Pith. sign in

REVIEW 5 major objections 5 minor 56 references

Hybrid Quantum Generative Adversarial Networks To Inverse Design Metasurfaces For Incident Angle-Independent Unidirectional Transmission

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

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

Pith's one-line read A hybrid quantum GAN inverse-designs metasurface cells that hit 95% of target far-field profiles while using 30% less data.

desk verdict A useful QGAN inverse-design workflow undermined by missing reverse-direction simulations and internal inconsistencies in the headline claims. read the letter →

arxiv 2507.03518 v1 pith:KCCIG2X4 submitted 2025-07-04 physics.optics cond-mat.soft

Sreeraj Rajan Warrier , Jayasri Dontabhaktuni This is my paper
classification physics.opticscond-mat.soft
keywords inversedesignmetasurfacesquantumgenerativeadversarialnetworksunidirectionaltransmissionperovskitesolarcellsvariationalautoencoderangle-independentall-dielectricmetasurface
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper sets out to show that inverse design of a metasurface can be treated as a conditional image-generation problem and solved by a hybrid quantum-classical generative model, which the authors call LaSt-QGAN (latent style-based quantum GAN). The model pairs a pretrained variational autoencoder with a quantum generator and a classical discriminator, encoding each candidate unit cell as an RGB image whose channels carry the ambient refractive index, the resonator refractive index, and the resonator thickness. The authors report that this model reproduces target far-field radiation patterns at 95% fidelity, needs about 30% fewer training examples than a classical GAN baseline, and produces designs that maintain their response for incidence angles from $-60^\circ$ to $+60^\circ$. Embedding the generated bowtie and cubic metasurfaces into a perovskite solar cell is claimed to raise simulated conversion efficiency by up to 95% relative to a no-metasurface baseline.

What carries the argument

The central mechanism is the latent style-based quantum generator within LaSt-QGAN, paired with a pretrained variational autoencoder. The VAE first reduces each RGB-encoded unit-cell image to a compact latent code, so the quantum circuit generates latent features rather than raw pixels; in each layer the generator's rotation angles are set by an affine map of the concatenated noise vector and conditional radiation-profile vector, and a patch of sub-generator circuits applies $R_Y$ rotations with circular CNOT entanglements, with outputs read as Pauli-$\sigma_z$ expectations. This VAE-to-quantum-latent pipeline is what the authors credit with allowing training on 500 images and with avoiding the mode collapse typical of classical GANs.

What would settle it

Compute the reverse-direction transmission $S_{21}$ and compare it with $S_{12}$ for the generated unit cells over the full $-60^\circ$ to $60^\circ$ angle range; if the two are equal, or the asymmetry disappears at oblique incidence, the claimed unidirectionality is not realized. A second check is to brute-force sample the three-parameter RGB design space and see whether any point in it produces the target far-field profile, which would test whether the encoding is rich enough to contain the target at all.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a quantum-enhanced generative model can navigate a restricted, three-channel metasurface design space well enough to meet a demanding optical target from a small dataset. The unit cell is fixed at $5 \times 5 \times 5.5\ \mu\text{m}^3$ with two shape templates, and only the surrounding-medium refractive index, the resonator refractive index, and the resonator thickness are allowed to vary. A variational autoencoder compresses the $64 \times 64 \times 3$ encoded images into a low-dimensional latent representation, and a patch-style quantum generator built from $R_Y$ rotations and circular CNOT entanglements learns to produce latent features conditional on a five-value radiation-profile vector. The authors report 95% fidelity of simulated far-field patterns to the targets and a 30% reduction in data requirement relative to a classical GAN. They also supply a material look-up table that substitutes real materials such as GaN and SiC for the predicted refractive indices while retaining about 90% accuracy, and they embed the generated designs in a perovskite solar cell, reporting an efficiency gain from 9.36% (baseline) to 18.28% (bowtie metasurface) at normal incidence that is maintained at oblique angles.

Load-bearing premise

The design flow assumes that every useful metasurface can be described by only three varying quantities (ambient refractive index, resonator refractive index, and resonator thickness) on two fixed shape templates inside a fixed unit cell, and the paper never checks that this small space actually contains a design achieving angle-independent unidirectional transmission.

