NASA uses analytics to study the stars

Posted by Justin Hesser on February 27, 2014

Few organizations deal with bigger questions than NASA. While many companies have to deal with concerns about people thousands of miles apart, the distances that NASA is worried about stretch into the light-years. 

It's no surprise, then, that they would turn to Big Data. Custom web application development is a powerful tool, and for issues of the scope that they're dealing with, perhaps the only appropriate one. 

Round the clock, NASA has over 100 devices scattered throughout our solar system (including some here on Earth!), gathering information. They take in a lot of data — so much so, in fact, that at the Jet Propulsion Lab, parsing it, storing it and analyzing it is a full-time job. 

Because it gets so much data, sorting through it all can be a chore. The ultimate question scientists have to ask themselves is "what's signal, and needs to be kept, and what's just noise and can be readily discarded?"

To answer these questions, NASA turns to the use of complex algorithms. These computational shortcuts allow scientists not to have to observe every batch of data by hand, a task that would be literally impossible in the volumes that they're collecting. Instead, once the unmanned rovers, satellites and telescopes transmit their information to the servers, the process of scanning it and sorting it is automated. 

Kiri Wagstaff, who works in the machine learning sector of the propulsion lab, described the process that her division goes through when collecting knowledge. 

"Our software picks it up and makes a decision whether something interesting happened in the past five milliseconds. Our detection algorithm looks for correlations amongst all of the radio telescopes. If [an object is] popping up on all of them and not just one, they come together to make a quick decision whether or not to send it into classification," she told Forbes.

She's even more optimistic for the future of these devices. Wagstaff revealed to Forbes that there was room for improvements in the algorithms, especially in the area of better categorizing false events. As the algorithms get better, they will "learn" how not to prioritize these events. In addition, they will be able to move through this entire process significantly more quickly. 

For most organizations, the sky is the limit with what they can do with Big Data. For NASA, the possibilities stretch far beyond.