---

📡 Graph Neural Networks for Joint Wireless Power Control and Spectrum Allocation

Welcome to the official repository for:

Graph Neural Networks Approach for Joint Wireless Power Control and Spectrum Allocation
By Maher Marwani and Georges Kaddoum
Published in IEEE Transactions on Machine Learning in Communications and Networking, 2024

---

🔬 Overview

This work introduces a novel Graph Neural Network (GNN)-based framework for tackling the joint power control and spectrum allocation problem in wireless communication networks. It targets Device-to-Device (D2D) communication in complex interference environments where traditional resource allocation techniques fall short due to the non-Euclidean structure of wireless topologies.

---

📊 Results & Visualizations

📍 Simulation Setup:

• Area: 500m x 500m
• Topology: 50 D2D links
• Minimum data rate constraint: 1e3
• Bandwidth: 10 Resource Blocks (RBs) of 500 Hz each

📈 Performance Visualization

The following animation showcases the evolution of data rates during the simulation:

This visual captures the adaptive behavior of the GNN-based solution, effectively managing interference, power, and spectrum resources over time.

---

📄 Citation

If you find this work useful in your research or applications, please cite it using the following BibTeX entry:

@ARTICLE{10545547,
  author={Marwani, Maher and Kaddoum, Georges},
  journal={IEEE Transactions on Machine Learning in Communications and Networking}, 
  title={Graph Neural Networks Approach for Joint Wireless Power Control and Spectrum Allocation}, 
  year={2024},
  volume={2},
  pages={717-732},
  doi={10.1109/TMLCN.2024.3408723}
}

---

🧠 Abstract

The rising complexity of wireless environments driven by modern applications and user demands challenges traditional Radio Resource Management (RRM) frameworks. Although Deep Learning (DL) approaches offer adaptive solutions, most are limited to Euclidean data structures, ignoring the graph-based nature of wireless topologies.

This work proposes a GNN-based model that directly operates on non-Euclidean representations of wireless networks, enabling efficient joint optimization of power control and spectrum allocation. The framework:

• Adapts to varying interference conditions
• Supports flexible bandwidth allocation
• Maintains robust performance in imperfect channel conditions
  Experimental results demonstrate clear advantages in convergence speed, generalization, and robustness over existing solutions.

---

⚙️ Installation & Usage

Follow the steps below to install dependencies and run simulations:

🔧 Setup

1. Clone the Repository

   git clone https://github.com/mahermarwani/Graph-Neural-Networks-Approach-for-Joint-Wireless-Power-Control-and-Spectrum-Allocation.git
   cd Graph-Neural-Networks-Approach-for-Joint-Wireless-Power-Control-and-Spectrum-Allocation
   

2. Install Required Packages

   pip install -r requirements.txt
   

---

🚀 Running the Code

1. Generate Wireless Network Topology

   python wireless/Network.py
   

2. Generate Training & Testing Data

   python data_generation.py
   

3. Train the GNN Model

   python training.py
   

4. Evaluate the Model

   python testing.py
   

5. Optional: Solve Using Genetic Algorithm

   python GA_solver.py
   

---

📬 Questions or Feedback?

Feel free to open an Issue or submit a Pull Request — contributions are welcome!

---