> Building on a foundation of simple but solid statistical algorithms from a decade ago

I wonder what algorithms they are talking about? Can't find any papers referenced :(

Looking at the clustering code it looks like they are using kd-trees with knn. Old skool!