pith. sign in

arxiv: 2606.18932 · v1 · pith:GE3BD2GWnew · submitted 2026-06-17 · 🌌 astro-ph.EP · astro-ph.IM· cs.AI· cs.LG

TransitNet: A Compact Attention-Augmented Deep Learning Framework for Low-SNR Transit Blind Searches

Xingchen Yan , Jian Ge , Qingtian Liu , Kevin Willis , Quanquan Hu , Jiapeng Zhu This is my paper

Pith reviewed 2026-06-26 19:17 UTC · model grok-4.3

classification 🌌 astro-ph.EP astro-ph.IMcs.AIcs.LG
keywords exoplanet transit detectionlow signal-to-noise ratiodeep learningKepler missionattention mechanismblind searchEarth-size planets
0
0 comments X

The pith

TransitNet recovers 93 percent of injected Earth-size transits in Kepler data where TLS and BLS recover only about 60 percent.

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

The paper develops TransitNet as a compact attention-augmented neural network to find planetary transits at low signal-to-noise ratios that traditional methods miss. It builds a unified dataset and benchmarking setup to test the model under blind-search conditions on unseen Kepler targets. The central demonstration is that the network reaches 95.2 percent accuracy in the SNR 6-8 range and recovers 93 percent of Earth-size and sub-Earth-size injected signals, while also estimating transit windows and running 4 to 25 times faster than CPU versions of the standard algorithms. On real confirmed planets the model recovers every selected case with a mean midpoint error of 1.24 hours. These results are offered as evidence that the approach can extend searches for intermediate-to-long-period small planets.

Core claim

TransitNet is a compact attention-augmented deep-learning model that attains 95.2 percent accuracy on low-SNR (6-8) recovery benchmarks from unseen Kepler targets, ROC-AUC of 0.974, and PR-AP of 0.982, while recovering 93.0 percent of injected Earth-size and sub-Earth-size transits; the same model recovers all 34 selected confirmed Kepler planets on real data with 1.24-hour mean midpoint error and supplies attention-based transit-window estimates that fully cover 97.4 percent of injected cases.

What carries the argument

TransitNet, an attention-augmented convolutional-recurrent network that processes light curves to output both a detection score and attention-weighted transit-window estimates.

If this is right

  • The method can be applied directly to existing Kepler data to search for additional Earth-size planets at periods longer than those already catalogued.
  • Because the model runs 4-5 times faster than CPU-BLS, full blind searches of large photometric surveys become computationally practical on modest hardware.
  • Attention maps provide per-light-curve transit midpoint and duration estimates without a separate fitting step.
  • The compact 1.5 MB size allows deployment on edge devices or in real-time pipelines for upcoming surveys.

Where Pith is reading between the lines

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

  • Extending the same training and benchmarking pipeline to TESS or PLATO data would test whether the performance advantage holds for shorter-cadence or higher-noise observations.
  • If the attention mechanism reliably localizes transits, it could be combined with period-folding routines to improve orbital-parameter recovery for marginal detections.
  • The reported speed-up suggests that re-processing the entire Kepler archive at higher sensitivity thresholds is now feasible within existing compute budgets.

Load-bearing premise

The statistical properties of the injected signals and noise used to train and test the network match those of actual Kepler light curves without systematic biases that would favor the neural network.

What would settle it

A blind search on a fresh set of Kepler or TESS light curves containing only confirmed non-transit signals or planets with periods and depths outside the training distribution that yields substantially lower recovery rates than reported here.

Figures

Figures reproduced from arXiv: 2606.18932 by Jian Ge, Jiapeng Zhu, Kevin Willis, Qingtian Liu, QuanQuan Hu, XingChen Yan.

