I don't know if this observation is that helpful, but to me it almost seems as if in this process you are trying to construct the smallest k n-dimensional "spheres" that encompass the data. (I say "spheres" loosely-a spheroid or ellipsoid might be more appropriate, upon thinking a little further about it.) When projected into one dimension, this would seem to appear as a range; in 2 dimensions, as a circle; and in 3+ dimensions, as a sphere. The "amorphous blob" effect you mentioned in the CB when I made this observation earlier could be a result of overlaps in the spheres, and choosing one to encompass a particular data point than another.
As I said, I don't know how helpful the observation may be, but hope it helps.
In reply to Re: Making sense of data: Clustering OR A coding challenge
by atcroft
in thread Making sense of data: Clustering OR A coding challenge
by belg4mit
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |