arxiv: 2604.26110 · v1 · submitted 2026-04-28 · 🪐 quant-ph

Recognition: unknown

A Comprehensive Analysis of Accuracy and Robustness in Quantum Neural Networks

Ban Q. Tran , Duong M. Chu , Hai T.D. Pham , Viet Q. Nguyen , Quan A. Pham , Susan Mengel

Authors on Pith no claims yet

Pith reviewed 2026-05-07 16:23 UTC · model grok-4.3

classification 🪐 quant-ph

keywords quantum neural networksQCNNQRNNQViTrobustnessaccuracyNISQquantum noise

0 comments

The pith

Quantum neural networks perform well on low-feature datasets like MNIST but degrade on high-feature data, with vision transformers showing superior robustness to quantum noise.

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

The paper compares three hybrid quantum-classical architectures—QCNN, QRNN, and QViT—built around variational quantum circuits, evaluating their generalization, accuracy, and robustness to noise. It shows that all three achieve strong results on simple low-dimensional data but lose effectiveness as feature count rises, with convolutional models faring particularly poorly on complex inputs. Adversarial noise affects every architecture, yet recurrent and convolutional variants retain more resilience, while the transformer architecture alone holds up well against measurement noise, channel noise, and finite-shot sampling. These patterns highlight the need to match model type to both data characteristics and the specific noise environment of current quantum hardware.

Core claim

While these models exhibit exceptional performance on low-feature datasets such as MNIST, their learning efficacy degrades significantly when transitioned to high-feature datasets. Additionally, while all models are susceptible to adversarial noise, traditional architectures demonstrate superior resilience. In the presence of quantum noise, the transformer-based architecture maintains high robustness against measurement noise, channel noise, and finite-shot effects.

What carries the argument

Comparative testing of QCNN, QRNN, and QViT hybrid architectures on accuracy, generalization, and resilience to adversarial versus quantum noise sources.

If this is right

Architecture choice for quantum neural networks must weigh data dimensionality to avoid sharp accuracy losses.
Recurrent or convolutional QNNs are preferable when adversarial robustness is the primary concern.
Transformer-based QNNs are the stronger option inside NISQ devices dominated by quantum channel and measurement noise.
Model selection in quantum machine learning should be tailored to the dominant noise type rather than treated as interchangeable.
Current QNN designs remain limited by dataset complexity and therefore require further architecture-specific refinements.

Where Pith is reading between the lines

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

These results suggest that hybrid QNN benchmarks should routinely include both low- and high-dimensional test sets to prevent inflated performance estimates.
Future designs could combine the adversarial strength of recurrent layers with the quantum-noise tolerance of transformers.
The observed patterns imply that scaling quantum machine learning to realistic high-dimensional tasks will need explicit noise-type matching rather than generic architecture reuse.

Load-bearing premise

The selected datasets, noise models, and implementation details for the three architectures provide a fair and representative comparison without unstated biases in circuit depth, optimization, or hyperparameter choices.

What would settle it

Repeating the experiments on new high-feature datasets while enforcing identical circuit depths, optimizer settings, and hyperparameter budgets across all three models and checking whether the reported accuracy drop and differential noise robustness still appear.

Figures

Figures reproduced from arXiv: 2604.26110 by Ban Q. Tran, Duong M. Chu, Hai T.D. Pham, Quan A. Pham, Susan Mengel, Viet Q. Nguyen.

**Figure 1.** Figure 1: The overall architecture of the hybrid classical-quantum neural networks utilized in this study. The view at source ↗

**Figure 2.** Figure 2: Compare the ASR of QCNN with the four main adversarial attack methods: FGSM, PGD, APGD, and view at source ↗

**Figure 3.** Figure 3: An overview of the analytical methodology and evaluation framework employed in this research. view at source ↗

**Figure 4.** Figure 4: 2D and 3D loss landscapes of the QViT optimizer for Cat and Dog classification with 10,000 samples, view at source ↗

**Figure 5.** Figure 5: The generalization bound was measured on the Cat vs. Dog pair of the CIFAR-10 dataset for the view at source ↗

**Figure 6.** Figure 6: Experimental verification of the claims regarding the alignment between theoretical generalization view at source ↗

**Figure 7.** Figure 7: The measurement results of the variation in Lipschitz bound across training epochs on the MNIST and view at source ↗

