pith. sign in

arxiv: 2606.19295 · v1 · pith:NNGJED6Bnew · submitted 2026-06-17 · 🌌 astro-ph.SR

A New Methodology for Classifying Eclipsing Binaries with Kepler Data and Deep Learning

Mousam Mondal , Patricia Cruz , Hugh R. A. Jones , M. C. G\'alvez-Ortiz , John F. Aguilar This is my paper

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

classification 🌌 astro-ph.SR
keywords eclipsing binariesKepler light curveschi-square analysisconvolutional neural networksPHOEBE modelingstellar classificationmagnetic activitystellar variability
0
0 comments X

The pith

Chi-square versus box-size plots from Kepler light curves classify eclipsing binaries at 90 percent accuracy.

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

The paper develops an automated way to sort eclipsing binaries into contact, detached, and semi-detached categories by turning phase-folded Kepler light curves into chi-square versus box-size plots that display class-specific shapes. A simple fit of a polynomial damped sinusoidal function to these plots uses the resulting period as a feature and reaches 86.5 percent accuracy overall. Training a convolutional neural network on the same plots lifts total accuracy to 90 percent, with 47 percent on the hardest semi-detached class; adding PHOEBE-generated simulations further raises contact-versus-detached separation to 99 percent. The work also isolates a subset of systems whose chi-square signatures change across quarters, defines a statistical threshold based on normalized period spread to flag them as temporally varying, and reports four such systems not previously noted, linking the enhanced variability to magnetic activity that appears stronger among cooler stars.

Core claim

Phase-folded Kepler light curves produce chi-square versus box-size plots whose morphologies are distinct for contact, detached, and semi-detached eclipsing binaries; these morphologies are first summarized by the period of a fitted polynomial damped sinusoidal function to achieve 86.5 percent classification accuracy, then supplied as input to a convolutional neural network that reaches 90 percent accuracy overall and 99 percent when distinguishing only contact from detached systems after supplementation with PHOEBE simulations, while a subset of systems exhibit quarter-to-quarter changes in chi-square trends that are isolated by a normalized-spread threshold and attributed to magnetic activ

What carries the argument

The chi-square versus box-size plot, obtained by comparing flux values in phase-folded light curves against median flux, whose class-specific shape serves as the primary input feature for both damped-sinusoid fitting and convolutional neural network classification.

If this is right

  • The CNN reaches 47 percent accuracy on semi-detached systems and 99 percent on the contact-versus-detached distinction after PHOEBE augmentation.
  • Chi-square morphology correlates strongly with orbital period.
  • Cooler late-F, G, K, and M stars show systematically higher chi-square variability than hotter stars.
  • Four previously unreported temporally varying systems are identified whose magnetic activity requires further study.

Where Pith is reading between the lines

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

  • The same chi-square construction could be applied directly to light curves from TESS or other wide-field surveys to classify many more binaries without new training data.
  • The measured rise in variability with decreasing stellar temperature suggests the plots could serve as an automatic flag for magnetic activity across large catalogs.
  • Quarterly changes in the chi-square period may allow time-resolved tracking of starspot or flare evolution using only the existing photometry.

Load-bearing premise

The chi-square versus box-size plots from actual Kepler observations display stable class-specific shapes that PHOEBE simulations can reproduce closely enough to support reliable feature extraction and network training without introducing new biases.

What would settle it

Manual reclassification of several hundred Kepler eclipsing binaries by independent visual inspection or an established catalog yields an overall accuracy below 80 percent when compared against the CNN outputs.

Figures

Figures reproduced from arXiv: 2606.19295 by Hugh R. A. Jones, John F. Aguilar, M. C. G\'alvez-Ortiz, Mousam Mondal, Patricia Cruz.

