LinearSolve.jl
LinearSolve.jl copied to clipboard
low rank update on `cacheval`
If you're reusing a LinearCache to solve a slightly perturbed system, it would be useful to do a low-rank update on cacheval
instead of recomputing the factorization? something like a sherman-morrison formula?
ref https://infoscience.epfl.ch/record/161468?ln=en https://en.wikipedia.org/wiki/Woodbury_matrix_identity
It could be. We could add it to the interface.