Pith. sign in

REVIEW 2 major objections 73 references

Compact and Stable Representation of Real-Frequency Spectral Functions for Machine Learning

T0 review · 2 major / 0 minor · reviewed 2026-07-14 · grok-4.5

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

Pith's one-line read A fixed sequence of Cayley-mapped trigonometric moments is a compact, positivity-preserving learning target for real-frequency spectral functions that matches or beats dense grids and avoids analytic continuation.

desk verdict Solid methods paper: Cayley-mapped trigonometric moments give a fixed-size, positivity-constrained real-frequency ML target that works on the reported DMFT and matrix-impurity benchmarks. read the letter →

arxiv 2607.11190 v1 pith:CJYHVC2A submitted 2026-07-13 physics.comp-ph

Xinyang Dong , Ido Zemach , Lei Zhang , Guy Cohen , Emanuel Gull This is my paper · ORCID
classification physics.comp-ph
keywords spectralfunctionstrigonometricmomentsCayleytransformmachinelearningDMFTquantumimpuritymodelsESPRITreal-frequencyGreen's
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

Machine-learning models for quantum impurity and DMFT problems need a target for real-frequency spectra that is neither a wasteful dense grid nor an imaginary-axis Green’s function that must be analytically continued. This paper claims that a finite sequence of Cayley-mapped trigonometric moments, with the Jacobian kept so the zeroth moment is the physical spectral weight, meets that need. The sequence is fixed-dimensional once the cutoff is chosen, stays tied to a positive matrix-valued measure, and can be mapped back to poles by ESPRIT. Normalization and positivity can therefore be imposed directly in the learning space. On single-orbital DMFT, antiferromagnetic DMFT, and a two-orbital impurity model, a graph-attention network trained on these moments matches or exceeds direct frequency-domain learning, recovers density and staggered magnetization, and keeps self-energy reconstruction stable under Dyson inversion. The practical payoff is a learning representation that is compact, physically constrained, and usable inside real-frequency self-consistency loops without analytic continuation.

What carries the argument

Cayley-mapped trigonometric moments with Jacobian: moments mk = ∫ A(ω) [f(ω)]^k dω of the spectral measure under the Cayley map of the real axis to the unit circle, so m0 is the physical weight, the block-Toeplitz matrix of moments is positive semidefinite, and ESPRIT recovers a pole representation.

What would settle it

Train the same network on moments and on a dense real-frequency grid for a multi-orbital DMFT loop that requires Dyson inversion; if the moment route produces systematically larger density or magnetization errors, noncausal self-energies, or fails undamped self-consistency while the grid route succeeds, the central claim fails.

Watch

Extended reading notes

Core claim

A finite sequence of Cayley-mapped trigonometric moments (Jacobian included) is a fixed-dimensional learning target for real-frequency Green’s functions, hybridizations, and dynamical self-energies that preserves spectral-weight normalization, encodes positivity via a positive-semidefinite block-Toeplitz condition, admits systematic ESPRIT pole reconstruction, and yields accuracy matching or exceeding direct frequency-domain learning on single-orbital DMFT, antiferromagnetic DMFT, and a two-orbital impurity model.

Load-bearing premise

That turning a truncated positive moment sequence back into a spectrum with ESPRIT stays accurate and causal enough for undamped real-frequency DMFT, even though the reconstruction does not strictly guarantee positivity.

Editorial extensions

If this is right

  • Real-frequency ML impurity solvers can replace dense grids with a fixed-length moment vector and still enforce normalization and positivity by construction.
  • Density and order parameters that are sensitive to high-frequency weight become learnable without Matsubara analytic continuation.
  • Self-energy reconstruction inside real-frequency DMFT can be regularized by projecting dynamical moments onto the positive Toeplitz cone before pole reconstruction.
  • The same fixed-dimensional target extends from scalar to matrix-valued spectra with orbital mixing, so multi-orbital problems share one representation size once the cutoff is set.