Figure 1
Figure 1. Figure 1: Box plots illustrated the 𝑃orb values of C (black), SD (orange), and D (blue) EBs. Each box represented the interquartile range (IQR), which encompassed 50 percent of the data. The horizontal yellow line within each box indicated the median 𝑃orb. The whiskers extended to 1.5 times the IQR, and the dots above the whiskers represented outliers — data points that fell outside this range, highlighting extreme … view at source ↗
Figure 2
Figure 2. Figure 2: The chi-square vs. box size plots for nine different EBs are shown. The top row (a, b, c panels) displays three different DEBs with Kepler IDs (KIC): 1026032, 1873513, and 2166200, respectively. The middle row (d, e, f panels) displays three different SDEBs with KIC: 2577756, 5560831, and 5962514, respectively. Lastly, the bottom row (g, h, i panels) displayed three CEBs systems with KIC: 1433410, 2012362,… view at source ↗
Figure 3
Figure 3. Figure 3: Random Forest feature-importance ranking of the six free param￾eters of the PDS function for distinguishing CEBs, DEBs, and SDEBs. The vertical axis lists the six fit parameters: 𝜔—the angular frequency; 𝐶—the constant vertical offset parameter of the fit; 𝜙—the phase angle of the sinu￾soidal component; 𝐵—the exponential damping coefficient; 𝐴—the overall amplitude of the function; and 𝑛—the exponent of th… view at source ↗
Figure 4
Figure 4. Figure 4: Chi-square plot examples of three EB classes: (a) DEB (KIC 8299947); (b) SDEB (KIC 2577756); and (c) CEB (KIC 1433410). The chi-square values are Min–Max normalized to the range [0, 1], and the red curve shows the PDS function fitting 10 14 18 22 26 30 34 38 42 46 50 PDF Threshold Values 0 20 40 60 80 100 Classification Accuracy (a) C ( 10) D ( 10) C ( 50) D ( 50) Binary Class 0 10 20 30 40 50 60 Proportio… view at source ↗
Figure 5
Figure 5. Figure 5: (a) Plot shows how different 𝑃PDS values can separate visually identified CEBs (black) and DEBs (blue). The green dashed line represented the average classification accuracy for both CEBs and DEBs, the red dashed line indicated the accuracy difference between CEBs and DEBs, and the purple vertical line marked the threshold value of 30, which was the best 𝑃PDS value for classification. (b) The x-axis in the… view at source ↗
Figure 6
Figure 6. Figure 6: The confusion matrix of the 1D-CNN model using chi-square plots from Kepler data is shown here. (a) Displays the confusion matrix for all three EB classes: CEB, DEB, and SDEB. (b) Displays the confusion matrix for the binary classification involving only the two major classes: CEB and DEB. 4.2.2 CNN using Kepler and PHOEBE data combined PHOEBE (Prša et al. 2016) is an EB modelling code based on the Wilson–… view at source ↗
Figure 7
Figure 7. Figure 7: PHOEBE mesh, phase-folded light curve, and chi-square plots for D (a, b, c panels), SD (d, e, f panels), and C (g, h, i panels) binaries. All the data were generated from the CALEB catalog. an increased overall accuracy of 98 percent. Out of 2,831 DEBs in the test set, 2,804 were well classified, resulting in an accuracy of 99 percent. For the 2,468 CEBs in the test sample, 2,433 were well classified, with… view at source ↗
Figure 8
Figure 8. Figure 8: The confusion matrix of the 1D-CNN model using chi-square plots from the combined Kepler and PHOEBE data is shown here. Panel (a) shows the confusion matrix for all three eclipsing binary classes: CEB, DEB, and SDEB. Panel (b) shows the confusion matrix for the binary classification involving only the two major classes: CEBs and DEBs. between 0.5 and 0.7. CEBs had 𝑐 values between 0.7 and 0.8, and ellipsoi… view at source ↗
Figure 9
Figure 9. Figure 9: Box plot of 𝑃PDS values and the EB morphological classification. The black, yellow, and blue box plots represent the correctly classified C, SD, and D systems, respectively. The misclassified C, SD, and DEBs are shown with a hatch pattern. The green line inside each box indicates their median value (Q2). Dots beyond the whiskers denote outliers. values. The green lines inside the boxes indicate the median … view at source ↗
Figure 10
Figure 10. Figure 10: Distribution of EB targets in the parameter space defined by 𝑃PDS and 𝑃orb. Panels show the three EB classes separately: (a) DEBs, (b) SDEBs, and (c) CEBs. Targets for which the visual classification agrees with the PDS classification are shown as solid blue markers, while targets where the two classifications disagree are represented by open black circle markers. applied to data with a different cadence,… view at source ↗
Figure 11
Figure 11. Figure 11: A varying SD system, KIC 2577756, and a stable SD system, KIC 10920314, are shown here. Panel (a) illustrates the chi-square plot of quarter 9 for KIC 2577756, and Panel (b) shows its phase-folded LC for the same quarter. Panel (c) presents the chi-square plot of quarter 14 for the same system, while Panel (d) shows the corresponding phase-folded LC. Panel (e) displays the chi-square plot of quarter 9 for… view at source ↗
Figure 12
Figure 12. Figure 12: Distribution of Δ𝑃PDS/𝑃PDS for the test subset. The x-axis repre￾sents Δ𝑃PDS/𝑃PDS, shown with logarithmically spaced bins. The red dashed line marks the median of the distribution, while the black dashed line indi￾cates the median−MAD threshold. The solid blue and green vertical lines correspond to KIC 2577756 (Δ𝑃PDS/𝑃PDS = 1.234) and KIC 7284688 (Δ𝑃PDS/𝑃PDS = 2.612), respectively. in K and M dwarfs (Dave… view at source ↗
Figure 13
Figure 13. Figure 13: Distribution of TV systems with 𝑃PDS on the x-axis and 𝑃orb on the y-axis. The solid red line shows equation (2), while the dashed red line shows equation (3). For each system, quarter-wise 𝑃PDS values were connected with gray lines to indicate the spread across quarters. Visually classified CEBs are shown in black, DEBs in blue, and SDEBs in orange. The minima were shown as hollow circles, while the maxi… view at source ↗
Figure 14
Figure 14. Figure 14: (a) Time-domain Kepler light curve of KIC 2577756, plotted as normalized flux versus BJD, showing strong variability and clear flare signatures consistent with its high Δ𝑃PDS/𝑃PDS value. (b) Time-domain Kepler light curve of KIC 7284688, plotted as normalized flux versus BJD, showing strong variability and varying O’Connell effect due to asynchronous rotation of starspots consistent with its high Δ𝑃PDS/𝑃P… view at source ↗
read the original abstract

