Proof of Theorem 5.2, p. 25 — Γ nonsingular ⇒ S(α, Γ) = {x_*} and ¼cα(α − 1)‖x − x_*‖² ≤ r_{α,Γ}(x) − v(α, Γ)
ProvedProbMetricStab.Portfolio.nonsingular_quadratic_growthgrowth-conditionp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1stochastic-programmingstrong-convexity
Let be a nonsingular Borel probability measure on the unit sphere , and let be a constant (depending only on and ) with for all . Then for every stability index problem (22), , has a unique solution , i.e. , and
This quadratic growth of the risk around its unique minimizer is what makes the stability modulus of Theorem 5.2 of order .
Formalization Note The constant is the constant of the norm bound from the preceding sentence of the proof (the milestone nonsingular_norm_bound supplies one), so the statement quantifies over every with that property; it is fixed before , as on the page.
Preamble
import Mathlib import Definitions.Def_ProbMetricStab_Portfolio_Setting open MeasureTheory open scoped ENNReal
Formal statement
namespace ProbMetricStab.Portfolio
theorem nonsingular_quadratic_growth {s : ℕ} (Γ : Measure (Rs s)) (hΓ : IsSpectral Γ)
(hns : Nonsingular Γ) :
∀ c : ℝ, 0 < c → (∀ x : Rs s, c * ‖x‖ ^ 2 ≤ ∫ ξ in unitSphere s, |(inner ℝ x ξ : ℝ)| ^ 2 ∂Γ) →
∀ α : ℝ, 1 < α → α < 2 → ∃ xs : Rs s, solSet α Γ = {xs} ∧ ∀ x ∈ simplex s,
(1 / 4) * c * α * (α - 1) * ‖x - xs‖ ^ 2 ≤ risk α Γ x - optVal α Γ := by sorry
end ProbMetricStab.Portfolio
Source
Rachev & Römisch, Quantitative stability in stochastic programming: The method of probability metrics, preprint (edoc.hu-berlin.de), p. 25, proof of Theorem 5.2 ("We conclude that ... for each x ∈ X")
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.