Re: DM: equal-size clustering

[Jonathan A. Marshall]

> > > From: Hukan <hukan@cs.hku.hk>
> > >    I have a special clustering problem. Given a set of points 
>in the
> > > multidimensional space, we want to cluster these points under 
>the
> > > limition that the sizes of clusters are (almost) equal. Could 
>anyone
> > > give me some suggestions?

David Dowe wrote:
> > I would do this by MML (Minimum Message Length), and would use 
>Snob
> > http://www.cs.monash.edu.au/~dld/Snob.html
> > modified so that the relative class abundances had to be (almost) 
>equal,
> > and I would try to quantify "(almost) equal" with the best 
>Bayesian priors
> > I could.
> > No doubt, others will come up with alternative suggestions.

Warren Sarle wrote:
> K-means and numerous similar methods implicitly assume that the
> population mixing probabilities are exactly equal. One popular
> method for forcing the sample mixing proportions to be more nearly
> equal is given in: 
>    Desieno, D. (1988), "Adding a conscience to competitive 
>learning,"
>    Proc. Int. Conf. on Neural Networks, I, 117-124, IEEE Press. 

DeSieno's conscience method works by varying the sensitivity of each
classifying neuron according to the frequency with which the neuron is
activated.  Another alternative suggestion is to vary the interaction
strength between *pairs* of classifying neurons, as described in my 
paper

  Marshall JA, "Adaptive perceptual pattern recognition by 
self-organizing
  neural networks: Context, uncertainty, multiplicity, and scale." 
Neural
  Networks, 8:335-362, April 1995. 

The theory and benefits of this method are described in a new paper,

  Marshall JA, Gupta VS, "Generalization and exclusive allocation of 
credit
  in unsupervised category learning."  Submitted for journal 
publication,
  34 pp., August 1997. 

Both papers are available via my web page, 
http://www.cs.unc.edu/~marshall.

Another method is Nigrin's SONNET-2, which sets the the interactions 
between
pairs of *input elements*.  This is described in Nigrin's 1993 book, 
Neural
Networks for Pattern Recognition (MIT Press).

--Jonathan

   _____
  /     \     Jonathan A. Marshall                       
marshall@cs.unc.edu
  -------     Dept. of Computer Science      
http://www.cs.unc.edu/~marshall
  | | | |     CB 3175, Sitterson Hall
  | | | |     Univ. of North Carolina                 Office 
+1-919-962-1887
  =======     Chapel Hill, NC 27599-3175, USA            Fax 
+1-919-962-1799