We present a new method for the automated classification of eclipsing binaries, into contact, detached, and semi-detached types using Kepler data. Phase-folded light curves are generated and chi-square vs. box size plots are constructed by comparing flux values to the median flux, revealing distinct class patterns. These patterns were first modelled using a polynomial damped sinusoidal function, whose period served as classification feature, achieving an overall accuracy of 86.5 percent. To capture more features and enhance accuracy, we trained a convolutional neural network, which improved the total accuracy to 90 percent, including 47 percent for the challenging semi-detached systems. However, several binaries displayed irregular chi-square signatures. To mitigate this, we incorporated simulated light curves generated with the PHOEBE modelling code, achieving 99 per cent accuracy in distinguishing contact and detached binaries. The resulting chi-square morphologies show a strong correlation with orbital period, and a subset of systems exhibit quarterly variability in their light curves and chi-square trends. We designate these as Temporally Varying systems. By measuring the normalized spread of the chi-square period across quarters, we define a statistical threshold that separates these systems from stable binaries. We reported four Temporally Varying systems not previously noted in the literature with magnetic activity that requires further investigation. Furthermore, cooler stars, namely late-F, G, K, and M types, display systematically higher variability than hotter stars. Cross-matching with catalogues of magnetically active stars indicates that stellar flares and starspots are the most likely causes of this enhanced variability.

