Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!news.kei.com!bloom-beacon.mit.edu!spool.mu.edu!howland.reston.ans.net!swrinde!elroy.jpl.nasa.gov!sdd.hp.com!hp-pcd!hpcvsnz!briand
From: briand@cv.hp.com (Brian Dixon)
Subject: Re: how does it work in brain?
Sender: news@hpcvsnz.cv.hp.com (News )
Message-ID: <D9t6M6.7ID@hpcvsnz.cv.hp.com>
Date: Wed, 7 Jun 1995 15:20:30 GMT
References: <kovskyD93ArJ.BHs@netcom.com>
Nntp-Posting-Host: hpcvsgen.cv.hp.com
Organization: Hewlett-Packard
X-Newsreader: TIN [version 1.1 PL9.4]
Lines: 74

Bob Kovsky (kovsky@netcom.com) wrote:
: In article <3p8h6i$sck@nntp.interaccess.com>,
: Corey Sweeney <corey@interaccess.com> wrote:
: >i am trying to understand all this stuff.  Basicaly ANN's are a loose 
: >attempt at emulating the human brian. 
: ...
: ...
: > ...But i fail to 
: >see how some people could model reality (eg many scientists or computer 
: >programmers will work with a patern of logic, rather then a "i like this, 
: >i don't like that" approach.  Since this is not what people are hardwired 
: >for, how could this be?

You should be reading Bart Kosko on fuzzy logic and how 'logic' is actually
a subset or extreme case of fuzzy logic.  Programmers and traditional 
scientists may pick traditional logic because "it's easier." (Bart Kosko's
quote, but I agree that the attitude is there and built into the 'system'.)

: >(note: i'm not trying to make any kind of statement about what kinds 
: >of people are right here, i'm just trying to figure out the neurological 
: >model, so please don't respond about that kind of stuff, only respond 
: >about neurological kind of stuff)

: 	When hydraulic systems were in vogue, people said the brain was a 
: hydraulic system.  When clockwork was in vogue, people said the brain was 
[snip]
: computer.  Now artificial neural networks are in vogue...  

: 	Of course, some neural networkers will say that the architecture 
: of the neural network mimics that of the brain.  Indeed, that's why the 
: name "neural network" was chosen.  T'ain't so. Artificial neural networks 
: ignore half of brain function:  the chemical environment which modifies 
: the electrical activity and which is, apparently, modified in turn.  

You missed the point.  The adjustment of weights and bias' in an ANN is
*designed to mimic neuronal elasticity*.  Study the text From Neuron to
Brain by Nichols, Martin, and Wallace.  The rest of the "electrical
activity", the connectionist views and transmission issues, are indeed
modeled by ANNs.  The primary difference between how ANNs and NNs such as
the brain is *how* learning takes place, that is, how neuronal elasticity is
accomplished.  In ANNs, weights and bias' are modified by an artificially
created algorithm, sic back propagation.  In NNs, elasticity is accomplished
by changes in synaptic efficacy, which in turn is changed by means of 
changing availabilities of neurotransmitters in the presynaptic cell.

: Artifical neural networks, in general, incorporate energy functions that 
: have no correspondence to brain function.  To quote Hertz, Krogh & 

You don't *know* this.  How the brain (as a NN) achieves memory is only
partially known and is still in research.

[snip]
: evolves according to its dynamical rule."  I am not aware of any ANN 
: whose operations resemble the high energy consumption of the brain (the 
: brain consumes about 20% of the body's energy).  This is related to the 
: fact that ANN's are relaxing to stable states, contrary to the activity 
: of the brain.

Wrong again.  Check your CPU usage while you sit back watching hundreds or
thousands of learning epochs go cranking by....see how much is left for
other processes running on the same machine.  You cannot compare an average
energy consumption of a brain to anything.  You must be able to differentiate
between modes of operation: inferring or learning.  I think you'll see
the similarities appear if you'll take the time to do that.

: 	In other words, ANN's may superficially look like a brain, but 
: they don't talk like a brain and they don't walk like a brain.

All of life is a belief system.  Choose what you believe, then find the
evidence to support it.  Is that how it works???

--
Brian Dixon, Machine Vision Engineer, Hewlett Packard (Corvallis, Oregon)
503-715-3143 (wk), briand@cv.hp.com (email). "Opinions & attitudes are mine!"
