Haven't tried all too many methods, a lot of the standard approaches go out the window due to the sheer size of the graph means it can't be loaded into most tools. Got 12 million nodes and 40 million edges. Tried to load it into numpy, but that was no good at all. Gonna have to code up some dimensionality reduction algorithm if I want standard tools like that.

Although you can get away with some pretty surprising stuff due to the extreme sparseness of the matrix.