FV-PINNs migrates away from using pre-built automatic differentiation kernels to differentiate the neural network w/r to the governing partial differential equations. Instead we utilize underlying finite volume schemes to calculate gradients of fluxs and classic time integration schemes such as RK-X methods. I coin this method Finite Volume Physics-Informed Neural Networks. Instead of the physics coming from the underlying PDE, the physics come from the numerical discretization scheme.
First, clone repository:
git clone https://github.com/alexpapados/FV-PINNs/
Once the repository is cloned locally, run:
bash setup.sh
If you do not have bash on your machine, try:
chmod u+x setup.sh; ./setup.sh
All FV-PINNs code was written using Python. The libraries used are:
- PyTorch
- Pandas
- SciencePlots
- NumPy
- ScriPy
- Time
Each script provides a detailed description of the problem being solved and how to run the program
Preferably using an IDE such as PyCharm, and once all libraries are downloaded, users may simply run the code and each case as described in individual scripts.