Create cell arrays from doubles

Hi everyone,

I would like to ask how I can combine 3 doubles into one cell array. I want to get a cell array 2x1, say abc_i, where the firt cell will retrieve the values from a(1,1:t) b(1,1:t) and c(1,1:t). The second cell array should contain the respective values for a(2,1:t) b(2,1:t) and c(2,1:t). The code below delivers a cell array 2x1 but both cell contain the same values, retrieved from a(2,1:t) b(2,1:t) and c(2,1:t).

aaa = cell(2,1);
bbb = cell(2,1);
ccc = cell(2,1);
aaa = cellfun(@(x) randn(5,2), aaa, 'uni', false);  
bbb = cellfun(@(x) randn(5,2), bbb, 'uni', false);  
ccc = cellfun(@(x) randn(5,2), ccc, 'uni', false);

    for t=1:5;
        for i=1:2
  a(i,t)=mean(aaa{i}(t,:),2);    
  b(i,t)=mean(bbb{i}(t,:),2);         
  c(i,t)=mean([aaa{i}(t,:) bbb{i}(t,:)],2); 
    abc_i = cell(2,1);
    abc_i = cellfun(@(x) [a(i,1:t); b(i,1:t); c(i,1:t)],abc_i , 'uni', false);   

        end
    end

Any help would be much appreciated. Thanks in advance.

[Image Analyst]

Why on earth would you want to do that? Why use more memory and complicate your life with cell arrays when you don't have to?

[gsourop]

How else could I do it? In addition, mind that the numbers are greater than 2 and 5 and also I need a compact way to process further my results.

[Image Analyst]

Cell arrays are NOT compact. For large arrays they use far, far more memory than a regular double array. I'm not sure what you want to do because you removed, or never added, comments. If you want a 3-D array with all matrixes in it, make one with cat:

array3d = cat(aaa, bbb, ccc);

I'm not sure what you're doing with the means. Please explain in English. Is it the mean of the arrays, a mean of rows or columns, or a local/sliding mean, or something else.

What is the purpose of a, b, c, and abc_i?

[gsourop]

I am using neural networks and i are the number of hidden unit, ranging for 1:20 (for simplicity in the exercise above I used 2). I have j=90 predictors and produce the forecasts for 2000 observations (for simplicity I used 5). Hence, I created a 3 cell arrays of 30 predictors (because they can be divided like this according to the theory ; in the exercise above denoted as aaa,bbb,ccc) . Each cell array has the form {i}(t,j). I want to create 3 additional predictor, extracted from the mean of predictors for every t and every i (a,b and c). Then, I want to combine these means into a single cell array (abc_i). The reason I want to do this is because otherwise the code will be full of doubles. Is it clear now? I apologize for my English but I am not a native speaker.

[Greg Heath]

Go with the code full of doubles.

Greg

Answer 1

[Guillaume]

Same as Image Analyst, I think that you're probably needlessly complicating things and that you don't need cell arrays. For example, following your demo, aaa would be better stored as a 3D matrix:

aaa_m = cat(3, aaa{:});

it is then trivial to calculate your mean in one line without a single loop (there was never any point to your t loop in any case)

a = permute(mean(aaa_m, 2), [3 2 1])

The permute above is only there so that a is the same shape as in your demo. It's not actually needed if you're going to do further processing with a.

If you really want your a, b, c together in a cell array, you will first have to concatenate them into a 3D matrix, then split that matrix into a cell array. Again, there's no point in the cell array. Keeping it as a 3D matrix would be simpler:

abc_m = cat(3, a, b, c); %this is probably a lot more useful than the cell array
abc_i = squeeze(num2cell(permute(abc_m, [3 2 1]), [1 2]))

Again, the squeeze and permute is to get the arrays the same shape as you asked. There's little point in them.