Estimation of the second-order coherence function using quantum reservoir and ensemble methods

Amir Rahmani; Andrzej Opala; Dogyun Ko; Micha{\l} Matuszewski; Stanis{\l}aw \'Swierczewski

REVIEW 4 major objections 7 minor 1 cited by

Estimation of the second-order coherence function using quantum reservoir and ensemble methods

T0 review · 4 major / 7 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a hybrid quantum reservoir computer and classical decision-tree ensemble can estimate the zero-time second-order coherence function $g^{(2)}(0)$ of a quantum light source from average intensity measurements alone…

desk verdict An honest numerical proof-of-principle for estimating g2(0) from reservoir intensities with ensemble trees; the cross-system generalization analysis is a strength, but the lack of noise-injection tests and code/data limits the practical claim. read the letter →

arxiv 2504.18205 v1 pith:4EZF5OZR submitted 2025-04-25 quant-ph physics.optics

Dogyun Ko , Stanis{l}aw Świerczewski , Andrzej Opala , Micha{l} Matuszewski , Amir Rahmani This is my paper · ORCID

classification quant-phphysics.optics

keywords quantumreservoircomputingg(2)(0)estimationsecond-ordercoherencedecisiontreeensemblephotonstatisticsmachinelearninginopticsintensity-basedmeasurementgeneralization

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

Measuring the zero-time second-order coherence $g^{(2)}(0)$ — the standard signature of photon bunching and antibunching — normally requires single-photon detectors and coincidence counting. The paper claims that a small, fixed quantum reservoir coupled to the source, followed by a classical decision-tree ensemble, can instead estimate $g^{(2)}(0)$ from time-resolved average occupation numbers of the reservoir nodes. Across four model quantum sources the reported test errors are small (e.g., 0.0035 for three-beam mixing, $7.47\times10^{-5}$ for coherent light mixed with a two-level emitter), and in several cases the reservoir decisively outperforms regression on the source intensity alone. The authors also show that a trained model can interpolate to unseen parameters of the same source, while transfer between fundamentally different sources remains limited.

What carries the argument

The central object is the quantum reservoir processor: a fixed, fully connected network of fermionic nodes with random couplings normalized to the spectral radius, initialized in vacuum, possibly incoherently pumped, and driven by the source state through a unidirectional cascade coupling described by a Lindblad master equation. Its job is to map each input quantum state into a higher-dimensional nonlinear response; the features used for regression are the time-resolved average occupation numbers $\langle \hat b_j^\dagger \hat b_j\rangle$ of the reservoir nodes. The final trainable stage is a bagging ensemble of decision trees (Random Forest or Extra-Trees) that learns the map from those occupation features to $g^{(2)}(0)$. The key observation is that this fixed nonlinear preprocessor often turns an ill-posed regression problem — identical source intensities with different $g^{(2)}(0)$ — into one that a simple tree ensemble can solve.

What would settle it

Take a trained reservoir-plus-ensemble model and feed it experimental intensity traces from a source whose $g^{(2)}(0)$ is independently measured by coincidence counting (e.g., resonance fluorescence from a single two-level emitter or an attenuated coherent state); if the model's predictions deviate from the coincidence-measured values by more than the reported test MSE once detection efficiency, dark counts, and timing jitter are introduced, the central claim that average-intensity measurements suffice would collapse.

Watch

Extended reading notes

Core claim

The paper's central discovery is that the information needed to fix $g^{(2)}(0)$ is present, in a learnable form, in the average intensity dynamics of a tiny quantum reservoir, even when the source intensity alone does not determine the correlation. Using a two-node reservoir of fermionic modes coupled unidirectionally to the source, the scheme records time-resolved occupations of the nodes, feeds those classical features to a bagged decision-tree regressor, and labels each sample with the analytically or master-equation-computed $g^{(2)}(0)$. On the test splits the estimator achieves MSE 0.0035 for a three-mode mixture of Fock, coherent, and thermal states (vs 0.154 without the reservoir), 0.037 for an emitter in a driven-dissipative cavity (vs 1.17 without the reservoir), and $7.47\times10^{-5}$ for a coherent state mixed with a two-level emitter at a beam splitter (vs 0.0089 without the reservoir). For a single photon-added squeezed state the source intensity already suffices (MSE about $1.03\times10^{-6}$); the reservoir matters when samples from several $m$ values are pooled. The model trained on detunings $\Delta_b \in \{1.0,1.6,1.8\}\gamma_a$ predicts the held-out curve $\Delta_b = 1.4\gamma_a$ with MSE 0.036, while cross-source transfer as tabulated is unreliable.