**Figure 8.** Figure 8: Comparison of Average Fidelity across QNN models. For quantum metrics, the average fidelity is used to measure the similarity between the quantum state of the image before and after the attack. It can be observed that after the attack, the quantum state changes differently across the models. In image a), QCNN demonstrates relatively good resilience against adversarial attacks, as the average fidelity value… view at source ↗

**Figure 9.** Figure 9: Impact of quantum noise on the performance of the QRNN model. view at source ↗

read the original abstract

Quantum Machine Learning (QML) has recently emerged as a highly promising research frontier. Within this domain, Quantum Neural Networks (QNNs),characterized by Variational Quantum Circuits (VQCs) at their core and featuring layers of quantum gates optimized by classical algorithms, have garnered significant attention. However, a rigorous and exhaustive evaluation of their practical performance remains largely incomplete. In this study, we conduct a comprehensive comparative analysis of three prominent hybrid classical-quantum architectures: Quantum Convolutional Neural Networks (QCNN), Quantum Recurrent Neural Networks (QRNN), and Quantum Vision Transformers (QViT), focusing on the critical dimensions of generalization, accuracy, and robustness. Our findings provide novel insights that address previous evaluative gaps. Notably, while these models exhibit exceptional performance on low-feature datasets such as MNIST, their learning efficacy degrades significantly when transitioned to high-feature datasets. Furthermore, convolutional-based models like QCNN appear less effective on high-dimensional data than other machine learning architectures. Additionally, while all models are susceptible to adversarial noise, traditional architectures, such as recurrent and convolutional networks, demonstrate superior resilience. Conversely, in the presence of quantum noise, the transformer-based architecture proves its strength by maintaining high robustness against measurement noise, channel noise, and finite-shot effects, whereas other architectures suffer marked performance declines. These results provide a granular perspective on the current state of the field and underscore the critical importance of tailoring model selection to the constraints of contemporary Noisy Intermediate-Scale Quantum (NISQ) environments.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Comparative simulations of QCNN, QRNN, and QViT show data-feature sensitivity and noise-type differences, but require checks on whether the models were given equivalent resources.

read the letter

The punchline is that this is a simulation-based comparison of three quantum neural network architectures—QCNN, QRNN, and QViT—showing strong results on low-feature data like MNIST but clear accuracy loss on higher-feature inputs, with QViT holding up better under quantum noise while the others do better against adversarial noise. The paper does a decent job of pulling together tests on generalization to high-dimensional data and on two different kinds of noise. That combination is not entirely new, but it gives a practical snapshot that could help someone decide which architecture to try first on NISQ hardware. The abstract lays out the findings without much hype, and the work stays within what simulations can show. The main concern is whether the architectures were compared on equal footing. The stress-test note correctly identifies the risk that differences in circuit depth, parameter count, or tuning effort could explain the robustness gaps rather than the architectural choices themselves. The abstract gives no numbers on these factors, so it is impossible to rule out that the QViT advantage under quantum noise comes from a more favorable implementation. In addition, there are no error bars or details on run counts, which weakens the statements about significant degradation and marked declines. These are fixable issues, but they matter for how much weight the results can carry. This paper is for researchers who work on variational quantum circuits and need comparative data points rather than new theory. A reader already familiar with QML will see it as an incremental empirical study that asks reasonable questions about model selection. It does not introduce new methods or resolve open theoretical issues, but it engages the existing literature directly and reports concrete observations. I recommend sending it to peer review. The referees will likely request the missing implementation details and statistical information, and once those are added the paper could serve as a useful reference for the community.

Referee Report

2 major / 2 minor

Summary. The paper conducts an empirical comparative analysis of three hybrid quantum-classical neural network architectures—QCNN, QRNN, and QViT—on accuracy/generalization for low-feature (e.g., MNIST) versus high-feature datasets and on robustness to adversarial noise versus quantum noise (measurement, channel, finite-shot). It claims superior low-feature performance across models with degradation on high-feature data (especially for QCNN), greater adversarial resilience for traditional architectures, and superior quantum-noise robustness for the transformer-based QViT.

