Skip to content

PinnDE: A Python library for solving differential equations with physics informed neural networks and deep operator networks

License

Notifications You must be signed in to change notification settings

JB55Matthews/PinnDE

Repository files navigation

PinnDE

Physics Informed Neural Networks for Differential Equations (PinnDE) is an open-source library in Python 3 for solving ordinary and partial differential equations (ODEs and PDEs) using both physics informed neural networks (PINNs) and deep operator networks (DeepONets). The goal of PinnDE is to provide a user-friendly library as an alternative to the more powerful but more complex alternative packages that are available within this field.This library provides simple, user-friendly interfacing of solving methods which can easily be used in collaboration with non-profficent users of python or the library, where collaborators should be able to understand the contents of the code quickly and without having to learn the library themselves. We also propose the use of PinnDE for education use. Methods in this field that may be taught by educators at a low level may be understandable to students, but the code to implement these ideas can be large and more difficult to grasp. PinnDE provides simple implementations where students can expmeriment with different variations of model parameters and training methods without needing to delve into low level implementations.

The documentation can be found here

Installation

This package requires numpy, tensorflow, jax/flax/optax, matplotlib, and pyDOE. These are all installed with the package. If version of a package already installed which is above the reqiements for PinnDE, then currently package won't be upgraded when installed.

Installing can simply be done with pip in the command line with

pip install pinnde

Citing

If PinnDE is used in academic research, please cite the paper found here, or with the corresponding BibTex citation

@article{matthews2024pinnde,
        title={PinnDE: Physics-Informed Neural Networks for Solving Differential Equations},
        author={Matthews, Jason and Bihlo, Alex},
        journal={arXiv preprint arXiv:2408.10011},
        year={2024}
}

About

PinnDE: A Python library for solving differential equations with physics informed neural networks and deep operator networks

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages