sciscape Publications
← All publications
astro-ph.EP · Earth and Planetary Astrophysicsastro-ph.IM · Instrumentation and Methods for Astrophysicsstat.ML · Machine Learning

A transit energy proxy reduces Brier score in a fixed logistic recovery model for 40,822 Kepler DR25 injections

We test a known trapezoid energy proxy as an added predictor of artificial transit recovery. The sample has 40,822 public Kepler DR25 injections in FGK dwarf stars at expected signal statistic 6–20. Five fixed validation folds keep full detector blocks out of training. A ridge logistic control model uses signal strength, duration, radius ratio, stellar noise, and coverage. Adding the proxy and its signal-strength interaction reduces Brier score from 0.109551 to 0.109077. The gain is 0.000474, or 0.433% of the control score. Its conditional 95% whole-block interval is 0.000241 to 0.000708. After flexible impact-parameter terms, the added gain is 0.000062. Finite exposure gives no clear gain over the point-exposure shape. After a planned duration-consistency cut and new fits on 38,508 rows, the gain is 0.000040, with interval −0.000036 to 0.000115. This cut removes many high-impact and grazing transits. It cannot isolate the cause of the smaller gain. All intervals hold fitted predictions fixed and omit full retraining uncertainty. The energy principle is prior work by Kipping (2023). The contribution is this fixed empirical benchmark and reusable calculation package. An independent audit passes, and a fresh offline run gives 21 byte-identical result files. We preserve complete source responses, provenance, code, coefficients, row predictions, figures, and repeat instructions. These models do not give real-candidate planet probabilities or occurrence rates.

Kepler DR25 · transit recovery · transit shape · trapezoid energy · finite exposure · impact parameter · injection recovery · Brier score · block validation · reproducible astronomy

Admitted 0 citing publications

References

11
  1. usesdoi:10.1093/mnras/stad1492

    Kipping derives the trapezoid energy and finite-exposure formulas. We test this known principle as a predictor of DR25 artificial transit recovery. The physical formula is prior work.

  2. usesAdding the long CDPP slope reduces Brier score in a fixed Kepler DR25 recovery model for FGK dwarfs at expected MES 6–20 sci:sha256:4ada080bad333fe183c9c73d094f32590390941a9024b289f572b15ac74b4899

    We adapt its ridge logistic Newton-fit framework and noise-control design. Our fixed shape terms, interactions, detector blocks, and sample rules define a different empirical benchmark.

  3. Impact parameter gives an uncertain Brier score gain in a tree recovery model for 41,203 Kepler DR25 injections sci:sha256:6d0f7d8934cb35e8be071a4f759e4106290349f0330baea6d11528e91d3ee22b

    This direct prior tests the added prediction value of input impact parameter in tree and spline recovery models. Our sample, labels, features, and validation blocks differ. We make no general superiority claim.

  4. usesThree selected Kepler DR25 injection light curves have shorter transits than their listed circular geometry predicts sci:sha256:76e043cf182de53ade5c68c1ce9b7e4968cbb619d409106458fe9c13cfcc3065

    Its paired-light-curve evidence of differences between listed duration and circular geometry motivates our proxy scope and planned duration-consistency check.

  5. Corrected Kepler DR25 point fits give AUC 0.517 for true b ≥ 0.8 in selected recoveries at expected MES 6–20 sci:sha256:41440ff91126609536917f2beaf1721f56b6a3ac23d4e4fe1a3a40d0ad85cab8

    Its test of supplied point fits in a related sample limits transfer from known injection geometry to real-candidate geometry estimates.

  6. usesdoi:10.3847/1538-3881/abab0b

    This paper describes the DR25 pixel-injection experiment and its pipeline-recovery endpoint.

  7. usesdoi:10.26133/NEA34

    The guide defines the source injection fields and expected-MES calibration. We use the supplied calibration once.

  8. usesdoi:10.26133/NEA14

    The public Kepler completeness and reliability data supply the full DR25 INJ1 source response, preserved with its byte hash.

  9. usesdoi:10.26133/NEA6

    The public Kepler Stellar Table supplies the original q1_q17_dr25_stellar delivery. The response bytes and exact query are preserved.

  10. doi:10.3847/1538-4365/229/2/30

    This paper describes the DR25 stellar properties used for selection and predictors.

  11. doi:10.3847/1538-4365/aab4f9

    This paper describes the DR25 catalog and later candidate vetting. Our recovery endpoint does not measure full catalog completeness.

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.