FedCHG: Graph Autoencoder Enhanced Federated Learning for Cross-Domain Heterogeneous Graph This repository contains the official PyTorch implementation of FedCHG, a novel Federated Graph Learning (FGL) framework designed to tackle Cross-Domain Heterogeneity

FedCHG_Repo/ ├── run_exp.py # [Key Script] Automated scheduler ├── requirements.txt # Python environment dependencies ├── README.md # You are here ├── scripts/ │ └── download_data.sh # Helper script for downloading large datasets (DGraph/Reddit) └── src/ ├── init.py ├── main.py # Main entry point for single-scenario experiments
├── client.py # Client-side logic ├── server.py # Server-side logic ├── models.py # Model definitions ├── features.py # Structural Feature Extraction ├── data_loader.py # Dataset loading, splitting, and heterogeneity simulation ├── training.py # Training └── utils.py # Metrics, Logging, T-Test, and Visualization tools

🛠️ Environment Setup Dependencies are sensitive for Graph Neural Networks. Please install strictly according to the versions below to avoid torch_geometric compatibility issues. Create a Conda Environment: Bash conda create -n fedchg python=3.8.13 conda activate fedchg

Install Dependencies: Bash pip install -r requirements.txt

Key libraries: torch-geometric==2.0.4, numpy, scipy, scikit-learn, seaborn.