Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma (p. 628) — discount optimal stationary policies along α_n → 1 have a subsequence with a stationary limit point

Proved
Sennott1989.AvgCost.lemma_limit_point

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

discounted-costmarkov-decision-processp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1stationary-policy

Consider a Markov decision chain on the states 0,1,2,…0,1,2,\dots0,1,2,… with finite action sets. Let (αn)(\alpha_n)(αn​) be a sequence of discount factors in (0,1)(0,1)(0,1) converging to 111, and for each nnn let fαnf_{\alpha_n}fαn​​ be a stationary policy determined by the right side of the discount optimality equation

Vαn(i)=min⁡a∈Ai{C(i,a)+αn∑jPij(a)Vαn(j)},V_{\alpha_n}(i)=\min_{a\in A_i}\Big\{C(i,a)+\alpha_n\sum_jP_{ij}(a)V_{\alpha_n}(j)\Big\},Vαn​​(i)=a∈Ai​min​{C(i,a)+αn​j∑​Pij​(a)Vαn​​(j)},

i.e. fαn(i)f_{\alpha_n}(i)fαn​​(i) attains the minimum at every iii. Then there exist a subsequence (βn)(\beta_n)(βn​) of (αn)(\alpha_n)(αn​) and a stationary policy fff that is a limit point of (fβn)(f_{\beta_n})(fβn​​): for every state iii there is an integer N(i)N(i)N(i) with

fβn(i)=f(i)for all n≥N(i).f_{\beta_n}(i)=f(i)\qquad\text{for all }n\ge N(i).fβn​​(i)=f(i)for all n≥N(i).

The policy fff is the candidate for average cost optimality in the paper's Theorem.

Formalization Note The conclusion is IsLimitPoint of the published Sennott (1999) development: a strictly increasing index map rrr with fαr(k)(i)=f(i)f_{\alpha_{r(k)}}(i)=f(i)fαr(k)​​(i)=f(i) for all large kkk, for each iii.

Preamble
import Mathlib
import Definitions.Def_Sennott1989_AvgCost_Assumptions

open scoped ENNReal NNReal
open Filter Topology
Formal statement
namespace Sennott1989.AvgCost

open SennottDP.Discounted

/-- Sennott (1989), §2, Lemma (unnumbered), p. 628. Let `α_n` be any sequence of discount factors
in `(0, 1)` converging to `1`, and let `f_{α_n}` be the associated discount optimal stationary
policies, i.e. `f_{α_n}` realizes the minimum in the discount optimality equation (2) at
`α = α_n`. Then there exist a subsequence `(β_n)` of `(α_n)` and a stationary policy `f` that is a
limit point of `(f_{β_n})`: for every state `i` there is an integer `N(i)` with
`f_{β_n}(i) = f(i)` for all `n ≥ N(i)`.

**Formalization Note** "There exist a subsequence and a limit point `f` of `(f_{β_n})`" is
`IsLimitPoint M fs f`: there is a strictly increasing `r : ℕ → ℕ` with `fs (r k) i = f i` for all
sufficiently large `k`, for each `i`. -/
theorem lemma_limit_point {Act : Type} (M : MDC ℕ Act) (αs : ℕ → ℝ≥0)
    (hα0 : ∀ n, 0 < αs n) (hα1 : ∀ n, αs n < 1) (hlim : Tendsto αs atTop (𝓝 1))
    (fs : ℕ → StationaryPolicy M)
    (hfs : ∀ n, Realizes M (fs n) (αs n) (valueFn M (αs n))) :
    ∃ f : StationaryPolicy M, IsLimitPoint M fs f := by sorry

end Sennott1989.AvgCost
Source
Sennott, Average Cost Optimal Stationary Policies in Infinite State Markov Decision Processes with Unbounded Costs, Oper. Res. 37(4):626–633 (1989), DOI 10.1287/opre.37.4.626, §2, Lemma (unnumbered), p. 628
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