pith. machine review for the scientific record. sign in

arxiv: 2603.14144 · v2 · submitted 2026-03-14 · 🪐 quant-ph

Recognition: 2 theorem links

· Lean Theorem

Fast Single Nitrogen-Vacancy Center Ramsey Characterization using a Physics-Informed Neural Network

Chao Shang , Gregory D. Fuchs
Authors on Pith no claims yet

Pith reviewed 2026-05-15 10:50 UTC · model grok-4.3

classification 🪐 quant-ph
keywords nitrogen-vacancy centersRamsey characterizationphysics-informed neural networkshyperfine couplingquantum sensing13C spinsdenoising
0
0 comments X

The pith

A physics-informed neural network reconstructs clean Ramsey waveforms from noisy minimal-sweep data on single NV centers and estimates their hyperfine couplings to 13C spins, achieving up to 40 times faster measurements.

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

Precise mapping of the local spin environment around single diamond NV centers is essential for quantum sensing and networking but requires long data averaging to beat down noise for standard fitting. NVRNet addresses this by using a machine learning model pretrained on physics simulations of spin dynamics to turn sparse noisy Ramsey traces into denoised signals while directly inferring the coupling strengths to nearby carbon-13 nuclei. Lightweight adapters are then fine-tuned on limited real data from each specific NV center to close the gap between simulation and experiment. The approach lowers the reconstruction error to 0.44-0.67 times the raw noise and produces forward simulations that match observed features with FFT errors of 0.10-0.19. This slashes the time needed for characterization by as much as 40 times, opening the way to higher throughput experiments.

Core claim

NVRNet is a physics-informed simulation-to-reality pipeline that employs a two-stage time-frequency U-Net denoiser augmented with an attention-based time-domain U-Net, pretrained on Hamiltonian spin simulations with calibrated noise, and uses parameter-efficient adapters fine-tuned on experimental data. A subsequent transformer extracts hyperfine parameters. Across three NV centers the fine-tuned model reduces median reconstruction error on held-out few-sweep traces to 0.44-0.67 times the experimental noise level, with normalized FFT errors of 0.10-0.19, supporting up to 40x faster Ramsey characterization.

What carries the argument

NVRNet pipeline: a U-Net based denoiser pretrained on simulated Ramsey signals from NV spin Hamiltonians and adapted via parameter-efficient fine-tuning to real data, paired with a transformer estimator for 13C hyperfine parameters.

If this is right

  • Fewer sweeps suffice to obtain usable data for hyperfine inference, directly cutting acquisition time.
  • Denoised waveforms and parameter estimates allow reliable forward modeling that reproduces key experimental signatures.
  • High-throughput screening of NV centers for quantum applications becomes practical.
  • The method provides a hardware-compatible path for autonomous characterization without extensive post-processing.

Where Pith is reading between the lines

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

  • Similar simulation-to-reality adapter strategies could accelerate characterization in other quantum sensing platforms like superconducting qubits or trapped ions.
  • The reduced data needs might allow measurements in shorter total times, minimizing sensitivity to slow drifts in the apparatus.
  • Extending the pipeline to include more complex spin environments or multi-qubit interactions would be a natural next step for broader applicability.

Load-bearing premise

The simulation-trained model can be adapted to match real NV center data sufficiently well using only small amounts of experimental data to tune lightweight adapters without retraining the entire network.

What would settle it

Run the adapted model on a new held-out set of few-sweep Ramsey traces from an NV center and verify if the median error stays below the raw experimental noise level as reported.

Figures

Figures reproduced from arXiv: 2603.14144 by Chao Shang, Gregory D. Fuchs.

