Hello,

Is their a better way to fill a sparse matrix that is known to be symmetric. Right now I'm using something like this:

Code: Select all

// M is an Eigen::DynamicSparseMatrix<double>
M.coeffRef(i,j) -= value;
M.coeffRef(j,i) -= value;

// Finally 'mat' is the SparseMatrix I use
Eigen::SparseMatrix<double> mat(M);


Thanks for your amazing work on Eigen!

This is what I do:

[code]
// Eigen::SparseMatrix<float,1> M;

//fill it out
M.startFill();
for (i=0;i<N;i++) {
// loop some j
M.fill(i,j) = value;
}
M = M + Eigen::SparseMatrix<float,1>(M.transpose());
[\code]

I am not sure which one is faster. Is there another way to do this?

ideally, if you know your matrix is symmetric, you should only fill one half (the upper or lower triangular part), and then use the selfadjointView API to view and use it as a symmetric/selfadjoint matrix.

Unfortunately I'm pretty sure that for the sparse module, the seladjointView API is very incomplete. But the goal is to have this working:

// fill the upper part of a
c = a.selfadjointView<Upper>() * b;

maybe the above already works with b a dense vector/matrix, I don't remember...