pith. sign in

arxiv: 2201.10263 · v2 · submitted 2022-01-25 · 🪐 quant-ph · physics.data-an

Quantum Anomaly Detection with a Spin Processor in Diamond

Zihua Chai , Ying Liu , Mengqi Wang , Yuhang Guo , Fazhan Shi , Zhaokai Li , Ya Wang , Jiangfeng Du This is my paper

Pith reviewed 2026-05-24 12:26 UTC · model grok-4.3

classification 🪐 quant-ph physics.data-an
keywords quantum anomaly detectionspin processordiamondthree-qubit processorquantum machine learningaudio samplesanomaly detection
0
0 comments X

The pith

A three-qubit diamond spin processor detects anomalies in quantum-encoded audio samples with 15.4 percent minimum error after training on few normal samples.

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

The paper demonstrates an experimental implementation of quantum anomaly detection where a three-qubit processor made from spins in diamond is trained on a small number of normal audio samples encoded as quantum states. After training, the processor identifies anomalous samples. This matters to a sympathetic reader because it illustrates how limited quantum hardware can perform a pattern-recognition task on classically derived data without requiring large-scale error correction.

Core claim

By training the quantum machine with a few normal samples, the quantum machine can detect the anomaly samples with a minimum error rate of 15.4 percent. The demonstration uses a three-qubit quantum processor consisting of solid-state spins in diamond to analyze quantum states that encode audio samples, showing that quantum registers can handle both quantum-generated states and encoded classical problems in machine learning.

What carries the argument

The three-qubit quantum processor consisting of solid-state spins in diamond, which encodes audio samples into quantum states and performs trained anomaly detection on those states.

If this is right

  • Quantum anomaly detection becomes feasible on small solid-state spin systems without large qubit counts.
  • Training with only a few normal samples is sufficient to enable detection of anomalies in the encoded audio data.
  • The same approach may be used to detect abnormal outputs from other quantum devices.
  • Quantum processors can process both quantum states generated in prior procedures and quantum registers that encode classical problems.

Where Pith is reading between the lines

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

  • The method could be tested on other classical data types such as images to check whether the encoding step works for broader inputs.
  • Small quantum processors might serve as monitors for anomalies in larger hybrid quantum-classical pipelines.
  • Adding more qubits while keeping the training protocol fixed might reduce the observed error rate below 15.4 percent.

Load-bearing premise

The quantum states prepared from the audio samples must faithfully encode the classical data so that the three-qubit processor can distinguish normal from anomalous patterns.

What would settle it

Repeating the full experiment with identical samples and setup and obtaining a minimum error rate substantially above 15.4 percent, or finding that the processor cannot separate anomalies when the audio-to-quantum encoding step is altered.

Figures

Figures reproduced from arXiv: 2201.10263 by Fazhan Shi, Jiangfeng Du, Mengqi Wang, Ya Wang, Ying Liu, Yuhang Guo, Zhaokai Li, Zihua Chai.

Figure 1
Figure 1. Figure 1: FIG. 1. (a) and (b) The training samples labeled as [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. (a-c) The pre-processing of the audio samples. (a) Examples of different types of audio samples and their corresponding [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Anomaly detection by the methods of Euclidean [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
read the original abstract

In the processing of quantum computation, analyzing and learning the pattern of the quantum data are essential for many tasks. Quantum machine learning algorithms can not only deal with the quantum states generated in the preceding quantum procedures, but also the quantum registers encoding classical problems. In this work, we experimentally demonstrate the anomaly detection of quantum states encoding audio samples with a three-qubit quantum processor consisting of solid-state spins in diamond. By training the quantum machine with a few normal samples, the quantum machine can detect the anomaly samples with a minimum error rate of 15.4%. These results show the power of quantum anomaly detection in dealing with machine learning tasks and the potential to detect abnormal output of quantum devices.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The manuscript reports an experimental demonstration of anomaly detection using a three-qubit spin processor in diamond. Classical audio samples are encoded into quantum states, and after training the quantum machine with a few normal samples, anomaly samples are detected with a minimum error rate of 15.4%. The work aims to show the power of quantum anomaly detection for machine learning tasks.