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 presents a methodology for automated classification of Kepler eclipsing binaries into contact, detached, and semi-detached types. Phase-folded light curves yield chi-square versus box-size plots whose morphologies are first fit by a polynomial damped sinusoidal function (using the period as a feature for 86.5% accuracy), then fed to a CNN (raising overall accuracy to 90%, with 47% on semi-detached systems). PHOEBE-generated simulations are added to reach 99% accuracy on contact versus detached binaries. A normalized-spread threshold on quarterly chi-square periods flags Temporally Varying systems; four previously unreported examples are identified and linked to magnetic activity, with cooler stars showing higher variability.

Significance. If the reported accuracies are shown to be robust on held-out real Kepler data without simulation augmentation and if the PHOEBE light curves are demonstrated to reproduce the relevant noise and systematics properties, the pipeline could supply a scalable classification tool for future photometric surveys. The identification of Temporally Varying systems and the reported correlation between variability and stellar effective temperature would constitute an incremental observational result worthy of follow-up.

major comments (3)
  1. [Abstract / CNN results] Abstract and results section on CNN training: the jump from 90% to 99% accuracy on contact/detached binaries after adding PHOEBE simulations is load-bearing for the headline performance claim, yet no quantitative comparison of simulated versus real Kepler noise properties (e.g., correlated noise, quarter-to-quarter systematics, or flare statistics) is provided, nor is an explicit test reported on a held-out set of real data only.
  2. [Abstract] Abstract: the 47% accuracy reported for semi-detached systems is low enough to indicate that the assumed class-specific morphologies in the chi-square versus box-size plots are not stable for this category; this directly weakens the claim that the method provides reliable classification across all three types.
  3. [Temporally Varying systems section] Section describing the Temporally Varying threshold: the normalized-spread threshold used to flag the four new systems is defined from the same chi-square period measurements whose stability is questioned by the simulation-augmentation step; without an independent validation set or cross-check against known active binaries, the threshold risks being tuned to simulation artifacts.
minor comments (2)
  1. [Abstract] Abstract: accuracy figures are given without error bars, cross-validation details, or the size of the training/test splits.
  2. [Methods] The manuscript should clarify whether the CNN was trained and tested on real data alone before the PHOEBE augmentation step, and whether any overlap exists between the simulation training set and the real test objects.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We respond point by point to the major comments below, indicating where revisions are planned.

read point-by-point responses

  1. Referee: [Abstract / CNN results] Abstract and results section on CNN training: the jump from 90% to 99% accuracy on contact/detached binaries after adding PHOEBE simulations is load-bearing for the headline performance claim, yet no quantitative comparison of simulated versus real Kepler noise properties (e.g., correlated noise, quarter-to-quarter systematics, or flare statistics) is provided, nor is an explicit test reported on a held-out set of real data only.

    Authors: We agree that a quantitative comparison of noise properties between the PHOEBE simulations and real Kepler data was not provided. In revision we will add such a comparison (e.g., via residual distributions and power spectra) and will also report CNN performance on a held-out real Kepler subset without any simulation augmentation. revision: yes

  2. Referee: [Abstract] Abstract: the 47% accuracy reported for semi-detached systems is low enough to indicate that the assumed class-specific morphologies in the chi-square versus box-size plots are not stable for this category; this directly weakens the claim that the method provides reliable classification across all three types.

    Authors: The lower accuracy for semi-detached systems is already noted in the manuscript as reflecting their morphological variability. We will revise the abstract to report per-class accuracies explicitly and qualify the overall claim to reflect stronger performance on contact and detached systems. revision: yes

  3. Referee: [Temporally Varying systems section] Section describing the Temporally Varying threshold: the normalized-spread threshold used to flag the four new systems is defined from the same chi-square period measurements whose stability is questioned by the simulation-augmentation step; without an independent validation set or cross-check against known active binaries, the threshold risks being tuned to simulation artifacts.

