Convexity of the optimal cost in the right-hand side
ProvedLinearOptimization.lp_optimal_cost_convex_in_rhs(Theorem 5.1, p. 213) The optimal cost is a convex function of on the set .
(Standing hypotheses in force, made explicit: standard form with the rows of linearly independent, and the dual feasible set nonempty — so that is finite for every .)
import Mathlib.Analysis.Convex.Function import Mathlib.LinearAlgebra.LinearIndependent.Defs import Definitions.Def_LinearOptimization_OptimalCostFunction open Matrix /-- **Bertsimas & Tsitsiklis, Theorem 5.1 (p. 213).** The optimal cost `F(b)` of the standard form problem `min c'x, Ax = b, x ≥ 0` (rows of `A` linearly independent, dual feasible set nonempty) is a convex function of `b` on the set `S` of feasible right-hand sides. -/
theorem LinearOptimization.lp_optimal_cost_convex_in_rhs {m n : ℕ}
(A : Matrix (Fin m) (Fin n) ℝ) (c : Fin n → ℝ) (F : (Fin m → ℝ) → ℝ)
(hrank : LinearIndependent ℝ (fun i => A i))
(hdual : (dualFeasibleStd A c).Nonempty)
(hF : ∀ b ∈ feasibleRhsSet A, ((F b : ℝ) : EReal) = lpOptimalCostRhs A c b) :
ConvexOn ℝ (feasibleRhsSet A) F := by
sorry
Read-back
What the Lean code literally says, in plain math · claude-fable-5
Data: , an real matrix , a cost vector , and a function . Hypotheses: (a) the rows of form a linearly independent family in (full row rank; this forces ); (b) the set is nonempty; (c) for every in (the right-hand sides making the standard-form program feasible), the real number , viewed as an extended real, equals — so hypothesis (c) in particular forces this optimal value to be finite at every , while is completely unconstrained outside . Conclusion: is convex on in the standard sense that includes convexity of the domain: is a convex set, and for all and all with .
Confirmed by the mission captain (proposal self-audit).