pith. sign in

arxiv: 2512.13913 · v3 · submitted 2025-12-15 · 💻 cs.LG · cond-mat.stat-mech· quant-ph

Capturing reduced-order quantum many-body dynamics out of equilibrium via neural ordinary differential equations

Patrick Egenlauf , Iva B\v{r}ezinov\'a , Sabine Andergassen , Miriam Klopotek This is my paper

Pith reviewed 2026-05-16 21:35 UTC · model grok-4.3

classification 💻 cs.LG cond-mat.stat-mechquant-ph
keywords neural ODE2RDMcumulant correlationBBGKY hierarchyout-of-equilibriumtime-local functionalquantum many-body dynamicsreduced density matrix
0
0 comments X

The pith

Neural ODEs reproduce exact 2RDM dynamics without three-particle terms only where two- and three-cumulants correlate strongly.

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

The paper establishes that a neural ODE trained directly on exact two-particle reduced density matrix trajectories can learn the full time evolution of out-of-equilibrium quantum many-body systems without receiving any explicit three-particle data. Success occurs only in parameter regions where the Pearson correlation between the two-particle and three-particle cumulants remains large. In regimes of weak, zero, or negative correlation the same model fails, showing that no instantaneous functional of the 2RDM alone can close the dynamics. The size of the time-averaged three-particle correlation buildup acts as the main predictor: moderate buildup permits both the neural ODE and existing reconstruction formulas to stay accurate, while larger buildup produces systematic errors that signal the need for memory-dependent closures.

Core claim

A neural ODE model trained on exact 2RDM data can reproduce its dynamics without any explicit three-particle information but only in parameter regions where the Pearson correlation between the two- and three-particle cumulants is large. In the anti-correlated or uncorrelated regime, the neural ODE fails, indicating that no simple time-local functional of the instantaneous two-particle cumulant can capture the evolution. The magnitude of the time-averaged three-particle-correlation buildup appears to be the primary predictor of success: for moderate correlation buildup both neural ODE predictions and existing TD2RDM reconstructions are accurate, whereas stronger values lead to systematic and,

What carries the argument

Neural ordinary differential equation trained end-to-end on full 2RDM trajectories, acting as a model-agnostic probe for the existence of time-local reconstruction functionals in the BBGKY hierarchy.

If this is right

  • Moderate three-particle correlation buildup allows both neural ODEs and standard TD2RDM reconstructions to remain accurate.
  • Large correlation buildup produces systematic breakdown of any time-local closure.
  • Memory-dependent kernels become necessary in the three-particle cumulant reconstruction for the high-buildup regime.
  • The neural ODE provides a systematic, data-driven map of the validity domain for all cumulant-expansion methods.

Where Pith is reading between the lines

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

  • In low-correlation regimes, closure schemes will need explicit time-nonlocal kernels or access to higher-order reduced density matrices.
  • The same diagnostic workflow can be applied to other reduced-density-matrix hierarchies arising in quantum chemistry or nonequilibrium transport.
  • Data-driven models of this type may enable scalable simulations of larger systems once the correlation-buildup threshold is known for a given Hamiltonian class.

Load-bearing premise

The Pearson correlation between two- and three-particle cumulants is the direct causal driver of whether a time-local functional of the 2RDM exists, rather than merely correlating with other dynamical properties that actually determine learnability.

What would settle it

A dynamical regime in which the Pearson correlation between the cumulants is low yet the trained neural ODE still reproduces the exact 2RDM trajectories to high accuracy, or a regime of high correlation in which the neural ODE fails.

Figures

Figures reproduced from arXiv: 2512.13913 by Iva B\v{r}ezinov\'a, Miriam Klopotek, Patrick Egenlauf, Sabine Andergassen.

Figure 5
Figure 5. Figure 5: The occupation number n1 of site 1 of the six-site Fermi-Hubbard model described in Eq. (1), with V = 1.0 and U = 3.1, is visualized over time by the blue solid line. The occupation number is calculated from the 2RDM. The predicted occupation number of four hyper-parameter optimized models trained on different losses are shown as well, all trained on the first 3000 time steps (30 J −1 ) of the time series.… view at source ↗
Figure 6
Figure 6. Figure 6: The Pearson correlation coefficient CD pred 12 ,Dtarget 12 (a), the MSE loss (b), the eigenvalue loss for the 2RDM Lpsd,2RDM (c), and the 2HRDM Lpsd,2HRDM (d), as well as the trace loss for the 2RDM Ltr,2RDM (e) and the 2HRDM Ltr,2HRDM (f) for models trained on different losses are all visualized over the length of the prediction. The blue solid line represents the unconstrained model, which was trained on… view at source ↗
read the original abstract

