Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Greedy dominance of the finite stack-coupling value

Proved
BanditAlgorithm.gittins_finite_stack_coupling_value_le_greedy

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

banditsgittins-indexprobability

Fix a finite horizon NNN and use the common product space of independent arm trajectories. Let the stack-coupling value of a policy be the finite sum, over action sequences, of the policy likelihood times the discounted interleaving of the corresponding prevailing-charge stacks. If π∗\pi^*π∗ always selects an arm of maximal Gittins index, then for every Markov bandit policy π\piπ,

VNstack(π)≤VNstack(π∗).V_N^{\mathrm{stack}}(\pi)\leq V_N^{\mathrm{stack}}(\pi^*).VNstack​(π)≤VNstack​(π∗).

The formula in the formal statement spells out the product trajectory measure, action likelihoods, prevailing-charge stacks, and finite discounted interleavings explicitly, so the result has no dependency on hidden proof infrastructure.

This is the stochastic common-coupling comparison in the proof of the Gittins-index theorem.

Preamble
import Definitions.Def_GittinsPrevailingChargeValue
Formal statement
open MeasureTheory ProbabilityTheory ENNReal

theorem BanditAlgorithm.gittins_finite_stack_coupling_value_le_greedy
    {k : ℕ} {S : Type*} [MeasurableSpace S]
    [StandardBorelSpace 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)
    (π πstar : MarkovBanditPolicy k S)
    (hπstar : IsGittinsIndexPolicy P r α πstar)
    (x : Fin k → S) (N : ℕ) :
    let stackValue := fun (ρ : MarkovBanditPolicy k S) ↦
      let count := fun (a : Fin N → Fin k) (i : Fin k) (t : ℕ) ↦
        ∑ s : Fin N, if (s : ℕ) < t ∧ a s = i then 1 else 0
      let historyBefore := fun (a : Fin N → Fin k)
          (ω : Fin k → ℕ → S) (t : Fin N) ↦
        ((fun u : Fin (t : ℕ) ↦
            ((fun i ↦ ω i (count a i u)),
              a ⟨u, lt_trans u.isLt t.isLt⟩)),
          fun i ↦ ω i (count a i t))
      let likelihood := fun (a : Fin N → Fin k) (ω : Fin k → ℕ → S) ↦
        ∏ t : Fin N, (ρ.select t) (historyBefore a ω t) {a t}
      let charge := fun (ω : Fin k → ℕ → S) (i : Fin k) (u : ℕ) ↦
        (Finset.range (u + 1)).inf'
          ⟨0, Finset.mem_range.2 (Nat.zero_lt_succ u)⟩
          (fun v ↦ gittinsIndex P r α (ω i v))
      let value := fun (a : Fin N → Fin k) (ω : Fin k → ℕ → S) ↦
        ∑ t : Fin N, α ^ (t : ℕ) *
          charge ω (a t) (count a (a t) t)
      let stackMeasure : Measure (Fin k → ℕ → S) :=
        Measure.pi (fun i ↦ markovChainMeasure P (x i))
      ∑ a : Fin N → Fin k,
        ∫ ω, value a ω ∂stackMeasure.withDensity (likelihood a)
    stackValue π ≤ stackValue πstar := 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