Add Golub Riley iteration method #123
Conversation
src/QRMumps.jl
Outdated
| * `Δy`: an auxiliary vector used to compute the solution, the size of this vector is n and can be uninitialized when the function is called. | ||
|
|
||
| The definition of the `shifted_spmat` may look odd at first sight. We use such a block matrix as an argument of `qrm_golub_riley!` because the QR factorization of this matrix has the property that | ||
| `RᵀR = AᵀA + αI` which is the system we need to iteratively solve from for the Golub-Riley iteration. |
There was a problem hiding this comment.
It would be clearer to give a brief description of the method.
There was a problem hiding this comment.
Why was this marked as resolved? I still think a brief description would be good here.
Also, "iteratively" here is confusing. We don't solve a system with an iterative method. We solve it repeatedly.
There was a problem hiding this comment.
Well I added a description below...
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #123 +/- ##
==========================================
+ Coverage 75.71% 79.36% +3.64%
==========================================
Files 5 6 +1
Lines 803 1071 +268
==========================================
+ Hits 608 850 +242
- Misses 195 221 +26 ☔ View full report in Codecov by Sentry. |
|
The documentation build fails. @MaxenceGollier |
|
Where do I see the documentation build @dpo ? (and how can I solve it ?) |
|
I don't see why |
|
Click on the documentation build below to see the details: https://github.com/JuliaSmoothOptimizers/QRMumps.jl/actions/runs/13348730277/job/37282696629?pr=123 |
@dpo.
I already tested this new feature for our prox, it works great ! I wasn't able to add an allocation test for this function here as there is some kind of runtime dispatch but it seems like it doesn't, at least for our prox.
Also, I added a new type
qrm_shifted_spmat, this is very useful and makes everything more readable in our prox implementation and it also convienently forces how we expect the entry matrices forqrm_golub_rileyto be.Let me know what you think when you have some time, thank you !