Pith. sign in

REVIEW 4 major objections 4 minor 57 references

Data-Driven Reconstruction and Characterization of Stochastic Dynamics via Dynamical Mode Decomposition

T0 review · 4 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read DMD modes can be read as ensemble-level weights that yield a PSD-like spectral fingerprint, a decoherence time, and stable extrapolation from short noisy trajectory ensembles.

desk verdict A clever, clearly written DMD-based noise fingerprint method that deserves peer review, but the central spectral mapping is asserted rather than proven and the validation is too soft to fully back the claims. read the letter →

arxiv 2507.05797 v3 pith:6SZQHFDS submitted 2025-07-08 quant-ph

Adva Baratz , Loris Maria Cangemi , Assaf Hamo , Sivan Refaely-Abramson , Amikam Levy This is my paper · ORCID
classification quant-ph
keywords dynamicalmodedecompositionstochastictrajectoryensemblesnoisespectralfingerprint1/fdecoherencetimeconstrainedextrapolationqubitdephasing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that Dynamical Mode Decomposition, a standard method for extracting oscillatory modes from time series, can be reinterpreted to characterize stochastic noise rather than stationary dynamics. By treating DMD modes as statistical weights over an ensemble of realizations and mapping their $\ell^1$ norms through a softmax nonlinearity, the method yields a normalized, PSD-like spectral fingerprint that separates white from correlated $1/f$ noise. The same DMD eigenvalue spectrum directly supplies the coherence time $T_2^*$, and a constrained reconstruction that caps eigenvalue magnitudes and uses the learned spectral weights as amplitudes stabilizes DMD extrapolation beyond the measurement window. The demonstration uses simulated qubit dephasing trajectories and requires no parametric noise model, no training, and no specialized control sequences.

What carries the argument

The load-bearing construction is the reinterpretation of DMD modes as ensemble-level statistical weights. Given matrices $X$ and $X'$ of $n$ stochastic trajectories shifted by one time step, DMD builds a reduced linear operator $A_r$ from the SVD of $X$; its eigenvectors define DMD modes $\phi_i$, each an $n$-dimensional vector in realization space. Summing each mode's complex magnitudes ($\|\phi_i\|_1$) and passing the result through the softmax function converts the mode norms into positive, normalized spectral weights $S_i$, so that the DMD eigenvalues' frequencies and the weights form a data-driven spectrum. The same eigenvalue set yields a decay time: with an odd rank, one real eigenvalue isolates the ensemble coherence envelope, providing $T_2^*$. For extrapolation, the eigenvalue magnitudes are rescaled to at most $|\lambda_{T_2^*}|$ while preserving phase, and the amplitudes $b_i$ are replaced by $S_i$, yielding the stabilized reconstruction formula of Eq. (8).

What would settle it

Run the pipeline on simulated ensembles with a known non-monotone noise spectrum, such as two well-separated Lorentzian bands of different strengths; if the extracted softmax weights do not place clear peaks at the true band frequencies with roughly the true height ratio, or if the weights visibly track the chosen DMD rank instead of the true spectrum, the spectral-fingerprint claim is refuted.

Watch

Extended reading notes

Core claim

The central claim is that the standard DMD eigenpairs $\{\lambda_i,\phi_i\}$ of a stochastic trajectory ensemble carry physical content if $\phi_i$ is read as a vector in realization space rather than in physical space. Taking the $\ell^1$-norm of each mode as an unnormalized score, the softmax mapping $S_i = \exp(\|\phi_i\|_1)/\sum_j \exp(\|\phi_j\|_1)$ produces a normalized spectral-weight distribution $\{\omega_i, S(\omega_i)\}$ that acts as a PSD-like fingerprint of the underlying noise: a sharp peak at the system frequency under weak white noise, a flat background under strong white noise, and low-frequency dominance under $1/f$ noise. The eigenvalue associated with a real DMD mode isolates the coherence envelope, giving $T_2^* = -1/\mathrm{Re}(\mu_{T_2^*})$ without fitting a decay function. Finally, the constrained reconstruction of Eq. (8), which caps all eigenvalue magnitudes by $|\lambda_{T_2^*}| = \exp(-\Delta t/T_2^*)$ and weights modes by $S_i$ instead of the initial-condition amplitudes $b_i$, removes the unstable growth that plagues standard DMD extrapolation and tracks the true ensemble-averaged dynamics beyond the analysis window.

Load-bearing premise

The entire spectral fingerprint rests on the unproven premise that, after truncating the DMD to a chosen rank, the $\ell^1$-norms of the DMD modes (before and after the softmax mapping) reflect the relative power of the underlying noise at each frequency; the only quantitative validation is a visual match, at one chosen rank, to the true $1/f$ spectrum after it has been smoothed by a Gaussian whose width is set by the paper's own formula.

