Hi Rick, Erin, all,
> ERIN: My understanding of neural networks is a little different. The
> older
> conceptual models had a node equalling a concept but in neural
> networks the
> info is distributed across a set of nodes. (I am not sure but I
> thought it was
> the features that were the nodes.) They thought the older models
> didn't make
> sense with the limited number of neurons a person has available. So
> saying a
> there is a cat neuron seemed silly and thus the neural network
> interest is hot
> but it does not want a node = concept.
What I refering to was the 'grandmotherneuron' theory, which is indeed a
silly model. It's saying that in order to recognize each thing or person
(e.g. your grandmother), there is one neuron in the top of the hierarchy
representing each thing or person. But it seems you're well aware of
this theory. I'm familiar with some neural network theories describing
stimulus response phenomena, by Schmajuk and others, and in order to
represent 'configural' cues consisting of different stimuli, a hidden
unit is added to represent each configural cue (e.g. the conf. cue
saying 'this is a light and a tone compound'). That seems to me still
the grandmotherneuron theory. To be honest, I'm not that familiar with
distributed neural networks, although it's intuitive a better model for
representing things. A representation of a CAT is then distributed among
different nodes, maybe with the features as elements. I'm on slippery
ice here, but it seems to me that MATHEMATICALLY there's no real
difference between the gr.m.theory and the distributed one. To
strenghten my intuition, consider the fact that distributed neural
networks are implemented on computers, or Turing machines. So one Turing
machine can perform all the functions a distr. neural network can. And a
Turing machine performs just one step at the time... that's silly as a
plausible model for representations, too, I think.
A lot of networks are trained to give some output after given some
input. The weights are randomly adjusted at first, and if the output
given by the network is closer to the desired output, the weights are
changed accordingly. I've read it's quite a struggle afterwards to try
to locate 'representations' in a trained network. But theoretically, I
deem, there are only few possible ways to do the job, and in most cases
you will have to infer the kind of representations I just mentioned.
Hm... could discuss all I know about the non-brittleness (spelling
correct?) of distributed networks, and that that is more plausible a
model too, but I think I'm straying already a bit.
>
>
> >RICK
> >Okay. I think I understand this (except maybe the 'prototype' idea).
> >
>
> ERIN: An example. A canary is more prototypical bird then an ostrich.
> It
> takes longer for somebody to answer is "is a canary a bird" then "is
> an
> ostrich a bird"
>
Yes.
RICK
Hey Patrick,
Always great to hear from you....
PATRICK
O, thanks!
PATRICK
That's psychology. Now MoQ.
Maybe the perception of the true color (without names) of things are
close to dynamic quality. A static quality emerges when a node or neuron
will fire saying this is 'green', 'blue', 'bleen', or 'glue'. That is a
static quality which we can use to think and/or do things. Concepts are
used
as tools in thought and action.
RICK
I'm not quite sure what this means. Is this a vote for
'unvisulaized'
or
'unrecognized' (I think you're saying 'unrecognized', but I can read it
either way).
I meant 'unrecognized'. Hm... I'm sorry to admit I haven't read all
mails in this thread, it's just too much for me to follow them all and I
just randomly scan some. But do you mean by 'unvisualized' that our
undivided field of consciousness (to use Nishida Kitaro's terminology),
uh, is indeed undivided and that the perception of a field of colors in
front of us might not exist if you move the static-qualities-filter VERY
close to DQ? Catch my drift? I mean, before I think in words I CAN
perceive a field of colors, that is, in perception I differentiate
between the different colors. Are there static qualities in perception
as well, and not only in the intellect? Incidentally, *what* is Dynamic
Quality? I asked Andrea if she meant that DQ perhaps is transcendental,
that is, beyond or before our field of perception, but I didn't get an
answer. I'm aware that it's a bit naive question, but sometimes it seems
in the MOQ that DQ is meant to be simply something that is dynamic in
the sense that there is some kind of *movement* in consciousness. That's
dynamic in the ordinary sense, but I don't think Pirsig was refering to
this use of the word!
RICK
The part about, "...a static quality which we can use to think
and/or do things. Concepts are used as tools in thought and action...."
sounds like Pragmatism... I like that.
PATRICK
The question is how we divide the field of perception or maybe
consciousness
in it's total to do/think with it as static qualities.
RICK
I don't get this...
Could you elaborate some more?
PATRICK
I was trying to say in different words what I wrote just before about
the pragmatism. Correct to name it that, by the way! Hm, I just tried to
elaborate here, but I ended up with a similar formulation as I made
above! But let me try it in a different way. The perception of a cat
scratching a tree or climbing in it is, uh, dynamical, as opposed to
static in the ordinary sense. Every moment the scenery of the tree and
the cat change: The tree is moving in the wind, the cat is moving on a
branch, you see it's head turn, sometimes you don't or do see it's eyes,
etc. We transform the scenery into static qualities by thinking 'The CAT
is in the TREE'. And when you know you're talking about a cat and a
tree, you can make propositions about these things. The thing is, you're
inclined to use the static qualities (maybe a prototype...) CAT and TREE
when thinking about them instead of seeing the moving cat and the tree
blowing in the wind; the things as they really are. Like the famous
Eastern (Buddhist?) saying: You shouldn't confuse the finger pointing to
the Moon with the Moon itself... but our language does invite us
strongly to confuse the finger and the moon, doesn't it?
Is this better?
With friendly greetings, Patrick.
__________________________________________________
Do You Yahoo!?
Try FREE Yahoo! Mail - the world's greatest free email!
http://mail.yahoo.com/
MOQ.ORG - http://www.moq.org
Mail Archive - http://alt.venus.co.uk/hypermail/moq_discuss/
MD Queries - horse@darkstar.uk.net
To unsubscribe from moq_discuss follow the instructions at:
http://www.moq.org/md/subscribe.html
This archive was generated by hypermail 2b30 : Sat Aug 17 2002 - 16:01:57 BST