Symbolic Hyperdimensional Vectors with Sparse Graph Convolutional Neural Networks

This is the repository for the code for the paper Symbolic Hyperdimensional Vectors with Sparse Graph Convolutional Neural Networks. The code is adapted from the code for the paper Optimization of Graph Neural Networks with Natural Gradient Descent, whose code can be found here. Their code has in turn adapted from github.com/rusty1s/pytorch_geometric and github.com/Thrandis/EKFAC-pytorch.

Run experiments

Clone the repository and change your current directory:

git clone https://github.com/filco306/GNN-random-indexing.git
cd GNN-random-indexing

Install

pip install -r requirements.txt

• Go to the experiments folder:

cd experiments
bash scripts/EXPERIMENTNAME

Acknowledgements

This repository is a modification of the implementation of the Optimization of Graph Neural Networks with Natural Gradient Descent. You might want to cite their paper too if you re-use this code :)