Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 6.2, (72), p. 45 — posterior sampling certifies dec^Sq_γ(µ, M̄) ≤ inf_{Δ>0}{2Δ + 24·sup_M θ(F_M − f^M, Δ, γ⁻¹; ρ_µ)·log²(γ)/γ} for γ ≥ e

Proved
StatComplexityDM.Disagreement.posterior_sampling_sq_dec_disagreement

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

decision-estimation-coefficientdisagreement-coefficientp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-bookp2o-v1posterior-sampling

Let FM⊆(Π→[0,1])\mathcal{F}_{\mathcal{M}} \subseteq (\Pi \to [0, 1])FM​⊆(Π→[0,1]) be the class of mean-reward functions of a model class M\mathcal{M}M. Let μ=∑iwiδMi\mu = \sum_i w_i \delta_{M_i}μ=∑i​wi​δMi​​ be a finitely supported prior over M\mathcal{M}M, where each MiM_iMi​ has mean reward fMi∈FMf^{M_i} \in \mathcal{F}_{\mathcal{M}}fMi​∈FM​ and optimal decision πMi∈arg⁡max⁡πfMi(π)\pi_{M_i} \in \arg\max_{\pi} f^{M_i}(\pi)πMi​​∈argmaxπ​fMi​(π). Let M‾\overline{M}M be a reference model, not necessarily in M\mathcal{M}M, with mean reward fM‾:Π→[0,1]f^{\overline{M}} : \Pi \to [0, 1]fM:Π→[0,1]. The posterior sampling strategy plays ρμ(π):=μ({πM=π})\rho_\mu(\pi) := \mu(\{\pi_M = \pi\})ρμ​(π):=μ({πM​=π}), the law of πM\pi_MπM​ under M∼μM \sim \muM∼μ. Then for all γ≥e\gamma \ge eγ≥e it certifies

dec‾γSq(μ,M‾)≤inf⁡Δ>0{2Δ+24 sup⁡M∈Mθ(FM−fM,Δ,γ−1;ρμ)log⁡2(γ)γ},\underline{\mathrm{dec}}^{\mathrm{Sq}}_\gamma(\mu, \overline{M}) \le \inf_{\Delta > 0} \left\{ 2\Delta + 24\, \frac{\sup_{M\in\mathcal{M}} \theta(\mathcal{F}_{\mathcal{M}} - f^M, \Delta, \gamma^{-1}; \rho_\mu) \log^2(\gamma)}{\gamma} \right\},dec​γSq​(μ,M)≤Δ>0inf​{2Δ+24γsupM∈M​θ(FM​−fM,Δ,γ−1;ρμ​)log2(γ)​},

that is, the objective of (70),

EM∼μ Eπ∼ρμ[fM(πM)−fM(π)−γ(fM(π)−fM‾(π))2],\mathbb{E}_{M\sim\mu}\, \mathbb{E}_{\pi\sim\rho_\mu}\Bigl[ f^M(\pi_M) - f^M(\pi) - \gamma \bigl(f^M(\pi) - f^{\overline{M}}(\pi)\bigr)^2 \Bigr],EM∼μ​Eπ∼ρμ​​[fM(πM​)−fM(π)−γ(fM(π)−fM(π))2],

with M∼μM \sim \muM∼μ and π∼ρμ\pi \sim \rho_\muπ∼ρμ​ independent, is at most the right-hand side. Here θ\thetaθ is the disagreement coefficient of Definition 6.3.

Theorem 6.2 is a prior-dependent upper bound on the dual square-loss Decision-Estimation Coefficient. Combined with a bound of the disagreement coefficient by the star number or the eluder dimension (Lemma 6.1) and the minimax theorem, it yields the paper's Theorem 6.1, which recovers the eluder-dimension regret bound of Russo and Van Roy as a special case.

Formalization Note The infimum over Δ>0\Delta > 0Δ>0 is stated as a bound for every Δ>0\Delta > 0Δ>0, and the supremum over M∈MM \in \mathcal{M}M∈M as a bound for every upper bound Θ\ThetaΘ of θ(FM−g,Δ,γ−1;ρμ)\theta(\mathcal{F}_{\mathcal{M}} - g, \Delta, \gamma^{-1}; \rho_\mu)θ(FM​−g,Δ,γ−1;ρμ​) over all g∈FMg \in \mathcal{F}_{\mathcal{M}}g∈FM​; both forms are equivalent to the page's and avoid Lean's default values for empty or unbounded infima and suprema. The bound is stated for the posterior sampling distribution itself, which upper-bounds the infimum over ppp in (70). Priors and ρμ\rho_\muρμ​ are finite weighted families (footnote 5 of the paper); models enter only through their mean rewards and optimal decisions, which is all the square-loss DEC depends on. log⁡\loglog is the natural logarithm and e=exp⁡(1)e = \exp(1)e=exp(1).

Preamble
import Mathlib
import Definitions.Def_StatComplexityDM_Disagreement_Coefficient
Formal statement
namespace StatComplexityDM.Disagreement

/-- Theorem 6.2, (72), p. 45 (arXiv:2112.13487v3). For a class `F_M ⊆ (Π → [0, 1])` of mean-reward
functions, a finitely supported prior `µ = Σ_i w i · δ_{M_i}` over models with `fM i = f^{M_i} ∈ F_M`
and `piM i = π_{M_i} ∈ arg max f^{M_i}`, and a reference model `M̄` (not necessarily in `M`) with
`fbar = f^{M̄} ∈ [0, 1]`, the posterior sampling strategy `ρ_µ(π) = µ({π_M = π})` (the weighted
family `Σ_j w j · δ_{piM j}`) certifies, for all `γ ≥ e`,
`dec^Sq_γ(µ, M̄) ≤ inf_{Δ>0} {2Δ + 24 sup_{M∈M} θ(F_M − f^M, Δ, γ⁻¹; ρ_µ) log²(γ)/γ}`:
the objective of (70) at `p = ρ_µ` is at most `2Δ + 24Θ log²(γ)/γ` for every `Δ > 0` and every
upper bound `Θ` of `θ(F_M − g, Δ, γ⁻¹; ρ_µ)` over all `g = f^M ∈ F_M`. -/
theorem posterior_sampling_sq_dec_disagreement {ι Act : Type*} [Fintype ι] (FM : Set (Act → ℝ))
    (hF : ∀ f ∈ FM, ∀ π, 0 ≤ f π ∧ f π ≤ 1) (w : ι → ℝ) (hw : StatComplexityDM.LowerBound.IsDist w) (fM : ι → Act → ℝ)
    (hfM : ∀ i, fM i ∈ FM) (piM : ι → Act) (hpiM : ∀ i π, fM i π ≤ fM i (piM i))
    (fbar : Act → ℝ) (hfbar : ∀ π, 0 ≤ fbar π ∧ fbar π ≤ 1) (γ : ℝ) (hγ : Real.exp 1 ≤ γ) :
    ∀ Δ : ℝ, 0 < Δ → ∀ Θ : ℝ,
      (∀ g ∈ FM, disagreementCoeff (shiftClass FM g) Δ γ⁻¹ w piM ≤ Θ) →
      sqDecObjective w fM piM fbar γ w piM ≤ 2 * Δ + 24 * Θ * Real.log γ ^ 2 / γ := by sorry

end StatComplexityDM.Disagreement
Source
arXiv:2112.13487v3, Theorem 6.2, (72), p. 45 (proof p. 106)
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