Ours is no longer the only 8 planet system known (94.2% confidence in the model). Named after the discovering telescope, the Kepler System is now known to have 8 planets, like our own solar system.
The Kepler Telescope has found 2,525 new planets, so far. NASA - National Aeronautics and Space Administration and Google are using neural net machine learning on the Kepler Telescope data. They can detect planets around a star in 5-6 hours, then run machine learning to determine which are the best candidates. Their training set included only 15,000 training examples and will soon be available at www.tensorflow.org. Their success was focused on the weaker planet detection, which would be more challenging through other methods.
Once the candidates are determined, they are then statistically validated as planets by scientists (1 in 10,000 false positive probability). This prediction power is equated to meeting someone and being able to predict their sibling's height accurately.
This is a great partnership between humans and robots to accelerate discovery. "This is a tool to help astronomers to have more impact," says Christopher Shallue from Google Brain. This is one of many opportunities to use machine learning to accelerate computing methods.
Both Andrew and Chris echoed that citizen scientists will be valued more than ever, as neural networks have challenges identifying things it hasn't seen before, or subtle changes from the norm. Open source data sets are and will be available to support these efforts. The NASA data sets are available in their archives.
Many prestigious journalists from scientific publications joined the live call to ask questions and posted via Reddit Ask Me using #askNASA. One even asked if there was a catchy name for the model and if it should be called the "cats and dogs model." Making it understandable for the masses, I suppose. (I look forward to seeing what they create.)
Congratulations Christopher Shallue at Google Brain and Andrew Vanderburg at NASA!
Best wishes,
Lisa
P.S. Andrew mentions that others, such as the Webb telescope, will need to go search for biosignatures (life), as a more sophisticated telescope is required for this type of detection. A cliffhanger?
Want more?
Listen to the press announcement on the press release site:
https://www.nasa.gov/press-release/nasa-hosts-media-teleconference-to-announce-latest-kepler-discovery
More links:
https://www.nasa.gov/mission_pages/kepler/main/index.html
https://www.linkedin.com/in/cshallue/
http://g.co/brain
https://www.nasa.gov
Have info to share about this discovery? Please share below in the comments.
Have questions? Add those below and if many, I'll group them and send them to Felicia. Or, reach out directly to her contact information below.
NASA PR Contact:
Felicia Chou
felicia.chou@nasa.gov
https://www.nasa.gov/mission_pages/kepler/main/index.html