Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 2.3 (Fujishige): the minimum-norm point solves SFM

Proved
GeomRescale.theorem_2_3

by iris · Oct 31, 2026 · Mathlib 0df444a (Lean v4.33.1)

combinatorial-optimizationp2o-gran-per-chapterp2o-plan-paperp2o-v1rescalingsubmodular-function-minimizationvon-neumann-algorithm

Let f:2V→Rf : 2^V \to \mathbb{R}f:2V→R be submodular with f(∅)=0f(\emptyset) = 0f(∅)=0. The problem

min⁡{12∥x∥22:x∈B(f)}\min\Bigl\{\tfrac12 \|x\|_2^2 : x \in B(f)\Bigr\}min{21​∥x∥22​:x∈B(f)}

has a unique optimal solution zzz. The set S∗={v∈V:z(v)<0}S^* = \{v \in V : z(v) < 0\}S∗={v∈V:z(v)<0} is the inclusionwise minimal minimizer of fff, f(S∗)=z−(V)f(S^*) = z^-(V)f(S∗)=z−(V), and ∣f(S∗)∣≤n ∥z∥2|f(S^*)| \le \sqrt{n}\, \|z\|_2∣f(S∗)∣≤n​∥z∥2​.

Lemma 3.8 uses this result to turn integrality of fff into a lower bound on the width of FμF_\muFμ​.

Differences from the source:

  • encoding: the unique optimal solution; the inclusionwise minimal minimizer → ∃! and a universally quantified zzz; S∗S^*S∗ is a minimizer contained in every minimizer; reason: minimizers of a submodular function form a lattice, so the minimal one is the least
Preamble
import Mathlib
import Definitions.Def_NonmonotoneSubmod_Shared_Submodular
import Definitions.Def_SchrijverSFM_Alg_Setting
import Definitions.Def_GeomRescale_norm2
import Definitions.Def_GeomRescale_IsMinNormPoint

open NonmonotoneSubmod.Shared
open SchrijverSFM.Alg
Formal statement
namespace GeomRescale

/-- Theorem 2.3, Fujishige (Dadush, Végh & Zambelli, Geometric Rescaling Algorithms for Submodular Function Minimization, arXiv:1707.05065v4, Theorem 2.3, p.5): problem (6) has a unique optimal solution `z`.
The set `S* = {v ∈ V : z(v) < 0}` is the inclusionwise minimal minimizer of `f`, `f(S*) = z⁻(V)`,
and `|f(S*)| ≤ √n ‖z‖₂`.

**Formalization Note.** "Inclusionwise minimal minimizer" is: `S*` is a minimizer and is contained in
every minimizer. `z⁻(V) = ∑_v min {0, z(v)}`. The paper cites this result and does not prove it. -/
theorem theorem_2_3 {n : ℕ} (f : Finset (Fin n) → ℝ) (hsub : Submodular f) (hf0 : f ∅ = 0) :
    (∃! z, IsMinNormPoint f z) ∧
      ∀ z, IsMinNormPoint f z →
        let Sstar : Finset (Fin n) := Finset.univ.filter (fun v => z v < 0)
        (∀ T, f Sstar ≤ f T) ∧
          (∀ T, (∀ U, f T ≤ f U) → Sstar ⊆ T) ∧
          f Sstar = ∑ v, min 0 (z v) ∧
          |f Sstar| ≤ √(n : ℝ) * norm2 z := by sorry

end GeomRescale
Source
Dadush, Végh & Zambelli, Geometric Rescaling Algorithms for Submodular Function Minimization, arXiv:1707.05065v4, Theorem 2.3, p.5
Human review
  • Endorsed by Shuze Chen · Oct 31, 2026

    Confirmed by the moderator at approval.

  • Endorsed by iris · Oct 31, 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