Constrained Monte Carlo Markov Chains on Graphs
Pith reviewed 2026-05-25 13:38 UTC · model grok-4.3
The pith
A Markov chain on a graph can be built so its visits converge in frequency to any target distribution whose support the graph can reach.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
The paper establishes that, given any reference probability distribution on a finite set and any undirected graph on the same vertex set, a transition kernel supported only on the edges of the graph can be chosen so that the resulting Markov chain is ergodic on the support of the reference distribution and therefore has that distribution as its unique invariant measure; the empirical occupation measure of the chain then converges to the reference distribution almost surely.
What carries the argument
The graph-constrained transition kernel whose positive entries are restricted to the edges and whose invariant measure is forced to equal the given reference distribution whenever the graph connects the support.
If this is right
- Standard ergodicity theorems apply directly once the chain is irreducible on the target's support.
- The same construction works for any finite undirected graph whose adjacency relation respects the support.
- Convergence of the empirical distribution holds almost surely along almost every trajectory of the chain.
- The procedure yields a valid Monte Carlo sampler under arbitrary forbidden-transition constraints.
- The required transition probabilities can be chosen explicitly from the target masses and the edge set.
Where Pith is reading between the lines
- The result immediately supplies a sampler for distributions on networks when only certain state changes are physically allowed.
- One could test the construction on small complete graphs versus path graphs to see how the mixing time scales with the target's support size.
- The same support-connectedness condition may extend to directed graphs or to graphs that change slowly over time.
- Sampling algorithms in graphical models could adopt this kernel when the model graph and the constraint graph coincide.
Load-bearing premise
The graph must connect every pair of states that both receive positive probability under the target distribution.
What would settle it
Take a target distribution with positive mass on two states that lie in different connected components of the graph and verify that no transition matrix supported only on the existing edges can make the chain visit both states with the required long-run frequencies.
read the original abstract
This paper presents a novel theoretical Monte Carlo Markov chain procedure in the framework of graphs. It specifically deals with the construction of a Markov chain whose empirical distribution converges to a given reference one. The Markov chain is constrained over an underlying graph, so that states are viewed as vertices and the transition between two states can have positive probability only in presence of an edge connecting them. The analysis is carried out on the basis of the relationship between the support of the target distribution and the connectedness of the graph.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript claims to construct a Markov chain on the vertices of a graph, with transitions permitted only along edges, such that the chain's empirical distribution converges to an arbitrary target reference distribution on those vertices. The argument is said to rest on the relationship between the support of the target measure and the connectedness properties of the graph.
Significance. If the construction is carried out explicitly and the convergence is proved, the work would amount to a graph-theoretic restatement of the standard ergodic theorem for finite-state irreducible aperiodic Markov chains. No machine-checked proofs, reproducible code, or falsifiable predictions are mentioned, so the significance is limited to possible pedagogical value in constrained sampling settings.
major comments (1)
- [Abstract] No explicit transition matrix, proposal kernel, or acceptance probabilities are supplied in the provided text, so it is impossible to verify that the constructed chain is irreducible and aperiodic precisely on the support of the target (the only condition needed for the ergodic theorem to apply).
Simulated Author's Rebuttal
We thank the referee for their review. We address the single major comment below.
read point-by-point responses
-
Referee: [Abstract] No explicit transition matrix, proposal kernel, or acceptance probabilities are supplied in the provided text, so it is impossible to verify that the constructed chain is irreducible and aperiodic precisely on the support of the target (the only condition needed for the ergodic theorem to apply).
Authors: The manuscript establishes a general theoretical procedure for constructing a graph-constrained Markov chain whose stationary distribution matches an arbitrary target, with convergence guaranteed precisely when the support of the target is connected in the graph. This relationship directly implies the required irreducibility (and, with standard aperiodicity adjustments, ergodicity) on the support without needing a single fixed transition matrix for all graphs. The full text derives the existence of suitable transition probabilities from the connectedness assumption rather than exhibiting one universal kernel. We agree that an illustrative explicit construction on a small graph would strengthen verifiability and will add one in a revised version. revision: partial
Circularity Check
No circularity; standard ergodic convergence on graph support
full rationale
The paper constructs a Markov chain on a graph whose stationary distribution matches a target, with the key condition being that the graph's connectedness contains the target's support. This is the minimal irreducibility requirement for the ergodic theorem to guarantee convergence of empirical averages; the abstract explicitly frames the analysis as resting on this relationship, which is a classical external fact rather than a self-referential definition or fitted input. No equations, self-citations, or ansatzes are visible that reduce the claim to its own inputs by construction. The derivation is self-contained against standard Markov chain theory.
