Proposition C.4.6 — on a positive recurrent class, convergence of π, of m_zz and of m_iG are equivalent, and then so are those of J, c_zz and c_iG
DisprovedSennottDP.ChainASM.as_class_convergence_equivalencesapproximating-sequenceaverage-costmarkov-chainp2o-batch-b23ap2o-gran-per-chapterp2o-plan-bookp2o-v1
Let be an approximating sequence for the Markov chain with costs on a denumerable state space , and let be a positive recurrent class of . Then the following are equivalent:
- for ;
- for some ;
- for any nonempty finite subset of and .
Assume that any (and hence all) of these hold. Then the following are equivalent:
- for ;
- for some ;
- for any nonempty finite subset of and .
Here is the (possibly infinite) average cost on . The result reduces convergence of steady state probabilities and average costs on a class to convergence of the first passage moments to a single state.
Formalization Note All limits are in ; the equivalences are stated with List.TFAE.
Preamble
import Mathlib import Definitions.Def_SennottDP_ChainASM_MarkovChain import Definitions.Def_SennottDP_ChainASM_ApproxSeq open scoped ENNReal NNReal open Filter Topology open Classical
Formal statement
namespace SennottDP.ChainASM
theorem as_class_convergence_equivalences {S : Type*} [Countable S] [Infinite S]
(Γ : MC S) (AS : ApproxSeq Γ) (R : Set S) (hR : Γ.IsPosRecClass R) :
List.TFAE
[∀ i ∈ R, Tendsto (fun N => AS.steadyStateN N i) atTop (𝓝 (Γ.steadyState i)),
∃ z ∈ R, Tendsto (fun N => AS.meanPassageN {z} N z) atTop
(𝓝 (Γ.meanPassage {z} z)),
∀ G : Finset S, G.Nonempty → (↑G : Set S) ⊆ R → ∀ i ∈ R,
Tendsto (fun N => AS.meanPassageN (↑G : Set S) N i) atTop
(𝓝 (Γ.meanPassage (↑G : Set S) i))] ∧
((∀ i ∈ R, Tendsto (fun N => AS.steadyStateN N i) atTop (𝓝 (Γ.steadyState i))) →
List.TFAE
[∀ i ∈ R, Tendsto (fun N => AS.avgCostN N i) atTop (𝓝 (Γ.classAvgCost R)),
∃ z ∈ R, Tendsto (fun N => AS.passageCostN {z} N z) atTop
(𝓝 (Γ.passageCost {z} z)),
∀ G : Finset S, G.Nonempty → (↑G : Set S) ⊆ R → ∀ i ∈ R,
Tendsto (fun N => AS.passageCostN (↑G : Set S) N i) atTop
(𝓝 (Γ.passageCost (↑G : Set S) i))]) := by sorry
end SennottDP.ChainASM
Source
Sennott, Stochastic Dynamic Programming and the Control of Queueing Systems (Wiley, 1999), p. 305, Proposition C.4.6
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.