Hi,

I'm trying to port some code which uses ARPACK and ARPACK++. The catch is that the problem (chemistry calculations) creates huge matrices, so people usually just solve for N eigenvalues, where N is supplied by the user (e.g., one calculation solves for the 3 smallest eigenvalues, then maybe a follow-up solving for 10 roots, etc.)

Is this possible now using Eigen? Most of the examples I've seen solve for *all* eigenvalues.

Thanks,
-Geoff

Hi,

actually there is even no eigenvalue decomposition for sparse matrices. We only have eigendecompositions for dense matrices, and yes those compute all eigenvalues/eigenvectors.

So then, for now, I should stick with ARPACK(++). OK, that's too bad, but I know there's always a lot to do and not enough time.