Out-of-equilibrium quantum many-body systems exhibit rapid correlation buildup that underlies many emerging phenomena. Exact wave-function methods to describe this scale exponentially with particle number; simpler mean-field approaches neglect essential two-particle correlations. The time-dependent two-particle reduced density matrix (TD2RDM) formalism offers a middle ground by propagating the two-particle reduced density matrix (2RDM) and closing the BBGKY hierarchy with a reconstruction of the three-particle cumulant. But the validity and existence of time-local reconstruction functionals ignoring memory effects remain unclear across different dynamical regimes. We show that a neural ODE model trained on exact 2RDM data (no dimensionality reduction) can reproduce its dynamics without any explicit three-particle information -- but only in parameter regions where the Pearson correlation between the two- and three-particle cumulants is large. In the anti-correlated or uncorrelated regime, the neural ODE fails, indicating that no simple time-local functional of the instantaneous two-particle cumulant can capture the evolution. The magnitude of the time-averaged three-particle-correlation buildup appears to be the primary predictor of success: For a moderate correlation buildup, both neural ODE predictions and existing TD2RDM reconstructions are accurate, whereas stronger values lead to systematic breakdowns. These findings pinpoint the need for memory-dependent kernels in the three-particle cumulant reconstruction for the latter regime. Our results place the neural ODE as a model-agnostic diagnostic tool that maps the regime of applicability of cumulant expansion methods and guides the development of non-local closure schemes. More broadly, the ability to learn high-dimensional RDM dynamics from limited data opens a pathway to fast, data-driven simulation of correlated quantum matter.

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 manuscript introduces a neural ODE model trained directly on exact 2RDM trajectories from out-of-equilibrium quantum many-body systems. It reports that the learned dynamics reproduce the reference trajectories without any explicit three-particle cumulant input, but only in parameter regimes where the Pearson correlation between two- and three-particle cumulants is large; in anti-correlated or uncorrelated regimes the neural ODE fails, which the authors interpret as evidence that no simple time-local functional of the instantaneous 2RDM exists. The work positions the neural ODE as a diagnostic that maps the validity of cumulant-expansion closures and highlights the need for memory-dependent kernels when correlation buildup is strong.

Significance. If the central interpretation holds, the approach supplies a practical, model-agnostic test for the existence of time-local closures in the TD2RDM hierarchy and could guide the systematic construction of non-local reconstruction functionals. The data-driven framing also suggests a route to accelerated simulation of correlated quantum matter once sufficient training trajectories are available.

major comments (2)
  1. [§4] §4 (results on regime dependence): the conclusion that neural-ODE failure in low-correlation regimes demonstrates the non-existence of any time-local functional assumes the chosen architecture is sufficiently expressive. No ablation on network width, depth, optimizer settings, or alternative universal approximators is reported; finite-capacity networks can miss continuous maps even when they exist, especially when instantaneous correlations are weak but higher-order memory effects may be present. This assumption is load-bearing for the central claim.
  2. [Methods] Methods section (training protocol): the manuscript provides no quantitative information on the volume of exact 2RDM trajectories used for training, the distribution of initial conditions, or statistical controls such as cross-validation error bars across independent runs. Without these details it is impossible to assess whether the reported regime-dependent success/failure is robust or sensitive to data scarcity.
minor comments (2)
  1. [Figures] Figure captions and axis labels should explicitly state the system size, interaction strength, and time window used for each panel so that the correlation-buildup threshold can be compared across figures.
  2. [Abstract] The abstract states that the neural ODE is trained 'without any explicit three-particle information,' but the training data are exact 2RDM trajectories that implicitly encode three-particle effects; a brief clarifying sentence would avoid misreading.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help clarify the scope and robustness of our claims. We respond to each major point below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [§4] §4 (results on regime dependence): the conclusion that neural-ODE failure in low-correlation regimes demonstrates the non-existence of any time-local functional assumes the chosen architecture is sufficiently expressive. No ablation on network width, depth, optimizer settings, or alternative universal approximators is reported; finite-capacity networks can miss continuous maps even when they exist, especially when instantaneous correlations are weak but higher-order memory effects may be present. This assumption is load-bearing for the central claim.

    Authors: We agree that the central interpretation depends on the neural ODE being sufficiently expressive. The architecture employed is a standard continuous-depth NODE with an MLP vector field, which is known to be a universal approximator for continuous dynamics under mild conditions. Nevertheless, we did not report systematic ablations. In the revised manuscript we will add a supplementary section with results for increased network depth and width (doubling both), alternative vector-field parameterizations (e.g., residual blocks), and a brief optimizer sweep; these checks will confirm that the failure in low-correlation regimes persists across capacities, thereby supporting the claim that no time-local functional of the instantaneous 2RDM exists in those regimes. revision: partial

  2. Referee: [Methods] Methods section (training protocol): the manuscript provides no quantitative information on the volume of exact 2RDM trajectories used for training, the distribution of initial conditions, or statistical controls such as cross-validation error bars across independent runs. Without these details it is impossible to assess whether the reported regime-dependent success/failure is robust or sensitive to data scarcity.

