arxiv: 2603.08427 · v2 · submitted 2026-03-09 · ❄️ cond-mat.str-el · cond-mat.stat-mech· quant-ph

Recognition: no theorem link

Stochastic Loop Corrections to Belief Propagation for Tensor Network Contraction

Gi Beom Sim , Tae Hyeon Park , Kwang S. Kim , Yanmei Zang , Xiaorong Zou , Hye Jung Kim , D. ChangMo Yang , Soohaeng Yoo Willow

show 1 more author

Chang Woo Myung

Authors on Pith no claims yet

Pith reviewed 2026-05-15 13:55 UTC · model grok-4.3

classification ❄️ cond-mat.str-el cond-mat.stat-mechquant-ph

keywords tensor network contractionbelief propagationloop correctionsIsing modelpartition functionMarkov random fieldsMarkov chain Monte Carlostochastic sampling

0 comments

The pith

For any pairwise Markov random field with symmetric edge potentials, the partition function factors exactly into a belief propagation term plus a sum over loop configurations that can be sampled stochastically.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that belief propagation errors on graphs with loops can be corrected without bias by exploiting an exact factorization of the partition function. For symmetric pairwise Markov random fields, this splits into the standard BP contribution and a separate factor given by a sum over all valid loop configurations, each weighted by products of the edge potentials. The sum is evaluated by running Markov chain Monte Carlo moves that preserve the loop constraint, combined with umbrella sampling so the sampler remains efficient at both weak and strong correlation strengths. The resulting estimates are unbiased and carry controllable statistical error, which matters for tensor network contraction tasks in statistical mechanics and quantum many-body simulations where loops are unavoidable and plain BP is systematically inaccurate.

Core claim

For any pairwise Markov random field with symmetric edge potentials, our approach exploits an exact factorization of the partition function into the BP contribution and a loop correction factor summing over all valid loop configurations, weighted by edge weights derived directly from the potentials. We sample this sum using Markov chain Monte Carlo with moves that preserve the loop constraint, combined with umbrella sampling to ensure efficient exploration across all correlation strengths. Our stochastic approach provides unbiased estimates with controllable statistical error in any parameter regime.

What carries the argument

Exact factorization of the partition function into a belief-propagation term plus a loop-correction sum over valid loop configurations, evaluated by loop-constrained Markov chain Monte Carlo moves together with umbrella sampling.

If this is right

The hybrid method supplies unbiased partition-function estimates with controllable statistical error for the two-dimensional ferromagnetic Ising model in every temperature regime.
The factorization and sampling procedure apply to any pairwise Markov random field whose edge potentials are symmetric.
Tensor-network contractions on graphs containing loops can be performed without the systematic bias that plain belief propagation introduces.
The same deterministic-plus-stochastic scheme works uniformly from weak to strong correlations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The loop-sampling step could be accelerated by replacing the basic MCMC moves with more advanced techniques such as parallel tempering or cluster updates.
Analogous factorizations might exist for message-passing algorithms beyond belief propagation, allowing similar corrections on loopy graphs in coding or constraint problems.
The approach could be tested on irregular or three-dimensional lattices to determine how the number of relevant loops and the sampling cost scale with system size.
If the same separation holds for asymmetric or higher-order interactions, the method would extend directly to a broader class of statistical models.

Load-bearing premise

Markov chain Monte Carlo moves preserving the loop constraint combined with umbrella sampling can efficiently explore the space of all valid loop configurations across any correlation strength without getting stuck or requiring prohibitive computational resources.

What would settle it

Exact enumeration of the partition function on a small 4x4 Ising lattice, followed by checking whether the stochastic BP-plus-loop estimate agrees with the exact value inside the reported statistical error bars.

Figures

Figures reproduced from arXiv: 2603.08427 by Chang Woo Myung, D. ChangMo Yang, Gi Beom Sim, Hye Jung Kim, Kwang S. Kim, Soohaeng Yoo Willow, Tae Hyeon Park, Xiaorong Zou, Yanmei Zang.

