How to accelerate matrix calculation with matlab?
Show older comments
I have 40800 linear sparse equations that have the following forms:
F(1)=2*x(1)+x(2)-10;
F(2)=x(9)+3*x(10)-7;......
F(40800)=x(40200)-x(1)+....-10;
I want to convert these equations to matrices by evaluating them at the columns of speye(N);
N=40200; %the number of variables
E=speye(N,N+1);
columns=cell(1,N);
beq=-evaluate_equations(E(:,N+1));
for i=1:N
columns{i}=evaluate_equations(E(:,i))+ beq;
end
Aeq=cell2mat(columns);
full(Aeq),
full(beq),
function F=evaluate_equations(x)
F(1)=2*x(1)+x(2)-10;
F(2)=x(9)+3*x(10)-7;......
F(40800)=x(40200)-x(1)+....-10;
F=F(:); %column vector
But Matlab is very slow. It can not build the matrices. However i use 8 workers to accelerate the computing. Have you an idea to overcome this problem? Thanks.
Answers (1)
Walter Roberson
on 7 Jan 2014
0 votes
If they are sparse linear equations, could you not code them as a matrix multiplication in the first place? If you did that you could also skip the "for i" loop and do it as a single matrix multiplication.
4 Comments
imed NASRI
on 7 Jan 2014
Walter Roberson
on 7 Jan 2014
A = [2 1 0 0 0 0 0 0 0 0 -10; %predetermined
0 0 0 0 0 0 0 0 1 3 -7]
x = [E(1:size(A,2)-1, :); ones(1, size(E,2))]; %column-oriented!
A * x
This is for general E. Considering that your particular E is spyeye(), the answer probably just comes out as A.
Thus, you would code them in matrix form instead of in equation form.
Are you generating the equations somehow, perhaps using the symbolic toolbox? If so then the symbolic toolbox can be used to convert into matrix form.
imed NASRI
on 8 Jan 2014
Edited: imed NASRI
on 8 Jan 2014
imed NASRI
on 8 Jan 2014
Edited: imed NASRI
on 8 Jan 2014
Categories
Find more on Creating and Concatenating Matrices in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!