Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

Standard quantum error-mitigation techniques do not systematically restore accuracy in hybrid quantum neural networks; their benefit is limited and depends strongly on the noise type and strength.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-02 15:54 UTC pith:ZLWQFF6C

load-bearing objection A useful negative-result benchmark with a real internal inconsistency: the summary table shows no mitigation gain anywhere, while the abstract claims selective improvements. the 4 major comments →

arxiv 2604.17515 v2 pith:ZLWQFF6C submitted 2026-04-19 quant-ph

Robustness Evaluation of Hybrid Quantum Neural Networks under Noise Models via System-Level Error Mitigation

Jesse Roberta Mingue Njiki , Nouhaila Innan , Alberto Marchisio , Muhammad Kashif , Jean-Michel Dricot , Muhammad Shafique This is my paper
classification quant-ph
keywords quantum machine learningquantum error mitigationhybrid quantum neural networksnoise robustnesszero-noise extrapolationprobabilistic error cancellationdigital dynamical decouplinglayerwise Richardson extrapolation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper tries to establish that, inside a full hybrid quantum-neural-network training loop, common error-mitigation techniques do not reliably recover the performance lost to realistic hardware noise. The authors simulate five quantum noise channels at eight strengths and apply four mitigation strategies over hundreds of training runs. Their central finding is that ZNE, DDD, and LRE track the unmitigated noisy baseline, while PEC helps only at low depolarizing noise. If correct, this means QNN practitioners should not expect a single default mitigation method to restore accuracy. The result matters because it challenges the default use of error mitigation in quantum machine learning and points toward noise-aware, context-specific strategies.

Core claim

The paper claims that the accuracy of a hybrid quantum neural network degrades in a way that depends jointly on the noise type and noise strength, and that none of the four evaluated error-mitigation methods—zero-noise extrapolation, probabilistic error cancellation, digital dynamical decoupling, and layerwise Richardson extrapolation—provides consistent recovery across conditions. In the experiments, depolarizing and amplitude-damping noise cause the strongest degradation, while phase-flip and phase-damping noise are comparatively benign. ZNE, DDD, and LRE generally follow the same degradation trend as the unmitigated baseline, with only small, selective deviations; PEC, evaluated only unde

What carries the argument

The experimental framework is the central mechanism: a hybrid quantum-classical training loop that injects static, memoryless quantum noise channels (depolarizing, amplitude damping, phase damping, bit flip, phase flip) before and after every gate, at probabilities ranging from 0.01 to 1.0, and then applies ZNE (noise amplification plus extrapolation), PEC (quasiprobabilistic inversion of the noise channel), DDD (insertion of identity-equivalent decoupling pulses), and LRE (layerwise Richardson extrapolation) inside the same training pipeline. This setup lets the authors attribute accuracy differences to the interaction of noise channel, noise strength, and mitigation method rather than to a

Load-bearing premise

The load-bearing premise is that static, memoryless quantum noise channels inserted before and after every gate faithfully represent real hardware noise; this premise is weakest for DDD, which is designed to suppress time-correlated low-frequency noise that these channels do not contain.

What would settle it

Run the same hybrid QNN training pipeline on real quantum hardware that exhibits temporally correlated dephasing (for example, 1/f noise), and compare validation accuracy with and without DDD. If DDD systematically and substantially improves accuracy over the unmitigated baseline in that setting, the paper's conclusion that mitigation benefits are generally limited and noise-dependent would be contradicted for the DDD case.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • QNN robustness cannot be summarized by a single noise rate: training and validation accuracy depend on both the type of noise and its strength.
  • Depolarizing and amplitude-damping noise are the most damaging regimes for this architecture, whereas phase-flip and phase-damping noise cause comparatively little degradation.
  • Deploying ZNE, DDD, or LRE as a default mitigation in QNN training will not reliably improve accuracy; these methods tend to follow the unmitigated baseline.
  • PEC, despite being theoretically exact for Pauli-type noise, pays off only in the low-noise depolarizing regime and at substantial sampling cost.
  • Mitigation effectiveness must be matched to the dominant physical error process; there is no one-size-fits-all mitigation method in this setting.