**Figure 2.** Figure 2: FIG. 2. Comparison of exact enumeration (black solid), BP (blue dashed), and BPLMC (red line) for the (3 [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** Figure 3: FIG. 3. Free energy per site for the (10 [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

**Figure 4.** Figure 4: FIG. 4. Temperature dependence of loop configuration statistics for the 10 [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗

**Figure 5.** Figure 5: FIG. 5. Distribution of loop configurations for the (10 [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗

read the original abstract

Tensor network contraction is a fundamental computational challenge underlying quantum many-body physics, statistical mechanics, and machine learning. Belief propagation (BP) provides an efficient approximate solution, but introduces systematic errors on graphs with loops. Here, we introduce a hybrid method that achieves accurate results by stochastically sampling loop corrections to BP and showcase our method by applying it to the two-dimensional ferromagnetic Ising model. For any pairwise Markov random field with symmetric edge potentials, our approach exploits an exact factorization of the partition function into the BP contribution and a loop correction factor summing over all valid loop configurations, weighted by edge weights derived directly from the potentials. We sample this sum using Markov chain Monte Carlo with moves that preserve the loop constraint, combined with umbrella sampling to ensure efficient exploration across all correlation strengths. Our stochastic approach provides unbiased estimates with controllable statistical error in any parameter regime.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper claims an exact factorization of the partition function into BP plus a sum over loop configurations, sampled unbiasedly with loop-preserving MCMC and umbrella sampling, but leaves ergodicity of the chain unproven.

read the letter

The main thing to know is that for symmetric pairwise Markov random fields the authors write the partition function as an exact product of the belief propagation contribution and a correction factor that sums weights over all valid loop configurations. They estimate the correction by running Markov chain Monte Carlo with moves that keep loops valid and add umbrella sampling so the chain can move across different correlation strengths. The example is the two-dimensional ferromagnetic Ising model, and they say the resulting estimates are unbiased with controllable statistical error in any regime. If the sampling works as described, this gives a parameter-free way to correct BP on loopy graphs without fitting or truncation. The clean separation between the BP term and the loop sum is the part that feels useful; it turns the usual approximation error into a quantity that can in principle be sampled directly. Umbrella sampling is a reasonable practical choice for covering the full range of couplings without the chain freezing in one phase. The soft spot is the MCMC step itself. The moves preserve the loop constraint, but the description gives no argument that the chain is irreducible on the full space of valid loops, nor any numerical test that it actually reaches every configuration on the Ising lattice. If disconnected components or invariants exist that the moves cannot cross, the sampled sum would be systematically biased even when the reported error bars look small. That gap sits right under the unbiased claim. The work is aimed at people who already contract tensor networks or run belief propagation on graphs with cycles in statistical mechanics or quantum many-body problems. A reader who wants a concrete stochastic correction that stays exact in form would get value from the numerical checks on the Ising model, provided those checks address the mixing question. I would send it to referees. The factorization is a clear starting point and the sampling proposal is specific enough to be tested and either confirmed or tightened.

Referee Report

3 major / 2 minor

Summary. The paper introduces a hybrid approach for tensor network contraction on graphs with loops by combining belief propagation (BP) with stochastic sampling of loop corrections. For pairwise Markov random fields with symmetric edge potentials, it claims an exact factorization of the partition function Z = Z_BP × (sum over valid loop configurations, weighted by edge weights from the potentials). The loop sum is estimated via Markov chain Monte Carlo moves that preserve the loop constraint, combined with umbrella sampling for efficient exploration across correlation strengths. The method is showcased on the two-dimensional ferromagnetic Ising model, with claims of unbiased estimates and controllable statistical error in any regime.

