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-the-art 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 which 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.

• The paper introduces a CNN that classifies Kepler transit signals with 98.8% accuracy in distinguishing exoplanets from false positives.
• It reports the discovery of a five-planet resonant chain in Kepler-80 and an eighth planet in Kepler-90, matching our Solar System’s count.
• The study demonstrates the scalability of deep learning for exoplanet detection, paving the way for future analyses of large astronomical datasets.

Identifying Exoplanets with Deep Learning: A Methodological Exploration

The identification of exoplanets has historically been a challenging task, requiring the meticulous sifting of massive datasets for signals of potential planets. In "Identifying Exoplanets with Deep Learning: A Five Planet Resonant Chain Around Kepler-80 and an Eighth Planet Around Kepler-90," Christopher J. Shallue and Andrew Vanderburg propose a robust methodology leveraging deep learning to improve the accuracy and efficiency of detecting exoplanets from data acquired by NASA's Kepler Space Telescope.

The Kepler telescope was originally designed to explore the prevalence of Earth-sized planets around Sun-like stars. However, detecting these planets is complex due to their small size and the noise present in observational data. The authors propose an automated approach that applies deep learning techniques to the classification of transit signals, distinguishing genuine planets from false positives induced by astrophysical or instrumental factors.

The core of the proposed method is the application of a deep convolutional neural network (CNN) trained using human-classified transit signals from Kepler data. This model is remarkable for its capacity to adaptively learn representations from raw light curves, automatically identifying distinguishing features between transit planet signals and false positives. Impressively, in the authors' test set, the CNN correctly ranks plausible planet signals higher than false positives 98.8% of the time, demonstrating the model's discriminative power.

As part of their analysis, Shallue and Vanderburg validate two new planetary discoveries. The first is a planet Kepler-80g, which belongs to an intricate five-planet resonant chain with orbits satisfying specific Laplacian relationships, affirming the resonance predictions. The second discovery, Kepler-90i, defines an eighth planet in the Kepler-90 system, which now matches our Solar System’s planet count and underscores the high-multiplicity nature of such systems.

The implications of this work span both theoretical and practical domains. Theoretically, it reaffirms the utility of deep learning in astronomical applications where model interpretability is often traded for performance. Practically, the advance suggests a scalable approach to analyze existing and future datasets from missions like the Transiting Exoplanet Survey Satellite (TESS) and the James Webb Space Telescope.

Future enhancements to this system could involve incorporating additional sources of astrophysical information into the CNN's learning process, such as stellar position metrics or auxiliary observational features that might better account for background noise and various error sources. Additionally, further incorporating simulated data into the training process may bolster model robustness in data-poor regimes.

While the research does not address the full extent of complexities involved with exoplanet discovery, particularly regarding weak signals and diverse noise characteristics, it sets a solid precedent for integrating advanced machine learning methods in the field of astronomy. The use of deep learning in this manner may herald a new era of planet discovery, maximizing the yield of astronomical datasets and enhancing our understanding of planetary systems.