Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

(14) — the Bregman drift of diagonal AdaGrad is at most ½ maxₜ‖x* − xₜ‖∞² Σᵢ‖g_{1:T,i}‖₂ − ½‖x* − x₁‖∞²⟨s₁, 1⟩

Proved
AdaGrad.Diag.eq_14

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

adaptive-methodsonline-learningp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1

Let g1,g2,⋯∈Rdg_1,g_2,\dots\in\mathbb R^dg1​,g2​,⋯∈Rd and x1,x2,…,x∗∈Rdx_1,x_2,\dots,x^*\in\mathbb R^dx1​,x2​,…,x∗∈Rd be arbitrary, δ∈R\delta\in\mathbb Rδ∈R, st,i=∥g1:t,i∥2s_{t,i}=\|g_{1:t,i}\|_2st,i​=∥g1:t,i​∥2​, and Bψt(y,z)=12∑i(δ+st,i)(yi−zi)2B_{\psi_t}(y,z)=\frac12\sum_i(\delta+s_{t,i})(y_i-z_i)^2Bψt​​(y,z)=21​∑i​(δ+st,i​)(yi​−zi​)2 the Bregman divergence of AdaGrad's ψt(x)=12⟨x,(δI+diag(st))x⟩\psi_t(x)=\frac12\langle x,(\delta I+\mathrm{diag}(s_t))x\rangleψt​(x)=21​⟨x,(δI+diag(st​))x⟩. For every T≥1T\ge1T≥1,

∑t=1T−1[Bψt+1(x∗,xt+1)−Bψt(x∗,xt+1)]≤12∑t=1T−1∥x∗−xt+1∥∞2⟨st+1−st,1⟩\sum_{t=1}^{T-1}\big[B_{\psi_{t+1}}(x^*,x_{t+1})-B_{\psi_t}(x^*,x_{t+1})\big]\le\frac12\sum_{t=1}^{T-1}\|x^*-x_{t+1}\|_\infty^2\langle s_{t+1}-s_t,\mathbf 1\ranglet=1∑T−1​[Bψt+1​​(x∗,xt+1​)−Bψt​​(x∗,xt+1​)]≤21​t=1∑T−1​∥x∗−xt+1​∥∞2​⟨st+1​−st​,1⟩ ≤12max⁡1≤t≤T∥x∗−xt∥∞2∑i=1d∥g1:T,i∥2−12∥x∗−x1∥∞2⟨s1,1⟩.\le\frac12\max_{1\le t\le T}\|x^*-x_t\|_\infty^2\sum_{i=1}^d\|g_{1:T,i}\|_2-\frac12\|x^*-x_1\|_\infty^2\langle s_1,\mathbf 1\rangle .≤21​1≤t≤Tmax​∥x∗−xt​∥∞2​i=1∑d​∥g1:T,i​∥2​−21​∥x∗−x1​∥∞2​⟨s1​,1⟩.

This bounds the Bregman-divergence drift term of the composite mirror descent regret bound (Proposition 3) for diagonal AdaGrad.

Formalization Note Both inequalities are stated, as a conjunction. δ\deltaδ cancels from the left-hand side and is arbitrary. The maximum is over t∈{1,…,T}t\in\{1,\dots,T\}t∈{1,…,T} (Finset.sup', well defined because T≥1T\ge1T≥1). ⟨st+1−st,1⟩=∑i(st+1,i−st,i)\langle s_{t+1}-s_t,\mathbf 1\rangle=\sum_i(s_{t+1,i}-s_{t,i})⟨st+1​−st​,1⟩=∑i​(st+1,i​−st,i​).

Preamble
import Mathlib
import Definitions.Def_AdaGrad_Diag_Setup
Formal statement
namespace AdaGrad.Diag

/-- Duchi, Hazan, Singer, JMLR 12 (2011), §3, inequality (14), p. 2131: for AdaGrad's diagonal
proximal functions `ψ_t(x) = ½⟨x, (δI + diag(s_t)) x⟩`, any points `x_1, x_2, … , x* ∈ ℝ^d`,
any subgradients `g_t` and any `T ≥ 1`,
`∑_{t=1}^{T−1} [B_{ψ_{t+1}}(x*, x_{t+1}) − B_{ψ_t}(x*, x_{t+1})]
   ≤ ½ ∑_{t=1}^{T−1} ‖x* − x_{t+1}‖∞² ⟨s_{t+1} − s_t, 1⟩
   ≤ ½ max_{t≤T} ‖x* − x_t‖∞² ∑_{i=1}^d ‖g_{1:T,i}‖₂ − ½ ‖x* − x_1‖∞² ⟨s_1, 1⟩`.
The maximum is over `t ∈ {1, …, T}`. -/
theorem eq_14 {d : ℕ} (δ : ℝ) (x g : ℕ → EuclideanSpace ℝ (Fin d))
    (xstar : EuclideanSpace ℝ (Fin d)) (T : ℕ) (hT : 1 ≤ T) :
    ∑ t ∈ Finset.Ico 1 T,
        (bregman (adaWeights δ g (t + 1)) xstar (x (t + 1))
          - bregman (adaWeights δ g t) xstar (x (t + 1)))
      ≤ (1 / 2) * ∑ t ∈ Finset.Ico 1 T,
          supNorm (xstar - x (t + 1)) ^ 2 * ∑ i, (s g (t + 1) i - s g t i) ∧
    (1 / 2) * ∑ t ∈ Finset.Ico 1 T,
          supNorm (xstar - x (t + 1)) ^ 2 * ∑ i, (s g (t + 1) i - s g t i)
      ≤ (1 / 2) * (Finset.Icc 1 T).sup' (Finset.nonempty_Icc.mpr hT)
            (fun t => supNorm (xstar - x t) ^ 2) * ∑ i, s g T i
        - (1 / 2) * supNorm (xstar - x 1) ^ 2 * ∑ i, s g 1 i := by sorry

end AdaGrad.Diag
Source
Duchi, Hazan, Singer, Adaptive Subgradient Methods for Online Learning and Stochastic Optimization, JMLR 12 (2011), p. 2131, §3, (14)
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 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