pith. sign in

arxiv: 2605.27231 · v1 · pith:NDDVL4KMnew · submitted 2026-05-26 · 🌌 astro-ph.CO · gr-qc· physics.comp-ph

A Fast Method to Compute Scalar Induced Gravitational Waves on a Lattice with Primordial Non-Gaussianities

Giovanni Piccoli This is my paper

Pith reviewed 2026-06-29 15:34 UTC · model grok-4.3

classification 🌌 astro-ph.CO gr-qcphysics.comp-ph
keywords scalar induced gravitational wavesprimordial non-Gaussianitieslattice computationfast Fourier transformconvolutionsgravitational wave spectrumradiation-dominated background
0
0 comments X

The pith

Recasting Fourier integrals as convolutions allows fast lattice computation of non-Gaussian scalar induced gravitational waves.

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

The paper develops a method to compute the spectrum of scalar induced gravitational waves that incorporates arbitrary primordial non-Gaussianities. Direct solution of the Fourier integrals arising from semi-analytic wave-equation solutions is intractable on a lattice. The integrals are rewritten as a sum of roughly fifty convolutions that can each be evaluated with fast Fourier transform techniques. The resulting power spectrum is extracted directly from the lattice field realization, and the procedure is realized in a GPU code that finishes in seconds while staying within ten percent error for a radiation-dominated background.

Core claim

Solving the wave equation semi-analytically produces Fourier integrals that cannot be evaluated directly; these integrals are recast as a sum of about fifty convolutions, each computed efficiently by FFT methods, so that the gravitational-wave power spectrum can be measured from the lattice realization and thereby include fully non-perturbative non-Gaussian effects.

What carries the argument

The representation of the wave-equation Fourier integrals as a sum of about fifty convolutions, each evaluated by FFT on the lattice.

If this is right

  • Arbitrary primordial non-Gaussian statistics can be included in SIGW spectrum calculations.
  • The computation finishes in seconds on modest GPU hardware.
  • The measured spectrum stays within ten percent of the expected result under the tested conditions.
  • A public implementation is provided for repeated use on different initial conditions.

Where Pith is reading between the lines

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

  • The same convolution reduction could be applied to backgrounds with equations of state other than w = 1/3.
  • Specific inflationary models that predict particular non-Gaussian shapes could be confronted directly with the lattice output.
  • The method supplies a practical route to test whether non-Gaussian corrections alter the amplitude or shape of the SIGW spectrum at observable frequencies.

Load-bearing premise

The semi-analytic wave-equation solutions remain accurate enough when their Fourier integrals are rewritten as convolutions.

What would settle it

Compute the spectrum for a simple non-Gaussian distribution where the original integrals can be evaluated by direct numerical quadrature and check whether the lattice result agrees within ten percent.

Figures

Figures reproduced from arXiv: 2605.27231 by Giovanni Piccoli.

Figure 1
Figure 1. Figure 1: FIG. 1. We show the relevant dynamical range of the kernels [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. We show the absolute value of the eigenvalues (nor [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. We compare the kernels with their reconstruction. We first adopted a grid built with [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. Gravitational wave spectrum induced by a Gaus [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. Gravitational wave spectrum induced by a non-Gaussian primordial curvature perturbation, expressed in terms of [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Gravitational wave spectrum induced by a non-Gaussian primordial curvature perturbation, expressed in terms of a [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
read the original abstract

Scalar Induced Gravitational Waves (SIGW) are generated at second order in perturbation theory and to achieve observational relevance, inflationary dynamics must evade the standard slow-roll scenario at small scales, generating large curvature perturbations following strongly non-Gaussian statistics. We propose a method to efficiently compute the SIGW spectrum including arbitrary non-Gaussianities. First, we solve the wave equation adopting semi-analytic methods; this results in an expression involving integrals in Fourier space which are impossible to solve directly on a lattice. We overcome this bottleneck by recasting these integrals as a sum of about 50 convolutions, each of which can be computed efficiently with FFT methods. Finally, the power spectrum is measured directly from the lattice realization. We implement this in FLAN-SIGW, a GPU-accelerated code capable of computing fully non-perturbative, non-Gaussian SIGW spectra in seconds with an error within 10% with modest computational resources. The code is made public at https://github.com/giovannipiccoli99/FLAN-SIGW. In this first implementation, in order to assess the performance of the method, we adopt a standard radiation-dominated background with $w = 1/3$.