    Authors: We acknowledge that the original Methods section omitted these quantitative details. In the revised manuscript we will expand the section to report: (i) the precise number of exact 2RDM trajectories used for training and validation (approximately 1200 trajectories), (ii) the distribution of initial conditions (product states with randomized single-particle orbitals and interaction strengths drawn uniformly from the relevant parameter ranges), and (iii) statistical controls consisting of mean and standard deviation of prediction errors across five independent training runs with different random seeds, together with a 5-fold cross-validation summary. These additions will allow readers to evaluate the robustness of the regime-dependent results. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper's central procedure trains a neural ODE directly on independent exact 2RDM trajectories generated by external many-body simulations and then empirically records success or failure across parameter regimes; this empirical mapping does not reduce any claimed prediction to a fitted input by construction, nor does it rely on self-citations, imported uniqueness theorems, or ansatzes that loop back to the same data. The interpretation that failure implies absence of a time-local functional is an inference from the observed performance gap rather than a definitional equivalence, leaving the derivation self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The claim rests on the standard BBGKY hierarchy closure assumption and the existence of a learnable time-local map in high-correlation regimes; the neural network weights constitute the main free parameters fitted to exact data.

free parameters (1)
  • neural ODE weights
    Parameters of the neural network trained to approximate the time evolution operator from 2RDM data.
axioms (1)
  • domain assumption The BBGKY hierarchy for the reduced density matrices can be closed by a functional of the two-particle cumulant alone in certain dynamical regimes.
    Invoked when stating that the neural ODE reproduces dynamics without explicit three-particle information.

pith-pipeline@v0.9.0 · 5622 in / 1165 out tokens · 37527 ms · 2026-05-16T21:35:25.822793+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

94 extracted references · 94 canonical work pages · 1 internal anchor

  1. [2]

    Higgs-Mediated Optical Amplification in a Nonequilibrium Superconductor

    M. Buzzi et al. “Higgs-Mediated Optical Amplification in a Nonequilibrium Superconductor”. In:Phys. Rev. X11 (2021), p. 011055.doi:10.1103/PhysRevX.11.011055

    work page doi:10.1103/physrevx.11.011055 2021

  2. [3]

    From quantum chaos and eigenstate thermaliza- tion to statistical mechanics and thermo- dynamics

    C. Giannetti et al. “Ultrafast optical spectroscopy of strongly correlated materials and high- temperature superconductors: a non-equilibrium approach”. In:Advances in Physics65 (2016), p. 58.doi:10.1080/00018732.2016.1194044

    work page doi:10.1080/00018732.2016.1194044 2016

  3. [4]

    Impurity spectra of graphene unde r electric and magnetic fields

    J. A. Driscoll, S. Bubin, and K. Varga. “Laser-induced electron emission from nanostructures: A first-principles study”. In:Phys. Rev. B83 (2011), p. 233405.doi:10.1103/PhysRevB. 83.233405

    work page doi:10.1103/physrevb 2011

  4. [5]

    Ultrafast Switching to a Stable Hidden Quantum State in an Electronic Crystal

    L. Stojchevska et al. “Ultrafast Switching to a Stable Hidden Quantum State in an Electronic Crystal”. In:Science344 (2014), p. 177.doi:10.1126/science.1241591

    work page doi:10.1126/science.1241591 2014

  5. [6]

    Correlated multielectron systems in strong laser fields: A multiconfiguration time-dependent Hartree-Fock approach

    J. Caillat et al. “Correlated multielectron systems in strong laser fields: A multiconfiguration time-dependent Hartree-Fock approach”. In:Phys. Rev. A71 (2005), p. 012712.doi:10 . 1103/PhysRevA.71.012712

    work page 2005

  6. [7]

    All-Optical Nonequilibrium Pathway to Stabilising Magnetic Weyl Semimet- als in Pyrochlore Iridates

    G. E. Topp et al. “All-Optical Nonequilibrium Pathway to Stabilising Magnetic Weyl Semimet- als in Pyrochlore Iridates”. In:Nature Communications9 (2018), p. 4452.doi:10.1038/ s41467-018-06991-8

    work page 2018

  7. [8]

    Non-Equilibrium Quantum Domain Reconfiguration Dynamics in a Two- Dimensional Electronic Crystal and a Quantum Annealer

    J. Vodeb et al. “Non-Equilibrium Quantum Domain Reconfiguration Dynamics in a Two- Dimensional Electronic Crystal and a Quantum Annealer”. In:Nature Communications15 (2024), p. 4836.doi:10.1038/s41467-024-49179-z

    work page doi:10.1038/s41467-024-49179-z 2024

  8. [9]

    Schusser, H

    J. Bedow, E. Mascot, and D. K. Morr. “Emergence and Manipulation of Non-Equilibrium Yu- Shiba-Rusinov States”. In:Communications Physics5 (2022), p. 281.doi:10.1038/s42005- 022-01050-7

    work page doi:10.1038/s42005- 2022

  9. [10]

    Evidence for Metastable Photo-Induced Superconductivity in K3C60

    M. Budden et al. “Evidence for Metastable Photo-Induced Superconductivity in K3C60”. In: Nature Physics17 (2021), p. 611.doi:10.1038/s41567-020-01148-1. 15

    work page doi:10.1038/s41567-020-01148-1 2021

  10. [11]

    Towards Single-Particle Spectroscopy of Small Metal Clusters

    A. Pohl, P.-G. Reinhard, and E. Suraud. “Towards Single-Particle Spectroscopy of Small Metal Clusters”. In:Phys. Rev. Lett.84 (2000), p. 5090.doi:10.1103/PhysRevLett.84. 5090

    work page doi:10.1103/physrevlett.84 2000

  11. [12]