Figure 1
Figure 1. Figure 1: Confirmed exoplanets on the 𝑅p–𝑃 plane (log scale), with marginal histograms of 𝑃 (top) and 𝑅p (right). Solar System planets are shown as labelled magenta stars; dashed lines mark 𝑅p = 1 𝑅⊕, 𝑃 = 100 days, and 𝑃 = 101.5 days (≈ 31.6 days). The transit-detected population becomes sparse beyond 𝑃 ≈ 31.6 days, particularly in the regime of Earth-size planets, leaving an unpopulated region around 𝑅p ∼ 1 𝑅⊕. We … view at source ↗
Figure 2
Figure 2. Figure 2: Architecture of TransitNet after multiple optimisations. Our network processes a batch of 1D sequences (Batch, 𝐿0 ) through three stages: Filtering, MHA, and FCN. In the FM, five convolution blocks (CB) expand the channel width (16, 32, 64, 128, 256) and produce features of shape (Batch, 256, 𝐿1 ). Two variants of CBs are used: one with Batch Normalization (BN) and one without, both of which consist of Con… view at source ↗
Figure 3
Figure 3. Figure 3: Schematic of the MHA computation. The input feature matrix 𝐹, encoding flux patterns from the folded transit profile, is linearly projected into queries 𝑄 (search templates for transit-like features), keys 𝐾 (reference catalogue of flux patterns), and values 𝑉 (actual flux information) using learnt weights 𝑊𝑄, 𝑊𝐾 , and 𝑊𝑉 . Attention weights 𝐴 are computed via scaled dot-product attention, softmax 𝑄𝐾⊤/ √ 𝑑… view at source ↗
Figure 4
Figure 4. Figure 4: End-to-end feature-map visualisation of TransitNet for a transit (left) and a non-transit (right) input. From top to bottom: input light curves; FM (rows 2-6); MHA — 𝑄, 𝐾, attention matrix 𝐴 = softmax(𝑄𝐾⊤/ √ 𝑑𝐾 ), 𝑉, and 𝐴𝑉 (rows 7–11); FCN (rows 12–18); detection score spectrum (bottom). MNRAS 000, 1–20 (2026) [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Distributions of transit parameters for Kepler CPs with orbital periods 𝑃 ∈ [30, 60] days and SNR ∈ [6, 15]. Panels show (left to right): orbital period 𝑃, transit duration 𝑇14, and transit depth 𝛿. Inset boxes display summary statistics (min, max, mean, median) for each distribution. These empirical distributions serve as the basis for generating synthetic transit signals in our training dataset. where 𝛿 … view at source ↗
Figure 6
Figure 6. Figure 6 [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Detection score distributions of transit and non-transit light curves on the Low-SNR Transit Recovery Set for BLS, TLS, and TransitNet. Tran￾sitNet exhibits the clearest separation between transit and non-transit popu￾lations, indicating improved recoverability of low-SNR transit signals while maintaining strong rejection of non-transit backgrounds. the Cross-KIC Recovery Set, another independent benchmark… view at source ↗
Figure 8
Figure 8. Figure 8: ROC and PR curves for BLS, TLS, and TransitNet on the Low-SNR Transit Recovery Set, evaluated in four SNR bins: [6, 8), [8, 10), [10, 12), and [12, 15]. Each bin contains equal numbers of transit and non-transit samples derived from the same KIC target. TransitNet achieves substantially higher detection performance in the low-SNR regime, while the performance differences diminish as SNR increases. overall … view at source ↗
Figure 11
Figure 11. Figure 11: Threshold selection for BLS, TLS, and TransitNet on the Cross￾KIC Recovery Set based on maximising Youden’s statistic, 𝐽 = TPR − FPR, and equivalently minimising the mean classification error, 1 2 (FPR + FNR). Metrics are computed independently for each KIC and then macro-averaged across all KICs. Vertical dashed lines indicate the adopted operating thresh￾olds (BLS: 6.76, TLS: 9.80, and TransitNet: 0.54)… view at source ↗
Figure 9
Figure 9. Figure 9: Violin plots of detection scores on the Low-SNR Transit Recovery Set for BLS, TLS, and TransitNet, stratified by SNR bins [6, 8), [8, 10), [10, 12), and [12, 15], with an additional column aggregating all non-transit light curves (rightmost). The light pink shaded band, termed the non-transit￾prone (NT-prone) region, is bounded above by the score threshold corre￾sponding to a FPR of 1%. TransitNet consiste… view at source ↗
Figure 10
Figure 10. Figure 10: Distribution of per-KIC ROC-AUC and PR-AP values on the Cross-KIC Recovery Set. For each KIC, the metrics are computed from the corresponding transit and non-transit detection scores spectrum. These dis￾tributions evaluate the transit recovery performance of different algorithms across previously unseen KICs. Distributions that are more concentrated near 1 indicate better and more stable performance under… view at source ↗
Figure 12
Figure 12. Figure 12: Confusion matrices showing the classification performance of BLS, TLS, and TransitNet on the Low-SNR Transit Recovery Set with SNR ∈ [6, 8). Classification is performed using the operating thresholds selected from the macro-averaged threshold optimisation analysis shown in [PITH_FULL_IMAGE:figures/full_fig_p014_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Detection score spectrum for BLS, TLS, and TransitNet on semi-synthetic ATLCs. The left column shows results for an injected transit signal with SNR = 8.8 and true period of 35.38 days, where all three algorithms successfully detect the target transit, with scores above the operating thresholds selected in [PITH_FULL_IMAGE:figures/full_fig_p014_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Recovery of injected Earth-size and sub-Earth-size transits in Kepler TMLCs. Top: planet radius versus period; blue and gray circles denote injected transits recovered and missed at algorithm-specific operating thresholds ( [PITH_FULL_IMAGE:figures/full_fig_p015_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Runtime scaling of transit-search algorithms as a function of period-grid size. Dashed lines denote CPU-based implementations, while solid lines denote GPU-accelerated implementations. The left panel shows the mean wall-clock inference time; the right panel shows the speed-up factor relative to CPU TLS, 𝑇CPU-TLS/𝑇. Curves compare CPU-TLS, CPU-BLS, GPU-BLS, and TransitNet. The horizontal axis gives 𝑁period… view at source ↗
Figure 16
Figure 16. Figure 16: Schematic illustration of the transit-region scoring scheme com￾paring an estimated transit window, [𝜏ˆ𝑎, 𝜏ˆ𝑏 ] (light blue), with the true in￾transit interval, [𝜏1, 𝜏4 ] (magenta). Black points show synthetic photometry generated from a Batman model (solid red curve), and the hatched region denotes the covered transit interval. The three panels illustrate partial cover￾age (left; 0 < 𝑠 < 1), full coverag… view at source ↗
Figure 17
Figure 17. Figure 17: Transit window overlap versus transit and observational param￾eters on a separate evaluation set. Points show bin-averaged overlap scores; shaded bands indicate mean ± standard error of the mean (SEM). The score measures agreement between the estimated ingress-egress interval [𝜏ˆ𝑎, 𝜏ˆ𝑏 ] and the true transit window. Panels: SNR (top left), orbital period in days (top right), transit duration in hours (bot… view at source ↗
Figure 18
Figure 18. Figure 18: Analysis of two examples. The top panel in each figure shows the normalised flux over time. The estimated transit midpoint (𝜏ˆ0) is indicated by the red dashed line, while the true 𝑇0 is marked by the blue dashed line. The bottom panel displays the attention weights, with peaks indicating the temporal segments most relevant for transit detection. estimation in a single forward pass. This end-to-end formul… view at source ↗
read the original abstract

