no I didn't hear that.I've heard that u can choose the NUMBER of hidden units in
a neural network (e.g. MLP neural network)using GA. it may give u better
results.

no, but in light of fadel response, I suppose
that one could use a GA to choose the number of hidden
nodes. This would of course be a lenghthly proccess as
changing the number of hidden nodes would invalidate the
wieghts, also I have no idea about the effectivness of
this approach, ergo I can make no promises.

Does anyone know of a book, website, etc that discusses AI algarithms for
understanding writing (sentences etc.)?

are you reffering to hand written characters or natural language?

Sorry, yes I am looking for a way to understand natural language typed into the
computer.  I am looking for a book, sample source code, web pages, etc.?

Title says it all. Could someone please tell me how they work? I have some ideas
but am still in a hazy place....<br><br>By the way I'm new here. Hello all! :)

None that I know of but I can tell you ones that
I have made.<br><br>They're based on what I could
gleen from what JUMBO is said to be in the book I am
reading again.<br><br>I just got a book called games and
decisions from DOVER. Good stuff so
far...<br><br>Spider_plant9

Well,<br><br>General description for JUMBO goes
as follows. Each piece of a word has a weight type.
This is a computer way to figure jumbles hence
"Jumbo". The way it works is that the weights for
sylabiles are set by likelihood of
combining.<br><br>ckysspear<br><br>Now, let's break it up "ck" is common so is
"ear".
Now, as the computer begins to match things up it
might come up with the word "skyscaper". The entire
thing is based on context associated weights. Although,
I think the writer may say otherwise. Who
knows?<br><br>The code is secret for mine (very pretective and all)
but the idea isn't mine so it wouldn't be fair not to
share. The program that I built is based on the same
principle but it is for sentense structure and paragraphs.
The context was built the way many of us learn
language by reading a lot. The program built the weight
system based on the probability of words being near
eachother (although not quite). Now, the codon phrases
(three word phrases) are matched with commands and
common usage groups kind of like a theasaurus. The
structure is based on bible quoting typing.<br><br>This is
loosly how it works most of this stuff is pretty simple.
Hey, Frozzy this should answer a lot of your questions
about how Haley works.<br><br>Spider_plant9

A neural network operates under a theory called
Parallel Distributed Processing (developed by Jay
McClelland and David Rumelhart- look up their papers for
detailed info. ) which is under a broader theory known as
Connectionism. If you are wondering what these theories mean,
Connectionism and PDP are theories that model the brain in
representing information. Thus a neural network refers to a
system that uses patterns of activation among its units
to process, retrieve and store information.
<br><br>The brain consists of neurons and other cells that
support neurons in their functions. Neurons are the only
cells directly responsible for the cognitive abilities
of the brain. A neuron has a receiving/input end and
a output end. In general the ends of neurons are
connected to other neurons. Neurons have the ability to
produce action potentials which can be basically
described sending a signal to another neuron or simply
"being turned on". If a neuron produces an action
potential or increases/decreases the rate of "firing", a
signal is sent to its connecting neurons and those
neurons must react. This is more complex than it sounds,
but let's just assume that when a neuron fires it
turns on its connection neurons. The brain represents
information by connecting neurons or circuits of neurons to
other circuits or neurons. Although the exact coding of
information is not understood, it is thought that the brain
uses both neuron-neuron connections and rate of firing
to code for information. Thus in a loose way this is
how neural networks function. <br><br>For example,
when you think of a labrador dog you activate the node
the represents the "labrador" but since that node is
connected to other nodes which could represent attributes
such that a labrador is a dog, good swimmer, has
floppy ears and that it is a mammal. The closer the
other nodes are to the original node the stronger the
association. Thus floppy ears are closer associated a labrador
dog than the fact it is a mammal. Here is a simple
diagram of the node connections :<br><br>MAMMAL --- DOGS
--- LABRADOR --- FLOPPY EARS<br> <br>It is worth
noting that since each node represents a "chunk" of
information and attributes of that node are represented by
links to other nodes, there is no differentiation
between hardware or software, that is, the software is
coded by the make up of the hardware.
<br><br>"Retrieving General and Specific Knowledge from Stored
Knowledge of Specifics" by J. L. McClelland, 1981 uses a
good example. <br><br><br>I hope this helps, I am by
no means an expert so others may present a more
accurate definition.<br><br>Loud-Amp

<EOM>

hi all,<br><br>having some problems lately..i'll
like to find out more on the use of fuzzy logic in
sensor fusion. how do i implement fuzzy logic using C?
does FUZZGEN helps? <br><br>thanks a lot
..<br><br>ride_9fe

what are you trying to do? What kind of sensor and what do you want to do with
the information?

