Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!ix.netcom.com!netcom.com!ddyer
From: ddyer@netcom.com (Dave Dyer)
Subject: Re: WANTED-Clustering Code
Message-ID: <ddyerCzMtvF.5pH@netcom.com>
Reply-To: ddyer@netcom.com
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
X-Newsreader: TIN [version 1.2 PL1]
References: <39tgt5$mid@venus.nist.gov>
Date: Mon, 21 Nov 1994 19:23:39 GMT
Lines: 24


I don't have C code handy, but there's an almost magical 
cluster finding algorithm that uses a n-dimensional peano 
curve to find clusters in n-dimensional space.   

In Phase 1, you histogram the data points with some resolution you
can afford. 

In Phase 2, you walk through the histogram using a peano curve as a
path; Clusters in the original N-dimensional space appear as easily
recognizable 1-dimensional clusters during the walk.


Here are the primary references:

  "Algorithms developed from R. J. Stevens, et.al.  
  "Manipulation and Presentation of Multidimensional Image Data Using the Peano Scan",
  IEEE Trans on Pattern Anal and Mach Intell, Sept. 1983.

  Multi-dimension Peano scan  developed from A. R. Butz,
  "Alternative Algorithm for Hilbert's Space-Filling Curve",
   IEEE Trans on Computer Graphics, April 1971.


