Authors: Valentin Duruisseaux, Joshua W. Burby, and Qi Tang
This repository provides a simplified version of the Python/TensorFlow code used to generate some of the results in our paper
Approximation of Nearly-Periodic Symplectic Maps via Structure-Preserving Neural Networks.
Valentin Duruisseaux, Joshua W. Burby, and Qi Tang.
Scientific Reports, vol. 13, no. 8351, 2023.
Collection on "Physics-informed Machine Learning and its real-world applications"
The code is provided in two different formats
NPMap_Learning.py NPMap_Learning.ipynb
These codes contain detailed explanations with equations in LaTeX which are better rendered in the Jupyter notebook version.
The directory ./TrainingWeights/
contains the weights of the trained model whose results are presented in Figure 5 of our paper.
These weights can be loaded into the model by setting
train = False
python ./NPMap_Learning.py
This code is also published and available at https://www.osti.gov/biblio/1972078/
If you use this code in your research, please consider citing:
@article{Duruisseaux2023NPMap,
author = {V. Duruisseaux, and J. W. Burby, and Q. Tang},
title = {Approximation of nearly-periodic symplectic maps via structure-preserving neural networks},
journal = {Scientific Reports, Collection on ``Physics-informed Machine Learning and its real-world applications"},
doi = {10.1038/s41598-023-34862-w},
year = {2023}
}
@article{Duruisseaux2023NPMapCode,
title = {Code Demonstration: Approximation of nearly-periodic symplectic maps via structure-preserving neural networks},
author = {V. Duruisseaux, and J. W. Burby, and Q. Tang},
doi = {10.2172/1972078},
url = {https://www.osti.gov/biblio/1972078},
year = {2023}
}
The software is available under the MIT License.