Pith. sign in

REVIEW 5 major objections 6 minor 101 references

Physics-inspired Generative AI models via real hardware-based noisy quantum diffusion

T0 review · 5 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A tuned mix of quantum and classical noise in a diffusion model's forward chain gives lower FID on MNIST than pure classical noise, and real four-qubit hardware noise can itself drive image generation.

desk verdict A genuine proof-of-concept that quantum dynamics can drive the forward process in diffusion models, with a real 4-qubit IBM demo, but the headline statistical claim about omega=0.3 is not actually established. read the letter →

arxiv 2505.22193 v1 pith:UPD62EOA submitted 2025-05-28 quant-ph cond-mat.dis-nncs.AIcs.CVcs.LG

Marco Parigi , Stefano Martina , Francesco Aldo Venturelli , Filippo Caruso This is my paper
classification quant-phcond-mat.dis-nncs.AIcs.CVcs.LG
keywords quantumdiffusionmodelsstochasticwalksgenerativeAInoisyintermediate-scaledevicesFréchetInceptionDistancenoiseasaresourcediscrete-timewalkMNISTimagegeneration
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 tries to show that quantum dynamics can be put to work inside generative diffusion models rather than being treated as an obstacle. It claims that a forward diffusion chain built from a quantum stochastic walk—a Lindblad master equation with a mixing parameter $\omega$ that interpolates between coherent quantum motion and classical random walking—produces MNIST digit sets with lower Fr\'echet Inception Distance (FID) than the fully classical chain, and that the improvement is statistically consistent over repeated runs. It also claims that the intrinsic noise of a real four-qubit quantum processor, running a discrete-time quantum walk on a cycle graph, can act as the forward noising process for a trained classical denoiser, generating 28\times28 gray-scale images. The broader point is that quantum noise on near-term hardware can be exploited as a resource for generative AI, not merely mitigated.

What carries the argument

The carrying object is a quantum stochastic walker on an 8-node cycle graph, one node per gray level, evolving under the Kossakowski\textendash Lindblad master equation $d\rho/dt = (1-\omega)i[H,\rho] + \omega \sum_j (L_j \rho L_j^\dagger - \tfrac{1}{2}\{L_j^\dagger L_j, \rho\})$, where $\omega$ interpolates from a pure quantum walk ($\omega=0$) to a classical random walk ($\omega=1$). The diagonal populations of $\rho$ become the categorical transition probabilities for each pixel in the forward chain. For the hardware version, the same cycle graph is realized as a discrete-time quantum walk using three position qubits plus one coin qubit; the graph's rotation invariance lets the authors run one forward trajectory and remap it by shifts to all gray values. The noise injected by the real device, scaled by a cosine-like delay schedule, plays the role of the diffusion schedule in a classical diffusion model.

What would settle it

Run the same discrete-time quantum walk circuit on the real processor for each of the eight possible initial positions (gray levels) and compare the eight measured output distributions: under the rotation-invariance assumption they must be cyclic shifts of one another. If qubit-dependent errors make them differ by more than the statistical uncertainty from the finite number of shots, the single-run forward chain is not a faithful model of all pixels, and generated images inherit a per-gray-level bias that this setup cannot detect.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the forward diffusion process need not be classical to be useful: a quantum stochastic walker on an 8-node cycle graph, with one node per gray level, defines a family of forward kernels indexed by $\omega$. At $\omega=0.3$ the hybrid dynamics gives a lower mean FID than the classical $\omega=1$ case over ten repetitions, with results clustered more tightly around the median, which the authors read as both better and more robust generation. The second discovery is that the same cycle-graph walk can be compiled onto a noisy real device: using an efficient circuit implementation of the discrete-time quantum walk, the hardware's intrinsic noise itself supplies the stochasticity that drives pixels toward a uniform distribution, and an MLP trained to reverse that hardware-defined forward chain produces new digit images with FID 352 on digit 0 using only four qubits. The authors frame both results as evidence that quantum noise can be a resource.

Load-bearing premise

The hardware result rests on the assumption that the noisy circuit on the real device still respects the cycle graph's rotation symmetry, so that measuring one initial state and shifting the outcome faithfully represents all eight gray levels; the paper does not test whether qubit-dependent gate and readout errors break this symmetry.

Editorial extensions

If this is right

  • At a specific hybrid mixing ($\omega=0.3$), the quantum stochastic walk forward chain yields a lower mean FID and a tighter FID distribution than the classical chain over ten runs.
  • The hardware forward chain on a real four-qubit processor generates 28\times28 eight-gray-level MNIST images with FID 352 for digit 0, using only four qubits independent of image size.
  • Because each pixel is an independent walker and graph symmetry permits one shared forward run, the circuit depth does not grow with the number of pixels.
  • Quantum noise can be used as the diffusion mechanism instead of being mitigated, potentially opening near-term hardware to generative tasks.

