Proof of Theorem 5, p. 18 — for fixed areas, the rates (17) minimize Σ A_k c/(μ_k − λA_k) under Σμ_k = B, with value c/(B−λ)(Σ√A_k)²
ProvedSpatialDemandEq.Optimal.mu_stepFix constants , and areas with , and write . Consider the rate problem (18) for a fixed partition: minimize
over rates with and for every . Let be the rates (17). Then
- is feasible: and for all ;
- its objective value is
- every other feasible has a strictly larger objective value.
This is the first step of the proof of Theorem 5: optimizing the service rates out of (15) leaves the partition problem (23) with the nonconvex .
Formalization Note The page derives the minimizer from the Lagrangian and states only the optimal value; the uniqueness clause 3 is part of the formal statement because Theorem 5's formula (17) for the rates needs it. The minimizer is also Kelly's square-root capacity allocation (Kelly 1979, Theorem 4.1, a referenced item of this mission) with arrival rates , unit costs and budget .
import Mathlib import Definitions.Def_SpatialDemandEq_Optimal_Model open MeasureTheory
namespace SpatialDemandEq.Optimal
theorem mu_step {K : ℕ} (c lam B : ℝ) (hc : 0 < c) (hlam : 0 < lam) (hB : lam < B)
(a : Fin K → ℝ) (ha : ∀ k, 0 < a k) (hsum : ∑ k, a k = 1) :
(∑ k, optRates lam B a k = B ∧ ∀ k, lam * a k < optRates lam B a k) ∧
rateCost c lam (optRates lam B a) a = c / (B - lam) * (∑ k, Real.sqrt (a k)) ^ 2 ∧
∀ μ : Fin K → ℝ, ∑ k, μ k = B → (∀ k, lam * a k < μ k) → μ ≠ optRates lam B a →
rateCost c lam (optRates lam B a) a < rateCost c lam μ a := by sorry
end SpatialDemandEq.Optimal
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.