Elias Jarlebring

Results 14 issues of Elias Jarlebring

When running `curve_fit` in bigfloat mode, I get problems when I want to use the `g_tol` kwarg from `levenberg_marquard`: ``` julia> p0 = rand(BigFloat,2); julia> curve_fit((x,p)-> p[1] .* x, big.(rand(2)),...

@JuliaRegistrator register()

A common use-case: We have a linear solver (eg gmres) for a nep, and want to carry out deflation. The original linear solver will not directly work for the deflated...

This PR adds the possibility to do damping in the newton-type methods. This is useful when you want further control of the convergence, e.g., if the eigenvalue jumps around too...

Here is a reimplementation of NLEIGS TOAR. This is inspired / follows the implementation in SLEPc: https://slepc.upv.es/slepc-master/src/nep/impls/nleigs/. In contrast to our other implementation of NLEIGS, this incorporates a compact representation...

Here is a "future" issue. After speaking a bit to @haampie, it seems `orthogonalize_and_normalize!` will get a new home in a new package. Once that package exists we should start...

keep an eye on

There is a new version of the online code from G Demesy that we base our nano-tutorial on. The tutorial still works, but you would need to go back to...

NLEIGS uses a cache to store factorization objects, e.g., in the code in `src/rk_helper/linsolvercache.jl`. In `NewLinSolvers.jl` the `FactorizeLinSolverCreator` provides the same functionality. Therefore, I propose we remove the functionality in...

Based on discussion in #120 we want to make a cleaner interface with fortran arpack, in order to lessen the number of corner cases which are making the code difficult...

Wish-list / todo-list for more matrix exponential methods * [ ] An implementation of *A. H. Al-Mohy and N. J. Higham, A new scaling and squaring algorithm for the matrix...