    N., Averitt, R

    D. N. Basov, R. D. Averitt, and D. Hsieh. “Towards properties on demand in quantum materials”. en. In:Nature Materials16 (2017), p. 1077.doi:10.1038/nmat5017

    work page doi:10.1038/nmat5017 2017

  12. [14]

    Strongly correlated electron–photon systems

    J. Bloch et al. “Strongly correlated electron–photon systems”. en. In:Nature606 (2022), p. 41.doi:10.1038/s41586-022-04726-w

    work page doi:10.1038/s41586-022-04726-w 2022

  13. [15]

    Nature493(7430), 70–74 (2013) https://doi.org/10.1038/nature11567

    A. Schiffrin et al. “Optical-field-induced current in dielectrics”. en. In:Nature493 (Jan. 2013), p. 70.doi:10.1038/nature11567

    work page doi:10.1038/nature11567 2013

  14. [16]

    Ab Initio Simulation of Electrical Currents Induced by Ultrafast Laser Excitation of Dielectric Materials

    G. Wachter et al. “Ab Initio Simulation of Electrical Currents Induced by Ultrafast Laser Excitation of Dielectric Materials”. In:Phys. Rev. Lett.113 (2014), p. 087401.doi:10.1103/ PhysRevLett.113.087401

    work page 2014

  15. [17]

    Controlling dielectrics with the electric field of light

    M. Schultze et al. “Controlling dielectrics with the electric field of light”. en. In:Nature493 (2013), p. 75.doi:10.1038/nature11720

    work page doi:10.1038/nature11720 2013

  16. [18]

    The speed limit of optoelectronics

    M. Ossiander et al. “The speed limit of optoelectronics”. en. In:Nature Communications13 (2022), p. 1620.doi:10.1038/s41467-022-29252-1

    work page doi:10.1038/s41467-022-29252-1 2022

  17. [19]

    Photoenhanced metastable c-axis electrodynamics in stripe-ordered cuprate La 1.885 Ba 0.115 CuO 4

    K. A. Cremin et al. “Photoenhanced metastable c-axis electrodynamics in stripe-ordered cuprate La 1.885 Ba 0.115 CuO 4”. en. In:Proceedings of the National Academy of Sciences 116.40 (Oct. 2019), pp. 19875–19879.doi:10.1073/pnas.1908368116

    work page doi:10.1073/pnas.1908368116 2019

  18. [20]

    Time-Dependent Density-Matrix Renormalization Group: A Systematic Method for the Study of Quantum Many-Body Out-of-Equilibrium Systems

    M. A. Cazalilla and J. B. Marston. “Time-Dependent Density-Matrix Renormalization Group: A Systematic Method for the Study of Quantum Many-Body Out-of-Equilibrium Systems”. In:Phys. Rev. Lett.88 (2002), p. 256403.doi:10.1103/PhysRevLett.88.256403

    work page doi:10.1103/physrevlett.88.256403 2002

  19. [21]

    First-principles electron dynamics simulation for optical breakdown of di- electrics under an intense laser field

    T. Otobe et al. “First-principles electron dynamics simulation for optical breakdown of di- electrics under an intense laser field”. In:Phys. Rev. B77 (2008), p. 165104.doi:10.1103/ PhysRevB.77.165104

    work page 2008

  20. [22]

    Time-dependent restricted-active-space configuration-interaction method for the photoionization of many-electron atoms

    D. Hochstuhl and M. Bonitz. “Time-dependent restricted-active-space configuration-interaction method for the photoionization of many-electron atoms”. In:Phys. Rev. A86 (2012), p. 053424. doi:10.1103/PhysRevA.86.053424

    work page doi:10.1103/physreva.86.053424 2012

  21. [23]

    Time-Dependent Multiconfiguration Methods for the Numerical Simulation of Photoionization Processes of Many-Electron Atoms

    D. Hochstuhl, C. M. Hinz, and M. Bonitz. “Time-Dependent Multiconfiguration Methods for the Numerical Simulation of Photoionization Processes of Many-Electron Atoms”. In:The European Physical Journal Special Topics223 (2014), p. 177.doi:10.1140/epjst/e2014- 02092-3

    work page doi:10.1140/epjst/e2014- 2014

  22. [24]

    Communication: Time-dependent optimized coupled-cluster method for mul- tielectron dynamics

    T. Sato et al. “Communication: Time-dependent optimized coupled-cluster method for mul- tielectron dynamics”. In:The Journal of Chemical Physics148 (2018), p. 051101.doi:10. 1063/1.5020633

    work page 2018

  23. [25]

    Symplectic integration and physical interpretation of time- dependent coupled-cluster theory

    T. B. Pedersen and S. Kvaal. “Symplectic integration and physical interpretation of time- dependent coupled-cluster theory”. In:The Journal of Chemical Physics150 (2019), p. 144106. doi:10.1063/1.5085390

    work page doi:10.1063/1.5085390 2019

  24. [27]

    Achieving the Scaling Limit for Nonequilibrium Green Functions Simulations

    N. Schl¨ unzen, J.-P. Joost, and M. Bonitz. “Achieving the Scaling Limit for Nonequilibrium Green Functions Simulations”. In:Phys. Rev. Lett.124 (2020), p. 076601.doi:10.1103/ PhysRevLett.124.076601

    work page 2020

  25. [28]

