Pith. sign in

REVIEW 3 major objections 3 minor 87 references

Machine-learning approaches to accelerating lattice simulations

T0 review · 3 major / 3 minor · reviewed 2026-08-09 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Four families of machine-learning methods can accelerate lattice Monte Carlo simulations without adding systematic bias.

desk verdict A clear, honest status report on ML-accelerated lattice simulations; the no-bias claim needs a held-out-data caveat. read the letter →

arxiv 2502.02670 v2 pith:DOMCBRHC submitted 2025-02-04 hep-lat

Scott Lawrence This is my paper · ORCID
classification hep-lat
keywords latticeQCDmachinelearningnormalizingflowscontourdeformationssignproblemcontrolvariatessignal-to-noisesurrogateobservables
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This review argues that four families of machine-learning techniques can accelerate lattice Monte Carlo simulations without compromising their first-principles character: normalizing flows that improve sampling, contour deformations that mitigate sign and signal-to-noise problems, control variates that reduce variance exactly, and surrogate observables that cheapen measurement. The organizing claim is that all four can be used without introducing any additional systematic bias into the Monte Carlo results. This matters because lattice QCD calculations are extremely expensive, and unbiased speed-ups translate directly into smaller error bars or lower computational cost at fixed physics. The paper also records where the evidence stands: flow-based sampling and surrogate observables have shown utility in four-dimensional gauge theories, while contour deformations and control variates have so far been demonstrated mainly in smaller systems.

What carries the argument

The unifying machinery is the loss-functional view of algorithmic improvement: an acceleration scheme is obtained by finding a function $f(x)$ for which a functional $L[f(x)]\approx 0$, then parameterizing $f$ and training it. The bias-free property, however, is carried by four specific mechanisms. A normalizing flow $\phi(z)$ obeys the change-of-variables identity $p[\phi(z)]\,\det(\partial\phi/\partial z)=\mathcal{N}(z)$, which makes flow-generated samples valid proposals or reweighted samples. Contour deformations rest on Cauchy's integral theorem, $\int_{\mathbb{R}^N} f(z)\,dz=\int_{\mathbb{R}^N} f(\phi(x))\,\det(\partial\phi/\partial x)\,dz$, so holomorphic expectation values are invariant while the average phase $\langle\sigma\rangle=Z/Z_Q$ can improve. Schwinger-Dyson relations $\langle\partial g\rangle=\langle g\,\partial S\rangle$ provide exactly zero-mean control variates. Surrogate observables use the decomposition $\langle O\rangle=\langle \tilde O\rangle+\langle O-\tilde O\rangle$, paying for exact evaluations only enough samples to control the low-variance correction.

What would settle it

Take an exactly solvable lattice model where exact expectation values are known and run all four families on it, using a trained flow with reweighting, a machine-learned contour, a Schwinger-Dyson control variate, and a surrogate-observable estimator; if any final estimator differs from the exact value by more than its statistical uncertainty, the claim that these methods introduce no systematic bias is falsified.

Watch

Extended reading notes

Core claim

The paper's central assertion, stated in its discussion section, is that four broad families of machine-learning algorithms for accelerating lattice simulations—normalizing flows for improving sampling, contour deformations chiefly for improving sign problems, control variates for improving signal-to-noise ratios, and surrogate observables for accelerating measurement—can all be used without introducing any additional systematic bias into Monte Carlo results. The review assembles the evidence for each family: flow-based samplers generate Markov-chain proposals or reweighted samples with short autocorrelation times; contour deformations exploit Cauchy's integral theorem so that expectation values are unchanged while the sampled integrand is tamer; Schwinger-Dyson control variates are exactly zero-mean by construction and can cut noise by an order of magnitude; and surrogate observables combine a cheap approximate estimator with a small exact correction so the final estimator is unbiased. The author also records two repeated patterns: methods with very few parameters often match or beat deep networks, and only flow-based sampling and surrogate observables have yet shown real utility in four-dimensional gauge theories such as QCD.

Load-bearing premise

The load-bearing premise is that the published proofs and demonstrations showing these methods are unbiased are correct and representative, because this review compiles those results rather than re-deriving them.

