Hi, I am sovling some non-linear optimization problem with big sparse matrix whoose dimension is about 756m x 762m. The locations of each entry are fixed which means that the row and column index vectors are fixed. I had tried to create this matrix by using 'sparse(i,j,v,m,n)' like,
M = sparse(I(:), J(:), V(:), m, n);
It takes 1.3s (it may vary depending on what computer uses). Also, I had tried another approach which seems more valid for me
Mt = sparse(I(:), J(:), 1:length(I(:)), m, n);
M = spfun(@(x) V(x), Mt);
It takes about 0.5s. Since the matrix M is created at every iteration, I think the latter approach is right for me.
However, I don`t under stand where this difference comes from. In the spfun(at line 22), there is a code
f = sparse(i,j,feval(fun,x),m,n);
which can be thought as identical approach in the first one above.
Is there any reason why those difference comes from? and any faster method than what i tried?