Editorial extensions

If this is right

  • If the 95% fidelity claim holds, the conditional generation pipeline can be redirected to arbitrary target far-field profiles, because the conditional vector already encodes the radiation response.
  • The claimed 30% data reduction means the same hybrid-VAE-plus-quantum-generator recipe could be used for other photonic inverse-design tasks where full-wave simulations are expensive and training sets are small.
  • The material look-up table makes the designs physically testable: substituting real materials such as GaN and SiC preserves about 90% of the simulated performance, so the generated unit cells are candidates for fabrication.
  • In the perovskite solar-cell embedding, the bowtie metasurface raises simulated efficiency to 18.28% from a 9.36% baseline and retains gains above 70% at 45° and 60° incidence.

Reading between the lines

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

  • A reverse-direction simulation computing $S_{21}$ under illumination from the opposite side is the decisive test the paper leaves unperformed; without it, the claimed unidirectionality is inferred from a single transmission direction and one far-field pattern.
  • Because every quantum run is paired with a variational autoencoder, the claimed data saving is not cleanly attributable to the quantum generator; replacing the quantum sub-generators with a classical network inside the same VAE-GAN would isolate the quantum contribution.
  • The three-channel RGB encoding restricts the search to two shape templates and three scalar values; extending the method to other functionalities would likely require more channels or shape parameters, and the 95% fidelity figure would need re-measurement in that larger space.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The manuscript presents LaSt-QGAN, a hybrid quantum-classical generative model combining a variational autoencoder with a quantum GAN, and applies it to inverse-design all-dielectric metasurface unit cells. The authors claim that the generated designs reproduce target far-field radiation profiles with 95% fidelity, require 30% less training data than classical GANs, and, when embedded in a perovskite solar cell, raise conversion efficiency by 95% independently of incidence angle from -60 to 60 degrees. The paper describes an RGB image encoding of the unit-cell parameters, the quantum-circuit architecture, a comparison with a DCGAN, a material lookup table, and COMSOL/CST simulations.

Significance. The application of a hybrid quantum GAN to metasurface inverse design is a potentially interesting direction, and the paper contains a concrete solar-cell application with tabulated performance metrics and a material-substitution study. However, the central functional claim is not demonstrated: no reverse-illumination simulation is reported, so unidirectional transmission is never established. In addition, the abstract's data-reduction and angle-independence claims are inconsistent with numbers in Section 3 and Table 2, and the 95% fidelity metric is undefined. As written, the paper does not support its headline conclusions, although the underlying methodology could be salvageable with substantial additional validation.

