Further blurring the lines between science and science-fiction, an artificial intelligence system leveraged by NASA has discovered two previously unknown exoplanets. One of the new exoplanets, a sizzling rocky world called Kepler-90i, is significant because it brings the known planets orbiting its star to eight, and it's the first time a numerical twin to our solar system has ever been detected.
"The Kepler-90 star system is like a mini version of our solar system," Andrew Vanderburg, a NASA Sagan Postdoctoral Fellow and astronomer at the University of Texas at Austin, said in a statement. "You have small planets inside and big planets outside, but everything is scrunched in much closer."
While the seven other planets in the Kepler-90 star system, located some 2,545 light-years from Earth, were previously detected by the Kepler Space Telescope, researchers had a hunch that others might be hiding in its archived dataset. Over the past four years, Kepler has detected over 35,000 possible planetary signals, with rudimentary automated tests and even human eyes serving as the only vetting tools available to mine the information.
In a paper posted at the Harvard-Smithsonian Center for Astrophysics, researchers Christopher Shallue and Andrew Vanderburg explained how they trained a neural network, a computing system that mimics the way the human brain works, to sift through the massive quantities of Kepler data and "pick out" signals indicating the presence of exoplanets.
"In my spare time, I started googling for ‘finding exoplanets with large data sets’ and found out about the Kepler mission and the huge data set available,” said Shallue, a senior software engineer with Google’s research team Google AI. "Machine learning really shines in situations where there is so much data that humans can't search it for themselves."
According to Shallue, Kepler's dataset was so large that it took him more than two weeks to download.
'... like sifting through rocks to find jewels'
To teach this intergalactic google how to search the stars for new worlds, Vanderburg and Shallue first trained the system on 15,000 previously vetted signals from Kepler. Once it had passed this first test (with an accuracy of 96 percent) the team directed the network to scan for weaker signals in 670 star systems that already had multiple planets.
"We got lots of false positives of planets, but also potentially more real planets," added Vanderburg, who plans to apply the neural network to Kepler's full set of more than 150,000 stars. "It’s like sifting through rocks to find jewels. If you have a finer sieve then you will catch more rocks but you might catch more jewels, as well."
After four years of scanning the same patch of sky, the Kepler is now operating on an extended mission and switching its field of view every 80 days. By leveraging the growing sophistication of neural networks, NASA researchers are hopeful that their new high-tech partner will further increase the rate at which they discover new worlds and solar systems like our own.
"Today's result highlights that planetary systems come in a variety of configurations -— and that there's a lot we don't yet know," Jessie Dotson of NASA Ames Research Center said in a Reddit AMA. "It also demonstrates the importance of developing and utilizing sophisticated algorithms in addition to designing and building more powerful telescopes."