Pith. sign in

REVIEW 2 major objections 5 minor 2 cited by

Input-dependence in quantum reservoir computing

T0 review · 2 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Distinct input sequences reach distinct reservoir states exactly when the reservoir map is state-input invertible and has the echo state property; for finite-dimensional quantum reservoirs this becomes a checkable rank condition on the…

desk verdict The local injectivity results and contracted-encoding analysis are salvageable, but the headline global criterion (Prop. 5) is false as stated, so the paper needs major revision before it can be relied on. read the letter →

arxiv 2412.08322 v3 pith:MOSE6GI4 submitted 2024-12-11 quant-ph

Rodrigo Martínez-Peña , Juan-Pablo Ortega This is my paper · ORCID
classification quant-ph
keywords quantumreservoircomputingfilterinjectivityechostatepropertyfadingmemorystate-affinesystemscontracted-encodingchannelsgeneralizedsynchronizationinputdependence
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 addresses a gap in quantum reservoir computing: even when a reservoir has the echo state property (each input has a unique reservoir history) and fading memory, its filter may still fail to tell different input sequences apart. The authors establish that injectivity of the filter—the one-to-one correspondence between input histories and reservoir state histories—is guaranteed once the state-update map is state-input invertible and the system has ESP. For finite-dimensional quantum reservoirs, which admit an affine (state-affine) representation in a Gell-Mann basis, this reduces to a rank condition on the derivatives of p and q that is readily checkable in many models. The relevance is that injective filters are a prerequisite for learning deterministic dynamical systems through injective generalized synchronization, and the paper shows numerically that losing injectivity, for example through periodic input encodings, degrades temporal tasks such as short-term memory.

What carries the argument

The load-bearing object is the state-affine system (SAS) representation of a reservoir in a generalized Gell-Mann basis, written x_t = p(z_t)x_{t-1} + q(z_t), which turns the quantum channel into an affine map on the Bloch vector; contractivity of p guarantees the echo state property and the absolutely convergent filter expansion U(z)_t = \sum_{j\geq 0} \left(\prod_{k=0}^{j-1} p(z_{t-k})\right) q(z_{t-j}). The argument is carried by Lemma 4, which reduces filter injectivity to state-input invertibility, and by the local injectivity theorem, which converts the rank condition DF_x(z) rank n into that invertibility. For contracted-encoding channels, the composition structure lets the rank condition be expressed through p_E, p_J(z), q_E, and q_J(z), with direct simplifications when E is invertible or J is unitary.

What would settle it

Symbolically evaluate the rank condition (22) for a state-affine quantum reservoir that has the echo state property, then compute the filter U_F on two distinct input sequences whose entries all lie in the input domain. Equality of the two state histories for any such pair would refute Proposition 5, since the proposition predicts no collision can occur when the rank condition holds at every reachable state; a random search over contractive p(z) and differentiable q(z) with a symbolic rank check would settle the sufficient condition directly.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central discovery is that filter injectivity for reservoir computers, quantum or not, follows from a local, finite-dimensional condition. Lemma 4 proves the general statement: if F is state-input invertible (each map F_x(z)=F(x,z) is injective) and has the echo state property, then the filter U_F is injective. In the state-affine form that every finite-dimensional quantum reservoir takes in a generalized Gell-Mann basis, state-input invertibility is implied by the rank condition that the linear maps (Dp(z)(·))x + Dq(z) have rank n for all reachable states x and all inputs z (Proposition 5). The paper extends the same logic to the widely used contracted-encoding architecture, where an input-encoding quantum channel is followed by a strictly contractive channel; there the injectivity conditions factor through the affine data of the encoding channel J and the contractive channel E, with simplifications when E is invertible or J is unitary. It also gives local injectivity results around constant sequences, characteizes when the filter is constant through the fixed-point map, and shows numerically that a periodic rotation encoding that breaks global injectivity at g=2π also destroys short-term memory capacity.

Load-bearing premise

The paper's main theorems rely on the imported characterization from Ref. [32] that, for compact input spaces, uniform operator-norm contraction of the reservoir map on traceless operators is equivalent to the echo state and fading memory properties; if that contractivity characterization fails, or the uniform epsilon-contraction needed for the series expansion does not hold, the filter expansion and the injectivity results built on it do not apply.