Editorial extensions

If this is right

  • From short, noisy trajectory ensembles, the method yields a normalized spectral fingerprint that separates broadband (white) from correlated $1/f$ noise without any parametric assumption or training.
  • The decoherence time $T_2^*$ emerges as a distinct real DMD eigenvalue, so a coherence-time estimate can be read directly from the data without fitting a decay model.
  • The constrained reconstruction of Eq. (8) suppresses the exponential blowup of standard DMD extrapolation, keeping predictions stable and close to the true ensemble-averaged dynamics beyond the measurement window.
  • Because the analysis is formulated at the trajectory-ensemble level, the same construction applies to any stochastic process with comparable phase-accumulation structure, including classical oscillators with fluctuating instantaneous frequency.
  • The spectral fingerprint is a finite-data descriptor: its resolution depends on ensemble size, sampling, window length, noise level, and DMD rank, so in practice the extracted features should be checked for stability under rank and window variations.

Reading between the lines

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

  • A testable implication the authors leave implicit: the method should recover known multi-band spectra, not only monotone white and $1/f$ shapes; a two-Lorentzian or bandpass noise test would directly probe whether the softmax weights track spectral shape or merely rank order.
  • The softmax temperature $\beta$ offers a tunable contrast knob ($\beta>1$ sharpens peaks); the authors fix $\beta=1$ to stay parameter-free, but the ability to vary $\beta$ makes the fingerprint method robust to the degeneracy they describe, and could be used to quantify uncertainty in the extracted weights across rank choices.
  • Since the normalization discards absolute power information, the method gives a relative spectral fingerprint; combining it with a separate estimate of total noise power (e.g., from the coherence envelope) would yield an absolute PSD estimate without leaving the DMD framework.
  • The $T_2^*$ extraction relies on the coherence envelope being associable with a single real DMD eigenvalue; for multi-timescale or non-exponential decay processes, the same idea would produce a distribution of decay rates rather than a single number, which may be a feature rather than a bug.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The manuscript proposes a Dynamical Mode Decomposition (DMD)-based framework for analyzing ensembles of stochastic trajectories. It reinterprets the L1 norms of DMD modes as unnormalized statistical weights, maps them through a softmax function (Eq. 5) to a normalized 'PSD-like spectral fingerprint' of the noise, and claims to extract the coherence time T2* directly from the DMD eigenvalue spectrum (Section 4.4). It then introduces a constrained reconstruction (Eq. 8) that replaces standard DMD amplitudes with these spectral weights and clips eigenvalue magnitudes according to the extracted T2*, with the aim of stabilizing long-time DMD extrapolation. The method is demonstrated on simulated qubit dephasing under 1/f and white noise, including comparisons with Welch and Tikhonov estimators in the Supplementary Material.

Significance. If the central mapping were rigorously established, the framework would be an appealing model-free complement to noise spectroscopy: it operates directly on trajectory ensembles, requires no parametric noise model, and yields a spectral descriptor plus a coherence-time estimate from short records. The paper is clearly written, the benchmark material in the Supplementary Material is a useful contribution, and the Discussion is appropriately cautious about finite-data limitations. However, the central spectral-fingerprint claim currently rests on an unproven identification of softmax-transformed DMD mode norms with the noise PSD, and the only quantitative validation uses a smoothing kernel and a rank criterion chosen after the fact. The extrapolation claim also suffers from a missing amplitude normalization in Eq. (8). These issues are load-bearing and require additional derivation, quantitative validation, and a non-monotonic spectral test before the paper's central claims can be accepted.