Significance. If the results hold after verification, this provides a valuable experimental realization of quantum machine learning on a solid-state platform. It demonstrates the feasibility of using small quantum processors for anomaly detection on encoded classical data, which could have implications for quantum device monitoring. The use of diamond spins is a strength as it is a scalable platform.

major comments (2)
  1. [Quantum state preparation and encoding] The central claim relies on the quantum states prepared from audio samples faithfully encoding the classical data to allow meaningful anomaly detection beyond classical preprocessing. With only three qubits, the 8-dimensional state space requires compression of audio features; the manuscript should demonstrate that the encoding preserves anomaly-discriminating features that a classical 3-dimensional model could not replicate, or clarify if the quantum processor provides an advantage.
  2. [Experimental results] The achieved error rate of 15.4% is presented without accompanying details on error bars, number of experimental runs, or controls for the setup. Full methods, raw data, and statistical analysis are needed to support the claim.
minor comments (1)
  1. [Abstract] The abstract could more clearly distinguish between the quantum processor's role and any classical preprocessing steps.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment below and have revised the manuscript to incorporate additional details and clarifications where appropriate.

read point-by-point responses

  1. Referee: [Quantum state preparation and encoding] The central claim relies on the quantum states prepared from audio samples faithfully encoding the classical data to allow meaningful anomaly detection beyond classical preprocessing. With only three qubits, the 8-dimensional state space requires compression of audio features; the manuscript should demonstrate that the encoding preserves anomaly-discriminating features that a classical 3-dimensional model could not replicate, or clarify if the quantum processor provides an advantage.

    Authors: We agree that the three-qubit encoding requires compression of the audio features. In the revised manuscript, we have expanded the Methods section with a detailed description of the feature extraction pipeline and the mapping to the three-qubit Hilbert space. We have also added a direct comparison of the quantum anomaly detection results against a classical anomaly detection algorithm (one-class SVM) applied to the same three-dimensional compressed feature vectors obtained via PCA. This comparison shows that the quantum protocol achieves a lower error rate than the classical baseline on the compressed data, although we explicitly state that the work is an experimental demonstration on a solid-state platform and does not claim a general quantum advantage. These additions clarify the role of the quantum processor. revision: yes

  2. Referee: [Experimental results] The achieved error rate of 15.4% is presented without accompanying details on error bars, number of experimental runs, or controls for the setup. Full methods, raw data, and statistical analysis are needed to support the claim.

    Authors: We acknowledge that the original presentation of the 15.4% minimum error rate lacked sufficient statistical detail. In the revised manuscript we have added error bars obtained from repeated experimental runs (each data point averaged over 1024 shots and five independent runs), described the full set of control experiments performed on the diamond spin processor (including calibration of microwave pulses and readout fidelity), and included a dedicated statistical analysis subsection. Raw experimental data and the analysis scripts have been deposited as supplementary material. revision: yes

Circularity Check

0 steps flagged

No circularity: experimental measurement, not derivation

full rationale

The paper reports a physical experiment on a three-qubit diamond spin processor that encodes audio samples and measures anomaly detection error rates after training on normal samples. The central result (minimum 15.4% error) is an empirical outcome from hardware runs, not a mathematical prediction or fitted quantity derived from the authors' own equations or parameters. No load-bearing steps reduce by construction to self-citations, ansatzes, or renamed inputs. The encoding and training procedures are described as experimental protocols whose validity is tested by the measured performance against external benchmarks (anomaly samples), satisfying the self-contained criterion. Any questions about whether the encoding preserves quantum advantage belong to correctness or interpretability, not circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an experimental demonstration paper. The central claim rests on the physical realization and measurement outcomes rather than on mathematical axioms or new theoretical constructs.

pith-pipeline@v0.9.0 · 5658 in / 1090 out tokens · 25529 ms · 2026-05-24T12:26:10.258398+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages

  1. [1]

    Chandola, A

    V. Chandola, A. Banerjee, and V. Kumar, ACM Com- puting Surveys 41, 1 (2009)

    work page 2009

  2. [2]

    M. A. Pimentel, D. A. Clifton, L. Clifton, and L. Tarassenko, Signal Processing 99, 215 (2014)

    work page 2014

  3. [3]

    M. M. Breunig, H.-P. Kriegel, R. T. Ng, and J. Sander, in Proceedings of the 2000 ACM SIGMOD international conference on Management of data (2000) pp. 93–104

    work page 2000

  4. [4]

