Predicting missing links in a graph
Link prediction, also known as graph completion, is a common problem when dealing with graphs. More precisely, from a partially observed graph—a graph where for a certain pair of nodes it is not possible to exactly know if there is (or there is not) an edge between them—we want to predict whether or not edges exist for the unknown status node pairs, as seen in Figure 5.1. Formally, let 
be a graph where 
is its set of nodes and 
is its set of edges. The set of edges 
are known as observed links, while the set of edges 
are known as unknown links. The goal of the link prediction problem is to exploit the information of 
and 
to estimate 
. This problem is also common when dealing with temporal graph data. In this setting, let 
be a graph observed at a given timepoint 
, where we want to predict the edges of this graph at a given timepoint 
. The partially observed graph can be seen here:
Figure 5.1 – Partially...