Editorial extensions

If this is right

  • Designers of quantum reservoir computers can certify that their reservoir distinguishes input histories by checking a finite-dimensional rank condition on the affine maps p(z) and q(z) over the reachable set, rather than comparing arbitrary input sequences.
  • For contracted-encoding channels, the injectivity check splits into the encoding channel and the contractive channel: an invertible E reduces it to the input-encoding map's derivatives, and a unitary encoding reduces it to p_E Dp_J(z)(\cdot)x with x\neq 0.
  • Local injectivity around constant sequences gives an easy-to-verify criterion: if (Dp(z)(\cdot))x^*(z)+Dq(z) has rank n at the fixed point x^*(z)=(I-p(z))^{-1}q(z), the filter is injective on a neighborhood of that constant input.
  • Non-injective filters cannot support injective generalized synchronization, so the rank conditions are prerequisites for learning deterministic dynamical systems by reservoir forecasting.
  • Periodic input encodings destroy global injectivity, and the paper's numerical experiment ties that loss to a drop in short-term memory capacity, making injectivity an explicit design constraint for temporal tasks.

Reading between the lines

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

  • Because Lemma 4 is purely set-theoretic and the SAS rank condition is not quantum-specific, the same injectivity criterion should apply to classical state-affine reservoirs; the paper's examples are quantum, but the mechanism is agnostic.
  • The rank condition is sufficient rather than necessary, so there is likely a larger class of injective reservoirs whose filters remain injective even where the Jacobian drops rank; mapping the boundary of the sufficient region could reveal cheaper design criteria.
  • Sampling the rank condition over the full state space instead of the reachable set can produce false negatives, since states that are never visited may violate the condition without affecting injectivity; automatic differentiation over simulated trajectories could estimate the reachable set in practice.
  • The connection to nonstationary echo state properties noted in the paper suggests that injectivity of the fixed-point map, together with contraction, may provide a route to formalizing when time-varying or nonstationary reservoirs can process sequences, going beyond the constant-input analysis.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper studies input-dependence of quantum reservoir computing filters. It builds on the state-affine system (SAS) representation of finite-dimensional quantum channels and proposes conditions under which the reservoir filter is injective as a map from input sequences to reservoir state sequences. The main results are Lemma 4 (SI-invertibility plus ESP implies filter injectivity), Proposition 5 (a rank condition on the SAS derivative that is claimed to guarantee SI-invertibility), a local injectivity theorem (Proposition 11) and its SAS specialization (Proposition 15, Corollary 16), and a classification of constant filters for contracted-encoding quantum channels (Theorem 19). The paper also contains several examples and a numerical demonstration that breaking global injectivity degrades a short-term memory task.

Significance. The motivation is sound: filter injectivity is a natural and practically relevant design goal for reservoir computing, and the connection to generalized synchronization and learnability in Remark 8 is useful. The paper has clear strengths: Lemma 4 is correct and concise; Proposition 12 correctly characterizes preimages of constant output sequences; the contracted-encoding family is well motivated; and the examples are concrete, with reproducible code for the numerical experiment. If a correct sufficient condition in the spirit of Proposition 5 existed, the paper would make a valuable design contribution. However, the central rank criterion is false as stated, and the local injectivity theorem has a serious proof gap, so the headline results cannot be accepted in their current form.

