Identifying Exoplanets with Deep Learning: A Five-planet Resonant Chain around Kepler-80 and an Eighth Planet around Kepler-90

Abstract

NASA's Kepler Space Telescope was designed to determine the frequency of Earth-sized planets orbiting Sun-like stars, but these planets are on the very edge of the mission's detection sensitivity. Accurately determining the occurrence rate of these planets will require automatically and accurately assessing the likelihood that individual candidates are indeed planets, even at low signal-to-noise ratios. We present a method for classifying potential planet signals using deep learning, a class of machine learning algorithms that have recently become state-of-theart in a wide variety of tasks. We train a deep convolutional neural network to predict whether a given signal is a transiting exoplanet or a false positive caused by astrophysical or instrumental phenomena. Our model is highly effective at ranking individual candidates by the likelihood that they are indeed planets: 98.8% of the time it ranks plausible planet signals higher than false-positive signals in our test set. We apply our model to a new set of candidate signals that we identified in a search of known Kepler multi-planet systems. We statistically validate two new planets that are identified with high confidence by our model. One of these planets is part of a five-planet resonant chain around Kepler-80, with an orbital period closely matching the prediction by three-body Laplace relations. The other planet orbits Kepler-90, a star that was previously known to host seven transiting planets. Our discovery of an eighth planet brings Kepler-90 into a tie with our Sun as the star known to host the most planets.

Searching the Entirety of Kepler Data. I. 17 New Planet Candidates Including One Habitable Zone World

“…So far, independent Kepler searches have focused on only a subset of light-curves (e.g., Kunimoto et al 2018;Shallue & Vanderburg 2018) or a more limited range of orbital periods than examined by the Kepler team (e.g., Wang et al 2015;Caceres et al 2019). In this paper, we present an independent systematic search of the entirety of Kepler data (∼ 200, 000 stars) for planets, using the same three-transit minimum detection criteria as the Kepler team.…”

“…S/N can often be underestimated, such as through the assumption of a box-shaped transit or the distortion of the transit shape from the detrending algorithm, causing a planet TC to be erroneously rejected. For these reasons, recent searches have begun to relax the noise floor, such as Shallue & Vanderburg (2018) in which a BLS algorithm was used to search as low as S/N = 5, and Kunimoto et al (2018), in which members of our team searched down to S/N = 6.…”

“…Even following the release of Kepler catalogues based on all four years of Kepler data (Coughlin et al 2016; Thompson et al 2018), independent authors have shown that there are still scientifically interesting candidates to be found. Shallue & Vanderburg (2018) represents the first application of machine learning to searching Kepler data for exoplanets, finding two new planets among a subset of Q1−Q17 KOIs including an eighth planet in a planetary system. Kunimoto et al (2018) also searched a subset of KOIs, finding four new candidates including a Neptune-sized candidate in the Habitable Zone.…”

Fourier series for eclipses on exoplanet binaries

“…The discovery of extensive multi-planet systems around other stars (Lovis et al 2011;Gillon et al 2017;Shallue & Vanderburg 2018) shows that the Solar System is not unique. Therefore, we may expect exoplanets to also have their own satellites, like the Solar-System planets; Mercury and Venus being exceptions (Namouni 2010;Ogihara & Ida 2012;Barr 2016).…”

Surveying the reach and maturity of machine learning and artificial intelligence in astronomy

“…Recent applications in astronomy utilizing ML and AI include: the discovery of extrasolar planets (Pearson, Palafox, & Griffith, 2018;Shallue & Vanderburg, 2018) and gravitationally lensed systems (Jacobs, Glazebrook, Collett, More, & McCarthy, 2017;Lanusse et al, 2018;Pourrahmani, Nayyeri, & Cooray, 2018); discovery and classification of transient objects (Connor & van Leeuwen, 2018;Farah et al, 2018;Mahabal et al, 2019); forecasting solar activity (Florios et al, 2018;Inceoglu et al, 2018;Nishizuka et al, 2017); assignment of photometric redshifts within large-scale galaxy surveys (Bilicki et al, 2018;Ruiz, Corral, Mountrichas, & Georgantopoulos, 2018;Speagle & Eisenstein, 2017); and the classification of gravitational wave signals and instrumental noise (George & Huerta, 2018a, 2018bPowell et al, 2017).…”