A PyTorch Implementation of Graph Convolutional Network for Edge Prediction

Summary

This repository provides a PyTorch implementation of Stanford's tutorial on GCN for edge prediction found at: Graph Convolutional Prediction of Protein Interactions in Yeast

Requirements

• numpy~=1.20.2
• scipy~=1.6.3
• networkx~=2.5.1
• torch~=1.8.1
• sklearn~=0.0
• scikit-learn~=0.24.2

Usage

The train.py script trains the model for a specified number of epochs and performs inference at the end of training on the test set. The dataset used for this experiment is provided in the /data folder due its manageable size.

To run the script use the following command in the /src directory:

python train.py --input_path <path to dataset> --epochs <number of training epochs> --learning_rate <well, LR :)> --model_path <path to save the trained model to> --hidden_dim <hidden layer dimension> --output_dim <graph embedding dimension>