Graph Data Representation Learning is not a new concept. Traditional
methods like graphlets, hand-engineered graph feature extraction then
applying different prediction task on those feature was there. Then there is Deepwalk and
node2vec which gives better idea of neighbourhood information of graph. But deep learning on graph data
gave more robust oppurtunities to explore new domain and redefine
graph features within themselves.
This repo is used to explore pytorch geometric, a library based on torch for graph deep learning by Matthias Fey.
A GCN with Karate network Dataset is implemented.
- Karate Network Dataset Preparation
- Visualizer with Networkx
- GCN Model
- Training
- Main Script
- Inference