Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Pathwise Gittins terminal-potential envelope

Proved
BanditAlgorithm.gittins_current_terminal_potential_le_stack_envelope

by Harry_Xu · Aug 1, 2026 · Mathlib 0df444a (Lean v4.33.1)

banditsgittins-indexprobability

On a common stack realization of the arms, consider any length-NNN interleaving. The sum of the arms' current Gittins retirement values after this history is bounded above by an explicit nonnegative envelope. For each arm, the envelope consists of the absolute discounted-reward values at its first N+1N+1N+1 stack states, together with a geometric factor multiplying its initial absolute value and the absolute rewards observed along the first NNN subsequent stack states.

UN(ω,a)≤EN(ω).U_N(\omega,a)\le E_N(\omega).UN​(ω,a)≤EN​(ω).

This pathwise bound is uniform over the action interleaving and supplies the domination needed to remove the terminal retirement term in the finite-horizon Gittins comparison.

Preamble
import Definitions.Def_GittinsTerminalPotential
Formal statement
open MeasureTheory ProbabilityTheory ENNReal

theorem BanditAlgorithm.gittins_current_terminal_potential_le_stack_envelope
    {k N : ℕ} {S : Type*} [MeasurableSpace S]
    (P : Kernel S S) [IsMarkovKernel P]
    {r : S → ℝ} (hr : Measurable r)
    {α : ℝ} (hα0 : 0 < α) (hα1 : α < 1)
    (hint : DiscountedRewardIntegrable P r α)
    (ω : Fin k → ℕ → S) (a : Fin N → Fin k) :
    let count := fun (i : Fin k) (t : ℕ) ↦
      ∑ s : Fin N, if (s : ℕ) < t ∧ a s = i then 1 else 0
    let stackHistory : MarkovBanditHistory k S N :=
      ((fun t ↦ ((fun i ↦ ω i (count i t)), a t)),
        fun i ↦ ω i (count i N))
    let absoluteValue := fun y : S ↦
      ∫ path, ∑' t : ℕ, α ^ t * |r (path t)|
        ∂markovChainMeasure P y
    let envelope := ∑ i : Fin k,
      ((∑ u ∈ Finset.range (N + 1), absoluteValue (ω i u)) +
        (∑' t : ℕ, α ^ t) *
          (absoluteValue (ω i 0) +
            ∑ v ∈ Finset.range N, |r (ω i (v + 1))|))
    currentGittinsRetirementPotential P r α stackHistory ≤ envelope := by
  sorry
Source
Tor Lattimore and Csaba Szepesvári, Bandit Algorithms, §35.4, proof of Lemma 35.10, printed pp. 452–453 (free PDF pp. 460–461), https://tor-lattimore.com/downloads/book/book.pdf

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