Figure 1
Figure 1. Figure 1: FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5 [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7 [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8 [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9 [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10 [PITH_FULL_IMAGE:figures/full_fig_p014_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: FIG. 11. Distribution of retained [PITH_FULL_IMAGE:figures/full_fig_p016_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: FIG. 12 [PITH_FULL_IMAGE:figures/full_fig_p017_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: FIG. 13 [PITH_FULL_IMAGE:figures/full_fig_p018_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: FIG. 14 [PITH_FULL_IMAGE:figures/full_fig_p022_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: FIG. 15 [PITH_FULL_IMAGE:figures/full_fig_p023_15.png] view at source ↗
read the original abstract

Precise characterization of the local spin environment of single diamond nitrogen-vacancy (NV) centers is crucial for advancing quantum sensing, quantum networking, and the optimization of quantum materials. However, single NV center fluorescence measurements requires long averaging times to obtain clean data that is suitable for conventional model fitting, and that constitutes a key experimental bottleneck for high-throughput characterization. To address this, we introduce \textsc{NVRNet}, a physics-informed simulation-to-reality machine learning pipeline that maps minimal-sweep, noisy Ramsey data to a denoised waveform while directly estimating the hyperfine coupling to proximal ${}^{13}\mathrm{C}$ nuclear spins. The pipeline's denoiser utilizes a two-stage time-frequency U-Net and an attention-augmented time-domain U-Net, pretrained on Hamiltonian-based spin-dynamics simulations with experimentally calibrated noise. To effectively bridge the simulation-to-reality gap, parameter-efficient adapters are attached to the backbone and fine-tuned on targeted experimental data. Across three distinct NV centers, this experimentally fine-tuned model reduces the median reconstruction error on held-out, few-sweep traces to $0.44\text{-}0.67\times$ of the raw experimental noise level. Subsequently, a transformer-based estimator extracts the underlying hyperfine parameters. Forward reconstructions derived from these inferred parameters faithfully reproduce the dominant experimental time- and frequency-domain features, yielding representative normalized fast Fourier transform (FFT) reconstruction errors of $0.10\text{-}0.19$. By reducing both the required data volume and acquisition time, \textsc{NVRNet} enables up to $\sim 40\times$ acceleration of the measurement process, establishing a fast, hardware-compatible pathway for robust hyperfine inference and autonomous qubit characterization.

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 introduces NVRNet, a physics-informed pipeline that pretrains a two-stage time-frequency U-Net denoiser and attention-augmented time-domain U-Net on Hamiltonian spin-dynamics simulations, attaches parameter-efficient adapters for fine-tuning on experimental Ramsey data from single NV centers, and uses a transformer estimator to extract 13C hyperfine couplings from few-sweep noisy traces. Across three NV centers it reports median reconstruction errors reduced to 0.44-0.67 times the raw noise level on held-out traces and normalized FFT reconstruction errors of 0.10-0.19, claiming up to 40x acceleration of the measurement process.

Significance. If the hyperfine estimates prove quantitatively accurate, the work would offer a practical route to high-throughput NV characterization by cutting acquisition time while preserving dominant time- and frequency-domain features, directly addressing a bottleneck in quantum sensing and networking experiments. The simulation-to-reality adapter strategy is a concrete strength that could generalize to other qubit platforms.

major comments (2)
  1. [Abstract] Abstract: the central claim that the transformer extracts accurate hyperfine couplings rests solely on forward-simulation fidelity (normalized FFT errors 0.10-0.19) matching experimental traces; no direct comparison is reported against hyperfine values obtained from conventional high-SNR Ramsey fits on the same three NV centers, leaving open the possibility that the estimator recovers only the dominant envelope while missing or biasing the actual couplings.
  2. [Methods / Results] Methods / Results (training protocol): the reported performance gains lack any description of training/validation splits, error bars on the 0.44-0.67x noise reduction, statistical significance tests, or ablation studies that isolate the contribution of the adapters versus the pretrained backbone, so the quantitative claims cannot be assessed for robustness.
minor comments (1)
  1. [Figures] Figure captions and text should explicitly state the number of experimental traces per NV center and the exact definition of 'held-out' data to allow reproduction.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their insightful comments on our manuscript. We address each major comment below and have revised the manuscript accordingly to strengthen the presentation and validation of our results.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that the transformer extracts accurate hyperfine couplings rests solely on forward-simulation fidelity (normalized FFT errors 0.10-0.19) matching experimental traces; no direct comparison is reported against hyperfine values obtained from conventional high-SNR Ramsey fits on the same three NV centers, leaving open the possibility that the estimator recovers only the dominant envelope while missing or biasing the actual couplings.

