Re: [PATCH] Quick Cluster: Code tidy up.

[John Darrington]

If you save my original email (say in a file called "johnspatch")
and then in the top level directory of the source code, type:

patch -p 1 < johnspatch

then it should apply the patch.  For details:   man patch


I've also been looking at the algorithm for the initial clusters.
I think the one given by IBM is correct, but with a slight modification
to break ties when a case is equidistant between two groups.

J'

On Sat, Nov 07, 2015 at 02:22:21PM -0600, Alan Mead wrote:
     John,
     
     How do I apply this as a patch so I can see the final outcome?  I think
     the changes you describe would be good.
     
     FWIW, in a previous email I described the intended operation of the
     Quick Cluster algorithm incorrectly.  There is no randomness in how the
     centers are chosen although the algorithm scans the database so the
     order of the data matters (it's not clear, but the algorithm might pick
     the same centers even with different orderings).

signature.asc
Description: Digital signature