"Dan Minette" <[EMAIL PROTECTED]> wrote:
>
> > True - but only in the most pedantic way. You typically can't write an
> > algorithm for what a neural network does that doesn't involve "insert a
> > neural network here" as a step.
>
>Can you, with enough time, actually walk through the steps that are made in
>a neural network? If not, why not? If so, then it is an algorithm.
As I said, yes. But that set of steps involves a complete description of the
neural network. Contrast this with the description of a binary adder at the
NAND gate level can be vastly simplified and abstracted.
Robert Shaw is making my point much better than I am on another part of this
thread. We can devise things like neural networks for which for which we
cannot describe with simple algorithms but which can perform tasks quickly
and with good but not perfect reliability. Contrast with simpler algorithms
which might have perfect reliability but take much longer - perhaps
untennably longer. Since humans aren't perfectly accurate, I object to
statements which say our processing capabilities must go beyond the
capabilities of a Turing machine.
(And if you think quantum randomness is vital, just have your code tie into
HotBits: http://www.fourmilab.ch/hotbits/)
Joshua
_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com
