Matrix operations without for structure

I'm making a program in Matlab and now I am trying to make it run faster.

In one step I need to make operations in a matrix with more than 60000 rows and 2 columns. Now, I have a for structure to go to every line and make the following operation: (line-mean)*s*(line-mean)'

line is the line that is selected in each iteration, mean, is the result of the mean of each column, and s is a matrix 2x2.

I would like to know if it's possible to make this operation without a for structure, so that the program runs faster.

Thanks for your atention

Answer 1

[Rick Rosson]

Please try:

   M = 60e3;
   N = 2;
   A = rand(M,N);
   s = rand(N,N);

   mu = ones(M,1)*mean(A);
   X = A - mu;

   R = X*s*X';

HTH.

[Pedro]

When I try to do that, it gives me the error: "Out of memory."

I think your solution only works with smaller matrices, or my computer can't make operations with big matrices like this one.

Answer 2

[Teja Muppirala]

M = 60e3;
N = 2;
A = rand(M,N);
s = rand(N,N);

Am = bsxfun(@minus,A,mean(A)); %subtract the column means

V = sum((Am*s).*Am,2); % Do the matrix operation