    Authors: We agree that a direct quantitative comparison between the hyperfine parameters inferred by the transformer estimator and those obtained from conventional high-SNR Ramsey fits on the same NV centers would provide additional validation. While the forward reconstructions from the inferred parameters faithfully reproduce the dominant features of the experimental data (as evidenced by the low normalized FFT errors), this does not explicitly confirm the accuracy of individual coupling values. In the revised manuscript, we will include such a comparison using the high-SNR data available for the three NV centers, reporting the differences in the extracted hyperfine couplings. This will help demonstrate that the estimator recovers accurate parameters rather than just the envelope. revision: yes

  2. Referee: [Methods / Results] Methods / Results (training protocol): the reported performance gains lack any description of training/validation splits, error bars on the 0.44-0.67x noise reduction, statistical significance tests, or ablation studies that isolate the contribution of the adapters versus the pretrained backbone, so the quantitative claims cannot be assessed for robustness.

    Authors: We acknowledge the need for more rigorous statistical reporting to assess the robustness of our quantitative claims. The current manuscript focuses on the overall performance across the three NV centers but does not detail the splits or provide error bars and ablations. In the revised version, we will add a dedicated subsection describing the training and validation splits used during pretraining and fine-tuning, include error bars (e.g., standard deviations across multiple runs or NV centers) on the noise reduction metrics, conduct statistical significance tests (such as paired t-tests) where applicable, and perform ablation studies to quantify the impact of the parameter-efficient adapters compared to the pretrained backbone alone. revision: yes

Circularity Check

0 steps flagged

No significant circularity in NVRNet derivation or validation

full rationale

The pipeline pretrains a U-Net denoiser and transformer estimator on Hamiltonian spin simulations (with known ground-truth hyperfine values), attaches and fine-tunes adapters on real experimental Ramsey traces from each NV center, then evaluates reconstruction error and normalized FFT match strictly on held-out few-sweep experimental data. These metrics are measured on traces excluded from both pretraining and fine-tuning, so reported error reductions (0.44-0.67× noise level, FFT errors 0.10-0.19) are empirical generalization results rather than quantities forced by construction from fitted parameters. No self-definitional equations, fitted-input-renamed-as-prediction steps, load-bearing self-citations, uniqueness theorems, or ansatz smuggling appear in the abstract or described chain. The forward-reconstruction check is a standard consistency test on independent held-out data and does not collapse the extracted hyperfine values to the input measurements by definition.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim depends on the fidelity of the Hamiltonian simulations used for pretraining and on the assumption that a modest number of experimental traces suffice to adapt the model to each new NV center.

free parameters (2)
  • adapter weights
    Parameter-efficient adapters are fine-tuned on targeted experimental data for each NV center; their values are learned from data rather than derived from first principles.
  • U-Net and transformer weights
    Backbone weights are pretrained on simulated data and then adapted; the final numerical values are fitted rather than analytically fixed.
axioms (1)
  • domain assumption Hamiltonian-based spin-dynamics simulations with experimentally calibrated noise accurately capture the dominant features of real NV Ramsey signals
    The entire pretraining stage rests on this modeling assumption stated in the abstract.

pith-pipeline@v0.9.0 · 5609 in / 1640 out tokens · 58753 ms · 2026-05-15T10:50:44.117497+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

74 extracted references · 74 canonical work pages · 4 internal anchors

  1. [1]

    large box

    Lattice construction, cutoff, and 13Cstatistics Diamond supercell enumeration.We generate a finite diamond simulation lattice by explicitly enumerating a conventional-cell basis over ann×n×nsupercell. Using the lattice constanta= 3.57 ˚A, atomic coordinates are constructed from integer “mod-4” basis pointsp µ (eight sites per conventional cell) and an FCC...

    work page

  2. [2]

    In the following derivation, we treatℏ= 1 for convinience

    Rotating-wave approximation for the NV center: elimination of perpendicular hyperfine terms and rotating-frame reduction This appendix derives the effective two-level rotating- frame Hamiltonian used in the Ramsey simulator and jus- tifies neglecting perpendicular (spin-flip) hyperfine terms under our experimental conditions. In the following derivation, ...

    work page 2000

  3. [3]

