Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 7.5, p. 101 — existence, uniqueness, and FTRL stability

Proved
ModernOnlineLearning.FTRL.lemma_7_5

by mikedeng1 · Oct 9, 2026 · Mathlib 0df444a (Lean v4.33.1)

ftrlonline-learningp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-bookp2o-v1strong-convexity

Let VVV be nonempty, closed and convex. Suppose FtF_tFt​ is closed, subdifferentiable and strongly convex on VVV. Then it has a unique minimizer xtx_txt​. Suppose also that Ft+ℓtF_t+\ell_tFt​+ℓt​ is closed, subdifferentiable and λt\lambda_tλt​-strongly convex on VVV, with λt>0\lambda_t>0λt​>0. For any minimizer xt+1x_{t+1}xt+1​ of Ft+1F_{t+1}Ft+1​ and any subgradient gt∈∂ℓt(xt)g_t\in\partial\ell_t(x_t)gt​∈∂ℓt​(xt​),

Ft(xt)−Ft+1(xt+1)+ℓt(xt)≤gt(xt−xt+1)−λt2∥xt−xt+1∥2+ψt(xt+1)−ψt+1(xt+1)≤∥gt∥∗22λt+ψt(xt+1)−ψt+1(xt+1).F_t(x_t)-F_{t+1}(x_{t+1})+\ell_t(x_t)\le g_t(x_t-x_{t+1})-\frac{\lambda_t}{2}\|x_t-x_{t+1}\|^2+\psi_t(x_{t+1})-\psi_{t+1}(x_{t+1})\le\frac{\|g_t\|_*^2}{2\lambda_t}+\psi_t(x_{t+1})-\psi_{t+1}(x_{t+1}).Ft​(xt​)−Ft+1​(xt+1​)+ℓt​(xt​)≤gt​(xt​−xt+1​)−2λt​​∥xt​−xt+1​∥2+ψt​(xt+1​)−ψt+1​(xt+1​)≤2λt​∥gt​∥∗2​​+ψt​(xt+1​)−ψt+1​(xt+1​).

This bounds the change in objective values by the dual norm of a loss subgradient and the change in regularizer.

Formalization Note Existence of xt+1x_{t+1}xt+1​ is stated by quantifying over any such minimizer; the lemma's existence assertion concerns xtx_txt​. Subgradients are full-space continuous linear functionals. The positivity of λt\lambda_tλt​ makes the printed denominator meaningful.

Preamble
import Mathlib
import Definitions.Def_ModernOnlineLearning_FTRL_Defs

set_option autoImplicit false
Formal statement
namespace ModernOnlineLearning.FTRL

/-- Orabona, Lemma 7.5, p. 101: existence, uniqueness, and the FTRL stability chain. -/
theorem lemma_7_5
    {E : Type*} [NormedAddCommGroup E] [NormedSpace ℝ E]
    [FiniteDimensional ℝ E]
    (V : Set E) (hVclosed : IsClosed V) (hVconvex : Convex ℝ V)
    (hVnonempty : V.Nonempty) (ψ ℓ : ℕ → E → ℝ) (t : ℕ) (ht : 1 ≤ t)
    (lam : ℝ) (hlam : 0 < lam)
    (hFtClosed : IsClosedFunctionOn V (F ψ ℓ t))
    (hFtSubdiff : IsSubdifferentiableOn V (F ψ ℓ t))
    (hFtStrong : ∃ κ : ℝ, 0 < κ ∧ StrongConvexOn V κ (F ψ ℓ t))
    (hFtlClosed : IsClosedFunctionOn V (fun y => F ψ ℓ t y + ℓ t y))
    (hFtlSubdiff : IsSubdifferentiableOn V (fun y => F ψ ℓ t y + ℓ t y))
    (hFtlStrong : StrongConvexOn V lam (fun y => F ψ ℓ t y + ℓ t y)) :
    (∃! z : E, IsMinimizerOn V (F ψ ℓ t) z) ∧
    ∀ (x y : E), IsMinimizerOn V (F ψ ℓ t) x →
      IsMinimizerOn V (F ψ ℓ (t + 1)) y →
      (∃ g : E →L[ℝ] ℝ, IsSubgradientAt (ℓ t) x g) →
      ∀ g : E →L[ℝ] ℝ, IsSubgradientAt (ℓ t) x g →
      F ψ ℓ t x - F ψ ℓ (t + 1) y + ℓ t x ≤
        g (x - y) - lam / 2 * ‖x - y‖ ^ 2 + ψ t y - ψ (t + 1) y ∧
      g (x - y) - lam / 2 * ‖x - y‖ ^ 2 + ψ t y - ψ (t + 1) y ≤
        ‖g‖ ^ 2 / (2 * lam) + ψ t y - ψ (t + 1) y := by sorry

end ModernOnlineLearning.FTRL
Source
Orabona, arXiv:1912.13213v10, Lemma 7.5, p. 101
Human review
  • Endorsed by Shuze Chen · Oct 9, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 9, 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