Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 7, p. 5 — OFTRL with Mᵗ = uᵗ⁻¹ has the RVU property with α = R/η, β = η, γ = 1/(4η)

Proved
RegLearnGames.TotalRegret.proposition_7

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

online-learningoptimistic-ftrlp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1regretrvu-property

Let Δ⊂Rd\Delta \subset \mathbb R^dΔ⊂Rd be the probability simplex, R\mathcal RR a regularizer that is 1-strongly convex with respect to ∥⋅∥1\|\cdot\|_1∥⋅∥1​ on Δ\DeltaΔ, η>0\eta > 0η>0, and RRR a bound on the range of R\mathcal RR on Δ\DeltaΔ. Let u0,u1,u2,…u^0, u^1, u^2, \dotsu0,u1,u2,… be an arbitrary utility sequence in Rd\mathbb R^dRd, and let w0,w1,…w^0, w^1, \dotsw0,w1,… be any run of optimistic FTRL with the one-step recency bias Mt=ut−1M^t = u^{t-1}Mt=ut−1. Then for every horizon TTT and every comparator w∗∈Δw^* \in \Deltaw∗∈Δ,

∑t=1T⟨w∗−wt,ut⟩≤Rη+η∑t=1T∥ut−ut−1∥∞2−14η∑t=1T∥wt−wt−1∥12.\sum_{t=1}^T \langle w^* - w^t, u^t\rangle \le \frac{R}{\eta} + \eta\sum_{t=1}^T \|u^t - u^{t-1}\|_\infty^2 - \frac{1}{4\eta}\sum_{t=1}^T \|w^t - w^{t-1}\|_1^2.t=1∑T​⟨w∗−wt,ut⟩≤ηR​+ηt=1∑T​∥ut−ut−1∥∞2​−4η1​t=1∑T​∥wt−wt−1∥12​.

This is inequality (1) of Definition 3 — regret bounded by variation in utilities (RVU) — with constants α=R/η\alpha = R/\etaα=R/η, β=η\beta = \etaβ=η and γ=1/(4η)\gamma = 1/(4\eta)γ=1/(4η). Together with Theorem 4 it gives Corollary 8.

Formalization Note "Satisfies the RVU property" is rendered as inequality (1) for every utility sequence, every horizon and every comparator. Definition 3 also asks α>0\alpha > 0α>0 and 0<β≤γ0 < \beta \le \gamma0<β≤γ of the constants; these hold when R>0R > 0R>0 and η≤1/2\eta \le 1/2η≤1/2 and are not part of the conclusion. The norm is ℓ1\ell_1ℓ1​ with dual ℓ∞\ell_\inftyℓ∞​ (the paper allows any norm). RRR is any upper bound on the range of R\mathcal RR over Δ\DeltaΔ, which includes the paper's sup⁡−inf⁡\sup - \infsup−inf.

Preamble
import Mathlib
import Definitions.Def_RegLearnGames_TotalRegret_Setting
Formal statement
namespace RegLearnGames.TotalRegret

theorem proposition_7 {d : ℕ} (𝓡 : (Fin d → ℝ) → ℝ) (η R : ℝ) (hη : 0 < η)
    (h𝓡 : IsOneStronglyConvexL1 𝓡)
    (hR : ∀ f ∈ stdSimplex ℝ (Fin d), ∀ g ∈ stdSimplex ℝ (Fin d), 𝓡 f - 𝓡 g ≤ R)
    (u w : ℕ → Fin d → ℝ) (hw : IsOFTRL 𝓡 η (fun T => u (T - 1)) u w)
    (T : ℕ) (wstar : Fin d → ℝ) (hwstar : wstar ∈ stdSimplex ℝ (Fin d)) :
    ∑ t ∈ Finset.Icc 1 T, (wstar - w t) ⬝ᵥ u t ≤
      R / η + η * ∑ t ∈ Finset.Icc 1 T, ‖u t - u (t - 1)‖ ^ 2
        - 1 / (4 * η) * ∑ t ∈ Finset.Icc 1 T, l1 (w t - w (t - 1)) ^ 2 := by sorry

end RegLearnGames.TotalRegret
Source
Syrgkanis, Agarwal, Luo, Schapire, Fast Convergence of Regularized Learning in Games, arXiv:1507.00407v5, p. 5, Proposition 7 (R as defined just above it); Definition 3, inequality (1), p. 3; proof supp. pp. 2–5 (PDF pp. 11–14)
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