wgmueller1 / fastgae

Source code from the article "FastGAE: Scalable Graph Autoencoders with Stochastic Subgraph Decoding" by G. Salha, R. Hennequin, J.B. Remy, M. Moussallam and M. Vazirgiannis (2020)

FastGAE: Scalable Graph Autoencoders with Stochastic Subgraph Decoding

This repository provides a Tensorflow implementation of the FastGAE framework, introduced in the article FastGAE: Scalable Graph Autoencoders with Stochastic Subgraph Decoding. This framework aims to speed up Graph Autoencoders (AE) and Graph Variational Autoencoders (VAE), and at scaling these models to large graphs with millions of nodes and edges.

This work has been accepted for publication in Elsevier's Neural Networks journal.

We provide the FastGAE and Variational FastGAE models from our paper, together with standard Graph Autoencoders and standard Variational Graph Autoencoders from the original article of Kipf and Welling (2016).

We evaluate our framework on the link prediction and node clustering tasks introduced in the paper. We provide the Cora, Citeseer, Pubmed and Google datasets in the data folder. The three additional graphs used in the paper (SBM, Patent and Youtube) will later be provided via external links due to size constraints. We refer to section 4 of the paper for more information about datasets.

Our code builds upon Thomas Kipf's original implementation of graph AE and VAE, and upon previous research works from our team.

Installation

python setup.py install

Requirements: tensorflow (1.X), networkx, numpy, scikit-learn, scipy

Run Experiments

cd fastgae
python train.py --model=gcn_ae --dataset=pubmed --task=link_prediction --fastgae=True --measure=degree --alpha=1.0 --nb_node_samples=1000
python train.py --model=gcn_ae --dataset=pubmed --task=link_prediction

The above commands will train a graph AE on the Pubmed dataset, with (line 2) and without (line 3) the FastGAE framework, decoding random subgraphs of 1000 nodes, drawn via degree-based sampling with alpha=1.0. Models are evaluated on the Link Prediction task, with all parameters set to default values.

As detailed in Table 2 of the paper, line 2 returns competitive average performances w.r.t. line 3 i.e. the standard graph AE, while being significantly faster. We recommend GPU usage for faster learning. Use the --nb_run option to average results over multiple runs.

Complete list of parameters

Parameter	Type	Description	Default Value
model	string	Name of the model, among: - gcn_ae: Graph AE from Kipf and Welling (2016), with 2-layer GCN encoder and inner product decoder - gcn_vae: Graph VAE from Kipf and Welling (2016), with Gaussian distributions, 2-layer GCN encoders for mu and sigma, and inner product decoder	gcn_ae
dataset	string	Name of the dataset, among: - cora: scientific publications citation network - citeseer: scientific publications citation network - pubmed: scientific publications citation network - google: google.com web graph You can specify any additional graph dataset, in edgelist format, by editing input_data.py	cora
task	string	Name of the Machine Learning evaluation task, among: - link_prediction: Link Prediction - node_clustering: Node Clustering See section 4 of paper for details about tasks	link_prediction
fastgae	boolean	Whether to use the FastGAE framework	False
nb_node_samples	int	For FastGAE: number of nodes to sample at each iteration, i.e. sampled subgraph size	1000
measure	string	For FastGAE: node importance measure used in sampling, among degree, core and uniform	degree
alpha	float	For FastGAE: alpha hyperparameter for core/degree-based sampling	2.0
replace	boolean	For FastGAE: whether to sample nodes with (True) or without (False) replacement	False
dropout	float	Dropout rate	0.
iterations	int	Number of training iterations	200
features	boolean	Whether to include node features in encoder	False
learning_rate	float	Initial learning rate (with Adam optimizer)	0.01
hidden	int	Number of units in GCN encoder hidden layer	32
dimension	int	Dimension of encoder output, i.e. embedding dimension	16
nb_run	integer	Number of model runs + tests	1
prop_val	float	Proportion of edges in validation set (for Link Prediction)	5.
prop_test	float	Proportion of edges in test set (for Link Prediction)	10.
validation	boolean	Whether to report validation results at each epoch (implemented for Link Prediction task)	False
verbose	boolean	Whether to print full comments details	True

Cite

Please cite our paper if you use this code in your own work:

@article{salha2021fastgae,
  title={FastGAE: Scalable graph autoencoders with stochastic subgraph decoding},
  author={Salha, Guillaume and Hennequin, Romain and Remy, Jean-Baptiste and Moussallam, Manuel and Vazirgiannis, Michalis},
  journal={Neural Networks},
  volume={142},
  pages={1--19},
  year={2021},
  publisher={Elsevier}
}