A curated collection of resources, libraries, papers, and learning materials on Graph Neural Networks (GNNs).
Graph Neural Networks (GNNs) are transforming AI by enabling deep learning on graph-structured data. This repository is a comprehensive hub for researchers, engineers, and enthusiasts who want to explore and advance in this field.
Here are some popular and well-maintained libraries for GNNs:
- PyG (PyTorch Geometric) – A powerful library for deep learning on graphs using PyTorch.
- DGL (Deep Graph Library) – A scalable library for deep learning on graphs supporting multiple backends (PyTorch, TensorFlow, MXNet).
- Spektral – A deep learning library for graphs built on TensorFlow and Keras.
- Graph Nets – DeepMind’s library for building graph-based neural networks in TensorFlow.
- NetworkX – Though not specifically for deep learning, NetworkX is essential for working with graph data structures.
- PyGOD - Graph anomaly/outlier detection.
(More libraries will be added. Feel free to suggest additions!)
Some must-read papers for understanding Graph Neural Networks:
- Graph Isomorphism Networks (GIN) – Xu et al., 2019
- Graph Attention Networks (GAT) – Veličković et al., 2018
- Semi-Supervised Classification with Graph Convolutional Networks (GCN) – Kipf & Welling, 2017
- Benchmarking GNNs – Dwivedi et al., 2020
- Graph Transformer Networks (GTN) – Yun et al., 2019
- Neural Message Passing – Gilmer et al., 2017
(More papers are continuously added. Feel free to contribute!)
Explainability is crucial for making GNNs interpretable and trustworthy. Here are key papers and tools in Explainable AI (XAI) for GNNs:
- A Survey on Explainability of Graph Neural Networks – Kakkad et al., 2023
- RCExplainer – Bajaj et al. 2021
- PGExplainer – Luo et al., 2020
- GraphLIME – Huang et al., 2020
- GNNExplainer – Ying et al., 2019
💡 Tools:
- Captum – PyTorch’s library for model interpretability.
- GNNExplainer Implementation – Official implementation of GNNExplainer.
Learn more about GNNs through these insightful videos:
- Stanford CS224W: Machine Learning with Graphs | 2021 | Lecture 1.1 - Why Graphs – Jure Leskovec
(Have a great video to suggest? Open a PR!)
If you're new to GNNs or want a structured learning path, check out these tutorials:
- Stanford CS224W: Machine Learning with Graphs – Course Materials
Benchmark datasets commonly used for training and evaluating GNNs:
- SNAP – Stanford Large Network Dataset Collection
- Open Graph Benchmark (OGB) – Standardized datasets for GNN research.
- TUDataset
- Planetoid – Commonly used citation datasets (Cora, Citeseer, PubMed).
Real-world applications of GNNs:
- Drug Discovery & Bioinformatics – GNNs help predict molecular interactions.
- Recommender Systems – Used by platforms like Pinterest and Amazon.
- Social Network Analysis – Understanding relationships in networks like Twitter and Facebook.
- Cybersecurity & Fraud Detection – Identifying fraudulent transactions and network intrusions.
We welcome contributions! Follow these steps to contribute:
1️⃣ Fork This Repository
Click the "Fork" button at the top right to create your own copy.
2️⃣ Clone Your Fork
git clone https://github.com/your-username/gnn-atlas.git
cd gnn-atlas3️⃣ Create a New Branch
git checkout -b feature/new-topic4️⃣ Make Your Changes & Commit
git add .
git commit -m "Added new GNN paper on XAI"5️⃣ Push to Your Fork
git push origin feature/new-topic6️⃣ Open a Pull Request (PR)
Go to the original repo and submit a Pull Request with a clear description of your changes.
📢 Looking for suggestions? Check the issues tab for tasks!
This repository is licensed under the MIT License.