Summary
In this chapter, we introduced graph representation learning, a fundamental concept in the domain of using ML on graph data. First, we discussed what representation learning is, in the general sense in ML. When concentrating solely on graphs, you learned that the primary objective of representation learning is to find embeddings that can emulate the structure of the graph, as well as learn important concepts that are necessary for the inference task, if any.
We also explored DeepWalk and Node2Vec, two popular graph representation learning approaches, which comprise a class of algorithms that use random walks to generate a neighborhood for a node. Based on this neighborhood, you can optimize the embedding values so that the embeddings of the nodes in the neighborhood are highly similar to those of the embeddings of the concerned node. Finally, we looked at the drawbacks of using these approaches in practice.
In the next chapter, we’ll concentrate on the most popular...
