NASA’s Kepler spacecraft uses a neural network to find new planets

NASA held a press conference today to announce the latest findings of the Kepler spacecraft. Kepler’s four-year mission to scan for planets in a certain region of space came to an end in 2013. Since then, it’s been scanning in many other directions, providing huge amounts of data to sift through.

That huge amount of data proved to be something of a problem for researchers. While Kepler offers up as much data as we could hope for, it also offers more than human beings could reasonably process. Kepler’s task is to monitor the light around stars. When planets pass in front of their star, they interrupt the strength of that star’s light. This isn’t enough to clearly see the shape of a planet, but it is often enough to show that one is there. With around 150,000 signals since Kepler got started, researchers had to be picky about what data they examined. Only the strongest, and therefore most likely, candidates were examined by the researchers in detail. Unfortunately, this left a massive amount of data which was simply being ignored.

Luckily, NASA, in conjunction with elements of Google’s AI division, were able to create a neural network to examine the data. This neural network is able to learn from existing data, in this case, the positive results from all of the human-found planets in the data. Based on those positive results, the neural network is then able to search through all of the weakest signals to find planets that had been overlooked. The beauty of this neural network is that, while it works much faster than human scientists, it also works alongside them. The network requires a backlog of data before it can function. This data naturally comes from human research and then the network can be aimed only at areas of research which humans couldn’t be expected to handle on their own.

The new neural network has recently identified two new planets, Kepler90-I and Kepler80-G. The star, Kepler90 is now the first star besides our own to be discovered with as many as eight planets. This means our own solar system is no longer the sole record holder, and it may also mean that our particular system isn’t as unusual as we may like to think. The newly discovered 90-I is the sixth planet to orbit its star and all of its system’s planets are packed unusually close to their sun. NASA scientists speculate that this may suggest that the planets formed further out from their star and were pulled inwards by gravitational forces. It could also mean that there are still more planets in that star system which have just escaped notice for the time being.

Kepler 90 and its 8 currently discovered exoplanets ties with our own solar system for the record.

NASA representatives were clear that the data being examined and the network’s capabilities can only really confirm or deny the presence of a planet. They can’t provide much further information about what phenomenon may have caused a negative finding. They also certainly can’t identify whether planets might have the potential for life.

Perhaps one of the most exciting things about NASA’s neural network is this: They’re about to share it with the world. The code for their planet-searching network is going to be released shortly. Kepler’s data is already publicly available and, according to NASA, no specialised hardware is needed to run their network. This means that all of the data we currently have about the stars and planets beyond our own is lying their for anyone to look through. The next planet to be identified could be found by anyone who feels like joining in on the search.

Ronan Daly

Ronan Daly is a staff writer for My Good Planet who specialises in Technology and Science. With a Masters Degree in English, and over a decade's experience as a teacher and writer, Ronan has brought a breezy, learned style to My Good Planet, making occasionally complex material accessible and understandable to all.

No Comments Yet

Leave a Reply

Your email address will not be published.