Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

QFGN: A Quantum Approach to High-Fidelity Implicit Neural Representations

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read The paper claims QFGN, a hybrid quantum-classical network with only 585 trainable parameters, outperforms current state-of-the-art implicit neural representations on low-resolution medical image reconstruction and super-resolution, and…

desk verdict A well-written hybrid QML paper whose central Fourier feature layer is, as written, an affine map, so the claimed frequency-balancing mechanism is unsupported, though the hardware benchmarking is a genuine effort. read the letter →

arxiv 2504.19053 v1 pith:WJPZOKFI submitted 2025-04-26 quant-ph cs.LG

Hongni Jin , Gurinder Singh , Kenneth M. Merz Jr This is my paper · ORCID
classification quant-phcs.LG
keywords quantummachinelearningimplicitneuralrepresentationFourierfeaturesspectralbiasmedicalimagereconstructionsuper-resolutionparameterizedcircuitsfrequencyspectrum
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

The paper proposes QFGN, a hybrid classical-quantum network for implicit neural representations, and claims that with far fewer trainable parameters it reconstructs medical images and performs super-resolution better than established classical and quantum baselines. The core idea is to pre-process input coordinates with a Fourier-Gaussian feature scaling layer that is meant to balance low and high frequencies before encoding them into an 8-qubit parameterized circuit. On the three low-resolution medical images tested, QFGN reports the highest PSNR and SSIM among all baselines; on noisy quantum hardware with error mitigation, its reconstruction quality is claimed to be comparable to the SIREN model. The paper argues this indicates a structural advantage of quantum circuits, which express functions as truncated Fourier series with potentially exponential frequency diversity.

What carries the argument

The central object is the Fourier-Gaussian feature scaling (FGFS) layer feeding a parameterized quantum circuit. As intended, coordinates are repeated, projected onto a fixed matrix $B$ of cosine entries $b_{k,j}=\cos(w_f s_j+\varphi_p)$, and transformed by $h_1=\Lambda B x_{\mathrm{rep}}+b$; a Gaussian factor $\varepsilon=\exp(-\gamma h_1^2)$ then attenuates large amplitudes so low frequencies do not dominate. The quantum circuit uses 8 qubits, 16 encoding gates and 256 trainable gates in a Super-Parallel ansatz where encoding and trainable gates alternate. The theoretical engine is the identity that a data re-uploading circuit's expectation value equals a sum over frequency differences $\Lambda_K-\Lambda_J$, with coefficients fixed by trainable unitaries and the observable, so diversifying encoding eigenvalues can in principle make the available frequency spectrum grow exponentially with input dimension.

What would settle it

Train QFGN on the same three images with the Fourier basis matrix $B$ replaced by a random fixed matrix of identical size, holding everything else fixed; if reconstruction quality does not drop meaningfully, the specific Fourier structure of $B$ is not the source of the reported gains. A direct check: sweep $x$ across $[0,1]$, record $h_1$, and test whether it oscillates like a sinusoid in $x$ rather than tracing a line.

Watch

Extended reading notes

Core claim

The central claim is that a quantum circuit used as an implicit neural representation can achieve high-fidelity signal reconstruction if its input encoding is pre-enriched by a classical layer that suppresses the dominance of low frequencies. The paper derives that a data re-uploading circuit's output is exactly a truncated Fourier series, and that this spectrum is redundant: L Pauli encoding gates yield only 2L+1 unique frequencies. QFGN's Fourier-Gaussian feature scaling layer is intended to supply a broad, balanced spectrum to the circuit, so the circuit does not need many repeated encoding gates. The paper reports that QFGN, with 585 trainable parameters, outperforms ReLU/Tanh MLPs, random Fourier features, SIREN, and QIREN on Pneumonia, Path, and Breast image reconstruction and super-resolution, and that on noisy hardware with combined error mitigation it reaches PSNR 31.532 dB and SSIM 0.955 on the Breast image, comparable to SIREN.

Load-bearing premise

The load-bearing premise is that equation (13), $h_1=\Lambda B x_{\mathrm{rep}}+b$, actually delivers sinusoidal Fourier features of the input coordinate $x$ to the quantum circuit; as written, it is a fixed linear projection onto cosine values evaluated at fixed sampling points, so if that premise fails the claimed frequency balancing and exponential frequency diversity do not reach the circuit.