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 / 2 minor

Summary. The paper proposes an efficient lattice-based method to compute scalar-induced gravitational wave (SIGW) spectra that incorporate arbitrary primordial non-Gaussianities. It solves the wave equation semi-analytically to obtain Fourier-space integrals, recasts those integrals as a sum of approximately 50 convolutions that can be evaluated via FFTs, and measures the resulting power spectrum directly from the lattice realization. The approach is implemented in the open-source GPU-accelerated FLAN-SIGW code, which is reported to produce results in seconds with an error within 10% for a radiation-dominated background (w=1/3).

Significance. If validated, the method would enable rapid, non-perturbative calculations of SIGW spectra for general non-Gaussian primordial curvature perturbations, which is relevant for connecting small-scale inflationary dynamics to potential observational signals. The public release of the code is a clear positive.

major comments (2)
  1. [method section describing the integral-to-convolution recasting (likely §3)] The central accuracy claim rests on recasting the semi-analytic Fourier integrals exactly as a finite sum of ~50 convolutions. The manuscript must supply the explicit decomposition (including any truncation), a derivation of the error bound, and direct numerical cross-checks against brute-force integration or known analytic limits for non-Gaussian cases; without these, the <10% error statement for arbitrary non-Gaussianities remains unverified and load-bearing for the efficiency-accuracy tradeoff.
  2. [§2 (wave-equation solution) and abstract] The abstract states that the results are 'fully non-perturbative' while simultaneously relying on semi-analytic solutions of the wave equation. The text should clarify precisely which aspects are treated non-perturbatively versus which retain perturbative or background assumptions, and demonstrate that the semi-analytic step does not reintroduce uncontrolled errors when non-Gaussianities are strong.
minor comments (2)
  1. [abstract and code-description section] The abstract mentions 'modest computational resources' but provides no concrete benchmarks (grid size, GPU model, wall-clock time, memory footprint); these should be added for reproducibility.
  2. [early theory section] Notation for the primordial non-Gaussianity (e.g., how the bispectrum or higher correlators enter the source term) should be defined explicitly before the convolution step.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed report. We address the two major comments below and will revise the manuscript accordingly to improve clarity and provide the requested supporting material.

read point-by-point responses

  1. Referee: [method section describing the integral-to-convolution recasting (likely §3)] The central accuracy claim rests on recasting the semi-analytic Fourier integrals exactly as a finite sum of ~50 convolutions. The manuscript must supply the explicit decomposition (including any truncation), a derivation of the error bound, and direct numerical cross-checks against brute-force integration or known analytic limits for non-Gaussian cases; without these, the <10% error statement for arbitrary non-Gaussianities remains unverified and load-bearing for the efficiency-accuracy tradeoff.

    Authors: We agree that the explicit decomposition, truncation details, error bound derivation, and numerical validation are necessary to substantiate the accuracy claim. In the revised version we will expand §3 to include the complete algebraic decomposition of the Fourier integrals into the finite sum of convolutions, the precise truncation criterion employed, and a derivation of the resulting error bound. We will also add direct numerical benchmarks comparing the FFT-based results against brute-force quadrature for representative non-Gaussian templates (including analytic limits where available) to confirm the reported ~10% accuracy level. revision: yes

  2. Referee: [§2 (wave-equation solution) and abstract] The abstract states that the results are 'fully non-perturbative' while simultaneously relying on semi-analytic solutions of the wave equation. The text should clarify precisely which aspects are treated non-perturbatively versus which retain perturbative or background assumptions, and demonstrate that the semi-analytic step does not reintroduce uncontrolled errors when non-Gaussianities are strong.

