Neural networks for Helmholtz--Hodge decomposition

In this project, we investigate physics-informed neural networks (PINNs) that are constructed in such way that they separate potential (curl-free) and solenoidal (divergence-free) vector fields, what is known as Helmholtz--Hodge decomposition.

Setup

1. Clone the project with git via SSH:

   git clone [email protected]:dmitry-kabanov/hhpinn.git
   

   or HTTPS:

   git clone https://github.com/dmitry-kabanov/hhpinn.git
   

2. Install conda environment for the project:

   conda env create -f environment.yml [-n env-name]
   

where optional argument env-name has the default value hhpinn.