|Posted by Mark Cantrell on December 15, 2017 at 1:45 PM|
Machines find worlds in star data
Artificial intelligence looks set to play an important role in the discovery of planets beyond our solar system, writes Mark Cantrell, after NASA scientists used machine learning techniques to discover a previously overlooked world orbiting the Kepler-90 star
Image courtesy of: NASA/Wendy Stenzel
WHEN it comes to cosmic wonder, we can look but never touch. The universe has clearly evolved to taunt us: we keep on discovering all these tantalizing exo-planets, but thanks to the laws of physics they are thus far forever out of reach. So much for boldly going.
NASA announced its latest planetary discovery yesterday. With a little help from Artificial Intelligence (AI), it has found a hitherto overlooked addition to the Kepler-90 star system, 2,545 light years away from Earth. As a result, the Solar System has gained a twin. Well, kind of.
If you want to discount the debate around Pluto’s status as a planet (or not), then the Kepler-90 system is now remarkably similar to our own, with eight planets, escalating in size the further out they are from the parent star. But that’s where the similarities end.
“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 Andrew Vanderburg, a NASA Sagan Postdoctoral Fellow and astronomer at the University of Texas, in Austin, who was part of the team involved in the discovery of the latest planet.
Kepler-90 is a Sol-like star in the constellation Draco, but NASA points out that the system is an unlikely candidate for harbouring life. The newly discovered planet – Kepler-90i – is described as a “sizzling hot” rocky planet that orbits its star once every 14.4 days.
About 30 percent larger than Earth, the planet is so close to its star that its average surface temperature is believed to exceed 800 degrees Fahrenheit, putting it on a par with Mercury. The system’s outermost planet, Kepler-90h, orbits at a similar distance to its star as Earth does to the Sun.
The new planet was discovered in data from NASA’s Kepler Space Telescope, which is where the AI comes in. More specifically, machine learning techniques from Google were deployed to trawl the vast troves of data gathered by the telescope and look for the tell-tale signs of exo-planets.
Minuscule changes in brightness are captured when a planet passes in front of – or transits – a star. Vanderburg and fellow researcher Christopher Shallue trained a computer to learn how to identify exo-planets in these light readings recorded by the space telescope. Their approach was inspired by the way neurons connect in the human brain. This artificial ‘neural network’ sifted through Kepler’s data and found weak transit signals from the previously missed eighth planet.
“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. “This finding shows that our data will be a treasure trove available to innovative researchers for years to come.”
Shallue is a senior software engineer with Google’s research team, Google AI, and came up with the idea to apply a neural network to Kepler data. His interest in exo-planet discovery was derived from his interest in large-scale data. When he learned that astronomy, like other branches of science, is rapidly being inundated with data as the technology for data collection from space advances, he took up the challenge of making sense of this ‘virtual cosmos’.
“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,” he said. "Machine learning really shines in situations where there is so much data that humans can't search it for themselves.”
Machine learning has previously been used in searches of the Kepler database, NASA explained, but this research is said to demonstrate that neural networks are a “promising tool” in finding some of the weakest signals of distant worlds.
Kepler’s four-year dataset consists of 35,000 possible planetary signals. Automated tests (and sometimes human eyes) are used to verify the most promising signals in the data. However, the weakest signals are often missed using these methods. Shallue and Vanderburg thought there could be more interesting exo-planet discoveries lurking in the data.
First, they trained the neural network to identify transiting exo-planets using a set of 15,000 previously vetted signals from the Kepler exo-planet catalogue. In the test set, the neural network correctly identified true planets and false positives 96 percent of the time. Then, with the neural network having ‘learned’ to detect the pattern of a transiting exo-planet, 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 exo-planets.
“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.”
Kepler-90i wasn’t the only ‘jewel’ this neural network sifted out. In the Kepler-80 system, they found a sixth planet. This one, the Earth-sized Kepler-80g, and four of its neighbouring planets form what is called a resonant chain – where planets are locked by their mutual gravity in a rhythmic orbital dance. The result is an extremely stable system.
After gazing at one patch of space for four years, the Kepler spacecraft has gathered a vast amount of data, and there is much more to come. It is now is operating on an extended mission and switches its field of view every 80 days, so there’s going to be plenty of material for AIs to hone their capabilities.
“These results demonstrate the enduring value of Kepler’s mission,” said Jessie Dotson, Kepler’s project scientist at NASA’s Ames Research Centre in California’s Silicon Valley. “New ways of looking at the data – such as this early-stage research to apply machine learning algorithms – promises to continue to yield significant advances in our understanding of planetary systems around other stars. I’m sure there are more firsts in the data waiting for people to find them.”
Hollywood sci-fi blockbusters have tended to give AI something of a bad rep, and current fears over their impact on jobs has only added to their image problems, but maybe they’re not all that bad if they can find us some bright new frontiers to explore.
Now, if only they could crack the laws of physics and come up with a workable means to breach the light-speed barrier. A hyperdrive would come in rather handy for a first-hand glimpse of all these brave new worlds…