When dealing with Neural nets, The weight is
changed by:<br><br>dw = a(t - y)v<br><br>Where dw =
change in weight, t is actual output, y is expected
output, and v is input vector, and a is learning
rate.<br><br>How do you change threshold?

Hey, I'm an ametur cyberneticist and I am more
than a bit annoyed at the very broad and unfocused
scope of most artificial intelligence groups. I require
one whose sole focus is solving the single problem of
cybernetic intelligence.<br><br>All these endless stupid
discussions of neural nets, expert systems, and the like are
a waste of time and effort. They only skirt around
the single worth-while problem and can only provide a
marginal benefit over other programming methods. <br><br>A
seperate list called "advanced programming techniques"
should be created and serve as a trash-heap for the
rubish that currently passes for AI. I will unsubscribe
in a few days if this group doesn't improve
substantialy.

first, I don't ussually dis people so to speak,
but in this case I will. First go learn what you are
talking about. You are not an ametaur anything, you an
uneducated, which isn't a problem and a moron, because you
fail to understand any concept of what you are talking
about. Your method is actually rebish, it was tried
orginally, AI is not that easy and thier are many concepts
involved in it. If you are not interested in the different
parts of it, give up, go away, and don't go insulting
people and and trying to pass your rubish for AI.
Otherwise, read what's already thier, learn to code it, and
combine what's thier.<br><br>wizard

"First go learn what you are talking
about.<br><br>Neural nets? I have read several books on
them.<br>Expert systems? They've been tried and didn't yield any
commercially viable results.<br><br>"You are not an ametaur
anything, you an uneducated, which isn't a
problem"<br><br>So I can't pass english 112, At least my spelling
and my grammar are decient. The only education which
I have acquired has been through books. I will
grant you that AI has not been a primary focus of my
studies but at least I have a good understanding of what
is relevant and what is stupid or
extraneous.<br><br>"and a moron, because you fail to understand any
concept of what you are talking about.<br><br>For
clarification, I'm the kind of moron who has scored in the 96th
percentile in a rigorous and scientfic IQ test.<br><br>I
understand a great many things. I only talk about things
that I don't understand in the context of an iniquiry.
<br><br>"Your method is actually rebish, it was tried
orginally,""<br><br>Yeah, and the people who attempted it didn't succede.
Unless you can prove that the reason for their failure
was infact the method they were trying to apply *and*
that my methods are no different than theirs then the
assertion that it is "rubish?" is false.<br><br>Infact,
aren't you here making a statement about what my methods
are before I have communicated them them to you in
any detail? I ask that you seek to understand what it
is you are talking about before making and
statements about it. ;)<br><br>"AI is not that easy and
thier are many concepts nvolved in it."<br><br>I would
be careful if I were you in making statments about
the nature of AI before a complete unified theory has
been proposed and demonstrated. It might turn out that
these two assertions are completely false!<br><br>"If
you are not interested in the different parts of
it,"<br><br>My assertion is that Neural networks, Expert
systems, Fuzzy logic, and the like are NOT in any way
parts of AI. The ball is now in your court to proove or
even show evidence that I am wrong in any part of this
assertion. <br><br>"give up,"<br><br>That would be a stupid
thing to do because I will need an extremely powerful
AI if I am to have any hopes of completing my life's
work. =\<br><br>"go away,"<br><br>Why you say this is
baffeling because it was my stated intention to leave the
list if I couldn't get past all the idiots such as
your self who degrade the value of the list so
drasticly.<br><br>"and don't go insulting people and and trying to pass
your rubish for AI."<br><br>Oh, and who was it who
began his posting with "I don't usually dis people".
<br><br>Who is it who is passing crap like CYC or Neural Nets
off as having any relevance to the field of AI? It
was >>> ME <<< who proposed to
remove the rubish from the list. Things like this
suggest that you agree with my intent but are too stupid
to see that we agree and infact accuse me of doing
the exact opposite of what I was attempting to
do.<br><br>If you aren't going to read my postings before
flaming me I suggest you shut up or remove yourself from
this group. <br><br>"Otherwise, read what's already
thier, learn to code it, and combine what's
thier."<br><br>Why would I do any of that? What's there doesn't
work. I will even go as far as to say that in the
context of this group, that stuff is *WORTHLESS*.
<br><br>I can code anything I desire and understand
algorithmicly. The only things I desire are: <br><br>- An
acceptable replacement for Windoze 3.11 <br>- An artifical
intelligence. <br><br>Since no algorithm for intelligence
exists I can't code it. =P If it did, and it can be
coded into a conventional computer, then I could code
it in any language ever developed. <br><br>I would
very much like to do that but you can't use my
inability to do that against me because you can't do it
either.

