Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!mit-eddie!ll-xn!ames!sdcsvax!ucsdhub!hp-sdd!artecon!donahue
From: donahue@artecon.artecon.UUCP (Brian D. Donahue)
Newsgroups: comp.ai
Subject: Re: Is Computer Science Science?
Message-ID: <598@artecon.artecon.UUCP>
Date: Tue, 15-Sep-87 17:31:39 EDT
Article-I.D.: artecon.598
Posted: Tue Sep 15 17:31:39 1987
Date-Received: Fri, 18-Sep-87 00:44:13 EDT
References: <5113@sunybcs.UUCP> <6195@apple.UUCP> <5068@jade.BERKELEY.EDU> <737@elmgate.UUCP>
Reply-To: donahue@artecon.UUCP (Brian D. Donahue)
Distribution: na
Organization: artecon
Lines: 41
Keywords: computers Science not bloody likely

In article <737@elmgate.UUCP> ram@elmgate.UUCP (Randy Martens) writes:
>I am of the firm opinion that there is NO such thing as
>computer science.  To quote (and I have forgotten the attribution)
>"Computer Science bears the same relationship to Real Science, that
>plumbing bears to Hydrodynamics."
>
>There is, however, Computer Engineering. (and Software Engineering,
>and Systems Engineering etc.).  Science is the discovery of the new.
>

From Webster's New World Dictionary:
science n. [<L. scire, know]
1. systematized knowledge derived from observation, study, etc.  
2. a branch of knowledge, esp. one that systematizes facts, principles, 
and methods  3. skill or technique

From the above definition, it seems fair & reasonable to call the field 
of computers, computer SCIENCE.  After all, it is a branch of knowledge.
On the other hand, there's
From Webster's New World Dictionary:
engineering n. 1. the putting of scientific knowledge in various branches 
to practical uses

According to the above definition, there is SCIENTIFIC knowledge before
engineering.

Not that any of this belongs in comp.ai, but can we get back to neural nets?
Does anyone know of a good introductory article/book to neural networks?
How do they "learn" anyways?  From what I understand, NN's basically tweak
the coefficients of your basic model/equation for the best fit to 
your "answer".  What if you've got several thousand sets of data/"answers",
and there really isnt a very good single "fit"?  Can they detect certain
characteristics that might indicate that answer X might be more probable
than answer Y?  How do nueral nets compare with multiple regression
models of statistics/probability?  

buenos adidas,
bd


---
no silly quotes here		{seismo, sdcsvax}!esosun!artecon!donahue