You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Because we can also directly provide estimates of xNorm2 and dNorm2, the final result can be completely wrong because
they are not always computed in Euclidean norm.
cc: @paraynaud
Be careful if you use it with preconditioners.
The text was updated successfully, but these errors were encountered:
The norm used for the trust-region should be modified if a preconditioner is provided.
We should use the Euclidean norm only when
M=I
in CG and CR.https://github.com/JuliaSmoothOptimizers/Krylov.jl/blob/main/src/krylov_utils.jl#L376-L392
Because we can also directly provide estimates of
xNorm2
anddNorm2
, the final result can be completely wrong becausethey are not always computed in Euclidean norm.
cc: @paraynaud
Be careful if you use it with preconditioners.
The text was updated successfully, but these errors were encountered: