Skip to content
Gumbel Graph Network (GGN) : A General Deep Learning Framework for Network Reconstruction
Branch: master
Clone or download
Latest commit f1d5070 Apr 10, 2019
Type Name Latest commit message Commit time
Failed to load latest commit information.
data update Apr 10, 2019
img add initial readme Mar 2, 2019
utils update Apr 10, 2019 update Apr 10, 2019 update Apr 10, 2019 update Apr 10, 2019 update train process Mar 11, 2019

A General Deep Learning Framework for Network Reconstruction

This repository will contain the official PyTorch implementation of:

A General Deep Learning Framework for Network Reconstruction.
Zhang Zhang, Yi Zhao, Jing Liu, Shuo Wang, Ruyue Xin and Jiang Zhang*(*: Corresponding author)


Recovering latent network structure and dynamics from observed time series data are important tasks in network science, and host a wealth of potential applications. In this work, we introduce Gumbel Graph Network (GGN), a model-free, data-driven deep learning framework to accomplish network reconstruction and dynamics simulation. Our model consists of two jointly trained parts: a network generator that generating a discrete network with the Gumbel Softmax technique; and a dynamics learner that utilizing the generated network and one-step trajectory value to predict the states in future steps. We evaluate GGN on Kuramoto, Coupled Map Lattice, and Boolean networks, which exhibit continuous, discrete, and binary dynamics, respectively. Our results show that GGN can be trained to accurately recover the network structure and predict future states regardless of the types of dynamics, and outperforms competing network reconstruction methods.


  • Python 3.6
  • Pytorch 0.4

Data Generation

To generate experimental data, you need to switch to the / data folder and run the corresponding file.

cd data

Run Experiment

You can replicate the experiment for Boolean Network by simply running the file


To replicate the experiment for Coupled Map Lattice and Kuramoto model, please run the

python --simulation-type cml --dims 1 --skip 0


python --simulation-type kuramoto --dims 2 --skip 1


If you use this code in your own work, please cite our paper:

  title={A General Deep Learning Framework for Network Reconstruction},
  author={Zhang, Zhang and Zhao, Yi and Liu, jing and Wang, Shuo and Xin, Ru-Yue and Zhang, Jiang},
  journal={arXiv preprint arXiv:1812.11482},
You can’t perform that action at this time.