    During training and inference, traces are processed in mini-batches

    Detailed architecture of the denoising network Input representation.Each Ramsey trace consists of 200 uniformly sampled time points and is treated as a single-channel one-dimensional signal. During training and inference, traces are processed in mini-batches. For a batch of input traces, the time-domain input to the denoiser is therefore represented as a ...

    work page

  4. [4]

    LetH∈ Rbatchsize×Lb×d denote the bottleneck representation written as a length-L b sequence ofd-dimensional tokens

    Multi-head self-attention used at the UNet bottlenecks At the bottleneck of each 1D UNet we apply stan- dard multi-head self-attention (MHA) [30]. LetH∈ Rbatchsize×Lb×d denote the bottleneck representation written as a length-L b sequence ofd-dimensional tokens. 20 MHA withHheads computes queries, keys, and values by learned linear projections Q=HW Q,K=HW...

    work page

  5. [5]

    Token construction, embeddings, and Transformer self-attention for the hyperfine predictor This appendix specifies the exact tokenization and Transformer operations used by the hyperfine frequency predictor (Transformer head) in Sec. III B. Per-trace normalization.For a single tracey∈R 200 we compute ey= y−µ(y) σ(y) +ϵ , µ(y) = 1 200 200X ℓ=1 yℓ, σ2(y) = ...

    work page

  6. [6]

    False-Positive Test on the denoised result To verify that the denoiser does not hallucinate Ramsey-like structure when no physical signal is present, we perform a false-positive control usingpure-noisein- puts. Specifically, we generate random traces with the 21 same length and scale as the experimental PL read- out and provide an uncertainty channel set ...

    work page

  7. [7]

    8), we provide additional qualita- tive results illustrating the consistency of the hyperfine- parameter estimator across the held-out experimental test set

    More result on Experimental Hyperfine Prediction and reconstruction To complement the representative examples shown in the main text (Fig. 8), we provide additional qualita- tive results illustrating the consistency of the hyperfine- parameter estimator across the held-out experimental test set. Figure 15 shows multiple randomly selected re- constructions...

    work page

  8. [8]

    M. W. Doherty, N. B. Manson, P. Delaney, F. Jelezko, J. Wrachtrup, and L. C. L. Hollenberg, The nitrogen- vacancy colour centre in diamond, Phys. Rep.528, 1 (2013)

    work page 2013

  9. [9]

    Jelezko and J

    F. Jelezko and J. Wrachtrup, Single defect centres in dia- mond: A review, Phys. Status Solidi A203, 3207 (2006)

    work page 2006

  10. [10]

    D. D. Awschalom, R. Hanson, J. Wrachtrup, and B. B. Zhou, Quantum technologies with optically interfaced solid-state spins, Nat. Photonics12, 516 (2018)

    work page 2018

  11. [11]

    C. L. Degen, F. Reinhard, and P. Cappellaro, Quantum sensing, Rev. Mod. Phys.89, 035002 (2017)

    work page 2017

  12. [12]

    J. R. Maze, P. L. Stanwix, J. S. Hodges, S. Hong, J. M. Taylor, P. Cappellaro, L. Jiang, M. V. G. Dutt, E. Togan, A. S. Zibrov, A. Yacoby, R. L. Walsworth, and M. D. Lukin, Nanoscale magnetic sensing with an individual electronic spin in diamond, Nature455, 644 (2008)

    work page 2008

  13. [13]

    J. M. Taylor, P. Cappellaro, L. Childress, L. Jiang, D. Budker, P. R. Hemmer, A. Yacoby, R. Walsworth, and M. D. Lukin, High-sensitivity diamond magnetome- ter with nanoscale resolution, Nat. Phys.4, 810 (2008)

    work page 2008

  14. [14]

    Maletinsky, S

    P. Maletinsky, S. Hong, M. S. Grinolds, B. Hausmann, M. D. Lukin, R. L. Walsworth, M. Lonˇ car, and A. Ya- coby, A robust scanning diamond sensor for nanoscale imaging with single nitrogen-vacancy centres, Nat. Nan- otechnol.7, 320 (2012)

    work page 2012

  15. [15]

    M. S. Grinolds, S. Hong, P. Maletinsky, L. Luan, M. D. Lukin, R. L. Walsworth, and A. Yacoby, Nanoscale mag- netic imaging of a single electron spin under ambient con- ditions, Nat. Phys.9, 215 (2013)

    work page 2013

  16. [16]