Editorial extensions

If this is right

  • If the review is right, normalizing flows can be dropped into a lattice Monte Carlo pipeline as proposal generators and trusted to leave equilibrium averages unbiased, with the benefit showing up as shorter autocorrelation times and smaller error bars.
  • Contour deformations can be applied to a sign problem without changing any physical expectation value, because holomorphic integrals are contour-independent; on the theories tried, they reduce the exponential cost of the sign problem, sometimes by more than an order of magnitude in the decay rate.
  • Schwinger-Dyson control variates can shrink the variance of a correlator by an order of magnitude, a speed-up of roughly a factor of 100 in the demonstrated case, while the zero-mean property remains an exact theorem rather than a numerical approximation.
  • Surrogate observables can cut the cost of expensive measurements in lattice QCD by 7 to 38 percent, with the systematic bias removed by evaluating the exact-minus-surrogate difference on a smaller set of configurations.
  • The unbiasedness claim does not by itself promise that every method works at QCD scale: contour deformations and control variates are proven in smaller systems, and extending them to four-dimensional gauge theories is the open step.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Not claimed in the paper, but a natural extension: because each family's unbiasedness is a theorem about the estimator rather than about the particular training run, the four methods should be composable, so flow-based sampling could generate configurations and control variates could measure them with the combined estimator remaining unbiased.
  • The paper observes that few-parameter ansätze often beat large networks, but does not spell out the practical corollary: for a new lattice model, the first thing to try may be a hand-built one- or two-parameter contour shift or flow, reserving deep networks for cases where simple deformations fail.
  • The no-contour existence proofs in the review are limited to an exactly solvable 1+1-dimensional Yang-Mills theory; extending this differential-form lower-bound technique to non-exactly-solvable theories would give practitioners a stopping rule for contour training, telling them when no amount of optimization can make the average phase survive the volume limit.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 3 minor

Summary. This proceedings contribution reviews machine-learning-inspired acceleration of lattice Monte Carlo in four areas: normalizing-flow proposals and samplers (Section 2), contour deformations for sign and signal-to-noise problems (Section 3), Schwinger-Dyson control variates (Section 4), and surrogate observables for cheaper measurements (Section 5). The stated central claim (Section 6) is that all four families can be used without introducing additional systematic bias, with normalizing flows and surrogate observables already useful in four-dimensional gauge theories and with contour deformations and control variates proven in smaller systems. The paper is an overview rather than a new derivation; its mathematical statements are standard and its numerical claims are drawn from the cited literature.

Significance. The review is well organized and useful as a compact entry point to a fast-moving area. It states correctly the basic identities (Cauchy's theorem, Schwinger-Dyson relations, and the Metropolis/reweighting correction for approximate flows), gives concrete performance benchmarks, and is honest about gaps such as the absence of perfect-contour bounds for non-solvable theories and the lack of large-scale QCD demonstrations for contour deformations and control variates. It also cites work by many groups. However, because it is a proceedings review, its reliability depends on the cited articles, several of which are by the author; the manuscript does not independently re-derive or re-run the empirical demonstrations. If the central claim is taken at face value, the review gives practitioners a helpful map, but it should state the statistical conditions under which each method preserves unbiasedness.

major comments (3)
  1. [Section 4, Eqs. (8)-(9) and Section 6] The treatment of control variates omits a held-out-data condition that is needed for the Section 6 claim that all four families 'can be used without introducing any additional systematic bias.' A control variate built from a Schwinger-Dyson relation has zero mean as a theorem, but when its coefficients are 'determined by a fit' on the same ensemble that later estimates <O - f>, the fitted coefficients are correlated with the sample mean of f and E[hat c * bar f] is generically nonzero at finite N. The text should require that the fit be performed on data independent of the ensemble used for the final estimator, or otherwise prove that the fitted variate retains exactly zero mean.
  2. [Section 5, Eq. (10)] The surrogate-observable estimator is unbiased only if the correction term <O - Otilde>_N is evaluated on configurations that are independent of the training set for Otilde. The manuscript states that configurations on which O is computed 'are used to train' the approximation, but it does not say that the samples N in Eq. (10) are held out; if they are not, the in-sample residual systematically underestimates <O - Otilde> and the estimator is biased. This condition should be stated explicitly.
  3. [Section 6, first paragraph] The blanket sentence 'All can be used without introducing any additional systematic bias into the Monte Carlo results' is stronger than what the preceding sections establish. For flows, unbiasedness requires either exactness of the flow or the use of the flow inside a Metropolis accept/reject or importance-reweighting step; for contour deformations, the deformed expectation value is exact only if the contour deformation is valid and the average phase is nonzero; for control variates and surrogates, the parameters must be fixed independently of the estimation ensemble. The discussion should be rephrased to present these conditions rather than an unconditional guarantee.
minor comments (3)
  1. [Section 3, Eq. (4)] In Eq. (4), the integration element on the right-hand side is written dz but should be dx; as written the equality is dimensionally inconsistent.
  2. [Section 2, after Eq. (2)] There is a typo: 'extened' should be 'extended' in the sentence describing applications to theories with fermions.
  3. [Section 5, Eq. (10)] The symbols N and Ntilde are used in Eq. (10) before their meaning is explained in the following sentence; define them before or immediately after the equation for readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the review's bias-free claims rest on standard external identities and cited demonstrations; the held-out-data caveat is a protocol omission, not a circular reduction.