major comments (5)
  1. [Section 2, Figure 1] Only the forward scattering parameter S12 and a forward far-field radiation pattern are reported; no S21 or reverse-illumination simulation is shown. Unidirectional transmission is defined by asymmetry under reversal of the propagation direction, and a single forward transmission spectrum cannot establish it. This omission is load-bearing because the title, abstract, and solar-cell interpretations all depend on the generated metasurfaces being unidirectional.
  2. [Abstract vs. Section 3, Table 1] The abstract claims a 30% reduction in data requirements, but the text states that the QGAN works with 500 samples while the DCGAN requires 20,000 samples, which is a 97.5% reduction. No source or calculation for the 30% figure is provided, and the two numbers cannot both be correct as stated.
  3. [Abstract and Table 2] The claim that the conversion efficiency 'improves by 95% and remains independent of incident angle' is contradicted by the tabulated data: bowtie efficiency falls from 18.282% at 0 degrees to 10.679% at 60 degrees, a 41.6% drop. The relative enhancement over the baseline is indeed close to 95% at all four angles, but that is a different claim and should be stated explicitly instead of claiming angle-independent efficiency.
  4. [Section 3, Figures 7-8] The 95% fidelity value is never defined, and the validation compares simulated far-field profiles of generated designs with the target profiles used as conditional inputs. This protocol measures conditional fit rather than independent inverse-design prediction. The authors should define the fidelity metric, clarify which targets were held out, and report the same metric for a classical baseline on the same held-out set.
  5. [Section 2, RGB encoding] The RGB encoding represents each design by only three scalar quantities (ambient refractive index, resonator refractive index, and thickness) within two fixed shape templates and a fixed 5 by 5 micrometer unit cell. No forward simulation is provided to show that this low-dimensional space contains any design with angle-independent unidirectional transmission, so the inverse search may be optimizing over an insufficient design space. The authors should validate the expressiveness of this parameterization, for example by forward-simulating a diverse set of designs under both forward and reverse illumination.
minor comments (5)
  1. [Figure 1] The text refers to the S12 spectrum and the far-field profile in panel (d), but the caption and panel labels are ambiguous; panel (c) should clearly be labeled as the transmission spectrum and panel (d) as the far-field radiation pattern.
  2. [Section 2] The unit-cell dimensions are given as '5 x 5 x 5.5 micrometers cubed' and later as '5 x 5 micrometers squared'; please state the substrate and air-layer thicknesses in a single consistent set of dimensions.
  3. [Equations (1)-(4)] There are typesetting errors in the displayed equations, such as the subscript in the expectation value in Eq. (1); a careful proofread of the math is needed.
  4. [Table 2] The abstract claims independence over -60 to 60 degrees, but Table 2 lists only positive angles 0, 30, 45, and 60 degrees; show negative-angle data or state the symmetry assumption explicitly.
  5. [Throughout] The manuscript contains numerous typographical errors, for example 'on a a GPU server' in Section 2, and would benefit from professional copyediting.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the inverse-design pipeline is evaluated by forward EM simulation of generated designs against held-out target profiles, which is independent of the trained model.

full rationale

The paper's central claims are empirical results of an inverse-design machine-learning pipeline, not derivations that reduce to their inputs. The QGAN is trained with a conditional vector γ encoding target radiation profiles and with RGB-encoded unit-cell images; the reported '95% fidelity' is assessed by comparing simulated far-field profiles of generated designs to target profiles (Section 3, Figures 7 and 8), including designs from a validation subset held out from training. This is a standard fit-quality and generalization check, not a circular prediction: the forward electromagnetic simulation in CST/COMSOL is an external evaluation independent of the network weights. The 30% data-reduction claim is an empirical comparison with a DCGAN baseline, and the material look-up table is an external post-processing step whose 'about 90% precision' is verified by re-simulation, not by construction. The solar-cell efficiency improvement is computed from COMSOL simulations comparing baseline and metasurface-embedded cells, and the claim of angle-independent enhancement is a relative improvement statement consistent with Table 2, not an equation that assumes its conclusion. No uniqueness theorem, ansatz, or load-bearing result is imported from the authors' own prior work; the cited LaSt-QGAN architecture [34] is external. The main weakness—that unidirectional transmission is never directly demonstrated via reverse-direction S21 simulation—is a correctness/falsifiability concern, not a circularity concern, and does not warrant a nonzero circularity score under the specified criteria.

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

The central claim rests on simulation accuracy, on the sufficiency of a highly compressed RGB design representation, and on the equivalence between a directional far-field pattern and true unidirectional transmission. All three are assumed rather than demonstrated, and the paper provides no external benchmarks or experimental checks.