this argument is circular, point is, if algorithm
doesn't exist you can not do the folowing :<br> a. say
anything which has had any success is worthless<br><br> b.
disregard it.<br><br>as for your iq, not impressed, I never
am. If your iq is so great that you can do anything,
then code it yourself, yes? if not, then by
disregarding what already has had marginally success you are
making a mistake.<br><br>Also, for what reason, if any,
have you decided that neuron nets are worthless? What
founding? Exactly, hmm?<br><br>Well, we are at it, what
proof have you it can be done on a conventional
computer, that it is procedural in nature? If I were like
you, I would go into groups telling them that ALL
PROCEDURAL ATTEMPTS OR STRAIGHT ALGORITHMIC ATTEMPTS ARE
WORTHLESS, I think you see my point.<br><br>If you intent to
go around telling people that what they believe is
worthless then be prepared for the very same. I really
don't care whether you believe neuron nets or
conectionist networks or any other the names for them that you
may like, are relevant or not, others believe they.
<br><br>Oh, and read a psycology paper, a lot of
them.<br><br><br>We are in agreement however, in two sections, Expert
systems are not a good example of soemthing moving
towards true AI in my believe, however, as they stand to
gain the most from it, they are still something that
should be considered. <br><br>"Expert systems? They've
been tried and didn't yield any commercially viable
results."<br><br>reason why we need AI is thier, thier is room for
improvement, AI techniques would be needed, of course people
in AI are going to discuss it. They are the applied
version of AI. Although, not the modern name for same, I
believe their called knowledge based systems now

still not done..........<br><br>your statement "I
can code any algorithm I desire", so what, so can I,
so can most other programmers, who cares, not the
point whether you can code an algorithm, you have to
come up with it first, that is what shows the skill.
So, the question have tried comming up with new
algorithms, have you looked at psycology books, you get the
point.<br><br>Ok, the apology for the force of the original
response, I was over reacting.<br><br>>>That would be
a stupid thing to do because I >>will need an
extremely powerful AI if I am to >>have any hopes of
completing my life's work. =\<br><br>On a final note I
should think that AI would take any body a life time to
complete, wouldn't you?

wizard_of_frozzbozz wrote:<br>>> point is,
if algorithm doesn't exist you can not do the
folowing :<br> a. say anything which has had any success
is <br>worthless<<<br><br>It's [nearly]
worthless to someone trying to understand intelligence, the
only truly useful purpose for an 'AI'
list.<br><br>> b. disregard it.<br><br>After I have reached an
inital understanding of what I am looking at and found
it to not satisfy my requirements then I have no
reason to pay any<br>further attention to it unless I
re-evaluate my requirements. =P<br> <br>>> If your iq
is so great that you can do anything, then code it
yourself, yes?<<br><br>What says I won't? ;)<br>As I
said, what I need to code it is an understanding of the
function/relation it performs. <br><br>>> if not, then by
disregarding what already has had marginally success you are
making a mistake.<br><<<br><br>No, I'm not.<br>I
may look at something that has shown some success but
I will immediately discard it in favor of something
that I hope will show complete sucess.<br><br>There is
no point in persuing a marginally suitable
technology. Would you promote the use of wooden railroad
tracks because they are marginally able to hold a
train's weight? NO!!! You would go hunting for metals to
refine...<br> <br>>> Also, for what reason, if any, have
you decided <br> that neuron nets are worthless? What
founding? Exactly, hmm?<br><<<br><br>Because, today,
they are used in place of a good theory. Like a magic
button that does some cool things but for no explicable
reason. And also<br>because they do not address the
cybernetic problem, only the perceptive portion thereof...
<br><br>( The cybernetic problem could be stated as the
quest to construct a system by which perception and
"efection" are carried out in such a<br>manner that useful
work can be accomplished.)

