SSE-GNN

Requirements

You will need the following Python packages:

• signatory
• torch
• torch_geometric
• torch_geometric_temporal
• pandas
• (plus standard packages such as numpy, matplotlib, etc.)

Install the required packages individually:

pip install signatory torch torch_geometric torch_geometric_temporal pandas

If you encounter issues with torch or torch_geometric, refer to their official installation guides for platform-specific wheels.

How Everything Works

• configs/: Contains configuration files for data and hyperparameters.
• datasets/: Stores datasets, labels, and interfaces used for training and evaluation.
• src/: Main source code directory.
  • config.py: Loads and manages configuration files.
  • dataloader.py: Handles data loading and preprocessing.
  • graph.py: Contains graph construction utilities.
  • model.py: Defines the neural network models.
  • signature.py: Implements signature-related computations (e.g., using signatory).
  • train.py: Training loop and evaluation logic.
• main.py: Entry point for running experiments or training.
• notebook.ipynb: Example notebook for interactive exploration.
• experiments/, logs/, results/: Output directories for experiment tracking, logs, and results.

Typical Workflow

1. Edit configuration files in configs/ as needed.
2. Prepare your data in datasets/.
3. Run training or experiments via:

   python main.py

4. Check logs and results in the respective folders.

For more details on each module, refer to the docstrings in the source files.