free parameters (4)
  • VAE latent dimension and quantum-circuit width/depth = unspecified
    Not reported; determines the capacity of the generative model and directly affects the fidelity claims.
  • Conditional vector gamma size = 5
    Set to 5 key far-field parameters because of qubit constraints; the specific five parameters are never listed.
  • beta and IWAE sample count = beta=2, k=5
    Hand-chosen VAE hyperparameters with no ablation or sensitivity analysis.
  • Refractive index of 'sample material' = n=9.27
    Introduced as a training-set material value without a named real material; later replaced by lookup-table materials.
assumptions (5)
  • domain assumption Maxwell's equations solved by CST and COMSOL accurately model the unit-cell and solar-cell responses
    All performance numbers come from commercial electromagnetic simulations, with no experimental validation.
  • ad hoc to paper The RGB encoding with only refractive index, thickness, and ambient index varying is a sufficient design representation
    Section 2 fixes the shape templates and unit-cell size; no proof is given that this low-dimensional space contains angle-independent unidirectional designs.
  • ad hoc to paper The far-field RCS main lobe direction and 3-dB width adequately define unidirectional transmission
    Unidirectionality requires asymmetric transmission (S12 vs S21); the paper only reports S12 and one radiation pattern.
  • domain assumption The nearest-material lookup with 'about 90% precision' preserves the designed function
    Section 3 material substitution; precision is asserted without quantitative error analysis.
  • ad hoc to paper The conditional vector of 5 parameters captures the essential target radiation profile
    Chosen due to qubit limits; the specific five parameters are not defined.
invented entities (1)
  • Hypothetical dielectric material with refractive index 9.27
    purpose: Serves as an extreme training example for the QGAN's continuous refractive-index channel; later replaced by real materials.
    No real material with this value is identified, and performance with this simulated index does not validate manufacturability.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hybrid Quantum Generative Adversarial Networks To Inverse Design Metasurfaces For Incident Angle-Independent Unidirectional Transmission." pith.science (2026). https://pith.science/paper/KCCIG2X4

@misc{pith2026250703518,
  author       = {Pith},
  title        = {Pith review of: Hybrid Quantum Generative Adversarial Networks To Inverse Design Metasurfaces For Incident Angle-Independent Unidirectional Transmission},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KCCIG2X4}},
  note         = {Machine review of arXiv:2507.03518}
}
abstract

Optimization of metasurface designs for specific functionality is a challenging problem due to the intricate relation between structural features and electromagnetic responses. Recently, many researchers resolved to inverse design of metasurfaces for efficient design parameters based on methods such as parameter optimization, evolutionary optimization and machine learning. In this paper a hybrid quantum machine learning method which uses quantum encoders to enhance the performance of a classical GAN is applied to implement inverse design of a metasurface. Aiming towards angle-independent unidirectional transmission, this approach combines a Quantum Generative Adversarial Network (QGAN) with a Variational Autoencoder (VAE) to optimize metasurface designs. Incident-angle independent uni-directional transmission has potential applications in efficient solar cells, thermal cooling, non-reciprocal devices, etc. However, it is very challenging to achieve such a metasurface design via forward methods and hence very few studies exist till now in this direction. The methodology employed in this work reduces the data requirement for inverse design by 30% compared to conventional GAN-based methods. More importantly, the developed metasurface designs show high fidelity of 95% with regard to the targeted far-field radiation patterns. We also provide a material look-up table for feasible substitutes of the obtained material design with real materials and yet maintaining performance accuracy. Further, we embed the inverse-designed metasurfaces into Perovskite solar cell layers to demonstrate the improvement in its performance. We observe that the conversion efficiency of the example perovskite solar cell improves by 95% and remains independent of incident angle in the range -60$^\circ$ to 60$^\circ$ within the desired frequency range.

Figures

Figures reproduced from arXiv: 2507.03518 by the authors.

Figure 1
Figure 1. Design and characterization of a high-Q metasurface unit cell: (a) Tellurium bow-tie metasurface cell of [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Encoding scheme for the unit cell of the metasurface in to RGB channels. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Schematic training architecture of a hybrid quantum generative adversarial network (QGAN) with a pretrained [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Workflow of testing and validation of a hybrid quantum GAN - from latent space sampling to the generation [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Architecture of the β-VAE and IWAE used in the paper. The loss functions for β-VAE and IWAE are given in eq 3 and eq 4 below: LBeta-VAE = X N i=1 ∥xi − x˜i∥ 2 | {z } Reconstruction Loss +β ·  − 1 2 X D j=1 [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Quantum neural network employed for the QGAN model. a) The Patch Quantum Generator connected to a [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Comparison of the far-field profiles and the unit cell designs of the target, DCGAN model, and LaSt-QGAN [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 8
Figure 8. Figure 8: Simulated 3D far-field profiles from 4 sets of La-St QGAN designs using pretrained [PITH_FULL_IMAGE:figures/full_fig_p010_8.png]
Figure 9
Figure 9. Figure 9: The target far-field radar cross-section (RCS) profile (red box) is compared with simulated results using [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Performance metrics Voc, Jsc, FF, and efficiency as a function of the angle of incidence for cubic and bowtie metasurface designs. 4 Conclusions This study employs Latent Style-based Quantum GAN (LaSt-QGAN) as a hybrid quantum-classical framework for the inverse desig…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