Where Pith is reading between the lines

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

  • Editorial inference: because the simulated noise channels are memoryless and static, DDD—whose theoretical basis is suppression of time-correlated low-frequency noise—may be significantly more effective on real hardware; the paper's negative DDD result could be an artifact of the noise model, not the technique.
  • Editorial inference: the findings motivate a practical selection rule—characterize the hardware's dominant noise channel before choosing a mitigation method—and suggest that a noise-aware switching policy could outperform any fixed method.
  • Editorial inference: the architecture tested is shallow (three qubits, four layers), so deeper circuits with more entangling gates may exhibit a different ordering of mitigation effectiveness; the conclusions should be stress-tested on deeper QNNs.
  • Editorial inference: the single small dataset and small model limit generalization; extending to higher-dimensional data would clarify whether the noise-mitigation interactions observed here persist in more complex learning tasks.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The manuscript benchmarks hybrid quantum neural networks (HQNNs) under five simulated noise channels (depolarizing, amplitude damping, phase damping, bit flip, phase flip) and four quantum error mitigation (QEM) techniques (ZNE, PEC, DDD, LRE). Using PennyLane, Qiskit Aer, and Mitiq, the authors train a 3-qubit, 4-layer QNN on the Iris dataset across eight noise probabilities, reporting noise-type- and strength-dependent accuracy degradation. The central claim is that mitigation benefits are limited and noise-dependent: ZNE, DDD, and LRE largely track the unmitigated baseline, while PEC shows limited gains only at low depolarizing noise. The paper concludes that no single mitigation method consistently restores QNN performance under the tested conditions.

Significance. If the negative result holds, the paper would provide a useful cautionary benchmark for the QML community, showing that standard QEM techniques do not automatically improve QNN training accuracy under memoryless simulated noise. The systematic sweep over five noise models and four mitigation methods in a unified training pipeline is a constructive contribution, especially in contrast to prior QEM studies focused on VQE/QAOA. The use of established open-source tools (PennyLane, Qiskit Aer, Mitiq) supports reproducibility, though no code release is mentioned. The paper's value depends heavily on the consistency of its reported data; currently, the headline claims are not supported by the paper's own summary table.