Motivated by the observational incompleteness of intermediate-to-long-period Earth-size planets, we present TransitNet, a compact attention-augmented deep-learning framework for low-SNR transit blind searches. To enable realistic method development and objective threshold calibration under blind-search conditions, we develop a unified dataset construction, benchmarking, and threshold-selection framework. On recovery benchmarks constructed from unseen Kepler targets, TransitNet attains 95.2 percent accuracy in the challenging SNR range of 6 to 8 and outperforms both TLS and BLS, achieving ROC-AUC and PR-AP values of 0.974 and 0.982, respectively. In an injected Earth-size and sub-Earth-size transit recovery experiment, TransitNet achieves a recovery rate of 93.0 percent, substantially exceeding those of TLS (63.1 percent) and BLS (60.0 percent). In addition to detection, TransitNet provides attention-based estimates of transit windows and midpoints. On an independent evaluation set, 97.4 percent of injected transits are fully covered by the estimated transit window. Applied to real Kepler observations, the model successfully recovers all 34 selected confirmed Kepler planets, with a mean absolute transit midpoint error of 1.24 hours. The model combines a compact footprint of about 1.5 MB with high inference efficiency, yielding speed-ups of about 12 to 25 times relative to CPU-TLS and about 4 to 5 times relative to CPU-BLS. These results demonstrate that TransitNet provides an accurate, scalable, and computationally efficient framework for low-SNR transit blind searches in the tested regime and motivate its extension to longer-period Earth-size planet searches.

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

2 major / 2 minor

Summary. The manuscript introduces TransitNet, a compact attention-augmented neural network for blind low-SNR transit searches in Kepler photometry. It develops a unified dataset construction, injection, and benchmarking pipeline and reports 95.2% accuracy (SNR 6–8), ROC-AUC 0.974 and PR-AP 0.982 on held-out Kepler targets, 93% recovery of injected Earth/sub-Earth transits (vs. 63.1% TLS, 60% BLS), 97.4% transit-window coverage on an independent set, recovery of all 34 tested confirmed planets (mean midpoint error 1.24 h), and 4–25× speed-up relative to CPU baselines. The model footprint is stated as ~1.5 MB.

Significance. If the benchmark results prove robust, TransitNet would supply a practical, scalable complement to traditional periodogram methods for recovering small, long-period planets that remain observationally incomplete. The attention-based window estimation and the unified injection-plus-threshold framework are constructive contributions that could aid future method comparisons. The computational efficiency is a clear practical strength for large-scale surveys.

major comments (2)
  1. [Dataset construction / unified benchmarking framework] § Dataset construction / unified benchmarking framework: The central performance claims (95.2% accuracy, 93% Earth-size recovery, large gaps versus TLS/BLS) rest on the premise that the injected signals plus underlying noise statistically match the distribution of real undetected low-SNR Kepler transits, including red noise and instrumental systematics. No quantitative validation (power-spectrum matching, variability statistics, or comparison against real non-detections) is reported; any systematic mismatch would directly inflate the reported advantage of the attention layers over BLS/TLS.
  2. [Baseline implementation details (methods section)] Baseline implementation details (methods section): The large reported gaps versus TLS and BLS require explicit confirmation that the traditional pipelines were run with equivalent search ranges, detrending choices, and threshold calibration on the identical light-curve set. Without these specifics it remains unclear whether the performance difference is intrinsic or arises from implementation disparities.
minor comments (2)
  1. The abstract repeatedly uses approximate phrasing (“about 1.5 MB”, “about 12 to 25 times”); exact values and hardware specifications would improve reproducibility.
  2. Clarify whether the 34 confirmed planets used for real-data validation were selected before or after model development to avoid inadvertent selection bias.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive review and the positive assessment of the work's potential impact. We address the two major comments point by point below and will incorporate clarifications and additional validation into the revised manuscript.

read point-by-point responses

  1. Referee: [Dataset construction / unified benchmarking framework] The central performance claims (95.2% accuracy, 93% Earth-size recovery, large gaps versus TLS/BLS) rest on the premise that the injected signals plus underlying noise statistically match the distribution of real undetected low-SNR Kepler transits, including red noise and instrumental systematics. No quantitative validation (power-spectrum matching, variability statistics, or comparison against real non-detections) is reported; any systematic mismatch would directly inflate the reported advantage of the attention layers over BLS/TLS.

    Authors: The injection procedure operates directly on real Kepler light curves pre-selected to contain no known transits, thereby retaining the actual red noise, instrumental systematics, and variability present in the survey data by construction. We agree, however, that explicit quantitative checks would strengthen the claim. In the revised manuscript we will add power-spectral-density comparisons and basic variability statistics between the injected training/validation sets and an independent sample of real non-detection Kepler light curves. revision: yes

  2. Referee: [Baseline implementation details (methods section)] The large reported gaps versus TLS and BLS require explicit confirmation that the traditional pipelines were run with equivalent search ranges, detrending choices, and threshold calibration on the identical light-curve set. Without these specifics it remains unclear whether the performance difference is intrinsic or arises from implementation disparities.

    Authors: We concur that full reproducibility of the baseline comparisons requires explicit parameter documentation. The original manuscript states that standard TLS and BLS implementations were applied to the same light-curve sets, but the revised methods section will list the precise search-period grids, detrending filters, and threshold-selection procedure (all performed on the identical held-out Kepler targets) to remove any ambiguity. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical ML performance on held-out data with no self-referential derivations

full rationale

The paper presents a trained neural network (TransitNet) and reports its detection performance on benchmarks constructed from unseen Kepler targets plus independent injected transit sets. These are standard held-out evaluation practices with no equations, parameters, or predictions that reduce by construction to quantities defined from the training data or fitted thresholds. No self-citations, uniqueness theorems, or ansatzes are invoked to justify the central claims; the reported accuracies, ROC-AUC, recovery rates, and midpoint errors are externally falsifiable empirical results rather than tautological outputs of the model's own training procedure.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central performance claims rest on the realism of the simulated dataset and the assumption that the chosen training/validation splits do not leak information about the test cases.

