Resolution invariant deep operator network for PDEs with complex geometries

This repository is the official implementation of Resolution invariant deep operator network for PDEs with complex geometries at Journal of Computational Physics

Preparing your dataset

1. Run data/step1_genereate_data.ipynb to generate the raw datasets
2. Run data/step2_split.ipynb to split the data to train/validation/test datasets

Training neural operators

Producing the prediction sets:

python main_train_model.py

with the following arguments:

• lx: the length correlation.
• model: the name of the model.
• r: the resolution of input field.
• integral_type: the type of integral used in the model.
• seed: the random seed for reproducibility.
• gpu: the ID of the GPU to use.

Citation

If you find this useful in your research, please consider citing:

@article{huang2024resolution,
title={Resolution invariant deep operator network for PDEs with complex geometries},
author={Huang, Jianguo and Qiu, Yue},
journal={arXiv preprint arXiv:2402.00825},
year={2024}
}