This repository contains code for the paper "Scale-Free, Attributed and Class-Assortative Graph Generation to Facilitate Introspection of Graph Neural Networks," published at KDD MLG 2020. In particular, it enables use of the CABAM (Class Assortative Barabasi Albert Model) for graph generation, which enhances the traditional Barabasi-Albert preferential attachment model with flexible class labels, feature distributions and assortativity (homophily) parameters.
You can install a cabam package with
pip install .
You can see ./Dataset Summaries.ipynb for reproducing dataset statistics mentioned in the paper.
Data referenced in this code is sourced from versions available at graph2gauss and GAug repositories.
Please run:
bash download_data.sh
in order to download all relevant data files referenced in scripts here (all files will be placed under ./data), before trying to run notebook commands yourself.
The below is a basic example, to use CABAM to generate a graph with with 1000 nodes and minimum degree of 3, with two balanced classes (default) and roughly equal numbers of homophilous and heterophilous edges (default).
from cabam import CABAM
model = CABAM()
G, degrees, labels, num_intra_edges, num_inter_edges = model.generate_graph(n=1000, m=3)
See generate_graph(...) in ./cabam/cabam.py for customizing use.
If you use the model, or graphs generated with the model for evaluation in your own work, please cite
@inproceedings{cabam2020shah,
author = {Shah, Neil},
title = {Scale-Free, Attributed and Class-Assortative Graph Generation to Facilitate Introspection of Graph Neural Networks},
booktitle = {KDD Mining and Learning with Graphs},
year = {2020}
}