56 extracted references · 22 canonical work pages

  1. [1]

    Flat optics with designer metasurfaces

    Nanfang Yu and Federico Capasso. Flat optics with designer metasurfaces. Nature materials, 13(2):139–150, 2014

    2014

  2. [2]

    Planar photonics with metasurfaces

    Alexander V Kildishev, Alexandra Boltasseva, and Vladimir M Shalaev. Planar photonics with metasurfaces. Science, 339(6125):1232009, 2013

    2013

  3. [3]

    Directional scattering switching from an all-dielectric phase change metasurface

    Gonzalo Santos, Maria Losurdo, Fernando Moreno, and Yael Gutiérrez. Directional scattering switching from an all-dielectric phase change metasurface. Nanomaterials, 13(3):496, 2023

    work page 2023

  4. [4]

    Eden Rephaeli, Aaswath Raman, and Shanhui H. Fan. Ultrabroadband photonic structures to achieve high- performance daytime radiative cooling. Nano Letters, 13(4):1457–1461, 2013

    work page 2013

  5. [5]

    Overvig, Sander A

    Adam C. Overvig, Sander A. Mann, and Andrea Alù. Thermal metasurfaces: Complete emission control by combining local and nonlocal light-matter interactions. Phys. Rev. X, 11:021050, Jun 2021

    work page 2021

  6. [6]

    Controlling thermal emission with metasurfaces and its applications

    Qiongqiong Chu, Fan Zhong, Xiaohe Shang, Ye Zhang, Shining Zhu, and Hui Liu. Controlling thermal emission with metasurfaces and its applications. Nanophotonics, 13(8):1279–1301, 2024

    work page 2024

  7. [7]

    Recent advances in planar optics: from plasmonic to dielectric metasurfaces

    Patrice Genevet, Federico Capasso, Francesco Aieta, Mohammadreza Khorasaninejad, and Robert Devlin. Recent advances in planar optics: from plasmonic to dielectric metasurfaces. Optica, 4(1):139–152, 2017

    2017

  8. [8]

    Computational scaling in inverse photonic design through factorization caching

    Ahmet Onur Dasdemir, Victor Minden, and Emir Salih Magden. Computational scaling in inverse photonic design through factorization caching. Applied Physics Letters, 123(22):221106, 2023

    2023

Show all 56 references
  1. [9]

    Inverse design in nanophotonics

    Sean Molesky, Zeyu Lin, Alexander Y Piggott, Weng Cho Chew Jin, Jelena Vuckovi´c, and Alejandro W Rodriguez. Inverse design in nanophotonics. Nature Photonics, 12:659–670, 2018

    2018

  2. [10]

    Global optimization of dielectric metasurfaces using a physics-driven neural network

    Jicheng Jiang and Jonathan A Fan. Global optimization of dielectric metasurfaces using a physics-driven neural network. Nano letters, 19(8):5366–5372, 2019

    2019

  3. [11]

    Knight, and Aaswath P

    Christopher Yeung, Ryan Tsai, Benjamin Pham, Brian King, Yusaku Kawagoe, David Ho, Julia Liang, Mark W. Knight, and Aaswath P. Raman. Global inverse design across multiple photonic structure classes using generative deep learning. Advanced Optical Materials, 9(20):2100548, 2021

    2021

  4. [12]

    Training deep neural networks for the inverse design of nanophotonic structures

    Deyu Liu, Yichen Tan, Ehsan Khoram, and Zongfu Yu. Training deep neural networks for the inverse design of nanophotonic structures. ACS Photonics, 5(4):1365–1369, 2018

    2018

  5. [13]

    Deep learning enabled inverse design in nanophotonics

    Sunae So, Trevon Badloe, Jaebum Noh, Jorge Bravo-Abad, and Junsuk Rho. Deep learning enabled inverse design in nanophotonics. Nanophotonics, 9(5):1041–1057, 2020

