Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 4.1, p. 798 — CMNS with an LP-optimal priority list is a (1+β)(1+1/β+max{1,(m−1)Δ/m})-approximation for P|r_j, prec|E[Σ w_j C_j]

Proved
StochSchedPrec.CMNS.theorem_4_1

by mikedeng1 · Oct 7, 2026 · Mathlib 0df444a (Lean v4.33.1)

approximation-algorithmlp-relaxationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1precedence-constraintsstochastic-scheduling

Consider an instance of the stochastic machine scheduling problem P ∣ rj,prec ∣ E[∑wjCj]\mathrm P\,|\,r_j,\mathit{prec}\,|\,\mathrm E[\sum w_jC_j]P∣rj​,prec∣E[∑wj​Cj​]: m≥1m\ge1m≥1 identical machines, acyclic precedence constraints, release dates rj≥0r_j\ge0rj​≥0 satisfying Assumption 2.1, weights wj≥0w_j\ge0wj​≥0, and independent nonnegative processing times PjP_jPj​ with CV[Pj]≤Δ\mathrm{CV}[P_j]\le\sqrt\DeltaCV[Pj​]≤Δ​ for all jjj and some Δ≥0\Delta\ge0Δ≥0. Let CLPC^{\mathrm{LP}}CLP be an optimal solution of the LP relaxation of §3, let LLL be a priority list according to nondecreasing CjLPC^{\mathrm{LP}}_jCjLP​, let β>0\beta>0β>0, and let σ\sigmaσ be the policy given by Algorithm CMNS with list LLL and thresholds β E[Pj]\beta\,\mathrm E[P_j]βE[Pj​]. Then the expected cost of CMNS is finite and, for every feasible nonanticipatory scheduling policy Π\PiΠ with integrable completion times,

E[∑j∈VwjCjCMNS(P)]≤α  E[∑j∈VwjCjΠ(P)],α=(1+β)(1+1β+max⁡{1,m−1mΔ}).\mathrm E\Big[\sum_{j\in V}w_jC^{\mathrm{CMNS}}_j(P)\Big]\le\alpha\;\mathrm E\Big[\sum_{j\in V}w_jC^\Pi_j(P)\Big],\qquad \alpha=(1+\beta)\Big(1+\frac1\beta+\max\Big\{1,\frac{m-1}m\Delta\Big\}\Big).E[j∈V∑​wj​CjCMNS​(P)]≤αE[j∈V∑​wj​CjΠ​(P)],α=(1+β)(1+β1​+max{1,mm−1​Δ}).

In particular CMNS is an α\alphaα-approximation. For NBUE processing times one may take Δ=1\Delta=1Δ=1, and β=1/2\beta=1/\sqrt2β=1/2​ gives α=3+22\alpha=3+2\sqrt2α=3+22​.

Formalization Note The comparison is against every admissible policy, which implies the paper's comparison with an optimal policy and does not presuppose that one exists. Nonanticipation and measurability of the CMNS policy are the paper's assertions (p. 795; §5, p. 800), taken as hypotheses; measurability, of the CMNS policy and of the comparators, means measurability for the law of PPP, which §5's universal measurability of every policy implies. The fact that the LP order is a linear extension of the precedence constraints is not assumed: it follows from E[Pj]>0\mathrm E[P_j]>0E[Pj​]>0 and the arc constraints of the LP.

Preamble
import Mathlib
import Definitions.Def_StochSchedPrec_CMNS_Model
import Definitions.Def_StochSchedPrec_CMNS_Algorithm

open MeasureTheory ProbabilityTheory
Formal statement
namespace StochSchedPrec.CMNS

theorem theorem_4_1 {V : Type*} [Fintype V] [DecidableEq V]
    {Ω : Type*} [MeasurableSpace Ω] {Pr : Measure Ω} [IsProbabilityMeasure Pr]
    (m : ℕ) (hm : 0 < m) (A : V → V → Prop) (hacyc : ∀ j, ¬ Relation.TransGen A j j)
    (r w : V → ℝ) (hr : ∀ j, 0 ≤ r j) (hw : ∀ j, 0 ≤ w j)
    (hAss : ∀ i j, Relation.TransGen A i j → r i ≤ r j)
    (P : V → Ω → ℝ) (hP : IsStochModel Pr P) (Δ : ℝ) (hΔ : 0 ≤ Δ) (hCV : HasCVBound Pr P Δ)
    (C : V → ℝ) (hC : IsLPOptimal m A Δ (fun j => ∫ ω, P j ω ∂Pr) w C)
    (L : Fin (Fintype.card V) ≃ V) (hL : IsSortedBy L C) (β : ℝ) (hβ : 0 < β)
    (σ : (V → ℝ) → V → ℝ) (hσ : IsCMNSPolicy m A r L β (fun j => ∫ ω, P j ω ∂Pr) σ)
    (hσna : IsNonanticipatory σ)
    (hσm : AEMeasurable σ (Pr.map (fun ω k => P k ω)))
    (pol : (V → ℝ) → V → ℝ) (hpol : IsAdmissiblePolicy m A r Pr P pol)
    (hpolm : AEMeasurable pol (Pr.map (fun ω k => P k ω))) :
    Integrable (fun ω => ∑ j, w j * (σ (fun k => P k ω) j + P j ω)) Pr ∧
    ∫ ω, ∑ j, w j * (σ (fun k => P k ω) j + P j ω) ∂Pr
      ≤ (1 + β) * (1 + 1 / β + max 1 (((m : ℝ) - 1) / m * Δ))
          * ∫ ω, ∑ j, w j * (pol (fun k => P k ω) j + P j ω) ∂Pr := by sorry

end StochSchedPrec.CMNS
Source
Skutella and Uetz, Stochastic machine scheduling with precedence constraints, SIAM J. Comput. 34(4) (2005) 788–802, p. 798, Theorem 4.1
Human review
  • Endorsed by Shuze Chen · Oct 7, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 7, 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