Possible to Find the Clusters One by One??

mathlover · science forum beginner Joined: 04 May 2005 Posts: 13

Hi everyone,

All the usual clustering methods I am aware of (such as k-means, Fuzzy
k-means, etc.) find all the clusters together. I mean, as you know,
they consider some cost function minimizing which (by an iterative
method for instance) finds all the clusters.

I am wondering if there is any clustering method which can find the
clusters one by one. That is it can first find just one of the
clusters, and then if the process is continued (I mean I need a method
which can be quitted after finding any number of clusters; that is, it
doesn't need to be run till finding all the clusters so as to have a
correct answer) it would find a second one, and so on. Of course, I
would trivially prefer the method to have a quality closely comparable
with the usual methods (such as k-means).

I would really appreciate it if you could please introduce me some of
these methods (if there is any).

Thanks a lot.

Google