major comments (4)
  1. [Table III / Sec. IV-C] Table III, the only quantitative summary of mitigation efficacy, shows that every mitigation method achieves strictly lower validation accuracy than 'No Mitigation' in all 15 rows. For example, Depolarizing Low: NM 1.0000 vs DDD 0.8158, LRE 0.8420, ZNE 0.8158, PEC 0.8421; Bit Flip Medium: NM 0.9474 vs ZNE 0.7368. This directly contradicts the abstract and Sec. IV-B prose claiming 'PEC shows limited gains only in the low-noise depolarizing regime' and ZNE 'small selective improvements.' The broad negative trend may be correct, but the 'limited gains' qualifier is unsupported by the paper's own aggregated data. The authors must either reconcile the table with the prose by presenting per-noise-level results that show those gains, or revise the claims to state that all tested methods were consistently worse than baseline in the aggregated data.
  2. [Sec. III-D / Sec. IV] Only 3 repetitions per configuration are used, and no variance, confidence intervals, or statistical tests are reported. Accuracy differences across methods are often small (e.g., 0.9737 vs 0.9211) and could easily arise from stochastic training and finite shots. The manuscript uses phrases like 'small selective improvements' and 'limited benefit' without quantifying uncertainty. To support the central claim, the authors should report the per-repetition results, standard deviations, and ideally a paired significance test (e.g., Wilcoxon or t-test across the three runs) or a per-configuration breakdown.
  3. [Sec. III-B / Sec. II-C.3] The noise injection model applies static, memoryless quantum channels before and after each gate. However, DDD's theoretical basis (Sec. II-C.3) is suppression of time-correlated low-frequency noise, which is absent from these memoryless channels. Thus the observed lack of DDD benefit is expected under this noise model and does not generalize to temporally correlated dephasing present on real hardware. The authors should either include a noise model with temporal correlations (e.g., time-dependent dephasing or noise with 1/f spectrum) or explicitly limit the DDD conclusion to memoryless noise, which would substantially narrow the scope of the claim.
  4. [Sec. IV-B.2 / Fig. 6] The claim that PEC 'shows limited gains' in the low-depolarizing-noise regime is not visible in Table III: at Depolarizing Low, PEC (0.8421) is below the unmitigated baseline (1.0000), and at Medium and High it is 0.5000 and 0.5263, respectively, versus baselines 0.9737 and 0.6053. Figure 6's caption labels PEC 'Effective under low noise,' which is inconsistent with the table. Either the figure shows a different metric or the table aggregation is misleading. This must be clarified before the PEC claim can be accepted.
minor comments (5)
  1. [Sec. III-D / Fig. 3] The stated total of 128 configurations appears inconsistent with the framework: 5 noise models × 8 levels gives 40 noisy baselines; ZNE, DDD, LRE over all 5 noise models add 3×5×8 = 120; PEC over depolarizing adds 8. That sums to 168, not 128 (and 384 runs would be 168×3 = 504 if truly triple repeated). Please verify the arithmetic and clarify what is counted as a configuration.
  2. [Table III] The dash symbols are inconsistent: '—' for PEC missing entries and '–' for other missing cells. Also consider reporting standard deviations in the table or a supplemental table.
  3. [Figures 5-8] The captions and axis text in Figures 5-8 are extremely small and difficult to read, especially the inline labels like 'Small improvement at low noise, bad as noise increase.' The captions should be enlarged and the legends clearly state which curve is the baseline versus each mitigation method.
  4. [References] Reference [48] lists 'U. Fund' as the author; this should be 'Unitary Fund.' Reference [43] appears to have garbled author names: 'G.-T. Tudor, H. Yousef, L. Ryan, M. Andrea, and Z. W. J.' likely corresponds to Giurgica-Tiron et al. Please correct.
  5. [Sec. IV-B] The prose descriptions for each mitigation method use hedging language ('small selective improvements', 'limited benefit') that, given the aggregate data in Table III, is too strong. Please align the qualitative descriptions with the actual numerical results shown in the table and figures.

Circularity Check

0 steps flagged

No significant circularity: the central claims are empirical measurements, not derivations from fitted inputs or self-cited constraints.

full rationale

This paper is an empirical benchmarking study: it injects five Qiskit Aer noise channels at specified strengths p, trains a PennyLane hybrid QNN with and without Mitiq-based ZNE/DDD/LRE (and PEC separately under depolarizing noise), and reports measured validation accuracy. There is no derivation chain in which an output is equivalent by construction to an input: no parameter is fitted to a subset of the results and then used to 'predict' another subset; the noise channels are standard CPTP maps; the mitigation techniques are library implementations rather than redefinitions of the measured quantity; and no uniqueness theorem or ansatz is imported from the authors' prior work to force the conclusions. The dense self-citation network in Sec. II provides background and motivation but is not load-bearing: the central 'limited and noise-dependent' claim rests on Table III and Figs. 4-8, which are independent observations under explicitly stated assumptions. One internal consistency issue exists (Sec. IV-B/abstract prose claims PEC shows 'limited gains only in the low-noise depolarizing regime' and ZNE shows 'small selective improvements,' whereas Table III shows every mitigation method below No Mitigation in every aggregated interval, e.g., Depolarizing Low: NM 1.0 vs PEC 0.8421), but this is a data-reporting or correctness concern, not circularity. The weakest assumption, that memoryless per-gate channels are a faithful proxy for hardware (especially for DDD), is a validity limitation rather than a circular step. Overall, the paper is self-contained against its own benchmark data, so the circularity score is 0.