full rationale

This is a review article, not a derivation, and the Section 6 claim that all four families 'can be used without introducing any additional systematic bias' summarizes standard or previously published unbiasedness arguments rather than deriving a new result from a fitted input. Flow-based sampling is unbiased when used as a Metropolis proposal by detailed balance (Section 2), an external argument. Contour deformations rest on Cauchy's integral theorem, Eqs. (3)-(4), where the trained contour merely selects the integration contour and does not alter the equality. Control variates rest on the Schwinger-Dyson identity (8): a fixed linear combination of total-derivative relations has exactly zero mean, and the paper even stresses that 'It is critical that <f>=0 be a theorem, rather than an empirical statement from Monte Carlo data.' Surrogate observables rest on the exact identity (10), with the correction term <O - O~> evaluated to remove bias. The closest issue is that Sections 4 and 5 describe fitting coefficients and training approximations without explicitly stating the standard held-out/independent-ensemble condition needed for finite-sample unbiasedness ('This control variate has V free coefficients, which must be determined by a fit'; 'The true observable O is computed on some set of configurations, which are used to train...'). That is a practical correctness caveat about protocol, not a circular step: the fitted objects are not defined in terms of the predicted expectation value, and no fitted parameter is renamed as a prediction. Author self-citations appear in the contour-deformation and control-variate sections, but they are demonstrations or auxiliary theorems (e.g., [74] on no-perfect-contour bounds), not the justification for the bias-free claim. Therefore no significant circularity is present.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

No free parameters are introduced by this review; all numerical results are quoted from the cited literature. The central mathematical tools are standard: Cauchy's integral theorem for contour deformations, vanishing of total derivatives for Schwinger-Dyson control variates, and change-of-variables for normalizing flows. The main domain assumption is holomorphicity of the Boltzmann weight, which Section 3.3 acknowledges does not hold for all applications. The review also assumes that the cited performance numbers accurately represent the underlying studies, which is a trust assumption since several of those studies share the present author. No new entities are postulated.

assumptions (5)
  • standard math Cauchy's integral theorem permits replacement of the real integration domain by a continuously deformed complex contour (Eqs. 3-4).
    Invoked in Section 3 for contour deformations; this is a standard theorem, but its applicability requires holomorphicity of the integrand.
  • domain assumption The Boltzmann weight e^{-S} is holomorphic on the complexified configuration space.
    Section 3.3 explicitly notes this fails for some applications (e.g., nuclear forces) and must be relaxed or replaced by a nearby holomorphic action; the main contour-deformation narrative assumes it.
  • standard math Expectation values of total derivatives vanish on compact field configuration spaces, yielding Schwinger-Dyson relations as exact zero-expectation control variates (Eq. 8).
    Invoked in Section 4; standard integration by parts, but on a lattice the boundary term vanishes only for compact or periodic spaces.
  • domain assumption A learned normalizing flow used as a Metropolis proposal or reweighting factor preserves unbiasedness if the acceptance or reweighting step is exact.
    Section 2 describes unbiased use of approximate flows via MCMC or reweighting; the review does not detail error control for the learned proposal.
  • domain assumption The reported performance numbers in Table 1 and Figures 1-7 accurately represent the cited studies.
    The review reproduces figures and numbers from prior work without re-analysis; this is a review-level trust assumption, and several cited works share the present author.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Machine-learning approaches to accelerating lattice simulations." pith.science (2026). https://pith.science/paper/DOMCBRHC

@misc{pith2026250202670,
  author       = {Pith},
  title        = {Pith review of: Machine-learning approaches to accelerating lattice simulations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DOMCBRHC}},
  note         = {Machine review of arXiv:2502.02670}
}
read the original abstract

The last decade has seen an explosive growth of interest in exploiting developments in machine learning to accelerate lattice QCD calculations. On the sampling side, generative models are a promising approach to mitigating critical slowing down and topological freezing. Meanwhile, signal-to-noise problems have been shown to be improvable by the use of optimized improved observables. Both techniques can be made free of bias, resulting in trustworthy but reduced statistical errors. This talk reviews recent developments in this field.

Figures

Figures reproduced from arXiv: 2502.02670 by the authors.

