pith. sign in

arxiv: 2508.11175 · v3 · submitted 2025-08-15 · 🪐 quant-ph · cs.LG· eess.SP

The Role of Entanglement in Quantum Reservoir Computing with Coupled Kerr Nonlinear Oscillators

Ali Karimi , Hadi Zadeh-Haghighi , Youssef Kora , Christoph Simon This is my paper

Pith reviewed 2026-05-18 23:33 UTC · model grok-4.3

classification 🪐 quant-ph cs.LGeess.SP
keywords quantum reservoir computingentanglementKerr nonlinear oscillatorstime-series predictionlogarithmic negativityNRMSEquantum machine learning
0
0 comments X

The pith

Moderate non-zero entanglement optimizes quantum reservoir performance in coupled Kerr oscillators

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

This paper explores quantum reservoir computing with two coupled Kerr nonlinear oscillators for time-series forecasting tasks. It finds that predictive accuracy, measured by normalized root mean square error, peaks at moderate but non-zero entanglement levels quantified by logarithmic negativity. Individual parameter sweeps over input drive, Kerr nonlinearity, and coupling, plus aggregated binned analysis, show this moderate entanglement links to the best average performance across the space. The pattern holds up to a threshold in input frequency and under moderate dissipation and dephasing, with higher dissipation sometimes improving results.

Core claim

In this quantum reservoir computing framework based on two coupled Kerr nonlinear oscillators, optimal performance for forecasting both linear and nonlinear time series occurs at moderate but non-zero entanglement. Using logarithmic negativity to quantify entanglement and NRMSE to evaluate accuracy, parameter sweeps and binned analysis across the space confirm that moderate entanglement consistently associates with superior average predictive performance, persisting up to a threshold in input frequency and under some dissipation and dephasing.

What carries the argument

Two coupled Kerr nonlinear oscillators as the quantum reservoir, with entanglement (logarithmic negativity) tied to computational performance (NRMSE) for temporal data processing.

If this is right

  • Tuning drive strength, nonlinearity, and coupling to target moderate entanglement can optimize forecasting accuracy.
  • The performance benefit persists under moderate dissipation and dephasing, indicating some robustness for physical implementations.
  • Higher dissipation rates can sometimes enhance rather than degrade predictive performance in this setup.
  • The approach supports design of quantum systems for efficient processing of temporal data in machine learning tasks.

Where Pith is reading between the lines

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

  • Similar performance peaks might appear in classical nonlinear oscillator networks if entanglement is not required for the effect.
  • Scaling to more than two oscillators could test whether the moderate-entanglement optimum changes with system size.
  • Hardware experiments with tunable coupling and controllable dissipation would directly check the predicted performance curve.

Load-bearing premise

That logarithmic negativity and NRMSE together capture the computationally relevant quantum features without other unmeasured correlations or classical effects driving the performance differences.

What would settle it

Repeated sweeps where minimal NRMSE occurs at zero entanglement or maximal entanglement instead of moderate levels would falsify the central claim.

Figures

Figures reproduced from arXiv: 2508.11175 by Ali Karimi, Christoph Simon, Hadi Zadeh-Haghighi, Youssef Kora.

Figure 1
Figure 1. Figure 1: Throughout this work, we choose δt = 100 and [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Time evolution of entanglement (top row) and corresponding time-series prediction performance (bottom row) at [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. NRMSE (left) and entanglement (right) versus input [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4. NRMSE (left) and Entanglement (right) versus Kerr [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5. NRMSE (left) and Entanglement (right) versus cou [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6. Prediction error (NRMSE) vs. entanglement in the case of zero dephasing ( [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7. Minimum NRMSE at optimal points vs. input fre [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8. Minimum NRMSE at optimal points vs. input fre [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10. Same as Fig [PITH_FULL_IMAGE:figures/full_fig_p010_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: FIG. 11. Phonon number Vs different input strengths for [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗
read the original abstract