    Dynamically screened ladder approximation: Simultaneous treatment of strong electronic correlations and dynamical screening out of equilibrium

    J.-P. Joost et al. “Dynamically screened ladder approximation: Simultaneous treatment of strong electronic correlations and dynamical screening out of equilibrium”. In:Phys. Rev. B 105 (2022), p. 165155.doi:10.1103/PhysRevB.105.165155

    work page doi:10.1103/physrevb.105.165155 2022

  26. [29]

    Real-Time G W : Toward anAb InitioDe- scription of the Ultrafast Carrier and Exciton Dynamics in Two-Dimensional Materials

    E. Perfetto, Y. Pavlyukh, and G. Stefanucci. “Real-Time G W : Toward anAb InitioDe- scription of the Ultrafast Carrier and Exciton Dynamics in Two-Dimensional Materials”. en. In:Physical Review Letters128 (2022), p. 016801.doi:10.1103/PhysRevLett.128.016801

    work page doi:10.1103/physrevlett.128.016801 2022

  27. [30]

    Time-linear scaling nonequilibrium Green’s function methods for real- time simulations of interacting electrons and bosons. I. Formalism

    Y. Pavlyukh et al. “Time-linear scaling nonequilibrium Green’s function methods for real- time simulations of interacting electrons and bosons. I. Formalism”. en. In:Physical Review B105 (2022), p. 125134.doi:10.1103/PhysRevB.105.125134

    work page doi:10.1103/physrevb.105.125134 2022

  28. [31]

    Propagating two-particle reduced density matrices without wave functions

    F. Lackner et al. “Propagating two-particle reduced density matrices without wave functions”. In:Phys. Rev. A91 (2015), p. 023412.doi:10.1103/PhysRevA.91.023412

    work page doi:10.1103/physreva.91.023412 2015

  29. [32]

    High-harmonic spectra from time-dependent two-particle reduced-density- matrix theory

    F. Lackner et al. “High-harmonic spectra from time-dependent two-particle reduced-density- matrix theory”. In:Phys. Rev. A95 (2017), p. 033414.doi:10.1103/PhysRevA.95.033414

    work page doi:10.1103/physreva.95.033414 2017

  30. [33]

    Schollwöck,The density-matrix renormalization group, Rev

    U. Schollw¨ ock. “The density-matrix renormalization group”. In:Rev. Mod. Phys.77 (2005), p. 259.doi:10.1103/RevModPhys.77.259

    work page doi:10.1103/revmodphys.77.259 2005

  31. [34]

    Haegeman , author C

    J. Haegeman et al. “Unifying time evolution and optimization with matrix product states”. In:Phys. Rev. B94 (2016), p. 165116.doi:10.1103/PhysRevB.94.165116

    work page doi:10.1103/physrevb.94.165116 2016

  32. [35]

    Time-dependent variational principle in matrix- product state manifolds: Pitfalls and potential

    B. Kloss, Y. B. Lev, and D. Reichman. “Time-dependent variational principle in matrix- product state manifolds: Pitfalls and potential”. In:Phys. Rev. B97 (2018), p. 024307.doi: 10.1103/PhysRevB.97.024307

    work page doi:10.1103/physrevb.97.024307 2018

  33. [36]

    Electro nic transport in two-dimensional graphene

    J. I. Cirac et al. “Matrix product states and projected entangled pair states: Concepts, sym- metries, theorems”. In:Rev. Mod. Phys.93 (2021), p. 045003.doi:10.1103/RevModPhys. 93.045003

    work page doi:10.1103/revmodphys 2021

  34. [37]

    Runge \ and\ author E

    E. Runge and E. K. U. Gross. “Density-Functional Theory for Time-Dependent Systems”. In:Phys. Rev. Lett.52 (1984), p. 997.doi:10.1103/PhysRevLett.52.997

    work page doi:10.1103/physrevlett.52.997 1984

  35. [38]

    Liesen, Z

    C. Ullrich.Time-Dependent Density-Functional Theory: Concepts and Applications. OUP Oxford, 2012. 541 pp.doi:10.1093/acprof:oso/9780199563029.001.0001

    work page doi:10.1093/acprof:oso/9780199563029.001.0001 2012

  36. [39]

    Stefanucci and R

    G. Stefanucci and R. van Leeuwen.Nonequilibrium Many-body Theory of Quantum Systems: A Modern Introduction. Cambridge University Press, Cambridge, 2025

    work page 2025

  37. [40]

    Nonequilibrium Correlation Dynamics in the One-Dimensional Fermi-Hubbard Model: A Testbed for the Two-Particle Reduced Density Matrix Theory

    S. Donsa et al. “Nonequilibrium Correlation Dynamics in the One-Dimensional Fermi-Hubbard Model: A Testbed for the Two-Particle Reduced Density Matrix Theory”. In:Physical Review Research5 (2023), p. 033022.doi:10.1103/PhysRevResearch.5.033022

    work page doi:10.1103/physrevresearch.5.033022 2023

  38. [41]

