https://doi.org/10.1140/epjs/s11734-025-02033-9
Regular Article
Characterizing non-transiting exoplanets through transit timing variations using deep learning
Department of Astronomy and Astrophysics, National Observatory, Rua General José Cristino 77, 20921-400, Rio de Janeiro, RJ, Brazil
a
This email address is being protected from spambots. You need JavaScript enabled to view it.
b
This email address is being protected from spambots. You need JavaScript enabled to view it.
Received:
29
April
2025
Accepted:
17
October
2025
Published online:
29
October
2025
Abstract
Using transit timing variations (TTV) to characterize the mass of planets in extrasolar systems has proven to be an efficient and consolidated technique. The usual approach involves to fit the TTVs time series to a predefined model of the system. This has been successfully applied, in particular, to estimate the mass of a non-transiting companion in a two-planet system. Here, we propose an alternative approach, consisting of using a deep convolutional neural network to classify the mass and period of such companion. We trained the popular residual network ResNet50, considering different architectures, to recognize images that encode the light curve of the transiting planet in the form of riverplots. We discuss the impact on the predictive capacity of the network by the inclusion of transit effects, like limb darkening and asteroseismic noise. We also discuss the impact of using different numbers of training classes in mass and period, either balanced or unbalanced. We present results of period and mass classification for a ResNet50 trained with about 500,000 synthetic riverplots. We obtained test accuracies of the order of 40% for mass and 60% for period classification, when applied to synthetic riverplots, and nearly 40% for mass and 25% for period classification, when applied to real Kepler planets riverplots.
Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
© The Author(s), under exclusive licence to EDP Sciences, Springer-Verlag GmbH Germany, part of Springer Nature 2025
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
