PetarV-/GAT
A TensorFlow implementation of Graph Attention Networks (GAT), a graph neural network layer that uses attention mechanisms for node classification tasks.
Not currently ranked — collecting fresh signals.
star history
This repository provides a reference implementation of the Graph Attention Network paper (ICLR 2018), featuring both dense and experimental sparse versions of the GAT layer in TensorFlow. It includes utilities for preprocessing the Cora dataset and PPI benchmark, along with a pretrained model achieving 84.4% accuracy on Cora’s test set. The implementation handles graph structure via adjacency matrices and uses multi-head attention for node representation learning.
Frequently asked
- What is PetarV-/GAT?
- A TensorFlow implementation of Graph Attention Networks (GAT), a graph neural network layer that uses attention mechanisms for node classification tasks.
- Is GAT open source?
- Yes — PetarV-/GAT is open source, released under the MIT license.
- What language is GAT written in?
- PetarV-/GAT is primarily written in Python.
- How popular is GAT?
- PetarV-/GAT has 3.5k stars on GitHub.
- Where can I find GAT?
- PetarV-/GAT is on GitHub at https://github.com/PetarV-/GAT.