major comments (4)
  1. [Section 4.1, Eq. (5)] The central identification of softmax(||phi_i||_1) with a PSD-like spectral weight distribution is asserted without a derivation or error bound. In addition, the DMD modes in Eq. (19) are eigenvectors of A_r, and the columns of W are only defined up to an arbitrary nonzero scale; unless a normalization convention for W (and hence phi_i) is specified, the L1 norms in Eq. (4) are not gauge-invariant and the softmax weights in Eq. (5) can change by an arbitrary amount under a rescaling of individual modes. Please specify the normalization, derive the claimed connection to the spectral density of the underlying stochastic process, and validate on a non-monotonic spectrum.
  2. [Section 4.3, Eq. (7), Fig. 5] The quantitative validation of the spectral fingerprint is weak. The true 1/f spectrum is convolved with a Gaussian whose FWHM is set by the authors' formula Eq. (7), and rank=20 is declared to give the 'best agreement' by visual inspection. With only ten natural frequencies at this rank and a smooth monotone 1/f curve, a wide family of monotone weight sets would pass the same test, so the comparison does not discriminate the method from alternative monotone weightings. Please provide a quantitative discrepancy metric and test the method on a non-monotonic or band-limited noise spectrum, for which the DMD resolution and the smoothing prescription do not predetermine the outcome.
  3. [Section 4.4, Fig. 6 (top)] The extraction of T2* as the single real eigenvalue of an odd-rank DMD is asserted without justification and without quantitative validation. DMD eigenvalue spectra of finite stochastic data need not contain a single real eigenvalue at odd rank, and even when one is present its identification with the ensemble dephasing envelope is not established by the by-eye comparison in Fig. 6. Please report the extracted T2* against the known simulation value for several ranks and noise realizations, with error bars, and discuss what happens when the real eigenvalue is not present or is not unique.
  4. [Section 4.5, Eq. (8)] The constrained reconstruction in Eq. (8) replaces the DMD amplitudes b_i with the softmax weights S_i, which by construction sum to one. As written, the formula therefore yields X_dmd-cons(t=0) = sum_i S_i = 1, not the initial amplitude of the physical observable, which appears in Fig. 6 only after the true and predicted dynamics are separately normalized to one. This missing overall amplitude factor means the extrapolation reproduces the shape of the decay but not its physical scale. Please include a data-driven amplitude prefactor (e.g., determined from the initial condition or from the reconstruction window) and assess the prediction without per-curve normalization.
minor comments (4)
  1. [Appendix A.1, Eqs. (14) and (15)] Equations (14) and (15) appear to be duplicated with identical content, including the repeated definition of the double-bracket ensemble average; please remove the duplicate.
  2. [Section 4.1, after Eq. (5)] The phrase 'parameter-free choice beta=1' is misleading because the method still relies on user selection of the rank r, the odd-rank choice for T2* extraction, and the smoothing FWHM in Eq. (7) for validation; please soften the wording or specify how these are determined from data-internal diagnostics.
  3. [Section 4.3, Fig. 5 caption] The caption states that rank=20 gives the 'closest agreement' but does not report a numerical error metric; adding a quantitative misfit (e.g., relative L2 error or KL divergence) for each rank would make the comparison reproducible.
  4. [General] No data or code availability statement is provided; making the simulation and DMD code available would strengthen the reproducibility of the claims.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the DMD spectral weights are a direct data transform validated against an external simulation, with only minor self-citation and a weak rank-dependent smoothing comparison.

full rationale

The derivation chain is not circular. The central spectral weights S_i in Eq. (5) are defined directly from the DMD mode norms ||phi_i||_1 and are never fitted to the true PSD; the comparison to the simulated 1/f spectrum is an external benchmark. The Gaussian smoothing in Eq. (7) adjusts the resolution of the reference spectrum, but it does not construct the DMD output from the reference, so no equation-level reduction occurs. The rank choice is partly guided by visual agreement in Fig. 5, which makes the validation test weak rather than circular, because the weights themselves are not optimized against the reference. The T2* extraction is an estimator of the decay rate of a DMD mode, and using that estimate as a bound in Eq. (8) is a regularization choice: the asymptotic envelope is constrained by the fitted T2*, but the extrapolated oscillation structure and spectral weighting are not algebraically equal to the input data. The only self-citation found ([1], supporting the generic statement that smaller singular values capture finer-scale features) is not load-bearing. Therefore, no step reduces to its own input by construction.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method introduces no new physical entities. Its outputs (spectral weights S_i, T2*, clamped eigenvalues) are derived from the DMD decomposition of the input data; the ledger lists the user-selected hyperparameters and the unproved assumptions that bridge DMD to the noise spectrum.

