Google’s artificial intelligence (AI) system that analyses data from NASA’s planet-hunting probe to identify the most promising signals has now been made public to help amateur scientists spot new worlds in the universe. The system has recently discovered two exoplanets by training a neural network to analyse data from NASA’s Kepler space telescope and accurately identify the most promising planet signals.
This was done by an initial analysis of about 700 stars. “We consider this a successful proof-of-concept for using machine learning to discover exoplanets, and more generally another example of using machine learning to make meaningful gains in a variety of scientific disciplines,” Chris Shallue, Senior Software Engineer with the Google Brain Team wrote in a blog post. “We are excited to release our code for processing the Kepler data, training our neural network model, and making predictions about new candidate signals,” Shallue wrote.
“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,” he wrote. The Kepler telescope hunts for planets by measuring the brightness of a star over time. When a planet passes in front of the star, it temporarily blocks some of the light, which causes the measured brightness to decrease and then increase again shortly.
However, other astronomical and instrumental phenomena can also cause the measured brightness of a star to decrease, including binary star systems, star spots, cosmic ray hits on Kepler’s photometer, and instrumental noise. To search for planets in Kepler data, scientists use automated software to detect signals that might be caused by planets, and then manually follow up to decide whether each signal is a planet or a false positive.
To avoid being overwhelmed with more signals than they can manage, the scientists apply a cutoff to the automated detections: those with signal-to-noise ratios above a fixed threshold are deemed worthy of follow-up analysis, while all detections below the threshold are discarded. Even with this cutoff, to date, over 30,000 detected Kepler signals have been manually examined, and about 2,500 of those have been validated as actual planets.
The Google Brain team applies machine learning to a diverse variety of data, from human genomes to sketches to formal mathematical logic. “Considering the massive amount of data collected by the Kepler telescope, we wondered what we might find if we used machine learning to analyse some of the previously unexplored Kepler data,” Shallue said. In collaboration with University of Texas at Austin in the US, the team developed a neural network to help search the low signal-to-noise detections for planets.
About 30,000 Kepler signals had already been manually examined and classified by humans. “We used a subset of around 15,000 of these signals, of which around 3,500 were verified planets or strong planet candidates, to train our neural network to distinguish planets from false positives,” Shallue said. The system was tested for its effectiveness by searching for new planets in a small set 670 stars.
“We chose these stars because they were already known to have multiple orbiting planets, and we believed that some of these stars might host additional planets that had not yet been detected,” Shallue said. “We allowed our search to include signals that were below the signal-to-noise threshold that astronomers had previously considered,” he said. The neural network rejected most of these signals as spurious detections, but a handful of promising candidates rose to the top, including two newly discovered planets: Kepler-90 i and Kepler-80 g. The model is now available to the public, allowing researchers to train it further and discover more planets.