50 New Exoplanets discovered by Kelper telescope. During the last decades we have made great strides in the search for planets outside the Solar System. For them, space telescopes like Kepler have been in charge of collecting a gigantic amount of data. Now, analyzing all that data is a tedious task for astronomers. What if an AI took care of them?
This is what researchers at the University of Warwick in collaboration with the Alan Turning Institute thought . So they trained an artificial intelligence to learn how to search the telescope data samples and thus find new exoplanets not yet identified. The result? 50 new exoplanets at once.
During the last two decades, more than 4,200 exoplanets have been discovered throughout the Universe. This is important because, by analogy, if they are similar to Earth when orbiting a star, they have a better chance of harboring life or being habitable. To discover these exoplanets , a technique called the transit method is generally used. The method is based on analyzing the amount of light coming from a star and how that amount temporarily decreases when an exoplanet comes between the star and the telescope, in other words, when it transits.
With this technique, astronomers can know that “there is something there”, but it is not a direct confirmation that it is an exoplanet orbiting that star. It can be a binary star system, it can be an asteroid close to the telescope that it has crossed, it can be a telescope error … That is why astronomers have to analyze all possible candidates one by one to see if they really do. It’s about an exoplanet or a false positive.
Discarding false positives
After being trained, the artificial intelligence analyzed a third sample with possible exoplanets that had not yet been analyzed. In analyzing the probabilities that they were false positives or not, the AI selected only those that were less than 1% false positives and classified them as exoplanets. With this he was able to discover a total of 50 new exoplanets at once.
One of the most important uses of machine learning algorithms is to analyze large samples of data in a much faster and more automated way. Essentially what this AI has done by observing the vast Universe filled with stars and possible exoplanets.
Find More On: Warwick University