    Rondin, J.-P

    L. Rondin, J.-P. Tetienne, T. Hingant, J.-F. Roch, P. Maletinsky, and V. Jacques, Magnetometry with nitrogen-vacancy defects in diamond, Rep. Prog. Phys. 77, 056503 (2014)

    work page 2014

  17. [17]

    Tetienne, R

    J.-P. Tetienne, R. W. de Gille, D. A. Broadway, T. Teraji, S. E. Lillie, J. M. McCoey, N. Dontschuk, L. T. Hall, A. Stacey, D. A. Simpson, and L. C. L. Hollenberg, Spin properties of dense near-surface ensembles of nitrogen- vacancy centers in diamond, Phys. Rev. B97, 085402 (2018)

    work page 2018

  18. [18]

    Schirhagl, K

    R. Schirhagl, K. Chang, M. Loretz, and C. L. Degen, Nitrogen-vacancy centers in diamond: Nanoscale sensors for physics and biology, Annu. Rev. Phys. Chem.65, 83 (2014)

    work page 2014

  19. [19]

    L. P. McGuinness, Y. Yan, A. Stacey, D. A. Simp- son, L. T. Hall, D. Maclaurin, S. Prawer, P. Mul- vaney, J. Wrachtrup, F. Caruso, R. E. Scholten, and L. C. L. Hollenberg, Quantum measurement and orien- tation tracking of fluorescent nanodiamonds inside living cells, Nat. Nanotechnol.6, 358 (2011)

    work page 2011

  20. [20]

    V. V. Soshenko, S. V. Bolshedvorskii, O. Rubinas, V. N. Sorokin, A. N. Smolyaninov, V. V. Vorobyov, and A. V. Akimov, Nuclear spin gyroscope based on the nitrogen vacancy center in diamond, Physical Review Letters126, 197702 (2021)

    work page 2021

  21. [21]

    Kuan and G

    J. Kuan and G. D. Fuchs, Optical readout of coherent nuclear spins in diamond coupled to electronic spins in a thermal state, Phys. Rev. Appl.24, 064059 (2025)

    work page 2025

  22. [22]

    Jarmola, S

    A. Jarmola, S. Lourette, V. M. Acosta, A. G. Birdwell, P. Bl¨ umler, D. Budker, T. Ivanov, and V. S. Malinovsky, Demonstration of diamond nuclear spin gyroscope, Sci- ence advances7, eabl3840 (2021)

    work page 2021

  23. [23]

    Wang, M.-T

    G. Wang, M.-T. Nguyen, and P. Cappellaro, Hyperfine- enhanced gyroscope based on solid-state spins, Phys. Rev. Lett.133, 150801 (2024)

    work page 2024

  24. [24]

    Ajoy and P

    A. Ajoy and P. Cappellaro, Stable three-axis nuclear- spin gyroscope in diamond, Physical Review A—Atomic, Molecular, and Optical Physics86, 062104 (2012)

    work page 2012

  25. [25]

    Childress, M

    L. Childress, M. V. G. Dutt, J. M. Taylor, A. S. Zi- brov, F. Jelezko, J. Wrachtrup, P. R. Hemmer, and M. D. Lukin, Coherent dynamics of coupled electron and nu- clear spin qubits in diamond, Science314, 281 (2006)

    work page 2006

  26. [26]

    T. H. Taminiau, J. J. T. Wagenaar, T. van der Sar, F. Jelezko, V. V. Dobrovitski, and R. Hanson, Detection and control of individual nuclear spins using a weakly coupled electron spin, Phys. Rev. Lett.109, 137602 (2012)

    work page 2012

  27. [27]

    T. H. Taminiau, J. Cramer, T. van der Sar, V. V. Do- brovitski, and R. Hanson, Universal control and error correction in multi-qubit spin registers in diamond, Nat. Nanotechnol.9, 171 (2014)

    work page 2014

  28. [28]

    E. R. MacQuarrie, T. A. Gosavi, A. M. Moehle, N. R. Jungwirth, S. A. Bhave, and G. D. Fuchs, Coherent con- trol of a nitrogen-vacancy center spin ensemble with a diamond mechanical resonator, Optica2, 233 (2015)

    work page 2015

  29. [29]