    Authors: The phrase 'fully non-perturbative' is intended to indicate that the primordial curvature perturbations are treated without any perturbative expansion in the non-Gaussian parameters; the lattice realization incorporates the complete non-Gaussian statistics. The wave equation itself is solved under the standard fixed-background approximation (radiation-dominated, w=1/3) that is common to all SIGW calculations. The semi-analytic Green's function is exact for that background and does not depend on the amplitude or shape of the non-Gaussianities. In the revision we will (i) rephrase the abstract and §2 to make this distinction explicit, (ii) add a short paragraph demonstrating that the semi-analytic step remains controlled for arbitrarily strong non-Gaussianities because the source term is evaluated exactly on the lattice, and (iii) note that the only background assumption is the fixed equation-of-state parameter, which is stated in the abstract. revision: yes

Circularity Check

0 steps flagged

Numerical method for SIGW spectra exhibits no circularity

full rationale

The paper describes a computational procedure: semi-analytic solution of the wave equation yields Fourier integrals that are exactly recast as a finite sum of convolutions evaluable via FFT on a lattice, followed by direct power-spectrum measurement. This is an algorithmic equivalence and implementation detail rather than a theoretical derivation in which a prediction is forced by its own inputs or by self-citation. No fitted parameters are introduced, no uniqueness theorems are invoked, and no load-bearing claim reduces to a prior result by the same authors. The stated 10% error bound is an empirical performance claim for the chosen radiation-dominated background, not a self-referential prediction. The derivation chain is therefore self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The method rests on the domain assumption of a radiation-dominated background and on the validity of the semi-analytic wave solutions; no free parameters or new physical entities are introduced in the abstract.

axioms (1)
  • domain assumption Standard radiation-dominated background with equation-of-state parameter w = 1/3
    Explicitly adopted to assess performance of the method.

pith-pipeline@v0.9.1-grok · 5748 in / 1106 out tokens · 29730 ms · 2026-06-29T15:34:04.275679+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. Matching second-order classical and 1-loop quantum tensor power spectra in de Sitter spacetime

    hep-ph 2026-05 unverdicted novelty 5.0

    Classical part of 1-loop tensor power spectrum in de Sitter is IR divergent but cancels with vacuum part, enabling non-perturbative renormalization to extract unaffected physical information.

Reference graph

Works this paper leans on

58 extracted references · 41 canonical work pages · cited by 1 Pith paper · 21 internal anchors

  1. [1]

    Ap = 0 simply leads toϱ= 1, which corresponds to a uniform discretization of theu−grid

    We start by considering the (yet un-normalized) den- sity of points to be uniform plus a bump centered over a pivotal point of interestu p: ϱ(u) = dn du = 1 +A p exp − (u−u p)2 2σ2p ; (59) written in terms of an amplitudeA p controlling the enhancement in density of sampling points around the pivot, and the relative width of said enhancement,σ p. Ap = 0 s...

  2. [2]

    All the lattice computations are performed decomposing each kernelN modes = 100, usingN α = 50 in the reconstruction step

    is omitted for clarity. All the lattice computations are performed decomposing each kernelN modes = 100, usingN α = 50 in the reconstruction step. A. Gaussian Initial Conditions As a first test, we consider the case where Φ i,k is a Gaussian field, to be able to compare the outcome of the numerical simulation with a relatively simple semi- analytic expres...