Axiom & Free-Parameter Ledger

1 free parameters · 4 axioms · 0 invented entities

No fitted parameters or new physical entities are introduced. The paper is an empirical benchmark, so the ledger records modeling assumptions and hand-chosen hyperparameters rather than derivation inputs.

free parameters (1)
  • Experimental hyperparameters (learning rate, epochs, batch size, shots, repetitions) = lr=0.3 (halved every 5 epochs), 20 epochs, batch=5, 8192 shots, 3 repetitions
    Hand-chosen constants; not fitted to data, but the central benchmark outcomes could depend on them.
axioms (4)
  • standard math Kraus/CPTP channels correctly describe the five noise models in Table I
    Section II-B relies on the standard Kraus-operator formalism for depolarizing, amplitude/phase damping, bit flip, and phase flip noise.
  • domain assumption Mitiq's ZNE, PEC, DDD, and LRE implementations correctly realize the published algorithms
    Section III-C integrates Mitiq without independent verification or code; incorrect integration would invalidate the negative results.
  • domain assumption Static memoryless channels applied before and after each gate are sufficient for testing DDD's decoherence suppression
    Section III-B uses memoryless noise, while DDD targets time-correlated noise (Sec. II-C); this mismatch may bias DDD negatively and is not acknowledged.
  • domain assumption The Iris dataset and the chosen 3-qubit architecture are representative enough to support general conclusions about QNN robustness
    Single dataset and one small architecture are used; generalization to other QNNs is assumed in the conclusions.

pith-pipeline@v1.3.0-alltime-deepseek · 36749 in / 14559 out tokens · 131500 ms · 2026-08-02T15:54:38.823532+00:00 · methodology

0 comments
read the original abstract

Quantum Neural Networks (QNNs) represent a promising direction within Quantum Machine Learning (QML), yet their realization on noisy intermediate-scale quantum (NISQ) devices remains constrained by decoherence, gate imperfections, crosstalk, and readout errors. This study provides a systematic evaluation of noise effects and mitigation strategies in hybrid quantum neural networks (HQNNs). Zero-Noise Extrapolation (ZNE), Digital Dynamical Decoupling (DDD), and Layerwise Richardson Extrapolation (LRE) are integrated into end-to-end QNN training pipelines developed with PennyLane, simulated under Qiskit Aer noise models, and integrated with the Mitiq framework, while Probabilistic Error Cancellation (PEC) is evaluated separately under depolarizing noise due to its computational cost. Experiments conducted on the Iris dataset with five representative noise channels show that the impact of noise and the effect of mitigation are strongly dependent on the noise model and its strength. The model maintains comparatively strong performance under phase-flip and phase-damping noise, while substantial degradation is observed under high depolarizing and amplitude-damping noise. Across the evaluated mitigation methods, the observed benefits remain limited and noise-dependent: ZNE, DDD, and LRE generally follow the same degradation trends as the unmitigated baseline, while PEC shows limited gains only in the low-noise depolarizing regime. These findings highlight the need for context-specific mitigation strategies to improve the robustness of QNNs in practical NISQ settings.

Figures

Figures reproduced from arXiv: 2604.17515 by Alberto Marchisio, Jean-Michel Dricot, Jesse Roberta Mingue Njiki, Muhammad Kashif, Muhammad Shafique, Nouhaila Innan.

Figure 1
Figure 1. Figure 1: Motivational analysis illustrating the behavior of the [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: General architecture of QNNs, illustrating data encoding, [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the proposed methodology for benchmarking QNN robustness under noise. The framework considers five [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Baseline QNN validation accuracy without mitigation $ !        $ !        [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Validation accuracy comparison for PEC and baseline                                          ! [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗

Review history (2 revisions) →

discussion (0)

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

Forward citations

Cited by 2 Pith papers

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

  1. Towards Fair Benchmarking of Quantum Transfer Learning for Visual Classification

    quant-ph 2026-05 unverdicted novelty 6.0

    Introduces a unified benchmarking methodology for quantum transfer learning in visual classification tasks, finding that no single method dominates and performance varies with dataset, encoding, and circuit design.