major comments (2)
  1. [III.A, Proposition 5 and Eq. (22)] The claimed sufficient condition is false. The rank condition (22) only gives, via the Local Injectivity Theorem, that for each fixed x and z there is a neighborhood on which F_x is injective; full rank at every point does not imply global injectivity of F_x, and SI-invertibility (Definition 3) is a global property. Concretely, take N=3, n=2, D=[0,4π]×[0,1], V a sufficiently large closed ball, and F(x,z) = r x + (cos z1, sin z1, z2) with 0<r<1. Then p(z)=rI and q(z)=(cos z1, sin z1, z2); Dp=0 and Dq has rank 2, so (22) holds for every x and z. The uniform contraction of p gives the ESP. Yet F_x(z1,z2)=F_x(z1+2π,z2) for every x, so F is not SI-invertible and the filter is not injective: the constant input sequences (0,0) and (2π,0) have identical fixed-point images. This invalidates the central design criterion of the paper, and the same issue propagates to the contracted-encoding sufficient conditions in Section III.B, in particular Eq. (48) and Eq. (50).
  2. [III.A, Proposition 11 proof] The compactness argument in the proof of Proposition 11 is not valid. The function f(x) = inradius(L_{z0}(x), z0) is asserted to attain its minimum because there is a sequence x_n with f(x_n) -> a and a convergent subsequence x_{n_k} -> l; the proof then says f(l) would otherwise contradict that convergence. This would require f to be lower semicontinuous, or at least that the neighborhoods L_{z0}(x) vary continuously in x, which is neither established nor generally true. An infimum of positive numbers over a compact set need not be attained for an arbitrary positive function, and inradius can drop at a limit point. Consequently the existence of the uniform neighborhood K_{z0} in Eq. (27), and with it Proposition 11(i) and the local injectivity results Proposition 15 and Corollary 16, is not proved as written.
minor comments (5)
  1. [Section II] The notation B0(H) for the space of traceless operators and B0 for the Gell-Mann basis of su(d) is confusing; please use different symbols for the basis and the subspace.
  2. [Eq. (14)] The zero block in the block matrix is written as 0_{d^2-1}; it should be identified explicitly as a row vector of length d^2-1 so that the dimensions of the four blocks are unambiguous.
  3. [Eq. (19)] The filter expansion is written for left-infinite sequences but the product order and the indices assume a specified time origin; please state explicitly whether the domain is (D^n)^{Z-} or (D^n)^Z and how the origin is chosen.
  4. [Proposition 5 statement] The statement 'Let F=V×D^n→V' is missing a colon after the domain, and the phrase 'have all rank n and hence are injective' should read 'have rank n for all ...' for clarity.
  5. [Example 24] The sentence 'showing U(z)_0 ≠ 0 for a given epsilon seems an unfeasible task' is vague; please state precisely which claim is being discarded and why the positivity of the filter output is relevant to the rank condition.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the injectivity criteria are derived from definitions, ESP, and standard analysis, with prior-work citations serving as independent support rather than as the conclusion.

full rationale

The paper's central claim is a sufficient condition for filter injectivity. Lemma 4 derives filter injectivity directly from the definition of state-input invertibility and the echo state property: if two input sequences produce the same reservoir state sequence, then applying SI-invertibility of F_x at each time step forces the inputs to be equal. This is a genuine derivation, not a restatement of the conclusion. Proposition 5 then translates the standard rank condition DF_x(z) having rank n into the SAS form (22); this is a sufficient design criterion, and the rank condition is not defined in terms of filter injectivity, so the result is not self-definitional. The paper's reliance on Ref. [32] for the SAS representation, the ESP/FMP characterization, and the constant-filter theorem is legitimate: those are parameter-free theorems with stated assumptions that do not already contain the injectivity conclusion, and they were established in a prior publication. Similarly, the filter expansion (19) comes from a general SAS reservoir result and is not fitted to make the injectivity theorems true. No parameter is fitted and no quantity is renamed as a prediction. The periodic obstruction highlighted in Example 24 is a mathematical correctness concern about whether pointwise rank conditions imply global injectivity, not a circularity: the hypothesis of Proposition 5 is not equivalent to its conclusion, and the paper does not assume filter injectivity in order to prove it. Accordingly, the derivation chain is self-contained with respect to circularity.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The new injectivity results are derived from standard fixed-point and differential-topology tools plus the authors' earlier SAS/ESP framework. No new physical entities are introduced, and no constants are fitted to data; the only hand-chosen numbers are example hyperparameters.

free parameters (1)
  • Example hyperparameters (epsilon, lambda, theta, gamma, Delta_tau, g)
    Chosen by hand to exhibit specific behaviors in Examples 17, 21-24 and Fig. 2; the central theorems do not depend on their values.
assumptions (5)
  • standard math CPTP maps on finite-dimensional Hilbert spaces are nonexpansive in trace norm, and strictly contractive channels have unique fixed points.
    Used to define fixed point functions and filters in Section II and Theorem 19.
  • domain assumption Proposition 3 in Ref. [32]: for compact D^n, uniform contraction |||T(·,z)|_{B0(H)}||| < 1-epsilon for all z is necessary and sufficient for ESP and FMP.
    Imported from authors' prior work; underpins the ESP guarantee and the geometric series filter (19) in Remarks 6 and 18.
  • standard math Local injectivity theorem (Ref. [58], Theorem 2.5.10): full-rank Frechet differential implies local injectivity.
    Basis for the rank conditions (20), (22), (29), and the local injectivity propositions.
  • domain assumption System isomorphisms between density-matrix and Bloch-vector SAS representations preserve ESP, FMP, and filters (Ref. [32], Proposition 2).
    Justifies replacing quantum channels by affine SAS dynamics (17) throughout.
  • domain assumption In Theorem 19(ii), E is assumed to have an inverse map on B(H).
    Explicit hypothesis needed for the equivalence of constant-filter conditions; many strictly contractive channels do not satisfy it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Input-dependence in quantum reservoir computing." pith.science (2026). https://pith.science/paper/MOSE6GI4

