Pith. sign in

REVIEW 4 major objections 5 minor 46 references

Stochastic Quantum Spiking Neural Networks with Quantum Memory and Local Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper claims that spiking neurons built on multi-qubit circuits with unmeasured memory qubits, trained by a local single-shot rule, beat classical and prior quantum spiking networks at matched parameter counts on standard and…

desk verdict Solid new neuron model and local learning rule, but the headline accuracy claims are produced by a surrogate classical simulator, not the hardware-compatible rule, so the paper needs a direct equivalence check before the main claim is credible. read the letter →

arxiv 2506.21324 v3 pith:FQDVYKB4 submitted 2025-06-26 cs.NE cs.LG

Jiechen Chen , Bipin Rajendran , Osvaldo Simeone This is my paper · ORCID
classification cs.NEcs.LG
keywords quantumneuromorphiccomputingspikingneuralnetworksstochasticneuronmemorylocallearningruleszero-thordergradientestimationsingle-shotmeasurementevent-drivencomputation
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 sets out to remove the two bottlenecks of existing quantum spiking neurons: classical memory on a single qubit, whose firing probability must be re-estimated through repeated measurements, and training by classical backpropagation. Its stochastic quantum spiking (SQS) neuron runs instead on $N$ input-output qubits plus $N_{\mathrm{mem}}$ memory qubits, leaves the memory qubits unmeasured so quantum information persists across time steps, and emits spikes from a single measurement of the input-output qubits. The authors build networks of such neurons and derive a local learning rule that needs only a constant number $M$ of global forward passes, after which each neuron estimates its own gradient from local perturbations and a scalar feedback signal. They report about 98% test accuracy on USPS with this rule on current noisy quantum hardware under error mitigation, and report that SQSNNs consistently beat a classical spiking network, a variational quantum classifier, and the prior QLIF spiking network on MNIST, FMNIST, KMNIST, and the neuromorphic MNIST-DVS dataset at matched trainable parameters. If these results hold, quantum spiking networks become trainable directly on device for event-driven tasks rather than only on classical simulators.

What carries the argument

The load-bearing object is the SQS neuron: $N$ input-output qubits plus $N_{\mathrm{mem}}$ memory qubits that jointly evolve through a trainable unitary each time step, after which only the input-output qubits are measured and the memory qubits' post-measurement state $\sigma^{\mathrm{mem}}_t$ is carried to the next step — effectively a dynamic parameterized quantum circuit with state space dimension $2^{N+N_{\mathrm{mem}}}$ replacing the single-qubit membrane potential of the QLIF neuron. The second piece of machinery is the local zero-th order learning rule: $M$ global forward passes supply hidden spike samples and scalar feedback losses $\ell_t(m)$; each neuron then estimates its own gradient locally, applying simultaneous perturbation stochastic approximation (SPSA) with Rademacher perturbations to its synaptic weights and the parameter-shift rule to its circuit angles. This moves the learning cost from one global forward pass per parameter to a constant $M$ global passes plus $2TM_p|S|(M_{\mathrm{som}}G+M_{\mathrm{syn}})$ local passes per iteration.

What would settle it

Compare the proposed local zero-th order rule — not the surrogate-gradient substitute — against the surrogate on the same SQSNN architecture for at least one of MNIST, FMNIST, or KMNIST at the paper's matched parameter counts, either on a noiseless classical simulation or on quantum hardware. If final test accuracy (or the distribution of output spike trains on held-out inputs) diverges by more than a few points, the claim that the local rule achieves the reported benchmark performance fails; a cheaper first screen is to compare the spike-train statistics produced by both rules on identical held-out inputs.

Watch

Extended reading notes

Core claim

The central claim is that a spiking neuron whose internal state lives on a multi-qubit Hilbert space, with a subset of qubits deliberately left unmeasured to carry quantum memory forward in time, is a strictly richer computational unit than the single-qubit QLIF neuron — and that networks of such neurons can be trained without global classical backpropagation. At each time step the SQS neuron encodes weighted presynaptic spikes as rotation angles $\varphi_t^n$, applies a trainable unitary $U(z_t,\theta)=V(\theta)(\bigotimes_n R_X(\varphi_t^n)\otimes I)$ to the whole register, and measures only the input-output qubits; the post-measurement state of the memory qubits becomes the input state of the next step. Because the joint spiking probability of the network factorizes over time and over neurons, the negative log-likelihood is bounded above by a Monte-Carlo-estimable objective, from which the authors derive a local zero-th order learning rule: hidden-neuron spike samples and scalar losses come from $M$ global forward passes, and each neuron then estimates its own gradient using simultaneous perturbation for synaptic weights and the parameter-shift rule for circuit angles. The experiments show the local rule matching an idealized exact rule (about 98% on USPS with error mitigation), and, switching to the classical surrogate-gradient simulator the paper introduces for larger tasks, the SQSNN is reported to achieve the highest accuracy among SNN, QNN, QLIF-SNN, and SQSNN on MNIST, FMNIST, and KMNIST, and to outperform both spiking baselines on MNIST-DVS at matched parameter counts while spiking less than the classical SNN.

