k-fold for nprtool

Apart from creating a NN and generating plots like confusion matrix ; how can one induce the k-fold CV method into the inputs data matrix for the NN;

specifically, for a 4-class classificaction problem with 10 samples inputs in each .

[yogini prabhu]

This does the job:-

cvFolds1 = crossvalind('Kfold',10, k); cvFolds2 = crossvalind('Kfold',10, k); cvFolds3 = crossvalind('Kfold',10, k); cvFolds4 = crossvalind('Kfold',10, k);

cvFolds=[cvFolds1;cvFolds2;cvFolds3;cvFolds4];

for i = 1:k

net = configure(net,input,target);

testIdx = (cvFolds == i);

trainIdx = ~testIdx;

trInd = find(trainIdx)

tstInd = find(testIdx)

net.trainParam.epochs = 100;

net.divideParam.trainInd = trInd

net.divideParam.testInd = tstInd