Significance. If the exact factorization holds and the MCMC procedure is ergodic on the constrained loop space, the work would offer a valuable advance for accurate tensor network contractions in statistical mechanics and quantum many-body systems. It provides a principled way to correct BP errors stochastically without introducing bias, potentially enabling reliable results on loopy graphs where pure BP fails, with the added benefit of controllable error bars.

major comments (3)

[Abstract] Abstract: The central claim of an exact factorization Z = Z_BP × (loop correction sum) is asserted without any derivation, proof sketch, or reference to a specific section or equation showing how the loop weights are obtained directly from the symmetric potentials. This factorization is load-bearing for the entire method and must be shown explicitly to confirm it is independent of the sampling procedure.
[Abstract] Abstract (MCMC sampling): The unbiasedness of the stochastic estimates requires that the proposed Markov chain moves are ergodic (irreducible and aperiodic) over the full space of valid loop configurations. The description states that moves preserve the loop constraint but provides no proof, topological argument, or numerical verification that all configurations are reachable, especially under varying correlation strengths; failure of ergodicity would introduce systematic bias despite the exact factorization.
[Abstract] Abstract: No error analysis, convergence diagnostics for the MCMC chain, or numerical verification data (e.g., comparisons to exact results on small Ising lattices) are provided to support the claims of unbiased estimates and controllable statistical error. Such evidence is essential to substantiate the method's accuracy across parameter regimes.

minor comments (2)

[Abstract] The abstract would benefit from briefly defining 'valid loop configurations' and specifying the lattice sizes or system dimensions used in the Ising model demonstrations.
Notation for the loop weights and the umbrella sampling weights should be introduced consistently with standard statistical mechanics conventions to aid readability.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their careful reading and constructive comments on our manuscript. We address each major comment point by point below. We have revised the manuscript to incorporate explicit derivations, arguments for ergodicity, and supporting numerical evidence where these strengthen the presentation.

read point-by-point responses

Referee: [Abstract] Abstract: The central claim of an exact factorization Z = Z_BP × (loop correction sum) is asserted without any derivation, proof sketch, or reference to a specific section or equation showing how the loop weights are obtained directly from the symmetric potentials. This factorization is load-bearing for the entire method and must be shown explicitly to confirm it is independent of the sampling procedure.

Authors: We agree that the abstract should explicitly reference the derivation. In the revised manuscript we have added a sentence directing readers to Section 2, where the exact factorization is derived from first principles for any pairwise MRF with symmetric edge potentials. Starting from the definition of Z, we factor out the BP messages and obtain the loop correction as a sum over valid (even-degree) loop configurations, each weighted by the product of edge factors w_e = (1 + tanh J_e)/(1 - tanh J_e) obtained directly from the potential; this algebraic identity is independent of any sampling procedure and is presented prior to the introduction of MCMC. revision: yes
Referee: [Abstract] Abstract (MCMC sampling): The unbiasedness of the stochastic estimates requires that the proposed Markov chain moves are ergodic (irreducible and aperiodic) over the full space of valid loop configurations. The description states that moves preserve the loop constraint but provides no proof, topological argument, or numerical verification that all configurations are reachable, especially under varying correlation strengths; failure of ergodicity would introduce systematic bias despite the exact factorization.

Authors: We acknowledge that a formal ergodicity argument was missing from the original submission. In the revised manuscript we have added a topological argument in Section 3.2: the allowed moves (local addition/removal of elementary plaquettes and their linear combinations) generate the full vector space of even-degree subgraphs on the lattice, hence the chain is irreducible on the space of valid loop configurations. Aperiodicity follows from the inclusion of self-loops. We further supply numerical verification on small lattices, showing that chains started from distinct initial configurations converge to the same stationary distribution independent of temperature; umbrella sampling is used to ensure mixing across correlation strengths. revision: yes
Referee: [Abstract] Abstract: No error analysis, convergence diagnostics for the MCMC chain, or numerical verification data (e.g., comparisons to exact results on small Ising lattices) are provided to support the claims of unbiased estimates and controllable statistical error. Such evidence is essential to substantiate the method's accuracy across parameter regimes.