free parameters (1)
  • neural network hyperparameters and training schedule
    Architecture depth, attention heads, learning rate, and regularization choices are selected or optimized on the training data to achieve the reported metrics.
axioms (1)
  • domain assumption Injected transit signals plus added noise accurately reproduce the statistical properties of real Kepler light curves at SNR 6-8.
    All recovery-rate comparisons depend on this simulation fidelity.

pith-pipeline@v0.9.1-grok · 5859 in / 1304 out tokens · 27757 ms · 2026-06-26T19:17:07.250243+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

85 extracted references · 14 canonical work pages

  1. [1]

    and Marcy, Geoffrey W

    Howard, Andrew W. and Marcy, Geoffrey W. and Bryson, Stephen T. and Jenkins, Jon M. and Rowe, Jason F. and Batalha, Natalie M. and Borucki2, William J. and Koch, David G. and Dunham, Edward W. and Gautier, Thomas N. and Van Cleve, Jeffrey and Cochran, William D. and Latham, David W. and Lissauer, Jack J. and Torres, Guillermo and Brown, Timothy M. and Gil...

  2. [2]

    and Coughlin, Jeffrey L

    Thompson, Susan E. and Coughlin, Jeffrey L. and Hoffman, Kelsey and Mullally, Fergal and Christiansen, Jessie L. and Burke, Christopher J. and Bryson, Steve and Batalha, Natalie and Haas, Michael R. and Catanzarite, Joseph and Rowe, Jason F. and Barentsen, Geert and Caldwell, Douglas A. and Clarke, Bruce D. and Jenkins, Jon M. and Li, Jie and Latham, Davi...

    work page doi:10.3847/1538-4365/aab4f9 2018

  3. [3]

    Youden, W. J. , title =. Cancer , volume =. doi:https://doi.org/10.1002/1097-0142(1950)3:1<32::AID-CNCR2820030106>3.0.CO;2-3 , url =. https://acsjournals.onlinelibrary.wiley.com/doi/pdf/10.1002/1097-0142 year =

    work page doi:10.1002/1097-0142(1950)3:1 1950

  4. [4]

    and Christiansen, Jessie and Dressing, Courtney D

    Fressin, François and Torres, Guillermo and Charbonneau, David and Bryson, Stephen T. and Christiansen, Jessie and Dressing, Courtney D. and Jenkins, Jon M. and Walkowicz, Lucianne M. and Batalha, Natalie M. , doi =. THE FALSE POSITIVE RATE OFKEPLERAND THE OCCURRENCE OF PLANETS , url =. ApJ , month =

  5. [5]

    and Coughlin, Jeffrey L

    Bryson, Steve and Kunimoto, Michelle and Kopparapu, Ravi K. and Coughlin, Jeffrey L. and Borucki, William J. and Koch, David and Aguirre, Victor Silva and Allen, Christopher and Barentsen, Geert and Batalha, Natalie M. and Berger, Travis and Boss, Alan and Buchhave, Lars A. and Burke, Christopher J. and Caldwell, Douglas A. and Campbell, Jennifer R. and C...

  6. [6]

    Pollacco, D. L. and Skillen, I. and Cameron, A. Collier and Christian, D. J. and Hellier, C. and Irwin, J. and Lister, T. A. and Street, R. A. and West, R. G. and Anderson, D. and Clarkson, W. I. and Deeg, H. and Enoch, B. and Evans, A. and Fitzsimmons, A. and Haswell, C. A. and Hodgkin, S. and Horne, K. and Kane, S. R. and Keenan, F. P. and Maxted, P. F....

  7. [7]

    and Siverd, Robert and James, David and Stassun, Keivan , doi =

    Pepper, Joshua and Kuhn, Rudolf B. and Siverd, Robert and James, David and Stassun, Keivan , doi =. The KELT-South Telescope1 , url =. PASP , month =. 2012 , bdsk-url-1 =

    2012

  8. [8]

    2022 , eprint=

    ET White Paper: To Find the First Earth 2.0 , author=. 2022 , eprint=

    2022

  9. [9]

    Ge, Jian and Zhang, Hui and Deng, Hongping and Howell, Steve B. , doi =. The ET mission to search for earth 2.0s , url =. The Innovation , month =

  10. [10]

    2022 , booktitle =

    The Earth 2.0 Space Mission for Detecting Earth-like Planets Around Solar Type Stars , author =. 2022 , booktitle =. doi:10.1117/12.2630656 , url =

    work page doi:10.1117/12.2630656 2022

  11. [11]

    Chinese Journal of Space Science , volume =

    Ge, Jian and Chen, Wen and Chen, Yonghe and Song, Zongxi and Wang, Jian and Zhang, Hui and Li, Yan and Zang, Weicheng and Zhou, Dan and Zhang, Yongshuai and Chen, Kun and Yang, Yingquan and Mao, Shude and Huang, Chelsea and Yao, Xinyu and Li, Xinglong and Jiang, Haijiao and Yu, Yong and Tang, Zhenghong and Dong, Feng and Gao, Wei and Zhang, Hongfei and Sh...

    2024

  12. [12]

    2024 , booktitle =

    The Earth 2.0 Mission: Space Telescopes and Instrumentation 2024 , author =. 2024 , booktitle =. doi:10.1117/12.3018669 , url =

    work page doi:10.1117/12.3018669 2024

  13. [13]

    Rauer, Heike and Aerts, Conny and Cabrera, Juan and Deleuil, Magali and Erikson, Anders and Gizon, Laurent and Goupil, Mariejo and Heras, Ana and Walloschek, Thomas and Lorenzo-Alvarez, Jose and Marliani, Filippo and Martin-Garcia, César and Mas-Hesse, J. Miguel and O’Rourke, Laurence and Osborn, Hugh and Pagano, Isabella and Piotto, Giampaolo and Pollacc...

