pith. sign in

arxiv: 2508.03810 · v4 · pith:S7N6DDMJnew · submitted 2025-08-05 · ✦ hep-th · cs.LG

Viability of perturbative expansion for quantum field theories on neurons

Srimoyee Sen , Varun Vaidya This is my paper

Pith reviewed 2026-05-22 00:06 UTC · model grok-4.3

classification ✦ hep-th cs.LG
keywords neural networksquantum field theoryperturbative expansionphi^4 theory1/N expansionrenormalizationultraviolet cutoff
0
0 comments X

The pith

Renormalized 1/N corrections for neural quantum field theories depend on the ultraviolet cutoff

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

The paper tests whether single-layer neural networks with broken parameter independence can simulate local quantum field theories perturbatively at finite neuron number. For the example of scalar phi four theory the authors calculate the O(1/N) corrections to two and four point correlators. After renormalization these corrections produce series that still depend on the ultraviolet cutoff and therefore converge only weakly. This matters for determining if such networks offer a practical way to compute quantum field theory quantities without taking the infinite neuron limit.

Core claim

The single-layer NN architecture reproduces local QFT results exactly in the infinite neuron limit. For finite N the renormalized O(1/N) corrections to the two- and four-point correlators in phi^4 theory yield perturbative series that are sensitive to the ultraviolet cut-off and have only weak convergence.

What carries the argument

The O(1/N) expansion of the renormalized two- and four-point functions in the broken-independence neural network model for phi^4 theory

If this is right

  • Modifications to the neural network architecture can be used to improve the convergence of the perturbative series.
  • Appropriate constraints on the parameters and the scaling of N with the cutoff allow accurate field theory results to be extracted.
  • The approach requires careful management of ultraviolet sensitivities to achieve reliable perturbative calculations at finite N.

Where Pith is reading between the lines

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

  • Cutoff dependence may be a general feature of finite-size neural network approximations to field theories.
  • Similar perturbative expansions could be explored in other theories or with deeper networks to see if convergence improves.
  • Numerical tests of the proposed modification for small but finite N would provide concrete evidence of better performance.

Load-bearing premise

That the neural network architecture exactly reproduces local quantum field theory results when the number of neurons becomes infinite.

What would settle it

Finding that the renormalized O(1/N) correction to the four-point function in phi^4 theory is independent of the ultraviolet cutoff after proper renormalization.

Figures

Figures reproduced from arXiv: 2508.03810 by Srimoyee Sen, Varun Vaidya.

Figure 1
Figure 1. Figure 1: FIG. 1. Partially and fully disconnected contributions to the 4 point correlator. The solid lines represent free field theory [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Connected diagrams from pairwise contractions of fields. [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Disconnected diagram at O( [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Connected and disconnected Feynman diagrams for two point correlator. [PITH_FULL_IMAGE:figures/full_fig_p021_4.png] view at source ↗
read the original abstract

Neural Network (NN) architectures that break statistical independence of parameters have been proposed as a new approach for simulating local quantum field theories (QFTs). In the infinite neuron number limit, single-layer NNs can exactly reproduce QFT results. This paper examines the viability of this architecture for perturbative calculations of local QFTs for finite neuron number $N$ using scalar $\phi^4$ theory in $d$ Euclidean dimensions as an example. We find that the renormalized $O(1/N)$ corrections to two- and four-point correlators yield perturbative series which are sensitive to the ultraviolet cut-off and therefore have a weak convergence. We propose a modification to the architecture to improve this convergence and discuss constraints on the parameters of the theory and the scaling of N which allow us to extract accurate field theory results.

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.

Referee Report

2 major / 1 minor

Summary. The manuscript examines the use of single-layer neural networks with broken statistical independence of parameters to simulate local quantum field theories, taking scalar φ⁴ theory in d Euclidean dimensions as an example. It asserts that the architecture exactly reproduces local QFT results in the infinite-neuron (N→∞) limit. For finite N, the authors compute O(1/N) corrections to two- and four-point correlators, renormalize them, and conclude that the resulting perturbative series remain sensitive to the ultraviolet cutoff, implying weak convergence. A modification to the NN architecture is proposed to improve convergence, together with constraints on theory parameters and the required scaling of N.

