Graph Learning Published

Graph Autoencoder Framework

Residual Variational Graph Autoencoders for Representation Learning

Indrit Nallbani, Aydin Ayanzadeh et al.
Istanbul Technical University
2020 - 2021
View Paper (arXiv) IEEE Paper

Overview

This project investigates residual connections in graph autoencoders for improved representation learning on graph-structured data, contributing to the field of graph neural networks and unsupervised learning.

By incorporating residual connections into variational graph autoencoders, the model achieves better gradient flow and learns more expressive node embeddings for downstream tasks like link prediction and node classification.

Key Contributions

Residual Graph Layers

Novel architecture combining residual connections with graph convolutions.

Link Prediction

State-of-the-art performance on graph link prediction benchmarks.

Node Embeddings

Learn rich, low-dimensional representations of graph nodes.

Variational Approach

Probabilistic framework for robust graph representation learning.

Technologies Used

PyTorch Geometric Python Graph Neural Networks Variational Autoencoders NetworkX Cora/CiteSeer

Publications

arXiv Preprint (2021)

Representation Learning Using Graph Autoencoders with Residual Connections.

View Paper

IEEE SIU Conference 2020

Graph Embedding For Link Prediction Using Residual Variational Graph Autoencoders.

View Paper

Interested in Graph Learning?

Let's discuss graph neural networks and their applications.

Get in Touch