HNO: Deep Neural Helmholtz Operators

Introduction

This repository provides code and (part of) data for the paper Deep Neural Helmholtz Operators for 3D Elastic Wave Propagation and Inversion.

File Description

• code:
  • util: Classes and functions involved
  • HNO_3D_train.ipynb: Workflow for training a 3D HNO
  • HNO_GNO_3D_train.ipynb: Workflow for training a 3D GNO-embedded HNO
  • HNO_GNO_3D_fwi.ipynb: Workflow for full-waveform inversion
  • HNO_3D_test_overthrust.ipynb: Generalization test with overthrust models
  • HNO_3D_test_super_resolution.ipynb: Generalization test with higher resolution (input_sr.npy that exceeds the size limit is here)
• data: Data for use with the code
• data_generation: Code for generating training data with Salvus
• model: Normalizers for data processing. For 3D HNO models please see below.

Dependencies

environment.yml

Pre-trained 3D Models

The 3D HNO models can be reproduced with code and data in this repository. Pre-trained models are also available here.

Example Prediction

Contact

We welcome any comments or questions regarding this work. If you find it helpful, please cite:

Zou, C., Azizzadenesheli, K., Ross, Z. E., & Clayton, R. W. (2024). Deep Neural Helmholtz Operators for 3-D Elastic Wave Propagation and Inversion. Geophysical Journal International, 239(3), 1469-1484.

Caifeng Zou
czou@caltech.edu