Load-bearing premise

The load-bearing premise is stated in the paper itself (Section V-D and Appendix A): after the small USPS hardware study, all larger benchmarks are run on a classical simulator that replaces stochastic spikes with sigmoid probabilities and trains by backpropagation, under the stated assumption that this surrogate-gradient scheme closely approximates the learning dynamics of the proposed local rule. If that surrogate dynamics diverges from what the local rule actually learns on hardware, the headline accuracy comparisons do not follow from the proposed training method.

Editorial extensions

If this is right

  • Training can move onto the quantum device: the rule needs only a constant number $M$ of global forward passes, after which every neuron updates from local measurements and a scalar feedback signal, with no classical backpropagation through the network.
  • Inference cost stops scaling with measurement shots: single-shot Born-rule spike emission replaces the multi-shot probability estimation of QLIF neurons, matching the event-driven principle that computation happens only when spikes occur.
  • Multi-qubit memory buys accuracy without proportional spiking cost: on FMNIST, enlarging the register from $(N=1, N_{\mathrm{mem}}=1)$ to $(N=2, N_{\mathrm{mem}}=2)$ raises accuracy while keeping spike density nearly flat.
  • In streaming and data-scarce regimes the architecture is reported to win on both axes: on MNIST-DVS and in the neuromorphic integrated sensing and communications scenario, SQSNN keeps higher accuracy than a classical SNN while emitting fewer spikes.
  • The probabilistic autoregressive formulation with a Jensen-bound objective carries over as a template for training other dynamic quantum circuits with mid-circuit measurements.

Reading between the lines

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

  • The surrogate-to-local validation shown in Fig. 7(b) is limited to the small USPS task; matching the surrogate-trained SQSNN against the same architecture trained by the local rule on at least one large benchmark would settle whether the accuracy gap over baselines is a property of the learning rule or only of the architecture.
  • If the surrogate assumption holds, the reported gains stem from the multi-qubit quantum memory itself: the natural control is the same SQSNN architecture trained with both rules from identical initializations, which would isolate the contribution of the training algorithm.
  • The noisier-but-better operating point ($M=1$, $M_{\mathrm{syn}}=5$, no error mitigation) suggests measurement noise rather than gradient variance was the binding constraint on the hardware run, so improvements in mid-circuit measurement or error mitigation could shift the optimal operating point and shrink the shot budget.
  • Because the network factorizes as a probabilistic autoregressive model with a three-factor local update, SQSNNs are structurally compatible with online and continual learning on streaming data — a direction the paper leaves for future work but that the derivation already supports.
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

4 major / 5 minor

Summary. This paper proposes a stochastic quantum spiking neuron (SQS) with multiple input-output qubits and unmeasured memory qubits, yielding an autoregressive spiking model with single-shot, event-driven spike generation. The authors derive a local zero-th order learning rule from an upper bound on the negative log-likelihood, using SPSA for synaptic weights and parameter-shift rules for PQC angles, with global scalar feedback to hidden neurons. They report a hardware demonstration on IBM Brisbane for binary USPS classification, and then use a classical surrogate-gradient simulator (Appendix A) to evaluate the model on MNIST, FMNIST, KMNIST, MNIST-DVS, and a neuromorphic integrated sensing and communications task, claiming that SQSNN outperforms QLIF-SNN, a classical SNN, and a QNN baseline at matched parameter counts.

Significance. The SQS neuron architecture and the local learning rule are novel, and the derivation of the local rule from a maximum-likelihood objective is principled. If the surrogate-to-local transfer were established, this would be a meaningful step toward trainable quantum neuromorphic systems with quantum memory, single-shot readout, and local training. The paper is commendably explicit about using a surrogate for the large-scale experiments and about the exact hardware configuration used. However, the current evidence does not yet establish the central claim, because the headline accuracy comparisons come from a different training algorithm than the proposed local rule.