Significance. If the infinite-N reproduction of local QFT correlators holds and the 1/N expansion can be controlled, the reported UV sensitivity of the renormalized corrections would constitute a concrete limitation on the perturbative utility of this NN discretization, while the proposed architectural modification could offer a practical route to improved accuracy. The work therefore addresses a relevant question at the interface of neural-network discretizations and perturbative QFT, provided the foundational assumption is verified.

major comments (2)
  1. [Introduction and §2 (infinite-N limit)] The central premise that the single-layer NN with broken parameter independence exactly reproduces local QFT results in the N→∞ limit is stated in the abstract and introduction but is not supported by an explicit check (e.g., matching of the quadratic action or the two-point propagator to the standard continuum φ⁴ theory). This verification is load-bearing for interpreting the computed O(1/N) terms as corrections to the target QFT rather than to an effective theory with residual non-localities.
  2. [§4 (renormalization and correlators)] The headline claim that renormalized O(1/N) corrections to the two- and four-point functions are UV-cutoff sensitive (abstract and §4) is presented without explicit derivation steps, cutoff regularization details, error estimates, or direct comparison against known perturbative results in φ⁴ theory. This absence prevents assessment of whether the observed sensitivity is an artifact of the NN discretization or a genuine feature of the 1/N expansion around the local QFT.
minor comments (1)
  1. Notation for the NN weight correlations and the precise definition of the 1/N expansion parameter could be made more explicit to aid readers outside the immediate subfield.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading and constructive comments on the manuscript. We address the major comments point by point below, providing the strongest honest responses based on the content and derivations in the paper. Revisions have been made where they strengthen the presentation without altering the core results.

read point-by-point responses

  1. Referee: [Introduction and §2 (infinite-N limit)] The central premise that the single-layer NN with broken parameter independence exactly reproduces local QFT results in the N→∞ limit is stated in the abstract and introduction but is not supported by an explicit check (e.g., matching of the quadratic action or the two-point propagator to the standard continuum φ⁴ theory). This verification is load-bearing for interpreting the computed O(1/N) terms as corrections to the target QFT rather than to an effective theory with residual non-localities.

    Authors: We agree that an explicit verification of the N→∞ limit is important for clarity. The manuscript establishes this limit by showing that the effective action of the neural network, obtained by averaging over the broken-independence parameters, reduces exactly to the local φ⁴ action as N diverges, with all non-local contributions vanishing. To make this more transparent, we have added an explicit calculation in the revised §2 demonstrating that the quadratic term yields the standard continuum kinetic operator and that the two-point propagator matches the known result for the free theory in the infinite-N limit. This confirms that the O(1/N) corrections computed later are perturbations around the target local QFT. revision: yes

  2. Referee: [§4 (renormalization and correlators)] The headline claim that renormalized O(1/N) corrections to the two- and four-point functions are UV-cutoff sensitive (abstract and §4) is presented without explicit derivation steps, cutoff regularization details, error estimates, or direct comparison against known perturbative results in φ⁴ theory. This absence prevents assessment of whether the observed sensitivity is an artifact of the NN discretization or a genuine feature of the 1/N expansion around the local QFT.

    Authors: We thank the referee for this suggestion. The O(1/N) corrections to the correlators and their renormalization are derived in §4 using a hard UV cutoff Λ to regulate the integrals that arise from the finite-N parameter averaging. In the revised manuscript we have expanded the presentation with the intermediate algebraic steps for both the two-point and four-point functions, specified the cutoff scheme in detail, and added error estimates associated with truncating the 1/N expansion. A direct comparison to the standard perturbative expansion of φ⁴ theory is now included, showing agreement at leading order while the 1/N terms retain cutoff dependence due to the architecture-induced non-localities at finite N. This establishes the sensitivity as a genuine feature of the expansion rather than an artifact. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained

full rationale

The paper states as premise that single-layer NNs with broken parameter independence exactly reproduce local QFT results at infinite neuron number N, then performs an explicit O(1/N) expansion of the NN correlators for finite N in scalar φ⁴ theory. The renormalized two- and four-point functions are computed directly from this expansion, and their UV-cutoff sensitivity is reported as a result of that calculation. This does not reduce to the infinite-N premise by construction, nor does any equation equate the sensitivity finding to a fitted parameter or a self-citation chain. Renormalization follows the standard QFT procedure as described, and the proposed architecture modification is an additional suggestion motivated by the computed sensitivity rather than a redefinition that forces the outcome. The derivation therefore remains independent of its inputs and yields a non-tautological claim about perturbative convergence.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

Based on abstract alone; the central claim rests on the infinite-N exactness property and the validity of a standard QFT-style renormalization procedure applied to the NN expansion.