Significance. If the architecture comparisons are controlled for circuit depth, parameter count, and optimization, the results would offer practical guidance for selecting QNN variants under NISQ constraints and highlight QViT's potential noise resilience as a distinguishing architectural feature. The work fills an evaluative gap by moving beyond single-architecture studies to head-to-head testing on both classical and quantum noise.

major comments (2)

[Experimental setup / results] Experimental setup (methods/results sections): The manuscript does not report or match the total number of variational parameters, circuit depths, qubit counts, or hyperparameter search grids across QCNN, QRNN, and QViT. Without explicit controls (e.g., a table of resource metrics or identical tuning protocols), the headline claims of relative accuracy degradation and QViT quantum-noise robustness cannot be attributed to architectural differences rather than unequal resources or optimization effort.
[Results] Results on high-feature datasets and noise robustness: The abstract asserts 'significant degradation' and 'marked performance declines' without citing error bars, number of independent runs, statistical tests, or the precise high-feature datasets used. These omissions make it impossible to assess whether the reported differences exceed run-to-run variance or implementation artifacts.

minor comments (2)

[Abstract] Abstract: missing space after comma in 'QNNs,characterized'.
[Throughout] Notation: ensure consistent use of 'finite-shot effects' versus 'shot noise' throughout; define all acronyms on first use.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thorough and constructive review. We address each major comment below, proposing revisions where appropriate to improve the manuscript's clarity and rigor.

read point-by-point responses

Referee: [Experimental setup / results] Experimental setup (methods/results sections): The manuscript does not report or match the total number of variational parameters, circuit depths, qubit counts, or hyperparameter search grids across QCNN, QRNN, and QViT. Without explicit controls (e.g., a table of resource metrics or identical tuning protocols), the headline claims of relative accuracy degradation and QViT quantum-noise robustness cannot be attributed to architectural differences rather than unequal resources or optimization effort.

Authors: We acknowledge the referee's point on the need for explicit controls to isolate architectural effects. The manuscript describes each model's implementation details individually in the Methods section, but we agree that a consolidated comparison is absent. In the revised manuscript, we will add a new table summarizing qubit counts, circuit depths, variational parameters, and hyperparameter tuning protocols for QCNN, QRNN, and QViT. This addition will allow readers to better evaluate the fairness of the comparisons and support attribution of the observed differences to architectural features. revision: yes
Referee: [Results] Results on high-feature datasets and noise robustness: The abstract asserts 'significant degradation' and 'marked performance declines' without citing error bars, number of independent runs, statistical tests, or the precise high-feature datasets used. These omissions make it impossible to assess whether the reported differences exceed run-to-run variance or implementation artifacts.

Authors: The abstract summarizes key findings from the detailed results section, which specifies the high-feature datasets employed and presents performance metrics accompanied by error bars from multiple independent runs. We will revise the abstract to explicitly reference the datasets and note the inclusion of standard deviations from repeated experiments. We will also ensure the results section cites any statistical tests performed. These changes will make the claims more precise and address concerns regarding variance and reproducibility. revision: yes

Circularity Check

0 steps flagged

No circularity: purely empirical comparative study with no derivations

full rationale

The manuscript conducts simulations comparing QCNN, QRNN, and QViT on accuracy, generalization, and robustness under adversarial and quantum noise. No equations, ansatzes, uniqueness theorems, or parameter-fitting steps are present that could reduce a claimed result to its own inputs by construction. All reported outcomes derive directly from executed circuits and measured performance metrics on standard datasets, with no self-citation load-bearing on core claims and no renaming of known results as novel derivations. The study is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Empirical study relying on standard quantum circuit simulation and classical optimization practices without introducing new free parameters, axioms beyond domain norms, or invented entities.

axioms (1)

domain assumption Standard assumptions of variational quantum circuit trainability and noise model fidelity in NISQ simulations
Invoked implicitly when reporting performance under quantum noise and finite-shot effects.

pith-pipeline@v0.9.0 · 5584 in / 1163 out tokens · 86601 ms · 2026-05-07T16:23:27.511740+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

46 extracted references · 24 canonical work pages · 3 internal anchors

[1]

Amira Abbas, David Sutter, Christa Zoufal, Aurélien Lucchi, Alessio Figalli, and Stefan Woerner. 2021. The power of quantum neural networks.Nature computational science1, 6 (2021), 403–409

2021
[2]