    Learning the non-Markovian features of subsys- tem dynamics

    M. Coppola, M. C. Ba˜ nuls, and Z. Lenarˇ ciˇ c. “Learning the non-Markovian features of subsys- tem dynamics”. In:SciPost Phys.19 (2025), p. 149.doi:10.21468/SciPostPhys.19.6.149

    work page doi:10.21468/scipostphys.19.6.149 2025

  39. [42]

    Neural network solution of non-Markovian quantum state diffusion and operator construction of quantum stochastic process

    J. Zhang, C. L. Benavides-Riveros, and L. Chen. “Neural network solution of non-Markovian quantum state diffusion and operator construction of quantum stochastic process”. In:The Journal of Chemical Physics163 (2025), p. 194103.doi:10.1063/5.0298594

    work page doi:10.1063/5.0298594 2025

  40. [43]

    Liu , author Z.-H

    W. Liu et al. “Predicting rate kernels via dynamic mode decomposition”. In:The Journal of Chemical Physics159 (2023), p. 144110.doi:10.1063/5.0170512. 17

    work page doi:10.1063/5.0170512 2023

  41. [44]

    Machine learning time-local generators of open quantum dynamics

    P. P. Mazza et al. “Machine learning time-local generators of open quantum dynamics”. In: Phys. Rev. Res.3 (2021), p. 023084.doi:10.1103/PhysRevResearch.3.023084

    work page doi:10.1103/physrevresearch.3.023084 2021

  42. [45]

    Learning quantum dissipation by the neural ordinary differential equa- tion

    L. Chen and Y. Wu. “Learning quantum dissipation by the neural ordinary differential equa- tion”. In:Physical Review A106 (2022).doi:10.1103/physreva.106.022201

    work page doi:10.1103/physreva.106.022201 2022

  43. [46]

    Unraveling quantum environments: Transformer-assisted learning in Lindblad dynamics

    C.-S. Chen and E.-J. Kuo. “Unraveling quantum environments: Transformer-assisted learning in Lindblad dynamics”. In:Phys. Rev. A112 (2025), p. 042227.doi:10.1103/gsxk-45mk

    work page doi:10.1103/gsxk-45mk 2025

  44. [47]

    Quantum-Tailored Machine-Learning Characterization of a Superconducting Qubit

    E. Genois et al. “Quantum-Tailored Machine-Learning Characterization of a Superconducting Qubit”. In:PRX Quantum2 (2021), p. 040355.doi:10.1103/PRXQuantum.2.040355

    work page doi:10.1103/prxquantum.2.040355 2021

  45. [48]

    Direct Entanglement Detection of Quantum Systems Using Machine Learn- ing

    Y. Huang et al. “Direct Entanglement Detection of Quantum Systems Using Machine Learn- ing”. In:npj Quantum Information11.1 (Feb. 20, 2025), p. 29.doi:10.1038/s41534-025- 00970-w

    work page doi:10.1038/s41534-025- 2025

  46. [49]

    From architectures to applications: a review of neural quantum states

    H. Lange et al. “From architectures to applications: a review of neural quantum states”. In: Quantum Science and Technology9 (2024), p. 040501.doi:10.1088/2058-9565/ad7168

    work page doi:10.1088/2058-9565/ad7168 2024

  47. [50]

    Simulating many-body open quantum systems by harnessing the power of artificial intelligence and quantum computing

    L. Ye, Y. Wang, and X. Zheng. “Simulating many-body open quantum systems by harnessing the power of artificial intelligence and quantum computing”. In:The Journal of Chemical Physics162.12 (Mar. 2025), p. 120901.doi:10.1063/5.0242648

    work page doi:10.1063/5.0242648 2025

  48. [51]

    Li, Y.-J

    Y. Chen and L. You. “Optimal Control of Unknown Collective Spin Systems via a Neural Network Surrogate”. In:Chinese Physics Letters42 (2025), p. 100601.doi:10.1088/0256- 307X/42/10/100601

    work page doi:10.1088/0256- 2025

  49. [52]

    Physics-Informed Neural Networks for Quantum Control

    A. Norambuena et al. “Physics-Informed Neural Networks for Quantum Control”. In:Phys. Rev. Lett.132 (2024), p. 010801.doi:10.1103/PhysRevLett.132.010801

    work page doi:10.1103/physrevlett.132.010801 2024

  50. [53]

    Dickerson

    P. Kidger et al. “Neural Controlled Differential Equations for Irregular Time Series”. In: Advances in Neural Information Processing Systems. Ed. by H. Larochelle et al. Vol. 33. Curran Associates, Inc., 2020, pp. 6696–6707.doi:https://doi.org/10.48550/arXiv. 2005.08926

    work page internal anchor Pith review doi:10.48550/arxiv 2020

  51. [54]

    Hamiltonian Generative Networks

    P. Toth et al. “Hamiltonian Generative Networks”. In:International Conference on Learning Representations. 2020.doi:https://doi.org/10.48550/arXiv.1909.13789

    work page doi:10.48550/arxiv.1909.13789 2020

  52. [55]

