December 28th, 2004 09:05 AM
Does anyone know the code in either perl or C++ for perceptrons?
Perceptrons, for those who don't know, are like electronic neurons in AI.
December 28th, 2004 02:56 PM
An introductory comment:
If you are just a starter into the world of neuronal networks,
I recommend also to study fluxes in a network in parallel. The
later can be used in large scale "ethernet" networks (in the
end possibly well paid ), and the correlations between the two
are large. For a start, check Robert Sedgewick's Algorithms.
Well, in any case, Sedgewick's book is a must.
A MATLAB comment:
I have no experience with the c/c++ packages I mention below,
but they made the impression to create a good fundament to built
upon. However, if you have available MATLAB I highly recommend the
neuronal network toolbox. It is easy to develop in this framework
and quite similar to c/c++. And it is an industrial standard.
Some c/c++ packages/libraries:
A collection of basic implementation/algorithms can be found at
neural-networks-at-your-fingertips. A typical project is
Annie, some more general machine-learning environment can
be found at Torch.
If you are already familiar with the backpropagation algorithms,
you may have noticed it is very slow/inefficient in realistic
scenarios. A good paper dealing with its optimization has been written
by Schiffmann (a google for "schiffmann neural network" also is interesting).
A kind of review can be found at the datamininglab, which also deal with
global network optimization.
You did not specify your current state of knowledge, so I tried
to cover the whole spectrum a little bit.
Have fun (it's been a while I played with these, but I really enjoyed
it - and got paid at the same time).
If the only tool you have is a hammer, you tend to see every problem as a nail.
(Abraham Maslow, Psychologist, 1908-70)
December 29th, 2004 08:19 AM
I am relatively completely new to this subject. This fascinates me nonetheless, though!