Load-bearing premise

The evaluation is entirely numerical and noise-free: the $g^{(2)}(0)$ labels come from the same analytic formulas and master equations that generate the source states, and reservoir node occupations are computed from ideal unitary-plus-Lindblad dynamics, so the reported low errors presume that experimental average-intensity detection adds no significant loss, background, or calibration error.

Editorial extensions

If this is right

A $g^{(2)}(0)$ measurement device could be built from a fixed two-node quantum reservoir plus classical post-processing, replacing single-photon coincidence counting with average-intensity readout.
For a known source class, the trained model can estimate $g^{(2)}(0)$ at parameter values it never saw, e.g., detuning $\Delta_b = 1.4\gamma_a$ with MSE 0.036, effectively acting as a black-box coherence meter for that source.
Cross-source transfer is not reliable: models trained on one source generalize to some configurations but fail on emitter-in-cavity states, so training data must be drawn from the target source's physical class.
The reservoir becomes essential when the source's mean occupation is not a single-valued function of $g^{(2)}(0)$, as in the emitter-in-cavity and mixed-$m$ photon-added cases; this motivates using reservoir dynamics whenever such ambiguities are present.
The authors propose extending the same framework from $g^{(2)}(0)$ to the time-delayed correlation function $g^{(2)}(\tau)$, which would capture Rabi oscillations and relaxation dynamics.

Reading between the lines

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

Because the reservoir is fixed and untrained, the scheme is essentially a physics-agnostic quantum feature extractor; the same trained ensemble could be recalibrated on experimental intensity traces from an existing setup, turning a numerical proof of principle into a lab calibration protocol.
A direct test of what the quantum reservoir adds would be to compare against classical nonlinear time-delay embeddings of the source intensity alone; if similar MSE is reached, the benefit may be comparable to a generic temporal feature map rather than specifically quantum.
Training with realistic detection inefficiencies and background noise, by sampling reservoir occupations with photon losses, is the natural next step; the paper's noise-free assumption is the main obstacle to direct experimental transfer.
The strong cross-system failures suggest that $g^{(2)}(0)$ is not a universal function of reservoir intensities; a practical system may need a small labeled calibration set for each new source architecture, e.g., via few-shot transfer.

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.

Desk Editor's Note

read the letter

This paper is a numerical proof-of-principle that a two-node quantum reservoir coupled to a source, with its average node occupations fed to random-forest/extra-trees, can estimate g^(2)(0) for four classes of quantum sources. The new bit is the specific combination of QRC with decision-tree ensembles for this task, and the systematic cross-system generalization test (Table I, Fig. 6). The work is honest: they include no-reservoir baselines, show that those baselines sometimes work (photon-added squeezed states) and sometimes fail (emitter in cavity), and they explicitly report that cross-system generalization is poor. The within-system interpolation test (Fig. 7) is a useful sanity check.

Credit where it is earned: the simulations are plausible, the MSE values are reported, and the generalization analysis is the kind of thing many authors skip. The paper does not oversell the cross-system claim.

The soft spots are real but manageable. The biggest one is the gap between the motivational language and the evidence: the abstract and conclusion suggest the method can replace single-photon detectors in practice, but all features come from noiseless expectation values of node occupations. A real intensity measurement has shot noise, loss, and dark counts. Decision-tree ensembles are piecewise constant, so they can be sensitive at leaf boundaries, and the low MSEs (e.g., 7e-5) likely reflect memorizing the deterministic map. The authors should inject noise into the features and show MSE degradation, or soften the claim to noise-free ideal detection. That said, this is a proof-of-principle paper, so the absence of an experiment is not fatal; the absence of a noise-injection test is a fixable omission.

Other issues: the normalized MSE metric is unconventional and makes cross-task comparison hard; no code or data is provided; no comparison to a classical reservoir or other nonlinear embeddings, so the "quantum" benefit is not isolated. I'd treat those as minor to moderate.

The citation pattern looks fine. No obvious issue. The fermionic reservoir modeling is not fully justified, but it's not a load-bearing flaw.

