Enter Note Done
Go to previous page in this tab
Session
  • Presentation | DI33A: Advances in Machine Learning for Solid Earth Geoscience I Oral
  • Oral
  • Share Icon
    Map Icon
    Bookmark Icon
  • DI33A-08: Towards Physics-based Machine Learning Framework of Seismic Wavefield Modeling and Full-waveform Inversion (invited)
    • Schedule
    Notes
  • Location Icon348-349
    NOLA CC
    Set Timezone
    •  
    View Map

Generic 'disconnected' Message
Author(s):
Shiqian Li, Institute for Artificial Intelligence, Peking University (First Author, Presenting Author)
Zhi Li, School of Earth and Space Sciences, Peking University
Zhancun Mu, Institute for Artificial Intelligence, Peking University
Shiji Xin, Harvard University
Zhixiang Dai, NVIDIA
Kuangdai Leng, Rutherford Appleton Laboratory
Ruihua Zhang, NVIDIA
Yixin Zhu, Institute for Artificial Intelligence, Peking University
Xiaodong Song, School of Earth and Space Sciences, Peking University


Understanding the Earth's interior is important for studying earthquakes and other geological processes. Scientists use seismic waves from natural earthquakes to create detailed models of what lies beneath the surface. A method called full-waveform inversion (FWI) helps with this but is very slow and requires massive computing power.


Our study introduces GlobalTomo, a new dataset that uses computer simulations of seismic waves to train artificial intelligence (AI) models. These AI models can quickly and accurately analyze the data, offering results much faster than traditional methods—up to 60,000 times faster in some cases.


This work shows how combining Earth science with AI can improve our ability to explore and understand the deep Earth.




Scientific Discipline
Suggested Itineraries
Neighborhood
Type
Main Session
Discussion