  14. [14]

    and Bodin, P

    Auvergne, M. and Bodin, P. and Boisnard, L. and Buey, J.-T. and Chaintreuil, S. and Epstein, G. and Jouret, M. and Lam-Trong, T. and Levacher, P. and Magnan, A. and Perez, R. and Plasson, P. and Plesseria, J. and Peter, G. and Steller, M. and Tiphène, D. and Baglin, A. and Agogué, P. and Appourchaux, T. and Barbet, D. and Beaufort, T. and Bellenger, R. an...

  15. [15]

    and Borucki, William J

    Koch, David G. and Borucki, William J. and Basri, Gibor and Batalha, Natalie M. and Brown, Timothy M. and Caldwell, Douglas and Christensen-Dalsgaard, Jørgen and Cochran, William D. and DeVore, Edna and Dunham, Edward W. and Gautier, Thomas N. and Geary, John C. and Gilliland, Ronald L. and Gould, Alan and Jenkins, Jon and Kondo, Yoji and Latham, David W....

  16. [16]

    and Sobeck, Charlie and Haas, Michael and Still, Martin and Barclay, Thomas and Mullally, Fergal and Troeltzsch, John and Aigrain, Suzanne and Bryson, Stephen T

    Howell, Steve B. and Sobeck, Charlie and Haas, Michael and Still, Martin and Barclay, Thomas and Mullally, Fergal and Troeltzsch, John and Aigrain, Suzanne and Bryson, Stephen T. and Caldwell, Doug and Chaplin, William J. and Cochran, William D. and Huber, Daniel and Marcy, Geoffrey W. and Miglio, Andrea and Najita, Joan R. and Smith, Marcie and Twicken, ...

  17. [17]

    and Winn, Joshua N

    Ricker, George R. and Winn, Joshua N. and Vanderspek, Roland and Latham, David W. and Bakos, Gáspár Á. and Bean, Jacob L. and Berta-Thompson, Zachory K. and Brown, Timothy M. and Buchhave, Lars and Butler, Nathaniel R. and Butler, R. Paul and Chaplin, William J. and Charbonneau, David and Christensen-Dalsgaard, Jørgen and Clampin, Mark and Deming, Drake a...

  18. [18]

    Exp Astron , volume =

    PLATO: PLAnetary Transits and Oscillations of Stars , author =. Exp Astron , volume =. 2009 , fjournal =. doi:10.1007/s10686-008-9122-9 , url =

    work page doi:10.1007/s10686-008-9122-9 2009

  19. [19]

    and Agol, Eric , doi =

    Carter, Joshua A. and Agol, Eric , doi =. THE QUASIPERIODIC AUTOMATED TRANSIT SEARCH ALGORITHM , url =. ApJ , month =

  20. [20]

    and Zucker, S

    Kovács, G. and Zucker, S. and Mazeh, T. , doi =. A box-fitting algorithm in the search for periodic transits , url =. A&A , month =

  21. [21]

    Optimized transit detection algorithm to search for periodic transits of small planets , url =

    Hippke, Michael and Heller, René , doi =. Optimized transit detection algorithm to search for periodic transits of small planets , url =. A&A , month =

  22. [22]

    and Mullally, Fergal and Coughlin, Jeff and Christiansen, Jessie L

    Thompson, Susan E. and Mullally, Fergal and Coughlin, Jeff and Christiansen, Jessie L. and Henze, Christopher E. and Haas, Michael R. and Burke, Christopher J. , doi =. A MACHINE LEARNING TECHNIQUE TO IDENTIFY TRANSIT SHAPED SIGNALS , url =. ApJ , month =

  23. [23]

    and Coughlin, Jeffery L

    Mullally, F. and Coughlin, Jeffery L. and Thompson, Susan E. and Christiansen, Jessie and Burke, Christopher and Clarke, Bruce D. and Haas, Michael R. , doi =. Identifying False Alarms in theKeplerPlanet Candidate Catalog , url =. PASP , month =

  24. [24]

    Armstrong, D. J. and Pollacco, D. and Santerne, A. , doi =. Transit shapes and self-organizing maps as a tool for ranking planetary candidates: application toKeplerandK2 , url =. MNRAS , month =

  25. [25]

    TSARDI: a Machine Learning data rejection algorithm for transiting exoplanet light curves , url =

    Mislis, D and Pyrzas, S and Alsubai, K A , doi =. TSARDI: a Machine Learning data rejection algorithm for transiting exoplanet light curves , url =. MNRAS , month =

  26. [26]

    Exoplanet validation with machine learning: 50 new validated Kepler planets , url =

    Armstrong, David J and Gamper, Jevgenij and Damoulas, Theodoros , doi =. Exoplanet validation with machine learning: 50 new validated Kepler planets , url =. MNRAS , month =

  27. [27]

    Auto-Vetting Transiting Planet Candidates Identified by the Kepler Pipeline , author =. Proc. IAU , volume =. 2012 , fjournal =. doi:10.1017/S1743921313012611 , url =

    work page doi:10.1017/s1743921313012611 2012

  28. [28]

    and Jenkins, Jon M

    McCauliff, Sean D. and Jenkins, Jon M. and Catanzarite, Joseph and Burke, Christopher J. and Coughlin, Jeffrey L. and Twicken, Joseph D. and Tenenbaum, Peter and Seader, Shawn and Li, Jie and Cote, Miles , doi =. AUTOMATIC CLASSIFICATION OFKEPLERPLANETARY TRANSIT CANDIDATES , url =. ApJ , month =

  29. [29]

