Avoid for loops, use cell operation to do: r2' * r1 * l1' * l2

Hi all,

Say I have 2 simple 1-by-2 cells, I'd like to perform operation like this: r2' * r1 * l1' * l2 to obtain a 2-by-2 matrix, sample code as follows:

clear; clc;

% the cells.
bl = {1 2};
br = {3 4};

% element 1.
l1 = bl{1, 1};
l2 = bl{1, 1};
r1 = br{1, 1};
r2 = br{1, 1};

otpt11 = r2' * r1 * l1' * l2;

% element 2.
l1 = bl{1, 2};
l2 = bl{1, 1};
r1 = br{1, 2};
r2 = br{1, 1};

otpt12 = r2' * r1 * l1' * l2;

% element 3.
l1 = bl{1, 1};
l2 = bl{1, 2};
r1 = br{1, 1};
r2 = br{1, 2};

otpt21 = r2' * r1 * l1' * l2;

% element 4.
l1 = bl{1, 2};
l2 = bl{1, 2};
r1 = br{1, 2};
r2 = br{1, 2};

otpt22 = r2' * r1 * l1' * l2;

% put all elements together.
otpt = [otpt11 otpt12; otpt21 otpt22];

I can probably use for loops to do it, but I'd like to use a compact operation such as cellfun to do it, which I believe is faster and more nature for cells.

Many thanks!

[KSSV]

cellfun uses loop inside. I think better is to use loops.

[Stephen23]

@Xh Du: why are you using cell arrays to hold numeric data? Why not use simple numeric arrays? If you put your data into numeric arrays then this would be a trivial operation.

[KSSV]

Yes.....that's correct..

[Xh Du]

Well, because in my real problem these numeric are vectors, so I put these vectors into cells to make it clearer rather than putting all vectors into arrays.

I know that this will work:

bl = [1 2];
br = [3 4];
b = bl .* br;
otpt = b' * b;

But for my case I have to keep bl and br separated, i.e. no .* operation, but to use "r2' * r1 * l1' * l2". Actually originally I have b = [3 8], bl and br are decomposed from b.

[Jan]

Use loops. They are fast and clean, if you pre-allocate the output.

[Stephen23]

Continued here:

https://www.mathworks.com/matlabcentral/answers/339488-avoid-for-loops-use-cell-operation