Reading between the lines

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

  • Editorial: if the $\omega=0.3$ advantage is not a statistical fluke, the mechanism may be faster-than-classical mixing to the uniform prior while preserving some coherent structure; a natural test is scanning $\omega$ on other discrete datasets to see whether the optimum tracks the graph's spectral properties.
  • Editorial: the unverified rotation-symmetry assumption on noisy hardware can be tested cheaply by comparing shifted runs; if it fails, the per-pixel bias would be correctable by running a small number of representative shifts and interpolating.
  • Editorial: the two protocols suggest a spectrum from continuous-time quantum stochastic walks to discrete-time hardware quantum walks; intermediate discrete-time simulations could isolate whether the FID gap between the hybrid and hardware chains comes from time discretization or from uncontrolled hardware noise.
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 / 6 minor

Summary. The paper presents two protocols for quantum diffusion models. In the first, the forward diffusion process is modeled by a quantum stochastic walk (QSW) on an 8-node cycle graph, with a mixing parameter ω interpolating between coherent quantum dynamics and classical random-walk dynamics. Training an MLP denoiser on MNIST digit 0, the authors report FID values across ω and claim that the hybrid value ω=0.3 yields lower and statistically more robust FID than the fully classical case ω=1. In the second protocol, a discrete-time quantum walk on a cycle graph is implemented with 4 qubits on the IBM device ibm_brisbane, using intrinsic hardware noise (injected via delay operations) as the forward process, and the resulting distributions are used to train an MLP to generate MNIST digits. The paper argues that quantum noise can be exploited as a resource rather than mitigated.

Significance. The hardware demonstration is a concrete proof-of-concept that a real NISQ processor can serve as the forward process of a diffusion model for image generation, using only 4 qubits for 8-gray-level MNIST pixels. The QSW framework is a natural and useful way to interpolate between classical and quantum dynamics, and the forward KL-convergence plots in Fig. 2 illustrate the expected damping of quantum oscillations by incoherent dynamics. If the claimed FID advantage at ω=0.3 were rigorously established, this would be an interesting step toward practical quantum generative models. However, the paper does not provide code or data, and the central statistical claim rests on a post-hoc scan without significance testing; the hardware protocol also relies on an untested symmetry assumption. These issues currently limit the strength of the conclusions.

major comments (5)
  1. [Section 2.2, Fig. 4] The central claim that the hybrid QSW dynamics at ω=0.3 produces 'statistically better' and 'statistically more robust' FID than the classical case (ω=1) is not supported by the reported evidence. The value ω=0.3 was selected after scanning 11 values with only 10 repetitions per value, and no significance test, confidence interval, or multiple-comparison correction is reported. Under post-hoc selection, the minimum of 11 noisy FID estimates is biased downward even under the null, so the qualitative box-plot comparison cannot establish the claim. Please provide effect sizes with uncertainties (e.g., bootstrap confidence intervals) and a test that accounts for the scan (e.g., a permutation test over the 11×10 runs, or a Dunnett/Bonferroni correction), or explicitly downgrade the claim to a preliminary observation.
  2. [Section 2.1, Eq. (4)] The QSW forward dynamics is defined only by the generic Kossakowski–Lindblad master equation; the specific Hamiltonian H and Lindblad operators L_j on the 8-node cycle graph used in all simulations are never specified. This makes the results in Figs. 2, 4, and 5 irreproducible and prevents a check of which 'interplay' of quantum and classical dynamics produces the ω=0.3 advantage. Please state H and L_j explicitly in the Methods (or include the simulation code), together with the numerical integration settings, including the meaning and value of δt=6×10^{-1} in Section 4.5.
  3. [Section 2.3, Fig. 7] The hardware protocol assumes that the 'rotation-invariant property of the cycle graph' allows one to run each walker from the same initial state and re-map outcomes by a shift operation. This is only valid if the noisy circuit on ibm_brisbane is covariant under the cyclic shift of position labels; however, the device error map in Fig. 7 shows qubit-dependent readout and gate errors that can break this symmetry. The paper does not test this assumption, for example by measuring the forward distribution for several distinct starting nodes and verifying that the shift-re-mapped distributions coincide. Without such a test, the single-run protocol may inject a per-pixel bias, so the hardware-generated images should be interpreted with this caveat.
  4. [Section 2.2 and Fig. S2] The headline FID comparison is performed only for MNIST digit 0. The abstract and discussion generalize the claim to 'MNIST images', but the supplementary results for digits 1–9 in Fig. S2 report FID values only for the ω=0.3 QSW model, with no classical baseline or statistical comparison. Please either extend the comparison to all digits or explicitly restrict the advantage claim to digit 0.
  5. [Section 2.2 and Section 4.5] The QSW simulation fixes δt=6×10^{-1} and T=20, and the hardware protocol fixes c=5×10^4 and T=20, but no sensitivity analysis is reported for these free parameters. Since the ω=0.3 advantage could in principle depend on these choices, a brief study varying δt and T (and c for the hardware protocol) would substantially strengthen the robustness of the central comparison.
minor comments (6)
  1. [Eq. (9)] The expression for 'scaling' is ambiguous in the current typesetting, as the exponent and the division by 8 are not clear from the text; please rewrite it unambiguously, for example as scaling = floor(8 sin^2(π t/(2(T−1))))/8 or an equivalent explicit formula.
  2. [Section 4.5] The library is named 'QuTiP' (not 'QuTip'), and the sentence about shot counts ('maximum available for the simulator, and reduced considering the computational resources available') should give the exact numbers for the simulator and the real device separately.
  3. [Fig. 8] Please state explicitly whether the FID value of 352 is computed on the same 6,903 generated samples used for the KL divergence in the bottom row, and clarify how the different shot counts (10^5 simulated vs 10^4 hardware) affect the comparison with the QSW-based FID of 114 in Fig. 5.
  4. [Section 2.2] The sentence 'box plots require at least 5 samples to provide solid statistical information' is a rule of thumb; it does not replace a significance test. Consider rewording to avoid implying that n=10 suffices for the 'statistically more robust' claim.
  5. [Abstract] The spelling of Fréchet is inconsistent between the abstract ('Frechet') and the text; please use a consistent notation.
  6. [Eq. (15)] In the ELBO decomposition, the term 'const' appears after the expectation without a clear definition; aligning the expression with the standard derivation (e.g., dropping the constant inside the expectation) would improve clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation; the reported FID comparison is an empirical scan, not a construction.