    Ovartchaiyapong, K

    P. Ovartchaiyapong, K. W. Lee, B. A. Myers, and A. C. Bleszynski Jayich, Dynamic strain-mediated coupling of a single diamond spin to a mechanical resonator, Nat. Commun.5, 4429 (2014)

    work page 2014

  30. [30]

    Teissier, A

    J. Teissier, A. Barfuss, P. Appel, E. Neu, and P. Maletinsky, Strain coupling of a nitrogen-vacancy cen- ter spin to a diamond mechanical oscillator, Phys. Rev. Lett.113, 020503 (2014)

    work page 2014

  31. [31]

    D. A. Hopper, H. J. Shulevitz, and L. C. Bassett, Spin readout techniques of the nitrogen-vacancy center in di- amond, Micromachines9, 437 (2018)

    work page 2018

  32. [32]

    Ronneberger, P

    O. Ronneberger, P. Fischer, and T. Brox, U-net: Con- volutional networks for biomedical image segmentation, inInternational Conference on Medical image computing and computer-assisted intervention(Springer, 2015) pp. 234–241

    work page 2015

  33. [33]

    J. Ho, A. Jain, and P. Abbeel, Denoising diffusion proba- bilistic models, Advances in neural information process- ing systems33, 6840 (2020)

    work page 2020

  34. [34]

    H. Wu, Z. Zhao, and Z. Wang, Meta-unet: Multi-scale efficient transformer attention unet for fast and high- accuracy polyp segmentation, IEEE Transactions on Au- tomation Science and Engineering21, 4117 (2023)

    work page 2023

  35. [35]

    Drozdzal, E

    M. Drozdzal, E. Vorontsov, G. Chartrand, S. Kadoury, and C. Pal, The importance of skip connections in biomedical image segmentation, inInternational workshop on deep learning in medical image analysis (Springer, 2016) pp. 179–187

    work page 2016

  36. [36]

    R. Azad, M. Heidari, Y. Wu, and D. Merhof, Contextual attention network: Transformer meets u-net, inInterna- tional workshop on machine learning in medical imaging (Springer, 2022) pp. 377–386

    work page 2022

  37. [37]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, At- tention is all you need, Advances in neural information processing systems30(2017)

    work page 2017

  38. [38]

    Bahdanau, K

    D. Bahdanau, K. Cho, and Y. Bengio, Neural machine translation by jointly learning to align and translate, in 25 3rd International Conference on Learning Representa- tions, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings, edited by Y. Bengio and Y. LeCun (2015)

    work page 2015

  39. [39]

    Shih, F.-K

    S.-Y. Shih, F.-K. Sun, and H.-y. Lee, Temporal pattern attention for multivariate time series forecasting, Ma- chine Learning108, 1421 (2019)

    work page 2019

  40. [40]

    Varona-Uriarte, C

    B. Varona-Uriarte, C. Munuera-Javaloy, E. Terradillos, Y. Ban, A. Alvarez-Gila, E. Garrote, and J. Casanova, Automatic detection of nuclear spins at arbitrary mag- netic fields via signal-to-image ai model, Physical Review Letters132, 150801 (2024)

    work page 2024

  41. [41]

    K. Jung, M. Abobeih, J. Yun, G. Kim, H. Oh, A. Henry, T. Taminiau, and D. Kim, Deep learning enhanced indi- vidual nuclear-spin detection, npj Quantum Information 7, 41 (2021)

    work page 2021

  42. [42]

    N. Xu, F. Zhou, X. Ye, X. Lin, B. Chen, T. Zhang, F. Yue, B. Chen, Y. Wang, and J. Du, Noise prediction and reduction of single electron spin by deep-learning- enhanced feedforward control, Nano Letters23, 2460 (2023)

    work page 2023

  43. [43]

    Training language models to follow instructions with human feedback

    L. Ouyang, J. Wu, X. Jiang, D. Almeida, C. L. Wain- wright, P. Mishkin, C. Zhang, S. Agarwal, K. Slama, A. Ray, J. Schulman, J. Hilton, F. Kelton, L. Miller, M. Simens, A. Askell, P. Welinder, P. Christiano, J. Leike, and R. Lowe, Training language models to follow instructions with human feedback (2022), arXiv:2203.02155 [cs.CL]