Tasnim Ahmed, Muhammad Kashif, Alberto Marchisio, and Muhammad Shafique. 2025. A comparative analysis and noise robustness evaluation in quantum neural networks.Scientific Reports15, 1 (2025), 33654

2025
[3]

Daniel Basilewitsch, João F Bravo, Christian Tutschku, and Frederick Struckmeier. 2025. Quantum neural networks in practice: a comparative study with classical models from standard data sets to industrial images.Quantum Machine Intelligence7, 2 (2025), 110

2025
[4]

Johannes Bausch. 2020. Recurrent Quantum Neural Networks. InAdvances in Neural Information Processing Systems (NeurIPS), Vol. 33. Curran Associates, Inc., Red Hook, NY, USA, 1368–1379

2020
[5]

Julian Berberich, Daniel Fink, Daniel Pranjić, Christian Tutschku, and Christian Holm. 2024. Training robust and generalizable quantum models.Physical Review Research6, 4 (2024), 043326. doi:10.1103/physrevresearch.6.043326

work page doi:10.1103/physrevresearch.6.043326 2024
[6]

PennyLane: Automatic differentiation of hybrid quantum-classical computations

Ville Bergholm, Josh Izaac, Maria Schuld, Christian Gogolin, Shahnawaz Ahmed, Vishnu Ajith, M. Sohaib Alam, Guillermo Alonso-Linaje, B. AkashNarayanan, Ali Asadi, et al. 2018. PennyLane: Automatic differentiation of hybrid quantum-classical computations.arXiv preprintabs/1811.04968, 1811.04968 (2018), 1–19. arXiv:1811.04968

work page internal anchor Pith review arXiv 2018
[7]

Jacob Biamonte, Peter Wittek, Nicola Pancotti, Patrick Rebentrost, Nathan Wiebe, and Seth Lloyd. 2017. Quantum machine learning.Nature549, 7671 (2017), 195–202. doi:10.1038/nature23474

work page doi:10.1038/nature23474 2017
[8]

Joseph Bowles, Shahnawaz Ahmed, and Maria Schuld. 2024. Better than classical? the subtle art of benchmarking quantum machine learning models.arXiv preprint arXiv:2403.07059(2024)

work page arXiv 2024
[9]

Matthias C Caro et al. 2022. Generalization in quantum machine learning from few training data.Nature Communica- tions13, 1 (2022), 4919. doi:10.1038/s41467-022-32550-3

work page doi:10.1038/s41467-022-32550-3 2022
[10]

Marco Cerezo, Andrew Arrasmith, Ryan Babbush, Simon C Benjamin, Suguru Endo, Keisuke Fujii, Jarrod R McClean, Kosuke Mitarai, Xiao Yuan, Lukasz Cincio, et al. 2021. Variational quantum algorithms.Nature Reviews Physics3, 9 (2021), 625–644

2021
[11]

Marco Cerezo, Guillaume Verdon, Hsin-Yuan Huang, Lukasz Cincio, and Patrick J Coles. 2022. Challenges and opportunities in quantum machine learning.Nature computational science2, 9 (2022), 567–576

2022
[12]

I-Chung Chen, Harmeet Singh, V. L. Anukruti, Beate Quanz, and Kavitha Yogaraj. 2024. A survey of classical and quantum sequence models. In2024 16th International Conference on Communication Systems and Networks (COMSNETS). 24 Tran et al. IEEE, Bengaluru, India, 1006–1011. doi:10.1109/comsnets59351.2024.10456721

work page doi:10.1109/comsnets59351.2024.10456721 2024
[13]

El Amine Cherrat, Iordanis Kerenidis, Natansh Mathur, Jonas Landman, Martin Strahm, and Yun Yvonna Li. 2024. Quantum vision transformers.Quantum8 (2024), 1265. doi:10.22331/q-2024-02-22-1265

work page doi:10.22331/q-2024-02-22-1265 2024
[14]

Carlo Ciliberto, Mark Herbster, Alessandro Davide Ialongo, Massimiliano Pontil, Andrea Rocchetto, Simone Severini, and Leonard Wossnig. 2018. Quantum machine learning: a classical perspective.Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences474, 2209 (2018), 20170551

2018
[15]

Iris Cong, Soonwon Choi, and Mikhail D Lukin. 2019. Quantum convolutional neural networks.Nature Physics15, 12 (2019), 1273–1278. doi:10.1038/s41567-019-0648-8

work page doi:10.1038/s41567-019-0648-8 2019
[16]