Editorial extensions

If this is right

  • If the central claim is right, quantum INR models can beat classical Fourier-based INRs on medical image reconstruction using roughly 585 trainable parameters, about 16% fewer than SIREN.
  • Frequency balancing at the input, attenuating large low-frequency amplitudes before quantum encoding, should transfer to other quantum Fourier models that suffer from spectral bias or vanishing high-frequency coefficients.
  • On one newer quantum processor, the combination of dynamical decoupling, twirling, readout error extinction, and zero-noise extrapolation reduced QFGN's error by roughly 24% compared with no mitigation, while some individual mitigation techniques made results worse.
  • The Fourier-series view implies that quantum circuits could access exponentially many frequency combinations as input dimension grows, a scaling that classical Fourier-feature networks do not share.

Reading between the lines

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

  • Editorial inference: the reported advantage is demonstrated only on three 32×32 medical images and a specific 8-qubit circuit, so the claim of general quantum advantage for INRs is not yet supported beyond low-resolution benchmarks.
  • Editorial inference: as written, Eq. (13) computes a fixed linear function of the coordinate, not a sine or cosine of the coordinate; replacing $B$ with a random or identity matrix would reveal whether the specific Fourier structure, rather than the added linear layer and nonlinearity, drives the reported gains.
  • Editorial inference: if the Fourier-basis assumption fails, the practical difference between QFGN and a classical feature-engineered network may be the hybrid circuit's optimization landscape rather than spectral balancing.
  • Editorial inference: the hardware result is a snapshot from one device generation with 50,000 shots per expectation value; broader claims about NISQ suitability would need repeated runs across devices and error regimes.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes QFGN, a hybrid classical-quantum implicit neural representation. A classical 'Fourier-Gaussian feature scaling' (FGFS) layer is intended to provide a spectrally balanced set of Fourier features for an 8-qubit parameterized quantum circuit, whose output is linearly mapped to pixel intensities. The authors report results on three MedMNIST images for image representation and a 64x64 'super-resolution' task, claiming that QFGN with fewer parameters outperforms SIREN, QIREN, and other baselines, and that on IBM hardware with error mitigation the model is comparable to SIREN. The paper's Section 2.2 competently summarizes the Fourier-series structure of data re-uploading quantum circuits, but the novel FGFS mechanism and the experimental claims are the main contributions under review.

Significance. If correct, QFGN would be a notable step toward quantum INRs, combining classical frequency shaping with quantum Fourier structure and demonstrating noise-mitigated inference on real hardware. The paper has credible strengths: the exposition of the quantum Fourier-series framework (Section 2.2) is accurate and properly attributed to Schuld, Mhiri, and Zhao; the hardware experiments address a real NISQ concern; and the parameter counts are honestly reported. However, the central mechanism of the new layer is not what the equations implement, the 'super-resolution' experiment does not perform super-resolution, and the empirical comparison rests on three images with best-of-five reporting and no error bars. These issues directly affect the paper's main claims, so the current significance is low.