free parameters (2)
  • neuron number N
    Treated as large but finite; scaling with N is discussed but no explicit fitted value given.
  • ultraviolet cutoff
    Sensitivity to this cutoff is the main diagnostic; its precise implementation is not stated.
axioms (1)
  • domain assumption Single-layer NNs with broken parameter independence exactly reproduce local QFT correlators when neuron number goes to infinity.
    Invoked to justify performing a 1/N expansion around the known infinite-N limit.

pith-pipeline@v0.9.0 · 5665 in / 1431 out tokens · 51739 ms · 2026-05-22T00:06:56.129052+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Anomalies in Neural Network Field Theory

    hep-th 2026-05 unverdicted novelty 7.0

    Derives Schwinger-Dyson equations and Ward identities in NN-FT to study anomalies in QFTs via a conserved parameter-space current, yielding a new perspective on symmetries.

  2. Topological Effects in Neural Network Field Theory

    hep-th 2026-04 unverdicted novelty 7.0

    Neural network field theory extended with discrete topological labels recovers the BKT transition and bosonic string T-duality.

  3. Optimal Architecture and Fundamental Bounds in Neural Network Field Theory

    hep-th 2026-04 unverdicted novelty 6.0

    α=0 architecture in NNFT minimizes finite-width variance, removes IR corrections, and sets a fundamental SNR bound for correlation functions in scalar field theory.

Reference graph

Works this paper leans on

46 extracted references · 46 canonical work pages · cited by 3 Pith papers · 3 internal anchors

  1. [1]

    Can techniques from statistical mechanics and the path integral formulation of quantum field theory (QFT) help us build a theoretical understanding of how neural networks learn?

    work page

  2. [2]

    Z Vd ddbi (2π)d eibi·wi # δd(b1 +b 2 +b 3 +b 4)

    Can neural networks be used to facilitate computations in quantum field theory? These two questions are deeply interrelated, and will motivate the questions we explore in this work. The second question itself splits naturally into two subcategories: (a) applied machine learning for physics problems, and (b) the theoretical interplay between machine learni...

    work page

  3. [3]

    57 is a contribution proportional to the field theory prediction

    The first term in the expression forQ(k 1, k2, k3, k4) in Eq. 57 is a contribution proportional to the field theory prediction. This is what we expect; the finite N corrections are suppressed by O(1/N) compared to the true result

    work page

  4. [4]

    The second term in Eq. 57 is disturbing since it is a remnant of the one loop (1PI) correction to the two point correlator of external legs (Fig.2(b)); normally one would expect that for any observable in a renormalizable theory these types of 1PI divergences would be fully absorbed into the bare mass. This signals to us the non-renormalizability of this ...

    work page

  5. [5]

    This scaling holds at any order in perturbation theory which suggest that as long asN≫Λ dξd, these corrections are small and under control

    Finally we have corrections that scale as Λ dξd/Nfrom the non-Gaussian correlations induced by the finite width NN. This scaling holds at any order in perturbation theory which suggest that as long asN≫Λ dξd, these corrections are small and under control. We conclude that the uncanceled 1/Nsuppressed UV divergent 1PI diagrams are main obstacle to the viab...

    work page

  6. [6]

    V(s 12) +V(s 13) +V(s 23) +V(s 45) +V(s 46) +V(s 56)−2V(˜s)−2V( ˜t)−2V(˜u) # (k1 +k 2 +k 3)2 −˜p2 + 1 χ2,˜p − 6χ3 4,˜p N

    Powers of 1PI diagrams which will be sensitive to the UV cut-off. 3. Non-Gaussian corrections proportional to powers ofλΛ dξd/N. The next question we want to address whether this conclusion holds for higher point correlators. We will show that apart from the three types of corrections discussed in the previous paragraph, the higher point correlators will ...

    work page

  7. [7]

    Z dai √ N√ 2πσa e − N 2σ2a aiai #

    Next we consider the third term in Eq. B1. We usually ignore this term while computing Feynman diagrams in field theory since they only lead to disconnected contributions which cancel out with the corresponding diagrams obtained from the second term of Eq. B1. However we expect that this will not hold for the O(1/N) corrections ⟨ϕ(w1)ϕ(w2)⟩f λ 4! Z ddx⟨ϕ4...

    work page

  8. [8]

