Get all unique combinations from cell array for use as functional arguments

Hello,

I have data stored in a n x n cell array. For example:

data = {rand(1000,1) rand(1000,1); rand(1000,1) rand(1000,1)};

I want to compute the mean squared error between all possible unique cell combinations. In this case, if computed manually, they would be the following:

immse(data{1,1}, data{1,2});
immse(data{1,1}, data{2,1});
immse(data{1,1}, data{2,2});
immse(data{1,2}, data{2,1});
immse(data{1,2}, data{2,2});
immse(data{2,1}, data{2,2});

I will work with larger cell arrays than in this example (e.g., 10 x 10 instead of 2 x 2). In those cases, I do not want to type out the combinations manually.

I am therefore trying to find a way / a function to automatize this process. Can anyone help?

Thanks a lot in advance.

Answer 1

[KSSV]

A 2*2 matrix means there are four elements i.e. four global indices.

idx = 1:4 ;
c = nchoosek(idx,2)

C has the possible combination's indices. You can use those indices to get the required error.

[Franziska von Albedyll]

Thank you very much for your answer!

Answer 2

[Jan]

This can be achieved by 2 loops using linear indexing:

for i1 = 1:numel(data)
    for i2 = 1:numel(data)
        immse(data{i1}, data{i2})
    end
end

ind2sub let you get the 2 indices from the single linear index on demand. But 4 loops are possible also:

[s1, s2] = size(data);
for i1 = 1:s1
    for i2 = 1:s2
        for i3 = 1:s1
            for i4 = 1:s2
                immse(data{i1, i2}, data{i3, i4})
            end
        end
    end
end

[Franziska von Albedyll]

Thank you very much for your answer!