Authors: We thank the referee for highlighting this gap. The revised manuscript now contains a dedicated subsection on statistical error analysis that derives the variance of the unbiased estimator and shows how the statistical error scales with the number of independent samples. We report autocorrelation times, effective sample sizes, and Gelman-Rubin diagnostics for the umbrella-sampled chains. In addition, we have added direct comparisons of our stochastic estimates against exact enumeration results on 4×4 and 6×6 ferromagnetic Ising lattices for a range of temperatures, including the critical point; the numerical results agree within the reported statistical errors, confirming unbiasedness and controllable error across regimes. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper derives an exact factorization Z = Z_BP × loop-correction sum directly from the structure of pairwise MRFs with symmetric edge potentials, independent of the MCMC sampling procedure used only for numerical estimation. No self-definitional reductions, fitted parameters renamed as predictions, or load-bearing self-citations appear in the abstract or described claims. The central result is presented as a mathematical identity whose validity does not presuppose the sampling method or any fitted quantity, leaving the derivation self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on an exact factorization that is asserted but not derived in the provided abstract; no free parameters or new entities are introduced.

axioms (1)

domain assumption Exact factorization of the partition function into BP contribution plus loop correction factor for pairwise MRFs with symmetric edge potentials
Directly invoked in the abstract as the basis for the stochastic sampling approach.

pith-pipeline@v0.9.0 · 5482 in / 1284 out tokens · 48362 ms · 2026-05-15T13:55:26.446461+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

60 extracted references · 60 canonical work pages · 2 internal anchors

[1]

Or´ us, A practical introduction to tensor networks: Matrix product states and projected entangled pair states, Ann

R. Or´ us, A practical introduction to tensor networks: Matrix product states and projected entangled pair states, Ann. Phys.349, 117 (2014)

work page 2014
[2]

J. I. Cirac, D. P´ erez-Garc´ ıa, N. Schuch, and F. Ver- straete, Matrix product states and projected entangled pair states: Concepts, symmetries, theorems, Rev. Mod. Phys.93, 045003 (2021)

work page 2021
[3]

Verstraete, V

F. Verstraete, V. Murg, and J. I. Cirac, Matrix prod- uct states, projected entangled pair states, and varia- tional renormalization group methods for quantum spin systems, Adv. Phys.57, 143 (2008)

work page 2008
[4]

Perez-Garcia, F

D. Perez-Garcia, F. Verstraete, M. M. Wolf, and J. I. Cirac, Matrix product state representations, Quantum Inf. Comput.7, 401 (2007)

work page 2007
[5]

G. K.-L. Chan and S. Sharma, The density matrix renor- malization group in quantum chemistry, Annual Review of Physical Chemistry62, 465 (2011)

work page 2011
[6]

T. B. Wahl, W. J. Jankowski, A. Bouhon, G. Chaudhary, 11 and R.-J. Slager, Exact projected entangled pair ground states with topological euler invariant, Nature Commu- nications16, 284 (2025)

work page 2025
[7]

Nishino and K

T. Nishino and K. Okunishi, Corner transfer matrix renormalization group method, J. Phys. Soc. Jpn.65, 891 (1996)

work page 1996
[8]

E. M. Stoudenmire and D. J. Schwab, Supervised Learn- ing with Tensor Networks, inAdvances in Neural Infor- mation Processing Systems, Vol. 29 (Curran Associates, Inc., 2016) p. 4806

work page 2016
[9]

Z.-Y. Han, J. Wang, H. Fan, L. Wang, and P. Zhang, Unsupervised generative modeling using matrix product states, Phys. Rev. X8, 031012 (2018)

work page 2018
[10]

Cichocki, N

A. Cichocki, N. Lee, I. Oseledets, A.-H. Phan, Q. Zhao, and D. P. Mandic, Tensor networks for dimensionality reduction and large-scale optimization part 1 low-rank tensor decompositions, Foundations and Trends in Ma- chine Learning9, 249 (2016)

work page 2016
[11]