Figure 1
Figure 1. Comparison of the autocorrelation time as a function of lattice size when using HMC updates, local Metropolis updates, and normalizing flows. Reproduced from [25]. Normalizing flows naturally generalize to multivariate distributions, where the defining equa￾tion is 𝑝[𝜙(𝑧)] det 𝜕𝜙 𝜕𝑧 = 1 √ 2𝜋 𝑒 −𝑧 2 /2 . (2) An early published example of a normalizing flow is the Box-Müller transform [24], which converts between a un… view at source ↗
Figure 2
Figure 2. Reproduced from [43], the charge of Perylene as a function of chemical potential. Computed with a contour deformation, defined by one parameter. for any continuous (and piecewise differentiable) 𝜙 : R 𝑁 → C 𝑁 parameterizing an 𝑁-dimensional integration contour. Suitable generalizations to functions defined on manifolds that only locally look like R 𝑁 are readily available. This includes the integrals over 𝑆𝑈(𝑁) whic… view at source ↗
Figure 3
Figure 3. From [42], investigations of lattice scalar field theory at complex coupling. At left is shown the partition function of 0 + 1-dimensional scalar (𝜆𝜙4 ) field theory as a function of complex anharmonicity 𝜆; the branch cut on R− can clearly be seen. At right, the average phase of 1 + 1-dimensional lattice scalar field theory at 𝜆 = 𝑖, as a function of volume. Both the defining contour R 𝑉 and the trained contour exh… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: From [70], evidence of exponential improvement in the signal-to-noise problem of two-dimensional 𝑆𝑈(2) Yang-Mills. Shown is the ratio of the variance of the estimator of the Wilson loop, to the the variance of a contour-improved estimator of the same quantity. 3.3 Theo…
Figure 5
Figure 5. Figure 5: No-go regions for 1 + 1-dimensional lattice Yang-Mills at complex coupling 𝛽, from [74]. Outside of the blue lines, perfect integration contours (on which ⟨𝜎⟩ = 1) were found numerically; within the red lines, proofs that no such contours existed were found. The black …
Figure 6
Figure 6. Figure 6: Two approaches to reducing the signal-to-noise problem associated with the correlator in scalar field theory, both based on Schwinger-Dyson control variates. At left [79], a complete basis of leading-order Schwinger-Dyson relations is used, with 𝐿1 regularization to mi…
Figure 7
Figure 7. Figure 7: Predicting 𝑅(𝑡 = 2) (ratio of 3-pt to 2-pt) of kaon quasi-PDF correlators at (𝑝pred, 𝑡pred, 𝑡sep) = (4, 4, 5) from measurements at (3, 4, 5). From [82]. The red curve shows the cost savings from the use of surrogate observables, with a maximum of around 20%. This metho…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

87 extracted references · 21 canonical work pages

  1. [1]

    Chen, H.T

    S.Y. Chen, H.T. Ding, F.Y. Liu, G. Papp and C.B. Yang,Machine learning spectral functions in lattice QCD, 2110.13521

    arXiv

  2. [2]

    Offler,A study of thermal NRQCD with machine learning methods, Ph.D

    S. Offler,A study of thermal NRQCD with machine learning methods, Ph.D. thesis, Swansea U., 2022

    2022

  3. [3]

    Fournier, L

    R. Fournier, L. Wang, O.V. Yazyev and Q. Wu,Artificial neural network approach to the analytic continuation problem, Phys. Rev. Lett.124 (2020) 056401

    2020

  4. [4]

    Kades, J.M

    L. Kades, J.M. Pawlowski, A. Rothkopf, M. Scherzer, J.M. Urban, S.J. Wetzel et al.,Spectral reconstruction with deep neural networks, Phys. Rev. D102 (2020) 096001

    2020

  5. [5]

    L. Wang, S. Shi and K. Zhou,Reconstructing spectral functions via automatic differentiation,Phys. Rev. D106(2022) L051502 [2111.14760]

    arXiv 2022

  6. [6]

    Carrasquilla and R.G

    J. Carrasquilla and R.G. Melko,Machine learning phases of matter, Nature Physics13 (2017) 431

    2017

  7. [7]

    Machine learning phases of an Abelian gauge theory

    J.-H. Peng, Y.-H. Tseng and F.-J. Jiang,Machine learning phases of an Abelian gauge theory, PTEP 2023(2023) 073A03 [2212.14655]

    work page Pith review arXiv 2023

  8. [8]

    Detection of phase transition via convolutional neural network

    A. Tanaka and A. Tomiya,Detection of phase transition via convolutional neural network, J. Phys. Soc. Jap.86 (2017) 063001 [1609.09087]

    work page Pith review arXiv 2017

Show all 87 references
  1. [9]

    van Nieuwenburg, Y.-H

    E.P.L. van Nieuwenburg, Y.-H. Liu and S.D. Huber,Learning phase transitions by confusion,Nature Phys.13(2017) 435 [1610.02048]

    2017 arXiv

  2. [10]

    Rodriguez-Nieva and M.S

    J.F. Rodriguez-Nieva and M.S. Scheurer,Identifying topological order through unsupervised machine learning,Nature Phys.15(2019) 790 [1805.05961]

    2019 arXiv

  3. [11]

    Broecker, J

    P. Broecker, J. Carrasquilla, R.G. Melko and S. Trebst,Machine learning quantum phases of matter beyond the fermion sign problem, Sci. Rep.7(2017) 8823. 14 Machine-learning approaches to accelerating lattice simulations Scott Lawrence

    2017

  4. [12]

    Shanahan, A

    P.E. Shanahan, A. Trewartha and W. Detmold,Machine learning action parameters in lattice quantum chromodynamics, Phys. Rev. D97(2018) 094506 [1801.05784]

