SHADOWCAST is a generative model based on a conditional generative adversarial network, capable of controlling graph generation while retaining the original graph's intrinsic properties.

Please cite our paper if you find this code useful for your own work:

@article{tann2020shadowcast,
	title={SHADOWCAST: Controllable graph generation},
	author={Tann, Wesley Joon-Wie and Chang, Ee-Chien and Hooi, Bryan},
	journal={arXiv preprint arXiv:2006.03774},
	year={2020}
}

Many times, data of various situations are not available in observed real-world networks. For example, email communications in an organization between various departments. Due to limited data, previously observed network information may be missing scenarios of intra-department email surge within either the Human Resources or Accounting departments.

Given an observed graph and some user-specified Markov model parameters, SHADOWCAST controls the conditions to generate desired graphs.

Installing package requirements:

pip install -r requirements.txt

Our datasets are in the data folder.

1. EUcore-top
2. Enron
3. Cora-ML

All the experiments of reported results are in the notebooks listed below:

• eggen-EUcoretop.ipynb
• eggen-enron.ipynb
• eggen-coraml.ipynb
• control-enron.ipynb