free parameters (3)
  • DMD rank r = r=20 chosen as optimal for 1/f comparison; r=15,25 for T2*; r=20-80 for white noise
    The spectral weights and T2* estimate depend on the chosen rank; rank is selected from SVD decay and reconstruction error, and for the 1/f comparison the rank that gives the best visual agreement with the smoothed true spectrum is declared optimal (Section 4.3).
  • Gaussian smoothing FWHM (Eq. 7) = Delta_f_in/(0.5*rank)
    Used to convolve the true 1/f PSD before comparison; the factor 0.5 and the explicit dependence on rank are chosen by the authors and directly control the apparent agreement.
  • Odd-rank choice for T2* extraction = 15 and 25
    A single real eigenvalue (identified as the T2* mode) appears only for odd DMD ranks; no stability test of the extracted T2* across ranks is provided.
assumptions (4)
  • domain assumption The ensemble of stochastic trajectories is approximately governed by a low-rank linear propagator X' ~= A~X, so DMD eigenvalues and modes carry physical meaning.
    Invoked in Section 3 and App. B; the paper itself acknowledges stochasticity challenges the stationarity assumption of standard DMD (Fig. 2 rank sensitivity).
  • ad hoc to paper softmax(||phi_i||_1) (Eq. 5) yields a PSD-like spectral-weight distribution of the noise.
    Introduced in Section 4.1 by analogy to machine-learning softmax; no derivation, bias, or consistency statement is given, and the output is a normalized relative weighting.
  • domain assumption For pure dephasing, every physical DMD mode should decay at least as fast as the T2* envelope, so clamping eigenvalue magnitudes to |lambda_T2*| is valid.
    Used in Section 4.5, Eq. (8); physically motivated for the simulated dephasing model but not derived for general stochastic dynamics.
  • ad hoc to paper With an odd DMD rank, the single real eigenvalue corresponds to the coherence decay mode T2*.
    Assumed in Section 4.4; validated only by visual comparison with the ensemble-average envelope (Fig. 6, top), with no quantitative metric or independence check.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Data-Driven Reconstruction and Characterization of Stochastic Dynamics via Dynamical Mode Decomposition." pith.science (2026). https://pith.science/paper/6SZQHFDS

@misc{pith2026250705797,
  author       = {Pith},
  title        = {Pith review of: Data-Driven Reconstruction and Characterization of Stochastic Dynamics via Dynamical Mode Decomposition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6SZQHFDS}},
  note         = {Machine review of arXiv:2507.05797}
}
abstract

Noise fundamentally limits the performance and predictive capabilities of classical and quantum dynamical systems by degrading stability and obscuring intrinsic dynamical characteristics. Characterizing such noise accurately is essential for enhancing measurement precision, understanding environmental interactions, and designing effective control strategies across diverse scientific and engineering domains. However, extracting the environment spectral features and associated characteristic decay or coherence times from limited and noisy datasets remains challenging. Here, we introduce a general, data-driven framework based on Dynamical Mode Decomposition (DMD) to analyze system dynamics under stochastic noise. We reinterpret DMD modes as statistical weights over an ensemble of stochastic trajectories and, via a nonlinear mapping, construct a PSD-like spectral fingerprint of the noise. This enables the identification of dominant frequency contributions in both broadband (white) and correlated ($1/f$) noise environments, as well as direct extraction of intrinsic characteristic decay times from DMD eigenvalues. To overcome instability in standard DMD-based extrapolation, we develop a constrained reconstruction method using extracted decay times as physical bounds and the learned noise as spectral weights. We demonstrate the effectiveness of this approach through simulations of quantum system dynamics subject to decoherence from noise, demonstrating its robustness and predictive capabilities and comparing it with standard methods. This methodology provides a trajectory-level framework for diagnostic and predictive analysis of stochastic processes from limited, noisy time-series data.

Figures

Figures reproduced from arXiv: 2507.05797 by the authors.

Figure 1
Figure 1. A schematic illustration of our data-driven approach shows how stochastic input data is transformed into physically informative properties. An ensemble of realizations measured over 0 ≤ 𝑡 ≤ 𝑡𝑚 is mapped by DMD into a reduced set of spatial-temporal modes, which reconstruct the original data with minimal error. From this, we extract three key physical properties: the noise spectral weights (S(𝜔)), coherence decay (𝑇 … view at source ↗
Figure 2
Figure 2. Comparison between the average over all stochastic realizations, 𝑋Avg (solid blue line), and the average of the reconstructed DMD trajectories, 𝑋 Avg DMD (dashed red line), obtained from Eq. (2), is shown for rank = 9 (upper-left), 10 (upper-right), and 20 (bottom-left). The case with rank = 20 yields the best alignment. The singular values from the SVD analysis reveal eight dominant components (bottom￾right panel).… view at source ↗
Figure 3
Figure 3. Schematic overview of our reinterpreted DMD-based noise characterization scheme. The upper block shows the input data matrix 𝑋𝑛×𝑚 , composed of n stochastic trajectories computed using Eq. (1), each representing the expectation value of 𝜎𝑥 over time. This matrix is decomposed into DMD eigenvalues and spatial modes {𝜆𝑖 , 𝜙𝑖 }, where each 𝜙𝑖 is interpreted as a vector in realization space. In the lower block, the 𝓁1 -… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: The spectral weights extracted using our DMD￾based scheme for a simulated quantum system with a charac￾teristic frequency 𝑓0 = 1 MHz, for white noise (left) and 1∕𝑓 noise (right) environments. Spectral reconstructions were gen￾erated using DMD with selected ranks: 20, …
Figure 5
Figure 5. Figure 5: Comparison between the DMD-derived 1∕𝑓 spectrum (black) and a Gaussian convolution of the true (simulated) 1∕𝑓 noise spectrum (red). The DMD-derived spectral weights are compared with an inverse-frequency trend evaluated at the DMD-derived natural frequencies, while th…
Figure 6
Figure 6. Figure 6: Top panel: Average stochastic evolution of the system (black solid line) compared with the dynamics of the DMD mode associated with 𝑇 ∗ 2 , given by 𝑏𝑇 ∗ 2 𝑒 𝜇𝑇 ∗ 2 𝑡 (gray dashed line). Results are shown for two odd DMD ranks: 𝑅 = 15 (left) and 𝑅 = 25 (right). Middle …
Figure 7
Figure 7. Figure 7: Root-mean-square reconstruction error as a func￾tion of DMD rank for six measurement durations Δ𝑡 = 2.0, 2.5, 3.0, 4.0, 5.0, 7.0 𝜇s. Longer acquisition windows exhibit their RMSE minima at higher ranks, indicating that extended observation times reveal additional coher…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 43 canonical work pages