  2. Meta-Quantum Ensemble Framework for Robust Network Intrusion Detection

    quant-ph 2026-05 unverdicted novelty 3.0

    A meta-ensemble fusing QSVM and QNN outputs with a Random Forest improves selected IDS metrics over standalone quantum models on TON IoT and CICIDS2017 datasets.

Reference graph

Works this paper leans on

48 extracted references · 3 linked inside Pith · cited by 2 Pith papers

  1. [1]

    A survey on quantum machine learning: Current trends, challenges, opportunities, and the road ahead,

    K. Zaman, A. Marchisio, M. A. Hanif, and M. Shafique, “A survey on quantum machine learning: Current trends, challenges, opportunities, and the road ahead,”arXiv preprint arXiv:2310.10315, 2023

    Pith/arXiv arXiv 2023

  2. [2]

    Quantum machine learning: a classical perspective,

    C. Ciliberto, M. Herbster, A. D. Ialongo, M. Pontil, A. Rocchetto, S. Severini, and L. Wossnig, “Quantum machine learning: a classical perspective,”Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, vol. 474, no. 2209, p. 20170551, 2018

    2018

  3. [3]

    A review of quantum neural networks: Methods, models, dilemmas,

    R. Zhao and S. Wang, “A review of quantum neural networks: Methods, models, dilemmas,”Engineering Reports, 2021

    2021

  4. [4]

    Demonstrating quantum advantage in hybrid quantum neural networks for model capacity,

    M. Kashif and S. Al-Kuwari, “Demonstrating quantum advantage in hybrid quantum neural networks for model capacity,” in2022 IEEE International Conference on Rebooting Computing (ICRC), 2022, pp. 36–44

    2022

  5. [5]

    A primer on quantum machine learning,

    S. Y . Chang and M. Cerezo, “A primer on quantum machine learning,” arXiv preprint arXiv:2511.15969, 2025

    arXiv 2025

  6. [6]

    Computational advantage in hybrid quantum neural networks: Myth or reality?

    M. Kashif, A. Marchisio, and M. Shafique, “Computational advantage in hybrid quantum neural networks: Myth or reality?” in2025 62nd ACM/IEEE Design Automation Conference (DAC). IEEE, 2025, pp. 1–7

    2025

  7. [7]

    Next- generation quantum neural networks: Enhancing efficiency, security, and privacy,

    N. Innan, M. Kashif, A. Marchisio, M. Bennai, and M. Shafique, “Next- generation quantum neural networks: Enhancing efficiency, security, and privacy,” in2025 IEEE 31st International Symposium on On-Line Testing and Robust System Design (IOLTS). IEEE, 2025, pp. 1–4

    2025

  8. [8]

    A survey and tutorial on security and resilience of quantum computing,

    A. A. Saki, M. Alam, K. Phalak, A. Suresh, R. O. Topaloglu, and S. Ghosh, “A survey and tutorial on security and resilience of quantum computing,” in2021 IEEE European Test Symposium (ETS). IEEE, 2021, pp. 1–10

    2021

  9. [9]

    Benchmarking quantum computers and the impact of quantum noise,

    S. Resch and U. R. Karpuzcu, “Benchmarking quantum computers and the impact of quantum noise,”ACM Computing Surveys, 2021

    2021

  10. [10]

    A comparative analysis and noise robustness evaluation in quantum neural networks,

    T. Ahmed, M. Kashif, A. Marchisio, and M. Shafique, “A comparative analysis and noise robustness evaluation in quantum neural networks,” Scientific Reports, vol. 15, no. 1, p. 33654, 2025

    2025

  11. [11]

    Investigating the effect of noise on the training performance of hybrid quantum neural networks,

    M. Kashif, E. Sychiuco, and M. Shafique, “Investigating the effect of noise on the training performance of hybrid quantum neural networks,” in2024 International Joint Conference on Neural Networks (IJCNN). IEEE, 2024, pp. 1–10

    2024

  12. [12]

    Noisy hqnns: A comprehensive analysis of noise robustness in hybrid quantum neural networks,

