§4.1, after (43) — under strict convexity, (43) has a unique solution , continuous in
ProvedGenCMu.HeavyTraffic.sec_4_1_unique_continuousconvex-optimizationp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1parametric-optimization
Let , let be continuous, let each be strictly convex, continuous and nondecreasing on , and let be continuous. Then for every the problem
has exactly one solution , and is continuous on .
This defines the map of the lower bound (45) and of Proposition 7.
Formalization Note The paper states uniqueness and continuity for " convex increasing". That is not enough: with , , , every point of is optimal. We state the claim under strict convexity, which Assumption 3 provides and Proposition 7 uses. The statement is for general continuous positive rates; the paper's case is , .
Preamble
import Mathlib import Definitions.Def_GenCMu_HeavyTraffic_Model import Definitions.Def_GenCMu_HeavyTraffic_Sequence import Definitions.Def_GenCMu_HeavyTraffic_Limits
Formal statement
namespace GenCMu.HeavyTraffic
open Filter Topology Finset
/-- §4.1, after (43) (p. 820), under strict convexity: for continuous positive rates and a continuous
nonnegative total `y`, problem (43) has a unique solution `g ∘ y(t)` for every `t ∈ [0, 1]`, and
`t ↦ g ∘ y(t)` is continuous. -/
theorem sec_4_1_unique_continuous {d : ℕ} (hd : 0 < d) (lam rho : ℝ → Fin d → ℝ)
(Cs : Fin d → ℝ → ℝ)
(hlam_cont : ContinuousOn lam (Set.Icc 0 1)) (hrho_cont : ContinuousOn rho (Set.Icc 0 1))
(hlam : ∀ t ∈ Set.Icc (0 : ℝ) 1, ∀ k, 0 < lam t k)
(hrho : ∀ t ∈ Set.Icc (0 : ℝ) 1, ∀ k, 0 < rho t k)
(hstrict : ∀ k, StrictConvexOn ℝ (Set.Ici 0) (Cs k))
(hcont : ∀ k, ContinuousOn (Cs k) (Set.Ici 0))
(hmono : ∀ k, MonotoneOn (Cs k) (Set.Ici 0))
(y : ℝ → ℝ) (hy : ContinuousOn y (Set.Icc 0 1)) (hy0 : ∀ t ∈ Set.Icc (0 : ℝ) 1, 0 ≤ y t) :
∃ G : ℝ → Fin d → ℝ, ContinuousOn G (Set.Icc 0 1) ∧
∀ t ∈ Set.Icc (0 : ℝ) 1, IsMin43 (lam t) (rho t) Cs (y t) (G t) ∧
∀ x : Fin d → ℝ, IsMin43 (lam t) (rho t) Cs (y t) x → x = G t := by sorry
end GenCMu.HeavyTraffic
Source
Van Mieghem, Dynamic Scheduling with Convex Delay Costs: The Generalized cμ Rule, Ann. Appl. Probab. 5(3) (1995), §4.1, after (43), p. 820
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.