How to sum different cells of cell array

Hi all,

I have the following problem.

I have a cell array (2x100) where each entry is a 11x74 table. What I would llike to do is to make the mean over those 100 tables of each value.

So if for instance I have a cell array (2x2) of say 2x3 matrices:

[1,2,3;4,5,6] [3,24,35;46,58,69]

[10,2,3;24,85,6] [11,22,33;54,15,16]

I would like to end up with the mean by row of each matrix so:

[(1+3)/2, (2+24)/2,(35+3)/2;(46+4)/2,58+5)/2,(69+6)/2]

[(10+11)/2,(2+22)/2,(3+33)/2;(24+54)/2,(85+15)/2,(6+16)/2]

Is there a way to do that?

Thank you

Answer 1

[Rik]

Something like this should work:

data={[1,2,3;4,5,6],[3,24,35;46,58,69];
    [10,2,3;24,85,6],[11,22,33;54,15,16]};
data=permute(data,[3 4 1 2]);
data=cell2mat(data);
data=mean(data,4)

data = 
data(:,:,1) =

    2.0000   13.0000   19.0000
   25.0000   31.5000   37.5000


data(:,:,2) =

   10.5000   12.0000   18.0000
   39.0000   50.0000   11.0000

You can reshape this output as needed.

[federico nutarelli]

Thank you for the answer @Rik. The problem is that my cell array is quite big. It i actually a (2x100) array. Can I usse prmute in this case too?

Also the following error arrise when using cell2mat:

Error using cat

Dimensions of arrays being concatenated are not consistent.

Error in cell2mat (line 118)

ct{mref{:}} = cat(cdim+1,c{mref{:},:});

I tried to delete empty cells but nothing changes:

data = Po1.Po(~cellfun('isempty',Po1.Po));

[Rik]

You could fill the empty arrays with NaN and use the 'omitnan' flag. 2x100 is not big. It is too cumbersome to display in text easily, but Matlab has a max array size far exceeding 2*100*11*74.

If the order of your cells doesn't matter, you could also use cat(3,data{:}) (assuming you have already removed all empty cells and the remainder has the same size.

Answer 2

[Star Strider]

I would do something like this:

C = {[1,2,3;4,5,6]    [3,24,35;46,58,69]
     [10,2,3;24,85,6]   [11,22,33;54,15,16]};

cat3 = zeros(size(C{1},1),size(C{1},2),size(C,2));
for k1 = 1:size(C,1)
    for k2 = 1:size(C,2)
        cat3(:,:,k2) = C{k1,k2};
    end
    cat3_mean{k1,:} = mean(cat3,3)
end
cat3_mean{:}                                            % Display Results

With the number of your matrices, two nested loops is likely the only possible solution. The cat function can do this, however it requires that the matrices to be concatenated be listed separately, and that does not appear to be possible here.