Who for? People working on quantum reservoir computing, optical coherence characterization, or ML for quantum optics. It deserves a serious referee; I would not desk-reject. My recommendation: send to peer review, but ask for noise-injection tests, code/data, and a classical baseline before considering publication.

Referee Report

4 major / 7 minor

Summary. The paper proposes a hybrid quantum-classical approach to estimate the zero-time second-order coherence function g^(2)(0) from average intensity measurements alone. A quantum reservoir (a small network of coupled modes) processes the source state, and the time-resolved average occupations of the reservoir nodes are used as features for a decision-tree ensemble (Random Forest/Extra-Trees) regressor. The authors test the method on four model systems: a statistical mixture of Fock, coherent, and thermal states; an emitter in a cavity; photon-added squeezed states; and the output of a beam splitter mixing a coherent state with a two-level emitter. They report low test MSE values for in-distribution tasks, and they analyze out-of-distribution generalization both across parameters of the same system and across different physical systems, finding that cross-system generalization is generally limited.

Significance. If the method works as claimed, it would offer a practical alternative to single-photon-detector-based measurements of g^(2)(0), requiring only average intensity detection. The paper's strengths include the diversity of source models considered, the honest treatment of cases where the reservoir is not needed (photon-added squeezed states), and the explicit evaluation of generalization to unseen parameters (Figure 7) and to different systems (Table I). The numerical simulations are plausible and the manuscript is generally well written. However, the practical claim is weakened by the absence of noise-injection tests and of a classical reservoir baseline, and the reported MSE metric is nonstandard. As a proof-of-principle simulation, the work is valuable, but the experimental relevance is not yet demonstrated.

major comments (4)

[Section III (all subsections) and IV] All reported MSE values are computed on noiseless simulated expectation values of reservoir occupations, e.g., the values 0.0035, 0.037, and 7.47e-5 in Sections III.A, III.B, and III.D. Since the central motivation is to replace single-photon detectors and time-correlated measurements with average intensity measurements, and real photodetection is subject to shot noise, inefficiency, and dark counts, the paper should include a noise-injection study (e.g., adding Poissonian or Gaussian noise to the features at realistic levels) to demonstrate that the estimator is robust. The piecewise-constant nature of decision-tree ensembles makes this test particularly important. Without it, the claim that average intensity measurements suffice is unsupported.
[Section II, Eq. (4)] The MSE metric defined in Eq. (4) is a normalized error whose denominator depends on the data, not the standard mean squared error. This makes the reported values difficult to interpret and to compare with standard benchmarks; for example, the value 7.47e-5 in Section III.D may appear deceptively small. The authors should report standard MSE or RMSE, and justify the choice of the normalized metric, or at least show that conclusions are unchanged under the standard metric.
[Sections III.B and III.C, and the title] The claim that the quantum reservoir is 'necessary' or that it enables learning is based only on a comparison to direct regression on the source occupation number without a reservoir (e.g., Fig. 3(c), Fig. 4(c)). A classical nonlinear feature extractor, such as a classical reservoir computer or a random nonlinear map, could provide a similar enhancement. Without such a baseline, the specific contribution of the quantum reservoir is not established, which is load-bearing given that the method is presented as QRC-based.
[Sections III.A through IV] The reported MSE values are point estimates. No error bars or standard deviations are given over random data splits, reservoir realizations, or tree ensemble seeds, and the reservoir hyperparameters (pump rate P, decay rates, pulse times, spectral radius) are chosen without a sensitivity analysis. Given the small dataset sizes (1000-8000 samples), the reader cannot assess the statistical reliability of the results or the robustness of the method to hyperparameter choice.

minor comments (7)

[Section II, Eq. (2)] The notation for the pulse functions is inconsistent: fk(t) and fsk(t) are both used; please clarify.
[Eq. (1)] The last term in the numerator, E(t), should be E(r,t) for consistency with the other operators.
[Section III.A] The text uses both 'ϕ' and 'φ' for the same spherical angle; please unify.
[Section IV, Figure 7] The text refers to 'panel (c)' for the testing result, but Figure 7 contains only panels (a) and (b); update the reference.
[Section III.C] The phrase 'sampled20% of it' should be 'sampled 20% of it'.
[Section I] The sentence 'allowing for estimation of g(2) or other correlation functions from averaged intensity measurements' appears twice; remove the duplicate.
[Acknowledgments] Typo: 'acknowldges' should be 'acknowledges'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the supervised regression is evaluated on held-out test data and does not derive its target from its inputs by construction.

