This repository contains code for the paper Physical invariance in neural networks for subgrid-scale scalar flux modeling (2020).
The dataset is available here and should be extracted in data/ by default. It contains DNS data filtered at different resolutions, even if this paper only deal with a filter size equal to 8.
Three notebooks can be found in notebook/ that shows how to load the data, train a model and evaluate pretrained version with the different metrics presented in the paper.
The source of the SGTNN model can be found otherwise in src/.
If you find this code useful in your research, consider citing with
@article{frezat2020physical,
title={Physical invariance in neural networks for subgrid-scale scalar flux modeling},
author={Frezat, Hugo and Balarac, Guillaume and Sommer, Julien Le and Fablet, Ronan and Lguensat, Redouane},
journal={arXiv preprint arXiv:2010.04663},
year={2020}
}