Quantum Reservoir Computing (QRC) uses quantum dynamics to efficiently process temporal data. In this work, we investigate a QRC framework based on two coupled Kerr nonlinear oscillators, a system well-suited for time-series prediction tasks due to its complex nonlinear interactions and potentially high-dimensional state space. We explore how its performance in forecasting both linear and nonlinear time-series depends on key physical parameters: input drive strength, Kerr nonlinearity, and oscillator coupling, and analyze the role of entanglement in improving the reservoir's computational performance, focusing on its effect on predicting non-trivial time series. Using logarithmic negativity to quantify entanglement and normalized root mean square error (NRMSE) to evaluate predictive accuracy, individual parameter sweeps show that optimal performance occurs at moderate but non-zero entanglement. Furthermore, an aggregated binned analysis reveals that this moderate entanglement is consistently associated with the optimal average predictive performance across the parameter space, an observation that persists up to a threshold in the input frequency. This relationship persists under some levels of dissipation and dephasing. In particular, we find that higher dissipation rates can enhance performance. These findings contribute to the broader understanding of quantum reservoirs for high performance, efficient quantum machine learning and time-series forecasting.

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 investigates quantum reservoir computing using two coupled Kerr nonlinear oscillators for time-series forecasting. It examines performance dependence on input drive strength, Kerr nonlinearity, and coupling strength, quantifying entanglement via logarithmic negativity and predictive accuracy via normalized root mean square error (NRMSE). Through individual parameter sweeps and an aggregated binned analysis, the central claim is that optimal performance occurs at moderate but non-zero entanglement levels, with this association persisting up to a threshold in input frequency and under some dissipation and dephasing; higher dissipation is also reported to enhance performance in some regimes.

Significance. If the reported association between moderate entanglement and optimal NRMSE is shown to be robust rather than an artifact of parameter correlations, the result would contribute to clarifying the computational utility of quantum features in reservoir computing, offering guidance for parameter regimes in quantum machine learning implementations. The numerical exploration of a physically motivated oscillator model and the use of standard entanglement and error metrics provide a concrete starting point for such investigations.

major comments (2)
  1. [Binned analysis] Binned analysis (results section): The aggregated binning of NRMSE versus logarithmic negativity across the (drive, Kerr, coupling) space does not report the distribution or range of the underlying parameters within each entanglement bin. Because the same parameters control both the generated entanglement and the nonlinear map/memory capacity of the reservoir, the observed minimum in average NRMSE at moderate negativity could arise from the dynamical regimes that happen to produce those negativity values rather than from entanglement itself. Fixed-parameter-slice binning or conditional analysis would be required to separate these effects.
  2. [Results and discussion] Performance evaluation: No baseline comparisons (e.g., against classical reservoirs or linear readouts) or explicit error bars/statistical tests on the binned averages are described, weakening the support for the claim that moderate entanglement is 'consistently associated with the optimal average predictive performance'.
minor comments (2)
  1. [Abstract] The abstract states that the moderate-entanglement optimum 'persists up to a threshold in the input frequency' and that 'higher dissipation rates can enhance performance'; these statements should be tied to specific quantitative thresholds or figure panels for clarity.
  2. [Methods] Notation for the input drive and frequency parameters should be defined explicitly when first introduced to avoid ambiguity in the parameter-sweep descriptions.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive feedback on our manuscript. We address each major comment below and have revised the manuscript to strengthen the presentation of the binned analysis and performance evaluation.

read point-by-point responses

  1. Referee: [Binned analysis] Binned analysis (results section): The aggregated binning of NRMSE versus logarithmic negativity across the (drive, Kerr, coupling) space does not report the distribution or range of the underlying parameters within each entanglement bin. Because the same parameters control both the generated entanglement and the nonlinear map/memory capacity of the reservoir, the observed minimum in average NRMSE at moderate negativity could arise from the dynamical regimes that happen to produce those negativity values rather than from entanglement itself. Fixed-parameter-slice binning or conditional analysis would be required to separate these effects.

    Authors: We agree that the shared dependence on the same parameters raises the possibility of confounding effects, and that reporting parameter distributions within bins would help clarify the analysis. In the revised manuscript we have added histograms showing the ranges and distributions of drive strength, Kerr nonlinearity, and coupling strength within each logarithmic-negativity bin. We have also included supplementary fixed-parameter-slice results at representative values of the other parameters to illustrate that the association between moderate negativity and lower NRMSE persists when individual parameters are held fixed. revision: yes

  2. Referee: [Results and discussion] Performance evaluation: No baseline comparisons (e.g., against classical reservoirs or linear readouts) or explicit error bars/statistical tests on the binned averages are described, weakening the support for the claim that moderate entanglement is 'consistently associated with the optimal average predictive performance'.