Francesco Croce and Matthias Hein. 2020. Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks. InProceedings of the 37th International Conference on Machine Learning (ICML) (Proceedings of Machine Learning Research, Vol. 119). PMLR, Vienna, Austria, 2206–2216

2020
[17]

Chen, Stefano Mangini, and Marcel Worring

Riccardo Di Sipio, Jiun-Hung Huang, Shih-Yuan C. Chen, Stefano Mangini, and Marcel Worring. 2022. The Dawn of Quantum Natural Language Processing. In2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, Singapore, 8612–8616. doi:10.1109/icassp43922.2022.9747675

work page doi:10.1109/icassp43922.2022.9747675 2022
[18]

Yinpeng Dong, Fangzhou Liao, Tianyu Pang, Hang Su, Jun Zhu, Xiaolin Hu, and Jianguo Li. 2018. Boosting Adversarial Attacks with Momentum. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, Salt Lake City, UT, USA, 9185–9193

2018
[19]

Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. 2021. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale.arXiv preprintabs/2010.11929, 2010.11929 (2021), 1–22. arXiv:2010.11929

work page internal anchor Pith review arXiv 2021
[20]

Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, et al. 2021. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. InInternational Conference on Learning Representations (ICLR). OpenReview.net, Vienna, Austria, 1–21

2021
[21]

Vedran Dunjko and Hans J Briegel. 2018. Machine learning and artificial intelligence in the quantum domain: a review of recent progress.Reports on Progress in Physics81, 7 (2018), 074001

2018
[22]

Edward Farhi and Hartmut Neven. 2018. Classification with Quantum Neural Networks on Near Term Processors. arXiv preprintabs/1802.06002, 1802.06002 (2018), 1–21. arXiv:1802.06002

work page Pith review arXiv 2018
[23]

Goodfellow, Jonathon Shlens, and Christian Szegedy

Ian J. Goodfellow, Jonathon Shlens, and Christian Szegedy. 2015. Explaining and Harnessing Adversarial Examples. In 3rd International Conference on Learning Representations (ICLR). OpenReview.net, San Diego, CA, USA, 1–11

2015
[24]

R. M. Goodman, J. W. Miller, and P. Smyth. 1991. Objective functions for neural network classifier design. InProceedings of 1991 IEEE International Symposium on Information Theory. IEEE, Budapest, Hungary, 87–87. doi:10.1109/ISIT.1991. 695123

work page doi:10.1109/isit.1991 1991
[25]

Maxwell Henderson, Samriddhi Shakya, Shashindra Pradhan, and Tristan Cook. 2020. Quanvolutional neural networks: powering image recognition with quantum circuits.Quantum Machine Intelligence2, 1 (2020), 2. doi:10.1007/s42484- 020-00012-y

work page doi:10.1007/s42484- 2020
[26]

Hsin-Yuan Huang, Michael Broughton, Masoud Mohseni, Ryan Babbush, Sergio Boixo, Hartmut Neven, and Jarrod R McClean. 2021. Power of data in quantum machine learning.Nature communications12, 1 (2021), 2631

2021
[27]

Alex Krizhevsky. 2009. Learning multiple layers of features from tiny images. Tech. Rep., Univ. Toronto. https: //www.cs.toronto.edu/~kriz/cifar.html

2009
[28]

Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. 2015. Deep learning.nature521, 7553 (2015), 436–444

2015
[29]

Yann LeCun, Corinna Cortes, and Christopher J. C. Burges. 2010. The MNIST handwritten digit database.AT&T Labs [Online]2, 5 (2010), 1–2. http://yann.lecun.com/exdb/mnist/ Available at http://yann.lecun.com/exdb/mnist/

2010
[30]

Gang Li, Xiaoliang Zhao, and Xiugang Wang. 2024. Quantum self-attention neural networks for text classification. Science China Information Sciences67, 4 (2024), 142501. doi:10.1007/s11432-023-3879-7

work page doi:10.1007/s11432-023-3879-7 2024
[31]

Yanan Li et al. 2023. Quantum recurrent neural networks for sequential learning.Neural Networks166 (2023), 148–161. doi:10.1016/j.neunet.2023.07.003

work page doi:10.1016/j.neunet.2023.07.003 2023
[32]