major comments (4)
  1. [V-D, Appendix A, Figs. 8-11] The headline accuracy results on MNIST, FMNIST, KMNIST, MNIST-DVS, and N-ISAC are obtained with the surrogate-gradient classical simulator described in Appendix A, not with the proposed local zero-th order learning rule (Algorithm 1). The paper states this directly: 'we henceforth employ classical simulations using a surrogate gradient-based approach to approximate the learning dynamics of the proposed rule.' The surrogate differs from Algorithm 1 in the objective (Eq. (37) replaces the likelihood-bound objective of Eq. (26) with a cross-entropy loss over spike rates), in the backward pass (Eq. (39) replaces stochastic spike sampling with a sigmoid surrogate), and in the use of global backpropagation instead of the local SPSA/PSR estimates of Eqs. (34)-(35). The only evidence that the surrogate tracks the local rule is Fig. 7(b), which uses a binary USPS task with two output neurons and no hidden layer. That small-scale comparison does not establish equivalence for the 1000-hidden-neuron rate-coded tasks or for the T=20 neuromorphic event-stream task. Consequently, Figs. 8-11 demonstrate the performance of a classically trained surrogate model, and the paper's central claim that the proposed hardware-compatible local rule yields state-of-the-art accuracy is not established.
  2. [V-D and V-C, Fig. 7] The hardware demonstration on IBM Brisbane only exercises a network with a single output layer of two neurons (Section V-C states 'For the USPS dataset, we use fully connected models with a single, output, layer of two neurons'). This means the hidden-neuron part of the local rule—the REINFORCE-style gradient in Eq. (33), the global feedback signals ℓ(m), and the local hidden-neuron updates—is never tested on quantum hardware. The hardware experiment therefore validates only output-neuron SPSA and parameter-shift estimates, not the 'local learning' mechanism that is the paper's main algorithmic contribution. To support the claim that SQSNNs can be trained on hardware with hidden layers, an experiment or a realistic noise-model simulation with at least one hidden SQS neuron is needed.
  3. [IV-D and Eq. (25), Appendix A Eq. (37)] The local rule minimizes an upper bound on the negative log-likelihood obtained by Jensen's inequality in Eq. (25). The tightness of this bound is never quantified, and the surrogate in Appendix A replaces the bound with yet another objective (cross-entropy on smoothed spike rates). The paper does not analyze whether the bound gap or the surrogate replacement changes the ordering of models or the qualitative conclusions. This compounds the transfer problem in the first comment: even a perfect match between the surrogate and the ideal local rule would not automatically validate the original likelihood-based objective as the quantity being optimized. At minimum, the authors should report bound tightness on a small task and justify why optimizing the bound is expected to yield the same model ranking as optimizing the true likelihood.
  4. [IV-F, Eq. (35)] The parameter-shift estimator in Eq. (35) is not fully specified for hardware implementation. It divides by Tr(ρI_{i,t} S_{i,t}) without saying how this denominator is estimated from local measurements; if it is estimated from the same finite shot counts as the numerator, the estimator is biased and can be undefined when the estimated probability is zero. Similarly, the SPSA gradient in Eq. (34) applies logarithms to estimated probabilities, which are problematic for zero-count estimates with few shots. These details are relevant to the claim that the rule is 'hardware-friendly', and they should be addressed by specifying shot allocations or regularized estimators.
minor comments (5)
  1. [Algorithm 1] The pseudocode in Algorithm 1 lists the local gradient estimation inside a loop over t, but Eqs. (30) and (33) sum over t after collecting M global trajectories; the ordering should be clarified to avoid the impression that gradients are computed sequentially in time before feedback is available.
  2. [Appendix A, Eq. (37)] In Eq. (37), the spike rate r_i is defined as a full-sequence average, while the loss sums over time steps with per-step targets o^n_{i,t}; please clarify whether the target is constant across t (as in rate encoding) or whether a time-dependent rate should be used.
  3. [V-E and V-F, Figs. 8-9] The box plots in Figs. 8 and 9 do not state the number of independent runs or seeds; please report these, and indicate whether the 'consistently highest' accuracy claims are supported by non-overlapping intervals or a statistical test.
  4. [V-E, Fig. 8(b)] The average spike count per time step is an aggregate over the whole network, but the models have different neuron counts (e.g., 1000 hidden SQS neurons vs adjusted SNN/QLIF neuron counts). The comparison would be easier to interpret if per-neuron spike rates were reported alongside the aggregate counts.
  5. [VI] There are minor typos, e.g., 'pertubation-based' in the conclusions, and inconsistent capitalization such as 'ibm brisbane'.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular step found: the learning rule is derived from a stated ML objective via Jensen's bound with standard SPSA/parameter-shift estimators written out in full; the acknowledged surrogate-gradient replacement for large-scale benchmarks is an external-validity caveat, not a by-construction reduction.

