Primary Convener:
Laura Laurenti, ETH Swiss Federal Institute of Technology Zurich

Convener:
Christopher Johnson, Los Alamos National Laboratory

Chair:
Laura Laurenti, Sapienza University of Rome
Christopher Johnson, Los Alamos National Laboratory

Machine learning is rapidly advancing scientific research and solid Earth geophysics has benefited in this data driven modeling transformation. Many existing models are task-specific without a generalization of the data, therefore limiting widespread application across different settings. Inspired by natural language processing and computer vision, we can extend the foundation model concept to seismology and geodesy. Foundation models are large-scale, pre-trained deep neural networks that learn general representations from vast amounts of data and easily adapt to a range of downstream task-specific applications. This session aims to highlight advancements in foundation models for solid Earth geophysics. We invite contributions that explore machine learning architectures, domain adaptation strategies, large-scale pretraining methodologies, evaluation methods, and downstream applications of foundation models (even existing ones) in seismology and geodesy. We aim to foster discussions on the impact of foundation models on seismology and geodesy research, as well as their contribution to scientific progress.

Index Terms
1299 General or miscellaneous
1942 Machine learning
4315 Monitoring, forecasting, prediction
7230 Seismicity and tectonics

Suggested Itineraries:
Disasters‚ Calamities and Extreme Events
Machine Learning and AI

Neighborhoods:
2. Earth Interior