full rationale

The paper's derivation chain is self-contained rather than circular. The claimed hybrid advantage is an empirical observation: Fig. 4 reports 10 FID simulations for each omega in {0,0.1,...,1}, and the text selects omega=0.3 as the best-seeming value. This is a post-hoc selection, and the absence of significance testing or multiple-comparison control is a legitimate statistical weakness, but it is not a circular reduction: omega itself is not fitted to the FID objective within the model, and no equation defines omega in terms of generated-image quality. The hardware-based section is likewise externally benchmarked: the forward noise is produced by an actual ibm_brisbane circuit (Fig. 6), the denoiser is a standard MLP trained with categorical cross-entropy, and the FID=352 is computed against the original MNIST digit-0 set, not against quantities fed into the model. The self-citations ([36-38] for QSW optimal mixing, [67] for QDMs) are used as background and inspiration, not as load-bearing proofs; no uniqueness theorem or fitted ansatz is imported from them. The rotation-invariance assumption in Sec. 2.3 is an untested symmetry condition on real-device noise, but even if it fails, the result would be biased or approximate, not tautological. No step in the manuscript reduces by construction to its own inputs.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claims rest on hand-chosen constants (omega, c, delta_t, T) and on assumptions about the validity of a Lindblad equation as a diffusion generator and about the shift symmetry of the noisy device.

free parameters (4)
  • omega (quantum-classical mixing parameter) = 0.3
    Selected after scanning 11 values because it gave the lowest mean FID in Fig. 4; the headline hybrid advantage depends on this choice.
  • c (noise delay coefficient) = 5e4
    Hand-chosen in Eq. (8) so the forward QW distribution converges to uniform within T=20; no optimization or error analysis is reported.
  • delta_t (Lindblad evolution time step) = 0.6
    Fixed in Section 4.5 without justification; affects the QSW transition probabilities and therefore the FID comparison.
  • T (number of diffusion steps) = 20
    Chosen a priori for both QSW and hardware experiments; the noise schedule is designed so that this T works, so T is not an independent test.
assumptions (5)
  • standard math The Lindblad master equation with omega=1 and Lij=Sij|i><j| reproduces the classical random walk on a graph.
    Invoked in Section 2.1 to define the classical limit of the QSW forward process.
  • domain assumption The evolution of the diagonal populations of the QSW after a finite step, followed by projective measurement and reset, defines a Markovian categorical forward process that can be reversed by training a neural network.
    Used in Eqs. (5)-(6) and in the MLP training pipeline; the standard discrete diffusion assumptions are assumed to extend to these transition probabilities.
  • domain assumption The cycle graph's rotation invariance lets the model run one quantum walk and remap outcomes to all pixel values.
    Stated in Section 2.3; the remapping procedure only works if the noisy circuit also has this symmetry.
  • ad hoc to paper The idle-delay noise on ibm_brisbane is homogeneous enough across qubits to preserve the per-pixel shift symmetry.
    The paper relies on this in Section 2.3 but Fig. 7 shows qubit-dependent readout and gate errors; no correction or test is provided.
  • ad hoc to paper The cosine-inspired scaling in Eq. (9) with c=5e4 makes the forward chain converge to the uniform distribution by t=T=20.
    The schedule is chosen to guarantee convergence (Section 2.3), rather than derived or validated independently.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Physics-inspired Generative AI models via real hardware-based noisy quantum diffusion." pith.science (2026). https://pith.science/paper/UPD62EOA

@misc{pith2026250522193,
  author       = {Pith},
  title        = {Pith review of: Physics-inspired Generative AI models via real hardware-based noisy quantum diffusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UPD62EOA}},
  note         = {Machine review of arXiv:2505.22193}
}
read the original abstract

Quantum Diffusion Models (QDMs) are an emerging paradigm in Generative AI that aims to use quantum properties to improve the performances of their classical counterparts. However, existing algorithms are not easily scalable due to the limitations of near-term quantum devices. Following our previous work on QDMs, here we propose and implement two physics-inspired protocols. In the first, we use the formalism of quantum stochastic walks, showing that a specific interplay of quantum and classical dynamics in the forward process produces statistically more robust models generating sets of MNIST images with lower Fr\'echet Inception Distance (FID) than using totally classical dynamics. In the second approach, we realize an algorithm to generate images by exploiting the intrinsic noise of real IBM quantum hardware with only four qubits. Our work could be a starting point to pave the way for new scenarios for large-scale algorithms in quantum Generative AI, where quantum noise is neither mitigated nor corrected, but instead exploited as a useful resource.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

