Pith. sign in

REVIEW 4 major objections 5 minor 91 references

Position: Quantum Kernel Machines Should Move Beyond Scalar-Valued Kernels to Realize Their Potential

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Quantum kernel machines should move beyond scalar-valued kernels to operator-valued kernels, whose entangled form can exploit non-factorizable input-output structure.

desk verdict The paper's QOVK construction is new and the circuit is concrete, but the missing psd proof and the lack of a classical OVK baseline leave the quantum advantage claim unsupported. read the letter →

arxiv 2506.03779 v2 pith:CZRKHOPV submitted 2025-06-04 quant-ph cs.LGstat.ML

classification quant-phcs.LGstat.ML
keywords quantumkernelmethodsoperator-valuedkernelsentanglementstructuredpredictionchannelestimationridgeregressionC*-algebraicmachinelearning
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

Quantum kernel research has stalled, in this paper's telling, because it has been testing scalar-valued kernels on standard classification and regression problems where classical kernel methods already excel. The paper argues that the field should move to operator-valued kernels, which return matrices rather than numbers and can represent structured outputs together with their internal dependencies. Its central proposal is the entangled quantum operator-valued kernel (QOVK), built from a non-separable unitary that couples input and output registers, allowing output correlations to depend on inputs in a joint, non-factorizable way. A proof-of-concept on quantum channel estimation shows the QOVK recovering bit-flip and dephasing channels far more accurately than a scalar-valued quantum kernel (Frobenius errors 0.101 and 0.096 versus 0.569 and 0.525). If this view is right, the meaningful testbed for quantum advantage in kernel machines is structured prediction, not scalar classification.

What carries the argument

The central object is the entangled quantum operator-valued kernel of Definition 3.1, $K(x,z)=\operatorname{Tr}_X[U_{YX}(\rho_Y\otimes\sigma^X_{x,z})U_{YX}^\dagger]$, in which $U_{YX}$ is a unitary that cannot be written as $A_Y\otimes B_X$, $\rho_Y$ is a density matrix on the output space, and $\sigma^X_{x,z}$ is a feature matrix extracted from inputs $x$ and $z$. The non-separable unitary carries the argument: it couples input and output subsystems so that correlations among outputs can depend on the input representation, a property separable kernels cannot express. Around this object the paper builds two supporting mechanisms: a swap-test quantum circuit that produces the feature matrix and evaluates the kernel on a quantum device, and a C*-algebraic generalization via reproducing kernel Hilbert C*-modules, in which kernel outputs can themselves be quantum operators such as gates and density matrices.

What would settle it

Run the same bit-flip and dephasing channel estimation task with a classical operator-valued kernel (for example a separable OVK with an output kernel learned by cross-validation) under identical training data, noise, and evaluation protocol; if it achieves errors near the QOVK's 0.101 and 0.096, the case for a quantum-specific entangled advantage collapses. A second check is to replace the hand-fixed CNOT+SWAP unitary $U$ with a randomly chosen non-separable unitary: if recovery errors jump back to QSVK levels, the result depends on the specific $U$ rather than the entangled class.

Watch

Extended reading notes

Core claim

The paper claims that scalar-valued quantum kernels lack the degrees of freedom needed to exploit intrinsically quantum resources, and that realizing the potential of quantum kernel machines requires operator-valued kernels. Its definition of an entangled QOVK is $K(x,z)=\operatorname{Tr}_X[U_{YX}(\rho_Y\otimes\sigma^X_{x,z})U_{YX}^\dagger]$, where $U_{YX}$ is a non-separable unitary, $\rho_Y$ is an output density matrix, and $\sigma^X_{x,z}$ is an input feature matrix; the non-separability is what lets the kernel represent input-dependent output correlations rather than factorizing them. The paper shows that quantum scalar-valued kernels, including fidelity kernels, are recovered as special separable cases of this construction. It also gives a swap-test quantum circuit that prepares the feature matrix and evaluates the kernel, and reports a proof-of-concept where kernel ridge regression with an entangled QOVK estimates single-qubit bit-flip and dephasing channels with much lower Frobenius error than a QSVK (0.101 and 0.096 versus 0.569 and 0.525). The intended contribution is a road map: develop entangled QOVKs, bring in C*-algebraic reproducing kernel Hilbert C*-modules, and concentrate applications on structured prediction.

