Board index » delphi » Neural network assgnmt

Neural network assgnmt

Hi

I definately need help here. I have to program a neural network in Pascal.
It's a very simple one but I just cannot seem to find a way to start.
Any clues? It's supposed to learn from a range of input and then when I
present new input, it's supposed to give the expected answers...

every help is appreciated...

Thanks, Michiel

cs...@luxor.latrobe.edu.au  

 

Re:Neural network assgnmt


Quote
In article <4neb59$...@lugb.latrobe.edu.au> Michiel Kromhout wrote:

>Hi

>I definately need help here. I have to program a neural network in Pascal.
>It's a very simple one but I just cannot seem to find a way to start.
>Any clues? It's supposed to learn from a range of input and then when I
>present new input, it's supposed to give the expected answers...

>every help is appreciated...

>Thanks, Michiel

There seems to be a lot of information missing.  Is this network
a perceptron, cognitron, neocognitron, back-propagation,
counter-propagation, Hopfield, etc.?

You can implement a network in an object oriented or traditional
manner.  The typical back-propagation non-oop method defines
vectors (arrays) for input, intermediate (if any), output, and
target (if learning).  

Convolution and learning are then processed through a series of for
loops.  The driver is itself a loop that applies each exemplar to the
input and target vectors, invokes the convolution routine, then the
learning routine. The process is repeated until all exemplars achieve
satisfactorily results or the network goes brain-dead.

You then apply the test set of exemplars w/o training, and if the
results are satisfactory you are done. If not, start again after
modifying the physical structure of the net and/or modifying the
various learning and acceleration parameters or simply creating a
different initial weight distribution.

OOP solutions depend upon whether you want the base object to be an
individual neuron or a layer and how input/output are to be handled.

--
RDon...@gnn.com
http://members.gnn.com/rdonais/index.html
--------------------------------------------------------------
Knowledge is one of the few things that you
can give away and still keep for yourself.

Re:Neural network assgnmt


See the book 'C++ Neural Networks and Fuzzy Logic' by Valluru and Hayagriva Rao.

Yes, the sample code is written in C++, but they explain it well, so you should have no trouble
translating it.

Malcolm J. Jones
TELUS Corporation

Other Threads