    2020

  6. [14]

    Inverse molecular design using machine learning: Genera- tive models for matter engineering

    Benjamin Sanchez-Lengeling and Alán Aspuru-Guzik. Inverse molecular design using machine learning: Genera- tive models for matter engineering. Science, 361(6400):360–365, 2018

    2018

  7. [15]

    Kun-Hua Tu, Hejin Huang, Sangho Lee, Wonmoo Lee, Zehao Sun, Alfredo Alexander-Katz, and Caroline A. Ross. Machine learning predictions of block copolymer self-assembly. Advanced Materials, 32(52):2005713, 2020

    2020

  8. [16]

    Deep-learning-enabled fast optical identification and characterization of 2d materials.Advanced Materials, 32(29), 2020

    Bingnan Han et al. Deep-learning-enabled fast optical identification and characterization of 2d materials.Advanced Materials, 32(29), 2020. 13 arXiv Template A PREPRINT

    2020

  9. [17]

    Simultaneous inverse design of materials and structures via deep learning: Demonstration of dipole resonance engineering using core–shell nanoparticles

    Sunae So, Jungho Mun, and Junsuk Rho. Simultaneous inverse design of materials and structures via deep learning: Demonstration of dipole resonance engineering using core–shell nanoparticles. ACS Applied Materials & Interfaces, 11(27):24264–24268, 2019

    2019

  10. [18]

    Adversarial autoencoders,

    Alireza Makhzani, Jonathon Shlens, Navdeep Jaitly, Ian Goodfellow, and Brendan Frey. Adversarial autoencoders,

  11. [19]

    Mode collapse in generative adversarial networks: An overview

    Youssef Kossale, Mohammed Airaj, and Aziz Darouichi. Mode collapse in generative adversarial networks: An overview. In 2022 8th International Conference on Optimization and Applications (ICOA) , pages 1–6, 2022

    2022

  12. [20]

    Challenges and opportunities in quantum machine learning

    Marco Cerezo, Guillaume Verdon, Hsin-Yuan Huang, et al. Challenges and opportunities in quantum machine learning. Nature Computational Science, 2:567–576, 2022

    2022

  13. [21]

    Unsupervised semantic-preserving adversarial hashing for image search

    Cheng Deng et al. Unsupervised semantic-preserving adversarial hashing for image search. IEEE Transactions on Image Processing, 28:4032–4044, 2019

    2019

  14. [22]

    Quantum Computing in the NISQ Era and Beyond

    John Preskill. Quantum Computing in the NISQ Era and Beyond. Quantum, 2:79, 2018

    2018

  15. [23]

    Quantum generative adversarial learning

    Seth Lloyd and Christian Weedbrook. Quantum generative adversarial learning. Physical Review Letters , 121(4):040502, 2018

    2018

  16. [24]

    Quantum generative adversarial networks

    Pierre-Luc Dallaire-Demers and Nathan Killoran. Quantum generative adversarial networks. Physical Review A, 98(1):012324, 2018

    2018

  17. [25]

    Xudong Mao, Qing Li, Haoran Xie, Raymond Y . K. Lau, Zhen Wang, and Stephen Paul Smolley. Least squares generative adversarial networks. In Proceedings of the IEEE International Conference on Computer Vision , pages 2794–2802, 2017

    2017

  18. [26]

    Perceptual adversarial networks for image-to-image transformation

    Chang Wang, Cheng Xu, Chunpeng Wang, and Dacheng Tao. Perceptual adversarial networks for image-to-image transformation. IEEE Transactions on Image Processing, 27:4066–4079, 2018

    2018

  19. [27]

    Sainath, Oriol Vinyals, Andrew Senior, and Hasim Sak

    Tara N. Sainath, Oriol Vinyals, Andrew Senior, and Hasim Sak. Convolutional, long short-term memory, fully connected deep neural networks. In 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 4580–4584. IEEE, 2015

    2015

  20. [28]

    Convex Optimization

    Stephen Boyd and Lieven Vandenberghe. Convex Optimization. Cambridge University Press, Cambridge, UK, 2004