    T. Ahmed, A. Marchisio, M. Kashif, and M. Shafique, “Noisy hqnns: A comprehensive analysis of noise robustness in hybrid quantum neural networks,” in2025 International Joint Conference on Neural Networks (IJCNN). IEEE, 2025, pp. 1–10

    2025

  13. [13]

    Quantum error mitigation,

    Z. Caiet al., “Quantum error mitigation,”Reviews of Modern Physics, vol. 95, no. 4, p. 045005, 2023

    2023

  14. [14]

    Benchmarking noisy intermediate scale quantum error mitigation strate- gies for ground state preparation of the hcl molecule,

    T. Weaving, A. Ralli, W. M. Kirby, P. J. Love, S. Succi, and P. V . Coveney, “Benchmarking noisy intermediate scale quantum error mitigation strate- gies for ground state preparation of the hcl molecule,”Physical Review Research, vol. 5, no. 4, p. 043054, 2023

    2023

  15. [15]

    Design space exploration of hybrid quantum–classical neural networks,

    M. Kashif and S. Al-Kuwari, “Design space exploration of hybrid quantum–classical neural networks,”Electronics, vol. 10, no. 23, p. 2980, 2021

    2021

  16. [16]

    Comparative performance analysis of quantum machine learning architectures for credit card fraud detection,

    M. El Alami, N. Innan, M. Shafique, and M. Bennai, “Comparative performance analysis of quantum machine learning architectures for credit card fraud detection,”Applied Intelligence, vol. 56, no. 3, p. 83, 2026

    2026

  17. [17]

    The power of quantum neural networks,

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

    2021

  18. [18]

    Quantum machine learning,

    J. Biamonte, P. Wittek, N. Pancotti, P. Rebentrost, N. Wiebe, and S. Lloyd, “Quantum machine learning,”Nature, vol. 549, no. 7671, pp. 195–202, 2017

    2017

  19. [19]

    Resource allocation optimization in 5g networks using variational quantum regressor,

    P. Pathak, V . Oad, A. Prajapati, and N. Innan, “Resource allocation optimization in 5g networks using variational quantum regressor,” in 2024 International Conference on Quantum Communications, Networking, and Computing (QCNC). IEEE, 2024, pp. 101–105

    2024

  20. [20]

    Qnn-vrcs: A quantum neural network for vehicle road cooperation systems,

    N. Innan, B. K. Behera, S. Al-Kuwari, and A. Farouk, “Qnn-vrcs: A quantum neural network for vehicle road cooperation systems,”IEEE Transactions on Intelligent Transportation Systems, 2025

    2025

  21. [21]

    Sentiqnf: A novel approach to sentiment analysis using quantum algorithms and neuro-fuzzy systems,

    K. Dave, N. Innan, B. K. Behera, Z. Mumtaz, S. Al-Kuwari, and A. Farouk, “Sentiqnf: A novel approach to sentiment analysis using quantum algorithms and neuro-fuzzy systems,”IEEE Transactions on Computational Social Systems, 2025

    2025

  22. [22]

    Quantum bayesian networks for machine learning in oil-spill detection,

    O. I. Siddiqui, N. Innan, A. Marchisio, M. Bennai, and M. Shafique, “Quantum bayesian networks for machine learning in oil-spill detection,” in2025 International Joint Conference on Neural Networks (IJCNN), 2025, pp. 1–8

    2025

  23. [23]

    Fedqnn: Federated learning using quantum neural networks,

    N. Innan, M. A.-Z. Khan, A. Marchisio, M. Shafique, and M. Bennai, “Fedqnn: Federated learning using quantum neural networks,” in2024 International Joint Conference on Neural Networks (IJCNN), 2024, pp. 1–9

    2024

  24. [24]

    Financial fraud detection using quantum graph neural networks,

    N. Innan, A. Sawaika, A. Dhor, S. Dutta, S. Thota, H. Gokal, N. Patel, M. A.-Z. Khan, I. Theodonis, and M. Bennai, “Financial fraud detection using quantum graph neural networks,”Quantum Machine Intelligence, vol. 6, no. 1, p. 7, 2024

