Data - Apps - AI - Cybernetics

Discovery of 69 New Exoplanets Using Machine Learning

By Keith Cowing
Press Release
May 23, 2023
Filed under , , , , , ,
Discovery of 69 New Exoplanets Using Machine Learning
Planet radius vs. orbital period for confirmed tran- siting planets and validated planets in this paper. The CPs are indicated by the discovery source with Kepler indicated by ‘+’, K2 by a diamond ‘⋄’, and TESS by ‘*’. ExoMiner V1.2-boosted validated planets are indicated by a magenta square. For reference, Earth is indicated by a black circle. — astro-ph.EP

In a groundbreaking achievement, a team of machine learning scientists and astronomers from Universities Space Research Association (USRA), the SETI Institute, and NASA discovered 69 new exoplanets using advanced machine learning techniques.

The findings have been accepted for publication in the Astronomical Journal. This significant breakthrough was made possible by harnessing the power that artificial intelligence promises to expand our understanding of the universe and pave the way for future discoveries.

Exoplanets or planets outside our solar system have long captured the curiosity of scientists and public alike. Various approaches have been used for exoplanet discovery, including the transit method which led to the discovery of a majority of exoplanets. The Kepler and TESS mission efforts for example, were based on the transit method in which a target star is monitored for periodic dimming of its brightness known as a transiting event. However, not all detections resulting from transit events are exoplanets and could be due to different sources of false positives, such eclipsing binary stars.

Traditionally, to ensure that the signal detected as an exoplanet is not due a false positive source, complementary observations are used to rule out those false positives. In contrast, however, statistical and machine learning techniques have advanced so that they rely on a new process called “validation” developed to discover new exoplanets. Instead of relying on the new observations to complement the transit method, the newly discovered 69 exoplanets are validated using the previously developed deep learning called ExoMiner and a concept called multiplicity. Astronomers strongly believe that multiplicity increases the probability and therefore the level of confidence that a new detected signal around a star that already has exoplanets is much higher than the ones that do not have one.

According to Dr. Hamed Valizadegan, a machine learning scientist at Universities Space Research Association and lead author of the paper, “By utilizing the information related to how many exoplanets have already been discovered for a given star, we could boost the ExoMiner’s confidence in ruling out false positives and validate 69 new exoplanets. The 69 newly discovered exoplanets vary widely in their characteristics, including size, orbital period and proximity to their host stars and can improve our understanding of the population of exoplanets in the universe. “

The team has recently developed ExoMiner, a new deep neural network, that was used in 2021 to validate 301 new exoplanets. However, existing transit signal classifiers, including ExoMiner, do not use information regarding the configuration of a planetary system, e.g., number of existing confirmed planets or false positive signals. Utilizing the configuration of the system can help improve the confidence of a classifier to validate new exoplanets.

The existing validation techniques ignore the multiplicity boost information. In the latest work, the team has used the proposed multiplicity boost framework for ExoMiner V1.2, which addresses some of the shortcomings of the original ExoMiner classifier (Valizadegan et al. 2022), and validates 69 new exoplanets for systems with multiple KOIs from the Kepler catalog.

The discovery of 69 new exoplanets using machine learning marks a pivotal milestone in exploratory research and propels us closer to answering fundamental questions about our place in the cosmos. As we continue to explore the vast depths of space, collaborations between astronomy and artificial intelligence are expected to redefine our understanding of the universe.

Additional resources:

Multiplicity Boost of Transit Signal Classifiers: Validation of 69 New Exoplanets Using The Multiplicity Boost of ExoMiner,

Team members, Collaborators, and Customers

Hamed Valizadegan (USRA), Miguel Martinho (USRA), Jon M Jenkins (NASA Ames), Douglas A Caldwell (SETI Institute), Joseph D Twicken (SETI Institute), Stephen T Bryson (NASA Ames

Hamed Valizadegan and Miguel Martinho are supported through NASA NAMS contract NNA16BD14C, TESS GI Cycle 4 contract 80NSSC22K0184, and NASA ROSES XRP proposal 22-XRP22_2-0173. Douglas Caldwell and Joseph Twicken are supported through NASA Cooperative Agreement 80NSSC21M0079.

About USRA

Founded in 1969, under the auspices of the National Academy of Sciences at the request of the U.S. Government, the Universities Space Research Association (USRA) is a nonprofit corporation chartered to advance space-related science, technology and engineering. USRA operates scientific institutes and facilities, and conducts other major research and educational programs. USRA engages the university community and employs in-house scientific leadership, innovative research and development, and project management expertise. More information about USRA is available at


Explorers Club Fellow, ex-NASA Space Station Payload manager/space biologist, Away Teams, Journalist, Lapsed climber, Synaesthete, Na’Vi-Jedi-Freman-Buddhist-mix, ASL, Devon Island and Everest Base Camp veteran, (he/him) 🖖🏻