X
Innovation

NASA uses Google machine learning for exoplanet detection

Neural networks have thrown up lots of false positives, but also previously undetected exoplanets.
Written by Chris Duckett, Contributor

An eighth planet orbiting a Sun-like star over 2,500 light years away called Kepler-90 has been detected by running the data from NASA's Kepler Space Telescope through a Google neural network.

The network was trained using 15,000 previously vetted signals from the Kepler exoplanet catalogue, NASA explained, before it moved on to learning how to detect weaker signals.

"We got lots of false positives of planets, but also potentially more real planets," said NASA Sagan postdoctoral fellow Andrew 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."

In addition to the new planet around Kepler-90, the network found a new Earth-sized planet orbiting Kepler-80.

NASA's unsung heroes: The Apollo coders who put men on the moon (PDF download) | NASA shows the world its 20-year virtual reality experiment to train astronauts: The inside story (TechRepublic) | NASA Hybrid Reality Lab combines VR and the real world

It is intended for the network to examine the full dataset from Kepler, which consists of more than 150,000 stars.

"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."

In a paper, the research team say the neural network ranks planet candidates above false positives 98.8 percent of the time.

"A technique like ours could be used in the future to make more accurate estimates of planetary occurrence rates. In particular, the occurrence rate of Earth-like planets, colloquially called 'η-Earth', is one of the most important and exciting open questions in exoplanet research -- it is directly proportional to the estimated fraction of planets that might harbor life as we know it on Earth," the paper said.

On a more terrestrial level, Google is pushing its machine learning into additional areas.

In October, the search giant announced a partnership with Rolls Royce to work on autonomous ships. The deal sees Rolls Royce use Google's Cloud Machine Learning Engine to train its object classification system for detecting, identifying, and tracking the objects that a vessel can encounter at sea.

Google is also continuing to push machine learning in its more regular product set, with its Sheets app now able to suggest pivot tables from a simple natural language query.

Foxconn looking at AI for its factories

Andrew Ng, co-founder of some of Google's most prominent artificial intelligence projects, has launched a new venture with iPhone assembler Foxconn to bring artificial intelligence onto the factory floor.

At a press briefing in San Francisco two days before Ng's Landing.ai venture is introduced, he demonstrated an example of using AI for visual inspection in a factory's quality control efforts.

In many factories, workers look over parts coming off an assembly line for defects.

Ng said that while typical computer vision systems might require thousands of sample images to become trained, Landing.ai's system would take only five training images, making it easier to adapt to different tasks in a factory.

Ng said Landing.ai had been approached by investors but had not accepted outside capital yet. Foxconn, known more formally as Hon Hai Precision Industry Co Ltd, is Landing.ai's first strategic partner. Ng said the startup has been working with Foxconn since July.

He said he understands that his firm's technology is likely to displace factory workers but that Landing.ai is already working on how to train workers for higher-skilled, higher paying factory work involving computers.

"I would love to help displaced workers gain the skills they need to succeed," Ng told reporters.

Ng, a Stanford University professor, co-founded Google Brain in 2011, an effort that strung together thousands of computers that learned to identify objects like cats purely from watching YouTube videos.

Ng in 2014 moved to Chinese internet giant Baidu Inc to head up its artificial intelligence research group. He resigned from Baidu in March.

With AAP

Related Coverage

Video: How machine learning's big data loop works

ZDNet contributor George Anadiotis explains how big data helps train artificial intelligence and machine learning algorithms to optimize business software productivity.

Will 2018 be the big year for machine learning?

A new report from Deloitte says the tipping point for the technology has arrived.

Getting started with artificial intelligence and machine learning

AWS head of architecture Glenn Gore offers up advice on where to start with artificial intelligence and machine learning.

Adobe Lightroom adds AI, machine learning for better auto settings

Adobe has rolled out AI and machine learning to analyze your photo and compare it to tens of thousands of professionally edited photos to make auto corrections.

How machine learning can be used to catch a hacker (TechRepublic)

SecBI's Alex Vaystikh and Gilad Peleg explain how AI can be used to strengthen threat detection systems.

Machine learning business use will double by end of 2018, report says (TechRepublic)

The technology's projected growth shows its emergence as a business aid, according to Deloitte research.

Editorial standards