101 extracted references · 75 canonical work pages

  1. [1]

    Communications of the ACM 63(11), 139–144 (2020)

    Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Bengio, Y.: Generative adver- sarial networks. Communications of the ACM 63(11), 139–144 (2020)

    2020

  2. [2]

    Foundations and Trends® in Machine Learning 12(4), 307– 392 (2019) 12

    Kingma, D.P., Welling, M.: An introduction to variational autoencoders. Foundations and Trends® in Machine Learning 12(4), 307– 392 (2019) 12

    2019

  3. [3]

    In: Bach, F., Blei, D

    Rezende, D., Mohamed, S.: Variational infer- ence with normalizing flows. In: Bach, F., Blei, D. (eds.) Proceedings of the 32nd Inter- national Conference on Machine Learning. Proceedings of Machine Learning Research, vol. 37, pp. 1530–1538. PMLR, Lille, France (2015)

    2015

  4. [4]

    In: Bach, F., Blei, D

    Sohl-Dickstein, J., Weiss, E., Mah- eswaranathan, N., Ganguli, S.: Deep unsupervised learning using nonequilibrium thermodynamics. In: Bach, F., Blei, D. (eds.) Proceedings of the 32nd International Con- ference on Machine Learning. Proceedings of Machine Learning Research, vol. 37, pp. 2256–2265. PMLR, Lille, France (2015)

    2015

  5. [5]

    In: Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M.F., Lin, H

    Ho, J., Jain, A., Abbeel, P.: Denoising diffu- sion probabilistic models. In: Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M.F., Lin, H. (eds.) Advances in Neural Information Processing Systems, vol. 33, pp. 6840–6851 (2020)

    2020

  6. [6]

    Machine Learning: Science and Technology 5(3), 035041 (2024)

    Paiano, M., Martina, S., Giannelli, C., Caruso, F.: Transfer learning with gener- ative models for object detection on lim- ited datasets. Machine Learning: Science and Technology 5(3), 035041 (2024)

    2024

  7. [7]

    In: Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Vaughan, J.W

    Dhariwal, P., Nichol, A.: Diffusion models beat gans on image synthesis. In: Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Vaughan, J.W. (eds.) Advances in Neu- ral Information Processing Systems, vol. 34, pp. 8780–8794 (2021)

    2021

  8. [8]

    In: Interna- tional Conference on Learning Representa- tions (2021)

    Kong, Z., Ping, W., Huang, J., Zhao, K., Catanzaro, B.: Diffwave: A versatile diffu- sion model for audio synthesis. In: Interna- tional Conference on Learning Representa- tions (2021)

    2021

Show all 101 references
  1. [9]

    : Photorealistic text-to-image diffusion models with deep language under- standing

    Saharia, C., Chan, W., Saxena, S., Li, L., Whang, J., Denton, E.L., Ghasemipour, K., Gontijo Lopes, R., Karagol Ayan, B., Sali- mans, T., et al. : Photorealistic text-to-image diffusion models with deep language under- standing. Advances in Neural Information Processing System...

    2022

  2. [10]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Ommer, B.: High-resolution image synthesis with latent diffusion models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 10684–10695 (2022)

    2022

  3. [11]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Lugmayr, A., Danelljan, M., Romero, A., Yu, F., Timofte, R., Van Gool, L.: Repaint: Inpainting using denoising diffusion prob- abilistic models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 11461– 11471 (2022)

    2022

  4. [12]

    In: Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Vaughan, J.W

    Austin, J., Johnson, D.D., Ho, J., Tarlow, D., Berg, R.: Structured denoising diffusion mod- els in discrete state-spaces. In: Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Vaughan, J.W. (eds.) Advances in Neural Information Processing Systems, vol. 34, pp. 17981–179...

    2021

  5. [13]

    In: Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Vaughan, J.W

    Tashiro, Y., Song, J., Song, Y., Ermon, S.: Csdi: Conditional score-based diffusion models for probabilistic time series impu- tation. In: Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Vaughan, J.W. (eds.) Advances in Neural Information Pro- cessing Systems, vol. 34,...

    2021

  6. [14]

    https://stability.ai/ stable-image

    Stable Diffusion. https://stability.ai/ stable-image

  7. [15]

    https://dalle4ai.com

    DALL- E4. https://dalle4ai.com

  8. [16]

    In: Proceedings 35th Annual Symposium on Foundations of Computer Science, pp

    Shor, P.W.: Algorithms for quantum com- putation: discrete logarithms and factoring. In: Proceedings 35th Annual Symposium on Foundations of Computer Science, pp. 124–

  9. [17]

    Shor, P.W.: Polynomial-time algorithms for prime factorization and discrete logarithms on a quantum computer. SIAM J. Comput. 26(5), 1484–1509 (1997)

    1997

  10. [18]

    In: Proceed- ings of the Twenty-Eighth Annual ACM Symposium on Theory of Computing

    Grover, L.K.: A fast quantum mechanical algorithm for database search. In: Proceed- ings of the Twenty-Eighth Annual ACM Symposium on Theory of Computing. STOC ’96, pp. 212–219. Association for Computing 13 Machinery, New York, NY, USA (1996)

    1996

  11. [19]

    Georgescu, I.M., Ashhab, S., Nori, F.: Quan- tum simulation. Rev. Mod. Phys.86, 153–185 (2014)

    2014

  12. [20]

    O’Malley, P.J.J., Babbush, R., Kivlichan, I.D., Romero, J., McClean, J.R., Barends, R., Kelly, J., Roushan, P., Tranter, A., Ding, N., Campbell, B., Chen, Y., Chen, Z., Chiaro, B., Dunsworth, A., Fowler, A.G., Jeffrey, E., Lucero, E., Megrant, A., Mutus, J.Y., Nee- ley, M., Ne...

