GCN for Cora Dataset

This project implements a Graph Convolutional Network (GCN) to classify papers in the Cora citation network dataset using PyTorch Geometric. The implementation is provided in a Jupyter Notebook.

Overview

• Implements a GCN model for node classification
• Uses the Cora dataset from PyTorch Geometric
• Visualizes training loss and node embeddings

Requirements

• Python 3.x
• Jupyter Notebook
• PyTorch
• PyTorch Geometric
• Matplotlib
• Seaborn

Usage

1. Install the required packages:

   pip install jupyter torch torch_geometric matplotlib seaborn
   

2. Start Jupyter Notebook:

   jupyter notebook
   

3. Open the Cora_gnn.ipynb file in the Jupyter Notebook interface.

4. Run all cells in the notebook to train the model and generate visualizations.

Results

• Test Accuracy: 0.703
• The notebook generates and displays:
  • A plot of the training loss
  • A GIF showing the evolution of node embeddings

Visualizations

Training Loss

Node Embeddings Evolution

Future Work

• Experiment with different GNN architectures
• Apply to other citation network datasets

Author

chirag1701

License

This project is open-source and available under the MIT License.