full rationale

Walked end-to-end, the paper's derivation chain contains no step that reduces to its own inputs. The training objective is the standard maximum-likelihood log-loss (22)-(23); the tractable objective is obtained by Jensen's inequality in (25), a mathematical identity applied to the model's own defined conditional probabilities (19)-(21) and written out explicitly; and the local estimators (30), (33), (34), (35) are textbook SPSA and parameter-shift rules applied to the gradient of log Tr(ρ S), a quantity defined from the neuron's quantum state (14)-(17) independently of any test label. Self-citations are present—[18], [19], [33], [41], and [46] share co-author Simeone—but none is load-bearing in a circular sense: the REINFORCE gradient (31), the Jensen bound (25), and the QRNN-style unitary (13) are all stated in full in this paper, so the cited works supply provenance rather than proof. The step that weakens the paper is not definitional but evidential: Section V-D explicitly hands off all large-scale benchmarks to a classical surrogate-gradient simulator, stating 'we henceforth employ classical simulations using a surrogate gradient-based approach to approximate the learning dynamics of the proposed rule (see Appendix),' with the surrogate defined in Appendix A as cross-entropy on spike rates (37) plus a sigmoid backward pass (39). That surrogate is not fitted to reproduce the local rule's gradient estimates or outputs (no calibration to the local rule is described), and its equivalence to the local rule is checked only on the binary USPS task (Fig. 7(b)); therefore the MNIST/FMNIST/KMNIST, MNIST-DVS, and N-ISAC accuracy figures are genuine predictions of a differently trained model, not results forced by construction. Presenting those figures under the single name 'SQSNN' risks over-generalization, but that is a validity concern, not a circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The central derivation relies on standard probability factorization, Jensen's inequality, and standard quantum mechanics (Born rule, density matrix evolution). No new physical entities or exotic axioms are introduced. The main non-standard assumption is the equivalence between the surrogate-gradient classical simulator and the proposed local rule, which is stated explicitly in the paper. The free parameters are the learned weights and PQC angles, plus the regularization strength; the ansatz choice itself is a design decision rather than a fitted constant.

free parameters (4)
  • PQC ansatz parameters θ (CRX angle plus two RX angles per neuron) = Learned on training data; final values not reported
    These variational parameters define the quantum circuit transformation and are optimized on the training set. The ansatz choice (CRX plus single-qubit RX rotations) is a design choice in Section V-C.
  • Synaptic weights wn per link = Learned on training data; final values not reported
    Real-valued learnable weights converting presynaptic spike counts into rotation angles via Eq. (11)-(12). The rotation mapping is fixed as fs(z)=πz, but the weights themselves are free parameters.
  • Sparsity regularization strength λ in Eq. (36) = Varied over a range; values plotted in Fig. 9
    Hyperparameter controlling the spiking rate regularizer, chosen per dataset/experiment.
  • USPS input encoding probability cap at 0.5 = 0.5
    The rate-encoding scheme caps the Bernoulli firing probability at 0.5. This is an encoding choice that affects the effective input statistics across all compared models equally, so it is a modeling parameter rather than a fitted constant.
assumptions (5)
  • standard math The factorization in Eq. (20) of the joint spike distribution over time and neurons, conditioned on the history of each neuron's parents, is a valid probability model for a directed graph of stochastic neurons.
    This is the standard chain-rule factorization assumption for probabilistic spiking networks, consistent with [19,33]. It is invoked to derive the local learning rule.
  • standard math Jensen's inequality bound in Eq. (25) is a legitimate upper bound on the negative log-likelihood and the surrogate objective is reasonable to minimize.
    The math is standard, but the paper optimizes the bound, not the exact log-likelihood, and does not characterize the tightness gap. This is a genuine assumption about optimization quality.
  • domain assumption The quantum device used in hardware experiments (ibm_brisbane) supports mid-circuit measurements and resets as required by the QRNN-style memory update (Eq. 17).
    The SQS neuron requires measuring input-output qubits while preserving memory qubit states. The paper cites IBM hardware and specific platforms (trapped ions, neutral atoms) as supporting this, but does not verify that the production device used has the required connectivity/measurement capabilities.
  • domain assumption Surrogate-gradient classical simulation faithfully approximates the learning dynamics of the proposed local zero-th order rule for large datasets.
    This is stated openly in Section V-D: 'we henceforth employ classical simulations using a surrogate gradient-based approach to approximate the learning dynamics of the proposed rule.' The only direct evidence for this equivalence is the USPS comparison in Fig. 7b.
  • domain assumption The single-shot Born-rule sample of the input-output qubits gives a useful training signal for the local learning rule without needing expectation values at inference.
    The paper's motivation is event-driven single-shot inference, but the local gradient estimations in Eqs. (34)-(35) still require Mp shots to estimate expectations during training. The paper acknowledges this with its shot-count analysis. The assumption is that single-shot inference preserves accuracy after training with multiple prediction shots.
invented entities (1)
  • SQS neuron: multi-qubit spiking unit with N input-output qubits and Nmem memory qubits
    purpose: Realizes a spiking neuron with internal quantum memory and single-shot probabilistic spike generation for quantum neuromorphic models.
    This is a new model component defined by the paper, not an independently discovered physical entity. Its usefulness is evidenced only by the paper's simulations and small-scale hardware test.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Stochastic Quantum Spiking Neural Networks with Quantum Memory and Local Learning." pith.science (2026). https://pith.science/paper/FQDVYKB4