    Neural Rough Differential Equations for Long Time Series

    J. Morrill et al. “Neural Rough Differential Equations for Long Time Series”. In:Proceedings of the 38th International Conference on Machine Learning, ICML 2021, 18-24 July 2021, Virtual Event. Ed. by Marina Meila and Tong Zhang. Vol. 139. Proceedings of Machine Learning Research. PMLR, 2021, p. 7829.doi:https://doi.org/10.48550/arXiv.2009.08295

    work page doi:10.48550/arxiv.2009.08295 2021

  53. [56]

    Learning effective stochastic differential equations from microscopic simu- lations: Linking stochastic numerics to deep learning

    F. Dietrich et al. “Learning effective stochastic differential equations from microscopic simu- lations: Linking stochastic numerics to deep learning”. In:Chaos33 (2023), p. 023121.doi: 10.1063/5.0113632

    work page doi:10.1063/5.0113632 2023

  54. [57]

    Inferring Markovian quantum master equations of few-body observables in interacting spin chains

    F. Carnazza et al. “Inferring Markovian quantum master equations of few-body observables in interacting spin chains”. In:New Journal of Physics24 (2022), p. 073033.doi:10.1088/ 1367-2630/ac7df6

    work page 2022

  55. [58]

    Machine learning stochastic differential equations for the evolution of order parameters of classical many-body systems in and out of equilibrium

    F. Carnazza et al. “Machine learning stochastic differential equations for the evolution of order parameters of classical many-body systems in and out of equilibrium”. In:Machine Learning: Science and Technology5 (2024), p. 045002.doi:10.1088/2632-2153/ad7ad7

    work page doi:10.1088/2632-2153/ad7ad7 2024

  56. [59]

    Kobrin, Z

    M. Schmitt and M. Heyl. “Quantum Many-Body Dynamics in Two Dimensions with Artificial Neural Networks”. In:Phys. Rev. Lett.125 (2020), p. 100503.doi:10.1103/PhysRevLett. 125.100503. 18

    work page doi:10.1103/physrevlett 2020

  57. [60]

    Scaling of Neural-Network Quantum States for Time Evolution

    S.-H. Lin and F. Pollmann. “Scaling of Neural-Network Quantum States for Time Evolution”. In:physica status solidi (b)259 (2022).doi:10.1002/pssb.202100172

    work page doi:10.1002/pssb.202100172 2022

  58. [61]

    Zhao et al.Learning effective Hamiltonians for adaptive time-evolution quantum algo- rithms

    H. Zhao et al.Learning effective Hamiltonians for adaptive time-evolution quantum algo- rithms. 2024. arXiv:2406.06198.url:https://arxiv.org/abs/2406.06198

    work page arXiv 2024

  59. [62]

    Deep learning of many-body observables and quantum information scram- bling

    N. Mohseni et al. “Deep learning of many-body observables and quantum information scram- bling”. In:Quantum8 (2024), p. 1417.doi:10.22331/q-2024-07-18-1417

    work page doi:10.22331/q-2024-07-18-1417 2024

  60. [63]

    How Sophisticated Are Neural Networks Needed to Predict Long-Term Nonadiabatic Dynamics?

    H. Zeng, Y. Kou, and X. Sun. “How Sophisticated Are Neural Networks Needed to Predict Long-Term Nonadiabatic Dynamics?” In:Journal of Chemical Theory and Computation20 (2024). PMID: 39540684, p. 9832.doi:10.1021/acs.jctc.4c01223

    work page doi:10.1021/acs.jctc.4c01223 2024

  61. [64]

    Dual-Capability Machine Learning Models for Quantum Hamiltonian Parameter Estimation and Dynamics Prediction

    Z. An et al. “Dual-Capability Machine Learning Models for Quantum Hamiltonian Parameter Estimation and Dynamics Prediction”. In:Phys. Rev. Lett.134 (2025), p. 120202.doi:10. 1103/PhysRevLett.134.120202

    work page 2025

  62. [65]

    Self-accelerating beam dynamics in the space fractional schr¨ odinger equation.Phys

    R. Kaneko et al. “Forecasting long-time dynamics in quantum many-body systems by dynamic mode decomposition”. In:Phys. Rev. Res.7 (2025), p. 013085.doi:10.1103/PhysRevResearch. 7.013085

    work page doi:10.1103/physrevresearch 2025

  63. [66]

    Probing prethermal nonergodicity through measurement outcomes of mon- itored quantum dynamics

    Z.-H. Sun et al. “Probing prethermal nonergodicity through measurement outcomes of mon- itored quantum dynamics”. In:Phys. Rev. B112 (2025), p. L180306.doi:10.1103/v4xv- 74s7

    work page doi:10.1103/v4xv- 2025

  64. [67]

    M. S. Schmitt et al.Information theory for data-driven model reduction in physics and biology

    work page

  65. [68]

    arXiv:2312.06608.url:https://arxiv.org/abs/2312.06608

    work page arXiv

  66. [69]

    Cemin, M

    G. Cemin, M. Schmitt, and M. Bukov.Learning to stabilize nonequilibrium phases of matter with active feedback using partial information. 2025. arXiv:2508 . 06612.url:https : / / arxiv.org/abs/2508.06612

    work page arXiv 2025

  67. [70]