    2024

  25. [25]

    Quantum state tomography using quantum machine learning,

    N. Innan, O. I. Siddiqui, S. Arora, T. Ghosh, Y . P. Ko c ¸ak, D. Paragas, A. A. O. Galib, M. A.-Z. Khan, and M. Bennai, “Quantum state tomography using quantum machine learning,”Quantum Machine Intelligence, vol. 6, no. 1, p. 28, 2024

    2024

  26. [26]

    HQNN-FSP: A hybrid classical-quantum neural network for regression-based financial stock market prediction,

    P. K. Choudhary, N. Innan, M. Shafique, and R. Singh, “HQNN-FSP: A hybrid classical-quantum neural network for regression-based financial stock market prediction,”arXiv preprint arXiv:2503.15403, 2025

    Pith/arXiv arXiv 2025

  27. [27]

    Optimizing low-energy carbon iiot systems with quantum algorithms: Performance evaluation and noise robustness,

    K. Dave, N. Innan, B. K. Behera, S. Mumtaz, S. Al-Kuwari, and A. Farouk, “Optimizing low-energy carbon iiot systems with quantum algorithms: Performance evaluation and noise robustness,”IEEE Internet of Things Journal, vol. 12, no. 17, pp. 34 653–34 662, 2025

    2025

  28. [28]

    Quiet- sr: Quantum image enhancement transformer for single image super- resolution,

    S. Dutta, N. Innan, K. Najafi, S. B. Yahia, and M. Shafique, “Quiet- sr: Quantum image enhancement transformer for single image super- resolution,”arXiv preprint arXiv:2503.08759, 2025

    arXiv 2025

  29. [29]

    Qfnn-ffd: Quantum federated neural network for financial fraud detection,

    N. Innan, A. Marchisio, M. Bennai, and M. Shafique, “Qfnn-ffd: Quantum federated neural network for financial fraud detection,” in2025 IEEE International Conference on Quantum Software (QSW). IEEE, 2025, pp. 41–47

    2025

  30. [30]

    Lep-qnn: Loan eligibility prediction using quantum neural networks,

    N. Innan, A. Marchisio, M. Bennai, and M. Shafique, “Lep-qnn: Loan eligibility prediction using quantum neural networks,” in2025 IEEE International Conference on Quantum Computing and Engineering (QCE), vol. 1. IEEE, 2025, pp. 1864–1872

    2025

  31. [31]

    Quantum vs. classical machine learning: A benchmark study for financial prediction,

    R. Ahmad, M. Kashif, N. Innan, and M. Shafique, “Quantum vs. classical machine learning: A benchmark study for financial prediction,”arXiv preprint arXiv:2601.03802, 2026

    arXiv 2026

  32. [32]

    Recent developments and applications in quantum neural network: A review,

    S. K. Jeswal and S. Chakraverty, “Recent developments and applications in quantum neural network: A review,”Archives of Computational Methods in Engineering, 2018

    2018

  33. [33]

    The disparate impact of noise on quantum learning algorithms,

    A. Angrisani, “The disparate impact of noise on quantum learning algorithms,” Ph.D. dissertation, Sorbonne Universit ´e, 2023, doctoral dis- sertation. [Online]. Available: https://theses.hal.science/tel-04511706v1

    2023

  34. [34]

    Resqnets: a residual approach for mitigating barren plateaus in quantum neural networks,

    M. Kashif and S. Al-Kuwari, “Resqnets: a residual approach for mitigating barren plateaus in quantum neural networks,”EPJ Quantum Technology, vol. 11, no. 1, pp. 1–28, 2024

    2024

  35. [35]

    The impact of cost function globality and locality in hybrid quantum neural networks on nisq devices,

    M. Kashif and S. Al-Kuwari, “The impact of cost function globality and locality in hybrid quantum neural networks on nisq devices,”Machine Learning: Science and Technology, vol. 4, no. 1, p. 015004, 2023

    2023

  36. [36]

