Artificial Intelligence discovers 50 exoplanets, can be trained for future planet-hunting missions: Experts
The technology, developed by researchers from the UK, is trained to address a problem often encountered by astronomers: differentiating real exoplanets from fake ones
Artificial intelligence has verified the existence of 50 new alien worlds, according to a new study. The technology, developed by researchers from the UK is trained to address a problem often encountered by astronomers: differentiating real exoplanets from fake ones.
The AI scoured through oceans of data gathered by NASA's Kepler mission to confirm these worlds. Of the 50 confirmed exoplanets, some take a day to orbit their stars while others take 200 days to do the same. Their sizes range from bodies bigger than Neptune to smaller than Earth. After confirming their existence, researchers are hoping to study them further.
Typically, astronomers discover new exoplanets by analyzing data gathered by telescopes. For instance, they studied data from the now-retired NASA's Kepler mission, looking for a dip in the brightness of stars. The dimness in light could indicate a potential planet passing in front of a star. Of the 4201 known exoplanets, the mission confirmed 2,331, according to the space agency. But the technique — transit — is not foolproof due to background interference, camera errors and other interfering factors.
"Almost 30% of the known planets to date have been validated using just one method, and that’s not ideal," says Dr David Armstrong from the University of Warwick Department of Physics in a statement. "Developing new methods for validation is desirable for that reason alone."
So the UK team developed machine learning, a form of artificial intelligence. "In terms of planet validation, no one has used a machine learning technique before. Rather than saying which candidates are more likely to be planets, we can now say what the precise statistical likelihood is. Where there is less than a 1% chance of a candidate being a false positive, it is considered a validated planet," says Dr Armstrong.
The machine learning algorithm is faster, automated and can perform better with time, say experts. Machine learning has found applications from training self-driving cars to predicting volcanoes. Dr Armstrong and his team trained the machine learning algorithm by using data on confirmed planets and fake ones gathered by the Kepler space telescope.
Then, the researchers tested it by feeding it with data on potential planets. Towards the end of the exercise, it managed to confirm 50 exoplanets. There is scope for improvement with further training. "We still have to spend time training the algorithm. But once that is done, it becomes much easier to apply it to future candidates. You can also incorporate discoveries to progressively improve it," explains Dr Armstrong.
In 2018, NASA launched another telescope hunting for exoplanets: TESS. It has confirmed 67 exponents and identified 1318 candidates that are awaiting confirmation. The machine learning algorithm can help here. "Fast, automated systems like this that can take us all the way to validated planets in fewer steps let us do that efficiently," notes Dr Armstrong.
Earlier this month, TESS covered 75% of the skies, accomplishing its primary mission. "TESS is producing a torrent of high-quality observations, providing valuable data across a wide range of science topics," says Patricia Boyd, the project scientist for TESS at NASA's Goddard Space Flight Center, in a statement. "As it enters its extended mission, TESS is already a roaring success." The study was published in the Monthly Notices of the Royal Astronomical Society.