Room: Exhibit Hall
Date: 5/3/2024
Session Time: 8:00 AM to 5:45 PM (local time)
Machine learning (ML) is quickly changing the landscape of how we approach seismic inverse problems, including full waveform inversion (FWI) where the seismic waveforms are directly used to solve for properties such as seismic velocities, migrated images, source locations, moment tensors and more. ML can potentially overcome some of the outstanding challenges associated with conventional FWI techniques by increasing computational efficiency, automating to reduce human labor and expertise requirements, mitigating cycle skipping, parameterization and convergence issues, implementing uncertainty quantification through deep learning-based approaches, and reducing the need for a suitable starting model. The wide breadth of ML methods and emerging scientific ML, deep learning architectures, and optimization algorithms that are rapidly expanding warrant a review of current application of these technologies in the seismic inverse domain.
We encourage submissions ranging from ML methods and tools that assist conventional physics-based FWI to full end-to-end deep learning FWI methods that estimate variety of inverted properties. All ML approaches are welcome, including but not limited to deep neural networks, generative methods, decision trees, unsupervised dimensionality reduction and clustering, physics-informed ML, and application of various learning algorithms including supervised, self-supervised and unsupervised learning.
Conveners:
Jennifer L. Harding, Sandia National Laboratories (jlhardi@sandia.gov)
Mrinal K. Sen, University of Texas at Austin (mrinal@utexas.edu)
Hongkyu Yoon, Sandia National Laboratories (hyoon@sandia.gov)
Poster Presentations
Participant Role | Details | Action |
---|---|---|
Submission | Towards a Practical Physics-Informed Neural Network Method for End-to-End Full Waveform Inversion | View |
Submission | Physics-Guided Neural Network for Full Waveform Inversion With Structural Enhancement | View |
Submission | WITHDRAWN Physics-Guided Unsupervised Deep Learning Approach for the Inversion of Receiver Functions in Dipping and Anisotropic Media | View |
Submission | An Autoencoder-Based Prior for Bayesian Full Waveform Inversion | View |
Machine Learning for Full Waveform Inversion: From Hybrid to End-to-End Approaches [Poster Session]
Description