    2018 arXiv

  5. [13]

    Carleo and M

    G. Carleo and M. Troyer,Solving the quantum many-body problem with artificial neural networks, Science 355 (2017) 602

    2017

  6. [14]

    D.-L. Deng, X. Li and S.D. Sarma,Machine learning topological states,Phys. Rev. B96 (2017) 195145

    2017

  7. [15]

    D. Luo, G. Carleo, B.K. Clark and J. Stokes,Gauge Equivariant Neural Networks for Quantum Lattice Gauge Theories, Phys. Rev. Lett.127 (2021) 276402 [2012.05232]

    2021 arXiv

  8. [16]

    Holland, A

    K. Holland, A. Ipp, D.I. Müller and U. Wenger,Machine learning a fixed point action for SU(3) gauge theory with a gauge equivariant convolutional neural network, 2401.06481

    arXiv

  9. [17]

    Kanwar,Flow-based sampling for lattice field theories, PoSLATTICE2023(2024) 114 [2401.01297]

    G. Kanwar,Flow-based sampling for lattice field theories, PoSLATTICE2023(2024) 114 [2401.01297]

    2024 arXiv

  10. [18]

    Huang and L

    L. Huang and L. Wang,Accelerated Monte Carlo simulations with restricted Boltzmann machines,Phys. Rev. B95 (2017) 035105 [1610.02746]

    2017 arXiv

  11. [19]

    Tanaka and A

    A. Tanaka and A. Tomiya,Towards reduction of autocorrelation in HMC by machine learning, 1712.03893

    arXiv

  12. [20]

    L. Wang, G. Aarts and K. Zhou,Diffusion models as stochastic quantization in lattice field theory, JHEP 05(2024) 060 [2309.17082]

    2024 arXiv

  13. [21]

    L. Wang, G. Aarts and K. Zhou,Generative Diffusion Models for Lattice Field Theory, in 37th Conference on Neural Information Processing Systems, 11, 20232311.03578

    arXiv

  14. [22]

    K. Zhou, G. Endrődi, L.-G. Pang and H. Stöcker,Regressive and generative neural networks for scalar field theory, Phys. Rev. D100 (2019) 011501 [1810.12879]

    2019 arXiv

  15. [23]

    Pawlowski and J.M

    J.M. Pawlowski and J.M. Urban,Reducing Autocorrelation Times in Lattice Simulations with Generative Adversarial Networks,Mach. Learn. Sci. Tech.1(2020) 045011 [1811.03533]

    2020 arXiv

  16. [24]

    Box and M.E

    G.E. Box and M.E. Muller,A note on the generation of random normal deviates, The annals of mathematical statistics29 (1958) 610

    1958

  17. [25]

    Albergo, G

    M.S. Albergo, G. Kanwar and P.E. Shanahan,Flow-based generative models for Markov chain Monte Carlo in lattice field theory, Phys. Rev. D100 (2019) 034515 [1904.12072]

    2019 arXiv

  18. [26]

    Nicolai,On a New Characterization of Scalar Supersymmetric Theories,Phys

    H. Nicolai,On a New Characterization of Scalar Supersymmetric Theories,Phys. Lett. B89 (1980) 341

    1980

  19. [27]

    Luscher,Trivializing maps, the Wilson flow and the HMC algorithm, Commun

    M. Luscher,Trivializing maps, the Wilson flow and the HMC algorithm, Commun. Math. Phys. 293 (2010) 899 [0907.5491]

    2010 arXiv

  20. [28]

    L. Dinh, D. Krueger and Y. Bengio,NICE: Non-linear Independent Components Estimation, 1410.8516. 15 Machine-learning approaches to accelerating lattice simulations Scott Lawrence

    arXiv

  21. [29]

    L. Dinh, J. Sohl-Dickstein and S. Bengio,Density estimation using Real NVP, 1605.08803

    arXiv

  22. [30]

    Kingma and P

    D.P. Kingma and P. Dhariwal,Glow: Generative Flow with Invertible 1x1 Convolutions, 1807.03039

    arXiv

  23. [31]

    Caselle, E

    M. Caselle, E. Cellini and A. Nada,Sampling the lattice Nambu-Goto string using Continuous Normalizing Flows, JHEP 02(2024) 048 [2307.01107]

    2024 arXiv

  24. [32]

    de Haan, C

    P. de Haan, C. Rainone, M.C.N. Cheng and R. Bondesan,Scaling Up Machine Learning For Quantum Field Theory with Equivariant Continuous Flows, 2110.02673

    arXiv

  25. [33]

    Nicoli, S

    K.A. Nicoli, S. Nakajima, N. Strodthoff, W. Samek, K.-R. Müller and P. Kessel, Asymptotically unbiased estimation of physical observables with neural samplers, Phys. Rev. E 101 (2020) 023304 [1910.13496]

    2020 arXiv

  26. [34]

