Graph embeddings in action

Now that we are past the comfort of community detection, we are getting into some weird territory with graph embeddings. The simplest way I think of graph embeddings is just the deconstruction of a complex network into a format more suitable for ML tasks. It’s the translation of a complex data structure into a less complex data structure. That’s a simple way of thinking about it.

Some unsupervised ML models will create more dimensions (more columns/features) of embeddings than others, as you will see in this section. In this section, we are going to create embeddings, inspect nodes that have similar embeddings, and then use the embeddings with supervised ML to predict “revolutionary or not,” like our “Spot the Revolutionary” game from the last chapter.

We’re going to quickly run through the use of several different models – this chapter would be hundreds of pages long if I went into great detail...