For the data set I was working with it really was no complexity at all. It certainly isn't the biggest reduction of the data but it's a fairly quick reduction to do. You could certainly partition it better but this was a quick and dirty way to do it that worked for my application.