    Nicoli, C.J

    K.A. Nicoli, C.J. Anders, L. Funcke, T. Hartung, K. Jansen, P. Kessel et al.,Estimation of Thermodynamic Observables in Lattice Field Theories with Deep Generative Models,Phys. Rev. Lett.126 (2021) 032001 [2007.07115]

    2021 arXiv

  27. [35]

    D.Boyda,G.Kanwar,S.Racanière,D.J.Rezende,M.S.Albergo,K.Cranmeretal., Sampling using𝑆𝑈(𝑁) gauge equivariant flows,Phys. Rev. D103 (2021) 074504 [2008.05456]

    2021 arXiv

  28. [36]

    Kanwar, M.S

    G. Kanwar, M.S. Albergo, D. Boyda, K. Cranmer, D.C. Hackett, S. Racanière et al., Equivariant flow-based sampling for lattice gauge theory, Phys. Rev. Lett.125(2020) 121601 [2003.06413]

    2020 arXiv

  29. [37]

    Kanwar,Machine Learning and Variational Algorithms for Lattice Field Theory, Ph.D

    G. Kanwar,Machine Learning and Variational Algorithms for Lattice Field Theory, Ph.D. thesis, MIT, 2021.2106.01975

    2021

  30. [38]

    Abbott et al.,Gauge-equivariant flow models for sampling in lattice field theories with pseudofermions,Phys

    R. Abbott et al.,Gauge-equivariant flow models for sampling in lattice field theories with pseudofermions,Phys. Rev. D106(2022) 074506 [2207.08945]

    2022 arXiv

  31. [39]

    Albergo, D

    M.S. Albergo, D. Boyda, K. Cranmer, D.C. Hackett, G. Kanwar, S. Racanière et al., Flow-based sampling in the lattice Schwinger model at criticality,Phys. Rev. D106(2022) 014514 [2202.11712]

    2022 arXiv

  32. [40]

    Abbott et al.,Normalizing flows for lattice gauge theory in arbitrary space-time dimension, 2305.02402

    R. Abbott et al.,Normalizing flows for lattice gauge theory in arbitrary space-time dimension, 2305.02402

    arXiv

  33. [41]

    Bacchio, P

    S. Bacchio, P. Kessel, S. Schaefer and L. Vaitl,Learning trivializing gradient flows for lattice gauge theories,Phys. Rev. D107 (2023) L051504 [2212.08469]

    2023 arXiv

  34. [42]

    Lawrence, H

    S. Lawrence, H. Oh and Y. Yamauchi,Lattice scalar field theory at complex coupling,Phys. Rev. D106 (2022) 114503 [2205.12303]

    2022 arXiv

  35. [43]

    Rodekamp, E

    M. Rodekamp, E. Berkowitz, C. Gäntgen, S. Krieg, T. Luu, J. Ostmeyer et al.,Single Particle Spectrum of DopedC20H12-Perylene, 2406.06711. 16 Machine-learning approaches to accelerating lattice simulations Scott Lawrence

    arXiv

  36. [44]

    Alexandru, G

    A. Alexandru, G. Basar, P.F. Bedaque and N.C. Warrington,Complex paths around the sign problem, Rev. Mod. Phys.94(2022) 015006 [2007.05436]

    2022 arXiv

  37. [45]

    Witten,Analytic Continuation Of Chern-Simons Theory,AMS/IP Stud

    E. Witten,Analytic Continuation Of Chern-Simons Theory,AMS/IP Stud. Adv. Math.50 (2011) 347 [1001.2933]

    2011 arXiv

  38. [46]

    AuroraSciencecollaboration, New approach to the sign problem in quantum field theories: High density QCD on a Lefschetz thimble,Phys. Rev. D86 (2012) 074506 [1205.3996]

    2012 arXiv

  39. [47]

    Alexandru, P.F

    A. Alexandru, P.F. Bedaque, H. Lamm and S. Lawrence,Deep Learning Beyond Lefschetz Thimbles,Phys. Rev. D96(2017) 094505 [1709.01971]

    2017 arXiv

  40. [48]

    Y. Mori, K. Kashiwa and A. Ohnishi,Toward solving the sign problem with path optimization method,Phys. Rev. D96 (2017) 111501 [1705.05605]

    2017 arXiv

  41. [49]

    Alexandru, P.F

    A. Alexandru, P.F. Bedaque, H. Lamm and S. Lawrence,Finite-Density Monte Carlo Calculations on Sign-Optimized Manifolds,Phys. Rev. D97 (2018) 094510 [1804.00697]

    2018 arXiv

  42. [50]

    Alexandru, P.F

    A. Alexandru, P.F. Bedaque, H. Lamm, S. Lawrence and N.C. Warrington,Fermions at Finite Density in 2+1 Dimensions with Sign-Optimized Manifolds,Phys. Rev. Lett.121 (2018) 191602 [1808.09799]