    Authors: The PHOEBE simulations were used exclusively to augment CNN training for contact/detached separation and played no role in computing the chi-square periods or the normalized-spread threshold on real Kepler quarters. The four systems and the temperature correlation were identified in real data, supported by cross-matching to active-star catalogs. We will clarify this separation in the revised text. revision: partial

Circularity Check

0 steps flagged

Classification pipeline is self-contained with no circular reductions

full rationale

The paper constructs chi-square vs. box-size plots from phase-folded Kepler light curves, fits a damped sinusoidal function to extract a period feature (yielding 86.5% accuracy), trains a CNN on these plus PHOEBE simulations (to 90% overall, 99% contact/detached), and defines a variability threshold via normalized spread of the chi-square period across quarters. These steps rely on external Kepler observations, standard fitting, external simulation code, and empirical accuracy measurements. No equations equate a reported result to its own fitted inputs by construction, and no self-citations are invoked as load-bearing uniqueness theorems. The derivation chain is independent of the target classifications.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 1 invented entities

The method depends on standard assumptions about Kepler photometry quality and the fidelity of PHOEBE simulations; it introduces one new designation (Temporally Varying systems) whose definition rests on a statistical threshold derived from the same data.

free parameters (2)
  • polynomial damped sinusoidal function coefficients
    Fitted to chi-square plots to extract the period feature used for initial classification.
  • normalized spread threshold for Temporally Varying label
    Statistical cutoff chosen to separate variable from stable systems.
axioms (2)
  • domain assumption Phase-folded Kepler light curves yield reliable median fluxes and chi-square deviations that reflect binary morphology.
    Invoked when constructing the chi-square versus box-size plots from observed data.
  • domain assumption PHOEBE-generated light curves reproduce the statistical properties of real Kepler eclipsing-binary observations sufficiently for training augmentation.
    Used to justify the accuracy gain from 90 percent to 99 percent.
invented entities (1)
  • Temporally Varying systems no independent evidence
    purpose: Label for binaries showing quarterly changes in chi-square morphology attributed to magnetic activity.
    New category defined by normalized spread of chi-square period across quarters; no independent confirmation outside the Kepler dataset is provided.

pith-pipeline@v0.9.1-grok · 5835 in / 1744 out tokens · 30900 ms · 2026-06-26T19:17:56.271879+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

46 extracted references · 26 canonical work pages · 12 internal anchors

  1. [1]

    , title =

    Andersen, J. , title =. A&ARv , year =. doi:10.1007/BF00873538 , url =

    work page doi:10.1007/bf00873538

  2. [2]

    2010, A&A Rv, 18, 67, doi: 10.1007/s00159-009-0025-1 van der Walt, S., Colbert, S

    Torres, G. and Andersen, J. and Gim. Accurate masses and radii of normal stars: modern results and applications , journal =. 2010 , volume =. doi:10.1007/s00159-009-0025-1 , url =

    work page doi:10.1007/s00159-009-0025-1 2010

  3. [3]

    Revisiting Physical Parameters of the Benchmark Brown Dwarf LHS 6343 C through a Hubble Space Telescope/WFC3 Secondary-eclipse Observation , journal =

    Frost, William and Albert, Lo. Revisiting Physical Parameters of the Benchmark Brown Dwarf LHS 6343 C through a Hubble Space Telescope/WFC3 Secondary-eclipse Observation , journal =. 2024 , month = sep, volume =

    2024

  4. [4]

    and Milone, E

    Kallrath, J. and Milone, E. F. , title =. 2009 , doi =

    2009

  5. [5]

    Sarro, L. M. and S. Automatic classification of eclipsing binaries light curves using neural networks , journal =. 2006 , month =. doi:10.1051/0004-6361:20052830 , archivePrefix =. astro-ph/0511346 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361:20052830 2006

  6. [6]

