Open afraser3 opened 2 years ago
I have used eigenvectors = ((EP.solver.eigenvectors.T)[EP.evalues_index]).T
in the past, which I believe reliably sorts the low-res solver's eigenvectors to line up with EP.evalues. Perhaps this is all there is to it. I guess I'm still worried about differences between the low-res and high-res solvers.
Sometimes I want to do matrix manipulations that involve using a matrix whose columns are various eigenvectors from the EVP solver. A good example can be found here in
calc_ps()
, specifically lines 399, 402, and 414. If I just doEP.solver.eigenvectors
, then those eigenvectors don't necessarily match up with the eigenvalues inEP.evalues
, which are the eigenvalues ofEP.hires_solver
(assumingreject=True
). This can cause unintended behavior.It would be great if I could just get
EP.evectors
and have that be the eigenvectors that correspond toEP.evalues
whetherreject=True
orreject=False
. And in case it's True, I'm thinking it would be great forEP.evectors
to be the higher-resolution eigenvectors, but maybe other users have a reason why they'd also want access to the lower-resolution eigenvectors.