    2016

  13. [21]

    Babbush, R., Wiebe, N., McClean, J., McClain, J., Neven, H., Chan, G.K.-L.: Low- depth quantum simulation of materials. Phys. Rev. X 8, 011044 (2018)

    2018

  14. [22]

    Harrow, A.W., Hassidim, A., Lloyd, S.: Quantum algorithm for linear systems of equations. Phys. Rev. Lett. 103, 150502 (2009)

    2009

  15. [23]

    In: 2016 IEEE 57th Annual Symposium on Foun- dations of Computer Science (FOCS), pp

    Crosson, E., Harrow, A.W.: Simulated quan- tum annealing can be exponentially faster than classical simulated annealing. In: 2016 IEEE 57th Annual Symposium on Foun- dations of Computer Science (FOCS), pp. 714–723 (2016)

    2016

  16. [24]

    Farhi, E., Harrow, A.W.: Quantum Supremacy through the Quantum Approximate Optimization Algorithm (2019)

    2019

  17. [25]

    In: Introduction to Quantum Compu- tation and Information, pp

    Preskill, J.: Fault-tolerant quantum compu- tation. In: Introduction to Quantum Compu- tation and Information, pp. 213–269. World Scientific (1998)

    1998

  18. [26]

    Quantum 2, 79 (2018)

    Preskill, J.: Quantum Computing in the NISQ era and beyond. Quantum 2, 79 (2018)

    2018

  19. [27]

    https://www

    IBM Quantum Prcessing Unit. https://www. ibm.com/think/topics/qpu

  20. [28]

    https://www.quera.com/glossary/ processing-unit

    QuEra. https://www.quera.com/glossary/ processing-unit

  21. [29]

    Knill, E., Laflamme, R.: Theory of quantum error-correcting codes. Phys. Rev. A 55, 900– 911 (1997)

    1997

  22. [30]

    Nature Reviews Physics 6(3), 160–161 (2024)

    Campbell, E.: A series of fast-paced advances in quantum error correction. Nature Reviews Physics 6(3), 160–161 (2024)

    2024

  23. [31]

    Nature (2024)

    AI, G.Q., Collaborators: Quantum error cor- rection below the surface code threshold. Nature (2024)

    2024

  24. [32]

    Aharonov, Y., Davidovich, L., Zagury, N.: Quantum random walks. Phys. Rev. A 48, 1687–1690 (1993)

    1993

  25. [33]

    Childs, A.M., Goldstone, J.: Spatial search by quantum walk. Phys. Rev. A 70, 022314 (2004)

    2004

  26. [34]

    In: Proceedings of the Sixteenth Annual ACM-SIAM Sympo- sium on Discrete Algorithms

    Ambainis, A., Kempe, J., Rivosh, A.: Coins make quantum walks faster. In: Proceedings of the Sixteenth Annual ACM-SIAM Sympo- sium on Discrete Algorithms. SODA ’05, pp. 1099–1108. Society for Industrial and Applied Mathematics, USA (2005)

    2005

  27. [35]

    SIAM Journal on Computing 40(1), 142–164 (2011)

    Magniez, F., Nayak, A., Roland, J., Santha, M.: Search via quantum walk. SIAM Journal on Computing 40(1), 142–164 (2011)

    2011

  28. [36]

    New Journal of Physics 16(5), 055015 (2014)

    Caruso, F.: Universally optimal noisy quan- tum walks on complex networks. New Journal of Physics 16(5), 055015 (2014)

    2014

  29. [37]

    Nature Communications 7(1), 11682 (2016)

    Caruso, F., Crespi, A., Ciriolo, A.G., Scia- rrino, F., Osellame, R.: Fast escape of a quantum walker from an integrated photonic maze. Nature Communications 7(1), 11682 (2016)

    2016

  30. [38]

    Quantum Machine Intelligence 4(1), 11 (2022)

    Dalla Pozza, N., Buffoni, L., Martina, S., Caruso, F.: Quantum reinforcement learn- ing: the maze problem. Quantum Machine Intelligence 4(1), 11 (2022)

    2022

  31. [39]

    Scientific Reports 10(1), 1930 (2020)

    Abd El-Latif, A.A., Abd-El-Atty, B., Amin, M., Iliyasu, A.M.: Quantum-inspired cas- caded discrete-time quantum walks with 14 induced chaotic dynamics and cryptographic applications. Scientific Reports 10(1), 1930 (2020)

    2020

  32. [40]

    npj Quantum Information 7(1), 25 (2021)

