Marco Z
Marco Z
I think this boils down to lack of preconditioning
Discussion and algorithms for preconditioned GMRES, TFQMR, BiCGSTAB, QMRCGSTAB : https://arxiv.org/pdf/1607.00351.pdf "A Comparison of Preconditioned Krylov Subspace Methods for Nonsymmetric Linear Systems" - Ghai, Lu, Jiao 2016
Uhh, this one is bad. Thanks again @GregorySchwartz , will fix this soon. BTW What are you using the library for, if I may? I'm in the middle of a...
@GregorySchwartz no, apparently it's not fixed. Could you provide a minimal repro and/or perhaps send a patch? Thanks
Thanks a lot @Jaxan ! I'll look into this asap. For the time being, why did you have to change the bounds on `base` and the Stackage LTS versions? Thanks!
Great points! Yes, working over Z or Q would enable exact algorithms, but then we'd have to rewrite all the parts that use square roots (such as the L2 norm),...
NB: Givens' rotation in the complex plane: Saad sec. 6.5.9
Current implementation of (real-valued) Givens' rotation introduces a `Ord` constraint due to the comparison between the `abs` of two values. (Whereas `magnitude a < magnitude b` does not).
Depends on QR factorization working over C (#27) , since both QR method (duh) and Rayleigh iteration (which uses `` which in turn depends on GMRES) do.
Alternating Anderson-Richardson method: An efficient alternative to preconditioned Krylov methods for large, sparse linear systems https://arxiv.org/abs/1606.08740