Researchers at Japan's RIKEN Prediction Science Laboratory applied deep learning to predict the impact of tsunamis in less than a second; conventional modeling yields forecasts in about 30 minutes.
The researchers trained the system on more than 3,000 computer-generated tsunami events, then tested it with 480 other tsunami models and three actual events.
The machine learning-based model required just 1% of the computational effort of conventional models to produce comparable accuracy.
RIKEN's Iyan Mulia said the method is only accurate for large tsunamis that exceed approximately 1.5 meters (about 5 feet), although he explained, "you can apply this method to any kind of disaster predictions where the time constraint is very limited."
From RIKEN (Japan)
View Full Article
Abstracts Copyright © 2023 SmithBucklin, Washington, DC, USA