Quenched Quantum Feature Maps

Alejandro Gomez Cadavid; Anton Simen; Archismita Dalal; Carlos Flores-Garrigos; Enrique Solano; Gabriel Dario Alvarado Barrios; Jos\'e D. Mart\'in-Guerrero; Juan F. R. Hern\'andez; Murilo Henrique de Oliveira; Narendra N. Hegade

REVIEW 5 major objections 6 minor 29 references

Quenched Quantum Feature Maps

T0 review · 5 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Encoding tabular data into a disordered spin glass and quenching it for about 20 nanoseconds on a quantum annealer produces feature representations that improve classical machine-learning models by up to 210% on toxicity and medical-diagnos

desk verdict A clean empirical study of quench-based quantum feature maps on real data, but the 'quantum advantage' claim rests on an unexamined classical-hardness assumption. read the letter →

arxiv 2508.20975 v1 pith:AZF3ZXX7 submitted 2025-08-28 quant-ph

Anton Simen , Carlos Flores-Garrigos , Murilo Henrique De Oliveira , Gabriel Dario Alvarado Barrios , Juan F. R. Hernández , Qi Zhang , Alejandro Gomez Cadavid , Yolanda Vives-Gilabert

show 4 more authors

José D. Martín-Guerrero Enrique Solano Narendra N. Hegade Archismita Dalal

This is my paper

classification quant-ph

keywords quantumfeaturemapsquenchdynamicsspinglassesannealersmachinelearningtabularclassificationadvantageanalogcomputing

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper sets out to show that a quantum feature map made by encoding tabular data into a spin-glass Hamiltonian and quenching the system for about 20 nanoseconds on a quantum annealer yields feature representations that classical machine-learning models cannot match using the original features. The technique measures single-qubit expectation values right after the fast, nonadiabatic evolution and feeds those as new features into standard classifiers. On molecular toxicity, myocardial-infarction, and drug-safety benchmarks, every tested metric improves, with relative gains up to 210% and the best results at anneal times of 20–30 ns, near the critical point. If the claim holds, this is a practical demonstration of quantum-derived advantage in machine learning on existing hardware.

What carries the argument

The central object is the analog quantum feature map produced by a quench of the transverse-field Ising Hamiltonian H(s) = −A(s) Σ σ^x_i + B(s) H(x). A classical feature vector x is embedded in the longitudinal fields and couplings of H(x); instead of waiting for adiabatic ground-state preparation, the system is driven quickly so that nonadiabatic excitations entangle the state with the full instantaneous spectrum. The new features are the single-qubit expectation values ⟨ψ_f(x)|σ^z_i|ψ_f(x)⟩. This map carries the argument: the structure of the final state is claimed to be classically intractable, so the resulting features are claimed to contain patterns that no classical feature map of the

What would settle it

Run the same quench in classical simulation (tensor-network time evolution or Schrödinger evolution with sparse observables) for the 111-to-200-qubit instances at 20 ns and train models on the simulated expectation values; if those match or beat the hardware features, the advantage claim collapses. Alternatively, feed a classical nonlinear feature map of the same dimension, such as random Fourier features, into the same classifiers and require the quantum features to beat it.

Watch

Extended reading notes

Core claim

The central claim is that the finite-time, nonadiabatic quench of a disordered Ising spin glass encodes classically hard correlations of the input data into single-qubit expectation values, so that replacing a data row x by the vector of measured expectation values improves downstream supervised learning. The authors encode each row into the longitudinal fields and couplings of a problem Hamiltonian H(x), run a fast transverse-field quench, and use the measured expectation values as features. Across three tabular datasets requiring 111 to 200 qubits, the quantum-enhanced models beat the same classical architectures on accuracy, precision, recall, F1, and AUC, with gains as large as 210% and

Load-bearing premise

The load-bearing premise is that the 20-nanosecond quench of a 111-to-200-qubit spin glass yields single-qubit expectation values that classical computers cannot reproduce at comparable cost, so the measured gains come from genuine quantum complexity; the paper imports this intractability from earlier beyond-classical simulations but never checks classically whether the same features could be generated.

