Eigencuts version of spectral clustering

[Andrew Musselman]

Looks like this is finished as of May of this year, but is there still the
"bottleneck" performance issue with it?  I.e., is it useful in production?

Thanks
Andrew