@misc{pith2026241208322,
  author       = {Pith},
  title        = {Pith review of: Input-dependence in quantum reservoir computing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MOSE6GI4}},
  note         = {Machine review of arXiv:2412.08322}
}
read the original abstract

Quantum reservoir computing is an emergent field in which quantum dynamical systems are exploited for temporal information processing. In previous work, it was found a feature that makes a quantum reservoir valuable: contractive dynamics of the quantum reservoir channel toward input-dependent fixed points. These results are enhanced in this paper by finding conditions that guarantee a crucial aspect of the reservoir's design: distinguishing between different input sequences to ensure a faithful representation of temporal input data. This is implemented by finding a condition that guarantees injectivity in reservoir computing filters, with a special emphasis on the quantum case. We provide several examples and focus on a family of quantum reservoirs that is much used in the literature; it consists of an input-encoding quantum channel followed by a strictly contractive channel that enforces the echo state and the fading memory properties. This work contributes to analyzing valuable quantum reservoirs in terms of their input dependence.

Figures

Figures reproduced from arXiv: 2412.08322 by the authors.

Figure 1
Figure 1. FIG. 1: Vector norm for the rank condition with [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2: Memory capacity for the short-term memory [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. Robust quantum reservoir computers for forecasting chaotic dynamics: generalized synchronization and stability

    quant-ph 2025-06 conditional novelty 6.0 of 10

    Recurrence-free quantum reservoir computers have a constant, contractive Jacobian, which guarantees the echo state property and enables accurate inference of Lyapunov spectra and attractor dimensions.

  2. Dissipation alters modes of information encoding in small quantum reservoirs near criticality

    quant-ph 2024-12 conditional novelty 6.0 of 10

    In a two-oscillator quantum reservoir, near the critical coupling J=|Δ|, information encoding switches from redundant to synergistic, with synergy aiding short-term and dissipation aiding longer-term memory.

Reference graph

Works this paper leans on

76 extracted references · 53 canonical work pages · cited by 2 Pith papers

  1. [1]

    Garc ´ ıa-Beni, G

    J. Garc ´ ıa-Beni, G. Luca Giorgi, M. C. Soriano, and R. Zambrini, Squeezing as a resource for time series pro- cessing in quantum reservoir computing, Optics Express 32, 6733 (2024)

    work page 2024

  2. [2]

    Garc ´ ıa-Beni, I

    J. Garc ´ ıa-Beni, I. Paparelle, V. Parigi, G. L. Giorgi, M. C. Soriano, and R. Zambrini, Quantum machine 15 learning via continuous-variable cluster states and tele- portation, EPJ Quantum Technology12, 1 (2025)

    work page 2025

  3. [3]

    Retrieving past quantum features with deep hybrid classical-quantum reservoir computing

    J. Nokkala, G. L. Giorgi, and R. Zambrini, Re- trieving past quantum features with deep hybrid classical-quantum reservoir computing, arXiv preprint arXiv:2401.16961 (2024)

    work page Pith review arXiv 2024

  4. [4]

    De Prins, G

    R. De Prins, G. Van der Sande, and P. Bienstman, A recurrent gaussian quantum network for online process- ing of quantum time series, Scientific Reports14, 12322 (2024)

    work page 2024

  5. [5]

    Mart ´ ınez-Pe˜ na, G

    R. Mart ´ ınez-Pe˜ na, G. L. Giorgi, J. Nokkala, M. C. So- riano, and R. Zambrini, Dynamical phase transitions in quantum reservoir computing, Physical Review Letters 127, 100502 (2021)

    work page 2021

  6. [6]

