chapter eleven

11 Graph representation learning and graph neural networks

 

This chapter covers

  • Understanding graph representation learning and its role in scaling machine learning on graphs
  • Automating feature engineering with deep learning
  • Understanding graph embeddings and their applications
  • Working with graph neural networks

In chapters 9 and 10, we explored the fundamental concepts of machine learning (ML) on graphs, demonstrating how these techniques can solve complex tasks like node classification, link prediction, and community detection. We showed how manual feature engineering can effectively capture graph properties and relationships to power downstream ML tasks. These approaches provide insights into what makes graph-based ML work, offering transparency into how our models make decisions.

However, even simple classification tasks require significant effort to design and implement effective features. Manual approaches excel at interpretability and help build intuition, but they face significant challenges when scaled to real-world knowledge graphs (KGs) containing millions of nodes and relationships.

11.1 Embeddings in graph representation learning

11.1.1 Understanding graph embeddings: From discrete to continuous

11.1.2 Real-world applications and examples

11.2 The encoder–decoder model

11.2.1 The encoder: Converting graph structure to vectors

11.2.2 The decoder: Reconstructing graph properties

11.2.3 The power of the framework

11.2.4 Node2Vec: An example of an encoder–decoder framework

11.3 Shallow embeddings: A first approach to graph representation

11.3.1 Understanding shallow embeddings

11.3.2 Limitations of shallow embeddings

11.4 Embeddings in knowledge graphs

11.4.1 Loss function

11.4.2 Multirelationship decoder

11.5 Message passing and graph neural networks

11.5.1 The message-passing framework: A neural conversation

11.5.2 Motivation and intuition: Why message passing works

11.5.3 The basic GNN model

11.5.4 Message passing with self-loops

11.6 Generalized aggregation and update methods