@misc{pith2026250621324,
  author       = {Pith},
  title        = {Pith review of: Stochastic Quantum Spiking Neural Networks with Quantum Memory and Local Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FQDVYKB4}},
  note         = {Machine review of arXiv:2506.21324}
}
read the original abstract

Neuromorphic and quantum computing have recently emerged as promising paradigms for advancing artificial intelligence, each offering complementary strengths. Neuromorphic systems built on spiking neurons excel at processing time series data efficiently through sparse, event-driven computation, consuming energy only upon input events. Quantum computing, on the other hand, operates on state spaces that grow exponentially in dimension with the number of qubits -- as a consequence of tensor-product composition -- with quantum states admitting superposition across basis states and entanglement between subsystems. Hybrid approaches combining these paradigms have begun to show potential, but existing quantum spiking models have important limitations. Notably, they implement classical memory mechanisms on single qubits, requiring repeated measurements to estimate firing probabilities, while relying on conventional backpropagation for training. In this paper, we propose a novel stochastic quantum spiking (SQS) neuron model that addresses these challenges. The SQS neuron uses multi-qubit quantum circuits to realize a spiking unit with internal quantum memory, enabling event-driven probabilistic spike generation in a single shot during inference. Furthermore, we study networks of SQS neurons, dubbed SQS neural networks (SQSNN), and demonstrate that they can be trained via a hardware-friendly local learning rule, eliminating the need for global classical backpropagation. The proposed SQSNN model is shown via experiments with both conventional and neuromorphic datasets to improve over previous quantum spiking neural networks, as well as over classical counterparts, when fixing the overall number of trainable parameters, highlighting its potential for event-driven applications such as neuromorphic integrated sensing and communications (N-ISAC).

Figures

Figures reproduced from arXiv: 2506.21324 by the authors.

Figure 1
Figure 1. Schematic of an SQS neuron with N = 3 input-output qubits (green) and Nmem = 3 memory qubits (orange), receiving input from P presynaptic neurons. At each time t, incoming presynaptic spikes are weighted into input currents zt = [z 1 t , . . . , zN t ]. The currents determine rotation angles φt applied to the input qubits (blue gates). A parameterized quantum circuit (PQC, gray box) with trainable parameters θ entan… view at source ↗
Figure 2
Figure 2. Schematic of the QLIF neuron with P presynaptic inputs [11]. A single qubit (green) serves as the neuron’s internal state, evolving in discrete time steps via a measure-and-prepare scheme. At each time t, input spikes from P presynaptic neurons are weighted (blue arrows) and encoded as a rotation φt on the qubit. The qubit is measured multiple times to estimate a spike probability (red arrow), which is compared agai… view at source ↗
Figure 3
Figure 3. Illustration of the connection between a set of [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Illustration of an SQS-based neural network (SQSNN) [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Illustration of the proposed local zero-th order lear [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Illustration of the quantum circuit for a SQS neuron w [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]
Figure 7
Figure 7. Figure 7: (a) Test accuracy versus training iterations on a NIS [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 8
Figure 8. Figure 8: Test accuracy and average number of spike per time of S [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: Accuracy versus average number of spikes per time ste [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 11
Figure 11. Figure 11: Test decoding accuracy, sensing accuracy, and aver [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 38 canonical work pages

  1. [1]

    Towards efficient and reliable AI through neuromorphic principles,

    B. Rajendran, O. Simeone, and B. M. Al-Hashimi, “Towards efficient and reliable AI through neuromorphic principles, ” Philosophical Transactions A of the Royal Society , 2025

    work page 2025

  2. [2]

    Quantum neuromorphic com put- ing,

    D. Markovi´ c and J. Grollier, “Quantum neuromorphic com put- ing,” Applied physics letters , vol. 117, no. 15, 2020

    work page 2020

  3. [3]

    Simeone, Classical and Quantum Uncertainty, Information, and Correlation

    O. Simeone, Classical and Quantum Uncertainty, Information, and Correlation. Cambridge University Press, 2026

    work page 2026

  4. [4]

    Quantum advantage w ith shallow circuits,

    S. Bravyi, D. Gosset, and R. K¨ onig, “Quantum advantage w ith shallow circuits,” Science, vol. 362, no. 6412, pp. 308–311, 2018

    work page 2018

  5. [5]

    M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information . Cambridge university press, 2010

    work page 2010

  6. [6]

