MATLAB: Isn’t the network splitting between training and testing data

neural networks

I'm using divideFcn = 'divideind' and splitting by index. This usually works but it doesn't here and I don't know why. Here's my code
% import training data
filename = 'train.txt';
delimiter = ' ';
headerlines = 0;
A = importdata(filename, delimiter, headerlines);
% split training data into inputs and targets
inputTraining = A(1:2:end)';
targetTraining = A(2:2:end)';
% import testing data
filename = 'test.txt';
delimiter = ' ';
headerlines = 0;
A = importdata(filename, delimiter, headerlines);
% split testing data
inputTesting = A(1:2:end)';
targetTesting = A(2:2:end)';
% initialize the NRBF network with 1-hidden layer as usual. Change the
% function from RBF to NRBF. And set the training function to gradient
% descent
squaredErrGoal = .01;
spread = .1;
input = [inputTraining inputTesting];
target = [targetTraining targetTesting];
net = newrbe(input,target,spread);
net.layers{1}.transferFcn = 'radbasn';
net.trainFcn = 'traingd';
net = init(net);
view(net);
numTraining = length(inputTraining);
numTesting = length(inputTesting);
net.divideFcn = 'divideind';
net.divideParam.trainInd = 1:numTraining;
net.divideParam.valInd = numTraining+1:numTesting;
[net,tr] = train(net,input,target);
plotperf(tr); 
figure(); %to create a new figure for the plotfit.
plotfit(net,input,target); % how many inputs and outputs can it handle?
xlabel('input');
legend({'target','output'})

Best Answer

DO NOT USE THE FUNCTIONS NEWRBE, NEWGRNN OR NEWPNN!!!
They do not have a reasonable amount of flexibility,
If you want a RADIAL BASIS FUNCTION NET for regression or classification, use NEWRB!!!
Even NEWRB has many more deficiencies than FITNET(regression) and PATTERNNET(classification), however, I have shown how it can be used without too much frustration.
It has been some time since I've used it. SO, you will have to search for my posts and tutorials in the NEWSGROUP and ANSWERS.
Search backwards in both NEWSGROUP and ANSWERS using
        greg newrb
Hope this helps.
Thank you for formally accepting my answer
Greg
Related Question