Convexity of the perturbation value
ProvedVectorSpaceOpt.perturbationValue_convexLet be convex on and let be cone-convex there with respect to . For the extended-real perturbation value , the finite-height epigraph
is convex. This epigraph formulation expresses convexity without performing ambiguous arithmetic between and and includes infeasible and unbounded perturbations. It is the source-faithful extended-value version of Luenberger's assertion that is convex. The milestone makes the geometry of right-hand-side perturbations available to later supporting-functional and multiplier arguments and can be reused for general conic value functions. No finite-value hypothesis is hidden in the statement.
import Definitions.Def_VectorSpaceOpt_coneConvexOn import Definitions.Def_VectorSpaceOpt_perturbationValue
namespace VectorSpaceOpt
/-- Luenberger, Chapter 8, §8.3, Proposition 1, expressed by convexity of the
finite-height epigraph of the extended-real perturbation value. -/
theorem perturbationValue_convex
{X Z : Type*}
[AddCommGroup X] [Module ℝ X]
[NormedAddCommGroup Z] [NormedSpace ℝ Z]
(P : ConvexCone ℝ Z) (Ω : Set X) (f : X → ℝ) (G : X → Z)
(hf : ConvexOn ℝ Ω f) (hG : ConeConvexOn P Ω G) :
Convex ℝ {p : Z × ℝ |
perturbationValue P Ω f G p.1 ≤ (p.2 : EReal)} := by sorry
end VectorSpaceOptRead-back
What the Lean code literally says, in plain math · gpt-5
For a real vector space , a real normed vector space , a real convex cone in , a set , and functions and , assume is convex on and assume is convex with for all and all with . Define as the infimum of over satisfying . Then the finite-height epigraph , where is embedded into , is a convex subset of . If no finite contributes a pair above , while if every real does; if is empty then is identically and the displayed epigraph is empty. No nonemptiness or closedness hypothesis is present.
Confirmed by the mission captain (proposal self-audit).