Every time we think we’re in a class by ourselves, it’s only a matter of time. Used to be our solar system had the most planets around a single star. But thanks to new data from NASA’s Kepler spacecraft, we’re now tied with the Kepler-90 system, a sun-like star 2,545 light years away in the constellation Draco.
Researchers recently uncovered Kepler-90i, a sizzling-hot 8th planet that orbits Kepler-90 once every 14.4 days. To find it, they used a unique method: machine learning. Machine learning is a type of computer science that gives machines (computers) the ability to go beyond strict programming and learn on their own. In this case, computers learned to identify planets by finding instances in the Kepler data where the telescope recorded signals from exoplanets beyond our solar system.
“Just as we expected, there are exciting discoveries lurking in our archived Kepler data, waiting for the right tool or technology to unearth them,” said Paul Hertz, director of NASA’s Astrophysics Division in Washington in a press release.
The discovery came about after researchers trained a computer to learn how to identify exoplanets in the light readings recorded by Kepler – the tiny change in brightness captured when a planet passed in front of, or transited, a star (see below). Inspired by the way neurons connect in the human brain, this artificial “neural network” sifted through Kepler data and found weak transit signals from a previously-missed eighth planet orbiting Kepler-90.
About 30% larger than Earth, Kepler-90i is so close to its star that its average surface temperature may exceed 800°F (426° C), on par with Mercury and Venus and likely too hot to support life.
“The Kepler-90 star system is like a mini version of our solar system. You have small planets inside and big planets outside, but everything is scrunched in much closer,” said Vanderburg, an astronomer at the University of Texas at Austin.
Kepler’s four years of data consists of 35,000 possible planetary signals. Previously, automated tests, and sometimes human eyes, are used to verify the most promising signals in the data, but the weakest signals often are missed. Researchers Christopher Shallue and Andrew Vanderburg thought there could be more exoplanet discoveries faintly lurking in the data.
First, they trained the neural network to identify transiting exoplanets by feeding it 15,000 previously studied signals from the Kepler exoplanet catalogue. In the test set, the network correctly identified true planets and false positives 96% of the time. Once the neural network “learned” to detect the pattern of a transiting exoplanet, the researchers directed their model to search for weaker signals in 670 star systems that already had multiple known planets. Their assumption was that multiple-planet systems would be the best places to look for more exoplanets.
“We got lots of false positives of planets, but also potentially more real planets,” said Vanderburg. “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.”
We’re teaching our machines rudimentary thinking … and it’s starting to show results.
Shallue and Vanderburg plan to apply their neural network to Kepler’s full set of more than 150,000 stars. Here’s their research paper.