  3. [3]

    A. H. Guth, Phys. Rev. D23, 347 (1981)

    1981

  4. [4]

    A. A. Starobinsky, Phys. Lett. B91, 99 (1980)

    1980

  5. [5]

    V. F. Mukhanov and G. V. Chibisov, JETP Lett.33, 532 (1981)

    1981

  6. [6]

    The Atacama Cosmology Telescope: DR6 Constraints on Extended Cosmological Models

    E. Calabreseet al.(Atacama Cosmology Telescope), JCAP11, 063 (2025), arXiv:2503.14454 [astro-ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  7. [7]

    Tomita, Prog

    K. Tomita, Prog. Theor. Phys.37, 831 (1967)

    1967

  8. [8]

    Matarrese, O

    S. Matarrese, O. Pantano, and D. Saez, Phys. Rev. D 13 47, 1311 (1993)

    1993

  9. [9]

    General Relativistic Dynamics of Irrotational Dust: Cosmological Implications

    S. Matarrese, O. Pantano, and D. Saez, Phys. Rev. Lett. 72, 320 (1994), arXiv:astro-ph/9310036

    work page internal anchor Pith review Pith/arXiv arXiv 1994

  10. [10]

    Relativistic second-order perturbations of the Einstein-de Sitter Universe

    S. Matarrese, S. Mollerach, and M. Bruni, Phys. Rev. D 58, 043504 (1998), arXiv:astro-ph/9707278

    work page internal anchor Pith review Pith/arXiv arXiv 1998

  11. [11]

    K. N. Ananda, C. Clarkson, and D. Wands, Phys. Rev. D75, 123518 (2007), arXiv:gr-qc/0612013

    work page internal anchor Pith review Pith/arXiv arXiv 2007

  12. [12]

    Baumann, K

    D. Baumann, K. Ichiki, P. J. Steinhardt, and K. Taka- hashi, Phys. Rev. D76, 084019 (2007), arXiv:hep- th/0703290

    work page arXiv 2007

  13. [13]

    Domenech, Universe7, 398 (2021), arXiv:2109.02581

    G. Domenech, Universe7, 398 (2021), arXiv:2109.02581

    work page arXiv 2021

  14. [14]

    Tristramet al., Astron

    M. Tristramet al., Astron. Astrophys.647, A128 (2021), arXiv:2010.01139 [astro-ph.CO]

    work page arXiv 2021

  15. [15]

    Probing the inflaton: Small-scale power spectrum constraints from measurements of the CMB energy spectrum

    J. Chluba, A. L. Erickcek, and I. Ben-Dayan, Astrophys. J.758, 76 (2012), arXiv:1203.2681 [astro-ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2012

  16. [16]

    Shedding light on the small-scale crisis with CMB spectral distortions

    T. Nakama, J. Chluba, and M. Kamionkowski, Phys. Rev. D95, 121302 (2017), arXiv:1703.10559 [astro- ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  17. [17]

    Science with CMB spectral distortions

    J. Chluba, in49th Rencontres de Moriond on Cosmology (2014) pp. 327–334, arXiv:1405.6938 [astro-ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  18. [18]

    Silk damping at a redshift of a billion: a new limit on small-scale adiabatic perturbations

    D. Jeong, J. Pradler, J. Chluba, and M. Kamionkowski, Phys. Rev. Lett.113, 061301 (2014), arXiv:1403.3697 [astro-ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2014

  19. [19]

    Probing small-scale non-Gaussianity from anisotropies in acoustic reheating

    A. Naruko, A. Ota, and M. Yamaguchi, JCAP05, 049 (2015), arXiv:1503.03722 [astro-ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  20. [20]

    Revisiting constraints on small scale perturbations from big-bang nucleosynthesis

    K. Inomata, M. Kawasaki, and Y. Tada, Phys. Rev. D 94, 043527 (2016), arXiv:1605.04646 [astro-ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  21. [21]

    Nakama, T

    T. Nakama, T. Suyama, and J. Yokoyama, Phys. Rev. Lett.113, 061302 (2014)

    2014

  22. [22]

    Probing the small-scale primordial power spectrum via relic neutrinos and acoustic reheating

    G. Piccoli, S. Vagnozzi, and J. Silk, (2026), arXiv:2605.11956 [hep-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  23. [23]

    Secondary isocurvature perturbations from acoustic reheating

    A. Ota and M. Yamaguchi, JCAP06, 022 (2018), arXiv:1705.05196 [astro-ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  24. [24]

