Machine learning algorithm confirms 50 new exoplanets

Hyperaxion Aug 26, 2020

An algorithm developed by British researchers confirmed, through precise calculations, the existence of new exoplanets.

A new machine learning algorithm developed by scientists at the University of Warwick, UK, confirmed the existence of 50 planets outside the Solar System.

The discovery was published last week in the Monthly Notices of the Royal Astronomical Society.

Machine learning algorithm confirms 50 new exoplanets
(Credit: NASA / JPL-Caltech / R. Hurt).

This is the first time that astronomers have used the method to analyze a sample of potential planets and determine which of them are real and which are not, calculating the probability of each candidate being a real planet.

“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,” said Dr. David Armstrong, from the University of Warwick Department of Physics.

Many studies try to confirm the existence of exoplanets using data from telescopes and analyzing the interaction of the object with its host star.

As useful as it may be, this method is not fully effective and the existence of a detected planet needs to be validated to confirm its existence.

In the new research, the scientists created an algorithm based on machine learning that can distinguish real planets from fake planets. Artificial intelligence was trained from two large samples of data from the now retired Kepler mission.

“Rather than saying which candidates are more likely to be planets, we can now say what the precise statistical likelihood is,” Armstrong explained. “Where there is less than a 1% chance of a candidate being a false positive, it is considered a validated planet.”

The sizes of these 50 planets vary: from worlds as large as Neptune to smaller than Earth, with orbits ranging from 200 days to just one.

According to astronomers, by confirming that these exoplanets are real, they can now prioritize them for future observations with telescopes.

In addition, scientists hope that, once trained, the algorithm will be faster than existing techniques for planet validation, and can still be fully automated.

“Almost 30% of the known planets to date have been validated using just one method, and that’s not ideal,” Armstrong said. “Machine learning also lets us do it very quickly and prioritise candidates much faster.”


Leave a Reply

Notify of