Using artificial intelligence, researchers from the University of Warwick identified new exoplanets. The team combined a machine learning algorithm and old data from NASA's Kepler Space Telescope, which retired in 2018.

The collaborative research by Warwick's Departments of Physics and Computer Science and The Alan Turning Institute has been published in the Monthly Notice of the Royal Astronomical Society journal. 50 new planets were identified after sorting through thousands of potential candidates.

(Photo: Downloaded from NASA official website)

It was the AI algorithm that identified the distant planets outside the solar system, which may otherwise be mistaken for as background interference or camera errors. The algorithm learned to separate real planets from false positives from data that the Kepler Space Telescope collected for the nine years that it was operational.

The real exoplanets orbit around stars from distant galaxies and range from about the size of Neptune to about the size of Earth and smaller. Larger planets complete their orbit with roughly 200 days while the smaller ones in about one day.


Validating Planets

This is the first time that a machine learning technique was used to validate planets, shared David Armstrong of the University of Warwick. 'Machine learning has been used for ranking planetary candidates but never in a probabilistic framework, which is what you need to truly validate a planet.' Validated planets are confirmed after there is a less than 1% chance that the reading is a false positive.

The team's new algorithm can hopefully be used someday in future telescope missions. Other groups such as the Backyard Worlds: Planet 9 project is also involved in identifying exoplanets and other celestial objects.

Dr. Theo Damoulas shared that 'Probabilistic approaches to statistical machine learning are especially suited for an exciting problem like this in astrophysics that requires [the] incorporation of prior knowledge - from experts like Dr. Armstrong - and quantification of uncertainty in predictions. A prime example when the additional computational complexity of probabilistic methods pays off significantly.'

Read Also: Scientists Plan to Study Dead Stars and Exoplanets to Look for Fingerprints and Other Signs of Life


Using AI for Future Missions

Armstrong explained that only about one-third of all known planets have been validated using a single method. The team believes that astronomers should be using multiple techniques to validate exoplanets.

The trained AI is so efficient that it can 'validate thousands of unseen candidates in seconds,' wrote the authors. It can also be automated and validate exoplanets on its own. Now, the team is working on continually improving machine learning to become even more effective. 'We still have to spend time training the algorithm, but once that is done it becomes much easier to apply it to future candidates,' said Dr. Armstrong.

Hopefully, the new algorithm can be applied to missions conducted by NASA's Transiting Exoplanet Survey Satellite (TESS) and the ESA's Planetary Transits and Oscillations of Stars (PLATO). Since 2018, TESS has identified 66 exoplanets and more than 2,000 potential candidates and its mission has been extended until 2022.

Read Also: Scientists: Two Newly-Discovered 'Super-Earth' Exoplanets Are Habitable


Check out more news and information on Exoplanets on Science Times.