Load-bearing premise

The load-bearing premise is that the dramatic improvement in the channel-estimation experiment comes from the entangled quantum operator-valued kernel itself, yet the experiment compares QOVK only against a scalar-valued quantum kernel, so the gain could in principle come from the operator-valued output structure alone and be matched by a classical operator-valued kernel.

Editorial extensions

If this is right

  • Quantum kernel machines gain a natural application regime, structured prediction, where outputs carry dependencies that scalar kernels cannot encode, so quantum advantages can be assessed where classical methods are known to struggle.
  • Entanglement becomes a concrete design resource in kernels: non-separable unitaries generate kernel classes strictly larger than fidelity and separable kernels, so experiments can test whether entanglement improves learning.
  • The operator-valued kernel matrix grows from $n\times n$ to $np\times np$, giving quantum linear-system solvers and amplitude-amplification routines a larger matrix dimension on which their polylogarithmic scaling could, in principle, translate into a practical edge.
  • The C*-algebraic RKHM route lets kernel outputs be quantum objects such as density operators and gates, which would make tasks like quantum channel learning native to the kernel framework.
  • Quantum channel estimation is proposed as a first benchmark where the QOVK clearly separates from the QSVK, giving the community a concrete task for further study.

Reading between the lines

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

  • Beyond the paper: a classical operator-valued kernel with a good output kernel may reproduce the channel-estimation gains, which would show the advantage comes from operator-valued outputs rather than quantum entanglement; the paper does not test this control.
  • Beyond the paper: because the unitary $U$ is fixed by hand as CNOT+SWAP, the result may depend on matching $U$ to the channel structure; learning $U$ from data or averaging over random non-separable unitaries would clarify whether the entangled class, not a single gate, is responsible.
  • Beyond the paper: the operator-valued setting could connect quantum kernel methods to quantum neural tangent kernels and variational feature learning, since learned unitaries on a coupled input-output system would interpolate between kernel machines and trained quantum models; the authors only gesture at learned feature maps.
  • Beyond the paper: if the QOVK advantage persists under a classical OVK control, the framework suggests a concrete scaling test, increase output dimension $p$ and channel complexity and ask whether the gap between entangled QOVK and classical OVK grows, which would indicate where quantum resources start to matter.
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

4 major / 5 minor

Summary. This position paper argues that research on quantum kernel machines should move from scalar-valued quantum kernels (QSVKs) to operator-valued quantum kernels (QOVKs), which can encode non-factorizable input-output interactions and exploit entanglement. The paper defines an entangled QOVK in Definition 3.1 via a partial trace over a dilated input-output system with a non-separable unitary U, sketches a quantum circuit implementation in Appendix D, and reports a proof-of-concept on quantum channel estimation in Section 5.1 showing that a QOVK recovers bit-flip and dephasing channels with much lower Frobenius error than a QSVK (Table 1). The paper also proposes a roadmap with four actions: quantum implementations of OVKs, entangled QOVKs, C*-algebraic kernels, and quantum structured prediction.

Significance. If the central construction can be made rigorous, this is a timely and potentially influential proposal: it reframes the quantum kernel debate around structured prediction and operator-valued outputs, where classical kernel methods are less dominant. The paper has several strengths: Definition 3.1 is explicit, the circuit derivation in Appendix D is concrete, the proof-of-concept is a genuine simulation without fitted constants, and the authors acknowledge that entanglement alone does not guarantee advantage. However, the current manuscript does not establish that the proposed object is a valid kernel in the operator-valued sense, and the empirical support does not isolate a quantum or entanglement-specific effect. These issues are load-bearing because the proof-of-concept is the main quantitative evidence for the position.

