Neural Network retrieve test dataset to plot

I am wanting to retrieve the data in the 'test' plot from training so that I can plot the data myself in a different format.

I can access the training, validation, test indices using [tr] but how do I find the network outputs for these to plot against the targets?

Answer 1

[Greg Heath]

From either

 [ net tr y e ] = train( net, x, t );

 % or
 % [ net tr ] = train( net, x, t );
 % y = net(x);
 % e = t - y ; 

 ttrn = t(:,tr.trainInd);
 ytrn = y(:,tr.trainInd);

and similarly for yval and ytst.

Hope this helps.

Thank you for formally accepting my answer

Greg

[Lucy]

Thanks Greg, this looks promising.

However, I am slightly confused - y is the size of the trainInd not the full training set. Therefore the testInd are outside the range of y.

Lucy

[Greg Heath]

No.

Similarly for yval and ytst.

[Lucy]

I am interested in the test component of the training set. y appears to contains the output for train, but for val and test. How can I access the test output? i.e. ytst=y(:,tr.testInd); Index exceeds matrix dimensions Thanks

[Greg Heath]

No. if train(net,x,t) contains all x and t, then y contains ytrn, yval and ytst.

Obviously you are doing something wrong. Please post code and results of using one of MATLAB's example data sets in

 help nndatasets
 doc nndatasets.

HTH.

GGreg