Editorial extensions

If this is right

Quantum-enhanced features can be dropped directly into off-the-shelf classifiers; identical model architectures improve on all tested metrics, including a 40%+ accuracy gain on molecular toxicity and recall gains above 100% for atrial fibrillation prediction.
Operating the annealer in the 20–30 ns coherent regime, near the critical point, is essential; longer anneals (100 ns) lose the advantage because decoherence and thermalization erase the useful complexity.
The same feature-extraction recipe should transfer to other quantum platforms—neutral atoms, superconducting circuits, or bosonic sampling—wherever fast nonadiabatic quenches can be run.
For medical and drug-safety tasks, the recall improvement implies more true positives and fewer missed cases, i.e., earlier intervention or safer screening.
Higher-order multi-qubit expectations, adaptive schedules, and digital implementations are listed by the authors as immediate extensions that could scale the method beyond 200 qubits.

Reading between the lines

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

The paper does not check whether classical simulation of the 20 ns quench, or a classical nonlinear feature map such as random Fourier features, could reproduce the same expectation values and gains; until that is ruled out, the quantum-advantage-level claim rests on the assumed intractability of the simulation, not on a direct comparison.
Because all demonstrations use one annealer family, part of the gain could be idiosyncratic to that platform's noise and calibration rather than a generic property of spin-glass quenches; this is testable on another programmable quantum simulator.
If the improvement scales with qubit count, the method could become a practical hybrid workflow: quantum hardware generates a fixed feature layer for tabular data, and classical models do the rest.
The 210% figure is a relative gain on one metric and model comparison; the paper's own tables show smaller gains on other tasks, so the headline should be read as the best-case relative improvement, not the typical one.

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.

Desk Editor's Note

read the letter

Colleague,

Here's my read of arXiv:2508.20975. The paper is a straightforward, well-presented experiment: encode tabular data into Ising couplings, quench on a D-Wave annealer for 10–100 ns, read out single-qubit expectation values, and feed those as extra features to classical ML models. That specific protocol—fast non-adiabatic quench near the critical point as a feature map, with anneal-time sweeps—is a genuine novelty, and the empirical sweep across times (Fig. 3) is a useful addition to the reservoir-computing literature. The consistent gains over raw-feature baselines across three datasets and four ML models are credible as an internal comparison; the 10x5 stratified CV with median reporting is a sound methodology.

The soft spot is the central claim. The paper labels these features "quantum-advantage-level" and says they "cannot be efficiently reproduced by a classical method," but never tests that. The only support is Ref. [8], a 5,000-qubit beyond-classical sampling/simulation result; that says nothing about whether single-qubit marginals of a 111–200 qubit spin glass at 20 ns are hard to approximate classically. Tensor networks and mean-field methods can sometimes do exactly that. The paper reports no classical simulation of the same quench, and no comparison with classical nonlinear feature maps like polynomial expansions, random kitchen sinks, or random Fourier features. Without those baselines, the improvements are equally consistent with the hardware implementing an unspecified (and possibly classically reproducible) nonlinear transformation.

Two more issues are worth flagging, though they are secondary. Feature selection via mutual information is done on the full dataset before cross-validation, which leaks information. And the optimal anneal time (20 ns) is chosen by looking at the same CV performance that is later quoted as the headline; that is a form of post-selection, even if the CV curve is reported. The "210%" improvement in the abstract is a recall change on a test set of 40 samples, and no significance tests are reported anywhere.

So my verdict: the empirical phenomenon is probably real—the quantum features do help these classical models. But the paper does not establish that the help comes from quantum intractability. That's a load-bearing gap, not a minor omission. The fix is feasible: simulate the quench classically at these sizes (or approximate it), compare against classical nonlinear feature maps, and perform nested feature selection and significance tests. If the gains survive, that's a strong result; if not, it's a nice engineering demonstration.

This deserves a serious referee—the protocol is new, the data is real, and the question of classical baselines is exactly what peer review should sort out. I'd send it to a competent QML referee with instructions to demand the classical baselines. I'd bring it to a reading group as a case study in how quantum-advantage claims can outrun their evidence.

Best.

