Re: how would you write this?? (clustering algorithm)

(message (Hello 'Raffael)
(you :wrote :on '(Sat, 5 May 2007 17:26:50 -0400))
(

??>> however, k-means does not satisfy me, because i have to pass it K, and
??>> results are barely predictable.

RC> <http://en.wikipedia.org/wiki/Data_clustering>

i've read this article several times. do you know algorithm that will find
optimal K for given max radius, and will work faster than k-means?
this article mentions QT algorithm -- that will do what i want, but it would
be terribly slow, since it does each-to-each comparison many times.
agglomerative hierarchical clustering won't work for same reason -- it
involves each-to-each comparisons.

some partitional hierarchical clustering won't work because i have pretty
nasty objects -- 400-dimensional vectors with distance defined as angle
between them. this space looks very different from our usual 2 and 3
dimensional spaces -- most vectors are almost orthogonal.
i've already implemented two kinds of trees that did kinda hierarchical
clustering, and that didn't work well -- it identified some clusters, but
their radiuses were far from optimal, that's not acceptable for me.

i've analyzed data i'm working with for several months, and i came to
conclusion that some variation of k-means algorithm should work best with
it. but primitive k-means does not fully satisfy me, so i need a variation
that is better for my purposes.

)
(With-best-regards '(Alex Mizrahi) :aka 'killer_storm)
"I am everything you want and I am everything you need")


.

Relevant Pages

  • Re: NP-hardness of k-means clustering
    ... > I checked the compendium first of all, ... This one is the closest to k-means. ... clustering algorithm as implied in the *references*, ... Eray Ozkural ...
    (comp.theory)
  • Re: NP-hardness of k-means clustering
    ... > I checked the compendium first of all, ... This one is the closest to k-means. ... clustering algorithm as implied in the *references*, ... Eray Ozkural ...
    (sci.math)
  • Re: K-means and SOM
    ... Greg Heath wrote: ... >> I would like to know, in what way clustering achieved by K-means is ... there are several K-means objective ... > You'll have to search for that algorithm, ...
    (comp.soft-sys.matlab)
  • k-means clustering and data type
    ... clustering: Is it all right that if k has the same number like OTUs ... then k-means will work like UPGMA? ... Which data type needs k-means cluster algorithm? ...
    (sci.bio.misc)
Loading