Graph Embedding For Link Prediction Using Residual Variational Graph Autoencoders

Schematic of RVGAE: a residual graph convolutional encoder produces node embeddings, and the decoder reconstructs the adjacency matrix. This is a schematic, not the paper's figure.

Abstract

Graphs are usually represented by high dimensional data. Hence, graph embedding is an essential task, which aims to represent a graph in a lower dimension while protecting the original graph's properties. In this paper, we propose a novel graph embedding method called Residual Variational Graph Autoencoder (RVGAE), which boosts variational graph autoencoder's performance utilizing residual connections. Our method's performance is evaluated on the link prediction task. The results demonstrate that our model can achieve better results than graph convolutional neural network (GCN) and variational graph autoencoder (VGAE).

Framework

Residual encoder

Residual connections are added to the variational graph autoencoder's encoder.

Low-dimensional embedding

The graph is embedded in a lower dimension while protecting its original properties.

Link prediction

Evaluated on link prediction, RVGAE beats both GCN and VGAE.

Resources

Publisher: IEEE Xplore
DOI: 10.1109/SIU49456.2020.9302271

BibTeX

@inproceedings{keser2020graph,
  title = {Graph Embedding For Link Prediction Using Residual Variational Graph Autoencoders},
  author = {Keser, Reyhan Kevser and Nallbani, Indrit and {\c{C}}al{\i}k, Nurullah and Ayanzadeh, Aydin and T{\"o}reyin, Beh{\c{c}}et U{\u{g}}ur},
  booktitle = {2020 28th Signal Processing and Communications Applications Conference (SIU)},
  pages = {1--4},
  year = {2020},
  organization = {IEEE},
  doi = {10.1109/SIU49456.2020.9302271}
}