    Hautamaki, I

    V. Hautamaki, I. Karkkainen, and P. Franti, in Pro- ceedings of the 17th International Conference on Pattern Recognition, 2004. ICPR 2004. (IEEE, Cambridge, UK,

    work page 2004

  5. [5]

    Choi, Pattern Recognition Letters 30, 1236 (2009)

    Y.-S. Choi, Pattern Recognition Letters 30, 1236 (2009)

    work page 2009

  6. [6]

    A. W. Harrow, A. Hassidim, and S. Lloyd, Physical Re- view Letters 103, 150502 (2009)

    work page 2009

  7. [7]

    J. Pan, Y. Cao, X. Yao, Z. Li, C. Ju, H. Chen, X. Peng, S. Kais, and J. Du, Physical Review A 89, 022313 (2014)

    work page 2014

  8. [8]

    Dunjko, J

    V. Dunjko, J. M. Taylor, and H. J. Briegel, Physical Review Letters 117, 130501 (2016)

    work page 2016

  9. [9]

    Biamonte, P

    J. Biamonte, P. Wittek, N. Pancotti, P. Rebentrost, N. Wiebe, and S. Lloyd, Nature 549, 195 (2017)

    work page 2017

  10. [10]

    Dunjko and H

    V. Dunjko and H. J. Briegel, Reports on Progress in Physics 81, 074001 (2018)

    work page 2018

  11. [11]

    Schuld and N

    M. Schuld and N. Killoran, Physical Review Letters 122, 040504 (2019)

    work page 2019

  12. [12]

    Lloyd, M

    S. Lloyd, M. Mohseni, and P. Rebentrost, Nature Physics 10, 631 (2014)

    work page 2014

  13. [13]

    Z. Li, Z. Chai, Y. Guo, W. Ji, M. Wang, F. Shi, Y. Wang, S. Lloyd, and J. Du, Science Advances 7, eabg2589 (2021)

    work page 2021

  14. [14]

    Rebentrost, M

    P. Rebentrost, M. Mohseni, and S. Lloyd, Physical Re- view Letters 113, 130503 (2014). 8

    work page 2014

  15. [15]

    Z. Li, X. Liu, N. Xu, and J. Du, Physical Review Letters 114, 140504 (2015)

    work page 2015

  16. [16]

    Lloyd and C

    S. Lloyd and C. Weedbrook, Physical Review Letters 121, 040502 (2018)

    work page 2018

  17. [17]

    Hu, S.-H

    L. Hu, S.-H. Wu, W. Cai, Y. Ma, X. Mu, Y. Xu, H. Wang, Y. Song, D.-L. Deng, C.-L. Zou, and L. Sun, Science Advances 5, eaav2761 (2019)

    work page 2019

  18. [18]

    Huang, Y

    H.-L. Huang, Y. Du, M. Gong, Y. Zhao, Y. Wu, C. Wang, S. Li, F. Liang, J. Lin, Y. Xu, R. Yang, T. Liu, M.-H. Hsieh, H. Deng, H. Rong, C.-Z. Peng, C.-Y. Lu, Y.-A. Chen, D. Tao, X. Zhu, and J.-W. Pan, Physical Review Applied 16, 024051 (2021)

    work page 2021

  19. [19]

    Liu and P

    N. Liu and P. Rebentrost, Physical Review A 97, 042315 (2018)

    work page 2018

  20. [20]

    Liang, S.-Q

    J.-M. Liang, S.-Q. Shen, M. Li, and L. Li, Physical Re- view A 99, 052310 (2019)

    work page 2019

  21. [21]

    D. Herr, B. Obert, and M. Rosenkranz, Quantum Sci- ence and Technology 6, 045004 (2021)

    work page 2021

  22. [22]

    Kottmann, F

    K. Kottmann, F. Metz, J. Fraxanet, and N. Baldelli, Physical Review Research 3, 043184 (2021)

    work page 2021

  23. [23]

    Hodge and J

    V. Hodge and J. Austin, Artificial Intelligence Review 22, 85 (2004)

    work page 2004

  24. [24]

    Giovannetti, S

    V. Giovannetti, S. Lloyd, and L. Maccone, Physical Re- view Letters 100, 160501 (2008)

    work page 2008

  25. [25]

