Manipulating dimensions without using loops

I would like to know if there is a way to manipulate the following high dimensional array with permute or bsxfun or other functions without using so many loops, here iTe1, iTe2,... are known indexing matrices. Thanks in advance!

for i1 = 1:nE1
    for i2 = 1:nE2
        for i3 = 1:nE3
            for i4 = 1:nE4
                outputM(:,:,:,:,i1,i2,i3,i4) = inputM(iTe1(:,i1),iTe2(:,i2),iTe3(:,i3),iTe4(:,i4));
            end
        end
    end
end

[James Tursa]

What is inputM? A variable or a function? What are the dimensions of all of your variables? How are the values in the iTe1,...,iTe4 variables created? Any pattern to them?

[Boyi Ye]

Hi, inputM and outputM are both high dimensional arrays. Basically I want to reshape inputM to outputM. Let's say the dimension of inputM is and I want to reshape it into outputM with dimension . iTe1,...,iTe4 are indexes as. So I want to do like , , etc.

which each of the above can be considered as a block of dimension .

Thank!

Answer 1

[James Tursa]

It looks like the indexing is consecutive, it is just that you have a mixture of reshaping (which doesn't change memory order of data) and permuting (which does channge memory order of the data) the dimensions. E.g., for your specific example:

inputM = 22 x 22 x 22 x 22 array
r = reshape(inputM,[2,11,2,11,2,11,2,11]);
outputM = permute(r,[1 3 5 7 2 4 6 8]);

Perhaps you can use a variation of this two-step scheme for your actual problem.

[Boyi Ye]

Yes, that's exactly I want. Thank you so much!!