Choosing data as training, validation and testing ANN

Hi everyone,

I am a very new in MATLAB and ANN, I want to know how choose the data as training, validation, and testing.

if I have data

A = [1:100] % Input,,,,,,,, B = A.^2 % Target

I want to choose 80% data as training, 10% validation and 10% also as testing.. And how if I want to take it randomly..

If the questions are not very clear, please inform me. Thank you very much

Muammar

Answer 1

[Walter Roberson]

The below code would need to be adjusted a bit if A is not a vector.

numA = length(A);
trainsize = floor(0.8 * numA);
valsize = floor(0.8 * numA);
testsize = numA - trainsize - valsize;
ridx = randperm(numA);

traindata = A(ridx(1:trainsize));
valdata = A(ridx(trainsize + 1 : trainsize + valsize));
testdata = A(ridx(end-testsize+1:end));

[Muammar]

Thank you.. I will try..

Muammar

Answer 2

[Greg Heath]

See the input parameter list of the newff documentation

help newff

Hope this helps.

Greg

[Muammar]

Hii Greg...

I use newff, but i dont know how to divide the data manually. Example, I want to use A(1:80) as training, A(81:90) as validation and A(91:100) as testing. Thank you..

Muammar

Answer 3

[Greg Heath]

It is straightforward:

ptrn = A(1:80);

pval = A(81:90);

ptst = A(91:100);

Similarly for t and B.

Or did I miss something?

Hope this helps.

Greg