Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma A.2, pp. 33–34 — concentration of the epoch averages v̄ᵢ,ℓ of Algorithm 1 (6/(Nℓ), 4/(Nℓ), 3/(Nℓ))

Proved
MNLBandit.UCB.lemma_A_2

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

concentration-inequalitymnl-banditp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1upper-confidence-bound

Consider the MNL-Bandit with NNN products, parameters 0≤vi≤v0=10\le v_i\le v_0=10≤vi​≤v0​=1, revenues ri∈[0,1]r_i\in[0,1]ri​∈[0,1], a feasible family S\mathcal SS given by totally unimodular constraints (2.3) and closed under subsets, and Algorithm 1 run with any argmax tie-breaking rule. For an epoch index ℓ≥1\ell\ge1ℓ≥1 and a product iii, write Ti(ℓ)T_i(\ell)Ti​(ℓ) and vˉi,ℓ\bar v_{i,\ell}vˉi,ℓ​ for the number of the first ℓ\ellℓ epochs that offered iii and the average of their purchase counts of iii, and λℓ=log⁡(Nℓ+1)\lambda_\ell=\log(\sqrt N\ell+1)λℓ​=log(N​ℓ+1). Then, for every horizon TTT, with probabilities over the first TTT customers and restricted to histories in which epoch ℓ\ellℓ is completed and Ti(ℓ)≥1T_i(\ell)\ge1Ti​(ℓ)≥1,

Pπ(∣vˉi,ℓ−vi∣>48vˉi,ℓλℓTi(ℓ)+48λℓTi(ℓ))≤6Nℓ,Pπ(∣vˉi,ℓ−vi∣>24viλℓTi(ℓ)+48λℓTi(ℓ))≤4Nℓ,Pπ(vˉi,ℓ>3vi2+48λℓTi(ℓ))≤3Nℓ.\begin{aligned} &\mathbb P_\pi\Big(|\bar v_{i,\ell}-v_i|>\sqrt{48\bar v_{i,\ell}\frac{\lambda_\ell}{T_i(\ell)}}+\frac{48\lambda_\ell}{T_i(\ell)}\Big)\le\frac{6}{N\ell},\\ &\mathbb P_\pi\Big(|\bar v_{i,\ell}-v_i|>\sqrt{24v_i\frac{\lambda_\ell}{T_i(\ell)}}+\frac{48\lambda_\ell}{T_i(\ell)}\Big)\le\frac{4}{N\ell},\\ &\mathbb P_\pi\Big(\bar v_{i,\ell}>\frac{3v_i}{2}+\frac{48\lambda_\ell}{T_i(\ell)}\Big)\le\frac{3}{N\ell}. \end{aligned}​Pπ​(∣vˉi,ℓ​−vi​∣>48vˉi,ℓ​Ti​(ℓ)λℓ​​​+Ti​(ℓ)48λℓ​​)≤Nℓ6​,Pπ​(∣vˉi,ℓ​−vi​∣>24vi​Ti​(ℓ)λℓ​​​+Ti​(ℓ)48λℓ​​)≤Nℓ4​,Pπ​(vˉi,ℓ​>23vi​​+Ti​(ℓ)48λℓ​​)≤Nℓ3​.​

These bounds are the probabilistic core of the upper confidence bounds (3.4): they give Lemma 4.1 directly.

Formalization Note.

  1. The paper prints 48log⁡(ℓ+1)/Ti(ℓ)48\log(\ell+1)/T_i(\ell)48log(ℓ+1)/Ti​(ℓ) as the second term of the first bound; its proof (Corollary D.1 and (D.22)) and every use ((A.4)–(A.5)) have 48log⁡(Nℓ+1)/Ti(ℓ)48\log(\sqrt N\ell+1)/T_i(\ell)48log(N​ℓ+1)/Ti​(ℓ), which is stated here.
  2. vˉi,ℓ\bar v_{i,\ell}vˉi,ℓ​ is undefined when Ti(ℓ)=0T_i(\ell)=0Ti​(ℓ)=0; the events require Ti(ℓ)≥1T_i(\ell)\ge1Ti​(ℓ)≥1.
  3. The paper's statement is for the infinite-horizon process ("for every epoch ℓ\ellℓ"). Here it is stated for every finite horizon TTT, on the event that epoch ℓ\ellℓ has been completed by customer TTT; uniformity in TTT is the infinite-horizon statement.
  4. The standing assumptions of §4 (TU and down-closed S\mathcal SS, ri∈[0,1]r_i\in[0,1]ri​∈[0,1]) are carried as hypotheses.
Preamble
import Mathlib
import Definitions.Def_ChoiceCDLP_MNL_mnlObjective
import Definitions.Def_MNLBandit_UCB_Setting
import Definitions.Def_MNLBandit_UCB_Algorithm1
Formal statement
namespace MNLBandit.UCB

open ChoiceCDLP.MNL

theorem lemma_A_2 {N : ℕ} (v r : Fin N → ℝ)
    (𝒮 : Finset (Finset (Fin N))) (sel : (Finset (Fin N) → ℝ) → Finset (Fin N))
    (hTU : IsTU 𝒮) (hdc : DownClosed 𝒮) (hne : 𝒮.Nonempty) (hsel : IsArgmaxSel 𝒮 sel)
    (hv0 : ∀ i, 0 ≤ v i) (hv1 : ∀ i, v i ≤ 1) (hr : ∀ i, r i ∈ Set.Icc (0 : ℝ) 1)
    (T ℓ : ℕ) (hℓ : 1 ≤ ℓ) (i : Fin N) :
    probT v (alg1 r sel) T (fun h =>
        ℓ ≤ epochsDone r sel h ∧ 1 ≤ Tcount r sel h i ℓ ∧
        |vbar r sel h i ℓ - v i| >
          Real.sqrt (48 * vbar r sel h i ℓ * Real.log (Real.sqrt N * ℓ + 1)
              / Tcount r sel h i ℓ)
            + 48 * Real.log (Real.sqrt N * ℓ + 1) / Tcount r sel h i ℓ)
        ≤ 6 / (N * ℓ) ∧
    probT v (alg1 r sel) T (fun h =>
        ℓ ≤ epochsDone r sel h ∧ 1 ≤ Tcount r sel h i ℓ ∧
        |vbar r sel h i ℓ - v i| >
          Real.sqrt (24 * v i * Real.log (Real.sqrt N * ℓ + 1) / Tcount r sel h i ℓ)
            + 48 * Real.log (Real.sqrt N * ℓ + 1) / Tcount r sel h i ℓ)
        ≤ 4 / (N * ℓ) ∧
    probT v (alg1 r sel) T (fun h =>
        ℓ ≤ epochsDone r sel h ∧ 1 ≤ Tcount r sel h i ℓ ∧
        vbar r sel h i ℓ >
          3 * v i / 2 + 48 * Real.log (Real.sqrt N * ℓ + 1) / Tcount r sel h i ℓ)
        ≤ 3 / (N * ℓ) := by sorry

end MNLBandit.UCB
Source
Agrawal, Avadhanula, Goyal, Zeevi, MNL-Bandit: A Dynamic Learning Approach to Assortment Selection, arXiv:1706.03880v2, pp. 33–34, Lemma A.2 (proof p. 56)
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