Reading between the lines

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

  • For systems with tens of orbitals the compactness gain over frequency grids should grow, provided the needed moment cutoff does not rise as fast as the orbital count.
  • The same moment target could be used for other real-frequency response functions (e.g. susceptibilities) and for steady-state transport spectra where dense grids are equally wasteful.
  • If a positivity-preserving alternative to ESPRIT is substituted for the reconstruction step, the last empirical gap between moment space and guaranteed causal spectra would close.
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

2 major / 0 minor

Summary. The manuscript proposes a fixed-dimensional learning target for real-frequency Green’s functions, hybridizations, and dynamical self-energies based on Cayley-mapped trigonometric moments with the Jacobian included (Eqs. 2–4). The zeroth moment retains spectral-weight normalization, the sequence is tied to a positive matrix-valued measure via the block-Toeplitz condition (Eq. 5), and positivity/normalization are imposed by an autocorrelation parameterization (Eq. 8) or by projection onto the PSD Toeplitz cone. Truncated moments are converted to a pole representation by ESPRIT for spectral reconstruction. Using a graph-attention network with FiLM conditioning, the authors benchmark the representation on half-filled and doped single-orbital Bethe-lattice DMFT, antiferromagnetic cubic-lattice DMFT (with Dyson inversion and moment-space self-energy filtering), and a two-orbital NCA impurity model with orbital mixing. Reported results show accuracy matching or exceeding dense frequency-grid learning, density and staggered-magnetization errors of order 10^{-3}, stable undamped ML-DMFT loops, and recovery of matrix-valued spectra including off-diagonal components.

Significance. If the claims hold, the work supplies a practical, physically constrained alternative to both dense real-frequency grids and Matsubara-based learning that requires analytic continuation. The combination of fixed dimension, built-in normalization/positivity, and a systematic ESPRIT route back to poles is well matched to surrogate impurity solvers and real-frequency DMFT. Strengths include controlled external benchmarks (tensor-network DMFT and NCA), explicit comparison to frequency-domain targets (Fig. 5), undamped self-consistency tests (Fig. 6), held-out doping density accuracy (Fig. 7), AFM order-parameter and spin-resolved spectra (Fig. 8), and a matrix-valued worst-case reconstruction (Fig. 10). The authors also document free parameters (ω_p, K, R, ESPRIT tolerances) and openly note that ESPRIT does not strictly preserve positivity. This is a solid methods contribution for computational many-body physics and ML for Green’s functions.

major comments (2)
  1. §II.B and Conclusions: ESPRIT reconstruction from a truncated PSD block-Toeplitz sequence is not guaranteed to preserve positivity, yet undamped real-frequency DMFT self-consistency (Figs. 6, 8) and Dyson inversion of Σ rely on the reconstructed spectra remaining causal enough. The paper controls this empirically (PSD projection before reconstruction; elevated ESPRIT tolerances under noise in Fig. 3) and reports no observed violations, but a load-bearing claim for the method’s use inside self-consistency loops would be strengthened by a quantitative diagnostic—e.g., the fraction of reconstructed poles with Im ξ_ℓ ≥ 0, residual negative spectral weight after reconstruction, or a comparison against a positivity-preserving truncated-moment solver—on the same DMFT/NCA datasets.
  2. §III.A–C and Appendices B–E: Free parameters (ω_p, K, R=⌊1.5K⌋, ESPRIT tolerances, off-diagonal rescaling q) are chosen empirically per benchmark, and the main accuracy claims (matching/exceeding frequency-domain learning; density and m errors ~10^{-3}) are demonstrated only for those choices. A short sensitivity study—varying K and ω_p around the reported values and reporting KL or Frobenius error and DMFT convergence—would show that the central claim is not tied to a narrow hyperparameter island and would make the representation more transferable.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: moment representation and ML accuracy claims are definitional constructions plus independent numerical benchmarks against external solvers.

full rationale

The paper defines Cayley-mapped trigonometric moments (Eqs. 2–4) with Jacobian so that m0 equals physical spectral weight by construction, and enforces block-Toeplitz positivity either by the autocorrelation map (Eq. 8) or by explicit projection onto the PSD cone. These are transparent design choices for a fixed-dimensional learning target, not predictions derived from themselves. ESPRIT pole recovery (Eqs. 6–7) is presented as a systematic reconstruction route whose positivity is not guaranteed a priori (explicitly noted in §II.B and Conclusions); the authors control it empirically via elevated tolerances and pre-projection, then evaluate reconstructed spectra against independent real-frequency data from tensor-network DMFT and NCA solvers. Accuracy, density, magnetization, and self-energy claims in §III are therefore ordinary supervised-learning metrics on held-out solver trajectories, not tautological. Self-citations supply prior pole/ESPRIT tooling and do not underwrite the measured errors. The derivation chain is self-contained against external benchmarks.

