IterativeSolvers.jl

IterativeSolvers.jl copied to clipboard

Dear all, thank you for the great work. I am using IterativeSolvers.jl in some recent work. In particular, I called it from the QuantumOptics.jl toolbox, the function https://docs.qojulia.org/api/#QuantumOptics.steadystate.iterative which is...

WouterJRV

Merge DFTK's implementation of LOBPCG into IterativeSolvers

6

Hi! Since the first time we discussed about the implementation of the LOBPCG algorithm (#246), we (and by that I mean @mfherbst and @antoine-levitt) have made some improvements on the...

GVigne

When I use the powm function, I found sometimes it does not return the correct answer. For example ```Julia kernel = [3.850631961339304 -0.010072156399475053 -0.010258539801313745 -0.010293814829149936 -0.010306218470668904 -0.010311969710167707 -0.010315096531477644 -0.01031698282267339 -0.010318207467439057...

LiangtaoPeng

lsqr! not minimizing norm

1

According to the documentation lsqr! should return the minimum norm solution if the matrix is ambiguous. This doesn't seem to work: ``` julia using IterativeSolvers #- A = zeros(5,5) b...

FHell

I have a question. Why is the default value of `reltol` often picked as `sqrt(eps)` and not e.g. `eps`? Is there a particular reason? https://github.com/JuliaLinearAlgebra/IterativeSolvers.jl/blob/40d9e9db7e0c6ee2047d121125005b217023baf9/src/cg.jl#L122 cc @Elise-palethorpe

navidcy

use deterministic RNG initial state

7

Closes #303. Uses a deterministic RNG of `MersenneTwister(seed)` for generating initial vectors in iterative algorithms that use random initial vectors, controllable via an `rng` keyword. (In some cases, the `rng`...

stevengj

I would love to see that IDR(s) joins the iterator-interface like the other solvers. That's also important for SciML interoperability where IDR(s) is missing from http://linearsolve.sciml.ai/dev/solvers/solvers/#IterativeSolvers.jl .

mschauer

From the docs for `powm!`: > `tol::Real = eps(real(eltype(B))) * size(B, 2) ^ 3`: stopping tolerance for the residual norm that is, the tolerance scales as the cube of the...

danielwe

Look-ahead Lanczos Biorthogonalization

1

## Background This PR introduces `LookAheadLanczosDecomp`, an iterable which performs the Look-ahead Lanczos process. There is a companion PR #322 for Quasi-minimal residual method (QMR) for solving linear systems which...

platawiec

This is a companion PR to #321 . This will be rebased on top of that PR once merged (couldn't figure out how to stack the PRs in this case)...

platawiec