Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 6.1 — a limit of δ-locally optimal progressive hedging is a stationary point of (P) and a local solution of (P̃)

Proved
ProgHedging.Nonconvex.theorem_6_1

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

nonconvex-optimizationnonsmooth-analysisp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1progressive-hedgingstochastic-programming

Consider the scenario model under its standing assumptions, with no convexity. Fix r>0r>0r>0 and δ>0\delta>0δ>0, and let (Xν)(X^\nu)(Xν), (Wν)(W^\nu)(Wν) be generated by the progressive hedging algorithm in such a way that in each iteration Xν+1(s)X^{\nu+1}(s)Xν+1(s) is δ\deltaδ-locally optimal for the scenario subproblem

(Psν)minimize fs(x)+x⋅Wν(s)+12r∣x−X^ν(s)∣2 over x∈Cs,(P^\nu_s)\qquad \text{minimize } f_s(x)+x\cdot W^\nu(s)+\tfrac12 r|x-\hat X^\nu(s)|^2 \text{ over } x\in C_s ,(Psν​)minimize fs​(x)+x⋅Wν(s)+21​r∣x−X^ν(s)∣2 over x∈Cs​,

with X^ν=JXν\hat X^\nu=JX^\nuX^ν=JXν and Wν+1=Wν+rKXν+1W^{\nu+1}=W^\nu+rKX^{\nu+1}Wν+1=Wν+rKXν+1. Suppose Xν→X∗X^\nu\to X^*Xν→X∗ and Wν→W∗W^\nu\to W^*Wν→W∗. Then:

  1. X∗X^*X∗ and W∗W^*W∗ satisfy the optimality conditions of Theorem 4.1: X∗∈NX^*\in\mathcal NX∗∈N, X∗∈CX^*\in\mathcal CX∗∈C, W∗∈MW^*\in\mathcal MW∗∈M and
−W∗(s)∈∂fs(X∗(s))+NCs(X∗(s))for all s∈S;-W^*(s)\in\partial f_s(X^*(s))+N_{C_s}(X^*(s))\qquad\text{for all } s\in S;−W∗(s)∈∂fs​(X∗(s))+NCs​​(X∗(s))for all s∈S;

in this sense X∗X^*X∗ is a stationary point of (P); 2. X∗X^*X∗ is a locally optimal solution of the problem (P~)(\tilde P)(P~) obtained from (P) by replacing each fsf_sfs​ by

f~s(x)=fs(x)+12r∣x−X∗(s)∣2,(6.1)\tilde f_s(x)=f_s(x)+\tfrac12 r|x-X^*(s)|^2, \tag{6.1}f~​s​(x)=fs​(x)+21​r∣x−X∗(s)∣2,(6.1)

that is, ∑spsf~s(X∗(s))≤∑spsf~s(X(s))\sum_s p_s\tilde f_s(X^*(s))\le\sum_s p_s\tilde f_s(X(s))∑s​ps​f~​s​(X∗(s))≤∑s​ps​f~​s​(X(s)) for all X∈C∩NX\in\mathcal C\cap\mathcal NX∈C∩N near X∗X^*X∗; 3. W∗W^*W∗ is a corresponding multiplier vector for (P~)(\tilde P)(P~): −W∗(s)∈∂f~s(X∗(s))+NCs(X∗(s))-W^*(s)\in\partial\tilde f_s(X^*(s))+N_{C_s}(X^*(s))−W∗(s)∈∂f~​s​(X∗(s))+NCs​​(X∗(s)) for all s∈Ss\in Ss∈S.

Here ∂\partial∂ is Clarke's generalized gradient and NCN_CNC​ Clarke's normal cone. Without convexity one cannot expect the algorithm to solve (P) globally; the theorem says that whenever it converges, its limit is as good a "solution" as can be hoped for.

Formalization Note No convexity and no constraint qualification are assumed: the standing assumptions of the model (nonempty closed CsC_sCs​, locally Lipschitz fsf_sfs​, bounded level sets) are the only hypotheses besides the run and its convergence. The δ\deltaδ-neighbourhood of (Psν)(P^\nu_s)(Psν​) is the closed Euclidean ball. Convergence of policies and "near X∗X^*X∗" refer to the (norm-independent) topology of the finite-dimensional space E\mathcal EE. The algorithm starts from X0∈CX^0\in\mathcal CX0∈C and W0∈MW^0\in\mathcal MW0∈M as on p. 9.

Preamble
import Mathlib
import Definitions.Def_ProgHedging_Convex_Problem
import Definitions.Def_ProgHedging_Nonconvex_Algorithm
import Definitions.Def_ClarkeGradients_Shared_generalizedGradient
import Definitions.Def_ClarkeGradients_FlowInvariance_normalCone

open scoped Pointwise
open Filter Topology
Formal statement
namespace ProgHedging.Nonconvex

/-- Theorem 6.1, p. 30: if progressive hedging with `δ`-locally optimal subproblem solutions
generates `X^ν → X*` and `W^ν → W*`, then `X*`, `W*` satisfy the optimality conditions of
Theorem 4.1; moreover `X*` is locally optimal for, and `W*` a multiplier vector for, the problem
`(P̃)` obtained by replacing each `f_s` with `f̃_s(x) = f_s(x) + ½r|x − X*(s)|²` (6.1). -/
theorem theorem_6_1 {S : Type*} [Fintype S] {n T : ℕ} (pr : Problem S n T) {r δ : ℝ}
    (hr : 0 < r) (hδ : 0 < δ) (X W : ℕ → ProgHedging.Convex.Policy S n) (hseq : pr.IsLocalPHSeq r δ X W)
    (Xs Ws : ProgHedging.Convex.Policy S n) (hX : Tendsto X atTop (𝓝 Xs)) (hW : Tendsto W atTop (𝓝 Ws)) :
    pr.OptCond41 Xs Ws ∧
    (∃ ε > 0, ∀ Y ∈ pr.adm ∩ pr.N, dist Y Xs < ε →
      ∑ s, pr.p s * pr.fTilde r Xs s (Xs s) ≤ ∑ s, pr.p s * pr.fTilde r Xs s (Y s)) ∧
    ∀ s, -Ws s ∈ ClarkeGradients.Shared.generalizedGradient (pr.fTilde r Xs s) (Xs s) +
      ClarkeGradients.FlowInvariance.normalCone (pr.C s) (Xs s) := by sorry

end ProgHedging.Nonconvex
Source
Rockafellar and Wets, Scenarios and policy aggregation in optimization under uncertainty, IIASA Working Paper WP-87-119 (1987), p. 30, Theorem 6.1
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 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