GeoFed

Environment Requirements

Python Version

• Python 3.8+

Required Packages

• PyTorch 1.12+
• PyTorch Geometric 2.0+
• NumPy
• SciPy
• scikit-learn
• scikit-network
• pymetis
• geoopt
• tqdm

Installation

pip install torch torch-geometric numpy scipy scikit-learn scikit-network pymetis geoopt tqdm

Quick Start

Basic Usage

python main.py

Configuration

The main configuration is set in main.py. You can modify the following parameters:

• args.dataset: Dataset to use (e.g., "sbm", "Cora", "CiteSeer")
• args.simulation_mode: Data partitioning method ("subgraph_fl_louvain" or "subgraph_fl_metis_plus")
• args.num_clients: Number of federated clients
• args.num_rounds: Number of federated learning rounds

Notes

• The project uses CUDA by default. Set args.use_cuda = False in main.py if you want to use CPU only.
• Dataset files should be placed in the dataset/ directory.