Scientists Use A New Deep Learning Method To Add 301 Planets to Kepler’s Total Count


Deep neural networks are machine learning systems that automatically learn a task if provided with necessary data.  An artificial neural network (ANN) having numerous layers between the input and output layers is known as a deep neural network (DNN). Neural networks are made available in various shapes and sizes. However, they all include the same essential components: neurons, synapses, weights, biases, and functions.

Recently, scientists have added a total of 301 validated exoplanets to the already existing exoplanet tally. The cluster of planets is the most recent addition to the 4,569 confirmed planets orbiting various faraway stars. This news has gotten everyone into pondering how possibly scientists could discover such a huge, all at a time. The answer lies in a Deep Neural Network called ExoMiner. The Kepler Science Office upgraded all 301 machine-validated planets to planet candidate status after discovering them via the Kepler Science Operations Center pipeline. None of this was possible before the implementation of ExoMiner.

ExoMiner is a recent development in deep neural networks that uses NASA’s Pleiades supercomputer to detect genuine exoplanets from various forms of “false positives.” Its design is based on the tests and qualities that human scientists employ to confirm the discovery of new exoplanets. It also learns from previously verified exoplanets as well as false-positive cases.

Combing through the data thoroughly and deciphering what isn’t an actual planet forms the basis of ExoMiner. There are thousands of stars in the range of view of projects like NASA’s Kepler and its follow-on mission, K2, each with the potential to host many possible exoplanets. Poring over enormous datasets is a time-consuming process. ExoMiner, on the other hand, deals with this as well. Exoplanet-detection machine learning programs, in general, are like black boxes; no one knows how they decide whether something is a planet or not. However, the scientists provided precise information about aspects in the data that led the neural network to validate or reject a planet while using ExoMiner.

ExoMiner discovered the 301 planets using data from the remaining set of potential planets – or candidates – in the Kepler Archive, according to a research article published in the Astrophysical Journal. The paper also reveals how ExoMiner is more precise and reliable in ruling out false positives and displaying the actual signs of planets orbiting their parent stars. It achieves all of this while allowing scientists to understand precisely how ExoMiner arrived at its conclusions.

In addition to confirming 301 planets, the research team has pointed out that none resemble earth. However, the newly discovered planets share similar characteristics to the overall population of confirmed exoplanets in our galactic neighborhood.


ExoMiner has also been trained with Kepler data. The knowledge gained can be applied to other missions, including TESS, the team’s current endeavor.

ExoMiner will have a chance to prove its mettle when missions like NASA’s Transiting Exoplanet Survey Satellite, or TESS, and the European Space Agency’s planned PLAnetary Transits and Oscillations of Stars, or PLATO, that use transit photometry open doors to more opportunities.