Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

All-history FTRL bound of GIA-FTRL against a comparison law (G.218)–(G.241)

Proved
IAFTRL.lemma_G_2_ftrl

by Canzhe Zhao · Nov 17, 2026 · Mathlib 0df444a (Lean v4.33.1)

bandit-algorithmslinear-banditsonline-learning

Setting. As in Proposition G.2: a GIA-FTRL run of T≥1T\ge1T≥1 rounds on a compact generator B\mathcal BB of A⊆Bd\mathcal A\subseteq\mathbb B^dA⊆Bd, against an adaptive environment. Write βt=16d+t\beta_t=16d+\sqrt tβt​=16d+t​ and let μ\muμ be the reference measure (3.3).

Claim. Let u∈K=Conv(B)u\in\mathcal K=\mathrm{Conv}(\mathcal B)u∈K=Conv(B). Let qqq be a probability measure supported on K\mathcal KK with q≪μq\ll\muq≪μ and integrable log-density. Let ε≥0\varepsilon\ge0ε≥0 satisfy ∣(∫x q(dx)−u)⊤ϑ∣≤ε|(\int x\,q(dx)-u)^\top\vartheta|\le\varepsilon∣(∫xq(dx)−u)⊤ϑ∣≤ε for every ∥ϑ∥A≤1\|\vartheta\|_{\mathcal A}\le1∥ϑ∥A​≤1. Then

RegT(u)≤βT DKL(q,μ)+∑t=1T10dβt+Tε.\mathrm{Reg}_T(u)\le\beta_T\,D_{\mathrm{KL}}(q,\mu)+\sum_{t=1}^T\frac{10d}{\beta_t}+T\varepsilon .RegT​(u)≤βT​DKL​(q,μ)+t=1∑T​βt​10d​+Tε.

Proof outline ((G.218)–(G.241)).

  1. Telescope the potential Fβ(S)=−βlog⁡∫e−x⊤S/βμ(dx)F_\beta(S)=-\beta\log\int e^{-x^\top S/\beta}\mu(dx)Fβ​(S)=−βlog∫e−x⊤S/βμ(dx). It is nondecreasing in β\betaβ and satisfies Fβ(S)≤⟨q,S⟩+βDKL(q,μ)F_\beta(S)\le\langle q,S\rangle+\beta D_{\mathrm{KL}}(q,\mu)Fβ​(S)≤⟨q,S⟩+βDKL​(q,μ).
  2. Bound the per-round term pathwise: Dt≤βt∫χ(ht) dqtD_t\le\beta_t\int\chi(h_t)\,dq_tDt​≤βt​∫χ(ht​)dqt​.
  3. Condition on the latent action: Et−1[Dt]≤Ct(pt)≤2d/βt\mathbb E_{t-1}[D_t]\le C_t(p_t)\le2d/\beta_tEt−1​[Dt​]≤Ct​(pt​)≤2d/βt​.
  4. Use marginal domination: ⟨pt−qt,x⊤θt⟩≤2Γt=8d/βt\langle p_t-q_t,x^\top\theta_t\rangle\le2\Gamma_t=8d/\beta_t⟨pt​−qt​,x⊤θt​⟩≤2Γt​=8d/βt​.
  5. Use unbiased difference estimates.
Preamble
import Mathlib
import Definitions.Def_IAFTRL_LinEnv
import Definitions.Def_IAFTRL_GIAFTRL
Formal statement
namespace IAFTRL

open MeasureTheory

/-- **All-history FTRL bound against a comparison law (Proposition G.2, (G.218)–(G.241)).**
Zhao–Tan, *Best-of-Both-Worlds FTRL for Linear Bandits* (STOC submission version), proof of
Proposition G.2, pp.145–147: for a deterministic comparison law `q`,
`Reg_T(u) ≤ β_T D_KL(q, µ) + Σ_t E[D_t] + Σ_t 8d/β_t + Σ_t E[⟨q − δ_u, xᵀθ_t⟩]` (G.241), with
`E_{t−1}[D_t] ≤ C_t(p_t) ≤ 2d/β_t` (G.232).

For every execution of GIA-FTRL for `T ≥ 1` rounds, every comparator `u ∈ 𝒦`, every probability
measure `q` on `𝒦` with `q ≪ µ` and integrable log-density, and every `ε` with
`|(∫ x q(dx) − u)ᵀϑ| ≤ ε` whenever `‖ϑ‖_𝒜 ≤ 1`:
`Reg_T(u) ≤ β_T D_KL(q, µ) + Σ_{t=1}^T 10d/β_t + Tε`. -/
theorem lemma_G_2_ftrl {d : ℕ} (hd : 1 ≤ d) (𝒜 ℬ : Set (Fin d → ℝ)) (e : ℕ → Fin d → ℚ)
    (h𝒜 : IsCompact 𝒜) (hℬ : IsCompact ℬ) (hgen : IsGenerator 𝒜 ℬ)
    (hball : ∀ a ∈ 𝒜, a ⬝ᵥ a ≤ 1) (hspan : Submodule.span ℝ 𝒜 = ⊤) (hk : 1 ≤ kdim ℬ)
    (he : Function.Surjective e) {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω)
    [IsProbabilityMeasure P] (T : ℕ) (R : GRun d Ω) (hT : 1 ≤ T) (hR : IsGIAFTRLRun P 𝒜 ℬ e T R)
    (u : Fin d → ℝ) (hu : u ∈ Kset ℬ) (q : Measure (Fin d → ℝ)) [IsProbabilityMeasure q]
    (hqK : q (Kset ℬ)ᶜ = 0) (hqμ : q ≪ refMu ℬ e)
    (hKL : Integrable (fun x => Real.log (q.rnDeriv (refMu ℬ e) x).toReal) q) (ε : ℝ)
    (hε : ∀ ϑ : Fin d → ℝ, setNorm 𝒜 ϑ ≤ 1 → |((fun i => ∫ x, x i ∂q) - u) ⬝ᵥ ϑ| ≤ ε) :
    gRegretVs P T R u ≤ gbeta d T * ∫ x, Real.log (q.rnDeriv (refMu ℬ e) x).toReal ∂q +
      ∑ t ∈ Finset.range T, 10 * (d : ℝ) / gbeta d (t + 1) + T * ε := by sorry

end IAFTRL
Source
Zhao–Tan, Best-of-Both-Worlds FTRL for Linear Bandits: The Interplay of Information Acquisition, Corruption, and Geometry (STOC submission version; arXiv:2611.01299), Appendix G.6.3, proof of Proposition G.2, pp. 145–147

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