major comments (4)
  1. [Section 3, Eqs. (10)-(13)] The FGFS layer is not a Fourier feature map of the input. In Eq. (10), b_{k,j}=cos(w_f s_j + phi_p) is evaluated at fixed sampling points s_j on [-2pi,2pi], independent of the input coordinate x, and Eq. (13) then computes h1 = Lambda B xrep + b, which is affine in x. There is no term of the form cos(w_f x_i + phi_p) or sin(w_f x_i + phi_p) with the coordinate inside the sinusoid. The matrix B is therefore just a fixed linear operator, and the Gaussian window h2 = exp(-gamma h1^2) h1 does not turn this into a sinusoidal embedding. The claims in Section 3 and Figure 2 that the layer provides a 'uniformly distributed frequency spectrum' and penalizes low-frequency Fourier components are not derived from these equations. Since the frequency-balancing mechanism is the paper's main theoretical contribution, this is a load-bearing gap.
  2. [Section 4.2, Table 2] The experiment labeled 'image super-resolution' does not perform super-resolution. The text states that each original image is downsampled to 64x64 pixels as the ground truth and the model is given a grid of 64x64 inputs to construct 64x64 images. There is no low-resolution input and no upsampling, so this is another image-representation task, not super-resolution. Consequently the abstract and conclusion claims about super-resolution are unsupported.
  3. [Section 4.1, Tables 1 and 2] The empirical comparison is too weak to support the 'outperforms SOTA' claim. Only three MedMNIST images are used, all at 32x32 (Table 1) or 64x64 (Table 2); each model is trained five times and only the best run is reported, with no standard deviations, confidence intervals, or significance tests. Many PSNR margins are small (e.g., Table 1, Breast: 33.372 vs 32.649; Table 2, Breast: 26.479 vs 26.392), and the claimed percentage improvements (5.6%, 17.3%, 22.4%) do not match the values in the tables. The evaluation must report mean and standard deviation over seeds and preferably more images before any SOTA conclusion can be drawn.
  4. [Section 4.3, Tables 3 and 4] The hardware experiments use a single image (Breast) and report single-point MSE/PSNR/SSIM values without shot-noise or device-variability error bars; the 100-datapoint error-mitigation study in Table 3 shows MSE varying by up to roughly a factor of three across settings, which is not characterized statistically. Also, the 'comparable to SIREN' claim compares hardware QFGN to a simulator-trained classical SIREN, not to baselines executed on the same device, so it is not a controlled hardware comparison. This weakens contribution (3), the validation on real hardware.
minor comments (5)
  1. [Section 4.1] There is a typo: 'ReLU-MPL' should be 'ReLU-MLP'.
  2. [Section 3, Eq. (10)] The notation phi_p in {1,2,...,P} and w_f in {1,2,...,F} suggests integer indices; these should be defined as phase and frequency values, not integer sets, and the relationship to the sampling points s_j should be clarified.
  3. [Section 4.1] The sentence about a 16.5% parameter reduction is imprecise because the reduction depends on the chosen baseline (701 vs 585 for SIREN, 657 vs 585 for QIREN); the baseline should be specified in the comparison.
  4. [Appendix B.3] The use of 50,000 shots is stated, but no shot-noise analysis or standard errors are reported for the hardware expectation values.
  5. [General] No code or data availability statement is included, which would help reproducibility of the empirical results.

Circularity Check

0 steps flagged · score 0.0 of 10

No construction-level circularity: QFGN's empirical results come from standard supervised training against external baselines, and the FGFS equations' mismatch with the 'Fourier features' label is a derivation gap rather than a circular reduction.

full rationale

I walked the claimed derivation chain from Eqs. (10)-(15) through the quantum-circuit Fourier-series argument and the experimental tables. No step reduces a predicted quantity to a fitted input or defines the target result into existence. The FGFS layer defines B_{k,j}=cos(w_f s_j + phi_p) at fixed sample points s_j, so h1 = Lambda B xrep + b is affine in the input x; the paper's assertion that this provides Fourier features with a balanced spectrum is not supported by the equations. That is an internal derivation/correctness gap, not circularity: the model is still trained end-to-end with MSE and the reported numbers come from conventional supervised fitting, not from the frequency-balancing narrative. The quantum-circuit Fourier-series structure is imported from external prior work (Schuld, Zhao, Mhiri, Casas), not from the authors' own results. The only self-citation, [Singh et al. 2025], supports the generic statement that PQCs can represent functions with relatively few parameters and is not load-bearing for QFGN's central claims. No fitted parameter is renamed as a prediction, no uniqueness theorem from the authors' prior work is invoked, and no ansatz is smuggled in through the authors' own citations. The benchmark comparisons against SIREN and QIREN are independent of the paper's theoretical framing, so the appropriate finding is no significant circularity.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central design rests on hand-chosen classical feature-construction hyperparameters (gamma, F, P, Lambda) and on the prior literature's Fourier-series description of quantum circuits. No new physical entities are postulated. The ledger is dominated by an unverified link between the FGFS layer's fixed linear projection and the claimed frequency spectrum.

