How can I get the training output vector (a2) from newrb?

Greetings, I am using newrb in predicting the drilling rate of penetration (ROP). I am comparing the results with a multiple regression, thus I am using an F-test. I need both the training output and the simulated output for the comparison. In newrb function, the training output is a2. How can I get the value of a2 and use it in my code.

[Sean de Wolski]

Any chance you could provide a small example so we can reproduce it?

[Ibrahim]

P = [1 2 3];

T = [2.0 4.1 5.9];

net = newrb(P,T);

P = 1.5;

Y = sim(net,P)

% This will give the trained output (Associated with p=1.5). I want the training data associated with 1, 2, and 3.

Answer 1

[Greg Heath]

Apparently, there is a BUG in newrb. In previous versions

 [ net tr ] = newrb(Xtrain,Ttrain);

would provide training history details in the structure tr in addition to providing the trained net.

Now, only the single output "net" is allowed and training history data is unavailable.

%CONSTANT MODEL TRAINING PERFORMANCE REFERENCE

 [ O Ntrain ] = size(Ttrain)

 Ytrain00 = repmat(mean(Ttrain,2),1,Ntrain));

 MSEtrain00 = mse(Ttrain-Ytrain00)

%TRAINING:

 MSEgoal = MSEtrain00/200 % R2train >= 0.995

 net = newrb(Xtrain,Ytrain,MSEgoal)

 Ytrain = net(Xtrain);

 MSEtrain = mse(Ttrain-Ytrain)

 R2train  = 1- MSEtrain/MSEtrain00

%NONTRAINING SET OUTPUT

 Ytest = net(Xtest);

%NONTRAINING SET PERFORMANCE IF TARGETS ARE KNOWN

 MSEtest00 = repmat(mean(Ttest,2),1,Ntest)

 MSEtest = mse(Ttest-Ytest)

 R2test  = 1- MSEtest/MSEtest00