Out of memory during calculating covariance norm matrix

sumX=0;
for j=1:N
    sumX=sumX+X(:,j);
end
Xave=sumX/N;
Cov=0;
for j=1:N
    Cov=Cov+(X(:,j)-Xave)*(X(:,j)-Xave)';
end
Cov=Cov/N;
A=inv(Cov);

While running this code I am getting the error out of memory.I have tried the option of changing the settings in Preferences but still getting the same message.

Please help me to solve it out.

[Steven Lord]

What size is X and what is the value of N?

What release of MATLAB are you using? Is it a 32-bit release or a 64-bit?

On which line is the "Out of memory" error thrown? Is it the line where you add to Cov in the second for loop, on the line where you call inv, or some other line?

[K Ashwini]

X is 128541*2

N=length(X(1,:)) => N=2

Matlab R2016a

64-bit

Cov=Cov+(X(:,j)-Xave)*(X(:,j)-Xave)'; In this line the error is shown.

Answer 1

[Jan]

The wanted 128'541*128'541 matrix has 16'522'788'681 elements and it needs 132'182'309'448 Bytes of RAM: 132 GB. As a rule of thumb having the double size of memory is useful. Are you working on a 256 GB machine?

These claculations are very easy and should be the first you do, if you observe a message, that the RAM is exhausted.

By the way: You can simplify

sumX=0;
for j=1:N
    sumX=sumX+X(:,j);
end
Xave=sumX/N;

to:

Xave = sum(X, 2) / N;
% Or: Xave = mean(X, 2)

Please read: doc inv : An explicit inverting is rarely useful.