I have a square matrix with the dimension of $1600\times1600$. I need to compute the eigenvalues of this matrix and investigate its observability.
The usual way is to compute the eigenvalues of this matrix in MATLAB and proceed. But I know that the computation of eigenvalues is sensitive to numeric precision specially for large matrices.
My question is: What is the best (numerically most stable) method to computate the eigenvalues of a large non-symmetric matrix?
Thanks in advance, Reza