Re: Neural networks

new topic     » goto parent     » topic index » view thread      » older message » newer message

There is some guy(i forgot his name) woking on making NN's with
hardware...he has taken 3 transitors and did more then someone with a
few billion...the resion a computer canot truly do NN's is that it is
based on nothing more then 1's and 0's ...NN's are not based that way
but on a sliding scale...I have looked in to this a little and plan to
some more..I have thought about making a emulator that used 32 bit
number to represent 0 to 1(on a volt scale) and then make it
loop/add/subtract but i dont know just how to do it...might be a good
projeact for some one with more time on the subject....

Grape
>Greg P. Wrote:
><snip>
>> Students and professors at some university are using a genetic
algorithm
>> to create a new, useful, programming language.  Then, the computer,
>> using a neural network, would teach itself the language, and write an
>> improved genetic algorithm, to create an improved language, and an
>> improved neural net.
>
>> Which is scary.
>
>On the contrary, I think it's a wonderful thought !
>
>This reminds me of when I was a child. I had this idea: If I could
>jump a little bit up in the air, and then while still in the air,
perform
>another jump right away, I might be able to get higher and higher
>up until finally I would fly !
>
>As you can see, I was not a very bright child smile
>
>I don't think the idea you mentioned would work, though, not in a
>long while anyway. At any point, a human being would be able to
>work out a much better improvement of the language algorithm than
>the neural network in question. This of course due to the fact that
>our own NNs are far more powerful than what can be made with
>artificial means, and probably will be for another 50 years at least.
>
>But then, that's just MHO, and I'm still not very bright, so you may
>be right anyway... smile
>
>
>BTW: A month ago I asked whether anyone on the list was
>interested in, or was working with artificial neural networks,
>but I got no feedback. I'm trying to make some general Euphoria
>functions for a network with some kind of backpropagation in E.
>
>I'd still like to know whether anyone else is doing the same
>thing, or perhaps already has completed some functions.
>
>
>Regards,
>
>Tor Bernhard Gausen


______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com

new topic     » goto parent     » topic index » view thread      » older message » newer message

Search



Quick Links

User menu

Not signed in.

Misc Menu