Recently, 50 possible planets have been confirmed by an intriguing machine learning algorithm realized by the University of Warwick scientists.
Astronomers have utilized for the first time a procedure based on machine learning, a type of artificial intelligence, to examine a sample of possible planets. It was also used to determine which ones are the real deal and which are not. Here is what you need to know.
Machine Learning Algorithm Changed the Way of Finding Exoplanets
A team of researchers at Warwick’s Departments of Physics and Computer Science collaborated with the Alan Turing Institute to develop a machine learning-based algorithm that can spot the real planets from false ones in the large samples of thousands of candidates discovered by telescope missions.
The machine-learning algorithm was programmed to identify real planets using two large samples of confirmed planets and false positives from the former Kepler mission. Researchers then utilized the algorithm on a dataset of still unconfirmed planets from Kepler. Their findings were astonishing because they succeeded in discovering 50 new confirmed planets and the first to be validated by the new development.
These 50 planets are as large as Neptune or smaller than our world, with orbits of 200 days or only a single day. By validating these newly found planets, astronomers can now prioritize these for more observations with especially-developed telescopes.
“[…] we can now say what the precise statistical likelihood is,” stated Dr. David Armstrong, from the University of Warwick.
Once developed and programmed, the algorithm is faster than the actual methods and can be entirely automated. It is perfect for examing the potentially thousands of planetary candidates spotted in a current mission like NASA’s TESS. Also, developing new techniques for confirmation should be a continuous process, according to researchers. The work of finding new planets is more than just one mission, survey, or study.