    Palacios, R

    A. Palacios, R. Mart ´ ınez-Pe˜ na, M. C. Soriano, G. L. Giorgi, and R. Zambrini, Role of coherence in many- body quantum reservoir computing, arXiv preprint arXiv:2409.17734 (2024)

    arXiv 2024

  7. [7]

    W. Xia, J. Zou, X. Qiu, and X. Li, The reservoir learn- ing power across quantum many-body localization tran- sition, Frontiers of Physics17, 33506 (2022)

    work page 2022

  8. [8]

    M. N. Ivaki, A. Lazarides, and T. Ala-Nissila, Quantum reservoir computing on random regular graphs, arXiv preprint arXiv:2409.03665 (2024)

    arXiv 2024

Show all 76 references
  1. [9]

    Llodr` a, P

    G. Llodr` a, P. Mujal, R. Zambrini, and G. L. Giorgi, Quantum reservoir computing in atomic lattices, Chaos, Solitons & Fractals195, 116289 (2025)

    2025

  2. [10]

    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 Information9, 64 (2023)

    2023

  3. [11]

    Sannia, R

    A. Sannia, R. Mart ´ ınez-Pe˜ na, M. C. Soriano, G. L. Giorgi, and R. Zambrini, Dissipation as a resource for quantum reservoir computing, Quantum8, 1291 (2024)

    2024

  4. [12]

    Sannia, G

    A. Sannia, G. L. Giorgi, S. Longhi, and R. Zambrini, Skin effect in quantum neural networks, arXiv preprint arXiv:2406.14112 (2024)

    2024 arXiv

  5. [13]

    Abbas and I

    A. Abbas and I. S. Maksymov, Reservoir computing using measurement-controlled quantum dynamics, Electronics 13, 1164 (2024)

    2024

  6. [14]

    C. Zhu, P. J. Ehlers, H. I. Nurdin, and D. Soh, Prac- tical and scalable quantum reservoir computing, arXiv preprint arXiv:2405.04799 (2024)

    2024

  7. [15]

    Dasgupta, K

    S. Dasgupta, K. E. Hamilton, and A. Banerjee, Charac- terizing the memory capacity of transmon qubit reser- voirs, in2022 IEEE International Conference on Quan- tum Computing and Engineering (QCE)(IEEE, 2022) pp. 162–166

    2022

  8. [16]

    Mlika, S

    Z. Mlika, S. Cherkaoui, J. F. Laprade, and S. Corbeil- Letourneau, User trajectory prediction in mobile wire- less networks using quantum reservoir computing, IET Quantum Communication4, 125 (2023)

    2023

  9. [17]

    Suzuki, Q

    Y. Suzuki, Q. Gao, K. C. Pradel, K. Yasuoka, and N. Yamamoto, Natural quantum reservoir computing for temporal information processing, Scientific reports12, 1 (2022)

    2022

  10. [18]

    Kubota, Y

    T. Kubota, Y. Suzuki, S. Kobayashi, Q. H. Tran, N. Ya- mamoto, and K. Nakajima, Temporal information pro- cessing induced by quantum noise, Physical Review Re- search5, 023057 (2023)

    2023

  11. [19]

    Yasuda, Y

    T. Yasuda, Y. Suzuki, T. Kubota, K. Nakajima, Q. Gao, W. Zhang, S. Shimono, H. I. Nurdin, and N. Yamamoto, Quantum reservoir computing with re- peated measurements on superconducting devices, arXiv preprint arXiv:2310.06706 (2023)

    2023 arXiv

  12. [20]

    J. Chen, H. I. Nurdin, and N. Yamamoto, Temporal infor- mation processing on noisy quantum computers, Physical Review Applied14, 024065 (2020)

    2020

  13. [21]

    Molteni, C

    R. Molteni, C. Destri, and E. Prati, Optimization of the memory reset rate of a quantum echo-state network for time sequential tasks, Physics Letters A , 128713 (2023)

    2023

  14. [22]

    Pfeffer, F

    P. Pfeffer, F. Heyder, and J. Schumacher, Hybrid quantum-classical reservoir computing of thermal convec- tion flow, Physical Review Research4, 033176 (2022)

    2022

  15. [23]

