Reproducible material for Learned frequency-domain scattered wavefield solutions using neural operators - Xinquan Huang, Tariq Alkhalifah.

Project structure

This repository is organized as follows:

• 📂 asset: folder containing logo;
• 📂 data: folder containing data and the code used to generate the data;
• 📂 scripts: set of python scripts used to run multiple experiments;
• 📂 src: python library containing base code;

Getting started 👾 🤖

To ensure reproducibility of the results, we suggest using the opl.yml file when creating an environment.

Simply run:

./install_env.sh

It will take some time, if at the end you see the word Done! on your terminal you are ready to go.

Remember to always activate the environment by typing:

conda activate opl

Data Generation

The data are generated using the following scripts, which utilize the MATLAB API through Python. Got the folder scripts and run the following command:

bash data_generation.sh

After the data are generated, you will see the following files in the data folder:

Training

After the data are generated, we can start training the model. Got the folder scripts and run the following command:

bash train.sh

The training process will be logged in wandb. You can view the training process by going to wandb. The checkpoints will be saved in the results folder.

Disclaimer: All experiments have been carried on a Intel(R) Xeon(R) CPU @ 2.30GHz equipped with a single NVIDIA GEForce A100 GPU. Different environment configurations may be required for different combinations of workstation and GPU.

Cite us

@article{huang2025learned,
  title={Learned frequency-domain scattered wavefield solutions using neural operators},
  author={Huang, Xinquan and Alkhalifah, Tariq},
  journal={Geophysical Journal International},
  volume={241},
  number={3},
  pages={1467--1478},
  year={2025},
  publisher={Oxford University Press}
}