free parameters (3)
  • Gaussian shaping parameter gamma = 0.8
    Fixed shaping parameter in Eq. (14); controls amplitude attenuation; not optimized and not justified from data.
  • Fourier grid sizes F, P and repetition count n = not reported
    Defines the dense frequency-phase matrix B in Eqs. (10)-(12); central to the claimed frequency spectrum but the values are not given.
  • Fixed coefficient matrix Lambda = unspecified (fixed)
    Weight matrix in Eq. (13) that projects onto the Fourier-like basis; its initialization or distribution is not described and affects model behavior.
assumptions (3)
  • domain assumption A parameterized quantum circuit with data re-uploading is a truncated Fourier series in its inputs (Eq. 8).
    Taken from Schuld et al. 2021 and Mhiri et al. 2024 in Section 2.2; not re-derived, and the QFGN circuit's encoding Hamiltonians are not shown to satisfy the diagonal-eigenvalue decomposition required for Eqs. (5)-(8).
  • domain assumption Classical MLPs with Fourier feature mappings or sine activations represent truncated Fourier series and exhibit spectral bias where low frequencies dominate.
    Relying on Benbarka et al. and Rahaman et al. in Sections 2.1 and 3; used to justify why penalizing low-frequency amplitudes is beneficial.
  • ad hoc to paper Low-frequency features produce large amplitudes early in training, so an amplitude penalty on h1 preferentially removes spectral bias.
    Asserted in Section 3 after Eq. (15); no derivation connects the Gaussian factor exp(-gamma h1^2) specifically to low-frequency suppression, and as written h1 is not a Fourier expansion of x.

how reviews work

0 comments
Cite this review

Pith. "Pith review of QFGN: A Quantum Approach to High-Fidelity Implicit Neural Representations." pith.science (2026). https://pith.science/paper/WJPZOKFI

@misc{pith2026250419053,
  author       = {Pith},
  title        = {Pith review of: QFGN: A Quantum Approach to High-Fidelity Implicit Neural Representations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WJPZOKFI}},
  note         = {Machine review of arXiv:2504.19053}
}
read the original abstract

Implicit neural representations have shown potential in various applications. However, accurately reconstructing the image or providing clear details via image super-resolution remains challenging. This paper introduces Quantum Fourier Gaussian Network (QFGN), a quantum-based machine learning model for better signal representations. The frequency spectrum is well balanced by penalizing the low-frequency components, leading to the improved expressivity of quantum circuits. The results demonstrate that with minimal parameters, QFGN outperforms the current state-of-the-art (SOTA) models. Despite noise on hardware, the model achieves accuracy comparable to that of SIREN, highlighting the potential applications of quantum machine learning in this field.

Figures

Figures reproduced from arXiv: 2504.19053 by the authors.

Figure 1
Figure 1. The overall framework of QFGN. In the Fourier part [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The combination of low frequency and high frequency with different amplitudes. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. In all three image representation tasks, QFGN consistently achieved the lowest error in [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Our QFGN model consistently surpasses all baseline models. It achieves a maximum of [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 4
Figure 4. Figure 4: Image super- resolution of each model. 4.3 QFGN on Quantum Hardware All results presented in previous sections are obtained from a noiseless simulator via classical com￾puting. However, quantum circuits are designed for execution on quantum computers. In view of the cu…
Figure 5
Figure 5. Figure 5: The output from IBM-Sherbrooke. 5 Conclusion In this work, we present QFGN for quantum implicit neural representations. The intrinsic structural property of quantum circuits shows the quantum advantages of QFGN over classical neural networks as Fourier series. To mitig…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. CoQui: A Coordinate-Conditioned Quantum Implicit Generative Adversarial Network for End-to-End Image Generation

    quant-ph 2026-08 conditional novelty 6.0 of 10

    CoQui is a coordinate-conditioned quantum implicit GAN that reads each pixel from a dedicated color qubit's expectation value, decoupling qubit count from image resolution and reporting improved generation over amplit...