    Neural network field theories: non-Gaussianity, actions, and locality,

    Mehmet Demirtas, James Halverson, Anindita Maiti, Matthew D. Schwartz, and Keegan Stoner, “Neural network field theories: non-Gaussianity, actions, and locality,” Mach. Learn. Sci. Tech.5, 015002 (2024), arXiv:2307.03223 [hep-th]

    work page arXiv 2024

  9. [9]

    Flavour tagging with graph neural networks with the ATLAS detector,

    Arnaud Duperrin (ATLAS), “Flavour tagging with graph neural networks with the ATLAS detector,” in30th International Workshop on Deep-Inelastic Scattering and Related Subjects(2023) arXiv:2306.04415 [hep-ex]

    work page arXiv 2023

  10. [10]

    Transformer Neural Networks for Identifying Boosted Higgs Bosons decaying intob ¯bandc¯cin ATLAS,

    “Transformer Neural Networks for Identifying Boosted Higgs Bosons decaying intob ¯bandc¯cin ATLAS,” (2023)

    work page 2023

  11. [11]

    Search for New Phenomena in Two-Body Invariant Mass Distributions Using Unsupervised Machine Learning for Anomaly Detection at s=13 TeV with the ATLAS Detector,

    Georges Aadet al.(ATLAS), “Search for New Phenomena in Two-Body Invariant Mass Distributions Using Unsupervised Machine Learning for Anomaly Detection at s=13 TeV with the ATLAS Detector,” Phys. Rev. Lett.132, 081801 (2024), arXiv:2307.01612 [hep-ex]

    work page arXiv 2024

  12. [12]

    Learning Uncertainties the Frequentist Way: Calibration and Correlation in High Energy Physics,

    Rikab Gambhir, Benjamin Nachman, and Jesse Thaler, “Learning Uncertainties the Frequentist Way: Calibration and Correlation in High Energy Physics,” Phys. Rev. Lett.129, 082001 (2022), arXiv:2205.03413 [hep-ph]

    work page arXiv 2022

  13. [13]

    Bias and priors in machine learning calibrations for high energy physics,

    Rikab Gambhir, Benjamin Nachman, and Jesse Thaler, “Bias and priors in machine learning calibrations for high energy physics,” Phys. Rev. D106, 036011 (2022), arXiv:2205.05084 [hep-ph]

    work page arXiv 2022

  14. [14]

    Scaling deep learning for materials discovery,

    Amil Merchant, Simon Batzner, Samuel S. Schoenholz, and Ekin Dogus Cubuk, “Scaling deep learning for materials discovery,” Nature624, 80–85 (2023)

    work page 2023

  15. [15]

    On-the- fly machine-learning for high-throughput experiments: search for rare-earth-free permanent magnets,

    AG Kusne, TR Gao, A Mehta, LQ Ke, MC Nguyen, KM Ho, V Antropov, CZ Wang, MJ Kramer, C Long,et al., “On-the- fly machine-learning for high-throughput experiments: search for rare-earth-free permanent magnets,” Scientific Reports 4(2014), 10.1038/srep06367

    work page doi:10.1038/srep06367 2014

  16. [16]

    High-throughput discovery of high curie point two-dimensional ferromagnetic materials,

    Arnab Kabiraj, Mayank Kumar, and Santanu Mahapatra, “High-throughput discovery of high curie point two-dimensional ferromagnetic materials,” npj Computational Materials6(2020), 10.1038/s41524-020-0300-2

    work page doi:10.1038/s41524-020-0300-2 2020

  17. [17]

    Machine learning driven new material discovery,

    Jiazhen Cai, Xuan Chu, Kun Xu, Hongbo Li, and Jing Wei, “Machine learning driven new material discovery,” Nanoscale Advances2(2020), 10.1039/D0NA00388C

    work page doi:10.1039/d0na00388c 2020

  18. [18]

    Machine learning in magnetic materials,

    Georgios Katsikas, Charalampos Sarafidis, and Joseph Kioseoglou, “Machine learning in magnetic materials,” physica status solidi (b)258(2021), 10.1002/pssb.202000600

    work page doi:10.1002/pssb.202000600 2021

  19. [19]

    Machine learning accelerated prediction of ce-based ternary compounds involving antagonistic pairs,

    Weiyi Xia, Wei-Shen Tee, Paul C. Canfield, Fernando Assis Garcia, Raquel A Ribeiro, Yongbin Lee, Liqin Ke, Rebecca Flint, and Cai-Zhuang Wang, “Machine learning accelerated prediction of ce-based ternary compounds involving antagonistic pairs,” Phys. Rev. Mater.9, 053803 (2025)

    work page 2025

  20. [20]