    Advancing neuromorphic computing with loihi: A survey of results and outlook,

    M. Davies, A. Wild, G. Orchard, Y . Sandamirskaya, G. A. F. Guerra, P . Joshi, P . Plank, and S. R. Risbud, “Advancing neuromorphic computing with loihi: A survey of results and outlook,” Proceedings of the IEEE , vol. 109, no. 5, pp. 911– 934, 2021

    work page 2021

  7. [7]

    Modern neuromorphic AI: From intra-token t o inter-token processing,

    O. Simeone, “Modern neuromorphic AI: From intra-token t o inter-token processing,” arXiv preprint arXiv:2601.00245, 2026

    arXiv 2026

  8. [8]

    Low-power neuromorphic hardware for sign al processing applications: A review of architectural and sys tem- level design approaches,

    B. Rajendran, A. Sebastian, M. Schmuker, N. Srinivasa, a nd E. Eleftheriou, “Low-power neuromorphic hardware for sign al processing applications: A review of architectural and sys tem- level design approaches,” IEEE Signal Processing Magazine , vol. 36, no. 6, pp. 97–110, 2019

    work page 2019

Show all 46 references
  1. [9]

    Memristors—from in-memory com- puting, deep learning acceleration, and spiking neural net works to the future of neuromorphic and bio-inspired computing,

    A. Mehonic, A. Sebastian, B. Rajendran, O. Simeone, E. V a si- laki, and A. J. Kenyon, “Memristors—from in-memory com- puting, deep learning acceleration, and spiking neural net works to the future of neuromorphic and bio-inspired computing,” Advanced Intelligent Systems , vol...

    2020

  2. [10]

    Bayesian con tinual learning via spiking neural networks,

    N. Skatchkovsky, H. Jang, and O. Simeone, “Bayesian con tinual learning via spiking neural networks,” Frontiers in Computa- tional Neuroscience, vol. 16, p. 1037976, 2022

    2022

  3. [11]

    A quantum leaky integrate -and- fire spiking neuron and network,

    D. Brand and F. Petruccione, “A quantum leaky integrate -and- fire spiking neuron and network,” npj Quantum Information , vol. 10, no. 1, pp. 1–8, 2024

    2024

  4. [12]

    Quantum reservoir computing implementatio n on coherently coupled quantum oscillators,

    J. Dudas, B. Carles, E. Plouet, F. A. Mizrahi, J. Grollie r, and D. Markovi´ c, “Quantum reservoir computing implementatio n on coherently coupled quantum oscillators,” npj Quantum In- formation, vol. 9, no. 1, p. 64, 2023

    2023

  5. [13]

    From quantu m feature maps to quantum reservoir computing: perspectives and applications,

    C. Gyurik, F. Wudarski, E. Philip, A. Sannia, H. Sadeghi , O. Kyriienko, D. V enturelli, and A. A. Gentile, “From quantu m feature maps to quantum reservoir computing: perspectives and applications,” arXiv preprint arXiv:2510.01797 , 2025

  6. [14]

    Harnessing disordered-ense mble quantum dynamics for machine learning,

    K. Fujii and K. Nakajima, “Harnessing disordered-ense mble quantum dynamics for machine learning,” Physical Review Applied, vol. 8, no. 2, p. 024030, 2017

    2017

  7. [15]

    Modular quantum extreme reservoir computing,

    H. W. Lau, A. Hayashi, A. Sakurai, W. J. Munro, and K. Nemoto, “Modular quantum extreme reservoir computing,” arXiv preprint arXiv:2412.19336 , 2024

    2024

  8. [16]

    Feedback connections in quantum reservoir computing with mid-circuit measurements,

    J. Murauer, R. Krishnakumar, S. Tornow, and M. Geierhos , “Feedback connections in quantum reservoir computing with mid-circuit measurements,” arXiv preprint arXiv:2503.22380 , 2025

    2025

  9. [17]

    Learning temporal data with a variational quantum recurre nt neural network,

    Y . Takaki, K. Mitarai, M. Negoro, K. Fujii, and M. Kitaga wa, “Learning temporal data with a variational quantum recurre nt neural network,” Physical Review A , vol. 103, no. 5, p. 052414, 2021

    2021

  10. [18]

    Time-warping invariant quantum recurrent neural network s via quantum-classical adaptive gating,

    I. Nikoloska, O. Simeone, L. Banchi, and P . V eliˇ ckovi´ c, “Time-warping invariant quantum recurrent neural network s via quantum-classical adaptive gating,” Machine Learning: Science and Technology, vol. 4, no. 4, p. 045038, 2023

    2023

  11. [19]

    An intr o- duction to probabilistic spiking neural networks: Probabi listic models, learning rules, and applications,

    H. Jang, O. Simeone, B. Gardner, and A. Gruning, “An intr o- duction to probabilistic spiking neural networks: Probabi listic models, learning rules, and applications,” IEEE Signal Process- ing Magazine , vol. 36, no. 6, pp. 64–77, 2019