Reference graph

Works this paper leans on

33 extracted references · 7 canonical work pages · cited by 1 Pith paper

  1. [1]

    Vincent Sitzmann, Julien N. P. Martel, Alexander W. Bergman, David B. Lindell, and Gordon Wetzstein. Implicit neural representations with periodic activation functions. arXiv preprint arXiv:2006.09661, 2020. URL https://arxiv.org/abs/2006.09661

    arXiv 2006

  2. [2]

    Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Raghavan, Utkarsh Singhal, Ravi Ramamoorthi, Jonathan T

    Matthew Tancik, Pratul P. Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Raghavan, Utkarsh Singhal, Ravi Ramamoorthi, Jonathan T. Barron, and Ren Ng. Fourier features let networks learn high frequency functions in low dimensional domains. arXiv preprint arXiv:2006.10739, 2020. URL https://arxiv.org/abs/2006.10739

    arXiv 2006

  3. [3]

    Sinr: Spline-enhanced implicit neural representation for multi-modal registration

    Vasiliki Sideri-Lampretsa, Julian McGinnis, Huaqi Qiu, Magdalini Paschali, Walter Simson, and Daniel Rueckert. Sinr: Spline-enhanced implicit neural representation for multi-modal registration. In Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning, volume 250 of Proceedings of Machine Learning Research, pages 1462--1474....

    work page 2024

  4. [4]

    Implicit neural representation in medical imaging: A comparative survey

    Amirali Molaei, Amirhossein Aminimehr, Armin Tavakoli, Amirhossein Kazerouni, Bobby Azad, Reza Azad, and Dorit Merhof. Implicit neural representation in medical imaging: A comparative survey. arXiv preprint arXiv:2307.16142, 2023. URL https://arxiv.org/abs/2307.16142

    arXiv 2023

  5. [5]

    Neural radiance fields in medical imaging: A survey

    Xin Wang, Yineng Chen, Shu Hu, Heng Fan, Hongtu Zhu, and Xin Li. Neural radiance fields in medical imaging: A survey. arXiv preprint arXiv:2402.17797, 2025. URL https://arxiv.org/abs/2402.17797

    arXiv 2025

  6. [6]

    I-medsam: Implicit medical image segmentation with segment anything

    Xiaobao Wei, Jiajun Cao, Yizhu Jin, Ming Lu, Guangyu Wang, and Shanghang Zhang. I-medsam: Implicit medical image segmentation with segment anything. arXiv preprint arXiv:2311.17081, 2024. URL https://arxiv.org/abs/2311.17081

    arXiv 2024

  7. [7]

    Seeing Implicit Neural Representations as Fourier Series

    Nuri Benbarka, Timon Höfer, Hamd ul-moqeet Riaz, and Andreas Zell. Seeing implicit neural representations as fourier series. arXiv preprint arXiv:2109.00249, 2021. URL https://arxiv.org/abs/2109.00249

    work page Pith review arXiv 2021

  8. [8]

    Gurinder Singh, Hongni Jin, and Kenneth M. Merz Jr. Benchmarking medmnist dataset on real quantum hardware. arXiv preprint arXiv:2502.13056, 2025. URL https://arxiv.org/abs/2502.13056

    arXiv 2025

Show all 33 references
  1. [9]

    Effect of data encoding on the expressive power of variational quantum-machine-learning models

    Maria Schuld, Ryan Sweke, and Johannes Jakob Meyer. Effect of data encoding on the expressive power of variational quantum-machine-learning models. Phys. Rev. A, 103: 0 032430, 2021. doi:10.1103/PhysRevA.103.032430. URL https://link.aps.org/doi/10.1103/PhysRevA.103.032430

    2021 doi

  2. [10]

    Hamprecht, Yoshua Bengio, and Aaron Courville

    Nasim Rahaman, Aristide Baratin, Devansh Arpit, Felix Draxler, Min Lin, Fred A. Hamprecht, Yoshua Bengio, and Aaron Courville. On the spectral bias of neural networks. arXiv preprint arXiv:1806.08734, 2019. URL https://arxiv.org/abs/1806.08734

    2019 arXiv

  3. [11]

    Baraniuk

    Vishwanath Saragadam, Daniel LeJeune, Jasper Tan, Guha Balakrishnan, Ashok Veeraraghavan, and Richard G. Baraniuk. Wire: Wavelet implicit neural representations. arXiv preprint arXiv:2301.05187, 2023. URL https://arxiv.org/abs/2301.05187

    2023 arXiv

  4. [12]