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  44. [44]

    D. Guo, D. Yang, H. Zhang, J. Song, P. Wang, Q. Zhu, R. Xu, R. Zhang, S. Ma, X. Bi, X. Zhang, X. Yu, Y. Wu, Z. F. Wu, Z. Gou, Z. Shao, Z. Li, Z. Gao, A. Liu, B. Xue, B. Wang, B. Wu, B. Feng, C. Lu, C. Zhao, C. Deng, C. Ruan, D. Dai, D. Chen, D. Ji, E. Li, F. Lin, F. Dai, F. Luo, G. Hao, G. Chen, G. Li, H. Zhang, H. Xu, H. Ding, H. Gao, H. Qu, H. Li, J. Gu...

    work page 2025

  45. [45]

    Lambert, E

    N. Lambert, E. Gigu‘ere, P. Menczel, B. Li, P. Hopf, G. Su’arez, M. Gali, J. Lishman, R. Gadhvi, R. Agarwal, A. Galicia, N. Shammah, P. Nation, J. R. Johansson, S. Ahmed, S. Cross, A. Pitchford, and F. Nori, Qutip 5: The quantum toolbox in Python, Physics Reports1153, 1 (2026)

    work page 2026

  46. [46]

    Pearson, Liii

    K. Pearson, Liii. on lines and planes of closest fit to sys- tems of points in space, The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science2, 559 (1901)

    work page 1901

  47. [47]

    Hotelling, Analysis of a complex of statistical vari- ables into principal components, Journal of Educational Psychology24, 417 (1933)

    H. Hotelling, Analysis of a complex of statistical vari- ables into principal components, Journal of Educational Psychology24, 417 (1933)

    work page 1933

  48. [48]

    X. Wang, R. Girshick, A. Gupta, and K. He, Non- local neural networks, in2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2018) pp. 7794–7803

    work page 2018

  49. [49]

    Bello, B

    I. Bello, B. Zoph, A. Vaswani, J. Shlens, and Q. V. Le, Attention augmented convolutional networks, in2019 IEEE/CVF International Conference on Computer Vi- sion (ICCV)(2019)

    work page 2019

  50. [50]

    Yamamoto, E

    R. Yamamoto, E. Song, and J.-M. Kim, Parallel Wave- GAN: A fast waveform generation model based on gen- erative adversarial networks with multi-resolution spec- trogram, inProceedings of the 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)(2020) pp. 6199–6203

    work page 2020

  51. [51]

    University of Cambridge, Digital signal processing (lec- ture slides) (2024), see slide section defining DC com- ponent as the mean and AC component as signal minus mean

    work page 2024

  52. [52]

    Engel, L

    J. Engel, L. Hantrakul, C. Gu, and A. Roberts, Ddsp: Differentiable digital signal processing, inInternational Conference on Learning Representations (ICLR)(2020)

    work page 2020

  53. [53]

    Ioffe and C

    S. Ioffe and C. Szegedy, Batch normalization: Accelerat- ing deep network training by reducing internal covariate shift, inProceedings of the 32nd International Conference on Machine Learning (ICML), Proceedings of Machine Learning Research, Vol. 37 (PMLR, 2015) pp. 448–456

    work page 2015

  54. [54]

    Gaussian Error Linear Units (GELUs)

    D. Hendrycks and K. Gimpel, Gaussian error linear units (gelus), arXiv:1606.08415 (2016)

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  55. [55]

    G. E. P. Box and D. R. Cox, An analysis of transforma- tions, Journal of the Royal Statistical Society: Series B (Methodological)26, 211 (1964)

    work page 1964

  56. [56]

    W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.-Y. Fu, and A. C. Berg, Ssd: Single shot multi- box detector, inEuropean conference on computer vision (Springer, 2016) pp. 21–37

    work page 2016

  57. [57]

    Girshick, Fast r-cnn, inProceedings of the IEEE inter- national conference on computer vision(2015) pp

    R. Girshick, Fast r-cnn, inProceedings of the IEEE inter- national conference on computer vision(2015) pp. 1440– 1448

    work page 2015

  58. [58]

