Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

(31) — the descent inequality for an L_φ-smooth objective along θ − αg

Proved
CertDRO.SGD.eq_31

by mikedeng1 · Oct 8, 2026 · Mathlib 0df444a (Lean v4.33.1)

descent-lemmap2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1smoothness

Let F:Rd→RF:\mathbb R^d\to\mathbb RF:Rd→R be differentiable with LφL_\varphiLφ​-Lipschitz gradient, ∥∇F(θ)−∇F(θ′)∥2≤Lφ∥θ−θ′∥2\|\nabla F(\theta)-\nabla F(\theta')\|_2\le L_\varphi\|\theta-\theta'\|_2∥∇F(θ)−∇F(θ′)∥2​≤Lφ​∥θ−θ′∥2​. Then for every θ,g∈Rd\theta, g\in\mathbb R^dθ,g∈Rd and every α∈R\alpha\in\mathbb Rα∈R,

F(θ−αg)≤F(θ)−α(1−12Lφα)∥∇F(θ)∥22+α(1−Lφα)⟨∇F(θ),∇F(θ)−g⟩+Lφα22∥g−∇F(θ)∥22.F(\theta-\alpha g)\le F(\theta)-\alpha\Big(1-\tfrac12L_\varphi\alpha\Big)\|\nabla F(\theta)\|_2^2+\alpha(1-L_\varphi\alpha)\langle\nabla F(\theta),\nabla F(\theta)-g\rangle+\frac{L_\varphi\alpha^2}{2}\|g-\nabla F(\theta)\|_2^2 .F(θ−αg)≤F(θ)−α(1−21​Lφ​α)∥∇F(θ)∥22​+α(1−Lφ​α)⟨∇F(θ),∇F(θ)−g⟩+2Lφ​α2​∥g−∇F(θ)∥22​.

In the proof of Theorem 2 this is applied with θ=θt\theta=\theta^tθ=θt, g=gt=∇θℓ(θt;z^t)g=g^t=\nabla_\theta\ell(\theta^t;\hat z^t)g=gt=∇θ​ℓ(θt;z^t) and α=αt\alpha=\alpha_tα=αt​, so that the left side is F(θt+1)F(\theta^{t+1})F(θt+1); it is the progress guarantee of one step of Algorithm 1.

Formalization Note The statement is deterministic and holds for any stepsize; the paper's displayed chain of equalities after the first inequality is algebra.

Preamble
import Definitions.Def_CertDRO_SGD_model
Formal statement
namespace CertDRO.SGD

/-- (31), §B.3, p. 40: for a differentiable `F` on `ℝ^d` with `L_φ`-Lipschitz gradient, the step
`θ⁺ = θ − α g` satisfies
`F(θ⁺) ≤ F(θ) − α(1 − ½ L_φ α)‖∇F(θ)‖² + α(1 − L_φ α)⟨∇F(θ), ∇F(θ) − g⟩ + (L_φ α²/2)‖g − ∇F(θ)‖²`. -/
theorem eq_31 {d : ℕ} (F : Param d → ℝ) (Lφ : ℝ) (hF : Differentiable ℝ F)
    (hL : ∀ θ θ', ‖gradient F θ - gradient F θ'‖ ≤ Lφ * ‖θ - θ'‖)
    (θ g : Param d) (α : ℝ) :
    F (θ - α • g) ≤ F θ - α * (1 - 1 / 2 * Lφ * α) * ‖gradient F θ‖ ^ 2
      + α * (1 - Lφ * α) * inner ℝ (gradient F θ) (gradient F θ - g)
      + Lφ * α ^ 2 / 2 * ‖g - gradient F θ‖ ^ 2 := by sorry

end CertDRO.SGD
Source
Sinha, Namkoong, Volpi, Duchi, Certifying Some Distributional Robustness with Principled Adversarial Training, arXiv:1710.10571v5, p. 40, §B.3, (31)
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