Enable subsolver tolerances tuning, fix negative tolerance #93
Conversation
Codecov ReportPatch coverage:
Additional details and impacted files@@ Coverage Diff @@
## master #93 +/- ##
==========================================
+ Coverage 57.47% 58.10% +0.62%
==========================================
Files 11 11
Lines 1204 1222 +18
==========================================
+ Hits 692 710 +18
Misses 512 512
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here. ☔ View full report in Codecov by Sentry. |
|
Here are the |
dpo
left a comment
dpo
left a comment
There was a problem hiding this comment.
Thank you! Can we use TRDH inside LM and LMTR now?
|
|
||
| f = LSR1Model(model) | ||
| λ = norm(grad(model, zeros(model.meta.nvar)), Inf) / 10 | ||
| λ = 1.0e-1 # norm(grad(model, zeros(model.meta.nvar)), Inf) / 10 |
There was a problem hiding this comment.
Are the results very different if you leave the default lambda ?
There was a problem hiding this comment.
Not that much, I updated the tables and the benchmarks in the article.
|
|
||
| verbose = 0 # 10 | ||
| ϵ = 1.0e-6 | ||
| ϵ = 1.0e-5 |
There was a problem hiding this comment.
Yes, this is the tolerance for
Not at the moment be this would be a very minor change to make them work. I can do it in my next PR once this one is merged. |
|
Here are the |
| options3 = ROSolverOptions(spectral = false, psb = false, ϵa = ϵi, ϵr = ϵri, maxIter = maxIter_inner) | ||
| options3bis = ROSolverOptions(spectral = false, psb = false, ϵa = ϵi, ϵr = ϵri, maxIter = maxIter_inner, reduce_TR = false) | ||
| options4 = ROSolverOptions(spectral = true, ϵa = ϵi, ϵr = ϵri, maxIter = maxIter_inner) | ||
| options4bis = |
There was a problem hiding this comment.
Let's give all these options descriptive names and avoid "bis" (it's French and will confuse non-French speakers).
|
Here are the |
|
Here are the |
* tune inner solver parameters, fix error negative tol * improve tables, change test parameters, modif demos * keep default value bpdn * update tables parameters * update parameters name noredTR
2 participants
I changed the algorithms so that it is possible to set some specific tolerances in the
subsolver_options. I did a slight modification in the defaultsubsolver_optionsso that it keeps the existing behaviour.It allowed me to find better parameter for the benchmarks.
I also updated the stopping criterion in TRDH so that it handles negative (but small)
ξ1orξvalues, and fixed what I think is a bug in R2 with theneg_tolparameter (which is always positive).