  1. [1]

    Unsupervisedlearningapproachtoquantumwave-packetdynamicsfromcoupledtemporal-spatial correlations

    Baratz,A.,Cohen,G.,Refaely-Abramson,S.,2024. Unsupervisedlearningapproachtoquantumwave-packetdynamicsfromcoupledtemporal-spatial correlations. Physical Review B 110, 134304

    work page 2024

  2. [2]

    Adynamicmodedecompositionframeworkforglobalpowersystemoscillationanalysis

    Barocio,E.,Pal,B.C.,Thornhill,N.F.,Messina,A.R.,2015. Adynamicmodedecompositionframeworkforglobalpowersystemoscillationanalysis. IEEE Transactions on Power Systems 30, 2902–2912. doi:10.1109/TPWRS.2014.2368078

    work page arXiv 2015

  3. [3]

    Learning long-term dependencies with gradient descent is difficult

    Bengio, Y., Simard, P., Frasconi, P., 1994. Learning long-term dependencies with gradient descent is difficult. IEEE transactions on neural networks 5, 157–166

    work page 1994

  4. [4]

    Decoherence in qubits due to low-frequency noise

    Bergli, J., Altshuler, B., 2009. Decoherence in qubits due to low-frequency noise. New Journal of Physics 11, 025002

    work page 2009

  5. [5]

    Stochastic theory of line shape: Generalization of the kubo-anderson model

    Blume, M., 1968. Stochastic theory of line shape: Generalization of the kubo-anderson model. Phys. Rev. 174, 351–358

    work page 1968

  6. [6]

    Extracting spatial–temporal coherent patterns in large-scale neural recordings using dynamic mode decomposition

    Brunton, B.W., Johnson, L.A., Ojemann, J.G., Kutz, J.N., 2016. Extracting spatial–temporal coherent patterns in large-scale neural recordings using dynamic mode decomposition. Journal of Neuroscience Methods 258, 1–15. URL:https://www.sciencedirect.com/science/article/pii/ S0165027015003829, doi:https://doi.org/10.1016/j.jneumeth.2015.10.010

    doi:10.1016/j.jneumeth.2015.10.010 2016

  7. [7]

    Data-Driven Science and Engineering: Machine Learning, Dynamical Systems, and Control

    Brunton, S.L., Kutz, J.N., 2019. Data-Driven Science and Engineering: Machine Learning, Dynamical Systems, and Control. 1st ed., Cambridge University Press, USA

    work page 2019

  8. [8]

    Non-markovian qubit dynamics in the presence of 1/f noise

    Burkard, G., 2009. Non-markovian qubit dynamics in the presence of 1/f noise. Physical Review B—Condensed Matter and Materials Physics 79, 125317

    work page 2009

Show all 57 references
  1. [9]

    Noisespectroscopy through dynamical decoupling with a superconducting flux qubit

    Bylander,J.,Gustavsson,S.,Yan,F.,Yoshihara,F.,Harrabi,K.,Fitch,G.,Cory,D.G.,Nakamura,Y.,Tsai,J.S.,Oliver,W.D.,2011. Noisespectroscopy through dynamical decoupling with a superconducting flux qubit. Nature Physics 7, 565–570

    2011

  2. [10]

    Theory and experimental demonstration of quantum invariant filtering

    Cangemi, L.M., Woldiger, Y., Levy, A., Hamo, A., 2025. Theory and experimental demonstration of quantum invariant filtering. arXiv preprint arXiv:2506.15805