    2019

  12. [20]

    Noise resilience of bayesian quantum phase estimation tested on a si quantum photonic chip,

    A. Gentile, S. Paesani, R. Santagati, J. Wang, N. Wiebe, D. Tew, J. O’Brien, and M. Thompson, “Noise resilience of bayesian quantum phase estimation tested on a si quantum photonic chip,” in European Quantum Electronics Conference , p. EB 6 2, Optica Publishing Group, 2017

    2017

  13. [21]

    Characterizing midcircuit measurements on a superconduc ting qubit using gate set tomography,

    K. Rudinger, G. J. Ribeill, L. C. Govia, M. Ware, E. Niels en, K. Y oung, T. A. Ohki, R. Blume-Kohout, and T. Proctor, 15 “Characterizing midcircuit measurements on a superconduc ting qubit using gate set tomography,” Physical Review Applied , vol. 17, no. 1, p. 014014, 2022

    2022

  14. [22]

    Pre - serving a qubit during state-destroying operations on an ad jacent qubit at a few micrometers distance,

    S. Motlakunta, N. Kotibhaskar, C.-Y . Shih, A. V ogliano , D. McLaren, L. Hahn, J. Zhu, R. Habl¨ utzel, and R. Islam, “Pre - serving a qubit during state-destroying operations on an ad jacent qubit at a few micrometers distance,” Nature Communications, vol. 15, no. 1, p. 6575, 2024

    2024

  15. [23]

    Non- invasive mid-circuit measurement and reset on atomic qubit s,

    Z.-Y . Chen, I. Goetting, G. Toh, Y . Y u, M. Shalaev, S. Sah a, A. Kalakuntla, H. B. Shi, C. Monroe, A. Kozhanov, et al., “Non- invasive mid-circuit measurement and reset on atomic qubit s,” arXiv preprint arXiv:2504.12538 , 2025

    2025

  16. [24]

    Dynamic parameterized quantum circuits: expre s- sive and barren-plateau free,

    A. Deshpande, M. Hinsche, S. Najafi, K. Sharma, R. Sweke, and C. Zoufal, “Dynamic parameterized quantum circuits: expre s- sive and barren-plateau free,” arXiv preprint arXiv:2411.05760 , 2024

    2024 arXiv

  17. [25]

    Practical advantage of quantum machine learning in ghost imaging,

    T. Xiao, X. Zhai, X. Wu, J. Fan, and G. Zeng, “Practical advantage of quantum machine learning in ghost imaging,” Communications Physics , vol. 6, no. 1, p. 171, 2023

    2023

  18. [26]

    Quantum machin e learning with differential privacy,

    W. M. Watkins, S. Y .-C. Chen, and S. Y oo, “Quantum machin e learning with differential privacy,” Scientific Reports , vol. 13, no. 1, p. 2453, 2023

    2023

  19. [27]

    Quantum vs. classical: A comprehensive benchmark study for predict ing time series with variational quantum machine learning,

    T. Fellner, D. Kreplin, S. Tovey, and C. Holm, “Quantum vs. classical: A comprehensive benchmark study for predict ing time series with variational quantum machine learning,” arXiv preprint arXiv:2504.12416, 2025

    2025

  20. [28]

    The power of quantum neural networks,

    A. Abbas, D. Sutter, C. Zoufal, A. Lucchi, A. Figalli, an d S. Woerner, “The power of quantum neural networks,” Nature Computational Science , vol. 1, no. 6, pp. 403–409, 2021

    2021

  21. [29]

    Knowing when to stop: Delay-adaptive spiking neural network classifiers with rel iability guarantees,

    J. Chen, S. Park, and O. Simeone, “Knowing when to stop: Delay-adaptive spiking neural network classifiers with rel iability guarantees,” IEEE Journal of Selected Topics in Signal Process- ing, vol. 19, no. 1, pp. 88–102, 2025

    2025

  22. [30]

    State-space model inspired multiple-input multiple-output spiking neurons ,

    S. Karilanova, S. Dey, and A. ¨Ozc ¸elikkale, “State-space model inspired multiple-input multiple-output spiking neurons ,” arXiv preprint arXiv:2504.02591, 2025

    2025 arXiv

  23. [31]

    Bottom-up and top- down approaches for the design of neuromorphic processing systems: Tradeoffs and synergies between natural and artifi cial intelligence,

    C. Frenkel, D. Bol, and G. Indiveri, “Bottom-up and top- down approaches for the design of neuromorphic processing systems: Tradeoffs and synergies between natural and artifi cial intelligence,” Proceedings of the IEEE , vol. 111, no. 6, pp. 623– 652, 2023

    2023

  24. [32]