major comments (4)
  1. [Definition 3.1, Eq. (1), Appendix A, Section 5.2] The paper does not establish that the operator-valued function in Eq. (1) is a positive semi-definite operator-valued kernel. Appendix A (Definition A.1 and Theorem A.3) requires K(x,z)=K(z,x)* and block-wise positive semi-definiteness for a reproducing kernel to exist, but Definition 3.1 imposes only non-separability of U and makes no assumptions on the feature matrices sigma^{x,z}_X. Section 5.2 explicitly states that sigma^{x,z}_X is not required to be Hermitian and that the experiments use sigma^{i,j}_X = sigma_i sigma_j, which are generally non-Hermitian. For n=1, if sigma^{x,x}_X = -I, then K(x,x) = -Tr_X[U(rho_Y tensor I)U^dag] is negative semi-definite, so the proposed class is not psd without additional assumptions. Even for the specific product form used in Table 1, no psd proof is supplied; if the assembled np x np Gram matrix has negative eigenvalues, the kernel ridge regression solution reported in Section 5.1 is not a valid kernel estimate. The authors should either restrict Definition 3.1 to feature matrices satisfying a block-psd condition or prove psd for the constructions used in the experiments and the circuit.
  2. [Section 5.1, Table 1, Appendix B] The experiment compares the QOVK only against a scalar-valued quantum kernel (QSVK), so the observed improvement does not identify the source of the gain. The QOVK differs from the QSVK in two ways: it uses operator-valued outputs and it uses a non-separable unitary U. A classical operator-valued kernel with a suitable output kernel could plausibly achieve comparable or better recovery on this matrix-valued regression task, which would weaken the paper's central claim that the advantage comes from quantum operator-valued structure or entanglement. To support the position, the authors should add a classical OVK baseline (for example, k(sigma_i, sigma_j) L with a tuned positive definite output kernel L) and, ideally, a separable quantum OVK with the same input feature matrix, so that the effect of operator-valued outputs can be separated from the effect of the non-separable U.
  3. [Section 5.1 vs Appendix D, Eqs. (8) and (10)] There is an inconsistency between the kernel used in the experiments and the kernel for which a quantum circuit is provided. Appendix D derives a specific feature matrix sigma^{x,z}_X = (rho_x + rho_z + <psi_x|psi_z> |psi_x><psi_z| + <psi_z|psi_x> |psi_z><psi_x|) / (2(1+| <psi_z|psi_x> |^2)) in Eq. (8), which is Hermitian and positive semi-definite. Section 5.1, however, states that the experiments use sigma^{i,j}_X = sigma_i sigma_j, products of density matrices that are generally non-Hermitian. These are different constructions: the circuit prepares the former, while the experiments evaluate the latter. The paper should clarify which feature matrix is used in each place, and either provide a circuit for the experimental construction or validate the circuit construction experimentally.
  4. [Section 5.1, Appendix B] The proof-of-concept fixes the interaction unitary U to the composition of a CNOT and a SWAP gate and does not report any sensitivity analysis. Since the reported advantage may depend on this ad hoc choice, the authors should vary U (for example, random two-qubit unitaries or different entangling gates) and report the resulting error distribution. Without this, the experiment demonstrates only that one particular entangled construction works on one task, not that the QOVK framework generally provides the claimed structural advantage.
minor comments (5)
  1. [Section 2.2, Notation paragraph] The notation paragraph describes the inner product and tensor product in bra-ket notation, but the final sentence is slightly confusing because it lists the tensor product immediately after the definition of the inner product; consider separating these two definitions more clearly.
  2. [Section 5.1, Eq. (10) reference] Section 5.1 refers to 'Eq. (10)' for the QOVK definition, but Eq. (10) appears only in Appendix D, not in the main text. The authors should renumber or add a forward reference to the appendix.
  3. [Appendix B.1, dephasing channel definition] In the definition of the dephasing channel, diag(sigma) should be explicitly identified as the matrix obtained by setting all off-diagonal entries of sigma to zero; the current wording is acceptable but could be more precise.
  4. [Section 5.2] The sentence 'The input feature matrix sigma^{x,z}_X produced by the quantum circuit is a density matrix ... In contrast, the general definition ... does not require sigma^{x,z}_X to be Hermitian' highlights a tension that should be resolved in the main definition, not only in the circuit discussion.
  5. [Global] The paper promises that the code will be publicly available at a GitLab URL; adding a link to the actual repository and a short description of the software dependencies would improve reproducibility for the proof-of-concept.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular steps: the entangled QOVK is an ansatz with a genuine simulation; the main gaps are a missing PSD proof and a missing classical OVK baseline, which are correctness and design concerns, not circular reductions.