    2025 arXiv

  3. [11]

    Regularization of inverse problems

    Clason, C., 2021. Regularization of inverse problems. URL:https://arxiv.org/abs/2001.00617,arXiv:2001.00617

    2021 arXiv

  4. [12]

    Introductiontoquantumnoise,measurement,andamplification

    Clerk,A.A.,Devoret,M.H.,Girvin,S.M.,Marquardt,F.,Schoelkopf,R.J.,2010. Introductiontoquantumnoise,measurement,andamplification. Rev. Mod. Phys. 82, 1155–1208. URL:https://link.aps.org/doi/10.1103/RevModPhys.82.1155, doi:10.1103/RevModPhys.82.1155

    2010 doi

  5. [13]

    Quantumfilterdiagonalizationwithcompresseddouble-factorizedhamiltonians

    Cohn,J.,Motta,M.,Parrish,R.M.,2021. Quantumfilterdiagonalizationwithcompresseddouble-factorizedhamiltonians. PRXQuantum2,040352

    2021

  6. [14]

    Phase noise in oscillators: A unifying theory and numerical methods for characterization

    Demir, A., Mehrotra, A., Roychowdhury, J., 2000. Phase noise in oscillators: A unifying theory and numerical methods for characterization. IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications 47, 655–674. doi:10.1109/81.847872

    2000 doi

  7. [15]

    Machine learning & artificial intelligence in the quantum domain: a review of recent progress

    Dunjko, V., Briegel, H.J., 2018. Machine learning & artificial intelligence in the quantum domain: a review of recent progress. Reports on Progress in Physics 81, 074001. URL:https://dx.doi.org/10.1088/1361-6633/aab406, doi:10.1088/1361-6633/aab406

    2018 doi

  8. [16]

    Finding structure in time

    Elman, J.L., 1990. Finding structure in time. Cognitive science 14, 179–211

    1990

  9. [17]

    Initial decoherence in solid state qubits

    Falci, G., D’arrigo, A., Mastellone, A., Paladino, E., 2005. Initial decoherence in solid state qubits. Physical review letters 94, 167002

    2005

  10. [18]

    Bayesian quantum noise spectroscopy

    Ferrie, C., Granade, C., Paz-Silva, G., Wiseman, H.M., 2018. Bayesian quantum noise spectroscopy. New Journal of Physics 20, 123005. URL: https://doi.org/10.1088/1367-2630/aaf207, doi:10.1088/1367-2630/aaf207

    2018 doi

  11. [19]

    Handbook of Stochastic Methods for Physics, Chemistry and the Natural Sciences

    Gardiner, C., 1985. Handbook of Stochastic Methods for Physics, Chemistry and the Natural Sciences. Springer-Verlag, Berlin

    1985

  12. [20]

    Quantum Noise

    Gardiner, C., Zoller, P., 2004. Quantum Noise. Springer, Berlin

    2004

  13. [21]

    Handbook of stochastic methods

    Gardiner, C.W., et al., 2004. Handbook of stochastic methods. volume 3. springer Berlin

    2004

  14. [22]

    Universal dynamical decoherence control of noisy single-and multi-qubit systems

    Gordon, G., Erez, N., Kurizki, G., 2007. Universal dynamical decoherence control of noisy single-and multi-qubit systems. Journal of Physics B: Atomic, Molecular and Optical Physics 40, S75

    2007

  15. [23]

    Optimal dynamical decoherence control of a qubit

    Gordon, G., Kurizki, G., Lidar, D.A., 2008. Optimal dynamical decoherence control of a qubit. Physical review letters 101, 010403

    2008

  16. [24]

    Numericalanalyticcontinuation:Answerstowell-posedquestions

    Goulko,O.,Mishchenko,A.S.,Pollet,L.,Prokof’ev,N.,Svistunov,B.,2017. Numericalanalyticcontinuation:Answerstowell-posedquestions. Phys. Rev. B 95, 014102. doi:10.1103/PhysRevB.95.014102

    2017 doi

  17. [25]

    Colored noise in dynamical systems

    Hänggi, P., Jung, P., 1994. Colored noise in dynamical systems. Advances in chemical physics 89, 239–326

    1994

  18. [26]

    Long short-term memory

    Hochreiter, S., Schmidhuber, J., 1997. Long short-term memory. Neural computation 9, 1735–1780

    1997

  19. [27]

    Bayesianinferenceandtheanalyticcontinuationofimaginary-timequantummontecarlodata

    Jarrell,M.,Gubernatis,J.,1996. Bayesianinferenceandtheanalyticcontinuationofimaginary-timequantummontecarlodata. PhysicsReports269, 133–195. URL:https://www.sciencedirect.com/science/article/pii/0370157395000747, doi:https://doi.org/10.1016/0370-1573(95)00074-7