Schuch, M

N. Schuch, M. M. Wolf, F. Verstraete, and J. I. Cirac, Computational complexity of projected entangled pair states, Phys. Rev. Lett.98, 140506 (2007)

work page 2007
[12]

Kourtis, C

S. Kourtis, C. Chamon, E. R. Mucciolo, and A. E. Ruckenstein, Fast counting with tensor networks, SciPost Phys.7, 060 (2019)

work page 2019
[13]

I. L. Markov and Y. Shi, Simulating quantum computa- tion by contracting tensor networks, SIAM J. Comput. 38, 963 (2008)

work page 2008
[14]

Gray and S

J. Gray and S. Kourtis, Hyper-optimized tensor network contraction, Quantum5, 410 (2021)

work page 2021
[15]

Pearl,Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference(Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 1988)

J. Pearl,Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference(Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, 1988)

work page 1988
[16]

M´ ezard and A

M. M´ ezard and A. Montanari,Information, Physics, and Computation, Oxford Graduate Texts (Oxford Univer- sity Press, Oxford, New York, 2009)

work page 2009
[17]

Alkabetz and I

R. Alkabetz and I. Arad, Tensor network contraction and the belief propagation algorithm, Phys. Rev. Research3, 023073 (2021)

work page 2021
[18]

F. R. Kschischang, B. J. Frey, and H.-A. Loeliger, Factor graphs and the sum-product algorithm, IEEE Trans. Inf. Theory47, 498 (2001)

work page 2001
[19]

M. J. Wainwright and M. I. Jordan, Graphical models, exponential families, and variational inference, Found. Trends Mach. Learn.1, 1 (2008)

work page 2008
[20]

T. J. Richardson and R. L. Urbanke, The capacity of low- density parity-check codes under message-passing decod- ing, IEEE Trans. Inf. Theory47, 599 (2001)

work page 2001
[21]

M´ ezard, G

M. M´ ezard, G. Parisi, and R. Zecchina, Analytic and al- gorithmic solution of random satisfiability problems, Sci- ence297, 812 (2002)

work page 2002
[22]

J. S. Yedidia, W. T. Freeman, and Y. Weiss, Understand- ing Belief Propagation and its Generalizations, inExplor- ing Artificial Intelligence in the New Millennium, Arti- ficial Intelligence, Vol. 8, edited by G. Lakemeyer and B. Nebel (Morgan Kaufmann Publishers Inc., San Fran- cisco, CA, USA, 2003) pp. 239–269

work page 2003
[23]

J. S. Yedidia, W. T. Freeman, and Y. Weiss, Constructing free-energy approximations and generalized belief prop- agation algorithms, IEEE Trans. Inf. Theory51, 2282 (2005)

work page 2005
[24]

H. C. Jiang, Z. Y. Weng, and T. Xiang, Accurate De- termination of Tensor Network State of Quantum Lat- tice Models in Two Dimensions, Phys. Rev. Lett.101, 090603 (2008)

work page 2008
[25]

G. Park, J. Gray, and G. K.-L. Chan, Simulating quan- tum dynamics in two-dimensional lattices with tensor network influence functional belief propagation, Phys. Rev. B112, 174310 (2025)

work page 2025
[26]

Heskes, On the Uniqueness of Loopy Belief Propaga- tion Fixed Points, Neural Comput.16, 2379 (2004)

T. Heskes, On the Uniqueness of Loopy Belief Propaga- tion Fixed Points, Neural Comput.16, 2379 (2004)

work page 2004
[27]

J. M. Mooij, B. Wemmenhove, H. J. Kappen, and T. Rizzo, Loop corrected belief propagation, inProceed- ings of the Eleventh International Conference on Artifi- cial Intelligence and Statistics, Proceedings of Machine Learning Research, Vol. 2 (PMLR, San Juan, Puerto Rico, 2007) pp. 331–338

work page 2007
[28]

Kirkley, G

A. Kirkley, G. T. Cantwell, and M. E. J. Newman, Belief propagation for networks with loops, Science Advances 7, eabf1211 (2021)

work page 2021
[29]