    Zeuner, J., Pitsios, I., Tan, S.-H., Sharma, A.N., Fitzsimons, J.F., Osellame, R., Walther, P.: Experimental quantum homomorphic encryption. npj Quantum Information 7(1), 25 (2021)

    2021

  33. [41]

    Kasture, S., Acheche, S., Henriet, L., Henry, L.-P.: Multiparticle quantum walks for distin- guishing hard graphs (2025)

    2025

  34. [42]

    Lovett, N.B., Cooper, S., Everitt, M., Tre- vers, M., Kendon, V.: Universal quantum computation using the discrete-time quan- tum walk. Phys. Rev. A 81, 042330 (2010)

    2010

  35. [43]

    Scientific Reports 11(1), 11551 (2021)

    Singh, S., Chawla, P., Sarkar, A., Chan- drashekar, C.M.: Universal quantum comput- ing using single-particle discrete-time quan- tum walk. Scientific Reports 11(1), 11551 (2021)

    2021

  36. [44]

    Scien- tific Reports 13(1), 12078 (2023)

    Chawla, P., Singh, S., Agarwal, A., Srini- vasan, S., Chandrashekar, C.M.: Multi- qubit quantum computing using discrete- time quantum walks on closed graphs. Scien- tific Reports 13(1), 12078 (2023)

    2023

  37. [45]

    D¨ ur, W., Raussendorf, R., Kendon, V.M., Briegel, H.-J.: Quantum walks in optical lat- tices. Phys. Rev. A 66, 052319 (2002)

    2002

  38. [46]

    Science 336(6077), 55–58 (2012)

    Schreiber, A., G´ abris, A., Rohde, P.P., Laiho, K., ˇStefaˇ n´ ak, M., Potoˇ cek, V., Hamilton, C., Jex, I., Silberhorn, C.: A 2d quantum walk simulation of two-particle dynamics. Science 336(6077), 55–58 (2012)

    2012

  39. [47]

    Goyal, S.K., Roux, F.S., Forbes, A., Kon- rad, T.: Implementing quantum walks using orbital angular momentum of classical light. Phys. Rev. Lett. 110, 263602 (2013)

    2013

  40. [48]

    npj Quantum Information 4(1), 2 (2018)

    Lahini, Y., Steinbrecher, G.R., Bookatz, A.D., Englund, D.: Quantum logic using cor- related one-dimensional quantum walks. npj Quantum Information 4(1), 2 (2018)

    2018

  41. [49]

    Quantum Information Processing 19(12), 426 (2020)

    Acasiete, F., Agostini, F.P., Moqadam, J.K., Portugal, R.: Implementation of quantum walks on ibm quantum computers. Quantum Information Processing 19(12), 426 (2020)

    2020

  42. [50]

    Entropy 26(4) (2024)

    Razzoli, L., Cenedese, G., Bondani, M., Benenti, G.: Efficient implementation of discrete-time quantum walks on quantum computers. Entropy 26(4) (2024)

    2024

  43. [51]

    Physical Review A—Atomic, Molecular, and Optical Physics 81(2), 022323 (2010)

    Whitfield, J.D., Rodr ´ ıguez-Rosario, C.A., Aspuru-Guzik, A.: Quantum stochastic walks: A generalization of classical random walks and quantum walks. Physical Review A—Atomic, Molecular, and Optical Physics 81(2), 022323 (2010)

    2010

  44. [52]

    Academic Press (2014)

    Wittek, P.: Quantum machine learning: what quantum computing means to data mining. Academic Press (2014)

    2014

  45. [53]

    Nature 549(7671), 195–202 (2017)

    Biamonte, J., Wittek, P., Pancotti, N., Rebentrost, P., Wiebe, N., Lloyd, S.: Quan- tum machine learning. Nature 549(7671), 195–202 (2017)

    2017

  46. [54]

    Springer, (2021)

    Schuld, M., Petruccione, F.: Machine Learn- ing with Quantum Computers. Springer, (2021)

    2021

  47. [55]

    New Journal of Physics 18(2), 023023 (2016)

    McClean, J.R., Romero, J., Babbush, R., Aspuru-Guzik, A.: The theory of variational hybrid quantum-classical algorithms. New Journal of Physics 18(2), 023023 (2016)

    2016

  48. [56]

    Bharti, K., Cervera-Lierta, A., Kyaw, T.H., Haug, T., Alperin-Lea, S., Anand, A., Deg- roote, M., Heimonen, H., Kottmann, J.S., Menke, T., Mok, W.-K., Sim, S., Kwek, L.-C., Aspuru-Guzik, A.: Noisy intermediate-scale quantum algorithms. Rev. Mod. Phys. 94, 015004 (2022)

    2022

  49. [57]

    Quantum Machine Intelligence 5(1), 16 (2023)

    Das, S., Zhang, J., Martina, S., Suter, D., Caruso, F.: Quantum pattern recognition on real quantum processing units. Quantum Machine Intelligence 5(1), 16 (2023)

    2023

  50. [58]

    Quantum Machine Intelligence 4(2), 15 (2022) 15

    Geng, A., Moghiseh, A., Redenbach, C., Schladitz, K.: A hybrid quantum image edge detector for the nisq era. Quantum Machine Intelligence 4(2), 15 (2022) 15