    Ahmed, F

    O. Ahmed, F. Tennie, and L. Magri, Optimal training of finitely sampled quantum reservoir computers for fore- casting of chaotic dynamics, Quantum Machine Intelli- gence7, 1 (2025)

    2025

  16. [24]

    F. Hu, S. A. Khan, N. T. Bronn, G. Angelatos, G. E. Rowlands, G. J. Ribeill, and H. E. T¨ ureci, Overcoming the coherence time barrier in quantum machine learn- ing on temporal data, Nature Communications15, 7491 (2024)

    2024

  17. [25]

    S. Miao, O. T. Kulseng, A. Stasik, and F. G. Fuchs, Quantumreservoirpy: A software package for time series prediction, arXiv preprint arXiv:2401.10683 (2024)

    2024 arXiv

  18. [26]

    Sornsaeng, N

    A. Sornsaeng, N. Dangniam, and T. Chotibut, Quantum next generation reservoir computing: an efficient quan- tum algorithm for forecasting quantum dynamics, Quan- tum Machine Intelligence6, 57 (2024)

    2024

  19. [27]

    Mujal, R

    P. Mujal, R. Mart ´ ınez-Pe˜ na, G. L. Giorgi, M. C. Soriano, and R. Zambrini, Time-series quantum reservoir comput- ing with weak and projective measurements, npj Quan- tum Information9, 16 (2023)

    2023

  20. [28]

    Franceschetto, M

    G. Franceschetto, M. P lodzie´ n, M. Lewenstein, A. Ac ´ ın, and P. Mujal, Harnessing quantum back-action for time-series processing, arXiv preprint arXiv:2411.03979 (2024)

    2024 arXiv

  21. [29]

    Kobayashi, K

    K. Kobayashi, K. Fujii, and N. Yamamoto, Feedback- driven quantum reservoir computing for time-series anal- ysis, PRX Quantum5, 040325 (2024)

    2024

  22. [30]

    Chen and H

    J. Chen and H. I. Nurdin, Learning nonlinear input– output maps with dissipative quantum systems, Quan- tum Information Processing18, 1 (2019)

    2019

  23. [31]

    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 Physics4, 1 (2021)

    2021

  24. [32]

    Mart ´ ınez-Pe˜ na and J.-P

    R. Mart ´ ınez-Pe˜ na and J.-P. Ortega, Quantum reservoir computing in finite dimensions, Phys. Rev. E107, 035306 (2023)

    2023

  25. [33]

    Kobayashi, Q

    S. Kobayashi, Q. H. Tran, and K. Nakajima, Extend- ing echo state property for quantum reservoir computing, Physical Review E110, 024207 (2024)

    2024

  26. [34]

    Kobayashi, Q

    S. Kobayashi, Q. H. Tran, and K. Nakajima, Coherence influx is indispensable for quantum reservoir computing, arXiv preprint arXiv:2409.12693 (2024)

    2024 arXiv

  27. [35]

    Grigoryeva, A

    L. Grigoryeva, A. Hart, and J.-P. Ortega, Chaos on compact manifolds: Differentiable synchronizations be- yond the takens theorem, Physical Review E103, 062204 (2021)

    2021

  28. [36]

    Grigoryeva, A

    L. Grigoryeva, A. G. Hart, and J.-P. Ortega, Learning strange attractors with reservoir systems, Nonlinearity 36, 4674 (2023)

    2023

  29. [37]

    Grigoryeva, B

    L. Grigoryeva, B. Hamzi, F. P. Kemeth, Y. Kevrekidis, G. Manjunath, J.-P. Ortega, and M. J. Steynberg, Data- driven cold starting of good reservoirs, Physica D: Non- 16 linear Phenomena469, 134325 (2024)

    2024

  30. [38]

    Kocarev and U

    L. Kocarev and U. Parlitz, Generalized synchronization, predictability, and equivalence of unidirectionally cou- pled dynamical systems, Physical Review Letters76, 1816 (1996)

    1996

  31. [39]

    Jaeger, Short term memory in echo state networks, Fraunhofer Institute for Autonomous Intelligent Systems

    H. Jaeger, Short term memory in echo state networks, Fraunhofer Institute for Autonomous Intelligent Systems. Technical Report.152(2002)

    2002

  32. [40]