    2018 arXiv

  43. [51]

    Gäntgen, E

    C. Gäntgen, E. Berkowitz, T. Luu, J. Ostmeyer and M. Rodekamp,Fermionic sign problem minimization by constant path integral contour shifts, Phys. Rev. B109 (2024) 195158 [2307.06785]

    2024 arXiv

  44. [52]

    Bender, N

    C.M. Bender, N. Hassanpour, S.P. Klevansky and S. Sarkar,𝑃𝑇-symmetric quantum field theory in𝐷 dimensions, Phys. Rev. D98 (2018) 125003 [1810.12479]

    2018 arXiv

  45. [53]

    Romatschke,On the negative coupling O(N) model in 2d at high temperature, 2412.10496

    P. Romatschke,On the negative coupling O(N) model in 2d at high temperature, 2412.10496

    arXiv

  46. [54]

    Lawrence, R

    S. Lawrence, R. Weller, C. Peterson and P. Romatschke,Instantons, analytic continuation, and PT-symmetric field theory, Phys. Rev. D108 (2023) 085013 [2303.01470]

    2023 arXiv

  47. [55]

    Weller,Can negative bare couplings make sense? The®𝜙4 theory at large𝑁, 2310.02516

    R.D. Weller,Can negative bare couplings make sense? The®𝜙4 theory at large𝑁, 2310.02516

  48. [56]

    Tulipant, M

    Z. Tulipant, M. Giordano, K. Kapas, S.D. Katz and A. Pasztor,Exponential improvement of the sign problem via contour deformations in the 2+1D XY model at non-zero density,Phys. Rev. D106 (2022) 054512 [2202.07561]

    2022 arXiv

  49. [57]

    Kashiwa and Y

    K. Kashiwa and Y. Mori,Path optimization for𝑈(1) gauge theory with complexified parameters, Phys. Rev. D102 (2020) 054519 [2007.04167]

    2020 arXiv

  50. [58]

    Basar and J

    G. Basar and J. Marincel,Sign optimization and complex saddle points in one-dimensional QCD, Phys. Rev. D106 (2022) L091503 [2208.02072]. 17 Machine-learning approaches to accelerating lattice simulations Scott Lawrence

    2022 arXiv

  51. [59]

    Basar and J

    G. Basar and J. Marincel,Heavy-dense QCD, sign optimization, and Lefschetz thimbles, Phys. Rev. C109(2024) 045208 [2311.06343]

    2024 arXiv

  52. [60]

    Namekawa, K

    Y. Namekawa, K. Kashiwa, A. Ohnishi and H. Takase,Gauge invariant input to neural network for path optimization method,Phys. Rev. D105(2022) 034502 [2109.11710]

    2022 arXiv

  53. [61]

    Giordano, A

    M. Giordano, A. Pasztor, D. Pesznyak and Z. Tulipant,Alleviating the sign problem in a chiral random matrix model with contour deformations,Phys. Rev. D108 (2023) 094507 [2301.12947]

    2023 arXiv

  54. [62]

    Alexandru, G

    A. Alexandru, G. Başar, P.F. Bedaque, H. Lamm and S. Lawrence,Finite Density𝑄𝐸𝐷 1+1 Near Lefschetz Thimbles,Phys. Rev. D98 (2018) 034506 [1807.02027]

    2018 arXiv

  55. [63]

    Kashiwa, Y

    K. Kashiwa, Y. Namekawa, A. Ohnishi and H. Takase,Application of the path optimization method to a discrete spin system, Phys. Rev. D108 (2023) 094504 [2309.06018]

    2023 arXiv

  56. [64]

    Warrington,Real-time spin systems from lattice field theory, JHEP 12(2023) 156 [2310.19761]

    N.C. Warrington,Real-time spin systems from lattice field theory, JHEP 12(2023) 156 [2310.19761]

    2023 arXiv

  57. [65]

    Mooney, J

    T.C. Mooney, J. Bringewatt, N.C. Warrington and L.T. Brady,Lefschetz thimble quantum Monte Carlo for spin systems, Phys. Rev. B106 (2022) 214416 [2110.10699]

    2022 arXiv

  58. [66]

    Alexandru, G

    A. Alexandru, G. Basar, P.F. Bedaque and G.W. Ridgway,Schwinger-Keldysh formalism on the lattice: A faster algorithm and its application to field theory,Phys. Rev. D95 (2017) 114501 [1704.06404]

    2017 arXiv

  59. [67]

    Alexandru, G

    A. Alexandru, G. Basar, P.F. Bedaque, S. Vartak and N.C. Warrington,Monte Carlo Study of Real Time Dynamics on the Lattice, Phys. Rev. Lett.117 (2016) 081602 [1605.08040]

    2016 arXiv

  60. [68]