Referee Report

5 major / 6 minor

Summary. The paper proposes an analog quantum feature map (AQFM) in which tabular data are encoded into the fields and couplings of an Ising spin-glass Hamiltonian, evolved on a D-Wave quantum annealer in a fast, nonadiabatic quench, and then represented by measured single-qubit expectation values. These quantum-generated features are fed into standard classical ML models (SVM, random forest, gradient boosting, XGBoost). The authors benchmark on three datasets (molecular toxicity, myocardial infarction complications, drug-induced autoimmunity) and report consistent improvements over models trained on raw features, with the best performance at annealing times around 20–30 ns. The paper claims to present 'the first quantum ML demonstrations at the quantum-advantage level.'

Significance. If the central claim were established, this would be a practically important result: a hardware-implementable quantum feature map on existing annealers that improves real-world tabular classification, with potential impact in drug discovery and medical diagnostics. The paper has notable strengths: it uses real quantum hardware, includes several datasets and four classical base models, and includes an anneal-time dependence study. However, the quantum-advantage claim is not supported by the evidence presented. The paper lacks a classical simulation of the same quench, lacks classical nonlinear feature-map baselines, and does not provide uncertainty quantification or significance tests. These omissions are load-bearing because the observed gains may be reproducible by generic nonlinear feature transformations.

major comments (5)

[II, Eq. (2)] The claim that the feature map in Eq. (2) 'cannot be efficiently reproduced by a classical method' is asserted, not demonstrated. The cited Ref. [8] concerns 5,000-qubit beyond-classical simulation of a global state; it does not imply that single-qubit expectation values after a 20 ns 111–200 qubit quench are classically intractable. Such marginals may be estimable by tensor networks, mean-field methods, or other classical techniques. No classical simulation of the same quench is reported, so the central hardness premise is unsupported.
[IV, Tables I–IV] The improvements are measured only against models trained on raw features. There is no comparison with classical nonlinear feature maps such as random kitchen sinks, random Fourier features, polynomial expansions, or kernel PCA. The observed gains could therefore be produced by any nonlinear feature transformation. This alternative explanation must be excluded before the results can be attributed to quantum dynamics or quantum intractability.
[III and IV] The evaluation protocol has selection leakage. Mutual-information feature selection is applied to the full dataset before cross-validation, and the annealing time is chosen as 20 ns after inspecting Fig. 3 on the same dataset. The reported median cross-validation scores are therefore optimistically biased and do not provide an unbiased comparison. A nested cross-validation or a genuinely held-out test set is needed.
[IV, Tables II–IV] The text states 'clear and consistent improvements ... across all evaluation metrics,' but Tables II–IV contain counterexamples. In Table III, AQFM-GB recall is 0.35 vs. GB 0.51 and AQFM-RF accuracy/F1 are lower than RF; in Table IV, AQFM-XGB recall is 0.12 vs. 0.14. Only medians are reported, with no confidence intervals or significance tests for the 50 CV scores. The abstract's 'up to 210%' is also not derivable from the tables; the largest shown accuracy increase is roughly 39%.
[IV, Fig. 4] The DIA/Huk result is reported as a single test-set performance without error bars or repeated cross-validation, and the text states that the Huk solver 'provides the performance metrics from a single, given test set.' With no uncertainty quantification, the claimed gains in precision (8.8%) and AUC (7.5%) cannot be assessed. This is a weaker evaluation standard than the rest of the paper.

minor comments (6)

[Fig. 1 caption] 'A schematics' should be 'A schematic'.
[Fig. 2 caption] 'embbeded' should be 'embedded'.
[II] In Eq. (2), O_i is used before being defined. Please specify that O_i denotes the observable measured on qubit i (e.g., sigma_z^i) and clarify whether multi-qubit observables are also included.
[References] Reference [22] is a duplicate of Ref. [5] (Havlíček et al., Nature 2019). Please merge or distinguish them.
[II] The sentence about K(x_i, x_j) 'may not find overlaps' is unclear; explain what is meant by 'find overlaps' in the context of large final states.
[V] The phrase 'quantum-advantage level' in the conclusion repeats an unsupported claim. The empirical scope supports 'observable improvement over raw features on these datasets' but not 'quantum advantage.'