    Kepler Eclipsing Binary Stars. III. Classification of Kepler Eclipsing Binary Light Curves with Locally Linear Embedding

    Matijevi. Kepler Eclipsing Binary Stars. III. Classification of Kepler Eclipsing Binary Light Curves with Locally Linear Embedding , journal =. 2012 , month = may, volume =. doi:10.1088/0004-6256/143/5/123 , archivePrefix=. 1204.2113 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-6256/143/5/123 2012

  7. [7]

    Slawson, Robert W. and Pr. Kepler Eclipsing Binary Stars. II. 2165 Eclipsing Binaries in the Second Data Release , journal =. 2011 , month = nov, volume =. doi:10.1088/0004-6256/142/5/160 , archivePrefix=. 1103.1659 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1088/0004-6256/142/5/160 2011

  8. [8]

    and Saul, Lawrence K

    Roweis, Sam T. and Saul, Lawrence K. , title =. Science , volume =. 2000 , doi =

    2000

  9. [9]

    and Duin, R

    de Ridder, S. and Duin, R. , title =

  10. [10]

    2021 , month =

    Automatic classification of eclipsing binary stars using deep learning methods , journal =. 2021 , month =. doi:10.1016/j.ascom.2021.100488 , archivePrefix =. 2108.01640 , primaryClass =

    work page doi:10.1016/j.ascom.2021.100488 2021

  11. [11]

    Daza-Perilla, I. V. and Gramajo, L. V. and Lares, M. and Palma, T. and Ferreira Lopes, C. E. and Minniti, D. and Clari\'. Automated classification of eclipsing binary systems in the VVV Survey , journal =. 2023 , month =. doi:10.1093/mnras/stad141 , archivePrefix =. 2302.01200 , primaryClass =

    work page doi:10.1093/mnras/stad141 2023

  12. [12]

    and Lucas, P

    Minniti, D. and Lucas, P. W. and Emerson, J. P. and Saito, R. K. and Hempel, M. and Pietrukowicz, P. and Ahumada, A. V. and Alonso, M. V. and Alonso-Garcia, J. and Arias, J. I. and et al. , title =. New Astron. , year =

  13. [13]

    Saito, R. K. and Hempel, M. and Minniti, D. and Lucas, P. W. and Rejkuba, M. and Toledo, I. and Gonzalez, O. A. and Alonso-Garcia, J. and Irwin, M. J. and Gonzalez-Solares, E. and et al. , title =. A&A , year =. doi:10.1051/0004-6361/201118407 , archivePrefix =. 1111.5511 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/201118407

  14. [14]

    VISTA Variables in the V\'ia L\'actea (VVV): Halfway Status and Results

    Hempel, M. and Minniti, D. and D\'. VISTA Variables in the V\'. Messenger , year =. doi:10.48550/arXiv.1406.3241 , archivePrefix =. 1406.3241 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1406.3241

  15. [15]

    A new near-IR window of low extinction in the Galactic plane

    Minniti, D. and Saito, R. K. and Gonzalez, O. A. and Alonso-Garc\'. A new near-IR window of low extinction in the Galactic plane , journal =. 2018 , month =. doi:10.1051/0004-6361/201732099 , archivePrefix =. 1804.07785 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/201732099 2018

  16. [16]

    AJ , year =

    Ding, Xu and Ji, KaiFan and Cheng, QiYuan and Song, ZhiMing and Wang, JinLiang and Tian, XueFen and Wang, ChuanJun , title =. AJ , year =. doi:10.3847/1538-3881/adb846 , archivePrefix=. 2504.14612 , primaryClass =

    work page doi:10.3847/1538-3881/adb846

  17. [17]

    and Aguilar, J

    Cruz, P. and Aguilar, J. F. and Garrido, H. E. and Diaz, M. P. and Solano, E. , title =. MNRAS , year =. doi:10.1093/mnras/stac1707 , archivePrefix =. 2206.08708 , primaryClass =

    work page doi:10.1093/mnras/stac1707

  18. [18]