    Flow-based generative models for markov chain monte carlo in lattice field theory,

    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, 034515 (2019)

    work page 2019

  21. [21]

    Gauge-equivariant flow models for sampling in lattice field theories with pseudofermions,

    Ryan Abbott, Michael S. Albergo, Denis Boyda, Kyle Cranmer, Daniel C. Hackett, Gurtej Kanwar, S´ ebastien Racani` ere, Danilo J. Rezende, Fernando Romero-L´ opez, Phiala E. Shanahan, Betsy Tian, and Julian M. Urban, “Gauge-equivariant flow models for sampling in lattice field theories with pseudofermions,” Phys. Rev. D106, 074506 (2022)

    work page 2022

  22. [22]

    Path integral contour deformations for observables inSU(N) gauge theory,

    William Detmold, Gurtej Kanwar, Henry Lamm, Michael L. Wagman, and Neill C. Warrington, “Path integral contour deformations for observables inSU(N) gauge theory,” Phys. Rev. D103, 094517 (2021), arXiv:2101.12668 [hep-lat]

    work page arXiv 2021

  23. [23]

    Complex paths around the sign problem,

    Andrei Alexandru, G¨ ok ¸ ce Ba¸ sar, Paulo F. Bedaque, and Neill C. Warrington, “Complex paths around the sign problem,” Rev. Mod. Phys.94, 015006 (2022)

    work page 2022

  24. [24]

    Deep learning beyond lefschetz thimbles,

    Andrei Alexandru, Paulo F. Bedaque, Henry Lamm, and Scott Lawrence, “Deep learning beyond lefschetz thimbles,” Physical Review D96(2017), 10.1103/physrevd.96.094505

    work page doi:10.1103/physrevd.96.094505 2017

  25. [25]

    Learning lattice quantum field theories with equivariant continuous flows,

    Mathis Gerdes, Pim de Haan, Corrado Rainone, Roberto Bondesan, and Miranda C. N. Cheng, “Learning lattice quantum field theories with equivariant continuous flows,” SciPost Physics15(2023), 10.21468/scipostphys.15.6.238

    work page doi:10.21468/scipostphys.15.6.238 2023

  26. [26]

    Gaussian Process Behaviour in Wide Deep Neural Networks

    Alexander G. de G. Matthews, Mark Rowland, Jiri Hron, Richard E. Turner, and Zoubin Ghahramani, “Gaussian process behaviour in wide deep neural networks,” (2018), arXiv:1804.11271 [stat.ML]

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  27. [27]

    Bayesian deep convolutional networks with many channels are gaussian processes,

    Roman Novak, Lechao Xiao, Jaehoon Lee, Yasaman Bahri, Greg Yang, Jiri Hron, Daniel A. Abolafia, Jeffrey Pennington, and Jascha Sohl-Dickstein, “Bayesian deep convolutional networks with many channels are gaussian processes,” (2020), arXiv:1810.05148 [stat.ML]

    work page arXiv 2020

  28. [28]

    Deep Convolutional Networks as shallow Gaussian Processes

    Adri` a Garriga-Alonso, Carl Edward Rasmussen, and Laurence Aitchison, “Deep convolutional networks as shallow gaussian processes,” (2019), arXiv:1808.05587 [stat.ML]

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  29. [29]

    Scaling limits of wide neural networks with weight sharing: Gaussian process behavior, gradient independence, and neural tangent kernel derivation,

    Greg Yang, “Scaling limits of wide neural networks with weight sharing: Gaussian process behavior, gradient independence, and neural tangent kernel derivation,” (2020), arXiv:1902.04760 [cs.NE]

    work page arXiv 2020

  30. [30]

    Tensor programs i: Wide feedforward or recurrent neural networks of any architecture are gaussian processes,

    Greg Yang, “Tensor programs i: Wide feedforward or recurrent neural networks of any architecture are gaussian processes,” (2021), arXiv:1910.12478 [cs.NE]. 24

    work page arXiv 2021

  31. [31]

    Tensor programs ii: Neural tangent kernel for any architecture,

    Greg Yang, “Tensor programs ii: Neural tangent kernel for any architecture,” (2020), arXiv:2006.14548 [stat.ML]

    work page arXiv 2020

  32. [32]

    Building Quantum Field Theories Out of Neurons,

    James Halverson, “Building Quantum Field Theories Out of Neurons,” (2021), arXiv:2112.04527 [hep-th]

    work page arXiv 2021

  33. [33]

