Prove2Me
Navigate
DiscoverFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

Proof of Theorem 3.18, p. 291 — f(y_t) − f(x*) ≤ ((α + β)/2)‖x₁ − x*‖²(1 − 1/√κ)^{t−1}

Open
ConvexOptAlg.NesterovStrong.thm_3_18_rate

by mikedeng1 · Oct 5, 2026 · Mathlib 0df444a (Lean v4.33.1)

accelerated-gradientconvergence-rateconvex-optimizationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-bookp2o-v1

Let f:Rn→Rf:\mathbb R^n\to\mathbb Rf:Rn→R be α\alphaα-strongly convex and β\betaβ-smooth with α,β>0\alpha,\beta>0α,β>0, κ=β/α\kappa=\beta/\alphaκ=β/α, and let x∗x^*x∗ be a minimizer of fff on Rn\mathbb R^nRn. Let (xt),(yt)(x_t),(y_t)(xt​),(yt​) be a run of Nesterov's accelerated gradient descent. Then for every t≥1t\ge1t≥1,

f(yt)−f(x∗)≤α+β2 ∥x1−x∗∥2(1−1κ)t−1.f(y_t)-f(x^*)\le\frac{\alpha+\beta}2\,\|x_1-x^*\|^2\Big(1-\frac1{\sqrt\kappa}\Big)^{t-1}.f(yt​)−f(x∗)≤2α+β​∥x1​−x∗∥2(1−κ​1​)t−1.

This is the rate obtained by combining (3.18) at x=x∗x=x^*x=x∗ with (3.19); the exponential form of Theorem 3.18 follows from 1−u≤e−u1-u\le e^{-u}1−u≤e−u.

Formalization Note The existence of the minimizer x∗x^*x∗ is the book's standing assumption (p. 242). The exponent t−1t-1t−1 is a natural-number subtraction, guarded by t≥1t\ge1t≥1.

Preamble
import Mathlib
import Definitions.Def_OnlineConvexOpt_ConvexBasics_StronglyConvexOn
import Definitions.Def_ConvexOptAlg_NesterovStrong_Defs

open scoped InnerProductSpace
Formal statement
namespace ConvexOptAlg.NesterovStrong

/-- Bubeck, proof of Theorem 3.18, p. 291, the display combining (3.18) and (3.19) (first and
last members): for a `β`-smooth, `α`-strongly convex `f` on `ℝⁿ` with minimizer `x*` and a run
`(x, y)` of Nesterov's accelerated gradient descent, for every `t ≥ 1`,
`f(y_t) − f(x*) ≤ ((α + β)/2)‖x₁ − x*‖² (1 − 1/√κ)^{t−1}`. -/
theorem thm_3_18_rate {n : ℕ} (f : EuclideanSpace ℝ (Fin n) → ℝ)
    (g : EuclideanSpace ℝ (Fin n) → EuclideanSpace ℝ (Fin n)) (α β : ℝ)
    (hα : 0 < α) (hβ : 0 < β)
    (hsc : OnlineConvexOpt.ConvexBasics.StronglyConvexOn Set.univ f g α)
    (hsm : IsBetaSmooth f g β)
    (xstar : EuclideanSpace ℝ (Fin n)) (hmin : ∀ z, f xstar ≤ f z)
    (x y : ℕ → EuclideanSpace ℝ (Fin n)) (hrun : IsNesterovSCRun g α β x y)
    (t : ℕ) (ht : 1 ≤ t) :
    f (y t) - f xstar ≤
      (α + β) / 2 * ‖x 1 - xstar‖ ^ 2 * (1 - 1 / Real.sqrt (kappa α β)) ^ (t - 1) := by sorry

end ConvexOptAlg.NesterovStrong
Source
Bubeck, arXiv:1405.4980v2, proof of Theorem 3.18, p. 291, display after (3.19) (first and last members)

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