It would be great if there were more ways to evaluate the performance of clustering methods when they are calculated. This could be done very simply (offering default calculation of a small set of metrics) or with a more advanced wrapper (maybe letting someone configure their own scorer).