    Authors: We accept that explicit error bars and a baseline comparison would improve the robustness of the claim. We have added standard-deviation error bars to all binned-average NRMSE plots and performed a simple statistical comparison (two-sample t-test) between the moderate-negativity bin and the adjacent bins, reporting the p-values in the revised text. For baselines we have included a linear-readout comparison using the same reservoir states; a full classical reservoir benchmark lies outside the scope of the present study, which centers on the quantum oscillator system and the entanglement metric, but we now explicitly note this limitation in the discussion. revision: partial

Circularity Check

0 steps flagged

No significant circularity in numerical QRC analysis

full rationale

The paper reports results from direct numerical integration of the coupled Kerr oscillator dynamics, independently computing logarithmic negativity from the simulated density matrix and NRMSE from the reservoir readout against target time series. The binned analysis is a post-hoc aggregation of these two separately obtained quantities over the (drive, Kerr, coupling) parameter grid; neither metric is fitted to the other, nor does any central claim reduce by construction to a self-referential definition or prior self-citation. The observed association between moderate negativity and lower average NRMSE is an empirical finding, not a tautological re-expression of the input parameters.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The work relies on standard quantum optical modeling of the Kerr oscillator Hamiltonian and numerical solution of the open-system dynamics; no ad-hoc fitted constants or new entities are introduced.

axioms (1)
  • domain assumption The dynamics of the two coupled Kerr oscillators under coherent drive and dissipation are accurately captured by the standard Lindblad master equation for this system.
    Invoked implicitly when performing parameter sweeps and entanglement calculations.

pith-pipeline@v0.9.0 · 5755 in / 1144 out tokens · 69812 ms · 2026-05-18T23:33:22.490591+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

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

  • IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    Using logarithmic negativity to quantify entanglement and normalized root mean square error (NRMSE) to evaluate predictive accuracy, individual parameter sweeps show that optimal performance occurs at moderate but non-zero entanglement. Furthermore, an aggregated binned analysis reveals that this moderate entanglement is consistently associated with the optimal average predictive performance across the parameter space

  • IndisputableMonolith/Foundation/DimensionForcing.lean alexander_duality_circle_linking unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    The system can be described by the following Hamiltonian: H(t) = H_nl + H_int + H_drive with Kerr terms K N^2/2 and linear coupling g(a b† + a† b)

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

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages

  1. [1]

    quantumness

    analyzed how different drive and coupling regimes in a pair of coupled Kerr cavities affect information encod- ing. A key question driving QRC research is what intrinsic quantum features can provide computational advantages over classical reservoirs. Quantum systems offer two no- table features: (i) a state space that grows exponen- tially with the number...

    work page

  2. [2]

    quan- tumness

    to describe the nonlinear coupler. Note that we are using dimensionless units throughout this study. B. Learning and Simulation An input signal sk with a frequency scalef is generated as follows: sk = 20X i sin(2πfitk + 2πζ) (6) 20 frequencies fi are evenly spaced in the interval [f /5000, f /50]. Here tk denotes the time elapsed after k time steps, and ζ...

    work page

  3. [3]

    Bengio, P

    Y. Bengio, P. Simard, and P. Frasconi, Learning long- term dependencies with gradient descent is difficult, IEEE Transactions on Neural Networks 5, 157 (1994)

    work page 1994

  4. [4]

    Hestness, N

    J. Hestness, N. Ardalani, and G. Diamos, Beyond human- level accuracy: Computational challenges in deep learn- ing (2019), arXiv:1909.01736 [cs.LG]

    work page arXiv 2019

  5. [5]

    Tanaka, T

    G. Tanaka, T. Yamane, J. B. H´ eroux, R. Nakane, N. Kanazawa, S. Takeda, H. Numata, D. Nakano, and A. Hirose, Recent advances in physical reservoir comput- ing: A review, Neural Networks 115, 100 (2019)

    work page 2019

  6. [6]

    L. C. G. Govia, G. J. Ribeill, G. E. Rowlands, H. K. Krovi, and T. A. Ohki, Quantum reservoir computing with a single nonlinear oscillator, Phys. Rev. Res. 3, 013077 (2021)

    work page 2021

  7. [7]

    Paquot, F

    Y. Paquot, F. Duport, A. Smerieri, J. Dambre, B. Schrauwen, M. Haelterman, and S. Massar, Opto- electronic reservoir computing, Scientific Reports 2, 287 (2012)

    work page 2012

  8. [8]

