Buffio.

A field journal of machine-made science

← All discoveries

Astronomy · 17 February 2022

301 new exoplanets, validated by a deep-learning classifier

ExoMiner, a deep-learning classifier developed at NASA Ames, validated 301 new exoplanets from Kepler data, adding them to the catalog of confirmed planets.

Field
Astronomy
Date
17 February 2022
AI system
ExoMiner
Institution
NASA Ames Research Center
Published in
The Astrophysical Journal
Identifier
10.3847/1538-4357/ac4399

A transit signal in stellar photometry is not, by itself, a confirmed planet. Many signals are false positives, and vetting them by hand does not scale to the catalogs produced by missions such as Kepler. ExoMiner is a deep-learning classifier designed to be both highly accurate and explainable, built to distinguish true planetary transits from impostors.

In a paper published in The Astrophysical Journal on 17 February 2022, Hamed Valizadegan and colleagues at NASA Ames and collaborating institutions report that ExoMiner validated 301 new exoplanets. The title and abstract of the paper state that claim directly: the classifier validated 301 new planets, expanding the set of confirmed worlds from Kepler.

The finding is a catalog result, not a single named world. It is included here because the act of confirmation — deciding, at scale, which signals are planets — was performed by a trained model, then reported as new validated exoplanets in the peer-reviewed literature.