESS or our candidates. that upcoming missions, such as PLATO (Rauer et al. 2016) and the Nancy Grace Roman Space Telescope Roman, will present new challenges in th…
Figure 10
Figure 10. Figure 10: Light curves for TIC 232616346 in sectors 20 and 23, showing the same transit event with consistent depth, validating the detection of the same candidate in both sectors. Goodfellow I., Bengio Y., Courville A., 2016, Deep learning. MIT press Grieves N., et al., 2022, …
Figure 11
Figure 11. Figure 11: Three examples of single transiters from TESS observations for TIC 122522333, TIC 341687821 and TIC 233577004. Each candidate shows a single-transit event in the respective sector. While these candidates were also detected in other sectors (TIC 122522333 in sector 4, …
Figure 12
Figure 12. Figure 12: Lightcurves for TIC 221567884 across sectors 12, 39, and 66, illustrating transit events “b”, “c” and “d”, and a potential mutual transit event between candidates “c” and “d” in sector 66. A = Attention(Q, K, V) = softmax QK⊤ √︁ dk ! V (A2) The attention layer takes t…
Figure 13
Figure 13. Figure 13: Light curves for TIC 118798035 illustrating a system with two exoplanet candidates. In sector 13, transit signals for candidates “b” (orange line) and “c” (blue line) are detected. The depths are consistent across both sectors, with TTVs for candidate “c” in sector 39…
Figure 14
Figure 14. Figure 14: Light curve for TIC 193096383 in sector 9, showing the transits of candidates b (orange) and c (blue). The transits of candidate b have the same depth, which differs from the depth observed in the transit of candidate c, suggesting a potential multi-planet system. MNR…

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. HD 148797: A bright F-type star with two moderate-period low-density sub-Jovian planets. Compact multi-planet architectures are common in the Neptunian savanna

    astro-ph.EP 2026-07 conditional novelty 6.0 of 10

    HD 148797 hosts two ~40 ME, ~8.3 RE, low-density savanna planets near a 1.619 period ratio whose anti-correlated TTVs yield masses and support compact multi-planet architectures as typical in the savanna.

Reference graph

Works this paper leans on

71 extracted references · 44 canonical work pages · cited by 1 Pith paper

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTION or pop #1...

  2. [2]

    Agol E., Fabrycky D., 2017, arXiv preprint arXiv:1706.09849

    work page Pith review arXiv 2017

  3. [3]

    D., 2021, arXiv preprint arXiv:2105.06178

    Allam Jr T., McEwen J. D., 2021, arXiv preprint arXiv:2105.06178

    arXiv 2021

  4. [4]

    E., Mnih V., Leibo J

    Ba J., Hinton G. E., Mnih V., Leibo J. Z., Ionescu C., 2016, Advances in neural information processing systems, 29

    work page 2016

  5. [5]

    Bahdanau D., Cho K., Bengio Y., 2014, arXiv preprint arXiv:1409.0473

    arXiv 2014

  6. [6]

    P., et al., 2024, @doi [ ] 10.1051/0004-6361/202449307 , https://ui.adsabs.harvard.edu/abs/2024A&A...686A.230B 686, A230

    Battley M. P., et al., 2024, @doi [ ] 10.1051/0004-6361/202449307 , https://ui.adsabs.harvard.edu/abs/2024A&A...686A.230B 686, A230

    doi:10.1051/0004-6361/202449307 2024

  7. [7]

    J., et al., 2010, Science, 327, 977

    Borucki W. J., et al., 2010, Science, 327, 977

    2010

  8. [8]

    Brahm R., et al., 2023, @doi [ ] 10.3847/1538-3881/accadd , https://ui.adsabs.harvard.edu/abs/2023AJ....165..227B 165, 227

    doi:10.3847/1538-3881/accadd 2023

Show all 71 references
  1. [9]

    T., et al., 2020, Kepler Science Document KSCI-19081-003, p

    Bryson S. T., et al., 2020, Kepler Science Document KSCI-19081-003, p. 3

    2020

  2. [10]