full rationale

The paper trains a standard supervised estimator: reservoir-node intensities are the features, and analytically or numerically computed g(2)(0) values are the labels. The target g(2)(0) is not used to construct the input features, and the reported accuracies are measured on held-out test samples rather than on the training set. This is legitimate supervised learning on simulated data, not a derivation in which the conclusion is equivalent to the premise. The paper also includes cross-system generalization experiments and reports that performance across distinct systems is generally limited, which shows the evaluation is not a circular restatement of the training labels. The use of the prior QRC master equation from [8], which shares authors with the present work, is methodological rather than load-bearing: no uniqueness claim or alternative-forbidding theorem is invoked, and the reservoir dynamics are not adjusted to force the reported MSEs. The main limitation, that the simulations are noiseless and therefore do not prove robustness to real detector noise, is an external-validity concern rather than a circularity defect, and it is not evidence that the reported held-out MSE values are predetermined by the construction.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

No new physical entities are introduced; all degrees of freedom are standard bosonic or fermionic modes and density matrices. The free parameters and axioms listed here are the additional assumptions the reader must accept beyond the standard quantum-optics framework.

free parameters (4)

Trained decision-tree ensemble model W(s) = Learned from data for each source (Random Forest / Extra-Trees)
The estimator is a nonlinear regression function whose parameters are fit to training labels; reported test MSE measures this fit on held-out samples from the same distribution.
Reservoir random couplings J_ij and W_in = Not reported
Random couplings are normalized to the spectral radius, but exact values are not given, so the feature representation is underspecified.
Reservoir pump, decay, pulse, and measurement times = Not reported
Equation (2) requires values for gamma, p, eta, f_k(t), and measurement times; these hyperparameters determine the reservoir features and are not fully specified.
Source parameter sampling ranges = Not fully specified
Data distributions for theta, phi, mean photon number, alpha, r, Delta_a, and Delta_b are stated only partially; performance depends on these ranges.

assumptions (5)

domain assumption Ground-truth g2(0) labels from analytic formulas and master equations are correct.
Labels for supervised training are computed from Equations (8), (13), (14), and (10); any error in these models propagates into training and evaluation.
domain assumption The cascaded master equation Eq. (2) with fermionic Lindblad terms describes a physically realizable reservoir.
All QRC simulations rest on this evolution equation, including dissipators for fermionic nodes, without discussion of experimental realizability.
domain assumption Average reservoir occupation numbers are sufficient readout features.
Only diagonal occupation numbers are recorded; coherence information is discarded, so the method assumes this projection retains enough information for regression.
domain assumption A randomly coupled two-node reservoir provides adequate nonlinear feature expansion.
The method relies on the reservoir mapping input states to nonlinearly related intensity features; no proof or kernel analysis is given.
domain assumption Standard supervised learning generalization assumptions hold for decision-tree ensembles.
The paper assumes held-out test errors reflect performance on unseen samples from the same source and parameter distribution.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Estimation of the second-order coherence function using quantum reservoir and ensemble methods." pith.science (2026). https://pith.science/paper/4EZF5OZR

@misc{pith2026250418205,
  author       = {Pith},
  title        = {Pith review of: Estimation of the second-order coherence function using quantum reservoir and ensemble methods},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4EZF5OZR}},
  note         = {Machine review of arXiv:2504.18205}
}

read the original abstract

We propose a machine learning-based approach enhanced by quantum reservoir computing (QRC) to estimate the zero-time second-order correlation function g2(0). Typically, measuring g2(0) requires single-photon detectors and time-correlated measurements. Machine learning may offer practical solutions by training a model to estimate g2(0) solely from average intensity measurements. In our method, emission from a given quantum source is first processed in QRC. During the inference phase, only intensity measurements are used, which are then passed to a software-based decision tree-based ensemble model. We evaluate this hybrid quantum-classical approach across a variety of quantum optical systems and demonstrate that it provides accurate estimates of g2(0). We further extend our analysis to assess the ability of a trained model to generalize beyond its training distribution, both to the same system under different physical parameters and to fundamentally different quantum sources. While the model may yield reliable estimates within specific regimes, its performance across distinct systems is generally limited.