    1996

  20. [28]

    Principal Component Analysis

    Jolliffe, I.T., 2002. Principal Component Analysis. 2nd ed., Springer, New York. A. Baratz et al.:Preprint submitted to ElsevierPage 16 of 17 Stochastic dynamics characterization via DMD

    2002

  21. [29]

    URL:https://www.sciencedirect.com/science/article/pii/S0167947311004099, doi:https://doi.org/10

    Josse,J.,Husson,F.,2012.Selectingthenumberofcomponentsinprincipalcomponentanalysisusingcross-validationapproximations.Computational Statistics & Data Analysis 56, 1869–1879. URL:https://www.sciencedirect.com/science/article/pii/S0167947311004099, doi:https://doi.org/10. 1016/j...

    2012

  22. [30]

    Forecasting long-time dynamics in quantum many-body systems by dynamic mode decomposition.Phys.Rev.Res.7,013085

    Kaneko, R., Imada, M., Kabashima, Y., Ohtsuki, T., 2025. Forecasting long-time dynamics in quantum many-body systems by dynamic mode decomposition.Phys.Rev.Res.7,013085. URL:https://link.aps.org/doi/10.1103/PhysRevResearch.7.013085,doi:10.1103/PhysRevResearch.7.013085

    2025 doi

  23. [31]

    Numerical Solution to Stochastic Differential Equations

    Kloeden, P., Platen, E., 1999. Numerical Solution to Stochastic Differential Equations. Springer, Berlin

    1999

  24. [32]

    Artificial intelligence and machine learning for quantum technologies

    Krenn, M., Landgraf, J., Foesel, T., Marquardt, F., 2023. Artificial intelligence and machine learning for quantum technologies. Phys. Rev. A 107, 010101. URL:https://link.aps.org/doi/10.1103/PhysRevA.107.010101, doi:10.1103/PhysRevA.107.010101

    2023 doi

  25. [33]

    A stochastic theory of line shape, in: Shuler, K.E

    Kubo, R., 1969. A stochastic theory of line shape, in: Shuler, K.E. (Ed.), Advances in Chemical Physics: Stochastic Processes in Chemical Physics. John Wiley & Sons. volume 15, pp. 101–127. doi:10.1002/9780470143605.ch6

    1969 doi

  26. [34]

    Dynamic mode decomposition: data-driven modeling of complex systems

    Kutz, J.N., Brunton, S.L., Brunton, B.W., Proctor, J.L., 2016. Dynamic mode decomposition: data-driven modeling of complex systems. SIAM

    2016

  27. [35]

    Deep learning

    LeCun, Y., Bengio, Y., Hinton, G., 2015. Deep learning. Nature 521, 436–444. doi:10.1038/nature14539

    2015 doi

  28. [36]

    Lecture notes on the theory of open quantum systems.arXiv:1902.00967

    Lidar, D.A., 2020. Lecture notes on the theory of open quantum systems.arXiv:1902.00967

    2020 arXiv

  29. [37]

    Fdm: the filter diagonalization method for data processing in nmr experiments

    Mandelshtam, V.A., 2001. Fdm: the filter diagonalization method for data processing in nmr experiments. Progress in Nuclear Magnetic Resonance Spectroscopy 38, 159–196

    2001

  30. [38]

    Dynamic mode decomposition for analysis of time-series data

    Marusic, I., 2024. Dynamic mode decomposition for analysis of time-series data. Journal of Fluid Mechanics 1000, F7. doi:10.1017/jfm.2024.834

    2024 doi

  31. [39]

    Stochasticoptimizationmethodforanalyticcontinuation,in:Correlatedelectrons:frommodelstomaterials.Forschungszen- trum Jülich GmbH Institute for Advanced Simulation

    Mishchenko,A.S.,2012. Stochasticoptimizationmethodforanalyticcontinuation,in:Correlatedelectrons:frommodelstomaterials.Forschungszen- trum Jülich GmbH Institute for Advanced Simulation

    2012

  32. [40]

    Bound state eigenfunctions from wave packets: Time→energy resolution

    Neuhauser, D., 1990. Bound state eigenfunctions from wave packets: Time→energy resolution. The Journal of chemical physics 93, 2611–2616

    1990

  33. [41]

    Qubit noise spectroscopy for non-gaussian dephasing environments

    Norris, L.M., Paz-Silva, G.A., Viola, L., 2016. Qubit noise spectroscopy for non-gaussian dephasing environments. Physical review letters 116, 150503

    2016

  34. [42]