    Deep quanvolutional neural networks with enhanced trainability and gradient propagation,

    M. Kashif and M. Shafique, “Deep quanvolutional neural networks with enhanced trainability and gradient propagation,”Scientific Reports, vol. 15, no. 1, p. 21764, 2025

    2025

  37. [37]

    Investigating different barren plateaus mitigation strategies in variational quantum eigensolver,

    M. Atallah, N. Innan, M. Kashif, and M. Shafique, “Investigating different barren plateaus mitigation strategies in variational quantum eigensolver,” arXiv preprint arXiv:2512.11171, 2025

    arXiv 2025

  38. [38]

    Hqnet: Harnessing quantum noise for effective training of quantum neural networks in nisq era,

    M. Kashif and M. Shafique, “Hqnet: Harnessing quantum noise for effective training of quantum neural networks in nisq era,” in2025 IEEE International Conference on Quantum Artificial Intelligence (QAI). IEEE, 2025, pp. 387–394

    2025

  39. [39]

    Nrqnn: The role of observable selection in noise-resilient quantum neural networks,

    M. Kashif and M. Shafique, “Nrqnn: The role of observable selection in noise-resilient quantum neural networks,” inWorld Congress in Computer Science, Computer Engineering & Applied Computing. Springer, 2024, pp. 116–131

    2024

  40. [40]

    Quantum error mitigation by layerwise richardson extrapolation,

    R. Vincent and M. Andrea, “Quantum error mitigation by layerwise richardson extrapolation,”Physical Review A, vol. 110, no. 6, p. 062420, 2024

    2024

  41. [41]

    Optimization of richardson extrapolation for quantum error mitigation,

    K. Michael, T. Bj ¨orn, and C. Alessio, “Optimization of richardson extrapolation for quantum error mitigation,”Physical Review A, vol. 106, p. 062436, 2022

    2022

  42. [42]

    Quantum computing and its implications for cybersecurity: A comprehensive review of emerging threats and defenses,

    S. Khan, K. Palani, M. Goswami, and M. S. Arafath, “Quantum computing and its implications for cybersecurity: A comprehensive review of emerging threats and defenses,”Nanotechnology Perceptions, 2024

    2024

  43. [43]

    Digital zero noise extrapolation for quantum error mitigation,

    G.-T. Tudor, H. Yousef, L. Ryan, M. Andrea, and Z. W. J., “Digital zero noise extrapolation for quantum error mitigation,”arXiv preprint arXiv:2005.10921, 2021

    Pith/arXiv arXiv 2005

  44. [44]

    Probabilistic error cancellation with sparse pauli–lindblad models on noisy quantum processors,

    E. Van Den Berg, Z. K. Minev, A. Kandala, and K. Temme, “Probabilistic error cancellation with sparse pauli–lindblad models on noisy quantum processors,”Nature physics, vol. 19, no. 8, pp. 1116–1121, 2023

    2023

  45. [45]

    Dynamical decoupling for superconducting qubits: A performance survey,

    N. Ezzell, B. Pokharel, L. Tewala, G. Quiroz, and D. A. Lidar, “Dynamical decoupling for superconducting qubits: A performance survey,”Physical Review Applied, vol. 20, no. 6, p. 064027, 2023

    2023

  46. [46]

    Synergistic dynamical decoupling and circuit design for enhanced algorithm performance on near-term quantum devices,

    Y . Ji and I. Polian, “Synergistic dynamical decoupling and circuit design for enhanced algorithm performance on near-term quantum devices,” Entropy, vol. 26, no. 7, p. 586, 2024

    2024

  47. [47]

    UCI machine learning repository,

    D. Dua and C. Graff, “UCI machine learning repository,” 2019. [Online]. Available: http://archive.ics.uci.edu/ml

    2019

  48. [48]

    Mitiq documentation: A python toolkit for quantum error mitigation,

    U. Fund and Contributors, “Mitiq documentation: A python toolkit for quantum error mitigation,” https://mitiq.readthedocs.io/en/stable/, 2025, accessed: 2025-07-03

    2025