Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 2.5.2 — the monotone Propp–Wilson algorithm

Proved
PalmQueueing.Recurrence.monotone_propp_wilson

by mikedeng1 · Oct 2, 2026 · Mathlib 0df444a (Lean v4.33.1)

couplingexact-samplingmarkov-chainqueueingstationary-processstochastic-recurrence

Theorem 2.5.2. The monotone backwards coalescence time MMM is a.s. finite. Also, the random variables X0−M(1)=X0−M(r)X^{-M}_0(1) = X^{-M}_0(r)X0−M​(1)=X0−M​(r) have the distribution π\piπ.

Theorem 2.5.1 requires all rrr chains to meet, which is prohibitively time-consuming when the state space is large. If there is a partial order ⪯\preceq⪯ on EEE which the updating function preserves,

i⪯j  ⇒  h(i,ξ)⪯h(j,ξ)∀ξ,(2.5.11)i \preceq j \;\Rightarrow\; h(i,\xi) \preceq h(j,\xi) \quad \forall \xi , \tag{2.5.11}i⪯j⇒h(i,ξ)⪯h(j,ξ)∀ξ,(2.5.11)

with 1⪯2⪯⋯⪯r1 \preceq 2 \preceq \dots \preceq r1⪯2⪯⋯⪯r, and if a single updating sequence drives every chain (ξn(i)=ξn\xi_n(i) = \xi_nξn​(i)=ξn​ for all nnn and iii), then the two extremal chains funnel all the others: for all n∈Zn \in \mathbb{Z}n∈Z, all i∈Ei \in Ei∈E, Xnk(1)⪯Xnk(i)⪯Xnk(r)X^k_n(1) \preceq X^k_n(i) \preceq X^k_n(r)Xnk​(1)⪯Xnk​(i)⪯Xnk​(r) for all k≥nk \ge nk≥n. So the coalescence of two chains suffices, and the resulting sample is still exactly π\piπ-distributed.

The finiteness of MMM is immediate in this case: the forwards coupling time of the two chains {Xn0(1)}\{X^0_n(1)\}{Xn0​(1)} and {Xn0(r)}\{X^0_n(r)\}{Xn0​(r)} is dominated by the first time n≥0n \ge 0n≥0 at which Xn0(r)=1X^0_n(r) = 1Xn0​(r)=1, which is finite by the recurrence assumption.

Formalization Note. The model is CFTPMono, in which one i.i.d. uniform sequence drives all the chains. ⪯\preceq⪯ is a partial order (reflexive, transitive, antisymmetric) with a least element 111 and a greatest element rrr, which is what 1⪯2⪯⋯⪯r1 \preceq 2 \preceq \dots \preceq r1⪯2⪯⋯⪯r supplies to the funnelling argument; it is preserved by the update (2.5.11); and the recurrence assumption of the proof is the hypothesis that the chain started at time 000 in state rrr reaches state 111 in finite time a.s.

Preamble
import Mathlib
import Definitions.Def_PalmQueueing_Recurrence_ExactSampling
Formal statement
/-!
# Theorem 2.5.2: the monotone Propp-Wilson algorithm (§2.5.3, p.113)
-/

namespace PalmQueueing.Recurrence

open MeasureTheory

variable {Ω : Type*} [MeasurableSpace Ω]

/-- **Theorem 2.5.2** (p.113). The monotone backwards coalescence time `M` is a.s. finite. Also,
the random variables `X_0^{-M}(1) = X_0^{-M}(r)` have the distribution `π`.

Theorem 2.5.1 requires all `r` chains to meet, which is prohibitively slow when the state space is
large. If the state space carries a partial order `≼` that the updating function preserves,

`(2.5.11)  i ≼ j ⟹ h(i, ξ) ≼ h(j, ξ)  ∀ξ`,

with `1 ≼ 2 ≼ … ≼ r`, and if a **single** updating sequence drives every chain, then the two
extremal chains funnel all the others: `X^k_n(1) ≼ X^k_n(i) ≼ X^k_n(r)` for every `i` and every
`k ≥ n`. So the coalescence of two chains suffices, and the resulting sample is still **exactly**
`π`-distributed.

The hypotheses below are the book's. The model is `CFTPMono`, whose chains are all driven by
the single sequence `{ξ_n}`. `le` is the partial order `≼` (reflexive, transitive, antisymmetric),
and `bot` and `top` are its least and greatest elements, which is what the book's
`1 ≼ 2 ≼ … ≼ r` is used for in the funnelling argument; the update preserves `le` (2.5.11); and
`hrec` is the recurrence assumption the proof invokes: the chain started at the top state `r` at
time `0` reaches the bottom state `1` in finite time. -/
theorem monotone_propp_wilson {r : ℕ} (C : CFTPMono Ω r) (le : Fin r → Fin r → Prop)
    (hrefl : ∀ i, le i i) (htrans : ∀ i j k, le i j → le j k → le i k)
    (hantisymm : ∀ i j, le i j → le j i → i = j)
    (bot top : Fin r)
    (hbot : ∀ i, le bot i) (htop : ∀ i, le i top)
    (hpreserve : ∀ i j : Fin r, le i j → ∀ x : ℝ, le (C.h i x) (C.h j x))
    (hrec : ∀ᵐ ω ∂C.P, ∃ n : ℕ, 1 ≤ n ∧ C.X 0 n top ω = bot) :
    (∀ᵐ ω ∂C.P, ∃ n : ℕ, 1 ≤ n ∧ C.CoalescedExtremal bot top n ω) ∧
    ∀ Z : Ω → Fin r, Measurable Z →
      (∀ᵐ ω ∂C.P, ∀ n : ℕ, 1 ≤ n → C.CoalescedExtremal bot top n ω →
        Z ω = C.X (-(n : ℤ)) n bot ω) →
      ∀ j : Fin r, (C.P {ω | Z ω = j}).toReal = C.pi j := by sorry

end PalmQueueing.Recurrence
Source
Baccelli & Bremaud, Elements of Queueing Theory: Palm Martingale Calculus and Stochastic Recurrences, 2nd ed., Springer 2003, p. 113, Theorem 2.5.2
Human review
  • Endorsed by Shuze Chen · Oct 5, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 5, 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