stopping criteria update #120
Conversation
Codecov ReportAttention:
... and 1 file with indirect coverage changes 📢 Thoughts on this report? Let us know!. |
|
Here are the |
|
Could you please rebase this? |
dpo
force-pushed
the
xi_update_from_nls
branch
from
fa2adfc to
e02c02b
Compare
|
I rebased this PR for you. |
dpo
force-pushed
the
xi_update_from_nls
branch
from
e02c02b to
688df74
Compare
| @@ -115,11 +115,11 @@ function LMTR( | |||
|
|
|||
| if verbose > 0 | |||
| #! format: off | |||
| @info @sprintf "%6s %8s %8s %8s %7s %7s %8s %7s %7s %7s %7s %1s" "outer" "inner" "f(x)" "h(x)" "√ξ1" "√ξ" "ρ" "Δ" "‖x‖" "‖s‖" "1/ν" "TR" | |||
| @info @sprintf "%6s %8s %8s %8s %7s %7s %8s %7s %7s %7s %7s %1s" "outer" "inner" "f(x)" "h(x)" "√(ξ1/ν)" "√(ξ/ν)" "ρ" "Δ" "‖x‖" "‖s‖" "1/ν" "TR" | |||
There was a problem hiding this comment.
Compute
The other solvers don't compute or output
There was a problem hiding this comment.
I was stuck on this when I was rebasing on Friday. It doesn't make sense to me to only do
| @@ -179,7 +177,7 @@ function TR( | |||
| continue | |||
| end | |||
|
|
|||
| subsolver_options.ϵa = k == 1 ? 1.0e-5 : max(ϵ_subsolver, min(1e-2, sqrt_ξ1_νInv)) | |||
| subsolver_options.ϵa = k == 1 ? 1.0e-5 : max(ϵ_subsolver, min(1e-2, sqrt_ξ1_νInv) * ξ1) | |||
There was a problem hiding this comment.
@geoffroyleconte I reinstated this because it disappeared in ac255073a980ee340c02d7dff568dc0f2a07d265. It's not clear which produces the best results.
There was a problem hiding this comment.
It was intended, and this choice is mentioned in the section numerical results of indef-pg.
| @@ -125,6 +125,7 @@ function LM( | |||
|
|
|||
| σmax = opnorm(Jk) | |||
| νInv = (1 + θ) * (σmax^2 + σk) # ‖J'J + σₖ I‖ = ‖J‖² + σₖ | |||
| ν = 1 / νInv | |||
There was a problem hiding this comment.
Why do we bother doing this? Why not just set ν directly?
No description provided.