The perturbation value decreases when constraints are relaxed
ProvedVectorSpaceOpt.perturbationValue_antitoneLet be the extended-real optimal value under . If , so that the right-hand side relaxes the constraint at , then
The statement requires no convexity because it follows from the nesting of the two feasible sets, and its EReal inequality remains meaningful when either problem is infeasible or unbounded below. This is Luenberger's monotonicity proposition with the cone-order orientation made explicit. It is a reusable order-theoretic fact for perturbation bounds and checks the sign conventions shared by the later sensitivity and duality items. It also covers empty feasible sets without a separate case split.
import Definitions.Def_VectorSpaceOpt_perturbationValue
namespace VectorSpaceOpt
/-- Luenberger, Chapter 8, §8.3, Proposition 2. -/
theorem perturbationValue_antitone
{X Z : Type*}
[AddCommGroup Z] [Module ℝ Z]
(P : ConvexCone ℝ Z) (Ω : Set X) (f : X → ℝ) (G : X → Z)
(z₀ z₁ : Z) (hz : coneLE P z₀ z₁) :
perturbationValue P Ω f G z₁ ≤ perturbationValue P Ω f G z₀ := by sorry
end VectorSpaceOptRead-back
What the Lean code literally says, in plain math · gpt-5
For an arbitrary type , an additive commutative group with a real vector-space structure, a real convex cone in , set , function , map , and vectors , if , then in , where . These are extended-real infima: an empty feasible set gives and objective values unbounded below give , and the stated inequality includes those cases. No convexity of , , or , and no topology or norm, is assumed.
Confirmed by the mission captain (proposal self-audit).