The source code for the parfor-loop could not be found

Hi. I'm using the parfor in order to change some labels name:

Data = cell(1, length(DirList));
numberOfDbExamples = length(DirList);
parfor k = 1:numberOfDbExamples
    Data{k} = load(fullfile(Folder, DirList(k).name));
    [newLabelNames, newMask] = changeLabelNames(Data{k}.labelNames, Data{k}.mask);
    Data{k}.labelNames = newLabelNames;
    Data{k}.mask = newMask;
end

I get this error:

The source code (C:----) for the parfor-loop that is trying to execute on the worker could not be found.

Caused by:

Undefined function changeLabelNames for input arguments of type table.

Error using remoteParallelFunction

Do you have an idea of how fix it?

[Walter Roberson]

Where have you stored changeLabelNames relative to the current directory at the time of execution of the parfor ? And is it in its own file changeLabelNames.m or is it a local function inside your script ? Is it possibly a method of some object class?

Are you expecting your mat files to have a variable named labelNames and expecting it to be a table() object ?

[John]

The function is stored as changeLabelNames.m and it is located in the same folder of the script.

[Mehdi]

can you explain how you solved this problem? I faced similar error!

Answer 1

[Walter Roberson]

Sometimes you need to addAttachedFiles; https://www.mathworks.com/help/parallel-computing/parallel.pool.addattachedfiles.html

[Mehdi]

can you explain how you solved this problem? I faced similar error!

[Walter Roberson]

When there are hidden calls to a function, such as by

fmincon('NameOfFunction', x0)

then you might need to use addAttachedFiles

[Mehdi]

I build a function to put my parfor loop inside it and par for loop looks as below:

parfor i=1:M
    Hf = matlabFunction(H(i),'Vars',[xx yy]);
    Hf = @(xx,yy)Hf(xx,yy).*ones(size(xx));
    kdl(i,1) = k__d*a(2)*(b(2)/4)*integral2(Hf,-1,1,-1,1);
end

Where and how I have to use addAttachedFiles?

before the loop, inside, after???

[Walter Roberson]

I recommend using two different variable names for the two different anonymous functions.

Integral2 requires that the function accepts two arrays of the same size, and returns an array that size. Your second function handle contains the subexpression .*ones(size(xx)) . The multiplication will generate an error if the output of the matlabFunction-generated handle is not one of

• scalar
• row vector with the same number of rows as xx has
• column vector with the same number of columns as xx has
• 2d array with the same number of rows and columns as xx has

In the last case (array is already that size) then the multiplication by ones is redundant. In the other cases where the output of the generated function is being expanded to the proper size then the implication is that the output of the generated handle is not calculating each location individually, and so would be calculating incorrectly for the purposes of integral2()

Either you do not need the .*ones because the generated function already has the correct output size... or else your generated function is not suitable for use with integral2()

[Walter Roberson]

@Mehdi I just noticed your https://www.mathworks.com/matlabcentral/answers/1897595-why-receive-error-integrand-output-size-does-not-match-the-input-size?s_tid=srchtitle in which you are expecting that your function might return a constant and so not automatically vectorized by matlabFunction . There are better ways of dealing with that situation.

[Walter Roberson]

Over in https://www.mathworks.com/matlabcentral/answers/1897595-why-receive-error-integrand-output-size-does-not-match-the-input-size?s_tid=srchtitle the actual problem turned out to be the use of syms inside of parfor

Answer 2

[John]

I did something similar to what Walter mentioned, but I don't remember exactly.