Speedup Mean Shift Clustering

[Kevin]

Hello,

I've started to use the mean shift clustering algorithm and I am wondering
if it would be beneficial to include the functionality of
MeanShiftCanopyMapper.map(...) in the
MeanShiftCanopyCreatorMapper.map(...). This way the MapReduce job used to
create canopies from vector would not only transform all the vectors to
canopies, but also, in the same job, start merging canopies. The first
iteration of building the clusters would be done inside the
MeanShiftCanopyMapper instead of just transforming vectors to canopies.

Any inputs about this?

Thanks,
-Kevin