Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proof of Theorem 4, (56)–(59), PDF pp. 31–32 — the affine solution costs at most 3√m times the adaptive one

Proved
AffinePolicies.SqrtBound.constructed_cost_le

by mikedeng1 · Oct 5, 2026 · Mathlib 0df444a (Lean v4.33.1)

affine-policiesapproximation-boundp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1robust-optimization

Assume d≥0d\ge0d≥0 and U⊆R+m\mathcal U\subseteq\mathbb R^m_+U⊆R+m​. With μ\muμ, β1,…,βm\beta^1,\dots,\beta^mβ1,…,βm, a complete run u1,…,uKu^1,\dots,u^Ku1,…,uK, J1J_1J1​ of Algorithm A\mathcal AA, and a feasible solution (x∗,y∗)(x^*,y^*)(x∗,y∗) of ΠAdapt(U)\Pi_{Adapt}(\mathcal U)ΠAdapt​(U) as in the feasibility milestone, let x~=3m x∗\tilde x=3\sqrt m\,x^*x~=3m​x∗ and y~(b)=∑j∈J1bjμjy∗(βj)+y^\tilde y(b)=\sum_{j\in J_1}\frac{b_j}{\mu_j}y^*(\beta^j)+\hat yy~​(b)=∑j∈J1​​μj​bj​​y∗(βj)+y^​ with y^=2mK∑k=1Ky∗(uk)\hat y=\frac{2\sqrt m}{K}\sum_{k=1}^K y^*(u^k)y^​=K2m​​∑k=1K​y∗(uk). If ttt bounds the worst-case cost of (x∗,y∗)(x^*,y^*)(x∗,y∗), that is cTx∗+dTy∗(b)≤tc^Tx^*+d^Ty^*(b)\le tcTx∗+dTy∗(b)≤t for all b∈Ub\in\mathcal Ub∈U, then

cTx~+dTy~(b)≤3m⋅tfor all b∈U.c^T\tilde x+d^T\tilde y(b)\le 3\sqrt m\cdot t\qquad\text{for all } b\in\mathcal U .cTx~+dTy~​(b)≤3m​⋅tfor all b∈U.

Applied with ttt approaching zAdapt(U)z_{Adapt}(\mathcal U)zAdapt​(U) this is the bound cTx~+dTy~(b)≤3m⋅zAdapt(U)c^T\tilde x+d^T\tilde y(b)\le3\sqrt m\cdot z_{Adapt}(\mathcal U)cTx~+dTy~​(b)≤3m​⋅zAdapt​(U) of the paper.

Formalization Note Stated for every feasible (x∗,y∗)(x^*,y^*)(x∗,y∗) and every cost bound ttt, rather than for an optimal solution and t=zAdapt(U)t=z_{Adapt}(\mathcal U)t=zAdapt​(U); this avoids assuming attainment. Neither A≥0A\ge0A≥0 nor c≥0c\ge0c≥0 is needed and neither is assumed; μ>0\mu>0μ>0 is assumed.

Preamble
import Mathlib
import Definitions.Def_AffinePolicies_SqrtBound_Setting

open Matrix
Formal statement
namespace AffinePolicies.SqrtBound

theorem constructed_cost_le {m n₁ n₂ : ℕ} (A : Matrix (Fin m) (Fin n₁) ℝ)
    (B : Matrix (Fin m) (Fin n₂) ℝ) (c : Fin n₁ → ℝ) (d : Fin n₂ → ℝ) (U : Set (Fin m → ℝ))
    (hd : 0 ≤ d) (hUnn : ∀ b ∈ U, 0 ≤ b)
    (μ : Fin m → ℝ) (hμ : ∀ j, IsGreatest ((fun b : Fin m → ℝ => b j) '' U) (μ j))
    (hμpos : ∀ j, 0 < μ j)
    (bstar : Fin m → Fin m → ℝ) (hbstar : ∀ j, bstar j ∈ U ∧ bstar j j = μ j)
    (K : ℕ) (u : ℕ → Fin m → ℝ) (hrun : IsRun U μ K u)
    (xs : Fin n₁ → ℝ) (ys : (Fin m → ℝ) → Fin n₂ → ℝ) (hfs : AffinePolicies.Simplex.Feasible A B U xs ys) :
    ∀ t : ℝ, AffinePolicies.Simplex.CostLE c d U xs ys t →
      AffinePolicies.Simplex.CostLE c d U ((3 * Real.sqrt m) • xs)
        (AffinePolicies.Simplex.affinePolicy (thmPolicyP ys bstar μ (J1 μ u K)) (thmPolicyq ys u K))
        (3 * Real.sqrt m * t) := by sorry

end AffinePolicies.SqrtBound
Source
Bertsimas & Goyal, On the power and limitations of affine policies in two-stage adaptive optimization, Math. Program. Ser. A, DOI 10.1007/s10107-011-0444-4, Proof of Theorem 4, (56)–(59), PDF pp. 31–32
Human review
  • Endorsed by Shuze Chen · Oct 5, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 5, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me