    Diner: Disorder-invariant implicit neural representation

    Shaowen Xie, Hao Zhu, Zhen Liu, Qi Zhang, You Zhou, Xun Cao, and Zhan Ma. Diner: Disorder-invariant implicit neural representation. arXiv preprint arXiv:2211.07871, 2022. URL https://arxiv.org/abs/2211.07871

    2022 arXiv

  5. [13]

    Irem: High-resolution magnetic resonance (mr) image reconstruction via implicit neural representation

    Qing Wu, Yuwei Li, Lan Xu, Ruiming Feng, Hongjiang Wei, Qing Yang, Boliang Yu, Xiaozhao Liu, Jingyi Yu, and Yuyao Zhang. Irem: High-resolution magnetic resonance (mr) image reconstruction via implicit neural representation. arXiv preprint arXiv:2106.15097, 2021. URL https://ar...

    2021 arXiv

  6. [14]

    An arbitrary scale super-resolution approach for 3d mr images via implicit neural representation

    Qing Wu, Yuwei Li, Yawen Sun, Yan Zhou, Hongjiang Wei, Jingyi Yu, and Yuyao Zhang. An arbitrary scale super-resolution approach for 3d mr images via implicit neural representation. IEEE Journal of Biomedical and Health Informatics, 27 0 (2): 0 1004–1015, 2023. ISSN 2168-2208. ...

    2023

  7. [15]

    Convolutional occupancy networks

    Songyou Peng, Michael Niemeyer, Lars Mescheder, Marc Pollefeys, and Andreas Geiger. Convolutional occupancy networks. arXiv preprint arXiv:2003.04618, 2020. URL https://arxiv.org/abs/2003.04618

    2003 arXiv

  8. [16]

    Srinivasan, Matthew Tancik, Jonathan T

    Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view synthesis. arXiv preprint arXiv:2003.08934, 2020. URL https://arxiv.org/abs/2003.08934

    2003 arXiv

  9. [17]

    Neural implicit representation for highly dynamic lidar mapping and odometry

    Qi Zhang, He Wang, Ru Li, and Wenbin Li. Neural implicit representation for highly dynamic lidar mapping and odometry. arXiv preprint arXiv:2409.17729, 2024. URL https://arxiv.org/abs/2409.17729

    2024 arXiv

  10. [18]

    Pezzoli, F

    M. Pezzoli, F. Antonacci, and A. Sarti. Implicit neural representation with physics-informed neural networks for the reconstruction of the early part of room impulse responses. In Proceedings of the 10th Convention of the European Acoustics Association Forum Acusticum 2023, pa...

    2023

  11. [19]

    Piczak, Przemysław Spurek, Jacek Tabor, and Tomasz Trzciński

    Filip Szatkowski, Karol J. Piczak, Przemysław Spurek, Jacek Tabor, and Tomasz Trzciński. Hypersound: Generating implicit neural representations of audio signals with hypernetworks. arXiv preprint arXiv:2211.01839, 2024. URL https://arxiv.org/abs/2211.01839

    2024 arXiv

  12. [20]

    Hinerv: Video compression with hierarchical encoding-based neural representation

    Ho Man Kwan, Ge Gao, Fan Zhang, Andrew Gower, and David Bull. Hinerv: Video compression with hierarchical encoding-based neural representation. Advances in Neural Information Processing Systems, 36: 0 72692--72704, 2023

    2023

  13. [21]

    Joseph Shenouda, Yamin Zhou, and Robert D. Nowak. Relus are sufficient for learning implicit neural representations. arXiv preprint arXiv:2406.02529, 2024. URL https://arxiv.org/abs/2406.02529

    2024 arXiv

  14. [22]

    Finer: Flexible spectral-bias tuning in implicit neural representation by variable-periodic activation functions

    Zhen Liu, Hao Zhu, Qi Zhang, Jingde Fu, Weibing Deng, Zhan Ma, Yanwen Guo, and Xun Cao. Finer: Flexible spectral-bias tuning in implicit neural representation by variable-periodic activation functions. arXiv preprint arXiv:2312.02434, 2023. URL https://arxiv.org/abs/2312.02434

    2023 arXiv

  15. [23]

