Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma B.1, p. 289 — the value of s actions with and without a continuation policy differ by at most γ^s ℛ/(1 − γ)

Proved
QLearning.Convergence.lemmaB1_discounting_tail

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

discountingdynamic-programmingmarkov-decision-processp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Let (X,A,R,P)(X,\mathfrak A,\mathcal R,P)(X,A,R,P) be a finite controlled Markov process with ∣Rx(a)∣≤R|\mathcal R_x(a)|\le\mathcal R∣Rx​(a)∣≤R for all x,ax,ax,a, and let 0<γ<10<\gamma<10<γ<1. For a stationary policy π\piπ, a state xxx and actions a1,…,asa_1,\dots,a_sa1​,…,as​, let Qˉ(x,a1,…,as)\bar Q(x,a_1,\dots,a_s)Qˉ​(x,a1​,…,as​) be the expected discounted reward of the sss actions with 000 terminal reward and Qˉπ(x,a1,…,as)\bar Q^\pi(x,a_1,\dots,a_s)Qˉ​π(x,a1​,…,as​) the expected discounted reward of the same actions followed by π\piπ. Then

∣Qˉπ(x,a1,…,as)−Qˉ(x,a1,…,as)∣≤γs R1−γ.\bigl|\bar Q^\pi(x,a_1,\dots,a_s)-\bar Q(x,a_1,\dots,a_s)\bigr|\le\gamma^s\,\frac{\mathcal R}{1-\gamma}.​Qˉ​π(x,a1​,…,as​)−Qˉ​(x,a1​,…,as​)​≤γs1−γR​.

In particular the difference tends to 000 as s→∞s\to\inftys→∞, uniformly in xxx, π\piπ and the actions, which is the statement of Lemma B.1; the bound is the display in its proof, used in §3.2 to choose sss.

Formalization Note The page prints strict inequalities ∣Vπ∣<R/(1−γ)|V^\pi|<\mathcal R/(1-\gamma)∣Vπ∣<R/(1−γ) and ∣δ∣<γsR/(1−γ)|\delta|<\gamma^s\mathcal R/(1-\gamma)∣δ∣<γsR/(1−γ); with ∣r∣≤R|r|\le\mathcal R∣r∣≤R only ≤\le≤ holds (a constant reward R\mathcal RR gives equality), so the bound is stated with ≤\le≤. "Any arbitrary policy" is taken over stationary deterministic policies π:X→A\pi : X\to\mathfrak Aπ:X→A.

Preamble
import Mathlib
import Definitions.Def_QLearning_Convergence_Setting
open MeasureTheory ProbabilityTheory Filter Topology
Formal statement
namespace QLearning.Convergence

theorem lemmaB1_discounting_tail {X A : Type} [Fintype X] [DecidableEq X] [Fintype A] [DecidableEq A] [Nonempty A]
    (M : FiniteMDP X A) (γ : ℝ) (hγ0 : 0 < γ) (hγ1 : γ < 1)
    (Rbar : ℝ) (hR : ∀ x a, |M.R x a| ≤ Rbar)
    (π : X → A) (x : X) (bs : List A) :
    |seqThenPolicy M γ π x bs - seqValue M γ x bs| ≤ γ ^ bs.length * (Rbar / (1 - γ)) := by sorry

end QLearning.Convergence
Source
Watkins & Dayan, Technical Note: Q-Learning, Machine Learning 8 (1992), p. 289, Lemma B.1 and the display in its proof
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 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