Closed odow closed 3 months ago
If the PSD cone is given, do we error ?
No change to PSD. There's still written as affine cones
No change to PSD. There's still written as affine cones
Ah that's the current behavior. So it goes through the scalarize bridge ? So at the moment, conic variables are scalarized and PSD variables are ignored ?
Before this PR, the model is written out as:
min c' x
Ax - b in K
x free
All VectorOfVariables
constraints are accepted (no bridging) but are written as affine constraints.
After this PR, things are written as:
min c' x
Ax - b in K
x in some cones but not all
VectorOfVariables in Exponential, DualExponential, and PositiveSemidefiniteConeTriangle are still supported, but they are written as affine constraints.
Merging as an improvement for now. I'll look at doing EXP and PSD variables in a separate PR.
Part of #2457