Chapter 2. The Graph Machine Learning Pipeline
In Chapter 1, we gave you a glimpse into the world of graph learning in enterprise settings. We talked about its various use cases and wrapped up the chapter by highlighting the challenges it poses. As we move forward in the book, we’re going to give you a high-level overview of the entire GML pipeline that we’ll be exploring. From start to finish, this pipeline covers everything from collecting graph data to serving it for GML.
In this chapter, we’ll start with the foundation stage: the graph data pipeline. This initial phase, crucial for the entire process, refers to a series of processes designed to efficiently handle, process, and analyze data structured in the form of graphs. It sets the groundwork for everything that follows. Next, we’ll discuss the GML training pipeline, a series of steps designed to train and evaluate a chosen graph learning algorithm for a specific graph task, such as node classification or link prediction. Finally, we’ll go through the graph inference pipeline, which tests whether the new model is ready for production promotion. If it is, the model is deployed to be utilized by downstream systems. For example, in building an ecommerce graph recommender system, the model is integrated into the bigger ecommerce solution to provide personalized recommendations for the platform’s users. The three subpipelines are illustrated in Figure 2-1.
Figure 2-1. The graph learning pipeline is broken down into three ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access