    2022

  51. [59]

    Nature Communications 5(1), 4213 (2014)

    Peruzzo, A., McClean, J., Shadbolt, P., Yung, M.-H., Zhou, X.-Q., Love, P.J., Aspuru- Guzik, A., O’Brien, J.L.: A variational eigen- value solver on a photonic quantum pro- cessor. Nature Communications 5(1), 4213 (2014)

    2014

  52. [60]

    Nature 549(7671), 242– 246 (2017)

    Kandala, A., Mezzacapo, A., Temme, K., Takita, M., Brink, M., Chow, J.M., Gam- betta, J.M.: Hardware-efficient variational quantum eigensolver for small molecules and quantum magnets. Nature 549(7671), 242– 246 (2017)

    2017

  53. [61]

    PRX Quan- tum 3, 040326 (2022)

    Gonz´ alez-Garc ´ ıa, G., Trivedi, R., Cirac, J.I.: Error propagation in nisq devices for solving classical optimization problems. PRX Quan- tum 3, 040326 (2022)

    2022

  54. [62]

    Quantum Information Processing 20(7), 233 (2021)

    Zhang, K., Rao, P., Yu, K., Lim, H., Kore- pin, V.: Implementation of efficient quan- tum search algorithms on nisq computers. Quantum Information Processing 20(7), 233 (2021)

    2021

  55. [63]

    Farhi, E., Neven, H.: Classification with Quantum Neural Networks on Near Term Processors (2018)

    2018

  56. [64]

    Lloyd, S., Weedbrook, C.: Quantum gener- ative adversarial learning. Phys. Rev. Lett. 121, 040502 (2018)

    2018

  57. [65]

    Dallaire-Demers, P.-L., Killoran, N.: Quan- tum generative adversarial networks. Phys. Rev. A 98, 012324 (2018)

    2018

  58. [66]

    Quantum Science and Technology 4(1), 014001 (2018)

    Khoshaman, A., Vinci, W., Denis, B., Andriyash, E., Sadeghi, H., Amin, M.H.: Quantum variational autoencoder. Quantum Science and Technology 4(1), 014001 (2018)

    2018

  59. [67]

    Advanced Quantum Technologies, 2300401 (2024)

    Parigi, M., Martina, S., Caruso, F.: Quantum-noise-driven generative diffusion models. Advanced Quantum Technologies, 2300401 (2024)

    2024

  60. [68]

    Advanced Quantum Technologies, 2400171 (2024)

    Ma, H., Ye, L., Guo, X., Ruan, F., Zhao, Z., Li, M., Wang, Y., Yang, J.: Quantum genera- tive adversarial networks in a silicon photonic chip with maximum expressibility. Advanced Quantum Technologies, 2400171 (2024)

    2024

  61. [69]

    Communications Physics 7(1), 68 (2024)

    Hibat-Allah, M., Mauri, M., Carrasquilla, J., Perdomo-Ortiz, A.: A framework for demon- strating practical quantum advantage: com- paring quantum against classical generative models. Communications Physics 7(1), 68 (2024)

    2024

  62. [70]

    K¨ olle, M., Stenzel, G., Stein, J., Zielinski, S., Ommer, B., Linnhoff-Popien, C.: Quantum Denoising Diffusion Models (2024)

    2024

  63. [71]

    Quantum Machine Intelligence 6(2), 85 (2024)

    DeFalco, F., Ceschini, A., Sebastianelli, A., LeSaux, B., Panella, M.: Quantum latent dif- fusion models. Quantum Machine Intelligence 6(2), 85 (2024)

    2024

  64. [72]

    QTML 2024 Conference (2024)

    Cacioppo, A., Colantonio, L., Bordoni, S., Giagu, S.: Quantum diffusion models for quantum data learning in high-energy physics. QTML 2024 Conference (2024)

    2024

  65. [73]

    KI - K¨ unstliche Intelligenz (2024)

    De Falco, F., Ceschini, A., Sebastianelli, A., Le Saux, B., Panella, M.: Quantum hybrid diffusion models for image synthesis. KI - K¨ unstliche Intelligenz (2024)

    2024

  66. [74]

    Master’s thesis, University of Oslo (2024)

    Kivijervi, N.T.: Quantum diffusion model. Master’s thesis, University of Oslo (2024)

    2024

  67. [75]

    Zhang, B., Xu, P., Chen, X., Zhuang, Q.: Generative quantum machine learning via denoising diffusion probabilistic models. Phys. Rev. Lett. 132, 100602 (2024)

    2024

  68. [76]

    Chen, C., Zhao, Q., Zhou, M., He, Z., Sun, Z., Situ, H.: Quantum Generative Diffusion Model: A Fully Quantum-Mechanical Model for Generating Quantum State Ensemble (2024)

    2024

  69. [77]

    Cacioppo, A., Colantonio, L., Bordoni, S., Giagu, S.: Quantum Diffusion Models (2023)

    2023

  70. [78]

    Kwun, G., Zhang, B., Zhuang, Q.: Mixed- State Quantum Denoising Diffusion Proba- bilistic Model (2024)

    2024

  71. [79]

    In: Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Vaughan, J.W

    Hoogeboom, E., Nielsen, D., Jaini, P., Forr´ e, P., Welling, M.: Argmax flows and multi- nomial diffusion: Learning categorical distri- butions. In: Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Vaughan, J.W. 16 (eds.) Advances in Neural Information Pro- cessing Syst...

