Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Panopticon: a novel deep learning model to detect single transit events with no prior data filtering in PLATO light curves

H. G. Vivien, M. Deleuil, N. Jannsen, J. De Ridder, D. Seynaeve, M. -A. Carpine, Y. Zerah

TL;DR

Panopticon tackles the challenge of detecting single transit events in PLATO light curves without prior detrending, which can bias long or shallow transits. It uses a 1D Unet-based architecture to output a per-point probability map, enabling localization of transit events and extraction of their epochs and durations directly from unfiltered data. On a pixel-level simulated PLATO dataset with planetary and false-positive signals plus realistic noise, it achieves recovery around 90% at a 1% false-alarm rate (including Earth-analogs with >25% recovery) and >85% at FAR <0.01%, with transits deeper than ~180 ppm essentially guaranteed; inference is fast (~0.2 s per light curve) and training requires only a few hours per model. The approach offers a complementary, efficient path to identify transit candidates for follow-up, particularly benefiting long-period planet detection via mono-transits and reducing dependence on pre-processing steps.

Abstract

To prepare for the analyses of the future PLATO light curves, we develop a deep learning model, Panopticon, to detect transits in high precision photometric light curves. Since PLATO's main objective is the detection of temperate Earth-size planets around solar-type stars, the code is designed to detect individual transit events. The filtering step, required by conventional detection methods, can affect the transit, which could be an issue for long and shallow transits. To protect transit shape and depth, the code is also designed to work on unfiltered light curves. We trained the model on a set of simulated PLATO light curves in which we injected, at pixel level, either planetary, eclipsing binary, or background eclipsing binary signals. We also include a variety of noises in our data, such as granulation, stellar spots or cosmic rays. The approach is able to recover 90% of our test population, including more than 25% of the Earth-analogs, even in the unfiltered light curves. The model also recovers the transits irrespective of the orbital period, and is able to retrieve transits on a unique event basis. These figures are obtained when accepting a false alarm rate of 1%. When keeping the false alarm rate low (<0.01%), it is still able to recover more than 85% of the transit signals. Any transit deeper than 180ppm is essentially guaranteed to be recovered. This method is able to recover transits on a unique event basis, and does so with a low false alarm rate. Thanks to light curves being one-dimensional, model training is fast, on the order of a few hours per model. This speed in training and inference, coupled to the recovery effectiveness and precision of the model make it an ideal tool to complement, or be used ahead of, classical approaches.

Panopticon: a novel deep learning model to detect single transit events with no prior data filtering in PLATO light curves

TL;DR

Panopticon tackles the challenge of detecting single transit events in PLATO light curves without prior detrending, which can bias long or shallow transits. It uses a 1D Unet-based architecture to output a per-point probability map, enabling localization of transit events and extraction of their epochs and durations directly from unfiltered data. On a pixel-level simulated PLATO dataset with planetary and false-positive signals plus realistic noise, it achieves recovery around 90% at a 1% false-alarm rate (including Earth-analogs with >25% recovery) and >85% at FAR <0.01%, with transits deeper than ~180 ppm essentially guaranteed; inference is fast (~0.2 s per light curve) and training requires only a few hours per model. The approach offers a complementary, efficient path to identify transit candidates for follow-up, particularly benefiting long-period planet detection via mono-transits and reducing dependence on pre-processing steps.

Abstract

To prepare for the analyses of the future PLATO light curves, we develop a deep learning model, Panopticon, to detect transits in high precision photometric light curves. Since PLATO's main objective is the detection of temperate Earth-size planets around solar-type stars, the code is designed to detect individual transit events. The filtering step, required by conventional detection methods, can affect the transit, which could be an issue for long and shallow transits. To protect transit shape and depth, the code is also designed to work on unfiltered light curves. We trained the model on a set of simulated PLATO light curves in which we injected, at pixel level, either planetary, eclipsing binary, or background eclipsing binary signals. We also include a variety of noises in our data, such as granulation, stellar spots or cosmic rays. The approach is able to recover 90% of our test population, including more than 25% of the Earth-analogs, even in the unfiltered light curves. The model also recovers the transits irrespective of the orbital period, and is able to retrieve transits on a unique event basis. These figures are obtained when accepting a false alarm rate of 1%. When keeping the false alarm rate low (<0.01%), it is still able to recover more than 85% of the transit signals. Any transit deeper than 180ppm is essentially guaranteed to be recovered. This method is able to recover transits on a unique event basis, and does so with a low false alarm rate. Thanks to light curves being one-dimensional, model training is fast, on the order of a few hours per model. This speed in training and inference, coupled to the recovery effectiveness and precision of the model make it an ideal tool to complement, or be used ahead of, classical approaches.
Paper Structure (8 sections, 4 equations, 9 figures, 3 tables)

This paper contains 8 sections, 4 equations, 9 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Deep learning model
  3. Architectures
  4. Implementation
  5. Dataset and environment
  6. Performances
  7. Discussion
  8. Model architectures

Figures (9)

  • Figure 1: Theoretical input-output scheme of the model. A light curve, normalized between 0 and 1, is given as input to the model (top panel). We highlight the transit by the blue region. The model returns a classification map for the whole light curve (bottom panel).
  • Figure 2: Basic convolution block used in the model. For Unet and Unet++ each node is made up of two consecutive occurrence of this block, but a single one is used for the nodes of Unet3+. The dropout layer is optional, and if used is applied only once per block.
  • Figure 3: Histograms of the radii and periods of the bodies in our dataset. The left, center and right columns correspond respectively to the planets, the eclipsing binaries, and the background eclipsing binaries. The tree spikes in the planetary population correspond to an erroneous simulation run that didn't include the sampling of radii around the central values of the distribution.
  • Figure 4: The recovery capabilities of model C. Each panel shows a physical characteristic of our test population, in purple, as well as the fraction recovered by the model, in green. Panel c highlights the ability of the model to recover transits similar to that of Earth. Additionally, panel e shows that the ability to detect transits is not linked to the orbital period, and single transit are therefore detected.
  • Figure 5: Recovery and FAR for the best performing models. The top panel shows the receiver operating characteristic curve, that links the recovery percentage to its associated FAR. We show the epoch corresponding to each selected models, and mark the recovery/FAR balance highlighted in Table \ref{['tab:models_tested']}. The bottom panel shows model B, binned per transit depth. By discretizing the recovery, we can evaluate the performances of the model more thoroughly.
  • ...and 4 more figures