    MNRAS , volume =

    Sidra: A Blind Algorithm for Signal Detection in Photometric Surveys , author =. MNRAS , volume =. 2016 , fjournal =. doi:10.1093/mnras/stv2333 , url =

    work page doi:10.1093/mnras/stv2333 2016

  30. [30]

    and Feigelson, Eric D

    Melton, Elizabeth J. and Feigelson, Eric D. and Montalto, Marco and Caceres, Gabriel A. and Rosenswie, Andrew W. and Abelson, Cullen S. , doi =. DIAmante TESS AutoRegressive Planet Search (DTARPS). I. Analysis of 0.9 Million Light Curves , url =. AJ , month =

  31. [31]

    The detection of transiting exoplanets by Gaia , url =

    Panahi, Aviad and Zucker, Shay and Clementini, Gisella and Audard, Marc and Binnenfeld, Avraham and Cusano, Felice and Evans, Dafydd Wyn and Gomel, Roy and Holl, Berry and Ilyin, Ilya and de Fombelle, Grégory Jevardat and Mazeh, Tsevi and Mowlavi, Nami and Nienartowicz, Krzysztof and Rimoldini, Lorenzo and Shahaf, Sahar and Eyer, Laurent , doi =. The dete...

  32. [32]

    Exoplanet detection using machine learning , url =

    Malik, Abhishek and Moster, Benjamin P and Obermeier, Christian , doi =. Exoplanet detection using machine learning , url =. MNRAS , month =

  33. [33]

    2021 , eprint=

    Automation Of Transiting Exoplanet Detection, Identification and Habitability Assessment Using Machine Learning Approaches , author=. 2021 , eprint=

    2021

  34. [34]

    and Vanderburg, Andrew , doi =

    Shallue, Christopher J. and Vanderburg, Andrew , doi =. Identifying Exoplanets with Deep Learning: A Five-planet Resonant Chain around Kepler-80 and an Eighth Planet around Kepler-90 , url =. AJ , month =

  35. [35]

    and Palafox, Leon and Griffith, Caitlin A

    Pearson, Kyle A. and Palafox, Leon and Griffith, Caitlin A. , doi =. Searching for exoplanets using artificial intelligence , url =. MNRAS , month =

  36. [36]

    and Mayo, Andrew W

    Dattilo, Anne and Vanderburg, Andrew and Shallue, Christopher J. and Mayo, Andrew W. and Berlind, Perry and Bieryla, Allyson and Calkins, Michael L. and Esquerdo, Gilbert A. and Everett, Mark E. and Howell, Steve B. and Latham, David W. and Scott, Nicholas J. and Yu, Liang , doi =. Identifying Exoplanets with Deep Learning. II. Two New Super-Earths Uncove...

  37. [37]

    and Crossfield, Ian J

    Yu, Liang and Vanderburg, Andrew and Huang, Chelsea and Shallue, Christopher J. and Crossfield, Ian J. M. and Gaudi, B. Scott and Daylan, Tansu and Dattilo, Anne and Armstrong, David J. and Ricker, George R. and Vanderspek, Roland K. and Latham, David W. and Seager, Sara and Dittmann, Jason and Doty, John P. and Glidden, Ana and Quinn, Samuel N. , doi =. ...

  38. [38]

    and Shporer, Avi and Daylan, Tansu and Muthukrishna, Daniel and Vanderburg, Andrew and Dattilo, Anne and Ricker, George R

    Tey, Evan and Moldovan, Dan and Kunimoto, Michelle and Huang, Chelsea X. and Shporer, Avi and Daylan, Tansu and Muthukrishna, Daniel and Vanderburg, Andrew and Dattilo, Anne and Ricker, George R. and Seager, S. , doi =. Identifying Exoplanets with Deep Learning. V. Improved Light-curve Classification for TESS Full-frame Image Observations , url =. AJ , month =

  39. [39]

    and Sasdelli, Michele and Smith, Jeffrey C

    Ansdell, Megan and Ioannou, Yani and Osborn, Hugh P. and Sasdelli, Michele and Smith, Jeffrey C. and Caldwell, Douglas and Jenkins, Jon M. and Räissi, Chedy and Angerhausen, Daniel , doi =. Scientific Domain Knowledge Improves Exoplanet Transit Classification with Deep Learning , url =. ApJL , month =

  40. [40]

    Valizadegan, Hamed and Martinho, Miguel J. S. and Wilkens, Laurent S. and Jenkins, Jon M. and Smith, Jeffrey C. and Caldwell, Douglas A. and Twicken, Joseph D. and Gerum, Pedro C. L. and Walia, Nikash and Hausknecht, Kaylie and Lubin, Noa Y. and Bryson, Stephen T. and Oza, Nikunj C. , doi =. ExoMiner: A Highly Accurate and Explainable Deep Learning Classi...

  41. [41]

    Valizadegan, Hamed and Martinho, Miguel J. S. and Jenkins, Jon M. and Caldwell, Douglas A. and Twicken, Joseph D. and Bryson, Stephen T. , doi =. Multiplicity Boost of Transit Signal Classifiers: Validation of 69 New Exoplanets using the Multiplicity Boost of ExoMiner , url =. AJ , month =

  42. [42]

    Shallow Transits—Deep Learning

    Zucker, Shay and Giryes, Raja , doi =. Shallow Transits—Deep Learning. I. Feasibility Study of Deep Learning to Detect Periodic Transits of Exoplanets , url =. AJ , month =

  43. [43]

    One-Dimensional Convolutional Neural Networks for Detecting Transiting Exoplanets , url =

    Iglesias Álvarez, Santiago and Díez Alonso, Enrique and Sánchez Rodríguez, María Luisa and Rodríguez Rodríguez, Javier and Sánchez Lasheras, Fernando and de Cos Juez, Francisco Javier , doi =. One-Dimensional Convolutional Neural Networks for Detecting Transiting Exoplanets , url =. Axioms , month =

