Communications of the ACM
Deep Learning Accurately Forecasts Heat Waves, Cold Spells
Engineers at Rice University developed a system that used deep learning to teach itself to predict extreme weather events.
Credit: WLS-TV
Rice University engineers have developed a computer system that used deep learning to teach itself to accurately predict extreme weather events like heat waves and cold spells, based on a minimal amount of weather data.
Rice's capsule neural network studied hundreds of pairs of maps during training, with each map displaying surface temperatures and air pressures at five-kilometer elevations, several days apart. After training, the network could examine previously unseen maps and produce five-day extreme weather projections with 85% accuracy.
The researchers think the network eventually could function as an early warning system for weather forecasters, as well as a tool for learning more about the atmospheric precursors of extreme weather.
From Rice University
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA
No entries found