    2021

  72. [80]

    Reports on Mathematical Physics 3(4), 247– 274 (1972)

    Kossakowski, A.: On quantum statisti- cal mechanics of non-hamiltonian systems. Reports on Mathematical Physics 3(4), 247– 274 (1972)

    1972

  73. [81]

    Communications in Mathematical Physics 48(2), 119–130 (1976)

    Lindblad, G.: On the generators of quantum dynamical semigroups. Communications in Mathematical Physics 48(2), 119–130 (1976)

    1976

  74. [82]

    Journal of Mathe- matical Physics 17(5), 821–825 (1976)

    Gorini, V., Kossakowski, A., Sudarshan, E.C.G.: Completely positive dynamical semi- groups of n-level systems. Journal of Mathe- matical Physics 17(5), 821–825 (1976)

    1976

  75. [83]

    ATT Labs [Online]

    LeCun, Y., Cortes, C., Burges, C.: Mnist handwritten digit database. ATT Labs [Online]. Available: http://yann.lecun.com/exdb/mnist 2 (2010)

    2010

  76. [84]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp

    Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., Wojna, Z.: Rethinking the inception archi- tecture for computer vision. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 2818– 2826 (2016)

    2016

  77. [85]

    Nature Methods 11(2), 119–120 (2014)

    Krzywinski, M., Altman, N.: Visualizing sam- ples with box plots. Nature Methods 11(2), 119–120 (2014)

    2014

  78. [86]

    In: Meila, M., Zhang, T

    Nichol, A.Q., Dhariwal, P.: Improved denois- ing diffusion probabilistic models. In: Meila, M., Zhang, T. (eds.) Proceedings of the 38th International Conference on Machine Learning, ICML 2021, 18-24 July 2021, Vir- tual Event. Proceedings of Machine Learning Research, vol. 13...

    2021

  79. [87]

    Javadi-Abhari, A., Treinish, M., Krsulich, K., Wood, C.J., Lishman, J., Gacon, J., Martiel, S., Nation, P.D., Bishop, L.S., Cross, A.W., Johnson, B.R., Gambetta, J.M.: Quantum computing with Qiskit (2024)

    2024

  80. [88]

    https://www

    IBM Quantum Computing. https://www. ibm.com/quantum

  81. [89]

    In: Navab, N., Horneg- ger, J., Wells, W.M., Frangi, A.F

    Ronneberger, O., Fischer, P., Brox, T.: U- net: Convolutional networks for biomedical image segmentation. In: Navab, N., Horneg- ger, J., Wells, W.M., Frangi, A.F. (eds.) Medical Image Computing and Computer- Assisted Intervention – MICCAI 2015, pp. 234–241. Springer, Cham (2015)

    2015

  82. [90]

    In: 2nd International Con- ference on Learning Representations, ICLR 2014, Banff, AB, Canada, April 14-16, 2014, Conference Track Proceedings (2014)

    Kingma, D.P., Welling, M.: Auto-Encoding Variational Bayes. In: 2nd International Con- ference on Learning Representations, ICLR 2014, Banff, AB, Canada, April 14-16, 2014, Conference Track Proceedings (2014)

    2014

  83. [91]

    Weiss, G.H., Rubin, R.J.: Random Walks: Theory and Selected Applications, pp. 363–

  84. [92]

    Weiss, G.H.: Aspects and Applications of the Random Walk International Congress Series Random Materials and Processes, ISSN 0925-

  85. [93]

    Contemporary Physics 50(1), 339–359 (2009)

    Kempe, J.: Quantum random walks: an intro- ductory overview. Contemporary Physics 50(1), 339–359 (2009)

    2009

  86. [94]

    In: Pro- ceedings of the Thirty-Third Annual ACM Symposium on Theory of Computing

    Aharonov, D., Ambainis, A., Kempe, J., Vazi- rani, U.: Quantum walks on graphs. In: Pro- ceedings of the Thirty-Third Annual ACM Symposium on Theory of Computing. STOC ’01, pp. 50–59. Association for Computing Machinery, New York, NY, USA (2001)

    2001

  87. [95]

    https://www.ibm.com/ quantum/qiskit

    IBM Qiskit. https://www.ibm.com/ quantum/qiskit

  88. [96]

    https://qutip.org

    QuTip. https://qutip.org

  89. [97]

    https://pytorch.org

    PyTorch. https://pytorch.org

  90. [98]

    e-print arXiv:1412.6980 (2017) 17 Supplementary Material In Fig

    Kingma, D.P., Ba, J.: Adam: A method for stochastic optimization. e-print arXiv:1412.6980 (2017) 17 Supplementary Material In Fig. S1, we show forward and backward processes for image generation implemented by the simulatorfake brisbane for the digit 0 of the MNIST dataset. In...

    2017 arXiv

  91. [134]

    IEEE Computer Society, USA (1994)

    1994

  92. [505]

    John Wiley & Sons, Ltd (1982)

    1982

  93. [5850]

    North-Holland (1994)

    1994

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.