Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proof of Theorem A.1, p. 26 (corrected) — if R(λ̂) > α then R̂⁺(λ†) < α at λ† = sup{λ ∈ Λ : R(λ) > α}

Proved
RiskControl.UCB.ucb_lt_at_failure_point

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

calibrationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1risk-controlupper-confidence-bound

Let Λ⊆R‾\Lambda \subseteq \overline{\mathbb R}Λ⊆R be closed, let R:R‾→RR : \overline{\mathbb R} \to \mathbb RR:R→R be continuous on Λ\LambdaΛ, let α∈R\alpha \in \mathbb Rα∈R, and let r:R‾→Rr : \overline{\mathbb R} \to \mathbb Rr:R→R be any function (one realization of the upper confidence bound R^+\widehat R^+R+). Let

λ^=inf⁡{λ∈Λ:r(λ′)<α for all λ′∈Λ, λ′≥λ}\hat\lambda = \inf\{\lambda \in \Lambda : r(\lambda') < \alpha \text{ for all } \lambda' \in \Lambda,\ \lambda' \ge \lambda\}λ^=inf{λ∈Λ:r(λ′)<α for all λ′∈Λ, λ′≥λ}

as in (4). If this calibration set is nonempty and R(λ^)>αR(\hat\lambda) > \alphaR(λ^)>α, then the set {λ∈Λ:R(λ)>α}\{\lambda \in \Lambda : R(\lambda) > \alpha\}{λ∈Λ:R(λ)>α} is nonempty and, with λ†=sup⁡{λ∈Λ:R(λ)>α}\lambda^\dagger = \sup\{\lambda \in \Lambda : R(\lambda) > \alpha\}λ†=sup{λ∈Λ:R(λ)>α},

r(λ†)<α.r(\lambda^\dagger) < \alpha.r(λ†)<α.

This is a purely deterministic statement: whenever UCB calibration selects a parameter with risk above α\alphaα, the upper confidence bound is already below α\alphaα at the fixed point λ†\lambda^\daggerλ†. Together with R(λ†)≥αR(\lambda^\dagger) \ge \alphaR(λ†)≥α, the failure of calibration is contained in the failure of the pointwise bound (3) at λ†\lambda^\daggerλ†.

Formalization Note The printed proof argues "λ^<λ∗\hat\lambda < \lambda^*λ^<λ∗" and then "R^+(λ∗)<α\widehat R^+(\lambda^*) < \alphaR+(λ∗)<α" with λ∗=inf⁡{λ∈Λ:R(λ)≤α}\lambda^* = \inf\{\lambda \in \Lambda : R(\lambda) \le \alpha\}λ∗=inf{λ∈Λ:R(λ)≤α}; that route needs R(λ∗)=αR(\lambda^*) = \alphaR(λ∗)=α, false for non-interval Λ\LambdaΛ (see the failure-point milestone). This statement uses λ†\lambda^\daggerλ† instead. No monotonicity of RRR is assumed.

Preamble
import Mathlib
import Definitions.Def_RiskControl_UCB_Setting
Formal statement
namespace RiskControl.UCB

/-- Proof of Theorem A.1, p. 26, corrected: deterministically, for one realization `r` of
`λ ↦ R̂⁺(λ)`, if the calibration set of (4) is nonempty and `R(λ̂) > α`, then the set
`{λ ∈ Λ : R(λ) > α}` is nonempty and `R̂⁺(λ†) < α` at its supremum `λ†`. -/
theorem ucb_lt_at_failure_point (Λ : Set EReal) (hΛ : IsClosed Λ) (R : EReal → ℝ)
    (hcont : ContinuousOn R Λ) (r : EReal → ℝ) (α : ℝ)
    (hS : (calSet Λ r α).Nonempty) (hfail : α < R (lambdaHat Λ r α)) :
    {l ∈ Λ | α < R l}.Nonempty ∧ r (sSup {l ∈ Λ | α < R l}) < α := by sorry

end RiskControl.UCB
Source
Bates, Angelopoulos, Lei, Malik & Jordan, arXiv:2101.02703v3, Proof of Theorem A.1, p. 26, second and third sentences, corrected (λ† for λ*)
Human review
  • Endorsed by Shuze Chen · Oct 8, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 8, 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