Xref: utzoo sci.math:7876 comp.graphics:7507
Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!iuvax!rutgers!usc!bbn!bbn.com!ischick
From: ischick@bbn.com (Irvin C. Schick)
Newsgroups: sci.math,comp.graphics
Subject: Re: A graph theory question
Message-ID: <45682@bbn.COM>
Date: 15 Sep 89 22:30:35 GMT
References: <45600@bbn.COM>
Organization: Bolt Beranek and Newman Inc., Cambridge MA
Lines: 24

In article <45600@bbn.COM> ischick@bbn.com (Irvin C. Schick)
I wrote:
>
>Suppose you have a finite number of nodes, and a matrix of
>"distance" measures between all pairs of nodes.
>...
>I would like to know if it is possible to construct a 2-D
>representation such that the distance between pairs of
>nodes is somehow related to the values given by the matrix.
>...

Thanks to all those who have replied. The concensus is that if these
are true distances, this problem cannot be solved in the general case
(triangle inequality and all that), but approximations can be found by
using Multidimensional Scaling, and goodness-of-fit measures for such
approximations do exist.

As the quotation marks around "distance" and the rather loose term
"somehow related to the values given by the matrix" suggest, an
approximation was all I need, so my question has been answered.

Thanks again.

Irvin