Assumptions & free parameters 6 free parameters · 5 assumptions · 1 invented entities

The central claim rests on standard spectral/moment theory plus several empirical engineering choices (moment cutoff K, Cayley scale ω_p, ESPRIT tolerances, autocorrelation length R, network hyperparameters, off-diagonal rescaling). No new physical particles or forces are postulated; the ‘invented’ object is a representation construction. Positivity after ESPRIT is assumed from practice rather than proved.

free parameters (6)
  • Cayley scale ω_p = half-bandwidth (e.g. 2, 6, 10)
    Sets the frequency scale of the map; chosen as half-bandwidth (2, 6, or 10) per benchmark, not derived.
  • moment cutoff K = 30–150 depending on task
    Truncation length of the learning target; chosen empirically (30–150) by spectral resolution needs.
  • autocorrelation length R = ⌊1.5K⌋
    Hyperparameter R=⌊1.5K⌋ in the B_r→m_k map; chosen for flexibility, not derived.
  • ESPRIT reconstruction tolerances = task-dependent ~1e-3–8e-2
    Noise/error cutoffs (e.g. 2e-3 to 8e-2) control pole recovery and are tuned per noise/smoothness regime.
  • off-diagonal moment rescaling q = ⌊1/m_max⌋
    Integer prefactor q=⌊1/m_max⌋ balances small off-diagonal moments in the two-orbital loss; data-dependent.
  • network and training hyperparameters = see Appendix A
    Widths, heads, LR schedule, dropout, L2, epochs fixed by hand (Appendix A).
assumptions (5)
  • standard math A positive matrix-valued spectral measure has a positive-semidefinite block-Toeplitz sequence of trigonometric moments (and conversely a PSD Toeplitz sequence is realizable by some positive measure).
    Invoked in §II.B Eqs. (3)–(5); classical moment theory / Toeplitz forms.
  • standard math Including the Cayley Jacobian makes m0 equal the physical spectral weight and keeps moments bounded by that weight.
    §II.B Eqs. (2)–(4); standard change-of-variable under the Cayley map.
  • domain assumption A truncated trigonometric moment sequence can be converted to a useful complex-pole model via ESPRIT/Prony-type methods for the spectra considered.
    §II.B Eqs. (6)–(7); relies on prior pole/ESPRIT literature and empirical success on impurity/DMFT spectra.
  • ad hoc to paper ESPRIT reconstruction from PSD moments does not introduce practically harmful positivity violations for the benchmarks studied.
    Authors state they have not observed violations and that ESPRIT does not strictly preserve positivity (§II.B, Conclusions).
  • domain assumption NCA and complex-time tensor-network solvers produce spectra of sufficient quality to serve as supervised labels for the representation comparison.
    §II.E and Appendices D–E; NCA is approximate but used as a practical matrix-valued data source.
invented entities (1)
  • Cayley-mapped trigonometric moment ML target (Jacobian-included, fixed-K sequence with autocorrelation parameterization) independent evidence
    purpose: Provide a compact, normalized, positivity-constrained fixed-dimensional label for learning real-frequency G, Δ, and dynamical Σ.
    Construction assembled from known transforms and moment theory; novelty is the packaging as an ML target with by-construction constraints and ESPRIT recovery path. Not a new physical degree of freedom.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Compact and Stable Representation of Real-Frequency Spectral Functions for Machine Learning." pith.science (2026). https://pith.science/paper/CJYHVC2A

@misc{pith2026260711190,
  author       = {Pith},
  title        = {Pith review of: Compact and Stable Representation of Real-Frequency Spectral Functions for Machine Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CJYHVC2A}},
  note         = {Machine review of arXiv:2607.11190}
}
read the original abstract

