How statement and proof provenance work
The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.
- Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
- AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
- AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.
These labels describe origin, not correctness: citations and verification chips remain separate evidence.
Small-set coupling yields almost-everywhere total-variation convergence
Statement
Let be countably generated, -irreducible and aperiodic with invariant probability . If is -small, , and , then for -almost every .
Facts & Assumptions
Given: The kernel, irreducibility measure, aperiodicity, invariant law and small set.
Phi-irreducibility for a general-state Markov kernel: every positive- set is accessible from every state.
Aperiodicity for a general-state Markov kernel: no cyclic decomposition exists on the stationary irreducible part.
Invariant and stationary distribution for a Markov kernel: is invariant.
Invariant mass and almost-sure hitting of accessible sets: for any stationary kernel, if every start has positive chance to hit a target, stationary almost every start hits it almost surely.
Aperiodicity makes minorization times eventually consecutive: for all sufficiently large when and each state can reach the minorization.
Finite products of probability kernels are probability kernels: independent coordinates define product probability kernels.
Total variation duality, kernel contraction, and coupling bound: a coupling bounds total variation by its disagreement probability.
Canonical Markov chain on path space: the kernel defines its canonical path law under its stated AC convention.
Discrete strong Markov property: the pair chain restarts at discrete stopping times.
Measures agreeing on a generating pi-system are equal under an increasing finite-measure exhaustion from that pi-system: finite measures agreeing on measurable rectangles agree on the product sigma-algebra.
The Axiom of Choice: the path and split-kernel constructions use AC as specified by [F9].
Proof
Given: The hypotheses and facts [F1]–[F11], with [A1].
By [F5], . Integrating the small-set inequality over and using invariance gives for all , hence .
Fix . Irreducibility gives with , so . By [F6], for every sufficiently large . Thus for every sufficiently large , and in particular for every sufficiently large integer . This holds for every .
Let . For measurable rectangles , . By [F11], this proves that is invariant for . By step 2.1, every starting pair has positive probability of hitting under . Apply the stationary positive-hitting lemma [F5] to , invariant law , and target : the stationary pair hits almost surely from -almost every pair.
Start the pair with law and run independent transitions before each split attempt. By step 3.1 and [F5], the first hit of occurs almost surely. Its location law is absolutely continuous with respect to : for any -null set , the event that the first hit lies in is contained in , which has probability zero by stationarity. More generally, if the pair at the start of a waiting period has any subprobability law , then for every , because invariance gives for -almost every ; hence its next first-hit location is also absolutely continuous and the hit occurs almost surely by [F5]. At a visit with , use one common Bernoulli() coin: success draws a shared value and couples immediately, while failure draws independently from and . Each coordinate retains its marginal. For every -null , stationarity gives for -almost every , while the product expansion gives ; thus the failed endpoint law is absolutely continuous whenever the attempt-state law is. It therefore reaches almost surely again, with an absolutely continuous first-hit law. If , the first split succeeds and no failure branch occurs.
Each visit succeeds with probability , so successive failures have probability . The coupling succeeds almost surely for -almost every initial pair. Fubini gives that for -almost every , a chain from couples almost surely to an independent stationary chain. The coupling inequality gives convergence along times . For , , contraction by and invariance of gives convergence for every .
Depends on
- Canonical Markov chain on path space
- The Axiom of Choice
- Aperiodicity for a general-state Markov kernel
- Invariant and stationary distribution for a Markov kernel
- Phi-irreducibility for a general-state Markov kernel
- Small-set minorization and geometric drift
- Total variation distance for probability laws
- Total variation duality, kernel contraction, and coupling bound
- Aperiodicity makes minorization times eventually consecutive
- Finite products of probability kernels are probability kernels
- Invariant mass and almost-sure hitting of accessible sets
- Discrete strong Markov property
- Measures agreeing on a generating pi-system are equal under an increasing finite-measure exhaustion from that pi-system
- Tonelli's theorem for nonnegative measurable functions on a sigma-finite product
Used by
Dependency tree · two levels
43 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.
Sources
- Gareth O. Roberts and Jeffrey S. Rosenthal, General State Space Markov Chains and MCMC Algorithms (standard reference, not scraped)
- Randal Douc, Eric Moulines, Pierre Priouret, and Philippe Soulier, Markov Chains (standard reference, not scraped)