  44. [44]

    Osborn, H. P. and Ansdell, M. and Ioannou, Y. and Sasdelli, M. and Angerhausen, D. and Caldwell, D. and Jenkins, J. M. and Räissi, C. and Smith, J. C. , doi =. Rapid classification of TESS planet candidates with convolutional neural networks , url =. A&A , month =

  45. [45]

    and Burke, Christopher J

    Li, Jie and Tenenbaum, Peter and Twicken, Joseph D. and Burke, Christopher J. and Jenkins, Jon M. and Quintana, Elisa V. and Rowe, Jason F. and Seader, Shawn E. , year =. PASP , volume =. doi:10.1088/1538-3873/aaf44d , url =

    work page doi:10.1088/1538-3873/aaf44d

  46. [46]

    Searching for Possible Exoplanet Transits from BRITE Data through a Machine Learning Technique , url =

    Yeh, Li-Chin and Jiang, Ing-Guey , doi =. Searching for Possible Exoplanet Transits from BRITE Data through a Machine Learning Technique , url =. PASP , month =

  47. [47]

    PLoS ONE , volume =

    Deep Learning Exoplanets Detection by Combining Real and Synthetic Data , author =. PLoS ONE , volume =. 2022 , fjournal =. doi:10.1371/journal.pone.0268199 , url =

    work page doi:10.1371/journal.pone.0268199 2022

  48. [48]

    Detecting Exoplanet Transits through Machine-learning Techniques with Convolutional Neural Networks , url =

    Chintarungruangchai, Pattana and Jiang, Ing-Guey , doi =. Detecting Exoplanet Transits through Machine-learning Techniques with Convolutional Neural Networks , url =. PASP , month =

  49. [49]

    Identify Light-curve Signals with Deep Learning Based Object Detection Algorithm

    Cui, Kaiming and Liu, Junjie and Feng, Fabo and Liu, Jifeng , doi =. Identify Light-curve Signals with Deep Learning Based Object Detection Algorithm. I. Transit Detection , url =. AJ , month =

  50. [50]

    and Latham, David W

    Brown, Timothy M. and Latham, David W. and Everett, Mark E. and Esquerdo, Gilbert A. , doi =. KEPLER INPUT CATALOG: PHOTOMETRIC CALIBRATION AND STELLAR CLASSIFICATION , url =. AJ , month =. 2011 , bdsk-url-1 =

    2011

  51. [51]

    Shallow Transits—Deep Learning

    Dvash, Elad and Peleg, Yam and Zucker, Shay and Giryes, Raja , doi =. Shallow Transits—Deep Learning. II. Identify Individual Exoplanetary Transits in Red Noise using Deep Learning , url =. AJ , month =

  52. [52]

    2022 , eprint=

    Exoplanet Detection by Machine Learning with Data Augmentation , author=. 2022 , eprint=

    2022

  53. [53]

    Distinguishing a planetary transit from false positives: a Transformer-based classification for planetary transit signals , url =

    Salinas, Helem and Pichara, Karim and Brahm, Rafael and Pérez-Galarce, Francisco and Mery, Domingo , doi =. Distinguishing a planetary transit from false positives: a Transformer-based classification for planetary transit signals , url =. MNRAS , month =

  54. [54]

    2023 , eprint=

    Identification and Classification of Exoplanets Using Machine Learning Techniques , author=. 2023 , eprint=

    2023

  55. [55]

    Discovery of small ultra-short-period planets orbiting Kepler KG dwarfs with GPU phase folding and deep learning , url =

    Wang, Kaitlyn and Ge, Jian and Willis, Kevin and Wang, Kevin and Zhao, Yinan and Hu, Quanquan , doi =. Discovery of small ultra-short-period planets orbiting Kepler KG dwarfs with GPU phase folding and deep learning , url =. MNRAS , month =

  56. [56]

    The GPU phase folding and deep learning method for detecting exoplanet transits , url =

    Wang, Kaitlyn and Ge, Jian and Willis, Kevin and Wang, Kevin and Zhao, Yinan , doi =. The GPU phase folding and deep learning method for detecting exoplanet transits , url =. MNRAS , month =

  57. [57]

    and Kaiser,

    Vaswani, Ashish and Shazeer, Noam and Parmar, Niki and Uszkoreit, Jakob and Jones, Llion and Gomez, Aidan N. and Kaiser,. Attention Is All You Need , booktitle=

  58. [58]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Long, Jonathan and Shelhamer, Evan and Darrell, Trevor , booktitle =. Fully Convolutional Networks for Semantic Segmentation , year =. doi:10.1109/CVPR.2015.7298965 , langid =

    work page doi:10.1109/cvpr.2015.7298965 2015

  59. [59]

    2017 , eprint=

    MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications , author=. 2017 , eprint=

    2017

  60. [60]

    2019 , editor =

    Tan, Mingxing and Le, Quoc , booktitle =. 2019 , editor =

    2019

  61. [61]

    Xception: Deep Learning with Depthwise Separable Convolutions , year =

    Chollet, Fran. Xception: Deep Learning with Depthwise Separable Convolutions , year =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , doi =

  62. [62]

    Journal of Machine Learning Research , year =

    Dropout: A Simple Way to Prevent Neural Networks from Overfitting , author =. Journal of Machine Learning Research , year =

  63. [63]

    Proceedings of the 32nd International Conference on Machine Learning , pages =

    Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift , author =. Proceedings of the 32nd International Conference on Machine Learning , pages =. 2015 , editor =

    2015

  64. [64]

    International Conference on Learning Representations (ICLR) , year=

    Decoupled Weight Decay Regularization , author=. International Conference on Learning Representations (ICLR) , year=

