plot the performance of a neural network

Hi

Look at this code please:

P = [ -1 -1 2 2 ;0 5 0 5];

T = [-1 -1 1 1];

net = newff( minmax(P),[3 1],{'tansig','purelin'}, 'traingd']; [net, tr]= train(net, P, T);

a = sim(net, P);

After Running it, when i click to see the peformance (MSE versus epochs) it shows a plot comparing train , test and validation data. How can i chande train and test data and not having validation?

thanks in advance

Answer 1

[Greg Heath]

Replace 'traingd'] with 'traingd')

I only get the training curve.

Are you sure you are using that obsolete version of newff instead of the more recent obsolete version that uses the data division default?

net = newff( P,T,3,{'tansig','purelin'}, 'traingd');

I get 3 curves using this one.

To better understand data division go to the nnet documentation

doc nnet

and search on

dividefcn

With only 4 points you cannot afford to divide the data. The default yields 2 training points and 1 point each for validataion and testing.

Why don't you go to the nnet documentation and find the list of sample data sets. Then choose one for your purposes so that you can experiment with how to control the data division?

[ P, T] = simplefit_dataset;

Then divide the data any way you want and check the results.

Hope this helps.

Thank you for formally accepting my answer

Greg

[Greg Heath]

If you do not want validation and test data, either of these should work.

net.divideFcn = '';

or

net.divideFcn = dividetrain;