    Adrián Pérez-Salinas, Alba Cervera-Lierta, Elies Gil-Fuster, and José I. Latorre. Data re-uploading for a universal quantum classifier. Quantum, 4: 0 226, 2020. ISSN 2521-327X. doi:10.22331/q-2020-02-06-226. URL http://dx.doi.org/10.22331/q-2020-02-06-226

    2020 doi

  16. [24]

    Quantum implicit neural representations

    Jiaming Zhao, Wenbo Qiao, Peng Zhang, and Hui Gao. Quantum implicit neural representations. arXiv preprint arXiv:2406.03873, 2024. URL https://arxiv.org/abs/2406.03873

    2024 arXiv

  17. [25]

    Constrained and vanishing expressivity of quantum fourier models

    Hela Mhiri, Leo Monbroussou, Mario Herrero-Gonzalez, Slimane Thabet, Elham Kashefi, and Jonas Landman. Constrained and vanishing expressivity of quantum fourier models. arXiv preprint arXiv:2403.09417, 2024. URL https://arxiv.org/abs/2403.09417

    2024 arXiv

  18. [26]

    Classically approximating variational quantum machine learning with random fourier features

    Jonas Landman, Slimane Thabet, Constantin Dalyac, Hela Mhiri, and Elham Kashefi. Classically approximating variational quantum machine learning with random fourier features. arXiv preprint arXiv:2210.13200, 2022. URL https://arxiv.org/abs/2210.13200

    2022 arXiv

  19. [27]

    S. Shin, Y. S. Teo, and H. Jeong. Exponential data encoding for quantum supervised learning. Physical Review A, 107 0 (1), 2023. ISSN 2469-9934. doi:10.1103/physreva.107.012422. URL http://dx.doi.org/10.1103/PhysRevA.107.012422

    2023 doi

  20. [28]

    Generalization despite overfitting in quantum machine learning models

    Evan Peters and Maria Schuld. Generalization despite overfitting in quantum machine learning models. Quantum, 7: 0 1210, 2023. ISSN 2521-327X. doi:10.22331/q-2023-12-20-1210. URL http://dx.doi.org/10.22331/q-2023-12-20-1210

    2023 doi

  21. [29]

    Gentile, Youssef Achari Berrada, Elvira Shishenina, and Vincent E

    Ben Jaderberg, Antonio A. Gentile, Youssef Achari Berrada, Elvira Shishenina, and Vincent E. Elfving. Let quantum neural networks choose their own frequencies. Physical Review A, 109 0 (4), 2024. ISSN 2469-9934. doi:10.1103/physreva.109.042421. URL http://dx.doi.org/10.1103/Ph...

    2024 doi

  22. [30]

    Multidimensional fourier series with quantum circuits

    Berta Casas and Alba Cervera-Lierta. Multidimensional fourier series with quantum circuits. Physical Review A, 107 0 (6), 2023. ISSN 2469-9934. doi:10.1103/physreva.107.062612. URL http://dx.doi.org/10.1103/PhysRevA.107.062612

    2023 doi

  23. [31]

    Improved implicit neural representation with fourier reparameterized training

    Kexuan Shi, Xingyu Zhou, and Shuhang Gu. Improved implicit neural representation with fourier reparameterized training. arXiv preprint arXiv:2401.07402, 2024. URL https://arxiv.org/abs/2401.07402

    2024 arXiv

  24. [32]

    \'eliv\'agar: Efficient quantum circuit search for classification

    Sashwat Anagolum, Narges Alavisamani, Poulami Das, Moinuddin Qureshi, Eric Kessler, and Yunong Shi. \'eliv\'agar: Efficient quantum circuit search for classification. arXiv preprint arXiv:2401.09393, 2024. URL https://arxiv.org/abs/2401.09393

    2024 arXiv

  25. [33]

    Medmnist v2 - a large-scale lightweight benchmark for 2d and 3d biomedical image classification

    Jiancheng Yang, Rui Shi, Donglai Wei, Zequan Liu, Lin Zhao, Bilian Ke, Hanspeter Pfister, and Bingbing Ni. Medmnist v2 - a large-scale lightweight benchmark for 2d and 3d biomedical image classification. Scientific Data, 10 0 (1), 2023. ISSN 2052-4463. doi:10.1038/s41597-022-0...

    2023 doi

Pith tools

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