    Gonon, L

    L. Gonon, L. Grigoryeva, and J.-P. Ortega, Memory and forecasting capacities of nonlinear recurrent networks, Physica D414, 1 (2020)

    2020

  33. [41]

    Ballarin, L

    G. Ballarin, L. Grigoryeva, and J.-P. Ortega, Memory of recurrent networks: Do we compute it right?, Journal of Machine Learning Research25, 1 (2024)

    2024

  34. [42]

    Grigoryeva and J.-P

    L. Grigoryeva and J.-P. Ortega, Universal discrete-time reservoir computers with stochastic inputs and linear readouts using non-homogeneous state-affine systems, Journal of Machine Learning Research19, 1 (2018)

    2018

  35. [43]

    Grigoryeva and J.-P

    L. Grigoryeva and J.-P. Ortega, Echo state networks are universal, Neural Networks108, 495 (2018)

    2018

  36. [44]

    Gonon and J.-P

    L. Gonon and J.-P. Ortega, Reservoir computing univer- sality with stochastic inputs, IEEE transactions on neu- ral networks and learning systems31, 100 (2019)

    2019

  37. [45]

    Gonon, L

    L. Gonon, L. Grigoryeva, and J.-P. Ortega, Approxi- mation error estimates for random neural networks and reservoir systems, The Annals of Applied Probability33, 28 (2023)

    2023

  38. [46]

    Legenstein and W

    R. Legenstein and W. Maass, What makes a dynamical system computationally powerful?, inNew directions in statistical signal processing: from systems to brain, edited by S. Haykin (MIT Press, Cambridge, MA, 2007)

    2007

  39. [47]

    Maass, Liquid state machines: Motivation, theory, and applications, inComputability In Context: Compu- tation and Logic in the Real World, edited by S

    W. Maass, Liquid state machines: Motivation, theory, and applications, inComputability In Context: Compu- tation and Logic in the Real World, edited by S. S. Barry Cooper and A. Sorbi (World Scientific, 2011) Chap. 8, pp. 275–296

    2011

  40. [48]

    Boyd and L

    S. Boyd and L. Chua, Fading memory and the problem of approximating nonlinear operators with volterra se- ries, IEEE Transactions on circuits and systems32, 1150 (1985)

    1985

  41. [49]

    Kimura, The bloch vector for n-level systems, Physics Letters A314, 339 (2003)

    G. Kimura, The bloch vector for n-level systems, Physics Letters A314, 339 (2003)

    2003

  42. [50]

    M. S. Byrd and N. Khaneja, Characterization of the posi- tivity of the density matrix in terms of the coherence vec- tor representation, Physical Review A68, 062322 (2003)

    2003

  43. [51]

    Hantzko, L

    L. Hantzko, L. Binkowski, and S. Gupta, Pauli transfer matrices, arXiv preprint arXiv:2411.00526 (2024)

    2024 arXiv

  44. [52]

    Grigoryeva and J.-P

    L. Grigoryeva and J.-P. Ortega, Dimension reduction in recurrent networks by canonicalization, Journal of Geo- metric Mechanics13, 647 (2021)

    2021

  45. [53]

    J. R. Munkres,Topology, Vol. 2 (Prentice Hall Upper Saddle River, 2000)

    2000

  46. [54]

    Siewert, On orthogonal bases in the hilbert-schmidt space of matrices, Journal of Physics Communications6, 055014 (2022)

    J. Siewert, On orthogonal bases in the hilbert-schmidt space of matrices, Journal of Physics Communications6, 055014 (2022)

    2022

  47. [55]

    E. D. Sontag, Realization theory of discrete-time nonlin- ear systems: Part I-The bounded case, IEEE Transac- tions on Circuits and Systems26, 342 (1979)

    1979

  48. [56]

    Dang Van Mien and D

    H. Dang Van Mien and D. Normand-Cyrot, Nonlinear state affine identification methods: applications to elec- trical power plants, Automatica20, 175 (1984)

    1984

  49. [57]

    Manjunath, A

    G. Manjunath, A. de Clercq, and M. Steynberg, Universal set of observables for forecasting physical systems through causal embedding, arXiv preprint arXiv:2105.10759 (2021)

    2021 arXiv

  50. [58]

    Abraham, J

    R. Abraham, J. E. Marsden, and T. S. Ratiu,Mani- folds, Tensor Analysis, and Applications, Vol. 75 (Ap- plied Mathematical Sciences. Springer-Verlag, 1988)