    Decoherence and 1/f noise in josephson qubits

    Paladino, E., Faoro, L., Falci, G., Fazio, R., 2002. Decoherence and 1/f noise in josephson qubits. Physical review letters 88, 228304

    2002

  35. [43]

    1/f noise: Implications for solid-state quantum information

    Paladino, E., Galperin, Y., Falci, G., Altshuler, B., 2014. 1/f noise: Implications for solid-state quantum information. Reviews of Modern Physics 86, 361–418

    2014

  36. [44]

    A technique for the numerical solution of certain integral equations of the first kind

    Phillips, D.L., 1962. A technique for the numerical solution of certain integral equations of the first kind. J. ACM 9, 84–97. URL:https: //doi.org/10.1145/321105.321114, doi:10.1145/321105.321114

    1962

  37. [45]

    Dynamicmodedecompositionwithcontrol

    Proctor,J.L.,Brunton,S.L.,Kutz,J.N.,2016. Dynamicmodedecompositionwithcontrol. SIAMJournalonAppliedDynamicalSystems15,142–161

    2016

  38. [46]

    Discovering dynamic patterns from infectious disease data using dynamic mode decomposition

    Proctor, J.L., Eckhoff, P.A., 2015. Discovering dynamic patterns from infectious disease data using dynamic mode decomposition. International Health 7, 139–145

    2015

  39. [47]

    Optimal noise-canceling networks

    Ronellenfitsch, H., Dunkel, J., Wilczek, M., 2018. Optimal noise-canceling networks. Physical Review Letters 121, 208301

    2018

  40. [48]

    Spectralanalysisofnonlinearflows

    Rowley,C.W.,Mezić,I.,Bagheri,S.,Schlatter,P.,Henningson,D.S.,2009. Spectralanalysisofnonlinearflows. JournalofFluidMechanics641,115 – 127. URL:https://api.semanticscholar.org/CorpusID:1546931

    2009

  41. [49]

    Dynamic mode decomposition of numerical and experimental data

    Schmid, P.J., 2010. Dynamic mode decomposition of numerical and experimental data. Journal of fluid mechanics 656, 5–28

    2010

  42. [50]

    Dynamic mode decomposition and its variants

    Schmid, P.J., 2022. Dynamic mode decomposition and its variants. Annual Review of Fluid Mechanics 54, 225–254

    2022

  43. [51]

    Regularization of ill-posed problems

    Tikhonov, A., 1963. Regularization of ill-posed problems. Doklady Akademi Nauk USSR 152, 49–52

    1963

  44. [52]

    On dynamic mode decomposition: Theory and applications

    Tu, J.H., Rowley, C.W., Luchtenburg, D.M., Brunton, S.L., Kutz, J.N., 2014. On dynamic mode decomposition: Theory and applications. Journal of Computational Dynamics 1, 391–421. URL:https://www.aimsciences.org/article/id/1dfebc20-876d-4da7-8034-7cd3c7ae1161, doi:10.3934/jcd.20...

    2014 doi

  45. [53]

    Wall, M.R., Neuhauser, D., 1995. Extraction, through filter-diagonalization, of general quantum eigenvalues or classical normal mode frequencies fromasmallnumberofresiduesorashort-timesegmentofasignal.i.theoryandapplicationtoaquantum-dynamicsmodel. TheJournalofchemical physics...

    1995

  46. [54]

    The use of fast fourier transform for the estimation of power spectra: A method based on time averaging over short, modified periodograms

    Welch, P.D., 1967. The use of fast fourier transform for the estimation of power spectra: A method based on time averaging over short, modified periodograms. IEEE Transactions on Audio and Electroacoustics 15, 70–73. doi:10.1109/TAU.1967.1161901

    1967

  47. [55]

    Challenges in dynamic mode decomposition

    Wu, Z., Brunton, S.L., Revzen, S., 2021. Challenges in dynamic mode decomposition. Journal of the Royal Society Interface 18, 20210686

    2021

  48. [56]

    Entanglement evolution in a non-markovian environment

    Yu, T., Eberly, J., 2010. Entanglement evolution in a non-markovian environment. Optics Communications 283, 676–680. doi:https://doi.org/10. 1016/j.optcom.2009.10.042

    2010

  49. [57]

    Generative quantum machine learning via denoising diffusion probabilistic models

    Zhang, B., Xu, P., Chen, X., Zhuang, Q., 2024. Generative quantum machine learning via denoising diffusion probabilistic models. Phys. Rev. Lett. 132, 100602. URL:https://link.aps.org/doi/10.1103/PhysRevLett.132.100602, doi:10.1103/PhysRevLett.132.100602. A. Baratz et al.:Prep...

    2024 doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.