>> Well, we are at it, what proof have you
it can <br>be done on a conventional computer, that
it is procedural in nature?<br><<<br><br>I
have no definitive proof of any kind.<br><br>>>
If I were like you, I would go into groups <br>
telling them that ALL PROCEDURAL ATTEMPTS OR <br>STRAIGHT
ALGORITHMIC ATTEMPTS ARE WORTHLESS, I <br>think you see my
point. <<<br><br>I know of such a person. ( Lynn
H. Maxson ). He blindly and passionately asserts
that intelligence can't be emulated by a turing
machine, and by<br>extension a conventional computer. I
call this "Lynn's Postulate".<br><br>It is a very
interesting question. I have made a few attempts at showing
that the lambda calculus was incapable of expressing
solutions to several real-life problems, and provided
proposals for extensions to the same. Though I don't yet
know how to go about formulating a rigorous proof or
persue the project any further. <br><br>In any event, my
current focus is on making better computers so this kind
of work is easier... (Linux is too difficult for me.
I currently have no use for linux that would
justify the effort required on my part to master its
absurd complexity, therefore I need a suitable
alternative.)<br> <br>>> I really don't care whether you
believe neuron nets or conectionist networks or any other
the names for them that you may like, are relevant or
not, others believe they. [are]<<<br><br>If you
are going to assert that they are relevant to the
problem of general intelligence then I will insist uppon
*evidence* if not proof.<br> <br>[The Algorithm]<br>>>
you have to come up with it first, that is what shows
the skill. So, the question have tried comming up
with new algorithms, have you looked at psycology
books, you get the point. <<<br><br>I am more
interested in the physiology of the brain at this time.
Wheather I will succede in formulating an algorithm is yet
to be seen.<br><br>>> On a final note I should
think that AI would take any body a life time to
complete, wouldn't you? <<<br><br>I would hope not.
My hope is once it is understood in the general
case, special cases such as hearing, finger moving, and
others that aren't to<br>be found in nature can be
immediately derived and developed into working
solutions.<br><br>-- <br><a href=http://users.erols.com/alangrimes/
target=new>http://users.erols.com/alangrimes/</a> <my
website.

>Expert systems? They've been tried and didn't
>yield any commercially viable results.<br><br>Ever
heard of XCON? Just as an initial example of
commercially viable.<br><br>Since I've been involved since
1992 in developing commercial expert systems, I think
I have some difficulty with your
statement.<br><br>Julie

This is such a lame topic... let whomever wants to leave to do just that... why
complain?

I have read that if Moore’s law continues it’s
geometric progression, or even close to it, that the
complexity of computers will approximate, and then surpass
that of the human mind in approximately 20-30 years.
While this may seem a “chicken little” scenario to
some, should we not be thinking of ways to ensure that
if/when our machines become self-aware they do not
consider humans more of a nuisance than a boon?

Thats rich. You probably threatend to leave the
country if Bush was elected too. Why don't you try going
to college like many of the people in this group are
currently doing. You can research in AI until your heart is
content once you get into graduate school. Besides what
do you want? For people to cut and paste their
homework into the message fields? Anyway, this the
'general' AI club. Consider making a Yahoo! 'amateur
Cybernetics' club if thats what you want to talk about.

Alan,<br><br>your original message is annoying
and does not add anything useful to the group.<br>If
you are unhappy with this forum then please find (or
found) another one.<br><br>I will not comment on your
observations as I have better things to do with my time. Just
behave.<br><br>Dan

Also he does not spell very well.<br><br>Phil

should we not be<br> thinking of ways to ensure
that if/when our machines become self-aware<br> they
do not consider humans more of a nuisance than a
boon?<br><br> You have to consider that if your creating a life
you have to be aware that it is a life and not do the
human thing as to make it a slave or treat what is
created like an animal. As you said, If/when this happens
it would be best to treat the created life as
another person and not the hired hand. If the intent is
of slavery of Intelligent life then I daresay that
such a problem of robot life thinking of mankind as a
nuisance will be one thought of soon after creation, and
we're all in trouble....<br> But you do have a point, I
guess we'll find out how such life is treated when
created and shown to the world.

I know I haven't talked at all scince joining this club but I'm using a diff
name. I'm under Mighty_Grey_Dragon...

Just got this and I had to share it:<br><a
href=http://antwrp.gsfc.nasa.gov/apod/image/0011/earthlights_dmsp_big.jpg
target=new>http://antwrp.gsfc.nasa.gov/apod/image/0011/earthlights_dmsp_big.jpg<\
/a>

Linux is good to replace the DOS and Win9x crap as for the AI have a few what
would you be looking for.<br><br>Spider_plant9<br><a
href=Http://clubs.yahoo.com/compupsychoecology
target=new>Http://clubs.yahoo.com/compupsychoecology</a>