Oscar Dowson
Oscar Dowson
So KNITRO would be `VectorOfVariables` in `VariationalInequality{Nonnegatives}`
Related: https://github.com/jump-dev/MathOptInterface.jl/issues/1402 I closed because no one was really asking for it haha. It's do-able, but a little tricky to get right. Have you verified that it made a difference...
> But then why does PSDVAR even exist as part of the CBF specification Mosek used to (and maybe still does) require that PSD matrices are "special" and declared up...
I'm tempted to close this, but @blegat can decide. I think `Ax + b in K` is a reasonable thing to write. File formats don't have to be the most...
We're assuming that the solver implements a file reader for CBF and that it doesn't implement any support for this transition? We also don't (other than this SCIP-SDP solver) have...
So one sticking point is that when writing CBF files, we don't have a standard form that maps neatly onto CBF. Their PSDVar's can't be used in integrality, and they're...
> this is what MosekTools is doing Oh :cry: now I don't want to look
I guess at minimum, we could write variable cones of things other than PSDVAR as a preliminary improvement.
#2478 added support for some `x in K` constraints. I still need to look at `EXP` and `EXP*` cones, and PSDVAR block.
The EXP cone stuff is https://github.com/jump-dev/MathOptInterface.jl/pull/2482 So now all we're missing is PSDVAR.