  65. [65]

    International Conference on Learning Representations (ICLR) , year=

    Adam: A Method for Stochastic Optimization , author=. International Conference on Learning Representations (ICLR) , year=

  66. [66]

    and Caceres, Gabriel A

    Gondhalekar, Yash and Feigelson, Eric D. and Caceres, Gabriel A. and Montalto, Marco and Saha, Snehanshu , doi =. A Study of Two Periodogram Algorithms for Improving the Detection of Small Transiting Planets , url =. ApJL , month =. 2023 , bdsk-url-1 =

    2023

  67. [67]

    and Marcy, Geoffrey W

    Petigura, Erik A. and Marcy, Geoffrey W. and Howard, Andrew W. , doi =. A PLATEAU IN THE PLANET POPULATION BELOW TWICE THE SIZE OF EARTH , url =. ApJ , month =

  68. [68]

    GTLS: A method for speeding up periodic transit detection using GPU , year =

    Hu, Quanquan and Ge, Jian and Jin, Luoxi and Willis, Kevin , note =. GTLS: A method for speeding up periodic transit detection using GPU , year =

  69. [69]

    2026 , note =

    Michael Telesco and Jian Ge and Kevin Willis and Chuanfei Dong and Jun Yang and Beibei Liu and Luoxi Jin , title =. 2026 , note =

    2026

  70. [70]

    and Caldwell, Douglas A

    Jenkins, Jon M. and Caldwell, Douglas A. and Chandrasekaran, Hema and Twicken, Joseph D. and Bryson, Stephen T. and Quintana, Elisa V. and Clarke, Bruce D. and Li, Jie and Allen, Christopher and Tenenbaum, Peter and Wu, Hayley and Klaus, Todd C. and Middour, Christopher K. and Cote, Miles T. and McCauliff, Sean and Girouard, Forrest R. and Gunter, Jay P. ...

  71. [71]

    Software and Cyberinfrastructure for Astronomy , year = 2010, editor =

    Jenkins, Jon M. and Chandrasekaran, Hema and McCauliff, Sean D. and Caldwell, Douglas A. and Tenenbaum, Peter and Li, Jie and Klaus, Todd C. and Cote, Miles T. and Middour, Christopher , editor =. Transiting Planet Search in the Kepler Pipeline , booktitle =. 2010 , pages =. doi:10.1117/12.856764 , abstract =

    work page doi:10.1117/12.856764 2010

  72. [72]

    Experimental Astronomy , archivePrefix = "arXiv", eprint =

    Rauer, H. and Catala, C. and Aerts, C. and Appourchaux, T. and Benz, W. and Brandeker, A. and Christensen-Dalsgaard, J. and Deleuil, M. and Gizon, L. and Goupil, M. -J. and G. The PLATO 2.0 mission , url =. Experimental Astronomy , number =. 2014 , bdsk-url-1 =. doi:10.1007/s10686-014-9383-4 , id =

    work page doi:10.1007/s10686-014-9383-4 2014

  73. [73]

    Analytic Light Curves for Planetary Transit Searches , url =

    Mandel, Kaisey and Agol, Eric , doi =. Analytic Light Curves for Planetary Transit Searches , url =. ApJ , month =. 2002 , bdsk-url-1 =

    2002

  74. [74]

    and McElroy, Douglas L

    Christiansen, Jessie L. and McElroy, Douglas L. and Harbut, Marcy and Ciardi, David R. and Crane, Megan and Good, John and Hardegree-Ullman, Kevin K. and Kesseli, Aurora Y. and Lund, Michael B. and Lynn, Meca and Muthiar, Ananda and Nilsson, Ricky and Oluyide, Toba and Papin, Michael and Rivera, Amalia and Swain, Melanie and Susemiehl, Nicholas D. and Tam...

  75. [75]

    , keywords =

    ExoMiner++: Enhanced Transit Classification and a New Vetting Catalog for 2-Minute TESS Data , author =. 2025 , journal =. doi:10.3847/1538-3881/ae03a4 , eprint =

    work page doi:10.3847/1538-3881/ae03a4 2025

  76. [76]

    Martinho, Miguel J. S. and Valizadegan, Hamed and Jenkins, Jon M. and Caldwell, Douglas A. and Twicken, Joseph D. and Tofflemire, Ben and Jafariyazani, Marziye , title =. 2026 , eprint =

    2026

  77. [77]

    and Pichara, Karim and Ishitani Silva, Stela and Araujo, Vladimir , title =

    Salinas, Helem and Brahm, Rafael and Olmschenk, Greg and Barry, Richard K. and Pichara, Karim and Ishitani Silva, Stela and Araujo, Vladimir , title =. MNRAS , volume =. 2025 , doi =

    2025

  78. [78]

    Vivien, H. G. and Deleuil, M. and Jannsen, N. and De Ridder, J. and Seynaeve, D. and Carpine, M.-A. and Zerah, Y. , doi =. PANOPTICON: A novel deep learning model to detect single transit events with no prior data filtering in PLATO light curves , url =. A&A , month =

  79. [79]

    and Mohammed, Salman Arafath and Mohammed, Abdul Wase and Dessalegn, Adis Abebaw and Mittal, Mohit , title =

    Thomas, Bibin and Bhat, Vittal M. and Mohammed, Salman Arafath and Mohammed, Abdul Wase and Dessalegn, Adis Abebaw and Mittal, Mohit , title =. 2025 , eprint =

    2025

  80. [80]

    Choudhary, Anupma and Bandari, Sohith and Kushvah, B. S. and Swastik, C. , doi =. Exoplanet Classification Through Vision Transformers with Temporal Image Analysis , url =. The Astronomical Journal , month =. 2025 , bdsk-url-1 =

    2025

Showing first 80 references.