Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 16 — one step of composite mirror descent (diagonal quadratic proximal functions)

Proved
AdaGrad.Diag.lemma_16

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

mirror-descentonline-learningp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1

Let η>0\eta>0η>0, let X⊆Rd\mathcal X\subseteq\mathbb R^dX⊆Rd be convex, and let φ\varphiφ and ftf_tft​ be convex functions Rd→R\mathbb R^d\to\mathbb RRd→R. Let h∈Rdh\in\mathbb R^dh∈Rd have nonnegative entries and write ψt(y)=12∑ihiyi2\psi_t(y)=\frac12\sum_i h_iy_i^2ψt​(y)=21​∑i​hi​yi2​, so that Bψt(y,z)=12∑ihi(yi−zi)2B_{\psi_t}(y,z)=\frac12\sum_ih_i(y_i-z_i)^2Bψt​​(y,z)=21​∑i​hi​(yi​−zi​)2 and ∥g∥ψt∗2=∑igi2/hi\|g\|^2_{\psi_t^*}=\sum_i g_i^2/h_i∥g∥ψt∗​2​=∑i​gi2​/hi​. Let gtg_tgt​ be a subgradient of ftf_tft​ at xtx_txt​ with gt,i=0g_{t,i}=0gt,i​=0 whenever hi=0h_i=0hi​=0, and let xt+1∈Xx_{t+1}\in\mathcal Xxt+1​∈X minimize η⟨gt,y⟩+ηφ(y)+Bψt(y,xt)\eta\langle g_t,y\rangle+\eta\varphi(y)+B_{\psi_t}(y,x_t)η⟨gt​,y⟩+ηφ(y)+Bψt​​(y,xt​) over y∈Xy\in\mathcal Xy∈X (one step of the composite mirror descent update (4)). Then for every x∗∈Xx^*\in\mathcal Xx∗∈X,

η(ft(xt)−ft(x∗))+η(φ(xt+1)−φ(x∗))≤Bψt(x∗,xt)−Bψt(x∗,xt+1)+η22∥gt∥ψt∗2.\eta\big(f_t(x_t)-f_t(x^*)\big)+\eta\big(\varphi(x_{t+1})-\varphi(x^*)\big)\le B_{\psi_t}(x^*,x_t)-B_{\psi_t}(x^*,x_{t+1})+\frac{\eta^2}{2}\|g_t\|^2_{\psi_t^*}.η(ft​(xt​)−ft​(x∗))+η(φ(xt+1​)−φ(x∗))≤Bψt​​(x∗,xt​)−Bψt​​(x∗,xt+1​)+2η2​∥gt​∥ψt∗​2​.

This is the one-step inequality that, summed over rounds, gives the regret bound of composite mirror descent with time-varying proximal functions (Proposition 3).

Formalization Note Specialised to diagonal quadratic proximal functions ψt(y)=12∑iht,iyi2\psi_t(y)=\frac12\sum_ih_{t,i}y_i^2ψt​(y)=21​∑i​ht,i​yi2​; the paper's strong convexity of BψtB_{\psi_t}Bψt​​ with respect to ∥⋅∥ψt\|\cdot\|_{\psi_t}∥⋅∥ψt​​ is implied by the quadratic form. The comparator is required to lie in X\mathcal XX: the page says "for any x∗x^*x∗", but its proof applies the optimality condition of xt+1x_{t+1}xt+1​, valid for points of X\mathcal XX, at x=x∗x=x^*x=x∗. The condition "hi=0⇒gt,i=0h_i=0\Rightarrow g_{t,i}=0hi​=0⇒gt,i​=0" makes Lean's a/0=0a/0=0a/0=0 agree with the paper's dual norm, which is +∞+\infty+∞ otherwise.

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

open ShorNonsmooth.AlmostDiff

/-- Duchi, Hazan, Singer, JMLR 12 (2011), Lemma 16, p. 2154, specialised to diagonal quadratic
proximal functions `ψ_t(y) = ½ ∑_i h_i y_i²` with weights `h_i ≥ 0`: one step of the composite
mirror descent update (4). Let `g_t` be a subgradient of the convex `f_t` at `x_t`, and let
`x_{t+1} ∈ X` minimize `η⟨g_t, y⟩ + ηϕ(y) + B_{ψ_t}(y, x_t)` over the convex set `X`, where `ϕ` is
convex. If `h_i = 0` forces `g_{t,i} = 0`, then for every `x* ∈ X`,
`η(f_t(x_t) − f_t(x*)) + η(ϕ(x_{t+1}) − ϕ(x*))
   ≤ B_{ψ_t}(x*, x_t) − B_{ψ_t}(x*, x_{t+1}) + (η²/2) ‖g_t‖²_{ψ_t*}`,
with `‖g‖²_{ψ_t*} = ∑_i g_i² / h_i`. -/
theorem lemma_16 {d : ℕ} (η : ℝ) (hη : 0 < η) (X : Set (EuclideanSpace ℝ (Fin d)))
    (hXc : Convex ℝ X) (ϕ : EuclideanSpace ℝ (Fin d) → ℝ) (hϕ : ConvexOn ℝ Set.univ ϕ)
    (ft : EuclideanSpace ℝ (Fin d) → ℝ) (hft : ConvexOn ℝ Set.univ ft)
    (h : Fin d → ℝ) (hh : ∀ i, 0 ≤ h i)
    (xt xnext gt : EuclideanSpace ℝ (Fin d))
    (hhg : ∀ i, h i = 0 → gt i = 0)
    (hsub : IsSubgradient ft xt gt) (hmem : xnext ∈ X)
    (hmin : IsMinOn (fun y => η * inner ℝ gt y + η * ϕ y + bregman h y xt) X xnext)
    (xstar : EuclideanSpace ℝ (Fin d)) (hxstar : xstar ∈ X) :
    η * (ft xt - ft xstar) + η * (ϕ xnext - ϕ xstar)
      ≤ bregman h xstar xt - bregman h xstar xnext + η ^ 2 / 2 * dualNormSq h gt := by sorry

end AdaGrad.Diag
Source
Duchi, Hazan, Singer, Adaptive Subgradient Methods for Online Learning and Stochastic Optimization, JMLR 12 (2011), p. 2154, Lemma 16
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