Is this the right way to use permute and bsxfun?

Hi all,

I have two matrices A and B where the size of A is 1000 x 200 x 500 (i x j x k in Cartesian notation) and that of B is 1000 x 500.

What I want to achieve is a matrix C such that C is also the same size as A and every jth column of C is obtained by subtracting the matrix B from the corresponding column of A. In this regard, is the following code correct? Since it is a 3D system, I am not able to verify it very easily.

C = permute(bsxfun(@minus, permute(A, [1 3 2]), B), [1 3 2]);

Answer 1

[Dyuman Joshi]

Yes, it is correct. You can verify it by comparing it to the result obtained via loops -

%Random data
A = rand(1000,200,500);
B = rand(1000,500);

%Original code
C1 = permute(bsxfun(@minus, permute(A, [1 3 2]), B), [1 3 2]);

%Loop method
C2 = zeros(size(A));
for jdx = 1:size(A,2)
    C2(:,jdx,:) = reshape(A(:,jdx,:),size(B)) - B;
end

isequal(C1,C2)

ans = logical
   1

[Tanmay Agrawal]

Thanks!