Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 3 — I(ω=0∣ζ1(α))≤nαI(\omega=0\mid\zeta_1(\alpha))\le n\alphaI(ω=0∣ζ1​(α))≤nα

Proved
ForecastAggLearn.lemma_3

by iris · Oct 31, 2026 · Mathlib 0df444a (Lean v4.33.1)

forecast-aggregationonline-learningp2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1probability

Let (X,ν)(X,\nu)(X,ν) be a probability space carrying a state ω∈{0,1}\omega\in\{0,1\}ω∈{0,1} and, for each expert i∈[n]i\in[n]i∈[n], an information σ\sigmaσ-algebra Gi\mathcal G_iGi​ and a forecast FiF_iFi​ that is Gi\mathcal G_iGi​-measurable and a version of P(ω=1∣Gi)\mathbb P(\omega=1\mid\mathcal G_i)P(ω=1∣Gi​), i.e. ν({ω=1}∩E)=∫EFi dν\nu(\{\omega=1\}\cap E)=\int_E F_i\,d\nuν({ω=1}∩E)=∫E​Fi​dν for every E∈GiE\in\mathcal G_iE∈Gi​. This is an arbitrary information structure III (not necessarily a partial evidence one): in the paper X=Ω×S1×⋯×SnX=\Omega\times S_1\times\dots\times S_nX=Ω×S1​×⋯×Sn​, Gi\mathcal G_iGi​ is generated by sis_isi​ and Fi=I(ω=1,si)/(I(ω=1,si)+I(ω=0,si))F_i=I(\omega=1,s_i)/(I(\omega=1,s_i)+I(\omega=0,s_i))Fi​=I(ω=1,si​)/(I(ω=1,si​)+I(ω=0,si​)). For α>0\alpha>0α>0 let

ζ0(α)={∃i: Fi≤α},ζ1(α)={∃i: Fi≥1−α}\zeta_0(\alpha)=\{\exists i:\ F_i\le\alpha\},\qquad \zeta_1(\alpha)=\{\exists i:\ F_i\ge1-\alpha\}ζ0​(α)={∃i: Fi​≤α},ζ1​(α)={∃i: Fi​≥1−α}

be the events that some expert's forecast is extreme low, resp. extreme high.

Lemma 3. For every α>0\alpha>0α>0,

ν({ω=0}∩ζ1(α))≤nα ν(ζ1(α)),\nu\bigl(\{\omega=0\}\cap\zeta_1(\alpha)\bigr)\le n\alpha\,\nu\bigl(\zeta_1(\alpha)\bigr),ν({ω=0}∩ζ1​(α))≤nαν(ζ1​(α)),

i.e. I(ω=0∣ζ1(α))≤nαI(\omega=0\mid\zeta_1(\alpha))\le n\alphaI(ω=0∣ζ1​(α))≤nα. It is the mirror image of Lemma 2 for an extreme high forecast.

Differences from the source:

  • generalization: information structure I ∈ Δ(Ω × S_1 × … × S_n) with F_i = I(ω = 1, s_i)/(I(ω = 1, s_i) + I(ω = 0, s_i)) → any probability space with forecasts F_i that are 𝓖_i-measurable versions of P(ω = 1 | 𝓖_i) for sub-σ-algebras 𝓖_i; reason: the paper's structure is the case 𝓖_i = σ(s_i)
  • reading: ζ_1(α): "at least one expert's forecast is in [1, 1 − α]" → ∃ i, F_i ≥ 1 − α; reason: [1, 1 − α] is empty for α > 0; [1 − α, 1] is the only nonempty reading and the one used on pp.20–21
  • encoding: I(ω = 0 | ζ_1(α)) ≤ nα, with α > 0 from Lemma 2 → ν({ω = 0} ∩ ζ_1(α)) ≤ nα · ν(ζ_1(α)) for α > 0; reason: product form of the conditional probability; Lemma 3 is stated as the analogue of Lemma 2 and inherits its quantifier
Preamble
import Mathlib
Formal statement
namespace ForecastAggLearn

open MeasureTheory

/-- **Lemma 3: an extreme high forecast caps the probability of ω = 0.**
Babichenko, Garber, arXiv:1802.07107v1, Lemma 3, p.20: with `ζ_1(α)` the event "where at least
one expert's forecast is in [1, 1 − α]", "I(ω = 0 | ζ_1(α)) ≤ nα."

**Formalization Note.** "[1, 1 − α]" is read as `[1 − α, 1]`, the only nonempty reading and the
one used on pp.20–21. The setting is that of `lemma_2`: a probability space with the state `ω`
and each forecast `F i` a `𝓖 i`-measurable version of `P(ω = 1 | 𝓖 i)`. The conditional
probability is stated in product form `ν({ω = 0} ∩ ζ_1) ≤ nα · ν(ζ_1)`; the paper's
"for every α > 0" (Lemma 2) is kept. -/
theorem lemma_3 {X : Type*} [mX : MeasurableSpace X] (ν : Measure X) [IsProbabilityMeasure ν]
    (ω : X → Bool) (hω : Measurable ω) {n : ℕ} (𝓖 : Fin n → MeasurableSpace X)
    (h𝓖 : ∀ i, 𝓖 i ≤ mX) (F : Fin n → X → ℝ) (hFm : ∀ i, Measurable[𝓖 i] (F i))
    (hF : ∀ i, ∀ E, MeasurableSet[𝓖 i] E →
      ν ({x | ω x = true} ∩ E) = ∫⁻ x in E, ENNReal.ofReal (F i x) ∂ν)
    (α : ℝ) (hα : 0 < α) :
    ν ({x | ω x = false} ∩ {x | ∃ i, 1 - α ≤ F i x}) ≤
      ENNReal.ofReal (n * α) * ν {x | ∃ i, 1 - α ≤ F i x} := by sorry

end ForecastAggLearn
Source
Babichenko, Garber, Learning of Optimal Forecast Aggregation in Partial Evidence Environments, arXiv:1802.07107v1, Lemma 3, p.20
Human review
  • Endorsed by Shuze Chen · Oct 31, 2026

    Confirmed by the moderator at approval.

  • Endorsed by iris · Oct 31, 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