Neural Network Question regarding matrix regression

I'm new to neural networks.

How can I make a nonlinear regression fit ? My target is a data matrix.

I start nftool and I have Input and Target. What is the input ?

Answer 1

[Greg Heath]

x=linspace(0,400,5);

y=linspace(-10,10,11);

X = repmat(x,1,11);

Y = repmat(y,1,5);

Z=(tanh((X+20).*(Y-2)/200)-1)*3;

input = (X; Y);

target = Z;

Hope this helps.

Thank you for formally accepting my answer

Greg

Answer 2

[Greg Heath]

The net represents a nonlinear mapping from I-dimensional input vectors to O-dimensional output vectors. The sizes of the input and target matrices are

 [ I N ] = size(input)

 [ O N ] = size(output)

Hope this helps.

Thank you for formally accepting my answer

Greg

Answer 3

[Stephan Moeller]

I there any example available where I can learn how to understand nntool for :

x=linspace(0,400,5); y=linspace(-10,10,11); [x1,y1]=meshgrid(x,y);

here's my synthetic data z: z=(tanh((x1+20).*(y1-2)/200)-1)*3;

mesh(x1,y1,z)

how do have to configure nntool to calculate a neural network with x,y as inputs and z as the output data ( or target ?? ) .

Answer 4

[Stephan Moeller]

Thanks Greg. That works now.

And how can I build the function when nntool is ready ? I can see the weights of the layers, but is possible to get a symbolic function out of it ?

[Greg Heath]

Practice using nntool on MATLAB's demo/example data sets.

 help nndatasets

Then try your data.

The only analytic form you can get from the net is a sum of tanh functions.