We introduce a compact and stable moment representation for real-frequency Green's functions, hybridization functions, and self-energies for machine-learning applications, avoiding the inefficiency of dense frequency grids as well as the ill-posed analytic continuation of Matsubara approaches. The representation is constructed from Cayley-mapped trigonometric moments with the Jacobian included, which preserve spectral-weight normalization, tie the moment sequence to a positive matrix-valued spectral measure, and admit a systematic route to a pole representation via ESPRIT. This provides a fixed-dimensional learning target in which physical constraints such as normalization and positivity can be imposed directly. Using a graph-attention neural network with FiLM conditioning, we benchmark the representation on single-orbital DMFT, antiferromagnetic DMFT, and a two-orbital impurity model. The results demonstrate accuracy matching or exceeding that of direct frequency-domain learning, reliable reproduction of the density and staggered magnetization, stable self-energy reconstruction through Dyson equation inversion, and accurate recovery of matrix-valued spectra with orbital mixing.

Figures

Figures reproduced from arXiv: 2607.11190 by the authors.

Figure 1
Figure 1. FIG. 1. Ambiguity of the pole representation. (a) Spectral [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Robustness of the moment representation to noise. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. FIG. 4. Schematic of the neural network architecture. Left: [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figures from the paper (4 more)
Figure 5
Figure 5. Figure 5: FIG. 5. Comparison of moment-space and frequency-domain [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 8
Figure 8. Figure 8: FIG. 8. Antiferromagnetic order and spin-resolved spectral [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9. Learning curve for the two-orbital model. Average [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: FIG. 10. Worst-case test-sample reconstruction for the two [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

73 extracted references · 1 canonical work pages

  1. [1]

    In contrast, a recently developed minimal pole repre- sentation [25] and its matrix-valued generalization [27] provide a powerful compact representation on the real axis

    bases, are very compact but inaccurate in practice when applied to real spectra: while they provide precise imaginary-axis properties, they typically cannot repre- sent real-frequency functions accurately due to the ill- conditioned analytic continuation kernel. In contrast, a recently developed minimal pole repre- sentation [25] and its matrix-valued gen...

    2001

  2. [2]

    Georges, G

    A. Georges, G. Kotliar, W. Krauth, and M. J. Rozen- berg, Dynamical mean-field theory of strongly correlated fermion systems and the limit of infinite dimensions, Rev. 15 Mod. Phys.68, 13 (1996)

    1996

  3. [3]

    Kotliar, S

    G. Kotliar, S. Y. Savrasov, K. Haule, V. S. Oudovenko, O. Parcollet, and C. A. Marianetti, Electronic struc- ture calculations with dynamical mean-field theory, Rev. Mod. Phys.78, 865 (2006)

    2006

  4. [4]

    Onida, L

    G. Onida, L. Reining, and A. Rubio, Electronic exci- tations: density-functional versus many-body green’s- function approaches, Rev. Mod. Phys.74, 601 (2002)

    2002

  5. [5]

    Arsenault, A

    L.-F. Arsenault, A. Lopez-Bezanilla, O. A. von Lilienfeld, andA.J.Millis,Machinelearningformany-bodyphysics: The case of the Anderson impurity model, Phys. Rev. B 90, 155136 (2014)

    2014

  6. [6]

    Sheridan, C

    E. Sheridan, C. Rhodes, F. Jamet, I. Rungger, and C. Weber, Data-driven dynamical mean-field theory: An error-correction approach to solve the quantum many- body problem using machine learning, Phys. Rev. B104, 205120 (2021)

    2021

  7. [7]

    H. Lee, Z. Zhao, G. H. Booth, W. Ge, and C. Weber, Language-inspired machine learning approach for solving strongly correlated problems with dynamical mean-field theory, Phys. Rev. B112, 035165 (2025)

    2025

  8. [8]

    Agapov, O

    E. Agapov, O. Bertomeu, A. Carballo, C. B. Mendl, and A. Sander, Predicting interacting green’s functions with neural networks (2024), arXiv:2411.13644 [cond-mat.str- el]

    arXiv 2024

Show all 73 references
  1. [9]

