Index must be a positive integer less than or equal to the number of test sets.

Hi

i'm trying to create 10-fold cross validation for my CNN. To create the 10 batches I'm using a for loop, but it's giving this error again and again.

'Index must be a positive integer less than or equal to the number of test sets.'

Possibly how I can change the loop to initialize batches.?

selectBatch=cell(1,10);
for k=1:10
    %%Prep Data
    % 10 kfold crossvalidation indicies
    c = cvpartition(204,'kfold',10);
    testInd = test(c,selectBatch(k));
    trainInd = ~testInd;

    epochsTarget = zeros(204,1);
    epochsTarget(103:end)=1;
    epochsTarget= categorical(epochsTarget);
...............
end

[Image Analyst]

Read this link, then give us the full error. That means ALL the red text, including the line number and line of code it fails on.

[Joana]

Sorry if my question did not make sense.

This is the full error:

Error using internal.stats.cvpartitionImpl>checkindex (line 441)

Index must be a positive integer less than or equal to the number of test sets.

Error in internal.stats.cvpartitionImpl/test (line 229)

checkindex(i,cv.NumTestSets);

Error in cvpartition/test (line 229)

testIndices = test(cv.Impl,varargin{:});

And it is possible due to this line of the code:

testInd = test(c,selectBatch(k));

Because when I run the code withourt for loop, and with a fixed parameter of 'selectBatch' the code runs okay, but it doesn't work with the for loop.

[Image Analyst]

The error message says your line of code throwing the error is

testIndices = test(cv.Impl,varargin{:});

not

testInd = test(c,selectBatch(k));

When you set a breakpoint on line 229 of your program, what are the values of cv.Impl and varargin{:}? Why are you sending in varargin anyway???

Answer 1

[Adam]

selectBatch(k)

is a cell array, which clearly is not an integer.

selectBatch{k}

may be what you want instead.

That said, your cell array also appears to be empty anyway.