Add BurerMonteiro formulation - #32
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #32 +/- ##
==========================================
+ Coverage 98.07% 99.27% +1.20%
==========================================
Files 15 21 +6
Lines 415 969 +554
==========================================
+ Hits 407 962 +555
+ Misses 8 7 -1 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
| function NLPModels.grad!(model::BurerMonteiro, x::AbstractVector, g::AbstractVector) | ||
| grad!(model.model, x, g) | ||
| return g | ||
| end |
There was a problem hiding this comment.
I will have the same comment as usual: do you think it would be worth providing the hprod! as well? I have the feeling that Newton-CG can be faster than LBFGS, if done properly.
If you think that's more relevant, I can open a separate issue.
There was a problem hiding this comment.
Yes, it will be useful but too early in the development stage at the moment. The code is too unstable for now. I prefer stabilizing first :)
There was a problem hiding this comment.
hprod! is done now, so Percival is now an SDP solver ! We should now try MadNLP ;)
2 participants
It can then be parametrized with, say, Percival to yield a full-fledged Burer-Monteiro solver.
Refactored from kocvara/Loraine.jl#29
Ideally, https://github.com/luotuoqingshan/SDPLRPlus.jl could use the same interface and Loraine could use the same model as well.