    Neural Ordinary Differential Equations

    R. T. Q. Chen et al. “Neural Ordinary Differential Equations”. In:Advances in Neural In- formation Processing Systems. Ed. by S. Bengio et al. Vol. 31. Curran Associates, Inc., 2018.url:https : / / proceedings . neurips . cc / paper _ files / paper / 2018 / file / 69386f6bb1dfed68692a24c8686939b9-Paper.pdf

    work page 2018

  68. [71]

    Integrating Expert ODEs into Neural ODEs: Pharmacology and Disease Progression

    Z. Qian et al. “Integrating Expert ODEs into Neural ODEs: Pharmacology and Disease Progression”. In:Advances in Neural Information Processing Systems. Ed. by M. Ranzato et al. Vol. 34. Curran Associates, Inc., 2021, p. 11364.url:https://proceedings.neurips. cc/paper_files/paper/2021/file/5ea1649a31336092c05438df996a3e59-Paper.pdf

    work page 2021

  69. [72]

    Alonso-Gonz\' a lez , author A

    J. M. Worsham and J. K. Kalita. “A Guide to Neural Ordinary Differential Equations: Machine Learning for Data-Driven Digital Engineering”. In:Digital Engineering6 (2025), p. 100060.doi:10.1016/j.dte.2025.100060

    work page doi:10.1016/j.dte.2025.100060 2025

  70. [73]

    Improving hydrologic models for predictions and process understanding using neural ODEs

    M. H¨ oge et al. “Improving hydrologic models for predictions and process understanding using neural ODEs”. In:Hydrology and Earth System Sciences26 (2022), p. 5085.doi:10.5194/ hess-26-5085-2022

    work page 2022

  71. [74]

    Beyond predictions in neural odes: Identification and interventions

    H. Aliee, F. J. Theis, and N. Kilbertus.Beyond Predictions in Neural ODEs: Identification and Interventions. 2025. arXiv:2106.12430.url:https://arxiv.org/abs/2106.12430

    work page arXiv 2025

  72. [75]

    Parameterized Neural Ordinary Differential Equations: Applications to Computational Physics Problems

    K. Lee and E. J. Parish. “Parameterized Neural Ordinary Differential Equations: Applications to Computational Physics Problems”. In:Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences477 (2021), p. 20210162.doi:10.1098/rspa.2021.0162. 19

    work page doi:10.1098/rspa.2021.0162 2021

  73. [76]

    Deep Learning for Survival Analysis: A Review

    H. Niu et al. “On the Applications of Neural Ordinary Differential Equations in Medical Image Analysis”. In:Artificial Intelligence Review57 (2024), p. 236.doi:10.1007/s10462- 024-10894-0

    work page doi:10.1007/s10462- 2024

  74. [77]

    Physics-Enhanced Neural Ordinary Differential Equations: Application to Industrial Chemical Reaction Systems

    F. Sorourifar et al. “Physics-Enhanced Neural Ordinary Differential Equations: Application to Industrial Chemical Reaction Systems”. In:Industrial & Engineering Chemistry Research 62 (2023), p. 15563.doi:10.1021/acs.iecr.3c01471

    work page doi:10.1021/acs.iecr.3c01471 2023

  75. [78]

    Low-Dimensional Neural ODEs and Their Application in Pharmacoki- netics

    D. S. Br¨ am et al. “Low-Dimensional Neural ODEs and Their Application in Pharmacoki- netics”. In:Journal of Pharmacokinetics and Pharmacodynamics51 (2024), p. 123.doi: 10.1007/s10928-023-09886-4

    work page doi:10.1007/s10928-023-09886-4 2024

  76. [79]

    Learning quantum dynamics with latent neural ordinary differential equa- tions

    M. Choi et al. “Learning quantum dynamics with latent neural ordinary differential equa- tions”. In:Phys. Rev. A105 (2022), p. 042403.doi:10.1103/PhysRevA.105.042403

    work page doi:10.1103/physreva.105.042403 2022

  77. [80]

    Neural ordinary differential equation and holographic quantum chromodynamics

    Koji Hashimoto, Hong-Ye Hu, and Yi-Zhuang You. “Neural ordinary differential equation and holographic quantum chromodynamics”. In:Machine Learning: Science and Technology 2.3 (May 2021), p. 035011.doi:10.1088/2632-2153/abe527

    work page doi:10.1088/2632-2153/abe527 2021

  78. [81]

    Solving the quantum many-body Hamiltonian learning problem with neural differential equations

    T. Heightman, E. Jiang, and A. Ac´ ın. “Solving the quantum many-body Hamiltonian learning problem with neural differential equations”. In:Quantum Science and Technology10 (2025), p. 045072.doi:10.1088/2058-9565/ae0d79

    work page doi:10.1088/2058-9565/ae0d79 2025

  79. [82]

    Data-driven time propagation of quantum systems with neural networks

    J. Nelson et al. “Data-driven time propagation of quantum systems with neural networks”. In:Phys. Rev. B106 (2022), p. 045402.doi:10.1103/PhysRevB.106.045402

    work page doi:10.1103/physrevb.106.045402 2022

  80. [83]

    S. J. Wetzel et al.Interpretable Machine Learning in Physics: A Review. 2025. arXiv:2503. 23616.url:https://arxiv.org/abs/2503.23616

    work page arXiv 2025

Showing first 80 references.