Figures

Figures reproduced from arXiv: 2504.18205 by the authors.

**Figure 1.** FIG. 1. (a) Setup for the Hanbury Brown and Twiss experiment to measure second-order coherence. A signal is sent through [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

**Figure 2.** FIG. 2. (a) Three-mode mixing in a setup consisting of three [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** FIG. 3. (a) The coupling between a bosonic emitter mode and a cavity photon mode can lead to interesting statistical behavior. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: FIG. 4. (a): Setup for generating photon-added squeezed states, where a photon is added to the squeezed state [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

**Figure 5.** Figure 5: FIG. 5. (a): In a beam splitter, two modes can be mixed to produce an output with interesting features. For example, a [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

**Figure 6.** Figure 6: FIG. 6. Partition-based MSE analysis. The true data for each [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

**Figure 7.** Figure 7: FIG. 7. Example of generalization learning in the case of an [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

Phase-Space Framework for Noisy Intermediate-Scale Quantum Optical Neural Networks
quant-ph 2025-07 conditional novelty 6.0 of 10

A phase-space simulation method makes large bosonic quantum reservoir networks tractable and reveals that their performance peaks at just five to seven modes.

Reference graph

Works this paper leans on

61 extracted references · 41 canonical work pages · cited by 1 Pith paper

[1]

R. H. Brown and R. T. and, Lxxiv. a new type of inter- ferometer for use in radio astronomy, The London, Edin- burgh, and Dublin Philosophical Magazine and Journal of Science 45, 663 (1954)

work page 1954
[2]

Dravins, High Time Resolution Astrophysics, Astro- phys, 1st ed

D. Dravins, High Time Resolution Astrophysics, Astro- phys, 1st ed. (Springer, 2008)

work page 2008
[3]

C. C. Gerry and P. L. Knight, Introductory Quantum Optics, 2nd ed. (Cambridge University Press, 2023)

2023
[4]

Sempere-Llagostera, G

S. Sempere-Llagostera, G. S. Thekkadath, R. B. Patel, W. S. Kolthammer, and I. A. Walmsley, Reducing g(2)(0) of a parametric down-conversion source via photon- number resolution with superconducting nanowire detec- tors, Opt. Express30, 3138 (2022)

work page 2022
[5]

del Valle, A

E. del Valle, A. Gonzalez-Tudela, F. P. Laussy, C. Teje- dor, and M. J. Hartmann, Theory of frequency-filtered andtime-resolved n-photoncorrelations,Phys.Rev.Lett. 109, 183601 (2012)

work page 2012
[6]

Krapick, H

S. Krapick, H. Herrmann, V. Quiring, B. Brecht, H. Suche, and C. Silberhorn, An efficient integrated two- color source for heralded single photons, New Journal of Physics 15, 033010 (2013)

work page 2013
[7]

Ghosh, K

S. Ghosh, K. Nakajima, T. Krisnanda, K. Fujii, and T. C. H. Liew, Quantum neuromorphic computing with reservoir computing networks, Advanced Quantum Tech- 10 nologies 4, 2100053 (2021)

work page 2021
[8]

Ghosh, A

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

work page 2019

Show all 61 references

[9]

Tanaka, T

G. Tanaka, T. Yamane, J. B. Héroux, R. Nakane, N. Kanazawa, S. Takeda, H. Numata, D. Nakano, and A. Hirose, Recent advances in physical reservoir comput- ing: A review, Neural Networks115, 100 (2019)

2019
[10]

Fujii and K

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

2017
[11]

Mujal, R

P. Mujal, R. Martínez-Peña, J. Nokkala, J. García-Beni, G. L. Giorgi, M. C. Soriano, and R. Zambrini, Oppor- tunities in quantum reservoir computing and extreme learning machines, Advanced Quantum Technologies4, 2100027 (2021)

2021
[12]

Martínez-Peña, G

R. Martínez-Peña, 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)

2021
[13]

Čindrak, B

S. Čindrak, B. Donvil, K. Lüdge, and L. Jaurigue, En- hancing the performance of quantum reservoir comput- ing and solving the time-complexity problem by artificial memory restriction, Phys. Rev. Res.6, 013051 (2024)

2024
[14]

H. Dong, L. Jaurigue, and K. Lüdge, Time-multiplexed reservoir computing with quantum-dot lasers: Impact of charge-carrier scattering timescale, physica status solidi (RRL) – Rapid Research Letters , 2400433 (2025)