    A. S. Josan, A. M. Green, and K. A. Malik, Phys. Rev. D79, 103520 (2009), arXiv:0903.3184 [astro-ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2009

  25. [25]

    Improved constraints on the primordial power spectrum at small scales from ultracompact minihalos

    T. Bringmann, P. Scott, and Y. Akrami, Phys. Rev. D 85, 125027 (2012), arXiv:1110.2484 [astro-ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2012

  26. [26]

    Ultra slow-roll inflation demystified

    K. Dimopoulos, Phys. Lett. B775, 262 (2017), arXiv:1707.05644 [hep-ph]

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  27. [27]

    The NANOGrav 15-year Data Set: Search for Signals from New Physics

    A. Afzalet al.(NANOGrav), Astrophys. J. Lett. 951, L11 (2023), [Erratum: Astrophys.J.Lett. 971, L27 (2024), Erratum: Astrophys.J. 971, L27 (2024)], arXiv:2306.16219 [astro-ph.HE]

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  28. [28]

    Antoniadiset al.(EPTA, InPTA), Astron

    J. Antoniadiset al.(EPTA, InPTA), Astron. Astrophys. 685, A94 (2024), arXiv:2306.16227 [astro-ph.CO]

    work page arXiv 2024

  29. [29]

    Ellis, M

    J. Ellis, M. Fairbairn, G. Franciolini, G. H¨ utsi, A. Iovino, M. Lewicki, M. Raidal, J. Urrutia, V. Vaskonen, and H. Veerm¨ ae, Phys. Rev. D109, 023522 (2024), arXiv:2308.08546 [astro-ph.CO]

    work page arXiv 2024

  30. [30]

    Balaji, G

    S. Balaji, G. Dom` enech, and G. Franciolini, JCAP10, 041 (2023), arXiv:2307.08552 [gr-qc]

    work page arXiv 2023

  31. [31]

    Cecchini, G

    C. Cecchini, G. Franciolini, and M. Pieroni, Phys. Rev. D111, 123536 (2025), arXiv:2503.10805 [astro-ph.CO]

    work page arXiv 2025

  32. [32]

    Auclairet al.(LISA Cosmology Working Group), Living Rev

    P. Auclairet al.(LISA Cosmology Working Group), Living Rev. Rel.26, 5 (2023), arXiv:2204.05434 [astro- ph.CO]

    work page arXiv 2023

  33. [33]

    J. E. Gammalet al.(LISA Cosmology Working Group), JCAP05, 062 (2025), arXiv:2501.11320 [astro-ph.CO]

    work page arXiv 2025

  34. [34]

    G. C. Wick, Phys. Rev.80, 268 (1950)

    1950

  35. [35]

    R.-g. Cai, S. Pi, and M. Sasaki, Phys. Rev. Lett.122, 201101 (2019), arXiv:1810.11000 [astro-ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  36. [36]

    Perna, C

    G. Perna, C. Testini, A. Ricciardone, and S. Matarrese, JCAP05, 086 (2024), arXiv:2403.06962 [astro-ph.CO]

    work page arXiv 2024

  37. [37]

    A. J. Iovino, S. Matarrese, G. Perna, A. Ricciardone, and A. Riotto, Phys. Lett. B872, 140039 (2026), arXiv:2412.06764 [astro-ph.CO]

    work page arXiv 2026

  38. [38]

    X.-X. Zeng, Z. Ning, R.-G. Cai, and S.-J. Wang, (2025), arXiv:2508.10812 [astro-ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  39. [39]

    Lattice simulations of scalar-induced gravitational waves from inflation

    A. Caravano, G. Franciolini, and S. Renaux-Petel, (2026), arXiv:2604.03628 [astro-ph.CO]

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  40. [40]

    J. H. P. Jackson, H. Assadullahi, A. D. Gow, K. Koyama, V. Vennin, and D. Wands, JCAP05, 053 (2024), arXiv:2311.03281 [astro-ph.CO]

    work page arXiv 2024

  41. [41]

