Prove2Me
Navigate
DiscoverCollectionsFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

The self-scaled two-update direction is a descent direction

Proved
ShannoCG.SCONB.sconbDirection_descent

by mikedeng1 · Sep 29, 2026 · Mathlib 0df444a (Lean v4.33.1)

bfgsconjugate-gradientp2o-batch-p200bp2o-gran-per-chapterp2o-plan-paperp2o-v1quasi-newton

Let pt,yt,pk,yk,gk+1∈Rnp_t, y_t, p_k, y_k, g_{k+1} \in \mathbb R^npt​,yt​,pk​,yk​,gk+1​∈Rn with pt′yt>0p_t'y_t > 0pt′​yt​>0, pk′yk>0p_k'y_k > 0pk′​yk​>0 and gk+1≠0g_{k+1} \ne 0gk+1​=0, and let dk+1=−H^k+1gk+1d_{k+1} = -\hat H_{k+1} g_{k+1}dk+1​=−H^k+1​gk+1​ be the self-scaled two-update direction (H^k\hat H_kH^k​ from (37), H^k+1\hat H_{k+1}H^k+1​ its BFGS update (32) with (pk,yk)(p_k, y_k)(pk​,yk​)). Then dk+1d_{k+1}dk+1​ is a descent direction:

gk+1′ dk+1<0.g_{k+1}'\, d_{k+1} < 0.gk+1′​dk+1​<0.

No line-search exactness and no quadratic structure is required: the curvature condition p′y>0p'y > 0p′y>0 on each of the two updates suffices. This is why the method can be used with inexact line searches.

Formalization Note The paper's condition "pk′yk>0p_k'y_k > 0pk′​yk​>0" is imposed on every update, so it is assumed for both pairs (pt,yt)(p_t, y_t)(pt​,yt​) and (pk,yk)(p_k, y_k)(pk​,yk​). "Descent direction" is read as gk+1′dk+1<0g_{k+1}'d_{k+1} < 0gk+1′​dk+1​<0, which needs gk+1≠0g_{k+1} \ne 0gk+1​=0.

Preamble
import Mathlib
import Definitions.Def_ShannoCG_SCONB_sconbDirection

open Matrix
Formal statement
namespace ShannoCG.SCONB

/-- Shanno, *Conjugate Gradient Methods with Inexact Searches*, Math. Oper. Res. 3(3) (1978), §IV, p. 250 (PDF 7), the paragraph after (39): as the method uses two sequential
quasi-Newton updates, `p_k'y_k > 0` is sufficient to ensure that the search direction `d_{k+1}` is
always a descent direction. Stated here for the self-scaled algorithm (34), (38), (39), i.e.
`d_{k+1} = −Ĥ_{k+1} g_{k+1}` with `Ĥ_k` from (37) and `Ĥ_{k+1}` from (32).

**Formalization Note.** "`p_k'y_k > 0`" is the paper's condition on every update, so it is assumed
for both pairs `(p_t, y_t)` and `(p_k, y_k)`. "Descent direction" is `g_{k+1}' d_{k+1} < 0`, which
requires `g_{k+1} ≠ 0`. Generic vectors; no quadratic. -/
theorem sconbDirection_descent {n : ℕ} (pt yt pk yk g : Fin n → ℝ) (hpyt : 0 < pt ⬝ᵥ yt)
    (hpyk : 0 < pk ⬝ᵥ yk) (hg : g ≠ 0) :
    g ⬝ᵥ sconbDirection pt yt pk yk g < 0 := by sorry

end ShannoCG.SCONB
Source
Shanno, Conjugate Gradient Methods with Inexact Searches, Math. Oper. Res. 3(3) (1978) 244–256, DOI 10.1287/moor.3.3.244, p. 250 (PDF p. 7), §IV, the paragraph after eq. (39) (algorithm (34), (38), (39))
Human review
  • Endorsed by Shuze Chen · Oct 1, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 1, 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 paper
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me