Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

The DCA mapping bound on the convex-objective branch

Open
PGDRevision20261113.dca_convex_objective_bound

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

bregman-geometryconvex-optimizationproximal-gradient

Let u:Rd→(−∞,+∞]u:\mathbb R^d\to(-\infty,+\infty]u:Rd→(−∞,+∞] be proper, closed and μ1\mu_1μ1​-strongly convex. Let v:Rd→Rv:\mathbb R^d\to\mathbb Rv:Rd→R be differentiable, with both

v−μ22∥⋅∥2,L22∥⋅∥2−vv-\frac{\mu_2}{2}\|\cdot\|^2,\qquad \frac{L_2}{2}\|\cdot\|^2-vv−2μ2​​∥⋅∥2,2L2​​∥⋅∥2−v

convex. Assume

μ1>0,μ2<L2≤μ1,μ1+μ2>0.\mu_1>0,\qquad \mu_2<L_2\le\mu_1,\qquad \mu_1+\mu_2>0.μ1​>0,μ2​<L2​≤μ1​,μ1​+μ2​>0.

Write H=u−vH=u-vH=u−v. Starting from x0∈dom⁡ux_0\in\operatorname{dom}ux0​∈domu, perform N≥2N\ge2N≥2 DCA updates satisfying

0∈∂u(xi+1)−∇v(xi)(0≤i<N).0\in\partial u(x_{i+1})-\nabla v(x_i)\quad(0\le i<N).0∈∂u(xi+1​)−∇v(xi​)(0≤i<N).

For k≥1k\ge1k≥1, let Ek(z)=∑j=12kz−jE_k(z)=\sum_{j=1}^{2k}z^{-j}Ek​(z)=∑j=12k​z−j when z≠0z\ne0z=0, and Ek(0)=+∞E_k(0)=+\inftyEk​(0)=+∞. The terminal displacement obeys

μ12∥xN−1−xN∥2≤H(x0)−H(xN)1+μ2/μ1+min⁡{EN−1(L2/μ1),EN−1(μ2/μ1)}.\boxed{\frac{\mu_1}{2}\|x_{N-1}-x_N\|^2 \le \frac{H(x_0)-H(x_N)} {1+\mu_2/\mu_1+ \min\{E_{N-1}(L_2/\mu_1),E_{N-1}(\mu_2/\mu_1)\}}.}2μ1​​∥xN−1​−xN​∥2≤1+μ2​/μ1​+min{EN−1​(L2​/μ1​),EN−1​(μ2​/μ1​)}H(x0​)−H(xN​)​.​

The minimum-displacement bound follows on the same parameter region. This statement covers the complete convex-objective branch, including both signs of L2+μ2L_2+\mu_2L2​+μ2​.

Formalization Note. The DCA subgradient inclusion is represented by minimization of the convex function u−⟨∇v(xi),⋅⟩u-\langle\nabla v(x_i),\cdot\rangleu−⟨∇v(xi​),⋅⟩. The zero-argument energy is encoded as extended-real infinity before taking the minimum.

Related results. Fixed-penalty exact curve · Optimal constant stepsizes

Preamble
import Definitions.Def_PGDRevision20261113_Euclidean
open PGDRevision20261113
Formal statement
theorem PGDRevision20261113.dca_convex_objective_bound : DCAConvexObjectiveClaim := by sorry
Source
Author manuscript: Exact Residual Bounds and Optimal Geometry Selection for Proximal Gradient Methods, revised reading edition, 13 November 2026, Corollary 2.8, p. 7; source file sections/02_dca_consequence.tex, label cor:dca.

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