Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Sequential Halving optimized rank-tail average bound

Disproved
BanditAlgorithm.sequential_halving_rank_tail_average_bound_clog

by Harry_Xu · Aug 2, 2026 · Mathlib c5ea003 (Lean v4.30.0)

banditsconcentrationpure-explorationsequential-halving

Let L=⌈log⁡2k⌉L=\lceil\log_2 k\rceilL=⌈log2​k⌉, let mℓm_\ellmℓ​ be the deterministic number of active arms in phase ℓ\ellℓ, let qℓ=mℓ+1q_\ell=m_{\ell+1}qℓ​=mℓ+1​, and let Tℓ=⌊n/(Lmℓ)⌋T_\ell=\lfloor n/(L m_\ell)\rfloorTℓ​=⌊n/(Lmℓ​)⌋. Suppose positive gaps satisfy (i+1)/Δi2≤H2(i+1)/\Delta_i^2\le H_2(i+1)/Δi2​≤H2​. Then there is a rank cutoff t<qℓt<q_\ellt<qℓ​ such that the average of the one-arm tails above ttt obeys

∑i: i≥t, Δi>0exp⁡(−TℓΔi2/4)qℓ−t≤3exp⁡ ⁣(−n16H2L).\frac{\sum_{i:\,i\ge t,\,\Delta_i>0}\exp(-T_\ell\Delta_i^2/4)}{q_\ell-t}\le 3\exp\!\left(-\frac{n}{16H_2L}\right).qℓ​−t∑i:i≥t,Δi​>0​exp(−Tℓ​Δi2​/4)​≤3exp(−16H2​Ln​).

The cutoff is optimized rather than fixed at one quarter; this is what absorbs the natural-number floor in TℓT_\ellTℓ​ while retaining the book constant 161616.

Preamble
import Definitions.Def_SequentialHalving

open MeasureTheory
Formal statement
theorem BanditAlgorithm.sequential_halving_rank_tail_average_bound_clog
    {k n : ℕ} (gap : Fin k → ℝ)
    (hn : k * Nat.clog 2 k ≤ n) (H₂ : ℝ)
    (hH₂ : ∀ i : Fin k, 0 < gap i →
      ((i : ℕ) + 1 : ℝ) / gap i ^ 2 ≤ H₂)
    (ℓ : ℕ) (hℓ : ℓ < Nat.clog 2 k) :
    ∃ threshold : ℕ, threshold < BanditAlgorithm.seqHalvingCount k (ℓ + 1) ∧
      (∑ i : Fin k,
        if threshold ≤ (i : ℕ) ∧ 0 < gap i then
          Real.exp (-(BanditAlgorithm.seqHalvingPulls k n ℓ : ℝ) * gap i ^ 2 / 4)
        else 0) /
          ((BanditAlgorithm.seqHalvingCount k (ℓ + 1) - threshold : ℕ) : ℝ) ≤
        3 * Real.exp (-(n / (16 * H₂ * (Nat.clog 2 k : ℝ)))) := by
  sorry
Source
Lattimore and Szepesvari, Bandit Algorithms (2020), Theorem 33.10 p. 412 and Exercise 33.8(c)-(e) pp. 419-420; floor-aware analytic bridge for Algorithm 22 line 4.

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