Astronomers identify 50 new exoplanets using AI
This is the first time that a machine learning technique has been used to validate exoplanets
A machine learning algorithm, created by the researchers from the University of Warwick and the Alan Turing Institute, has helped astronomers to discover 50 new exoplanets in the data collected by NASA's ground-breaking Kepler space telescope.
Identifying an exoplanet - a planet orbiting a star other than the Sun - in the massive quantities of data gathered via telescopes is usually a complex process. To validate their existence, astronomers have to rely on planet transit phenomena; that is, passing of a planet between its host star and the Earth. Telescopes detect this as a fall in the amount of light coming from the star. However, such periodic dips in starlight could also be a result of equipment errors, interference from an object in the background or a binary star; therefore, astronomers have to take into account many other features, such as an object's size and shape.
In the new research, scientists developed a machine learning algorithm and trained it with two large samples of confirmed planets and false positives (fake planets) from NASA's now-defunct Kepler mission.
The team employed their algorithm on a new dataset of potential planetary candidates and were able to identify 50 new exoplanets, ranging in size from Neptune to Earth-like scales.
According to researchers, the objects with the smallest false-positive chance were selected as potential alien planets.
"Where there is less than a one per cent chance of a candidate being a false positive, it is considered a validated planet," said David Armstrong, a research fellow at the UK's University of Warwick- who is also the lead researcher of the study.
This is the first time that a machine-learning technique has been used to validate exoplanets. The researchers say that the algorithm, once trained, is faster than existing techniques for identifying exoplanets in telescopic data.
Moreover, it can be completely automated, making it ideal for studying thousands of planetary candidates observed in telescopic surveys.
The researchers now want to apply their algorithm to large samples of potential exoplanets from current and future space missions, like TESS and PLATO.
The team is also working to open-source their algorithm so that other scientists can use it to analyse data from other astronomical surveys.
The study's detailed findings are published in the Monthly Notices of the Royal Astronomical Society.