    Duport, B

    F. Duport, B. Schneider, A. Smerieri, M. Haelterman, and S. Massar, All-optical reservoir computing, Opt. Ex- press 20, 22783 (2012)

    work page 2012

  9. [9]

    Brunner, M

    D. Brunner, M. Soriano, C. Mirasso, and I. Fischer, Par- allel photonic information processing at gigabyte per sec- ond data rates using transient states, Nature communi- cations 4, 1364 (2013)

    work page 2013

  10. [10]

    Vandoorne, P

    K. Vandoorne, P. Mechet, T. Van Vaerenbergh, M. Fiers, G. Morthier, D. Verstraeten, B. Schrauwen, J. Dambre, and P. Bienstman, Experimental demonstration of reser- voir computing on a silicon photonics chip, Nature com- munications 5, 3541 (2014)

    work page 2014

  11. [11]

    C. Du, F. Cai, M. Zidan, W. Ma, S. Lee, and W. Lu, Reservoir computing using dynamic memristors for tem- poral information processing, Nature Communications 8 (2017)

    work page 2017

  12. [12]

    Kudithipudi, Q

    D. Kudithipudi, Q. Saleh, C. Merkel, J. Thesing, and B. Wysocki, Design and analysis of a neuromemristive reservoir computing architecture for biosignal processing, Frontiers in Neuroscience 9 (2016)

    work page 2016

  13. [13]

    Larger, A

    L. Larger, A. Bayl´ on-Fuentes, R. Martinenghi, V. S. Udaltsov, Y. K. Chembo, and M. Jacquot, High-speed photonic reservoir computing using a time-delay-based architecture: Million words per second classification, Phys. Rev. X 7, 011015 (2017)

    work page 2017

  14. [14]

    M. A. Nielsen and I. L. Chuang, Quantum Computa- tion and Quantum Information: 10th Anniversary Edi- tion (Cambridge University Press, 2010)

    work page 2010

  15. [15]

    Preskill, Quantum computing in the nisq era and be- yond, Quantum 2 (2018)

    J. Preskill, Quantum computing in the nisq era and be- yond, Quantum 2 (2018)

    work page 2018

  16. [16]

    Fujii and K

    K. Fujii and K. Nakajima, Harnessing disordered- ensemble quantum dynamics for machine learning, Phys. Rev. Appl. 8, 024030 (2017)

    work page 2017

  17. [17]

    Kutvonen, K

    A. Kutvonen, K. Fujii, and T. Sagawa, Optimizing a quantum reservoir computer for time series prediction, Scientific Reports 10, 14687 (2020)

    work page 2020

  18. [18]

    Suzuki, Q

    Y. Suzuki, Q. Gao, K. C. Pradel, K. Yasuoka, and N. Ya- mamoto, Natural quantum reservoir computing for tem- poral information processing, Scientific Reports 12, 1353 (2022)

    work page 2022

  19. [19]

    Ghosh, A

    S. Ghosh, A. Opala, M. Matuszewski, T. Paterek, and T. C. H. Liew, Quantum reservoir processing, npj Quan- tum Information 5, 35 (2019). 12

    work page 2019

  20. [20]

    L. C. G. Govia, G. J. Ribeill, G. E. Rowlands, and T. A. Ohki, Nonlinear input transformations are ubiquitous in quantum reservoir computing (2021), arXiv:2107.00147 [quant-ph]

    work page arXiv 2021

  21. [21]

    G. J. Milburn, Quantum and classical liouville dynam- ics of the anharmonic oscillator, Phys. Rev. A 33, 674 (1986)

    work page 1986

  22. [22]

    Kitagawa and Y

    M. Kitagawa and Y. Yamamoto, Number-phase minimum-uncertainty state with reduced number uncer- tainty in a kerr nonlinear interferometer, Phys. Rev. A 34, 3974 (1986)

    work page 1986

  23. [23]

    Haroche and J.-M

    S. Haroche and J.-M. Raimond, Exploring the Quantum: Atoms, Cavities, and Photons (Oxford University Press, 2006)

    work page 2006

  24. [24]

    Kanao and H

    T. Kanao and H. Goto, Fast elementary gates for univer- sal quantum computation with kerr parametric oscillator qubits, Phys. Rev. Res. 6, 013192 (2024)

    work page 2024

  25. [25]

