Most machines tend to carry out repetitive tasks. For example, on an assembly line they are limited to selecting parts and placing them in a predetermined location. Problems arise precisely when there is a situation for which they are not prepared. However, in recent years different artificial intelligence systems have appeared that allow them to ‘learn’ on their own and adapt to unforeseen scenarios for which they have not been programmed, which makes them much more efficient.
Now, a team of researchers coordinated by Dan Tamayo, an astronomer at the Center for Planetary Science at the University of Toronto, in Scarborough (Canada), has developed one of these algorithms with which it is possible to identify stable extrasolar planetary systems a thousand times faster than with the techniques used to date . To do this, it uses a tool that, in essence, is inspired by those used by Google to identify and classify spam or Netflix to recommend series and movies to users.
The methods used to determine the existence of exoplanets allow us to know their size and orbital period, but with the one presented by Tamayo and his collaborators in The Astrophysical Journal Letters , based on the XGBoost software, they can also be classified by their stability and studied. how elliptical is its orbit or the maximum mass of the aforementioned world. “It is encouraging to know that our findings have led to a tool that is not only accurate but works much faster than current ones,” Tamayo says.
Image: NASA / JPL-Caltech
