How speed up symbolic matrix calculation?

Hi, all

I try to calculate the following symbolic matrix('A').

A= s.'*(s.*repmat(census(:,6),1,k))-diag(sum(repmat(census(:,6),1,k).*s));

Here, s=438x611 Symbolic matrix, census=438x1 Numeric vector, k=611. And each element of matrix 's' is quite complicated because it includes exponential expression and division expression.

(Ultimately, I'll multiply 'A' and a vector(let's call it 'B'). And I'll differentiate the resulting vector('C=A*B') with respect to their parameters.)

I ran this code on Matlab. The problems are two things.

First, 'low memory' warning pops up and Matlab program is forced to end, even though I have 16G RAM. I checked windows task manager and found out that the above calculation used more than 16G memory. ---> What shall I do? Do I have to upgrade to 32G memory? Is there any other way to calculate 'A' which requires less memory?

Second, the calculation takes too long. It takes more than at least 2 hours. ---> Is there any faster way to calculate 'A'?

Could you help me? Thanks.

Thanks, Moonsik