mll-gh / DQGMRES.jl

modification of GMRES adapted from JuliaLinearAlgebra/IterativeSolvers.jl

DQGMRES

Direct Quasi-minimal GMRES
^{#numericallinearalgebra | #krylovmethods}

---

Fast Krylov subspace method for indefinite square systems

Use of incomplete orthogonalization¹ leads to reduced memory demands compared to the 'full' GMRES algorithm, which stores the entire Krylov Matrix. Adapted from GMRES implemetation in IterativeSolvers.jl, which is in fact a dependency. Initially developed as a project for CME 338² at Stanford University.

Footnotes

1. Saad, Y. and Wu, K., 1996. DQGMRES: A direct quasi‐minimal residual algorithm based on incomplete orthogonalization. Numerical linear algebra with applications, 3(4), pp.329-343. ↩

2. CME 338 / MSE 318: Large-Scale Numerical Optimization (spring quarter 2018). ↩