Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma A.8, p. 71 — D²_H(Rad(μ₁), Rad(μ₂)) ≤ ((μ₁−μ₂)²/2)(1/(2+μ₁+μ₂) + 1/(2−μ₁−μ₂)); ≤ (μ₁−μ₂)² on [−½, ½]; D²_H(Rad(μ), Rad(0)) ≤ ¾μ²

Proved
StatComplexityDM.LinearLB.rad_hellinger_bounds

by mikedeng1 · Oct 8, 2026 · Mathlib 0df444a (Lean v4.33.1)

hellinger-distanceinformation-theoryp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-bookp2o-v1rademacher-distribution

Let Rad(μ)\mathrm{Rad}(\mu)Rad(μ) be the Rademacher distribution on {−1,+1}\{-1,+1\}{−1,+1} with mean μ∈[−1,1]\mu \in [-1,1]μ∈[−1,1], i.e. Rad(μ)(+1)=(1+μ)/2\mathrm{Rad}(\mu)(+1) = (1+\mu)/2Rad(μ)(+1)=(1+μ)/2 and Rad(μ)(−1)=(1−μ)/2\mathrm{Rad}(\mu)(-1) = (1-\mu)/2Rad(μ)(−1)=(1−μ)/2, and let DH2(P,Q)=∑y(P(y)−Q(y))2D^2_{\mathrm{H}}(P,Q) = \sum_y (\sqrt{P(y)} - \sqrt{Q(y)})^2DH2​(P,Q)=∑y​(P(y)​−Q(y)​)2 be the squared Hellinger distance. Then:

  1. For all μ1,μ2∈[−1,+1]\mu_1, \mu_2 \in [-1, +1]μ1​,μ2​∈[−1,+1],
DH2(Rad(μ1),Rad(μ2))≤(μ1−μ2)22(12+μ1+μ2+12−μ1−μ2).D^2_{\mathrm{H}}\bigl(\mathrm{Rad}(\mu_1), \mathrm{Rad}(\mu_2)\bigr) \le \frac{(\mu_1 - \mu_2)^2}{2} \left( \frac{1}{2 + \mu_1 + \mu_2} + \frac{1}{2 - \mu_1 - \mu_2} \right).DH2​(Rad(μ1​),Rad(μ2​))≤2(μ1​−μ2​)2​(2+μ1​+μ2​1​+2−μ1​−μ2​1​).
  1. For all μ1,μ2∈[−12,+12]\mu_1, \mu_2 \in [-\tfrac12, +\tfrac12]μ1​,μ2​∈[−21​,+21​],  DH2(Rad(μ1),Rad(μ2))≤(μ1−μ2)2\ D^2_{\mathrm{H}}\bigl(\mathrm{Rad}(\mu_1), \mathrm{Rad}(\mu_2)\bigr) \le (\mu_1 - \mu_2)^2 DH2​(Rad(μ1​),Rad(μ2​))≤(μ1​−μ2​)2.
  2. For all μ∈[−1,+1]\mu \in [-1, +1]μ∈[−1,+1],  DH2(Rad(μ),Rad(0))≤34μ2\ D^2_{\mathrm{H}}\bigl(\mathrm{Rad}(\mu), \mathrm{Rad}(0)\bigr) \le \tfrac34 \mu^2 DH2​(Rad(μ),Rad(0))≤43​μ2.

The third bound is the information property of the hard family in the linear bandit lower bound (Proposition 6.2): a model whose mean differs from that of Rad(0)\mathrm{Rad}(0)Rad(0) by μ\muμ is at squared Hellinger distance O(μ2)O(\mu^2)O(μ2) from it.

Formalization Note In the first bound, a denominator 2±(μ1+μ2)2 \pm (\mu_1 + \mu_2)2±(μ1​+μ2​) vanishes only when μ1=μ2=∓1\mu_1 = \mu_2 = \mp 1μ1​=μ2​=∓1; there both sides are 000 (Lean's convention 1/0=01/0 = 01/0=0 gives right-hand side 0⋅14=00 \cdot \tfrac14 = 00⋅41​=0), so the literal division form is stated. Outcomes are encoded on Bool.

Preamble
import Mathlib
import Definitions.Def_FoundationsRL_GeneralDM_Divergences
import Definitions.Def_StatComplexityDM_LinearLB_Setting
Formal statement
namespace StatComplexityDM.LinearLB

open FoundationsRL.GeneralDM

/-- Lemma A.8 (arXiv:2112.13487v3, p. 71): squared Hellinger distance between Rademacher
distributions, all three displays. -/
theorem rad_hellinger_bounds :
    (∀ μ₁ μ₂ : ℝ, μ₁ ∈ Set.Icc (-1 : ℝ) 1 → μ₂ ∈ Set.Icc (-1 : ℝ) 1 →
      hellingerSq (rad μ₁) (rad μ₂) ≤
        (μ₁ - μ₂) ^ 2 / 2 * (1 / (2 + μ₁ + μ₂) + 1 / (2 - μ₁ - μ₂))) ∧
    (∀ μ₁ μ₂ : ℝ, μ₁ ∈ Set.Icc (-1 / 2 : ℝ) (1 / 2) → μ₂ ∈ Set.Icc (-1 / 2 : ℝ) (1 / 2) →
      hellingerSq (rad μ₁) (rad μ₂) ≤ (μ₁ - μ₂) ^ 2) ∧
    (∀ μ : ℝ, μ ∈ Set.Icc (-1 : ℝ) 1 →
      hellingerSq (rad μ) (rad 0) ≤ 3 / 4 * μ ^ 2) := by sorry

end StatComplexityDM.LinearLB
Source
arXiv:2112.13487v3, Lemma A.8, p. 71
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