Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem A.1, (A.2) and (A.5), p. 206, at ε = 0 as used in the proof of Theorem 2.4 — CG conjugacy and Krylov spans

Proved
TruncNewton.NegCurv.theorem_A_1_eps_zero

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

conjugacyconjugate-gradientkrylov-subspacep2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Let HHH be a symmetric linear operator on Rn\mathbb{R}^nRn, let g∈Rng\in\mathbb{R}^ng∈Rn, and let d0,d1,…d_0,d_1,\dotsd0​,d1​,… be the conjugate gradient directions of the TNCG minor iteration for HHH and ggg (Steps 1–4, p. 194). Let k≥0k\ge 0k≥0 and suppose that the curvature along every direction up to kkk is positive:

diTHdi>0,i=0,1,…,k.d_i^{\mathsf T}Hd_i>0,\qquad i=0,1,\dots,k.diT​Hdi​>0,i=0,1,…,k.

Then the directions are mutually HHH-conjugate and span the Krylov space of ggg:

  1. (A.2)   diTHdj=0\;d_i^{\mathsf T}Hd_j=0diT​Hdj​=0 for i≠ji\ne ji=j, i,j=0,1,…,ki,j=0,1,\dots,ki,j=0,1,…,k;
  2. (A.5)   [d0,d1,…,dk]=[g,Hg,…,Hkg]\;[d_0,d_1,\dots,d_k]=[g,Hg,\dots,H^kg][d0​,d1​,…,dk​]=[g,Hg,…,Hkg], where [⋅][\cdot][⋅] denotes the linear span.

This is the part of the Hestenes–Stiefel relations of Theorem A.1 that the proof of Theorem 2.4 uses: conjugacy with positive curvature, and the identification of the directions' span with the Krylov space, are what let one pass from the CG run to the spectral statement of Theorem A.5.

Formalization Note The paper states Theorem A.1 under "Let ε>0\varepsilon>0ε>0" with hypothesis diTHdi>εδid_i^{\mathsf T}Hd_i>\varepsilon\delta_idiT​Hdi​>εδi​; the proof of Theorem 2.4 applies (A.2) and (A.5) at ε=0\varepsilon=0ε=0, where that hypothesis reads diTHdi>0d_i^{\mathsf T}Hd_i>0diT​Hdi​>0, and this item states exactly that case. The page prints (A.5) as [d0,…,dk]=[g,Hg,…,Hk−1g][d_0,\dots,d_k]=[g,Hg,\dots,H^{k-1}g][d0​,…,dk​]=[g,Hg,…,Hk−1g], with k+1k+1k+1 vectors on the left and kkk on the right; this is a misprint, and the item states the correct [g,…,Hkg][g,\dots,H^kg][g,…,Hkg] (consistent with (A.23)). Symmetry of HHH is the paper's standing setting (HHH is a Hessian). The other relations (A.3), (A.4), (A.6), (A.7) of Theorem A.1 are not part of this item.

Preamble
import Mathlib
import Definitions.Def_TruncNewton_NegCurv_Setting
Formal statement
namespace TruncNewton.NegCurv

theorem theorem_A_1_eps_zero {n : ℕ}
    (H : EuclideanSpace ℝ (Fin n) →L[ℝ] EuclideanSpace ℝ (Fin n)) (hH : IsSelfAdjoint H)
    (g : EuclideanSpace ℝ (Fin n)) (k : ℕ)
    (hcurv : ∀ i ≤ k, 0 < inner ℝ ((cg H g i).d) (H (cg H g i).d)) :
    (∀ i j, i ≤ k → j ≤ k → i ≠ j → inner ℝ ((cg H g i).d) (H (cg H g j).d) = 0) ∧
    Submodule.span ℝ ((fun i => (cg H g i).d) '' Set.Iic k) =
      Submodule.span ℝ ((fun i => (H ^ i) g) '' Set.Iic k) := by sorry

end TruncNewton.NegCurv
Source
Dembo and Steihaug, Truncated-Newton algorithms for large-scale unconstrained optimization, Math. Programming 26 (1983), p. 206, Theorem A.1, (A.2) and (A.5), at ε = 0 as used in the proof of Theorem 2.4, p. 211
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