    Kepler Eclipsing Binary Stars. VII. The Catalog of Eclipsing Binaries Found in the Entire Kepler Data-Set

    Kirk, B. and Conroy, K. and Pr. Kepler Eclipsing Binary Stars. VII. The Catalog of Eclipsing Binaries Found in the Entire Kepler Data Set , journal =. 2016 , month =. doi:10.3847/0004-6256/151/3/68 , archivePrefix =. 1512.08830 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/0004-6256/151/3/68 2016

  19. [19]

    Morris, R. L. and Twicken, J. D. and Smith, J. C. and Clarke, B. D. and Jenkins, J. M. and Bryson, S. T. and Girouard, F. and Klaus, T. C. , title =. 2020 , editor =

    2020

  20. [20]

    and Christiansen, Jessie L

    Van Cleve, Jeffrey E. and Christiansen, Jessie L. and Jenkins, Jon M. and Caldwell, Douglas A. and Barclay, Thomas and Bryson, Stephen T. and Burke, Christopher J. and Cambell, Jennifer and Catanzarite, Joseph and Clarke, Bruce D. and et al. , title =. 2016 , month = dec, pages =

    2016

  21. [21]

    P., Habergham, S

    Murphy, Simon J. , title =. MNRAS , year =. doi:10.1111/j.1365-2966.2012.20644.x , url =

    work page doi:10.1111/j.1365-2966.2012.20644.x 2012

  22. [22]

    and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and van der Walt, St

    Virtanen, Pauli and Gommers, Ralf and Oliphant, Travis E. and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and van der Walt, St. Nature Methods , volume =. 2020 , doi =

    2020

  23. [23]

    Machine learning , volume=

    Random forests , author=. Machine learning , volume=. 2001 , publisher=

    2001

  24. [24]

    Scikit-learn: Machine Learning in Python , journal =

    Pedregosa, Fabian and Varoquaux, Ga. Scikit-learn: Machine Learning in Python , journal =. 2011 , volume =

    2011

  25. [25]

    , keywords =

    Evaluating Time-series Augmentation Techniques for Deep Learning Based Solar Flare Prediction. , keywords =. doi:10.3847/1538-4365/adfa2a , adsurl =

    work page doi:10.3847/1538-4365/adfa2a

  26. [26]

    and Ba, Jimmy , title =

    Kingma, Diederik P. and Ba, Jimmy , title =. arXiv e-prints , year =. 1412.6980 , archivePrefix=

    Pith/arXiv arXiv

  27. [27]

    Focal Loss for Dense Object Detection , journal =

    Lin, Tsung-Yi and Goyal, Priya and Girshick, Ross and He, Kaiming and Doll. Focal Loss for Dense Object Detection , journal =. 2020 , volume =

    2020

  28. [28]

    Prsa, Andrej and Matijevic, Gal and Latkovic, Olivera and Vilardell, Francesc and Wils, Patrick , title =

  29. [29]

    Physics of Eclipsing Binaries

    Pr. Physics of Eclipsing Binaries. II. Toward the Increased Model Fidelity , journal =. 2016 , volume =

    2016

  30. [30]

    and Devinney, Edward J

    Wilson, Robert E. and Devinney, Edward J. , title =. ApJ , volume =

  31. [31]

    Bradstreet, D. H. , title =

  32. [32]

    Scikit-learn: Machine Learning in Python , year =

    Pedregosa, Fabian and Varoquaux, Ga. Scikit-learn: Machine Learning in Python , year =. 1201.0490 , archivePrefix=

    Pith/arXiv arXiv

  33. [33]

    Turkish Journal of Astronomy and Astrophysics , year=

    A Deep Learning Neural Network Algorithm for Classification of Eclipsing Binary Light Curves , author=. Turkish Journal of Astronomy and Astrophysics , year=

  34. [34]