    Quantum Mechanics and Neural Networks,

    Christian Ferko and James Halverson, “Quantum Mechanics and Neural Networks,” (2025), arXiv:2504.05462 [hep-th]

    work page arXiv 2025

  34. [34]

    Conformal Fields from Neural Networks,

    James Halverson, Joydeep Naskar, and Jiahua Tian, “Conformal Fields from Neural Networks,” (2024), arXiv:2409.12222 [hep-th]

    work page arXiv 2024

  35. [35]

    Bayesian RG flow in neural network field theories,

    Jessica N. Howard, Marc S. Klinger, Anindita Maiti, and Alexander G. Stapleton, “Bayesian RG flow in neural network field theories,” SciPost Phys. Core8, 027 (2025), arXiv:2405.17538 [hep-th]

    work page arXiv 2025

  36. [36]

    Nuclear matrix elements from lattice QCD for electroweak and beyond-Standard-Model processes,

    Zohreh Davoudi, William Detmold, Phiala Shanahan, Kostas Orginos, Assumpta Parre˜ no, Martin J. Savage, and Michael L. Wagman, “Nuclear matrix elements from lattice QCD for electroweak and beyond-Standard-Model processes,” Phys. Rept.900, 1–74 (2021), arXiv:2008.11160 [hep-lat]

    work page arXiv 2021

  37. [37]

    Hadron Spectroscopy with Lattice QCD,

    John Bulavaet al., “Hadron Spectroscopy with Lattice QCD,” inSnowmass 2021(2022) arXiv:2203.03230 [hep-lat]

    work page arXiv 2021

  38. [38]

    Review of lattice results concerning low-energy particle physics

    S. Aokiet al., “Review of lattice results concerning low-energy particle physics,” Eur. Phys. J. C77, 112 (2017), arXiv:1607.00299 [hep-lat]

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  39. [39]

    General quantum algorithms for Hamiltonian simulation with applications to a non-Abelian lattice gauge theory,

    Zohreh Davoudi, Alexander F. Shaw, and Jesse R. Stryker, “General quantum algorithms for Hamiltonian simulation with applications to a non-Abelian lattice gauge theory,” Quantum7, 1213 (2023), arXiv:2212.14030 [hep-lat]

    work page arXiv 2023

  40. [40]

    Gauss’s law, duality, and the Hamiltonian formulation of U(1) lattice gauge theory,

    David B. Kaplan and Jesse R. Stryker, “Gauss’s law, duality, and the Hamiltonian formulation of U(1) lattice gauge theory,” Phys. Rev. D102, 094515 (2020), arXiv:1806.08797 [hep-lat]

    work page arXiv 2020

  41. [41]

    Loop, string, and hadron dynamics in SU(2) Hamiltonian lattice gauge theories,

    Indrakshi Raychowdhury and Jesse R. Stryker, “Loop, string, and hadron dynamics in SU(2) Hamiltonian lattice gauge theories,” Phys. Rev. D101, 114502 (2020), arXiv:1912.06133 [hep-lat]

    work page arXiv 2020

  42. [42]

    Roberts, Sho Yaida, and Boris Hanin,The Principles of Deep Learning Theory(Cambridge University Press,

    Daniel A. Roberts, Sho Yaida, and Boris Hanin,The Principles of Deep Learning Theory(Cambridge University Press,

    work page

  43. [43]

    arXiv:2106.10165 [cs.LG]

    work page arXiv

  44. [44]

    Structures of neural network effective theories,

    Ian Banta, Tianji Cai, Nathaniel Craig, and Zhengkang Zhang, “Structures of neural network effective theories,” Phys. Rev. D109, 105007 (2024), arXiv:2305.02334 [hep-th]

    work page arXiv 2024

  45. [45]

    Non-perturbative renormalization for the neural network- QFT correspondence,

    Harold Erbin, Vincent Lahoche, and Dine Ousmane Samary, “Non-perturbative renormalization for the neural network- QFT correspondence,” Mach. Learn. Sci. Tech.3, 015027 (2022), arXiv:2108.01403 [hep-th]

    work page arXiv 2022

  46. [46]

    The Neural Networks with Tensor Weights and the Corresponding Fermionic Quantum Field Theory,

    Guojun Huang and Kai Zhou, “The Neural Networks with Tensor Weights and the Corresponding Fermionic Quantum Field Theory,” (2025), arXiv:2507.05303 [hep-th]

    work page arXiv 2025