Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Corollary 3.5, p. 485 — d.c. duality transports a local minimizer x* of g − h to a local minimizer y* ∈ ∂h(x*) of h* − g*

Proved
TaoAnDCA.LocalOpt.corollary_3_5

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

conjugatedc-programmingdualitylocal-optimalityp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Let g,h∈Γ0(Rn)g, h \in \Gamma_0(\mathbb R^n)g,h∈Γ0​(Rn) satisfy the standing inclusions dom⁡g⊂dom⁡h\operatorname{dom} g \subset \operatorname{dom} hdomg⊂domh and dom⁡h∗⊂dom⁡g∗\operatorname{dom} h^* \subset \operatorname{dom} g^*domh∗⊂domg∗, with differences taken under the convention +∞−(+∞)=+∞+\infty - (+\infty) = +\infty+∞−(+∞)=+∞. Let x∗∈dom⁡gx^* \in \operatorname{dom} gx∗∈domg admit an open neighbourhood UUU such that

g(x)−h(x)≥g(x∗)−h(x∗)∀x∈U∩dom⁡g,g(x) - h(x) \ge g(x^*) - h(x^*) \qquad \forall x \in U \cap \operatorname{dom} g,g(x)−h(x)≥g(x∗)−h(x∗)∀x∈U∩domg,

so that x∗x^*x∗ is a local minimizer of g−hg - hg−h, and let y∗∈∂h(x∗)y^* \in \partial h(x^*)y∗∈∂h(x∗). If

y∗∈int⁡(dom⁡g∗)and∂g∗(y∗)⊂U,(12)y^* \in \operatorname{int}(\operatorname{dom} g^*) \quad\text{and}\quad \partial g^*(y^*) \subset U, \tag{12}y∗∈int(domg∗)and∂g∗(y∗)⊂U,(12)

then y∗y^*y∗ is a local minimizer of the dual d.c. objective h∗−g∗h^* - g^*h∗−g∗: the value h∗(y∗)−g∗(y∗)h^*(y^*) - g^*(y^*)h∗(y∗)−g∗(y∗) is finite and

h∗(y∗)−g∗(y∗)≤h∗(y)−g∗(y)for all y in a neighbourhood of y∗.h^*(y^*) - g^*(y^*) \le h^*(y) - g^*(y) \qquad \text{for all } y \text{ in a neighbourhood of } y^*.h∗(y∗)−g∗(y∗)≤h∗(y)−g∗(y)for all y in a neighbourhood of y∗.

This is the duality transportation of local minimizers: a local solution of the primal d.c. program (P) produces a local solution of the dual program (D), which matters when (D) is easier to solve locally than (P). It improves Toland's earlier result, which assumed g∗g^*g∗ differentiable on the whole space.

Formalization Note. The neighbourhood UUU of the local-minimum property is named because (12) refers to it, and it is taken open: the proof's upper-semicontinuity step needs an open set containing ∂g∗(y∗)\partial g^*(y^*)∂g∗(y∗). The page's "x∗∈dom⁡∂hx^* \in \operatorname{dom}\partial hx∗∈dom∂h" follows from y∗∈∂h(x∗)y^* \in \partial h(x^*)y∗∈∂h(x∗); x∗∈dom⁡gx^* \in \operatorname{dom} gx∗∈domg is the finiteness part of "local minimizer". The conclusion is the full local-minimizer predicate (4) for the pair (h∗,g∗)(h^*, g^*)(h∗,g∗), including finiteness.

Preamble
import Mathlib
import Definitions.Def_TaoAnDCA_LocalOpt_Setting

open Filter Topology
Formal statement
namespace TaoAnDCA.LocalOpt

/-- Corollary 3.5, p. 485 (d.c. duality transportation of a local minimizer): let `x*` be a
local minimizer of `g − h` with neighbourhood `U` (`g(x) − h(x) ≥ g(x*) − h(x*)` for all
`x ∈ U ∩ dom g`), and let `y* ∈ ∂h(x*)`. If `y* ∈ int (dom g*)` and `∂g*(y*) ⊂ U` (12), then
`y*` is a local minimizer of `h* − g*`. The neighbourhood `U` is taken open, as the proof's
upper-semicontinuity step requires. -/
theorem corollary_3_5 {n : ℕ} (g h : EuclideanSpace ℝ (Fin n) → EReal) (hgh : TaoAnDCA.GlobalOpt.DCStanding g h)
    (xs ys : EuclideanSpace ℝ (Fin n)) (hfin : xs ∈ TaoAnDCA.GlobalOpt.effDom g)
    (U : Set (EuclideanSpace ℝ (Fin n))) (hU : IsOpen U) (hxU : xs ∈ U)
    (hmin : ∀ x ∈ U, x ∈ TaoAnDCA.GlobalOpt.effDom g → TaoAnDCA.GlobalOpt.dcSub g h xs ≤ TaoAnDCA.GlobalOpt.dcSub g h x)
    (hy : ys ∈ TaoAnDCA.GlobalOpt.subdiff h xs)
    (h12a : ys ∈ interior (TaoAnDCA.GlobalOpt.effDom (CondatPD.FinDim.conj g)))
    (h12b : TaoAnDCA.GlobalOpt.subdiff (CondatPD.FinDim.conj g) ys ⊆ U) :
    IsDCLocalMin (CondatPD.FinDim.conj h) (CondatPD.FinDim.conj g) ys := by sorry

end TaoAnDCA.LocalOpt
Source
Pham Dinh & Le Thi, A d.c. optimization algorithm for solving the trust-region subproblem, SIAM J. Optim. 8 (1998), p. 485, Corollary 3.5, (12)
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