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the feature map is an empirical black-box transformation, and the quantum-advantage claim rests on an external result, not on a self-referential derivation.

full rationale

The paper's derivation chain is empirical rather than definitional. The quantum feature map (Eq. 2) is defined as single-qubit expectation values after a quench of the Ising Hamiltonian (Eq. 1); the ML models then train on these features and are compared with classical baselines on external UCI datasets. No equation defines the features in terms of the reported performance metrics, and no fitted parameter is renamed as a prediction. The choice of 20 ns anneal time is selected from cross-validated performance in Fig. 3, which is standard hyperparameter selection, not a construction that forces the later tables. The central 'quantum-advantage-level' claim is imported from Ref. [8] (King et al.), an external 5,000-qubit beyond-classical study, rather than from the authors' own prior results; whether that intractability transfers to 111-200 qubit single-qubit marginals is an evidentiary gap, not circularity. Self-citations such as the Huk press release [23] and the authors' earlier circuit papers [14,16] are peripheral and not load-bearing for the derivation. Therefore no circular step can be exhibited with the required specificity, and the appropriate score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

No new physical entities are postulated. The free parameters are data-dependent choices, and the main unexamined assumption is the transfer of beyond-classical claims to smaller, differently connected systems.

free parameters (3)

anneal_time = 20 ns (scanned 10-40 ns and 100 ns)
Anneal time is selected based on cross-validated performance on the molecular toxicity dataset (Section IV) and then fixed for the other datasets; optimality is therefore fitted to the same data.
mutual_information_feature_selection_thresholds = top 200 features for toxicity; MI > 0.005 for DIA
Feature filters are chosen by inspecting the target variable before cross-validation (Section III), so the thresholds are hand-fitted to the benchmark.
data-to-Hamiltonian encoding scale = unspecified in text
No formula is given for mapping normalized features to h_i and J_ij; this choice is a hand-tuned modeling parameter and prevents replication.

assumptions (4)

domain assumption D-Wave annealer realizes coherent transverse-field Ising quench dynamics for 10-40 ns as assumed
The entire feature extraction relies on the hardware evolving under Eq. (1) with negligible unwanted noise; no calibration or validation of the actual dynamics is provided.
domain assumption Beyond-classical result for 5,000-qubit spin-glass simulation applies to these smaller, sparse, logically embedded instances
Introduction and Section II import [8] to claim classically intractable features, but the systems here have 111-200 qubits and different connectivity and are run at D-Wave's fast regime.
ad hoc to paper Measured single-qubit expectation values retain the classically hard information of the final state
The paper projects the state down to n marginal expectation values (Eq. (2)) but gives no argument that these low-order marginals remain hard to compute classically.
standard math Standard Schrodinger evolution / transverse-field Ising model
Equation (1) is treated as the governing Hamiltonian, relying on standard quantum mechanics.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Quenched Quantum Feature Maps." pith.science (2026). https://pith.science/paper/AZF3ZXX7

@misc{pith2026250820975,
  author       = {Pith},
  title        = {Pith review of: Quenched Quantum Feature Maps},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AZF3ZXX7}},
  note         = {Machine review of arXiv:2508.20975}
}

read the original abstract

We propose a quantum feature mapping technique that leverages the quench dynamics of a quantum spin glass to extract complex data patterns at the quantum-advantage level for academic and industrial applications. We demonstrate that encoding a dataset information into disordered quantum many-body spin-glass problems, followed by a nonadiabatic evolution and feature extraction via measurements of expectation values, significantly enhances machine learning (ML) models. By analyzing the performance of our protocol over a range of evolution times, we empirically show that ML models benefit most from feature representations obtained in the fast coherent regime of a quantum annealer, particularly near the critical point of the quantum dynamics. We demonstrate the generalization of our technique by benchmarking on multiple high-dimensional datasets, involving over a hundred features, in applications including drug discovery and medical diagnostics. Moreover, we compare against a comprehensive suite of state-of-the-art classical ML models and show that our quantum feature maps can enhance the performance metrics of the baseline classical models up to 210%. Our work presents the first quantum ML demonstrations at the quantum-advantage level, bridging the gap between quantum supremacy and useful real-world academic and industrial applications.