    An introduction to quantum machine learning for engineers,

    O. Simeone et al. , “An introduction to quantum machine learning for engineers,” F oundations and Trends® in Signal Processing, vol. 16, no. 1-2, pp. 1–223, 2022

    2022

  25. [33]

    VOWEL: A loca l online learning rule for recurrent networks of probabilist ic spiking winner-take-all circuits,

    H. Jang, N. Skatchkovsky, and O. Simeone, “VOWEL: A loca l online learning rule for recurrent networks of probabilist ic spiking winner-take-all circuits,” in 2020 25th International Conference on Pattern Recognition (ICPR) , pp. 4597–4604, IEEE, 2021

    2020

  26. [34]

    Exponential c oncen- tration and symmetries in quantum reservoir computing,

    A. Sannia, G. L. Giorgi, and R. Zambrini, “Exponential c oncen- tration and symmetries in quantum reservoir computing,” arXiv preprint arXiv:2505.10062, 2025

    2025 arXiv

  27. [35]

    Multivariate stochastic approximation u sing a simultaneous perturbation gradient approximation,

    J. C. Spall, “Multivariate stochastic approximation u sing a simultaneous perturbation gradient approximation,” IEEE trans- actions on automatic control , vol. 37, no. 3, pp. 332–341, 1992

    1992

  28. [36]

    Schuld and F

    M. Schuld and F. Petruccione, Machine learning with quantum computers, vol. 676. Springer, 2021

    2021

  29. [37]

    An overview of the simultaneous perturbat ion method for efficient optimization,

    J. C. Spall, “An overview of the simultaneous perturbat ion method for efficient optimization,” Johns Hopkins apl technical digest, vol. 19, no. 4, pp. 482–492, 1998

    1998

  30. [38]

    Event-based vision: A survey,

    G. Gallego et al. , “Event-based vision: A survey,” IEEE trans- actions on pattern analysis and machine intelligence , vol. 44, no. 1, pp. 154–180, 2020

    2020

  31. [39]

    Training spiking neural networks using lessons from deep learning,

    J. K. Eshraghian, M. Ward, E. O. Neftci, X. Wang, G. Lenz, G. Dwivedi, M. Bennamoun, D. S. Jeong, and W. D. Lu, “Training spiking neural networks using lessons from deep learning,” Proceedings of the IEEE , vol. 111, no. 9, pp. 1016– 1054, 2023

    2023

  32. [40]

    Pok er-dvs and mnist-dvs. their history, how they were made, and other details,

    T. Serrano-Gotarredona and B. Linares-Barranco, “Pok er-dvs and mnist-dvs. their history, how they were made, and other details,” Frontiers in neuroscience , vol. 9, p. 481, 2015

    2015

  33. [41]

    Neuromorphi c wireless cognition: Event-driven semantic communication s for remote inference,

    J. Chen, N. Skatchkovsky, and O. Simeone, “Neuromorphi c wireless cognition: Event-driven semantic communication s for remote inference,” IEEE Transactions on Cognitive Communi- cations and Networking , vol. 9, no. 2, pp. 252–265, 2023

    2023

  34. [42]

    Liaf- net: Leaky integrate and analog fire network for lightweight and efficient spatiotemporal information processing,

    Z. Wu, H. Zhang, Y . Lin, G. Li, M. Wang, and Y . Tang, “Liaf- net: Leaky integrate and analog fire network for lightweight and efficient spatiotemporal information processing,” IEEE Transac- tions on Neural Networks and Learning Systems , vol. 33, no. 11, pp. 6249–6262, 2021

    2021

  35. [43]

    Circu it- centric quantum classifiers,

    M. Schuld, A. Bocharov, K. M. Svore, and N. Wiebe, “Circu it- centric quantum classifiers,” Physical Review A , vol. 101, no. 3, p. 032308, 2020

    2020

  36. [44]

    Quantum computing with qiskit,

    A. Javadi-Abhari, M. Treinish, K. Krsulich, C. J. Wood, J. Lish- man, J. Gacon, S. Martiel, P . D. Nation, L. S. Bishop, A. W. Cross, et al. , “Quantum computing with qiskit,” arXiv preprint arXiv:2405.08810, 2024

    2024 arXiv

  37. [45]

    Comparing experiments to the fault-tolerance threshold,

    R. Kueng, D. M. Long, A. C. Doherty, and S. T. Flammia, “Comparing experiments to the fault-tolerance threshold, ” Phys- ical review letters , vol. 117, no. 17, p. 170502, 2016

    2016

  38. [46]

    Neuromorphi c integrated sensing and communications,

    J. Chen, N. Skatchkovsky, and O. Simeone, “Neuromorphi c integrated sensing and communications,” IEEE Wireless Com- munications Letters , vol. 12, no. 3, pp. 476–480, 2023

    2023

Pith tools

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