    A., et al., 2020, Research Notes of the AAS, 4, 201

    Caldwell D. A., et al., 2020, Research Notes of the AAS, 4, 201

    2020

  3. [11]

    W., et al., 2020, The Astronomical Journal, 160, 53

    Carmichael T. W., et al., 2020, The Astronomical Journal, 160, 53

    2020

  4. [12]

    Cho K., Van Merri \"e nboer B., Gulcehre C., Bahdanau D., Bougares F., Schwenk H., Bengio Y., 2014, arXiv preprint arXiv:1406.1078

    2014 arXiv

  5. [13]

    pp 3642--3649

    Ciregan D., Meier U., Schmidhuber J., 2012, in 2012 IEEE conference on computer vision and pattern recognition. pp 3642--3649

    2012

  6. [14]

    Cui K., Liu J., Feng F., Liu J., 2021, The Astronomical Journal, 163, 23

    2021

  7. [15]

    I., Johnson J

    Dawson R. I., Johnson J. A., 2018, @doi [ ] 10.1146/annurev-astro-081817-051853 , https://ui.adsabs.harvard.edu/abs/2018ARA&A..56..175D 56, 175

    2018 doi

  8. [16]

    I., et al., 2021, @doi [ ] 10.3847/1538-3881/abd8d0 , https://ui.adsabs.harvard.edu/abs/2021AJ....161..161D 161, 161

    Dawson R. I., et al., 2021, @doi [ ] 10.3847/1538-3881/abd8d0 , https://ui.adsabs.harvard.edu/abs/2021AJ....161..161D 161, 161

    2021 doi

  9. [17]

    Dosovitskiy A., et al., 2020, arXiv preprint arXiv:2010.11929

    2020 arXiv

  10. [18]

    D., Hogg D

    Foreman-Mackey D., Morton T. D., Hogg D. W., Agol E., Sch \"o lkopf B., 2016, The Astronomical Journal, 152, 206

    2016

  11. [19]

    N., 2017, in International Conference on Machine Learning

    Gehring J., Auli M., Grangier D., Yarats D., Dauphin Y. N., 2017, in International Conference on Machine Learning. pp 1243--1252

    2017

  12. [20]

    Gill S., et al., 2020a, Monthly Notices of the Royal Astronomical Society, 491, 1548

  13. [21]

    Gill S., et al., 2020b, The Astrophysical Journal Letters, 898, L11

  14. [22]

    Gill S., et al., 2020c, @doi [ ] 10.3847/2041-8213/ab9eb9 , https://ui.adsabs.harvard.edu/abs/2020ApJ...898L..11G 898, L11

    doi

  15. [23]

    MIT press

    Goodfellow I., Bengio Y., Courville A., 2016, Deep learning. MIT press

    2016

  16. [24]

    Grieves N., et al., 2022, @doi [ ] 10.1051/0004-6361/202244077 , https://ui.adsabs.harvard.edu/abs/2022A&A...668A..29G 668, A29

    2022 doi

  17. [25]

    M., et al., 2021, The Astrophysical Journal Supplement Series, 254, 39

    Guerrero N. M., et al., 2021, The Astrophysical Journal Supplement Series, 254, 39

    2021

  18. [26]

    D., et al., 2019, @doi [ ] 10.3847/1538-3881/aaf8b6 , https://ui.adsabs.harvard.edu/abs/2019AJ....157...55H 157, 55

    Hartman J. D., et al., 2019, @doi [ ] 10.3847/1538-3881/aaf8b6 , https://ui.adsabs.harvard.edu/abs/2019AJ....157...55H 157, 55

    2019 doi

  19. [27]

    M., 2004, Journal of chemical information and computer sciences, 44, 1

    Hawkins D. M., 2004, Journal of chemical information and computer sciences, 44, 1

    2004

  20. [28]

    Hawthorn F., et al., 2024, Monthly Notices of the Royal Astronomical Society, 528, 1841

    2024

  21. [29]

    pp 770--778

    He K., Zhang X., Ren S., Sun J., 2016, in Proceedings of the IEEE conference on computer vision and pattern recognition. pp 770--778