Sirui Lu, Lu-Ming Duan, and Dong-Ling Deng. 2020. Quantum adversarial machine learning.Physical Review Research 2, 3 (2020), 033212. doi:10.1103/physrevresearch.2.033212

work page doi:10.1103/physrevresearch.2.033212 2020
[33]

Aleksander Madry, Aleksandar Makelov, Ludwig Schmidt, Dimitris Tsipras, and Adrian Vladu. 2018. Towards Deep Learning Models Resistant to Adversarial Attacks. In6th International Conference on Learning Representations (ICLR). OpenReview.net, Vancouver, BC, Canada, 1–28

2018
[34]

Michael A Nielsen. 2002. A simple formula for the average gate fidelity of a quantum dynamical operation.Physics Letters A303, 4 (2002), 249–252. doi:10.1016/s0375-9601(02)01272-0

work page doi:10.1016/s0375-9601(02)01272-0 2002
[35]

Susan R Sain and Vladimir N Vapnik. 1996. The nature of statistical learning theory.Technometrics38, 4 (1996), 409. doi:10.2307/1271324 A Comprehensive Analysis of Accuracy and Robustness in Quantum Neural Networks 25

work page doi:10.2307/1271324 1996
[36]

Richard M. Schmidt. 2019. Recurrent Neural Networks (RNNs): A Gentle Introduction and Overview.arXiv preprint abs/1912.05911, 1912.05911 (2019), 1–40. arXiv:1912.05911

work page arXiv 2019
[37]

Maria Schuld and Nathan Killoran. 2019. Quantum machine learning in feature Hilbert spaces.Physical review letters 122, 4 (2019), 040504

2019
[38]

Maria Schuld, Ilya Sinayskiy, and Francesco Petruccione. 2015. An introduction to quantum machine learning. Contemporary Physics56, 2 (2015), 172–185

2015
[39]

Yoshiki Takaki, Kosuke Mitarai, Makoto Negoro, Keisuke Fujii, and Masahiro Kitagawa. 2021. Learning temporal data with a variational quantum recurrent neural network.Physical Review A103, 5 (2021), 052414. doi:10.1103/physreva. 103.052414

work page doi:10.1103/physreva 2021
[40]

John R. Taylor. 1996.An Introduction to Error Analysis: The Study of Uncertainties in Physical Measurements(2 ed.). University Science Books, Sausalito, CA, USA

1996
[41]

Cordova-Esparza, Alan Ramirez-Pedraza, Esthela A

Juan Terven, Daniel M. Cordova-Esparza, Alan Ramirez-Pedraza, Esthela A. Chavez-Urbiola, and Jose A. Romero- Gonzalez. 2023. Loss Functions and Metrics in Deep Learning.arXiv preprintabs/2307.02694, 2307.02694 (2023), 1–35. arXiv:2307.02694

work page arXiv 2023
[42]

Tran, Chuong K

Ban Q. Tran, Chuong K. Luong, and Susan Mengel. 2025. Quantum Patches for Efficient Learning. InInternational Conference on Multi-disciplinary Trends in Artificial Intelligence (MIW AI). Springer, Springer Nature, Cham, Switzerland, 87–100

2025
[43]

Attention Is All You Need

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention Is All You Need.arXiv preprintabs/1706.03762, 1706.03762 (2017), 1–15. arXiv:1706.03762

work page internal anchor Pith review arXiv 2017
[44]

Guifre Vidal. 2008. Class of quantum many-body states that can be efficiently simulated.Physical Review Letters101, 11 (2008), 110501. doi:10.1103/physrevlett.101.110501

work page doi:10.1103/physrevlett.101.110501 2008
[45]

Nathan Wiebe, Alireza Kapoor, and Krysta M Svore. 2016. Quantum deep learning.Quantum Information and Computation16, 7-8 (2016), 541–587. doi:10.26421/qic16.7-8-1

work page doi:10.26421/qic16.7-8-1 2016
[46]

Kamila Zaman, Tasnim Ahmed, Muhammad Abdullah Hanif, Alberto Marchisio, and Muhammad Shafique. 2024. A comparative analysis of hybrid-quantum classical neural networks. InWorld Congress in Computer Science, Computer Engineering & Applied Computing. Springer, 102–115

2024