2025
[15]

T. K. Ho, Random decision forests, in Proceedings of 3rd International Conference on Document Analysis and Recognition, Vol. 1 (1995) pp. 278–282

1995
[16]

Geurts, D

P. Geurts, D. Ernst, and L. Wehenkel, Extremely ran- domized trees, Mach. Learn.63, 42 (2006)

2006
[17]

Hastie, R

T. Hastie, R. Tibshirani, and J. Friedman, The Ele- ments of Statistical Learning , 2nd ed. (Springer New York, 2009)

2009
[18]

M. M. Nieto, Displaced and squeezed number states, Physics Letters A229, 135 (1997)

1997
[20]

N. B. Grosse, T. Symul, M. Stobińska, T. C. Ralph, and P. K. Lam, Measuring photon antibunching from contin- uous variable sideband squeezing, Phys. Rev. Lett.98, 153603 (2007)

2007
[21]

Delteil, T

A. Delteil, T. Fink, A. Schade, S. Höfling, C. Schnei- der, and A. İmamoğlu, Towards polariton blockade of confined exciton–polaritons, Nature Materials 18, 219 (2019)

2019
[22]

Makarov, Theory for the beam splitter in quantum op- tics: Quantum entanglement of photons and their statis- tics, hom effect, Mathematics10, 4794 (2022)

D. Makarov, Theory for the beam splitter in quantum op- tics: Quantum entanglement of photons and their statis- tics, hom effect, Mathematics10, 4794 (2022)

2022
[23]

Sherstinsky, Fundamentals of recurrent neural net- work (rnn) and long short-term memory (lstm) network, Physica D: Nonlinear Phenomena404, 132306 (2020)

A. Sherstinsky, Fundamentals of recurrent neural net- work (rnn) and long short-term memory (lstm) network, Physica D: Nonlinear Phenomena404, 132306 (2020)

2020
[24]

echo state

H. Jaeger, The "echo state" approach to analysing and training recurrent neural networks , Tech. Rep. (GMD - German National Research Institute for Computer Sci- ence, 2001)

2001
[25]

Hülser, F

T. Hülser, F. Köster, L. Jaurigue, and K. Lüdge, Role of delay-times in delay-based photonic reservoir computing, Opt. Mater. Express12, 1214 (2022)

2022
[26]

H. J. Carmichael, Quantum trajectory theory for cas- caded open systems, Phys. Rev. Lett.70, 2273 (1993)

1993
[27]

J. C. L. Carreño and F. P. Laussy, Excitation with quan- tum light. i. exciting a harmonic oscillator, Phys. Rev. A 94, 063825 (2016)

2016
[28]

Cucchi, S

M. Cucchi, S. Abreu, G. Ciccone, D. Brunner, and H. Kleemann, Hands-on reservoir computing: a tutorial for practical implementation, Neuromorphic Computing and Engineering 2, 032002 (2022)

2022
[29]

Colliot, Machine Learning for Brain Disorders , Neu- romethods, Vol

O. Colliot, Machine Learning for Brain Disorders , Neu- romethods, Vol. 197 (Springer, 2023)

2023
[30]

Harrington, Machine Learning in Action (Manning Publications Co., USA, 2012)

P. Harrington, Machine Learning in Action (Manning Publications Co., USA, 2012)

2012
[31]

S. L. Braunstein and P. van Loock, Quantum informa- tion with continuous variables, Rev. Mod. Phys.77, 513 (2005)

2005
[32]

Spagnolo, C

N. Spagnolo, C. Vitelli, L. Aparo, P. Mataloni, F. Scia- rrino, A. Crespi, R. Ramponi, and R. Osellame, Three- photon bosonic coalescence in an integrated tritter, Na- ture Communications 4, 1606 (2013)

2013
[33]

K. J. Vahala, Optical microcavities, Nature 424, 839 (2003)

2003
[34]

M. P. C. Taverne, Y.-L. D. Ho, X. Zheng, L. Chen, C.- H. N. Fang, and J. Rarity, Strong light confinement in rod-connected diamond photonic crystals, Opt. Lett.43, 5202 (2018)

2018
[35]

