3d render of unreal Trappist-1 exoplanets system (This content is subject to copyright.)
Google has open-sourced their planet-hunting AI algorithm.
In December, NASA announced they had found two exoplanets hiding in plain sight. The discovery was made by a neural network trained to sift through data collected from the agency’s Kepler spacecraft.
Kepler was launched in 2009 specifically to search for exoplanets orbiting around distant stars. Astronomers detect exoplanets based on changes in the brightness of stars. If a star dims for a short period of time, it’s likely that a planet is passing in front of it.
In four years, Kepler observed 150,000 stars, which gave astronomers more data than they were able to sift through. So they only focused on the 30,000 strongest signals and managed to discover 2,500 exoplanets. But this left 120,000 signals ignored.
Google researchers then trained their AI to search through the 120,000 unanalyzed signals. They fed the machine 15,000 examples of NASA-confirmed exoplanet data in order to teach it how to spot the characteristics of an exoplanet.
Google has now released that code on Github, along with instructions on how to use it, so the public can try for their own celestial discovery. However, aspiring explorers will have an easier time navigating the AI if they’re familiar with coding in Python and Google’s machine-learning software, TensorFlow.
“We hope this release will prove a useful starting point for developing similar models for other NASA missions, like K2 (Kepler’s second mission) and the upcoming Transiting Exoplanet Survey Satellite mission,” Christopher Shallue, the lead engineer behind Google’s exoplanet AI, wrote in a blog post.
Shallue also wrote that he hopes this will encourage further analysis of the remaining Kepler data.
This story originally appeared in the New York Post.
Get a daily look at what’s developing in science and technology throughout the world.