    P. J. Huber, Robust estimation of a location parameter, inBreakthroughs in statistics: Methodology and distribu- tion(Springer, 1992) pp. 492–518

    work page 1992

  59. [59]

    Devlin, M.-W

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, BERT: Pre-training of deep bidirectional transformers for language understanding, inProceedings of the 2019 Conference of the North American Chapter of the Asso- ciation for Computational Linguistics: Human Language Technologies (NAACL-HLT), Volume 1(Association for Computational Linguistics, 2019) pp. 4171–4186

    work page 2019

  60. [60]

    Dosovitskiy, L

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, and N. Houlsby, An im- age is worth 16x16 words: Transformers for image recog- nition at scale, inInternational Conference on Learning Representations (ICLR)(2021). 26

    work page 2021

  61. [61]

    Touvron, M

    H. Touvron, M. Cord, M. Douze, F. Massa, A. Sablay- rolles, and H. J´ egou, Training data-efficient image trans- formers & distillation through attention, inProceedings of the 38th International Conference on Machine Learn- ing (ICML), Proceedings of Machine Learning Research, Vol. 139 (PMLR, 2021) pp. 10347–10357

    work page 2021

  62. [62]

    J. Bridle, Training stochastic model recognition algo- rithms as networks can lead to maximum mutual infor- mation estimation of parameters, Advances in neural in- formation processing systems2(1989)

    work page 1989

  63. [63]

    D. H. Ackley, G. E. Hinton, and T. J. Sejnowski, A learn- ing algorithm for boltzmann machines, Cognitive science 9, 147 (1985)

    work page 1985

  64. [64]

    J. J. Hopfield, Learning algorithms and probability dis- tributions in feed-forward and feed-back networks, Pro- ceedings of the national academy of sciences84, 8429 (1987)

    work page 1987

  65. [65]

    Baum and F

    E. Baum and F. Wilczek, Supervised learning of proba- bility distributions by neural networks, inNeural infor- mation processing systems(1987)

    work page 1987

  66. [66]

    Levin and M

    E. Levin and M. Fleisher, Accelerated learning in layered neural networks, Complex systems2, 3 (1988)

    work page 1988

  67. [67]

    D. E. Rumelhart, G. E. Hinton, and R. J. Williams, Learning representations by back-propagating errors, na- ture323, 533 (1986)

    work page 1986

  68. [68]

    Decoupled Weight Decay Regularization

    I. Loshchilov and F. Hutter, Decoupled weight decay reg- ularization (2019), arXiv:1711.05101 [cs.LG]

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  69. [69]

    Abobeih, J

    M. Abobeih, J. Randall, C. Bradley, H. Bartling, M. Bakker, M. Degen, M. Markham, D. Twitchen, and T. Taminiau, Atomic-scale imaging of a 27-nuclear-spin cluster using a quantum sensor, Nature576, 411 (2019)

    work page 2019

  70. [70]

    Abragam,The Principles of Nuclear Magnetism (Clarendon Press, 1961)

    A. Abragam,The Principles of Nuclear Magnetism (Clarendon Press, 1961)

    work page 1961

  71. [71]

    K. He, X. Zhang, S. Ren, and J. Sun, Deep residual learning for image recognition, in2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2016)

    work page 2016

  72. [72]

    J. L. Ba, J. R. Kiros, and G. E. Hinton, Layer normal- ization, inNIPS 2016 Deep Learning Workshop(2016)

    work page 2016

  73. [73]

    J. Chen, Y. Lu, Q. Yu, X. Luo, E. Adeli, Y. Wang, L. Lu, A. L. Yuille, and Y. Zhou, Transunet: Transformers make strong encoders for medical image segmentation, arXiv 10.48550/arXiv.2102.04306 (2021)

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2102.04306 2021

  74. [74]

    Hatamizadeh, Y

    A. Hatamizadeh, Y. Tang, V. Nath, D. Yang, A. My- ronenko, B. Landman, H. R. Roth, and D. Xu, Unetr: Transformers for 3d medical image segmentation, in2022 IEEE/CVF Winter Conference on Applications of Com- puter Vision (WACV)(2022) pp. 1748–1758

    work page 2022