Liu, Z.-K

R. Liu, Z.-K. Zhou, Y.-C. Yu, T. Zhang, H. Wang, G. Liu, Y. Wei, H. Chen, and X.-H. Wang, Strong light- matter interactions in single open plasmonic nanocavities atthequantumopticslimit,Phys.Rev.Lett. 118,237401 (2017)

2017
[36]

black box

and single-photon sources [37]. To model such physics, we may consider the following Hamiltonian Hs =∆a ˆa†ˆa + ∆b ˆb†ˆb +J ˆa†ˆb + ˆaˆb† + Ub 2 ˆb†ˆb†ˆbˆb + Ua 2 ˆa†ˆa†ˆaˆa + Ωaˆa + Ωbˆb +h.c., (9) where ˆa, ˆb are bosonic operators, ∆a,b/ℏ is the differ- ence between frequen...

2021
[37]

Maryam, A

W. Maryam, A. V. Akimov, R. P. Campion, and A. J. Kent, Dynamics of a vertical cavity quantum cascade phonon laser structure, Nature Communications4, 2184 (2013)

2013
[38]

T. T. H. Do, M. Nonahal, C. Li, V. Valuckas, H. H. Tan, A. I. Kuznetsov, H. S. Nguyen, I. Aharonovich, and S. T. Ha, Room-temperature strong coupling in a single- photon emitter-metasurface system, Nature Communica- tions 15, 2281 (2024)

2024
[39]

T. A. Khudaiberganov, I. Y. Chestnov, and S. S. Demirchyan, Photon statistics of semiconductor micro- cavity polaritons, Journal of Physics: Conference Series 951, 012032 (2018)

2018
[40]

Zhang, J

Z. Zhang, J. Ding, and H.-T. Wang, Entanglement and nonlocality in a coupled-cavity system, Photon. Res.5, 224 (2017)

2017
[41]

Dahan, G

D. Dahan, G. Arwas, and E. Grosfeld, Classical and quantum chaos in chirally-driven, dissipative bose- hubbardsystems,npjQuantumInformation 8,14(2022)

2022
[42]

P.D.DrummondandD.F.Walls,Quantumtheoryofop- tical bistability. i. nonlinear polarisability model, Journal of Physics A: Mathematical and General13, 725 (1980)

1980
[43]

G. S. Agarwal and K. Tara, Nonclassical properties of states generated by the excitations on a coherent state, Phys. Rev. A43, 492 (1991)

1991
[44]

Sperling, W

J. Sperling, W. Vogel, and G. S. Agarwal, Quantum state engineering by click counting, Phys. Rev. A89, 043829 (2014)

2014
[45]

Clausen, M

J. Clausen, M. Dakna, L. Knöll, and D. G. Welsch, Con- ditional quantum-state transformation at a beam split- ter, Journal of Optics B: Quantum and Semiclassical Op- tics 1, 332 (1999)

1999
[46]

Jiang and L

K. Jiang and L. Wei, Generating antibunching light by beam-splitting photon number resolved detection from the superbunching squeezed vacuum, Physics Letters A 11 403, 127396 (2021)

2021
[47]

T. Das, R. Prabhu, A. Sen(De), and U. Sen, Superiority of photon subtraction to addition for entanglement in a multimode squeezed vacuum, Phys. Rev. A93, 052313 (2016)

2016
[48]

N.Sangouard, C.Simon, H.deRiedmatten,andN.Gisin, Quantum repeaters based on atomic ensembles and linear optics, Rev. Mod. Phys.83, 33 (2011)

2011
[49]

C. K. Hong, Z. Y. Ou, and L. Mandel, Measurement of subpicosecond time intervals between two photons by in- terference, Phys. Rev. Lett.59, 2044 (1987)

1987
[50]

Zubizarreta Casalengua, J

E. Zubizarreta Casalengua, J. C. López Carreño, F. P. Laussy, and E. del Valle, Tuning photon statistics with coherent fields, Phys. Rev. A101, 063824 (2020)

2020
[51]

M. O. Scully and M. S. Zubairy,Quantum Optics (Cam- bridge University Press, 1997)

1997
[52]

Hanschke, L

L. Hanschke, L. Schweickert, J. C. L. Carreño, E. Schöll, K. D. Zeuner, T. Lettner, E. Z. Casalengua, M. Reindl, S.F.C.daSilva, R.Trotta, J.J.Finley, A.Rastelli, E.del Valle, F. P. Laussy, V. Zwiller, K. Müller, and K. D. Jöns, Origin of antibunching in resonance fluorescence,...

