Extracting from cell arrays using a nested loop

I have a 1x250 cell array. Each cell contains an 87001x16 matrix. I want to extract all of the first column from cell 1, column 1 from cell 2, column 1 from cell 3, ... column 1 from cell 250, to create a 87001x250 matrix. And I want to do this with all 16 columns, creating a separate 1x16 cell array, with each cell containing an 87001x250 matrix.

When I run the following code, I get a 1x16 cell array, with each cell being the first 16 columns of the 250th cell:

for k = 1:250
    Data{k} = Matrix(zapTime(k)-12000:zapTime(k)+75000, 2:end);   % Yields 1x250 cell array, with 87001x16 matrices.
    for c = 1:16
        subData{c} = Data{k}(:,c);   % Creates 1x16 cell array, with first 16 columns from Data cell #250.
    end
end

When I run this code, I get a 1x16 cell array, with each cell being the matrices from the first 16 cells.

for k = 1:250
    Data{k} = Matrix(zapTime(k)-12000:zapTime(k)+75000, 2:end);   % Yields 1x250 cell array, with 87001x16 matrices.
    for c = 1:16
        subData{c} = Data(:,c);   % Creates 1x16 cell array, with first 16 matrices from Data cell array.
    end
end

I'm able to code things separately just fine when I use the following code:

subData{k} = Data{k}(:,1);
subData{k} = Data{k}(:,2);
subData{k} = Data{k}(:,3);
...
    subData{k} = Data{k}(:,16);

But certainly a nested loop would be more efficiant. When I try this...:

for c = 1:16
    subData{k} = Data{k}(:,c);
end

...I get 250 cells of the data from columns 16.

Please help! Thank you!

Answer 1

[Stephen23]

The issue is that subData{c} is being overwritten on each iteration of k, whereas you need to index into both dimensions (c and k) simultaneously, building up each 87001x250 matrix column by column.

Instead use subData{c}(:,k) to assign the k-th column of each sub-matrix:

% Bild Data as before
for k = 1:250
    Data{k} = Matrix(zapTime(k)-12000:zapTime(k)+75000, 2:end);
end
% Pre-allocate and fill subData:
subData = repmat({nan(87001,250)},1,16); % preallocate
for k = 1:250
    for c = 1:16
        subData{c}(:,k) = Data{k}(:,c);  % k-th column of the c-th output matrix
    end
end

Here is another approach using cellfun and cell2mat (this won't be faster, just more compact code):

subData = cell(1,16);
for c = 1:16
    subData{c} = cell2mat(cellfun(@(x) x(:,c), Data, 'UniformOutput',false));
end

And for completeness, a solution all on one line (warning: creates a large intermediate 3D array!):

subData = reshape(num2cell(permute(cat(3, Data{:}), [1,3,2]), [1,2]), 1,16);

[Garyn]

Thank you so much! Do you recommend one of these methods over the others (for efficiency or speed)? I'm currently working with only part of our dataset for simplicity/time, but in the future I'll be working with larger datasets.

[Stephen23]

My guess is that in general the nested loops (together with suitable preallocation) would be the fastest approach, but rather than rely on my guesswork you could just test this yourself:

https://www.mathworks.com/help/matlab/ref/tic.html

https://www.mathworks.com/help/matlab/ref/toc.html

Answer 2

[Matt J]

Do not use a nested loop:

clear Data

for k=250:-1:1
 Data(:,k,:) = Matrix(zapTime(k)-12000:zapTime(k)+75000, 2:end) );
end

subData=reshape( num2cell(Data,[1,2]) ,1,[]) ;