    Kakizawa, S

    F. Kakizawa, S. Terasaki, and H. Shinaoka, Physics- informed neural network model for quantum impu- rity problems based on lehmann representation (2024), arXiv:2411.18835 [cond-mat.str-el]

    2024 arXiv

  2. [10]

    X. Dong, E. Gull, and L. Wang, Equivariant neural net- work for Green’s functions of molecules and materials, Phys. Rev. B109, 075112 (2024)

    2024

  3. [11]

    Valenti, I

    A. Valenti, I. Park, A. Georges, A. J. Millis, and O.Parcollet,Neural-networkquantumembeddingsolvers for correlated materials (2026), arXiv:2603.15741 [cond- mat.str-el]

    2026

  4. [12]

    Y. Zhu, P. Rosenberg, Z. Huang, H. Bassi, C. Yang, and S. Zhang, Transformer-based operator learning frame- work for self-energy in strongly correlated systems, Phys. Rev. B113, 245139 (2026)

    2026

  5. [13]

    E. J. Sturm, M. R. Carbone, D. Lu, A. Weichselbaum, and R. M. Konik, Predicting impurity spectral functions usingmachinelearning,Phys.Rev.B103,245118(2021)

    2021

  6. [14]

    Ren, R.-S

    X.-Y. Ren, R.-S. Han, and L. Chen, Learning impurity spectral functions from density of states, J. Phys.: Con- dens. Matter33, 495601 (2021)

    2021

  7. [15]

    Liu, R.-S

    T. Liu, R.-S. Han, and L. Chen, Prediction of impu- rity spectrum function by deep learning algorithm, Chin. Phys. B33, 057102 (2024)

    2024

  8. [16]

    Miles, M

    C. Miles, M. R. Carbone, E. J. Sturm, D. Lu, A. Weich- selbaum, K. Barros, and R. M. Konik, Machine learning of kondo physics using variational autoencoders and sym- bolic regression, Phys. Rev. B104, 235111 (2021)

    2021

  9. [17]

    F. Deng, Y. Lu, X. Cao, and Z. Zhong, Neural network impurity solver for real-frequency dynamical mean-field theory (2025), arXiv:2511.14505 [cond-mat.str-el]

    2025

  10. [18]

    Boehnke, H

    L. Boehnke, H. Hafermann, M. Ferrero, F. Lechermann, and O. Parcollet, Orthogonal polynomial representation of imaginary-time Green’s functions, Phys. Rev. B84, 075145 (2011)

    2011

  11. [19]

    E. Gull, S. Iskakov, I. Krivenko, A. A. Rusakov, and D. Zgid, Chebyshev polynomial representation of imaginary-time response functions, Phys. Rev. B98, 075127 (2018)

    2018

  12. [20]

    Shinaoka, J

    H. Shinaoka, J. Otsuki, M. Ohzeki, and K. Yoshimi, Compressing Green’s function using intermediate repre- sentation between imaginary-time and real-frequency do- mains, Phys. Rev. B96, 035147 (2017)

    2017

  13. [21]

    J. Kaye, K. Chen, and O. Parcollet, Discrete Lehmann representation of imaginary-time Green’s functions, Phys. Rev. B105, 235115 (2022)

    2022

  14. [22]

    Jarrell and J

    M. Jarrell and J. E. Gubernatis, Bayesian inference and the analytic continuation of imaginary-time quantum Monte Carlo data, Phys. Rep.269, 133 (1996)

    1996

  15. [23]

    Fei, C.-N

    J. Fei, C.-N. Yeh, and E. Gull, Nevanlinna analytical con- tinuation, Phys. Rev. Lett.126, 056402 (2021)

    2021

  16. [24]

    Fei, C.-N

    J. Fei, C.-N. Yeh, D. Zgid, and E. Gull, Analytical con- tinuation of matrix-valued functions: Carathéodory for- malism, Phys. Rev. B104, 165111 (2021)

    2021

  17. [25]

    Shao and A

    H. Shao and A. W. Sandvik, Progress on stochastic an- alytic continuation of quantum Monte Carlo data, Phys. Rep.1003, 1 (2023)