2020
[53]

M. C. Caro, H.-Y. Huang, M. Cerezo, K. Sharma, A. Sornborger, L. Cincio, and P. J. Coles, Generalization in quantum machine learning from few training data, Na- ture Communications 13, 4919 (2022)

2022
[54]

Killoran, T

N. Killoran, T. R. Bromley, J. M. Arrazola, M. Schuld, N. Quesada, and S. Lloyd, Continuous-variable quantum neural networks, Phys. Rev. Res.1, 033063 (2019)

2019
[55]

Bauwens, G

I. Bauwens, G. Van der Sande, P. Bienstman, and G. Ver- schaffelt, Using photonic reservoirs as preprocessors for deep neural networks, Frontiers in Physics10, 1051941 (2022)

2022
[56]

Nokkala, G

J. Nokkala, G. L. Giorgi, and R. Zambrini, Retriev- ing past quantum features with deep hybrid classical- quantum reservoir computing, Machine Learning: Sci- ence and Technology5, 035022 (2024)

2024
[57]

X.Lin, Z.Chen,andZ.Wei,Quantifyingquantumentan- glement via a hybrid quantum-classical machine learning framework, Phys. Rev. A107, 062409 (2023)

2023
[58]

M. C. Caro, H.-Y. Huang, N. Ezzell, J. Gibbs, A. T. Sornborger, L. Cincio, P. J. Coles, and Z. Holmes, Out- of-distribution generalization for learning quantum dy- namics, Nature Communications14, 3751 (2023)

2023
[59]

Haug and M

T. Haug and M. S. Kim, Generalization of quantum ma- chine learning models using quantum fisher information metric, Phys. Rev. Lett.133, 050603 (2024)

2024
[60]

K. M. Birnbaum, A. Boca, R. Miller, A. D. Boozer, T. E. Northup,andH.J.Kimble,Photonblockadeinanoptical cavity with one trapped atom, Nature436, 87 (2005)

2005
[61]

M. A. P. Marcos, E. Z. Casalengua, E. del Valle, and F. P. Laussy, Correlations in circular quantum cascades, Phys. Rev. A111, 023704 (2025)

2025
[62]

Lemonde, N

M.-A. Lemonde, N. Didier, and A. A. Clerk, Antibunch- ing and unconventional photon blockade with gaussian squeezed states, Phys. Rev. A90, 063824 (2014)

2014

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [1]

R. H. Brown and R. T. and, Lxxiv. a new type of inter- ferometer for use in radio astronomy, The London, Edin- burgh, and Dublin Philosophical Magazine and Journal of Science 45, 663 (1954)

work page 1954

[2] [2]

Dravins, High Time Resolution Astrophysics, Astro- phys, 1st ed

D. Dravins, High Time Resolution Astrophysics, Astro- phys, 1st ed. (Springer, 2008)

work page 2008

[3] [3]

C. C. Gerry and P. L. Knight, Introductory Quantum Optics, 2nd ed. (Cambridge University Press, 2023)

2023

[4] [4]

Sempere-Llagostera, G

S. Sempere-Llagostera, G. S. Thekkadath, R. B. Patel, W. S. Kolthammer, and I. A. Walmsley, Reducing g(2)(0) of a parametric down-conversion source via photon- number resolution with superconducting nanowire detec- tors, Opt. Express30, 3138 (2022)

work page 2022

[5] [5]

del Valle, A

E. del Valle, A. Gonzalez-Tudela, F. P. Laussy, C. Teje- dor, and M. J. Hartmann, Theory of frequency-filtered andtime-resolved n-photoncorrelations,Phys.Rev.Lett. 109, 183601 (2012)

work page 2012

[6] [6]

Krapick, H

S. Krapick, H. Herrmann, V. Quiring, B. Brecht, H. Suche, and C. Silberhorn, An efficient integrated two- color source for heralded single photons, New Journal of Physics 15, 033010 (2013)

work page 2013

[7] [7]

Ghosh, K

S. Ghosh, K. Nakajima, T. Krisnanda, K. Fujii, and T. C. H. Liew, Quantum neuromorphic computing with reservoir computing networks, Advanced Quantum Tech- 10 nologies 4, 2100053 (2021)

work page 2021

[8] [8]

Ghosh, A

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

work page 2019