According to the report of this paper, it was confirmed that domestic researchers developed their own typhoon route prediction system using artificial intelligence (AI) system. The short-term hit rate is expected to be of great help to the development of the typhoon forecast system in the future as it is comparable to the Korea Meteorological Administration’s forecast using supercomputers.

A team led by Yoo Dong-hyun, a professor of mechanical engineering at POSTECH, developed its own “Tropical Storm Prediction AI System” that can accurately predict typhoons with just one GPU and published it on May 3 in the international journal “IEEE Access” published by the International Electrical and Electronic Engineers Association (IEEE on May 3.

The technology that Professor Yoo’s team applied to the development of the typhoon prediction AI system is the “Generative Adversarial Network (GAN).” GAN is a deep learning technology that combines AI model ‘generator’ that makes sophisticated forged images with AI model ‘discriminator’ that determines authenticity of it. It combines these two AI competitions to create the most realistic “fake image.”

Professor Yoo’s team trained each AI model with atmospheric images taken by satellites and weather data of typhoons that occurred in the past. Based on this data, the creator AI created data on the creation and movement of future typhoons, and the discriminator AI selected only the most realistic data among them.

By fonuder

Leave a Reply