On Thu, 06 May 2004 14:56:45 -0400, Rajarshi Guha
<[EMAIL PROTECTED]> wrote:

> Hi,
>   I've been using neural networks mainly for regression. However, is it
>   valid to consider the neural network algorithm (say for a fully
> connected feed-forward 3 layer network) as form of dimensional scaling?
> 
> If the above is a valid point of view - does it make sense to compare
> neural networks by calculating some form of Sammon or Kruskal stress for a
> trained network?

A whole lot of things are encompassed by 'neural nets' and
some of them are varieties of regression.

Find the neural-nets FAQ by Sarle, and read it.
Google gives me -

... comp.ai.neural-nets FAQ, Part 1 of 7: Introduction Maintainer:
[EMAIL PROTECTED]
(Warren S. Sarle) FAQ Home Page: ftp://ftp.sas.com/pub/neural/FAQ.html
Last ... 
www.faqs.org/faqs/ai-faq/neural-nets/ - 7k - May 4, 2004 - Cached -
Similar pages 

If you still have a question, I expect you will get more responses
if you post to the neural-nets newsgroup, if there is one.
My ISP has 67000 groups listed, but not that one.  There were
several groups that appeared to be local neural-nets groups.

Hope this helps.

-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to