    W. D. Kalfus, G. J. Ribeill, G. E. Rowlands, H. K. Krovi, T. A. Ohki, and L. C. G. Govia, Hilbert space as a com- putational resource in reservoir computing, Phys. Rev. Res. 4, 033007 (2022)

    work page 2022

  26. [26]

    Dudas, B

    J. Dudas, B. Carles, E. Plouet, F. A. Mizrahi, J. Grol- lier, and D. Markovi´ c, Quantum reservoir computing im- plementation on coherently coupled quantum oscillators, npj Quantum Information 9, 64 (2023)

    work page 2023

  27. [27]

    Cheamsawat and T

    K. Cheamsawat and T. Chotibut, Dissipation alters modes of information encoding in small quantum reser- voirs near criticality, Entropy (Basel) 27 (2025)

    work page 2025

  28. [28]

    G¨ otting, F

    N. G¨ otting, F. Lohof, and C. Gies, Exploring quantum- ness in quantum reservoir computing, Physical Review A 108, 10.1103/physreva.108.052427 (2023)

    work page doi:10.1103/physreva.108.052427 2023

  29. [29]

    Nokkala, R

    J. Nokkala, R. Mart´ ınez-Pe˜ na, G. L. Giorgi, V. Pa- rigi, M. C. Soriano, and R. Zambrini, Gaussian states of continuous-variable quantum systems provide univer- sal and versatile reservoir computing, Communications Physics 4, 53 (2021)

    work page 2021

  30. [30]

    Y. Kora, H. Zadeh-Haghighi, T. C. Stewart, K. Heshami, and C. Simon, Frequency- and dissipation-dependent en- tanglement advantage in spin-network quantum reservoir computing, Phys. Rev. A 110, 042416 (2024)

    work page 2024

  31. [31]

    Garc´ ıa-Beni, G

    J. Garc´ ıa-Beni, G. L. Giorgi, M. C. Soriano, and R. Zam- brini, Squeezing as a resource for time series processing in quantum reservoir computing, Opt. Express 32, 6733 (2024)

    work page 2024

  32. [32]

    Y. Lu, A. Maiti, J. W. O. Garmon, S. Ganjam, Y. Zhang, J. Claes, L. Frunzio, S. M. Girvin, and R. J. Schoelkopf, High-fidelity parametric beamsplitting with a parity- protected converter, Nature Communications 14, 5767 (2023)

    work page 2023

  33. [33]

    Basilewitsch, Y

    D. Basilewitsch, Y. Zhang, S. M. Girvin, and C. P. Koch, Engineering strong beamsplitter interaction be- tween bosonic modes via quantum optimal control the- ory, Phys. Rev. Res. 4, 023054 (2022)

    work page 2022

  34. [34]

    B. J. Chapman, S. J. de Graaf, S. H. Xue, Y. Zhang, J. Teoh, J. C. Curtis, T. Tsunoda, A. Eickbusch, A. P. Read, A. Koottandavida, S. O. Mundhada, L. Frunzio, M. Devoret, S. Girvin, and R. Schoelkopf, High-on-off- ratio beam-splitter interaction for gates on bosonically encoded qubits, PRX Quantum 4, 020355 (2023)

    work page 2023

  35. [35]

    L. J. Bernstein, Quantizing a self-trapping transition, Physica D: Nonlinear Phenomena 68, 174 (1993)

    work page 1993

  36. [36]

    Chefles and S

    A. Chefles and S. M. Barnett, Quantum theory of two- mode nonlinear directional couplers, Journal of Modern Optics 43, 709 (1996)

    work page 1996

  37. [37]

    Vidal and R

    G. Vidal and R. F. Werner, Computable measure of en- tanglement, Phys. Rev. A 65, 032314 (2002)

    work page 2002

  38. [38]

    M. B. Plenio, Logarithmic negativity: A full entangle- ment monotone that is not convex, Phys. Rev. Lett. 95, 090503 (2005)

    work page 2005

  39. [39]

    Motamedi, H

    A. Motamedi, H. Zadeh-Haghighi, and C. Simon, Quan- tumness and learning performance in reservoir computing with a single oscillator (2024), arXiv:2304.03462 [quant- ph]

    work page arXiv 2024