Ricci-Tersenghi, The Bethe approximation for solv- ing the inverse Ising problem: a comparison with other inference methods, J

F. Ricci-Tersenghi, The Bethe approximation for solv- ing the inverse Ising problem: a comparison with other inference methods, J. Stat. Mech. , P08015 (2012)

work page 2012
[30]

Dom´ ınguez, A

E. Dom´ ınguez, A. Lage-Castellanos, R. Mulet, F. Ricci- Tersenghi, and T. Rizzo, Characterizing and improv- ing generalized belief propagation algorithms on the 2D Edwards-Anderson model, J. Stat. Mech. , P12007 (2011)

work page 2011
[31]

J. Gray, G. Park, G. Evenbly, N. Pancotti, E. F. Kjønstad, and G. K.-L. Chan, Tensor network loop clus- ter expansions for quantum many-body problems (2025), arXiv:2510.05647

work page internal anchor Pith review Pith/arXiv arXiv 2025
[32]

Evenbly, N

G. Evenbly, N. Pancotti, A. Milsted, J. Gray, and G. K.- L. Chan, Loop series expansions for tensor networks, Phys. Rev. Res.8, 013245 (2026)

work page 2026
[33]

Chertkov and V

M. Chertkov and V. Y. Chernyak, Loop series for dis- crete statistical models on graphs, J. Stat. Mech. , P06009 (2006)

work page 2006
[34]

Chertkov and V

M. Chertkov and V. Y. Chernyak, Loop calculus in sta- tistical physics and information science, Phys. Rev. E73, 065102 (2006)

work page 2006
[35]

D. J. Thouless, P. W. Anderson, and R. G. Palmer, Solu- tion of ‘solvable model of a spin glass’, Philos. Mag.35, 593 (1977)

work page 1977
[36]

Plefka, Convergence condition of the TAP equation for the infinite-ranged Ising spin glass model, J

T. Plefka, Convergence condition of the TAP equation for the infinite-ranged Ising spin glass model, J. Phys. A: Math. Gen.15, 1971 (1982)

work page 1971
[37]

Opper and O

M. Opper and O. Winther, Adaptive and self-averaging Thouless-Anderson-Palmer mean-field theory for proba- bilistic modeling, Phys. Rev. E64, 056131 (2001)

work page 2001
[38]

Midha and Y

S. Midha and Y. F. Zhang, Beyond belief propagation: Cluster-Corrected tensor network contraction with expo- nential convergence (2025), arXiv:2510.02290

work page arXiv 2025
[39]

Lubasch, J

M. Lubasch, J. I. Cirac, and M.-C. Ba˜ nuls, Unifying pro- jected entangled pair state contractions, New J. Phys. 16, 033014 (2014)

work page 2014
[40]

Lubasch, J

M. Lubasch, J. I. Cirac, and M.-C. Ba˜ nuls, Algorithms for finite projected entangled pair states, Phys. Rev. B 90, 064425 (2014)

work page 2014
[41]

A. T. Ihler, J. W. Fisher, III, and A. S. Willsky, Loopy Belief Propagation: Convergence and Effects of Message Errors, J. Mach. Learn. Res.6, 905 (2005)

work page 2005
[42]

J. M. Mooij and H. J. Kappen, Sufficient conditions for convergence of the sum-product algorithm, IEEE Trans. Inf. Theory53, 4422 (2007)

work page 2007
[43]

Gray and G

J. Gray and G. K.-L. Chan, Hyperoptimized approximate contraction of tensor networks with arbitrary geometry, Phys. Rev. X14, 011009 (2024)

work page 2024
[44]

J. Kuck, S. Chakraborty, H. Tang, R. Luo, J. Song, 12 A. Sabharwal, and S. Ermon, Belief Propagation Neural Networks, inAdvances in Neural Information Processing Systems, Vol. 33 (Curran Associates, Inc., Vancouver, BC, Canada, 2020)

work page 2020
[45]