    2023

  18. [26]

    Zhang and E

    L. Zhang and E. Gull, Minimal pole representation and controlled analytic continuation of matsubara response functions, Phys. Rev. B110, 035154 (2024)

    2024

  19. [27]

    A. A. Kananenka, A. R. Welden, T. N. Lan, E. Gull, and D. Zgid, Efficient temperature-dependent green’s func- tion methods for realistic systems: Using cubic spline in- terpolation to approximate matsubara green’s functions, Journal of Chemical Theory and Computation12, 2250 (2016)

    2016

  20. [28]

    Zhang, Y

    L. Zhang, Y. Yu, and E. Gull, Minimal pole representa- tion and analytic continuation of matrix-valued correla- tion functions, Phys. Rev. B110, 235131 (2024)

    2024

  21. [29]

    Zhang, A

    L. Zhang, A. Erpenbeck, Y. Yu, and E. Gull, Minimal pole representation for spectral functions, The Journal of Chemical Physics162, 214111 (2025)

    2025

  22. [30]

    Erpenbeck, Y

    A. Erpenbeck, Y. Zhu, Y. Yu, L. Zhang, R. Gerum, O. Goulko, C. Yang, G. Cohen, and E. Gull, Compact representation and long-time extrapolation of real-time data for quantum systems using the esprit algorithm, Phys. Rev. B113, 115129 (2026)

    2026

  23. [31]

    Gazizova, L

    D. Gazizova, L. Zhang, E. Gull, and J. P. F. LeBlanc, Feynman diagrammatics based on discrete pole represen- tations: A path to renormalized perturbation theories, Phys. Rev. B110, 075158 (2024)

    2024

  24. [32]

    N. I. Akhiezer,The Classical Moment Problem and Some Related Questions in Analysis(Oliver and Boyd, Edin- burgh, 1965)

    1965

  25. [33]

    Comanac,Dynamical mean field theory of corre- lated electron systems: New algorithms and applications to local observables, Ph.D

    A.-B. Comanac,Dynamical mean field theory of corre- lated electron systems: New algorithms and applications to local observables, Ph.D. thesis, Columbia University, New York (2007)

    2007

  26. [34]

    A. A. Rusakov, J. J. Phillips, and D. Zgid, Local hamilto- nians for quantitative green’s function embedding meth- ods, The Journal of Chemical Physics141, 194105 (2014)

    2014

  27. [35]

    full-frequency

    O. J. Backhouse and G. H. Booth, Constructing “full-frequency” spectra via moment constraints for coupled cluster green’s functions, Journal of Chem- ical Theory and Computation18, 6622 (2022), https://doi.org/10.1021/acs.jctc.2c00670

    2022 doi

  28. [36]

    Farid, Many-body perturbation expansions with- out diagrams

    B. Farid, Many-body perturbation expansions with- out diagrams. i. normal states (2021), arXiv:1912.00474 [cond-mat.str-el]

    2021 arXiv

  29. [37]

    Abbott, W

    R. Abbott, W. Jay, and P. Oare, Moment problems and spectral functions (2026), arXiv:2602.11260 [hep-lat]

    2026

  30. [38]

    Geronimus, On the trigonometric moment problem, 16 Annals of Mathematics47, 742 (1946)

    J. Geronimus, On the trigonometric moment problem, 16 Annals of Mathematics47, 742 (1946)

    1946

  31. [39]

    Grenander and G

    U. Grenander and G. Szegö,Toeplitz Forms and Their Applications(University of California Press, Berkeley, 1958)

    1958

  32. [40]

    G. R. de Prony, Essai experimental et analytique: sur les lois de la dilatabilite des fluides elastique et sur celles de la force expansive de la vapeur de l’eau et de la vapeur de l’alkool, a differentes temperatures, Journal Polytech- nique ou Bulletin du Travail fait a l’Ec...