    2004

  21. [29]

    Self-attention generative adversarial networks, 2018

    Han Zhang, Ian Goodfellow, Dimitris Metaxas, and Augustus Odena. Self-attention generative adversarial networks, 2018. Preprint at https://arxiv.org/abs/1805.08318

    2018 arXiv

  22. [30]

    Batch normalization: Accelerating deep network training by reducing internal covariate shift, 2015

    Sergey Ioffe and Christian Szegedy. Batch normalization: Accelerating deep network training by reducing internal covariate shift, 2015. Preprint at https://arxiv.org/abs/1502.03167

    2015 arXiv

  23. [31]

    Spectral normalization for generative adversarial networks, 2018

    Takeru Miyato, Toshiki Kataoka, Masanori Koyama, and Yuichi Yoshida. Spectral normalization for generative adversarial networks, 2018. Preprint at https://arxiv.org/abs/1802.05957

    2018 arXiv

  24. [32]

    G. N. Pande and John Middleton. NUMETA 90 Numerical Methods in Engineering: Theory and Applications: Numerical Techniques for Engineering Analysis and Design . CRC Press, Boca Raton, FL, USA, 2014

    2014

  25. [33]

    DCGAN Tutorial – PyTorch Tutorials

    Nathan Inkawhich. DCGAN Tutorial – PyTorch Tutorials. https://pytorch.org/tutorials/beginner/ dcgan_faces_tutorial.html, 2022. Accessed: May 2022

    2022

  26. [34]

    Latent style-based quantum gan for high-quality image generation, 2024

    Su Yeon Chang, Supanut Thanasilp, Bertrand Le Saux, Sofia Vallecorsa, and Michele Grossi. Latent style-based quantum gan for high-quality image generation, 2024. Preprint at https://arxiv.org/abs/2406.02668

    2024 arXiv

  27. [35]

    Kingma and Max Welling

    Diederik P. Kingma and Max Welling. Auto-Encoding Variational Bayes, 2013. Preprint at https://doi.org/ 10.48550/arXiv.1312.6114

    2013 doi arXiv

  28. [36]

    Experimental Quantum Generative Adversarial Networks for Image Generation

    He-Liang Huang, Yuxuan Du, Ming Gong, Youwei Zhao, Yulin Wu, Chaoyue Wang, Shaowei Li, Futian Liang, Jin Lin, Yu Xu, Rui Yang, Tongliang Liu, Min-Hsiu Hsieh, Hui Deng, Hao Rong, Cheng-Zhi Peng, Chao-Yang Lu, Yu-Ao Chen, Dacheng Tao, Xiaobo Zhu, and Jian-Wei Pan. Experimental Q...

    2021

  29. [37]

    CST Microwave STUDIO SUITE ®, 2019

    Dassault Systèmes, Vélizy-Villacoublay, France. CST Microwave STUDIO SUITE ®, 2019. Accessed: 2023-05-16

    2019

  30. [38]

    Kim, Lakshmi Raju, and Wenshan Cai

    Zhaocheng Liu, Dayu Zhu, Kyu-Tae Lee, Andrew S. Kim, Lakshmi Raju, and Wenshan Cai. Compounding meta-atoms into metamolecules with hybrid artificial intelligence techniques. Advanced Materials, 32(6):1904790, 2020

    2020

  31. [39]

    Designing nanophotonic structures using conditional deep convolutional generative adversarial networks

    So Sunae and Junsuk Rho. Designing nanophotonic structures using conditional deep convolutional generative adversarial networks. Nanophotonics, 8(7):1255–1261, 2019

    2019

  32. [40]

    COMSOL Multiphysics® v

    COMSOL AB, Stockholm, Sweden. COMSOL Multiphysics® v. 6.2, 2023. Accessed: 2024-05-29. 14 arXiv Template A PREPRINT

    2023

  33. [41]

    beta-V AE: Learning Basic Visual Concepts with a Constrained Variational Framework

    Irina Higgins, Loic Matthey, Arka Pal, Christopher Burgess, Xavier Glorot, Matthew Botvinick, Shakir Mohamed, and Alexander Lerchner. beta-V AE: Learning Basic Visual Concepts with a Constrained Variational Framework. In International Conference on Learning Representations . O...

