Contents

• erdoslib.py - A Python library for creating datasets of Erdos-Reyni graphs, sampling p from normal or uniform distributions with specified parameters
• train.py - Train the GNN on various datasets specified by the control dictionary, saving weights and plotting losses.

The architecture used is a two layer Graph Convolution Network followed by a pooling into a linear layer with Sigmoid non-linearlity.

Required Packages: DGL & NetworkX

License

The code is distributed under a Creative Commons Attribution 4.0 International Public License.