  33. [41]

    Roy and T

    R. Roy and T. Kailath, Esprit-estimation of signal parameters via rotational invariance techniques, IEEE Transactions on acoustics, speech, and signal processing 37, 984 (1989)

    1989

  34. [42]

    Potts and M

    D. Potts and M. Tasche, Parameter estimation for non- increasing exponential sums by prony-like methods, Lin- ear Algebra and its Applications439, 1024 (2013), 17th Conference of the International Linear Algebra Society, Braunschweig, Germany, August 2011

    2013

  35. [43]

    Ying, Pole recovery from noisy data on imaginary axis, Journal of Scientific Computing92, 107 (2022)

    L. Ying, Pole recovery from noisy data on imaginary axis, Journal of Scientific Computing92, 107 (2022)

    2022

  36. [44]

    Ying, Analytic continuation from limited noisy mat- subara data, Journal of Computational Physics469, 111549 (2022)

    L. Ying, Analytic continuation from limited noisy mat- subara data, Journal of Computational Physics469, 111549 (2022)

    2022

  37. [45]

    A. F. Kemper, C. Yang, and E. Gull, Denoising and ex- tension of response functions in the time domain, Phys. Rev. Lett.132, 160403 (2024)

    2024

  38. [46]

    Dette and J

    H. Dette and J. Wagener, Matrix measures on the unit circle, moment spaces, orthogonal polynomials and the geronimus relations, Linear Algebra and its Applications 432, 1609 (2010)

    2010

  39. [47]

    Ephremidze, G

    L. Ephremidze, G. Janashia, and E. Lagvilava, A simple proof of the matrix-valued fejér-riesz theorem, Journal of Fourier Analysis and Applications15, 124 (2009)

    2009

  40. [48]

    Simon,Orthogonal Polynomials on the Unit Circle, Part 1: Classical Theory, American Mathematical Soci- ety Colloquium Publications, Vol

    B. Simon,Orthogonal Polynomials on the Unit Circle, Part 1: Classical Theory, American Mathematical Soci- ety Colloquium Publications, Vol. 54 (American Mathe- matical Society, Providence, RI, 2005)

    2005

  41. [49]

    A. N. Rubtsov, V. V. Savkin, and A. I. Lichten- stein, Continuous-time quantum monte carlo method for fermions, Phys. Rev. B72, 035122 (2005)

    2005

  42. [50]

    Werner, A

    P. Werner, A. Comanac, L. de’ Medici, M. Troyer, and A. J. Millis, Continuous-time solver for quantum impu- rity models, Phys. Rev. Lett.97, 076405 (2006)

    2006

  43. [51]

    Werner and A

    P. Werner and A. J. Millis, Hybridization expansion im- purity solver: General formulation and application to kondo lattice and two-orbital models, Phys. Rev. B74, 155107 (2006)

    2006

  44. [52]

    E. Gull, P. Werner, O. Parcollet, and M. Troyer, Continuous-time auxiliary-field monte carlo for quantum impurity models, Europhysics Letters82, 57003 (2008)

    2008

  45. [53]

    E. Gull, A. J. Millis, A. I. Lichtenstein, A. N. Rubtsov, M. Troyer, and P. Werner, Continuous-time monte carlo methods for quantum impurity models, Rev. Mod. Phys. 83, 349 (2011)

    2011

  46. [54]

    Ganahl, M

    M. Ganahl, M. Aichhorn, H. G. Evertz, P. Thunström, K. Held, and F. Verstraete, Efficient dmft impurity solver using real-time dynamics with matrix product states, Phys. Rev. B92, 155132 (2015)

    2015

  47. [55]

    Bauernfeind, M

    D. Bauernfeind, M. Zingl, R. Triebl, M. Aichhorn, and H. G. Evertz, Fork tensor-product states: Efficient mul- tiorbital real-time dmft solver, Phys. Rev. X7, 031013 (2017)

    2017

  48. [56]

    X. Cao, Y. Lu, P. Hansmann, and M. W. Haverkort, Tree tensor-network real-time multiorbital impurity solver: Spin-orbit coupling and correlation functions insr2ruo4, Phys. Rev. B104, 115119 (2021)

    2021

  49. [57]

    X. Cao, Y. Lu, E. M. Stoudenmire, and O. Parcollet, Dynamical correlation functions from complex time evo- lution, Phys. Rev. B109, 235110 (2024)

    2024

  50. [58]