    2017

  34. [42]

    Importance Weighted Autoencoders, 2015

    Yuri Burda, Roger Grosse, and Ruslan Salakhutdinov. Importance Weighted Autoencoders, 2015. Preprint at https://arxiv.org/abs/1509.00519

    2015 arXiv

  35. [43]

    PyTorch: An Imperative Style, High-Performance Deep Learning Library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, Alban Desmaison, Andreas Kopf, Edward Yang, Zachary DeVito, Martin Raison, Alykhan Tejani, Sasank Chilamkurthy, Benoit Steiner, Lu ...

    2019

  36. [44]

    Solomon Kullback and Richard A. Leibler. On information and sufficiency. Annals of Mathematical Statistics , 22(1):79–86, 1951

    1951

  37. [45]

    Pennylane: Automatic differentiation of hybrid quantum-classical computations, 2018

    Ville Bergholm et al. Pennylane: Automatic differentiation of hybrid quantum-classical computations, 2018. Preprint at https://arxiv.org/abs/1811.04968

    2018 arXiv

  38. [46]

    Grabowska, and Stefano Carrazza

    Carlos Bravo-Prieto, Jacopo Baglio, Marco Cè, Anthony Francis, Dorota M. Grabowska, and Stefano Carrazza. Style-based quantum generative adversarial networks for monte carlo events. Quantum, 6:777, 2022

    2022

  39. [47]

    Variational quantum algorithms

    Marco Cerezo et al. Variational quantum algorithms. Nature Reviews Physics, 3(9):625–644, 2021

    2021

  40. [48]

    Love, Alán Aspuru- Guzik, and Jeremy L

    Alberto Peruzzo, Jarrod McClean, Peter Shadbolt, Man-Hong Yung, Xiao-Qi Zhou, Peter J. Love, Alán Aspuru- Guzik, and Jeremy L. O’Brien. A variational eigenvalue solver on a photonic quantum processor. Nature Communications, 5:4213, 2014

    2014

  41. [49]

    Classical and quantum algorithms for orthogonal neural networks, 2021

    Iordanis Kerenidis, Jonas Landman, and Natansh Mathur. Classical and quantum algorithms for orthogonal neural networks, 2021. Preprint at https://arxiv.org/abs/2106.07198

    2021 arXiv

  42. [50]

    Reluplex made more practical: Leaky ReLU

    Jin Xu, Zishan Li, Bowen Du, Miaomiao Zhang, and Jing Liu. Reluplex made more practical: Leaky ReLU. In 2020 IEEE Symposium on Computers and Communications (ISCC) , pages 1–7, 2020

    2020

  43. [51]

    Approximation by superpositions of a sigmoidal function

    George Cybenko. Approximation by superpositions of a sigmoidal function. Mathematics of Control, Signals and Systems, 2:303–314, 1989

    1989

  44. [52]

    M. N. Polyanskiy. Refractive index database. https://refractiveindex.info. (accessed: Aug 2024)

    2024

  45. [53]

    Polarization and incident angle independent multifunctional and multiband tunable thz metasurface based on VO2

    Rameez Iqbal, Usman Ul Rehman Qureshi, Chen Jie, Zia Ur Rahman, and Naseer Jafar. Polarization and incident angle independent multifunctional and multiband tunable thz metasurface based on VO2. Nanomaterials, 14(12):1048, 2024

    2024

  46. [54]

    Soma Zandi, Prateek Saxena, and Nima E. Gorji. Numerical simulation of heat distribution in RGO-contacted perovskite solar cells using comsol. Solar Energy, 197:105–110, 2020

    2020

  47. [55]

    Reference air mass 1.5 spectra

    National Renewable Energy Laboratory. Reference air mass 1.5 spectra. https://www.nrel.gov/grid/ solar-resource/spectra-am1.5. Accessed: 2025-02-22. 15

    2025

  48. [2015]

    Preprint at https://arxiv.org/abs/1511.05644

    arXiv

Pith tools

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