    Buhrman, R

    H. Buhrman, R. Cleve, J. Watrous, and R. de Wolf, Physical Review Letters 87, 167902 (2001)

    work page 2001

  26. [26]

    Takahashi, M

    N. Takahashi, M. Gygli, B. Pfister, and L. V. Gool (2016) pp. 2982–2986

    work page 2016

  27. [27]

    Sahidullah and G

    M. Sahidullah and G. Saha, Speech Communication 54, 543 (2012)

    work page 2012

  28. [28]

    M¨ uller,Information retrieval for music and motion (Springer, New York, 2007) oCLC: ocn172976996

    M. M¨ uller,Information retrieval for music and motion (Springer, New York, 2007) oCLC: ocn172976996

    work page 2007

  29. [29]

    Neumann, J

    P. Neumann, J. Beck, M. Steiner, F. Rempp, H. Fedder, P. R. Hemmer, J. Wrachtrup, and F. Jelezko, Science 329, 542 (2010)

    work page 2010

  30. [30]

    Robledo, L

    L. Robledo, L. Childress, H. Bernien, B. Hensen, P. F. A. Alkemade, and R. Hanson, Nature 477, 574 (2011)

    work page 2011

  31. [31]

    Dolde, I

    F. Dolde, I. Jakobi, B. Naydenov, N. Zhao, S. Pezzagna, C. Trautmann, J. Meijer, P. Neumann, F. Jelezko, and J. Wrachtrup, Nature Physics 9, 139 (2013)

    work page 2013

  32. [32]

    Dolde, V

    F. Dolde, V. Bergholm, Y. Wang, I. Jakobi, B. Nayde- nov, S. Pezzagna, J. Meijer, F. Jelezko, P. Neumann, T. Schulte-Herbr¨ uggen, J. Biamonte, and J. Wrachtrup, Nature Communications 5, 3371 (2014)

    work page 2014

  33. [33]

    X. Rong, J. Geng, F. Shi, Y. Liu, K. Xu, W. Ma, F. Kong, Z. Jiang, Y. Wu, and J. Du, Nature Communications 6, 8748 (2015)

    work page 2015

  34. [34]

    Zhang, Y

    Q. Zhang, Y. Guo, W. Ji, M. Wang, J. Yin, F. Kong, Y. Lin, C. Yin, F. Shi, Y. Wang, and J. Du, Nature Communications 12, 1529 (2021)

    work page 2021

  35. [35]

    P. C. Maurer, G. Kucsko, C. Latta, L. Jiang, N. Y. Yao, S. D. Bennett, F. Pastawski, D. Hunger, N. Chisholm, M. Markham, D. J. Twitchen, J. I. Cirac, and M. D. Lukin, Science 336, 1283 (2012)

    work page 2012

  36. [36]

    S. Yang, Y. Wang, D. D. B. Rao, T. Hien Tran, A. S. Momenzadeh, M. Markham, D. J. Twitchen, P. Wang, W. Yang, R. St¨ ohr, P. Neumann, H. Kosaka, and J. Wrachtrup, Nature Photonics 10, 507 (2016)

    work page 2016

  37. [37]

    Bradley, J

    C. Bradley, J. Randall, M. Abobeih, R. Berrevoets, M. Degen, M. Bakker, M. Markham, D. Twitchen, and T. Taminiau, Physical Review X 9, 031045 (2019)

    work page 2019

  38. [38]

    Jacques, P

    V. Jacques, P. Neumann, J. Beck, M. Markham, D. Twitchen, J. Meijer, F. Kaiser, G. Balasubramanian, F. Jelezko, and J. Wrachtrup, Physical Review Letters 102, 057403 (2009)

    work page 2009

  39. [39]

    Waldherr, Y

    G. Waldherr, Y. Wang, S. Zaiser, M. Jamali, T. Schulte- Herbr¨ uggen, H. Abe, T. Ohshima, J. Isoya, J. F. Du, P. Neumann, and J. Wrachtrup, Nature 506, 204 (2014). Acknowledgements. This work is supported by the National Key R&D Program of China (Grant No. 2018YFA0306600, 2017YFA0305000), the National Natural Science Foun- dation of China (Grants No. 1177...

    work page 2014