Figures

Figures reproduced from arXiv: 2508.20975 by the authors.

**Figure 1.** FIG. 1. A schematics of analog quantum feature mapping for ML. In the step of quantum feature mapping, (a) the information of a tabular [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

**Figure 2.** FIG. 2. Logical problems for both molecular toxicity and myocar [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** FIG. 3. Cross-validation performance comparison across multiple [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: FIG. 4. Performance metrics on the DIA dataset using our Huk ser [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

29 extracted references · 24 canonical work pages

[8]

Beyond- classical computation in quantum simulation

Andrew D King, Alberto Nocera, Marek M Rams, Jacek Dziar- maga, Roeland Wiersema, William Bernoudy, Jack Raymond, Nitin Kaushal, Niclas Heinsdorf, Richard Harris, et al. Beyond- classical computation in quantum simulation. Science, page eado6285, 2025

work page 2025
[1]

Quan- tum algorithm for linear systems of equations

Aram W Harrow, Avinatan Hassidim, and Seth Lloyd. Quan- tum algorithm for linear systems of equations. Physical review letters, 103(15):150502, 2009

work page 2009
[2]

Quan- tum support vector machine for big data classification

Patrick Rebentrost, Masoud Mohseni, and Seth Lloyd. Quan- tum support vector machine for big data classification. Phys. Rev. Lett., 113(13):130503, 2014

work page 2014
[3]

Bardin, Rami Barends, Rupak Biswas, and et al

Frank Arute, Kunal Arya, Ryan Babbush, Dave Bacon, Joseph C. Bardin, Rami Barends, Rupak Biswas, and et al. Quantum supremacy using a programmable superconducting processor. Nature, 574(7779):505–510, 2019

work page 2019
[4]

Han-Sen Zhong, Han Wang, Yu-Hao Deng, Ming-Cheng Chen, L. C. Peng, Yu-Heng Luo, Jian Qin, Di Wu, Yi Hu, and et al. Quantum computational advantage using photons. Science, 370(6523):1460–1463, 2020

work page 2020
[5]

C ´orcoles, Kirsten Temme, Aram W

V ojtˇech Havl ´ıˇcek, Antonio D. C ´orcoles, Kirsten Temme, Aram W. Harrow, Abhinav Kandala, Jerry M. Chow, and Jay M. Gambetta. Supervised learning with quantum-enhanced feature spaces. Nature, 567(7747):209–212, 2019. 6

work page 2019
[6]

Quantum machine learn- ing in feature hilbert spaces

Maria Schuld and Nathan Killoran. Quantum machine learn- ing in feature hilbert spaces. Physical Review Letters , 122(4):040504, 2019

work page 2019
[7]

Is quantum advantage the right goal for quantum machine learning? Prx Quantum, 3(3):030101, 2022

Maria Schuld and Nathan Killoran. Is quantum advantage the right goal for quantum machine learning? Prx Quantum, 3(3):030101, 2022

work page 2022

Show all 29 references

[9]

Quantum computation by adiabatic evolution, 2000

Edward Farhi, Je ffrey Goldstone, Sam Gutmann, and Michael Sipser. Quantum computation by adiabatic evolution, 2000

2000
[10]

Tameem Albash and Daniel A. Lidar. Adiabatic quantum com- putation. Rev. Mod. Phys., 90:015002, Jan 2018

2018
[11]

Transitionless quantum driving

M V Berry. Transitionless quantum driving. Journal of Physics A: Mathematical and Theoretical, 42(36):365303, aug 2009

2009
[12]

King, Jack Raymond, Trevor Lanting, Richard Har- ris, Alex Zucca, Fabio Altomare, Andrew J

Andrew D. King, Jack Raymond, Trevor Lanting, Richard Har- ris, Alex Zucca, Fabio Altomare, Andrew J. Berkley, Kelly Boothby, Sara Ejtemaee, Colin Enderud, Emile Hoskinson, Shuiyuan Huang, Eric Ladizinsky, Allison J. ˜R. MacDonald, Gaelen Marsden, Reza Molavi, Travis Oh, Gabri...