    , title =

    MacQueen, J. , title =. Proceedings of the 5th Berkeley Symposium on Mathematical Statistics and Probability, Volume 1 , editor =. 1967 , publisher =

    1967

  35. [35]

    and Caballero-Nieves, Saida M

    Knote, Matthew F. and Caballero-Nieves, Saida M. and Gokhale, Vayujeet and Johnston, Kyle B. and Perlman, Eric S. , title =. ApJS , year =. doi:10.3847/1538-4365/ac770f , url =

    work page doi:10.3847/1538-4365/ac770f

  36. [36]

    Weibull, Waloddi , title =. J. Appl. Mech. , year =

  37. [37]

    Davenport, James R. A. , title =. ApJ , year =. doi:10.3847/0004-637X/829/1/23 , archivePrefix=. 1607.03494 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/0004-637x/829/1/23

  38. [38]

    Tidal Synchronization and Differential Rotation of Kepler Eclipsing Binaries

    Lurie, John C. and Vyhmeister, Karl and Hawley, Suzanne L. and Adilia, Jamel and Chen, Andrea and Davenport, James R. A. and Juri. Tidal Synchronization and Differential Rotation of Kepler Eclipsing Binaries , journal =. 2017 , month = dec, volume =. doi:10.3847/1538-3881/aa974d , archivePrefix=. 1710.07339 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/1538-3881/aa974d 2017

  39. [39]

    Global stellar variability study in the field-of-view of the Kepler satellite

    Debosscher, J. and Blomme, J. and Aerts, C. and De Ridder, J. , title =. A&A , year =. doi:10.1051/0004-6361/201015647 , archivePrefix=. 1102.2319 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1051/0004-6361/201015647

  40. [40]

    Tracking the Stellar Longitudes of Starspots in Short-Period Kepler Binaries

    Balaji, Bhaskaran and Croll, Bryce and Levine, Alan M. and Rappaport, Saul , title =. MNRAS , year =. doi:10.1093/mnras/stv031 , archivePrefix=. 1412.8101 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.1093/mnras/stv031

  41. [41]

    White-Light Flares on Close Binaries Observed with Kepler

    Gao, Qing and Xin, Yu and Liu, Ji-Feng and Zhang, Xiao-Bin and Gao, Shuang , title =. ApJS , year =. doi:10.3847/0067-0049/224/2/37 , archivePrefix=. 1602.07972 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.3847/0067-0049/224/2/37

  42. [42]

    Armstrong, D. J. and G. A catalogue of temperatures for Kepler eclipsing binary stars , journal =. 2013 , month = dec, volume =. doi:10.1093/mnras/stt2146 , url =

    work page doi:10.1093/mnras/stt2146 2013

  43. [43]

    , title =

    Strassmeier, Klaus G. , title =. Astron. Astrophys. Rev. , year =. doi:10.1007/s00159-009-0020-6 , adsurl =

    work page doi:10.1007/s00159-009-0020-6

  44. [44]

    Vaiana, G. S. and Cassinelli, J. P. and Fabbiano, G. and Giacconi, R. and Golub, L. and Gorenstein, P. and Haisch, B. M. and Harnden, Jr., F. R. and Johnson, H. M. and Linsky, J. L. and Maxson, C. W. and Mewe, R. and Rosner, R. and Seward, F. and Topka, K. and Zwaan, C. , title =. ApJ , year =. doi:10.1086/158797 , adsurl =

    work page doi:10.1086/158797

  45. [45]

    J., Drake, J

    Wright, Nicholas J. and Drake, Jeremy J. and Mamajek, Eric E. and Henry, Gregory W. , title =. ApJ , year =. doi:10.1088/0004-637X/743/1/48 , url =

    work page doi:10.1088/0004-637x/743/1/48

  46. [46]

    AJ , year =

    Pan, Yang and Zhang, Xiaobin , title =. AJ , year =. doi:10.3847/1538-3881/accfa1 , url =

    work page doi:10.3847/1538-3881/accfa1