    Grundner, P

    M. Grundner, P. Westhoff, F. B. Kugler, O. Parcollet, and U. Schollwöck, Complex time evolution in tensor net- works and time-dependent green’s functions, Phys. Rev. B109, 155124 (2024)

    2024

  51. [59]

    Y. Yu, L. Zhang, E. Gull, X. Cao, and X. Dong, Multi- orbital dynamical mean-field theory with a complex-time solver, Phys. Rev. Res.8, 023142 (2026)

    2026

  52. [60]

    Eckstein and P

    M. Eckstein and P. Werner, Nonequilibrium dynamical mean-field calculations based on the noncrossing approx- imation and its generalizations, Physical Review B82, 115115 (2010)

    2010

  53. [61]

    Cohen, D

    G. Cohen, D. R. Reichman, A. J. Millis, and E. Gull, Green’s functions from real-time bold-line Monte Carlo, Physical Review B89, 115139 (2014)

    2014

  54. [62]

    Erpenbeck, E

    A. Erpenbeck, E. Gull, and G. Cohen, Revealing strong correlations in higher-order transport statistics: A non- crossing approximation approach, Physical Review B 103, 125431 (2021)

    2021

  55. [63]

    Erpenbeck and G

    A. Erpenbeck and G. Cohen, Resolving the nonequilib- rium kondo singlet in energy- and position-space using quantum measurements, SciPost Physics10, 142 (2021)

    2021

  56. [64]

    Zemach, A

    I. Zemach, A. Erpenbeck, E. Gull, and G. Cohen, Nonequilibrium steady state full counting statistics in the noncrossing approximation, The Journal of Chemi- cal Physics161, 164113 (2024)

    2024

  57. [65]

    Cohen, E

    G. Cohen, E. Gull, D. R. Reichman, and A. J. Millis, Taming the Dynamical Sign Problem in Real-Time Evo- lution of Quantum Many-Body Problems, Physical Re- view Letters115, 266802 (2015)

    2015

  58. [66]

    Eidelstein, E

    E. Eidelstein, E. Gull, and G. Cohen, Multiorbital Quantum Impurity Solver for General Interactions and Hybridizations, Physical Review Letters124, 206405 (2020)

    2020

  59. [67]

    Erpenbeck, E

    A. Erpenbeck, E. Gull, and G. Cohen, Quantum Monte Carlo Method in the Steady State, Physical Review Let- ters130, 186301 (2023)

    2023

  60. [68]

    Erpenbeck, T

    A. Erpenbeck, T. Blommel, L. Zhang, W.-T. Lin, G. Co- hen, and E. Gull, Steady-state properties of multi-orbital systems using quantum Monte Carlo, The Journal of Chemical Physics161, 094104 (2024)

    2024

  61. [69]

    Gilmer, S

    J. Gilmer, S. S. Schoenholz, P. F. Riley, O. Vinyals, and G. E. Dahl, Neural message passing for quantum chem- istry, inProceedings of the 34th International Conference on Machine Learning(PMLR, 2017) pp. 1263–1272

    2017

  62. [70]

    P. W. Battaglia, J. B. Hamrick, V. Bapst, A. Sanchez- Gonzalez, V. Zambaldi, M. Malinowski, A. Tacchetti, D. Raposo, A. Santoro, R. Faulkner,et al., Relational in- ductive biases, deep learning, and graph networks, arXiv preprint arXiv:1806.01261 (2018)

    2018 arXiv

  63. [71]

    Veličković, G

    P. Veličković, G. Cucurull, A. Casanova, A. Romero, P. Liò, and Y. Bengio, Graph attention networks, in International Conference on Learning Representations (2018)

    2018

  64. [72]

    Perez, F

    E. Perez, F. Strub, H. De Vries, V. Dumoulin, and A. Courville, Film: Visual reasoning with a general con- ditioning layer, inProceedings of the AAAI Conference on Artificial Intelligence, Vol. 32 (2018)

    2018

  65. [73]

    L. R. Mead and N. Papanicolaou, Maximum entropy 17 in the problem of moments, Journal of Mathematical Physics25, 2404 (1984)

    1984

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.