Graph Neural Networks: Merging Deep Learning With Graphs (Part I)
When It Comes to Node Classification
Recently, Graph Neural Networks (GNNs) have received a lot of attention. From marketing to social science to biology, they have been widely promoted as the new way of learning “smartly” from data. It’s more than a trend, though, as many research papers have proven that they can actually lead to more accurate and robust models.
What could possibly explain this? This is certainly due to their ability to combine graphical representation learning (which is used today for a wider variety of use cases) with the predictive power of deep learning models.
Objective
This article is the first part of three-part series that aims to provide a comprehensive overview of the most common applications of GNN models to real-world problems.
While the first focuses on node classification, the two others tackle link prediction and graph classification, respectively.
After reading this article, you will understand:
- What is graphical representation learning all about?
- What are the main mechanisms hidden under GNNs models?
- How can they be applied to real-world classification problems?