Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!gatech!concert!sas!mozart.unx.sas.com!saswss
From: saswss@hotellng.unx.sas.com (Warren Sarle)
Subject: Re: Salmon mappings and visualisation.
Originator: saswss@hotellng.unx.sas.com
Sender: news@unx.sas.com (Noter of Newsworthy Events)
Message-ID: <CzMsA9.L84@unx.sas.com>
Date: Mon, 21 Nov 1994 18:49:21 GMT
References: <1994Nov11.111331.118@rmcs.cranfield.ac.uk> <3agqg5$2im@bigfoot.wustl.edu>
Nntp-Posting-Host: hotellng.unx.sas.com
Organization: SAS Institute Inc.
Lines: 26


In article <3agqg5$2im@bigfoot.wustl.edu>, omar@wucs1.wustl.edu (Omar El-Ghazzawy) writes:
|> ...
|> It is commonly refered to as "Sammon's Algorithm" for non-linear mapping.
|> It is a non-linear algorithm for dimensionality reduction.  The goal is to
|> map datapoints from N-D down to 2 or 3-D such that inter-point distances
                                                      ^^^^^^^^^^^^^^^^^^^^^
Note that this is quite different from autoassociative dimensionality
reduction--not better or worse, just different.

|> are preserved - thus preserving the "structure" of the data while reducing
|> to a dimesionality that can be plotted on a 2-D or 3-D scatterplot.
|>
|> The algorithm is order N**2, so it becomes troublesome for mapping more than,
|> say, 1000 datapoints on most workstations and PC's.  But for smaller (or
|> sampled) datasets, it provides a better(??) mapping than Principal Components.

You can do essentially the same thing with any of the multidimensional
scaling programs available in various statistical software packages
and as stand-alone programs.

-- 

Warren S. Sarle       SAS Institute Inc.   The opinions expressed here
saswss@unx.sas.com    SAS Campus Drive     are mine and not necessarily
(919) 677-8000        Cary, NC 27513, USA  those of SAS Institute.
