Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 4.1 — UE regret ≤ a₄r‖z‖ + a₅r√T log^{3/2} T and Risk ≤ a₄r E‖Z‖ + a₅r√T log^{3/2} T

Proved
LinParamBandits.UEGeneral.ue_regret_bound_general

by mikedeng1 · Oct 7, 2026 · Mathlib 0df444a (Lean v4.33.1)

bayes-risklinear-banditp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1regretucb

Theorem 4.1 (Bounds for General Compact Sets of Arms). For all constants σ0,uˉ,λ0>0\sigma_0, \bar u, \lambda_0 > 0σ0​,uˉ,λ0​>0 there are constants a4,a5>0a_4, a_5 > 0a4​,a5​>0, depending only on σ0,uˉ,λ0\sigma_0, \bar u, \lambda_0σ0​,uˉ,λ0​, with the following property. For every dimension r≥2r \ge 2r≥2, every compact nonempty set of arms Ur⊂Rr\mathcal U_r \subset \mathbb R^rUr​⊂Rr, every noise model and arms b1,…,brb_1, \dots, b_rb1​,…,br​ satisfying Assumption 1 with these constants, every run of the Uncertainty Ellipsoid policy, and every horizon T≥r+1T \ge r+1T≥r+1:

  1. for every z∈Rrz \in \mathbb R^rz∈Rr,
Regret(z,T,UE)≤a4 r ∥z∥+a5 rTlog⁡3/2T;\mathrm{Regret}(z, T, \mathrm{UE}) \le a_4\,r\,\|z\| + a_5\,r\sqrt T\log^{3/2}T;Regret(z,T,UE)≤a4​r∥z∥+a5​rT​log3/2T;
  1. for every prior distribution of ZZZ with E∥Z∥<∞\mathbb E\|Z\| < \inftyE∥Z∥<∞, the regret is integrable in zzz and
Risk(T,UE)≤a4 r E[∥Z∥]+a5 rTlog⁡3/2T.\mathrm{Risk}(T, \mathrm{UE}) \le a_4\,r\,\mathbb E\big[\|Z\|\big] + a_5\,r\sqrt T\log^{3/2}T.Risk(T,UE)≤a4​rE[∥Z∥]+a5​rT​log3/2T.

The UE policy thus comes within a factor log⁡3/2T\log^{3/2}Tlog3/2T of the Ω(rT)\Omega(r\sqrt T)Ω(rT​) lower bound of Theorem 2.1, for any compact set of arms, without knowing the horizon TTT.

Formalization Note The quantifier order is the paper's: a4,a5a_4, a_5a4​,a5​ are chosen after σ0,uˉ,λ0\sigma_0, \bar u, \lambda_0σ0​,uˉ,λ0​ and before rrr, the arm set, the noise kernel, the arms bkb_kbk​, the UE run, TTT and zzz. Finite arm sets are included. log⁡3/2T\log^{3/2}Tlog3/2T is the real power Real.log T ^ (3/2 : ℝ), with log⁡T>0\log T > 0logT>0 since T≥3T \ge 3T≥3. The prior μ\muμ is a probability measure on Rr\mathbb R^rRr; E∥Z∥<∞\mathbb E\|Z\| < \inftyE∥Z∥<∞ is stated as integrability of ∥z∥\|z\|∥z∥ (the bound mentions E∥Z∥\mathbb E\|Z\|E∥Z∥), and integrability of z↦Regret(z,T,UE)z \mapsto \mathrm{Regret}(z, T, \mathrm{UE})z↦Regret(z,T,UE) is part of the conclusion, so the risk is never a junk value. The noise is independent of ZZZ, so the risk is the integral over the prior of the regret given Z=zZ = zZ=z.

Preamble
import Mathlib
import Definitions.Def_LinParamBandits_UEGeneral_Model
import Definitions.Def_LinParamBandits_UEGeneral_UEPolicy
Formal statement
namespace LinParamBandits.UEGeneral

open MeasureTheory ProbabilityTheory

/-- Theorem 4.1 (p. 21): bounds for general compact sets of arms. The constants `a₄, a₅`
depend only on `σ₀, ū, λ₀`: they are chosen before `r`, the arm set, the noise, the UE run,
`T` and `z`. -/
theorem ue_regret_bound_general :
    ∀ σ₀ ubar lam0 : ℝ, 0 < σ₀ → 0 < ubar → 0 < lam0 →
    ∃ a₄ a₅ : ℝ, 0 < a₄ ∧ 0 < a₅ ∧
    ∀ (r : ℕ), 2 ≤ r →
    ∀ (𝒰 : Set (LinParamBandits.LowerBound.Vec r)), IsCompact 𝒰 → 𝒰.Nonempty →
    ∀ (ν : Kernel (LinParamBandits.LowerBound.Vec r) ℝ), IsMarkovKernel ν →
    ∀ (b : Fin r → LinParamBandits.LowerBound.Vec r), Assumption1 𝒰 ν b σ₀ ubar lam0 →
    ∀ (ψ : Policy r), IsUE 𝒰 b σ₀ ubar lam0 ψ →
    ∀ (T : ℕ), r + 1 ≤ T →
      (∀ z : LinParamBandits.LowerBound.Vec r,
        regret ν 𝒰 ψ z T ≤
          a₄ * r * ‖z‖ + a₅ * r * Real.sqrt T * Real.log T ^ (3 / 2 : ℝ)) ∧
      (∀ μ : Measure (LinParamBandits.LowerBound.Vec r), IsProbabilityMeasure μ → Integrable (fun z => ‖z‖) μ →
        Integrable (fun z => regret ν 𝒰 ψ z T) μ ∧
        risk ν 𝒰 ψ μ T ≤
          a₄ * r * (∫ z, ‖z‖ ∂μ) + a₅ * r * Real.sqrt T * Real.log T ^ (3 / 2 : ℝ)) := by sorry

end LinParamBandits.UEGeneral
Source
Rusmevichientong, Tsitsiklis, Linearly Parameterized Bandits, arXiv:0812.3465v2, Theorem 4.1, p. 21 (proof App. B.2, p. 39)
Human review
  • Endorsed by Shuze Chen · Oct 7, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 7, 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