    Zeng, JCAP03, 065 (2026), arXiv:2510.02106 [astro-ph.CO]

    X.-X. Zeng, JCAP03, 065 (2026), arXiv:2510.02106 [astro-ph.CO]

    work page arXiv 2026

  42. [42]

    A. A. Kugarajh, Class. Quant. Grav.42, 127001 (2025), arXiv:2503.00083 [gr-qc]

    work page arXiv 2025

  43. [43]

    Dodelson and F

    S. Dodelson and F. Schmidt,Modern Cosmology(Aca- demic Press, 2020)

    2020

  44. [44]

    V. F. Mukhanov,Physical Foundations of Cosmology (Cambridge University Press, Oxford, 2005)

    2005

  45. [45]

    E. A. Coddington and N. Levinson,Theory of Ordi- nary Differential Equations(McGraw-Hill Book Com- pany, New York, 1955)

    1955

  46. [46]

    Franciolini, D

    G. Franciolini, D. Racco, and F. Rompineve, Phys. Rev. Lett.132, 081001 (2024), [Erratum: Phys.Rev.Lett. 133, 189901 (2024)], arXiv:2306.17136 [astro-ph.CO]

    work page arXiv 2024

  47. [47]

    Escriv` a, R

    A. Escriv` a, R. Inui, Y. Tada, and C.-M. Yoo, Phys. Rev. D111, 023528 (2025), arXiv:2404.12591 [astro-ph.CO]

    work page arXiv 2025

  48. [48]

    T. Kite, J. Chluba, A. Ravenni, and S. P. Patil, Mon. Not. Roy. Astron. Soc.509, 1366 (2021), arXiv:2107.13351 [astro-ph.CO]

    work page arXiv 2021

  49. [49]

    L. T. Witkowski, (2022), arXiv:2209.05296 [astro- ph.CO]

    work page arXiv 2022

  50. [50]

    Semianalytic Calculation of Gravitational Wave Spectrum Nonlinearly Induced from Primordial Curvature Perturbations

    K. Kohri and T. Terada, Phys. Rev. D97, 123532 (2018), arXiv:1804.08577

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  51. [51]

    Fast fourier transform (fft),

    J. W. Cooley, “Fast fourier transform (fft),” inEncyclo- pedia of Computer Science(John Wiley and Sons Ltd., GBR, 2003) p. 695–698

    2003

  52. [52]

    Efron,The Jackknife, the Bootstrap and Other Resam- pling Plans(Society for Industrial and Applied Mathe- matics, Philadelphia, PA, 1982)

    B. Efron,The Jackknife, the Bootstrap and Other Resam- pling Plans(Society for Industrial and Applied Mathe- matics, Philadelphia, PA, 1982)

    1982

  53. [53]

    J. W. Gibbs, Nature59, 200 (1898)

  54. [54]

    PyTorch: An Imperative Style, High-Performance Deep Learning Library

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, A. Desmaison, A. K¨ opf, E. Yang, Z. DeVito, M. Rai- son, A. Tejani, S. Chilamkurthy, B. Steiner, L. Fang, J. Bai, and S. Chintala, “Pytorch: An imperative style, high-performance deep learning library,” (2019), arXiv:1912.01703 [cs.LG]

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  55. [55]

    Briaud, R

    V. Briaud, R. Kawaguchi, and V. Vennin, JCAP12, 024 (2025), arXiv:2509.05124 [astro-ph.CO]

    work page arXiv 2025

  56. [56]

    S. M. Ahmadi and N. Ahmadi, (2026), arXiv:2602.00902 [gr-qc]

    work page arXiv 2026

  57. [57]

    J. D. Hunter, Computing in Science & Engineering9, 90 (2007)

    2007

  58. [58]

    Astrojacobli/smplotlib: v0.0.9,

    Jiaxuan Li, “Astrojacobli/smplotlib: v0.0.9,” (2023)

    2023