Communications of the ACM
The Rise of the Robo-Meteorologist
Microsoft researchers are using machine learning to make more accurate weather predictions as much as 24 hours in advance.
Credit: Getty Images Staff
Microsoft researchers Ashish Kapoor and Eric Horvitz are using machine learning to make more accurate weather predictions over a 24-hour period. Their system tries to "learn" from massive data sets of past weather events.
Kapoor and Horvitz use deep neural networks they had used to enhance artificial sight and speech. The researchers note their new models have much lower error rates when it comes to predicting wind, dew point, pressure zone locations such as geopotential height, and temperature, up to 24 hours in advance.
"The deep neural network strives to model the dependencies across variables without making explicit assumptions," Kapoor says. "You don't even need to encode that relationship, these models learn that relationship automatically."
Kapoor previously used machine learning to examine higher altitude wind patterns by using public U.S. Federal Aviation Administration data from tens of thousands of commercial airplane flights per day. "Using that function, you try to extrapolate what it's going to look like in the future," he notes.
Kapoor now is investigating how the system's period of accuracy can be extended beyond 24 hours. The researchers believe they can help scientists better understand the effects of climate change on weather patterns.
From Gizmodo
View Full Article
Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA
No entries found