Spectral Neural Networks for Solving PDEs

This repository contains the code for a Master's thesis research project on using Spectral Neural Networks (SNNs) for solving partial differential equations (PDEs). The framework is built on DeepXDE and PyTorch, with experiment tracking managed by MLflow.

Setup and Installation

1. Clone the repository:

   git clone https://github.com/aed6-ua/master2025.git
   cd master2025

2. Create a Python virtual environment:

   python -m venv venv
   source venv/bin/activate

3. Install the required packages:

   pip install -r requirements.txt

4. Start the MLflow Server (optional, for local tracking):

   mlflow server --host 127.0.0.1 --port 5000

   Then, open your browser to http://127.0.0.1:5000 to view the MLflow UI.

How to Run an Experiment

Experiments are launched using the run_experiment.py script, which requires a configuration file. To run a single experiment (e.g., the SpectralNN model on the wave equation), use the following command:

python run_experiment.py --config configs/wave_snn.yaml

You can add a random seed for reproducibility:

python run_experiment.py --config configs/wave_snn.yaml --seed 42