Graph-Powered Machine Learning

appendix C. Graphs for processing patterns and workflows

In many machine learning projects, including many of those described in this book, the graphs produced are large. The scale of these graphs makes processing them efficiently difficult. To deal with these challenges, a variety of distributed graph processing systems has emerged. In this appendix, we will explore one of these systems: Pregel, the first computational model (and still one of the most commonly used) for processing large-scale graphs.¹ This topic suits the purpose of the appendix for two main reasons:

• It defines a processing model that’s useful for providing an alternative implementation of some of the algorithms discussed in this book (both graph-based and non-graph-based). ...