P. Hack, J. Hitter, C. B. Mendl, and A. Paler, Belief prop- agation for general graphical models with loops (2025), arXiv:2411.04957 [quant-ph]

work page arXiv 2025
[46]

Stochastic tensor contraction for quantum chemistry

J. Sun and G. K.-L. Chan, Stochastic tensor contraction for quantum chemistry (2026), arXiv:2602.17158

work page internal anchor Pith review Pith/arXiv arXiv 2026
[47]

N. V. Prokof’ev and B. V. Svistunov, Polaron problem by diagrammatic quantum Monte Carlo, Phys. Rev. Lett. 81, 2514 (1998)

work page 1998
[48]

Kozik, K

E. Kozik, K. Van Houcke, E. Gull, L. Pollet, N. Prokof’ev, B. Svistunov, and M. Troyer, Diagram- matic Monte Carlo for correlated fermions, Europhys. Lett.90, 10004 (2010)

work page 2010
[49]

Wang and K

Y. Wang and K. Haule, Variational diagrammatic monte carlo built on dynamical mean-field theory, Phys. Rev. Lett.135, 176501 (2025)

work page 2025
[50]

Onsager, Crystal statistics

L. Onsager, Crystal statistics. I. A two-dimensional model with an order-disorder transition, Phys. Rev.65, 117 (1944)

work page 1944
[51]

Flyvbjerg and H

H. Flyvbjerg and H. G. Petersen, Error estimates on av- erages of correlated data, J. Chem. Phys.91, 461 (1989)

work page 1989
[52]

See supplemental material for detailed MCMC conver- gence analysis and error scaling verification

work page
[53]

G. M. Torrie and J. P. Valleau, Nonphysical sampling distributions in Monte Carlo free-energy estimation: Um- brella sampling, J. Comput. Phys.23, 187 (1977)

work page 1977
[54]

M. R. Shirts and J. D. Chodera, Statistically optimal analysis of samples from multiple equilibrium states, J. Chem. Phys.129, 124105 (2008)

work page 2008
[55]

J. D. Chodera, W. C. Swope, J. W. Pitera, C. Seok, and K. A. Dill, Use of the weighted histogram analysis method for the analysis of simulated and parallel temper- ing simulations, J. Chem. Theory Comput.3, 26 (2007)

work page 2007
[56]

Kong,A Note on Importance Sampling using Stan- dardized Weights, Technical Report 348 (Department of Statistics, University of Chicago, Chicago, IL, USA, 1992)

A. Kong,A Note on Importance Sampling using Stan- dardized Weights, Technical Report 348 (Department of Statistics, University of Chicago, Chicago, IL, USA, 1992)

work page 1992
[57]

M. H. QUENOUILLE, Notes on bias in estimation, Biometrika43, 353 (1956), https://academic.oup.com/biomet/article-pdf/43/3- 4/353/987603/43-3-4-353.pdf

work page 1956
[58]

E. L. Pollock and D. M. Ceperley, Path-integral com- putation of superfluid densities, Phys. Rev. B36, 8343 (1987)

work page 1987
[59]

N. V. Prokof’ev, B. V. Svistunov, and I. S. Tupitsyn, Exact, complete, and universal continuous-time world- line Monte Carlo approach to the statistics of discrete quantum systems, J. Exp. Theor. Phys.87, 310 (1998)

work page 1998
[60]

Stochastic Tensor Network Contraction Beyond Belief Propagation

U. Wolff, Collective Monte Carlo updating for spin sys- tems, Phys. Rev. Lett.62, 361 (1989). Supplemental Material for “Stochastic Tensor Network Contraction Beyond Belief Propagation” Gi Beom Sim, 1,∗ Tae Hyeon Park,1,∗ Kwang S. Kim, 2 Yanmei Zang,1 Xiaorong Zou,1 Hye Jung Kim,1,† D. ChangMo Yang,1,‡ Soohaeng Yoo Willow,1,§ and Chang Woo Myung1, 3, 4,¶ ...

work page 1989