2023
[13]

Svore, and Nathan Wiebe

Maria Schuld, Alex Bocharov, Krysta M. Svore, and Nathan Wiebe. Circuit-centric quantum classifiers. Physical Review A, 101(3):032308, 2020

2020
[14]

Evolutionary-enhanced quantum supervised learning model

Anton Simen, Rodrigo Bloot, Otto M Pires, and Erick G Sperandio Nascimento. Evolutionary-enhanced quantum supervised learning model. The Journal of Supercomputing , 81(2):388, 2025

2025
[15]

Iris Cong, Soonwon Choi, and Mikhail D. Lukin. Quantum convolutional neural networks. Nature Physics, 15(12):1273– 1278, 2019

2019
[16]

Digital-analog quantum convolutional neural networks for im- age classification

Anton Simen, Carlos Flores-Garrigos, Narendra N Hegade, Iraitz Montalban, Yolanda Vives-Gilabert, Eric Michon, Qi Zhang, Enrique Solano, and Jos ´e D Mart ´ın-Guerrero. Digital-analog quantum convolutional neural networks for im- age classification. Physical Review Research , ...

2024
[17]

Coherent quantum annealing in a programmable 2,000 qubit ising chain

Andrew D King, Sei Suzuki, Jack Raymond, Alex Zucca, Trevor Lanting, Fabio Altomare, Andrew J Berkley, Sara Ejtemaee, Emile Hoskinson, Shuiyuan Huang, et al. Coherent quantum annealing in a programmable 2,000 qubit ising chain. Nature Physics, 18(11):1324–1328, 2022

2022
[18]

Quantum criti- cal dynamics in a 5,000-qubit programmable spin glass.Nature, 617(7959):61–66, 2023

Andrew D King, Jack Raymond, Trevor Lanting, Richard Har- ris, Alex Zucca, Fabio Altomare, Andrew J Berkley, Kelly Boothby, Sara Ejtemaee, Colin Enderud, et al. Quantum criti- cal dynamics in a 5,000-qubit programmable spin glass.Nature, 617(7959):61–66, 2023

2023
[19]

Large-scale quantum reservoir computing using a gaussian boson sampler

Valeria Cimini, Mandar M Sohoni, Federico Presutti, Ben- jamin K Malia, Shi-Yuan Ma, Ryotatsu Yanagimoto, Tianyu Wang, Tatsuhiro Onodera, Logan G Wright, and Peter L McMa- hon. Large-scale quantum reservoir computing using a gaussian boson sampler. arXiv preprint arXiv:2505.13...

2025 arXiv
[20]

Large-scale quantum reservoir learning with an analog quantum computer

Milan Kornja ˇca, Hong-Ye Hu, Chen Zhao, Jonathan Wurtz, Phillip Weinberg, Majd Hamdan, Andrii Zhdanov, Sergio H Cantu, Hengyun Zhou, Rodrigo Araiza Bravo, et al. Large-scale quantum reservoir learning with an analog quantum computer. arXiv preprint arXiv:2407.02553, 2024

2024 arXiv
[21]

Ground state-based quantum feature maps

Chukwudubem Umeano and Oleksandr Kyriienko. Ground state-based quantum feature maps. arXiv preprint arXiv:2404.07174, 2024

2024 arXiv
[22]

Supervised learning with quantum-enhanced feature spaces

V ojtˇech Havl ´ıˇcek, Antonio D C ´orcoles, Kristan Temme, Aram W Harrow, Abhinav Kandala, Jerry M Chow, and Jay M Gambetta. Supervised learning with quantum-enhanced feature spaces. Nature, 567(7747):209–212, 2019

2019
[23]

Kipu quantum unveils the first industrial quan- tum machine learning service working at quantum advantage level, 2025

Kipu Quantum. Kipu quantum unveils the first industrial quan- tum machine learning service working at quantum advantage level, 2025. HUK press release

2025
[24]

Kipu quantum hub, 2025

Kipu Quantum. Kipu quantum hub, 2025. Kipu Quantum Hub

2025
[25]

