How to iterate over parfor

Hi everyone.

I need to apply some heavy signal processing to a third-order tensor (typcal set of data coming from thermography). The data has the structure of pixel-pixel-time.

I wanto to use the parfor, but I am struggling in implementing a for loop.

I am aware that this syntax does not work:

%% mock data
input signal = zeros(1,3000);
output_raw_signal = zeros(512,640,3000);
AA_filtered = []; %% no need to prealloc I know
parfor i=1:512
    for j=1:640
           AA_filtered(i,j,:) = xcorr(squeeze(output_raw_signal(i,j,:)),.input_signal) %% cross correlation for each   pixel in time
end
end

If I run this, I get the error "The variable in a parfor cannot be classified"

https://de.mathworks.com/matlabcentral/answers/223299-error-the-variable-in-a-parfor-cannot-be-classified

But what is a possible solution?

Greetings.

Luca

[Luca Pecoriello]

Hi. You are absolutely right about the third dimension of the output tensor. I have corrected it now.

why do you think the syntax does not work? what error are you getting?

I am getting "variable in a The variable in a parfor cannot be classified". Which should be expected since in a parfor with a nested for loop, the variables running one loop should not communicate with the other ones. They should be independent.

Answer 1

[Ridwan Alam]

I am not sure how much this would help, as I couldn't reproduce your error. Please let me know how it goes.

input_signal = zeros(1,3000);
output_raw_signal = zeros(512,640,3000);
AA_filtered = cell(512,1); 
parfor i=1:512
    tempvar = [];
    for j=1:640
        newvar = xcorr(squeeze(output_raw_signal(i,j,:)),input_signal);
        newvar = reshape(newvar,1,[]);
        tempvar = [tempvar;newvar]; 
    end
    AA_filtered{i} = tempvar;
end

even this ran fine:

input_signal = zeros(1,3000);
output_raw_signal = zeros(512,640,3000);
AA_filtered = [];
parfor i=1:512
    tempvar = [];
    for j=1:640
        newvar = xcorr(squeeze(output_raw_signal(i,j,:)),input_signal);
        newvar = reshape(newvar,1,[]);
        tempvar = [tempvar;newvar]; 
    end
    AA_filtered(i,:,:) = tempvar;
end

[Luca Pecoriello]

Hi. The second one works! I am about to test the first option you have written

[Luca Pecoriello]

For some reason your solution works for the case before, but in the following situation I get an error.

In one .m code I am defining the following command:

hyper_cube_filtered_lowpass = lowPassHyperCube(,LP); %% LP is a filter object

%% Then I define the following code in another external function that should process each signal from each pixel in time. The function receives 2 inputs: inputdata which is a 3rd order tensor and filterVar which is filter object

function [outputData] = lowPassHyperCube(inputData,filterVar)
outputData = [];
parfor i=1:size(inputData,1)
    tempvar = [];
    for j=1:size(inputData,2)
        newvar = filter(filterVar,squeeze(inputData(i,j,:))-mean(squeeze(inputData(i,j,:))))+mean(squeeze(inputData(i,j,:)));
        newvar = reshape(newvar,1,[]);
        tempvar = [tempvar;newvar];
    end
    outputData(i,:,:) = tempvar;
    disp(['Iteration n.', num2str(i)])
end
end

But if I run it I get the following error:

Error using lowPassHyperCube>(parfor body) (line 9)
An UndefinedFunction error was thrown on the workers for 'filter'.  This might be because the file containing 'filter' is not
accessible on the workers.  Use addAttachedFiles(pool, files) to specify the required files to be attached.  See the documentation
for 'parallel.Pool/addAttachedFiles' for more details.

Error in lowPassHyperCube (line 6)
parfor i=1:size(inputData,1)
Caused by:
    Undefined function 'filter' for input arguments of type 'struct'.

Any idea why?

Let me know you can try this out. Attached you have the mat file containing the filter object I have created but you shouldn't need any specific data, you can just try with fake ones.

Let me know

[Walter Roberson]

Where is B coming from?

[Luca Pecoriello]

sorry typo from another code. It's "filterVar" I have already corrected my comment.

[Ridwan Alam]

I don't think it's an issue of parfor, rather how you are calling filter(). The arguments are not compatible. Your LP is a filter object, as far as I understood, not a vector as required by filter(). As pointed out by Matlab:

Undefined function 'filter' for input arguments of type 'struct'.

Btw, you can open a new question for debugging that part, that way your question will get more visibility.

[Walter Roberson]

However: when there is an object transferred whose class is not known to the workers, it appears as a struct.

[Luca Pecoriello]

I have opened a new discussion.

https://it.mathworks.com/matlabcentral/answers/496202-iterate-to-filter-signal-in-a-parfor

Answer 2

[Walter Roberson]

%% mock data
input_signal = zeros(1,3000);
output_raw_signal = zeros(512,640,3000);
AA_filtered = []; %% no need to prealloc I know
parfor i=1:512
    output_raw_signal_i = output_raw_signal(i,:,:);
    for j=1:640
       AA_filtered_i(j,:) = xcorr(output_raw_signal_i(j,:).', input_signal) %% cross correlation for each   pixel in time
    end
    AA_filtered(i, :, :) = AA_filtered_i;
end

[Luca Pecoriello]

Hi

This option gives me the following error in the workspace.

Error: The variable AA_filtered in a parfor cannot be classified. See Parallel for Loops in MATLAB, "Overview".

[Walter Roberson]

Try removing the initial assignment to AA_filtered