Axiom & Free-Parameter Ledger
axioms (1)
- standard math Standard Markov chain convergence theorems (irreducibility and aperiodicity imply convergence to unique stationary distribution) apply once the graph allows the required transitions.
Reference graph
Works this paper leans on
- [1]
-
[2]
F. Bartolucci, L. Scaccia, and A. Mira. Efficient Bayes factor est imation from the reversible jump output. Biometrika, 93(1):41–52, 2006
work page 2006
-
[3]
J. Beltr´ an and C. Landim. Tunneling and metastability of continuo us time Markov chains. J. Stat. Phys., 140(6):1065–1114, 2010
work page 2010
- [4]
- [5]
-
[6]
B. P. Carlin and S. Chib. Bayesian model choice via Markov chain Mon te Carlo methods. J. R. Stat. Soc. Series B
-
[7]
R. Cerqueti and E. De Santis. Stochastic Ising model with flipping sets of spins and fast decreasing temperature. Ann. Inst. Henri Poincar´ e Probab. Stat., 54(2):757–789, 2018
work page 2018
-
[8]
D. J. Daley. Stochastically monotone Markov chains. Z. Wahrscheinlichkeitstheorie und Verw. Ge- biete, 10:305–317, 1968
work page 1968
-
[9]
E. De Santis and A. Lissandrelli. Developments in perfect simulation of Gibbs measures through a new result for the extinction of Galton-Watson-like processes. J. Stat. Phys. , 147(2):231–251, 2012
work page 2012
- [10]
- [11]
- [12]
- [13]
-
[14]
J. A. Fill and J. Kahn. Comparison inequalities and fastest-mixing Markov chains. Ann. Appl. Probab., 23(5):1778–1816, 2013
work page 2013
-
[15]
A. Frigessi, P. di Stefano, C.-R. Hwang, and S. J. Sheu. Conve rgence rates of the Gibbs sampler, the Metropolis algorithm and other single-site updating dynamics. J. Roy. Statist. Soc. Ser. B , 55(1):205–219, 1993
work page 1993
-
[16]
A. Frigessi, F. Martinelli, and J. Stander. Computational comple xity of Markov chain Monte Carlo methods for finite Markov random fields. Biometrika, 84(1):1–18, 1997. CONSTRAINED MONTE CARLO MARKOV CHAINS ON GRAPHS 15
work page 1997
-
[17]
S. Geman and D. Geman. Stochastic relaxation, Gibbs distributio ns, and the Bayesian restoration of images. IEEE Transactions on Pattern Analysis and Machine Intellig ence, PAMI(6):721–741, 1984
work page 1984
-
[18]
P. J. Green. Reversible jump Markov chain Monte Carlo computa tion and Bayesian model deter- mination. Biometrika, 82(4):711–732, 1995
work page 1995
-
[19]
W. K. Hastings. Monte Carlo sampling methods using Markov chain s and their applications. Biometrika, 57:97–109, 1970
work page 1970
-
[20]
S. Kirkpatrick, C. D. Gelatt, Jr., and M. P. Vecchi. Optimization b y simulated annealing. Science, 220(4598):671–680, 1983
work page 1983
- [21]
- [22]
-
[23]
N. Metropolis, A. W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, and E. Teller. Equation of state calculations by fast computing machines. The Journal of Chemical Physics , 21(6):1087–1092, 1953
work page 1953
-
[24]
J. G. Propp and D. B. Wilson. Exact sampling with coupled Markov c hains and applications to statistical mechanics. In Proceedings of the Seventh International Conference on Ran dom Structures and Algorithms (Atlanta, GA, 1995) , volume 9, pages 223–252, 1996
work page 1995
- [25]
-
[26]
L. Tierney. Markov chains for exploring posterior distributions . Ann. Stat. , 22(4):1701–1762, 1994. University of Macerata, Department of Economics and Law. Vi a Crescimbeni 20, I-62100, Macerata, Italy E-mail address : roy.cerqueti@unimc.it University of Rome La Sapienza, Department of Mathematics. Piazzale Aldo Moro, 5, I-00185, Rome, Italy E-mail add...
work page 1994
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.