JuliaLinearAlgebra / ArnoldiMethod.jl

The Arnoldi Method with Krylov-Schur restart, natively in Julia.
https://julialinearalgebra.github.io/ArnoldiMethod.jl/dev
MIT License
102 stars 18 forks source link

Purging: remove converged, unwanted vecs #140

Closed haampie closed 8 months ago

haampie commented 8 months ago

It looks like converged, unwanted vectors can actually be problematic for the Krylov-Schur restart, since they have tiny residuals, and we need to construct reflectors out of those values.

By reordering the converged, unwanted to the tail of the decomposition, they're removed, and we don't have to compute reflectors involving their residuals.

codecov[bot] commented 8 months ago

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Comparison is base (2a264db) 95.47% compared to head (9fd7e67) 95.52%.

Additional details and impacted files ```diff @@ Coverage Diff @@ ## master #140 +/- ## ========================================== + Coverage 95.47% 95.52% +0.05% ========================================== Files 10 10 Lines 839 849 +10 ========================================== + Hits 801 811 +10 Misses 38 38 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.