    Lawrence and Y

    S. Lawrence and Y. Yamauchi,Normalizing Flows and the Real-Time Sign Problem,Phys. Rev. D103 (2021) 114509 [2101.05755]

    2021 arXiv

  61. [69]

    Kanwar and M.L

    G. Kanwar and M.L. Wagman,Real-time lattice gauge theory actions: Unitarity, convergence, and path integral contour deformations,Phys. Rev. D104 (2021) 014513 [2103.02602]

    2021 arXiv

  62. [70]

    Detmold, G

    W. Detmold, G. Kanwar, H. Lamm, M.L. Wagman and N.C. Warrington,Path integral contour deformations for observables in𝑆𝑈(𝑁) gauge theory,Phys. Rev. D103 (2021) 094517 [2101.12668]

    2021 arXiv

  63. [71]

    Detmold, G

    W. Detmold, G. Kanwar, M.L. Wagman and N.C. Warrington,Path integral contour deformations for noisy observables, Phys. Rev. D102 (2020) 014514 [2003.05914]

    2020 arXiv

  64. [72]

    Kanwar, A

    G. Kanwar, A. Lovato, N. Rocco and M. Wagman,Mitigating Green’s function Monte Carlo signal-to-noise problems using contour deformations,Phys. Rev. C109(2024) 034317 [2304.03229]

    2024 arXiv

  65. [73]

    Lawrence, S

    S. Lawrence, S. Valgushev, J. Xiao and Y. Yamauchi,Contour deformations for non-holomorphic actions, 2401.16733. 18 Machine-learning approaches to accelerating lattice simulations Scott Lawrence

    arXiv

  66. [74]

    Lawrence and Y

    S. Lawrence and Y. Yamauchi,Convex optimization of contour deformations,Phys. Rev. D 110 (2024) 014508 [2311.13002]

    2024 arXiv

  67. [75]

    Abbott, A

    R. Abbott, A. Botev, D. Boyda, D.C. Hackett, G. Kanwar, S. Racanière et al.,Applications of flow models to the generation of correlated lattice QCD ensembles,Phys. Rev. D109 (2024) 094514 [2401.10874]

    2024 arXiv

  68. [76]

    Lawrence and Y

    S. Lawrence and Y. Yamauchi,Deep learning of fermion sign fluctuations, Phys. Rev. D107 (2023) 114505 [2212.14606]

    2023 arXiv

  69. [77]

    Lawrence,Perturbative Removal of a Sign Problem, Phys

    S. Lawrence,Perturbative Removal of a Sign Problem, Phys. Rev. D102 (2020) 094504 [2009.10901]

    2020 arXiv

  70. [78]

    Lawrence and Y

    S. Lawrence and Y. Yamauchi,Mitigating a discrete sign problem with extreme learning machines, 2312.12636

    arXiv

  71. [79]

    Bhattacharya, S

    T. Bhattacharya, S. Lawrence and J.-S. Yoo,Control variates for lattice field theory,Phys. Rev. D109 (2024) L031505 [2307.14950]

    2024 arXiv

  72. [80]

    Bedaque and H

    P.F. Bedaque and H. Oh,Leveraging neural control variates for enhanced precision in lattice field theory,Phys. Rev. D109(2024) 094519 [2312.08228]

    2024 arXiv

  73. [81]

    Lawrence,Schwinger-Dyson control variates for lattice fermions, 2404.10707

    S. Lawrence,Schwinger-Dyson control variates for lattice fermions, 2404.10707

    arXiv

  74. [82]

    R.Zhang,Z.Fan,R.Li,H.-W.LinandB.Yoon, Machine-learningpredictionforquasiparton distribution function matrix elements,Phys. Rev. D101 (2020) 034516 [1909.10990]

    2020 arXiv

  75. [83]

    B. Yoon, T. Bhattacharya and R. Gupta,Machine Learning Estimators for Lattice QCD Observables, Phys. Rev. D100 (2019) 014504 [1807.05971]

    2019 arXiv

  76. [84]

    J. Kim, G. Pederiva and A. Shindler,Machine learning mapping of lattice correlated data, Phys. Lett. B856 (2024) 138894 [2402.07450]

    2024 arXiv

  77. [85]

    Albergo, D

    M.S. Albergo, D. Boyda, D.C. Hackett, G. Kanwar, K. Cranmer, S. Racanière et al., Introduction to Normalizing Flows for Lattice Field Theory, 2101.08176

    arXiv

  78. [86]

    Tomiya and S

    A. Tomiya and S. Terasaki,GomalizingFlow.jl: A Julia package for Flow-based sampling algorithm for lattice field theory, 2208.08903

    arXiv

  79. [87]

    Nicoli, C.J

    K.A. Nicoli, C.J. Anders, L. Funcke, K. Jansen, S. Nakajima and P. Kessel,NeuLat: a toolbox for neural samplers in lattice field theories, PoS LATTICE2023(2024) 286. 19

    2024

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.