Toxicity

S ¸eref G¨ul and FATIH RAHIM. Toxicity. UCI Machine Learn- ing Repository, 2021. DOI: https://doi.org/10.24432/C59313

2021 doi
[26]

Golovenkin, V .A

S.E. Golovenkin, V .A. Shulman, D.A. Rossiev, P.A. Shesternya, S.Yu. Nikulina, Yu.V . Orlova, and V .F. V oino-Yasenetsky. My- ocardial infarction complications. UCI Machine Learning Repository, 2020. DOI: https://doi.org/10.24432/C53P5M

2020 doi
[27]

Interdia: Inter- pretable prediction of drug-induced autoimmunity through en- semble machine learning approaches

Lina Huang, Peineng Liu, and Xiaojie Huang. Interdia: Inter- pretable prediction of drug-induced autoimmunity through en- semble machine learning approaches. Toxicology, 511:154064, 2025

2025
[28]

Digitized counterdiabatic quantum critical dy- namics

Anne-Maria Visuri, Alejandro Gomez Cadavid, Balaganchi A Bhargava, Sebasti ´an V Romero, Andr ´as Grabarits, Pranav Chandarana, Enrique Solano, Adolfo del Campo, and Naren- dra N Hegade. Digitized counterdiabatic quantum critical dy- namics. arXiv preprint arXiv:2502.15100, 2025

2025
[29]

Universal defect statistics in counterdiabatic quantum critical dynamics

Andr ´as Grabarits and Adolfo del Campo. Universal defect statistics in counterdiabatic quantum critical dynamics. arXiv preprint arXiv:2503.22212, 2025

2025 arXiv

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [8]

Beyond- classical computation in quantum simulation

Andrew D King, Alberto Nocera, Marek M Rams, Jacek Dziar- maga, Roeland Wiersema, William Bernoudy, Jack Raymond, Nitin Kaushal, Niclas Heinsdorf, Richard Harris, et al. Beyond- classical computation in quantum simulation. Science, page eado6285, 2025

work page 2025

[2] [1]

Quan- tum algorithm for linear systems of equations

Aram W Harrow, Avinatan Hassidim, and Seth Lloyd. Quan- tum algorithm for linear systems of equations. Physical review letters, 103(15):150502, 2009

work page 2009

[3] [2]

Quan- tum support vector machine for big data classification

Patrick Rebentrost, Masoud Mohseni, and Seth Lloyd. Quan- tum support vector machine for big data classification. Phys. Rev. Lett., 113(13):130503, 2014

work page 2014

[4] [3]

Bardin, Rami Barends, Rupak Biswas, and et al

Frank Arute, Kunal Arya, Ryan Babbush, Dave Bacon, Joseph C. Bardin, Rami Barends, Rupak Biswas, and et al. Quantum supremacy using a programmable superconducting processor. Nature, 574(7779):505–510, 2019

work page 2019

[5] [4]

Han-Sen Zhong, Han Wang, Yu-Hao Deng, Ming-Cheng Chen, L. C. Peng, Yu-Heng Luo, Jian Qin, Di Wu, Yi Hu, and et al. Quantum computational advantage using photons. Science, 370(6523):1460–1463, 2020

work page 2020

[6] [5]

C ´orcoles, Kirsten Temme, Aram W

V ojtˇech Havl ´ıˇcek, Antonio D. C ´orcoles, Kirsten Temme, Aram W. Harrow, Abhinav Kandala, Jerry M. Chow, and Jay M. Gambetta. Supervised learning with quantum-enhanced feature spaces. Nature, 567(7747):209–212, 2019. 6

work page 2019

[7] [6]

Quantum machine learn- ing in feature hilbert spaces

Maria Schuld and Nathan Killoran. Quantum machine learn- ing in feature hilbert spaces. Physical Review Letters , 122(4):040504, 2019

work page 2019

[8] [7]

Is quantum advantage the right goal for quantum machine learning? Prx Quantum, 3(3):030101, 2022

Maria Schuld and Nathan Killoran. Is quantum advantage the right goal for quantum machine learning? Prx Quantum, 3(3):030101, 2022

work page 2022