Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Exact performance for separable profiles

Proved
PGDRevision20261113.separable_profiles

by Somehow · Nov 13, 2026 · Mathlib 0df444a (Lean v4.33.1)

bregman-geometryconvex-optimizationproximal-gradient

Let d≥1d\ge1d≥1, and let fj,gj,ϕjf_j,g_j,\phi_jfj​,gj​,ϕj​ be scalar profiles on open intervals IjI_jIj​, for 1≤j≤d1\le j\le d1≤j≤d. Fix one inverse stepsize p>0p>0p>0, and assume that every scalar profile satisfies the hypotheses of the exact-profile theorem: twice continuous differentiability, Aj=fj′′≥0A_j=f_j''\ge0Aj​=fj′′​≥0, Bj=gj′′≥0B_j=g_j''\ge0Bj​=gj′′​≥0, Aj+Bj>0A_j+B_j>0Aj​+Bj​>0, kj=ϕj′′>0k_j=\phi_j''>0kj​=ϕj′′​>0, and surjectivity of pϕj′+gj′p\phi_j'+g_j'pϕj′​+gj′​ onto R\mathbb RR.

Use separable functions f=∑jfjf=\sum_jf_jf=∑j​fj​, g=∑jgjg=\sum_jg_jg=∑j​gj​, ϕ=∑jϕj\phi=\sum_j\phi_jϕ=∑j​ϕj​ on the product interval and the actual coordinatewise implicit update. For Fℓ=f+g+⟨ℓ,x⟩F_\ell=f+g+\langle\ell,x\rangleFℓ​=f+g+⟨ℓ,x⟩, allow any ℓ∈Rd\ell\in\mathbb R^dℓ∈Rd and initial state with nonzero gradient. Define

ρj=sup⁡t∈Ij∣pkj(t)−Aj(t)pkj(t)+Bj(t)∣.\rho_j=\sup_{t\in I_j} \left|\frac{pk_j(t)-A_j(t)}{pk_j(t)+B_j(t)}\right|.ρj​=t∈Ij​sup​​pkj​(t)+Bj​(t)pkj​(t)−Aj​(t)​​.

For every finite N≥1N\ge1N≥1, the exact squared Euclidean gradient ratio is

sup⁡ℓ,x0: ∇Fℓ(x0)≠0∥∇Fℓ(xN)∥2∥∇Fℓ(x0)∥2=max⁡1≤j≤dρj2N.\boxed{ \sup_{\ell,x_0:\,\nabla F_\ell(x_0)\ne0} \frac{\|\nabla F_\ell(x_N)\|^2}{\|\nabla F_\ell(x_0)\|^2} =\max_{1\le j\le d}\rho_j^{2N}.}ℓ,x0​:∇Fℓ​(x0​)=0sup​∥∇Fℓ​(x0​)∥2∥∇Fℓ​(xN​)∥2​=1≤j≤dmax​ρj2N​.​

Infinite values are interpreted in the extended nonnegative reals. This is the product-profile extension of the scalar exact-performance result.

Related results. Exact profile performance (Theorem 4.1) · Geometry classification (Theorem 5.1) · Parameter localization (Theorem 7.1)

Preamble
import Definitions.Def_PGDRevision20261113_Profile
open PGDRevision20261113
Formal statement
theorem PGDRevision20261113.separable_profiles : SeparableProfilesClaim := by sorry
Source
Author manuscript: Exact Residual Bounds and Optimal Geometry Selection for Proximal Gradient Methods, revised reading edition, 13 November 2026, Corollary 4.2, p. 11; source file sections/04_bregman_framework.tex, label cor:bg_separable.

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