We propose a similarity classification for directed graphs. Two graphs are similar if the number of nodes with a given in-edges and out-edges are the same.
We discard graphs which contain nodes that do not participate in any edge, since those are equivalent to graphs with fewer nodes.