    1988

  51. [59]

    Grigoryeva and J.-P

    L. Grigoryeva and J.-P. Ortega, Differentiable reservoir computing, Journal of Machine Learning Research20, 1 (2019)

    2019

  52. [60]

    Jaeger and H

    H. Jaeger and H. Haas, Harnessing nonlinearity: Predict- ing chaotic systems and saving energy in wireless com- munication, Science304, 78 (2004)

    2004

  53. [61]

    Z. Lu, B. R. Hunt, and E. Ott, Attractor reconstruc- tion by machine learning, Chaos28, 10.1063/1.5039508 (2018), arXiv:1805.03362

    2018 arXiv

  54. [62]

    Pathak, B

    J. Pathak, B. Hunt, M. Girvan, Z. Lu, and E. Ott, Model- free prediction of large spatiotemporally chaotic systems from data: A reservoir computing approach, Physical Re- view Letters120, 24102 (2018)

    2018

  55. [63]

    Pathak, Z

    J. Pathak, Z. Lu, B. R. Hunt, M. Girvan, and E. Ott, Using machine learning to replicate chaotic attractors and calculate Lyapunov exponents from data, Chaos27, 10.1063/1.5010300 (2017), arXiv:1710.07313

    2017 arXiv

  56. [64]

    Lu and D

    Z. Lu and D. S. Bassett, Invertible generalized synchro- nization: A putative mechanism for implicit learning in neural systems, Chaos30, 10.1063/5.0004344 (2020)

    2020 doi

  57. [65]

    Verzelli, C

    P. Verzelli, C. Alippi, and L. Livi, Learn to synchronize, synchronize to learn, Chaos: An Interdisciplinary Journal of Nonlinear Science31, 083119 (2021)

    2021

  58. [66]

    Sauer, J

    T. Sauer, J. A. Yorke, and M. Casdagli, Embedology, Journal of Statistical Physics65, 579 (1991)

    1991

  59. [67]

    Takens, Detecting strange attractors in turbulence (Springer Berlin Heidelberg, 1981) pp

    F. Takens, Detecting strange attractors in turbulence (Springer Berlin Heidelberg, 1981) pp. 366–381. [68]https://github.com/RMPhys/input-dependence-QRC

    1981

  60. [69]

    Burgarth, G

    D. Burgarth, G. Chiribella, V. Giovannetti, P. Perinotti, and K. Yuasa, Ergodic and mixing quantum channels in finite dimensions, New Journal of Physics15, 073045 (2013)

    2013

  61. [70]

    Fujii and K

    K. Fujii and K. Nakajima, Harnessing disordered- ensemble quantum dynamics for machine learning, Phys- ical Review Applied8, 024030 (2017)

    2017

  62. [71]

    C. Gies, F. Lohof, and N. G¨ otting, Exploring quantum mechanical advantage for reservoir computing, inAI and Optical Data Sciences V(SPIE, 2024) p. PC129030S

    2024

  63. [72]

    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, Physical Review A110, 042416 (2024)

    2024

  64. [73]

    Monzani, E

    F. Monzani, E. Ricci, L. Nigro, and E. Prati, Leveraging non-unital noise for gate-based quantum reservoir com- puting, arXiv preprint arXiv:2409.07886 (2024)

    2024 arXiv

  65. [74]

    F. G. Fuchs, A. J. Stasik, S. Miao, O. T. Kulseng, and R. P. Bassa, Quantum reservoir computing using the stabilizer formalism for encoding classical data, arXiv preprint arXiv:2407.00445 (2024)

    2024 arXiv

  66. [75]

    Chen and H

    J. Chen and H. I. Nurdin, Nonlinear autoregression with convergent dynamics on novel computational platforms, IEEE Transactions on Control Systems Technology30, 2228 (2021)

    2021

  67. [76]

    Mifune, T

    S. Mifune, T. Kanao, and T. Tanamoto, Effects of dis- sipation in reservoir computing using a spin qubit array, arXiv preprint arXiv:2410.07481 (2024)

    2024 arXiv

  68. [77]

    Rivas and S

    A. Rivas and S. F. Huelga,Open quantum systems, Vol. 10 (Springer, 2012)

    2012

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.