From: Tiago Filgueiras Pereira on
Hello,

I have no experience with the software MATLAB nor in computer programming , and i need to create a neural network for wich the goal will be that with a matrix 299x101 as imput ,the output will be a matrix of 12x32. I have 21 imput matrices and their corresponding 21 output matrices to train the network. Do you think it´s possible to learn to use the matlab toolbox and create this neural network with a minimum of success within 2 to 3 months? Please , i need the most honest answer ´cause that´s all the time i have left and if i realize it´s impossible i´ll just forget about this project and think of something else for my final project of graduation...
thanks for th attention

Tiago
From: Charles on
Have a similar task to solve, someone help.
Charles.
From: Greg Heath on
On Apr 28, 10:56 am, "Tiago Filgueiras Pereira"
<tiago2newjer...(a)hotmail.com> wrote:
> Hello,
>
> I have no experience with the software  MATLAB nor in computer programming , and i need to create aneuralnetwork for wich the goal will be that with a matrix 299x101 as imput ,the output will be a matrix of 12x32.  I have 21 imput matrices and their corresponding 21 output matrices to train the network.  Do you think it´s possible to learn to use the matlab toolbox and create thisneuralnetwork with a minimum of success within 2 to 3 months? Please , i need the most honest answer ´cause that´s all the time i have left and if i realize it´s impossible i´ll just forget about this project and think of something else for my final project of graduation...
> thanks for th attention

If the data is not pathological, I don't see why it should take more
than several weeks.

The dimensions of the input and output target matrices for a
MLP FFNN with node topology I-H-O is

size(p) = [ I N] % training input
size(t) = [O N] % training target

There are tutorials and demos in the Toolbox documentation

help nnet

doc newff
help newff

Search within the newsgroup with

greg-heath pre training advice for newbies

Hope this helps.

Greg

P.S. Crossposted to comp.ai.neural-nets