    2016

  22. [30]

    Hendrycks D., Gimpel K., 2016, arXiv preprint arXiv:1606.08415

    2016 arXiv

  23. [31]

    Hod z i \'c V., et al., 2018, Monthly Notices of the Royal Astronomical Society, 481, 5091

    2018

  24. [32]

    pp 449--456

    Karita S., et al., 2019, in 2019 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU). pp 449--456

    2019

  25. [33]

    P., Ba J., 2014, arXiv preprint arXiv:1412.6980

    Kingma D. P., Ba J., 2014, arXiv preprint arXiv:1412.6980

    2014 arXiv

  26. [34]

    B., et al., 2019, The Astronomical Journal, 157, 124

    Kostov V. B., et al., 2019, The Astronomical Journal, 157, 124

    2019

  27. [35]

    Kov \'a cs G., Zucker S., Mazeh T., 2002, Astronomy & Astrophysics, 391, 369

    2002

  28. [36]

    E., 2012, Advances in neural information processing systems, 25

    Krizhevsky A., Sutskever I., Hinton G. E., 2012, Advances in neural information processing systems, 25

    2012

  29. [37]

    E., 2017, Communications of the ACM, 60, 84

    Krizhevsky A., Sutskever I., Hinton G. E., 2017, Communications of the ACM, 60, 84

    2017

  30. [38]

    M., Cettolo M., Federico M., 2018, arXiv preprint arXiv:1806.06957

    Lakew S. M., Cettolo M., Federico M., 2018, arXiv preprint arXiv:1806.06957

    2018 arXiv

  31. [39]

    LeCun Y., Bengio Y., Hinton G., 2015, nature, 521, 436

    2015

  32. [40]

    Lendl M., et al., 2020, Monthly Notices of the Royal Astronomical Society, 492, 1761

    2020

  33. [41]

    Magliano C., et al., 2023, Monthly Notices of the Royal Astronomical Society, 521, 3749

    2023

  34. [42]

    G., 2017, The Astrophysical Journal Letters, 837, L1

    M \'e ndez A., Rivera-Valent \' n E. G., 2017, The Astrophysical Journal Letters, 837, L1

    2017

  35. [43]

    H., Waldmann I., 2022, arXiv preprint arXiv:2207.02777

    Morvan M., Nikolaou N., Yip K. H., Waldmann I., 2022, arXiv preprint arXiv:2207.02777

    2022 arXiv

  36. [44]

    Olmschenk G., et al., 2021, The Astronomical Journal, 161, 273

    2021

  37. [45]

    P., et al., 2020, Astronomy & Astrophysics, 633, A53

    Osborn H. P., et al., 2020, Astronomy & Astrophysics, 633, A53

    2020

  38. [46]

    Paszke A., et al., 2019, Advances in neural information processing systems, 32

    2019

  39. [47]

    B., Hubickyj O., Bodenheimer P., Lissauer J

    Pollack J. B., Hubickyj O., Bodenheimer P., Lissauer J. J., Podolak M., Greenzweig Y., 1996, @doi [ ] 10.1006/icar.1996.0190 , https://ui.adsabs.harvard.edu/abs/1996Icar..124...62P 124, 62

    1996

  40. [48]

    M., 2020, arXiv preprint arXiv:2010.16061

    Powers D. M., 2020, arXiv preprint arXiv:2010.16061

    2020 arXiv

  41. [49]

    Pr s a A., et al., 2022, The Astrophysical Journal Supplement Series, 258, 16

    2022

  42. [50]

    J., 2010, arXiv preprint arXiv:1006.3727

    Ragozzine D., Holman M. J., 2010, arXiv preprint arXiv:1006.3727

    2010 arXiv

  43. [51]

    Rao S., Mahabal A., Rao N., Raghavendra C., 2021, Monthly Notices of the Royal Astronomical Society, 502, 2845

    2021

  44. [52]

    Rauer H., Aerts C., Cabrera J., Team P., et al., 2016, Astronomische Nachrichten, 337, 961

