Graph Representation Learning

Having explained why applying deep learning techniques to graph data is a worthy endeavor, let’s jump right into the thick of things. In this chapter, we’ll introduce you to graph representation learning.

First, we’ll examine representation learning from the perspective of traditional (tabular data-based) machine learning (ML) and then extend the idea to the graph data space. Following this, we’ll talk about the initial challenges that need to be addressed when you’re trying to learn features within graph data. Next, you’ll be introduced to a few simple graph representation learning algorithms, namely, Node2Vec and DeepWalk, and understand the differences between them. Finally, we’ll discuss the limitations of such shallow encoding techniques and why we need algorithms with more firepower to capture more complex relationships in graphs.

We’ll also introduce implementations of relevant algorithms...