Add (R)SOCtoNonConvexQuad Bridge #1046
Conversation
Codecov Report
@@ Coverage Diff @@
## master #1046 +/- ##
==========================================
+ Coverage 95.26% 95.28% +0.01%
==========================================
Files 100 101 +1
Lines 11436 11484 +48
==========================================
+ Hits 10895 10943 +48
Misses 541 541
Continue to review full report at Codecov.
|
| fq = MOI.ScalarQuadraticFunction(a_terms, q_terms, zero(T)) | ||
| quad = MOI.add_constraint(model, fq, MOI.LessThan(zero(T))) | ||
| fp = MOI.ScalarAffineFunction([MOI.ScalarAffineTerm(one(T),t)], zero(T)) | ||
| var_pos = MOI.add_constraint(model, fp, MOI.GreaterThan(zero(T))) |
There was a problem hiding this comment.
Ooooo. This is a nice way to avoid the double bound issue. Don't know why I didn't think of that for Gurobi et al. Nice job.
|
Bump, @mlubin can you take a quick look? |
| `ScalarAffineFunction`-in-`GreaterThan`. Indeed, the definition of the | ||
| second-order cone | ||
| ```math | ||
| t \\ge || x ||_2 \\ (1) |
There was a problem hiding this comment.
Do the (1), (2), (3) labels display in the REPL? (Is it correct markdown?)
There was a problem hiding this comment.
Will check the best way to do so.
There was a problem hiding this comment.
it is not proper markdown, but it renders as:
help?> MathOptInterface.Bridges.Constraint.RSOCtoNonConvexQuadBridge
RSOCtoNonConvexQuadBridge{T}
Constraints of the form VectorOfVariables-in-SecondOrderCone can be transformed into a
ScalarQuadraticFunction-in-LessThan and a ScalarAffineFunction-in-GreaterThan. Indeed, the definition of the
second-order cone
2tu \ge \lVert x \rVert_2^2, t,u \ge 0 (1)
is equivalent to
\sum x_i^2 \le 2tu (2)
with t,u \ge 0. (3)
WARNING This transformation starts from a convex constraint (1) and creates a non-convex constraint (2), because the
Q matrix associated with the constraint 2 has two negative eigenvalues. This might be wrongly interpreted by a
solver. Some solvers can look at (2) and understand that it is a rotated second order cone, but this is not a
general rule. For these reasons this bridge is not automatically added to the LazyBridgeOptimizer. Care is
recommended when adding this bridge to a optimizer.
Which is not bad in my opinion.
docs/src/apireference.md
Outdated
| [`Bridges.Constraint.SOCtoNonConvexQuadBridge`](@ref). | ||
|
|
||
| The first option is to add the specific bridges to a | ||
| `bridged_model` optimizer, with coefficient type `T`, using |
There was a problem hiding this comment.
You need to specify that it is a LazyBridgeOptimizer
|
I added docstring with ref to full_bridge, but there are doc errors: ┌ Warning: unable to get the binding for '[`Bridges.full_bridge_optimizer`](@ref)' in src/apireference.md from expression ':(Bridges.full_bridge_optimizer)' in module MathOptInterface.Bridges.Constraint
│ exception = UndefVarError: Bridges not defined
└ @ Documenter.CrossReferences ~/.julia/packages/Documenter/PLD7m/src/CrossReferences.jl:137you know how to reference functions from outside? |
|
Tests passed, there is som communication issue with travis |
Simpler version of: #478
Bridge not included by default in Lazy bridge optimizer, but can be used by advanced users.
Tried to include caveats in the doc strings.