    2016

  45. [53]

    Redmon J., Farhadi A., 2018, YOLOv3: An Incremental Improvement ( @eprint arXiv 1804.02767 ), https://arxiv.org/abs/1804.02767

    2018 arXiv

  46. [54]

    R., et al., 2014, Journal of Astronomical Telescopes, Instruments, and Systems, 1, 014003

    Ricker G. R., et al., 2014, Journal of Astronomical Telescopes, Instruments, and Systems, 1, 014003

    2014

  47. [55]

    R., et al., 2015, @doi [Journal of Astronomical Telescopes, Instruments, and Systems] 10.1117/1.JATIS.1.1.014003 , https://ui.adsabs.harvard.edu/abs/2015JATIS...1a4003R 1, 014003

    Ricker G. R., et al., 2015, @doi [Journal of Astronomical Telescopes, Instruments, and Systems] 10.1117/1.JATIS.1.1.014003 , https://ui.adsabs.harvard.edu/abs/2015JATIS...1a4003R 1, 014003

    2015 doi

  48. [56]

    Salinas H., Pichara K., Brahm R., Pérez-Galarce F., Mery D., 2023, @doi [Monthly Notices of the Royal Astronomical Society] 10.1093/mnras/stad1173 , 522, 3201

    2023 doi

  49. [57]

    F., Levrard B., Paillet J., Ribas I., Delfosse X., 2007, Astronomy & Astrophysics, 476, 1373

    Selsis F., Kasting J. F., Levrard B., Paillet J., Ribas I., Delfosse X., 2007, Astronomy & Astrophysics, 476, 1373

    2007

  50. [58]

    M., 2019, Journal of big data, 6, 1

    Shorten C., Khoshgoftaar T. M., 2019, Journal of big data, 6, 1

    2019

  51. [59]

    Srivastava N., Hinton G., Krizhevsky A., Sutskever I., Salakhutdinov R., 2014, The journal of machine learning research, 15, 1929

    2014

  52. [60]

    W., et al., 2015, The Astrophysical Journal, 809, 77

    Sullivan P. W., et al., 2015, The Astrophysical Journal, 809, 77

    2015

  53. [61]

    Tenenbaum P., et al., 2012, @doi [ ] 10.1088/0067-0049/199/1/24 , https://ui.adsabs.harvard.edu/abs/2012ApJS..199...24T 199, 24

    2012 doi

  54. [62]

    Tey E., et al., 2023, The Astronomical Journal, 165, 95

    2023

  55. [63]

    Valizadegan H., et al., 2022, The Astrophysical Journal, 926, 120

    2022

  56. [64]

    Vallenari A., et al., 2023, Astronomy & Astrophysics, 674, A1

    2023

  57. [65]

    N., Kaiser ., Polosukhin I., 2017, in Advances in neural information processing systems

    Vaswani A., Shazeer N., Parmar N., Uszkoreit J., Jones L., Gomez A. N., Kaiser ., Polosukhin I., 2017, in Advances in neural information processing systems. pp 5998--6008

    2017

  58. [66]

    Voulodimos A., Doulamis N., Doulamis A., Protopapadakis E., 2018, Computational intelligence and neuroscience, 2018

    2018

  59. [67]

    F., Chao L

    Wang Q., Li B., Xiao T., Zhu J., Li C., Wong D. F., Chao L. S., 2019, arXiv preprint arXiv:1906.01787

    2019 arXiv

  60. [68]

    Wang S., et al., 2021, The Astronomical Journal, 162, 50

    2021

  61. [69]

    F., et al., 2023, The Astrophysical Journal Supplement Series, 269, 5

    Wilson R. F., et al., 2023, The Astrophysical Journal Supplement Series, 269, 5

    2023

  62. [70]

    Yu L., et al., 2019, The Astronomical Journal, 158, 25

    2019

  63. [71]

    pp 2114--2124

    Zerveas G., Jayaraman S., Patel D., Bhamidipaty A., Eickhoff C., 2021, in Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. pp 2114--2124

    2021

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.