full rationale

Walking the claimed derivation chain, I find no step in which a labeled prediction or first-principles result is equivalent by construction to an input, and no fitted parameter is renamed as a prediction. Definition 3.1 is an explicit ansatz: K(x,z)=Tr_X[U_{YX}(rho_Y⊗sigma^{x,z}_X)U†_{YX}] with U non-separable. It is not derived from Definition A.1's definition of a psd operator-valued kernel, and the paper itself concedes in Section 5.2 that "the general definition of an entangled QOVK does not require sigma^{x,z}_X to be Hermitian" and that the channel-estimation experiments use sigma^{i,j}_X=sigma_i sigma_j, products of density matrices that are generally non-Hermitian. That omission means the block-Gram PSD property required by Definition A.1 and Theorem A.3 is never established; this is a genuine correctness risk for the claim that the object is a kernel, but it is not a circular reduction. The proof-of-concept in Section 5.1 and Table 1 is a genuine simulation on fixed bit-flip and dephasing channels, with error measured against ground-truth Choi matrices and regularization chosen by cross-validation; no fitted constant is subsequently reported as a prediction. The absence of a classical operator-valued kernel baseline weakens the inference that the observed gain is quantum-specific, but a missing control is not circularity. The framework relies on self-cited work (Huusari & Kadri 2021; Hashimoto et al. 2023a,b, 2024, 2026), but those citations provide background definitions and roadmap support rather than an unverified uniqueness theorem or the sole justification of the central claim. Accordingly, the derivation chain is self-contained in the sense relevant to circularity, and I score 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The central construction rests on standard RKHS theory, the swap test, and the unproven psd-ness of the general entangled QOVK; the experiments also depend on hand-chosen U and rho_Y. Free parameters are limited to design choices in the proof-of-concept, not fitted values in a derivation.

free parameters (3)
  • Interaction unitary U = CNOT composed with SWAP
    The kernel's expressivity and the experimental success depend on this hand-picked non-separable unitary; no learning or search over U is performed.
  • Output density matrix rho_Y = unspecified (pure state |phi><phi| in the circuit)
    The definition of QOVK in (1) includes rho_Y; the experiments do not state how rho_Y is chosen, so it acts as an unexamined design parameter.
  • Noise mixing parameter alpha = 0.1
    Set by hand to simulate noisy situations in the experiments; not central to the proposed kernel definition but influences the reported numbers.
assumptions (4)
  • standard math Aronszajn-Schwartz bijection between psd operator-valued kernels and vector-valued RKHS
    Invoked in Appendix A (Theorem A.3) as the theoretical foundation for operator-valued kernels.
  • standard math The swap test estimates the fidelity between two pure states
    Used in Appendices C and D to design the quantum circuit for both QSVK and QOVK.
  • domain assumption K(x,z)=Tr_X[U(rho_Y tensor sigma_x sigma_z)U^dagger] is a positive semidefinite operator-valued kernel
    The paper does not prove psd-ness for its general QOVK construction (Definition 3.1) despite using it in ridge regression; it only notes the non-Hermitian feature matrix in Section 5.2.
  • domain assumption C*-algebraic RKHM learning theory is valid and applicable
    Action 3 relies on Hashimoto et al. (2021, 2023a,b, 2024) for the existence and properties of reproducing kernel Hilbert C*-modules; these are self-cited references.
invented entities (1)
  • Entangled quantum operator-valued kernel (QOVK)
    purpose: Represent non-factorizable input-output dependencies and exploit entanglement in quantum kernel machines
    Only internal proof-of-concept simulation is provided; no external falsifiable prediction or benchmark outside the paper is offered.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Position: Quantum Kernel Machines Should Move Beyond Scalar-Valued Kernels to Realize Their Potential." pith.science (2026). https://pith.science/paper/CZRKHOPV

@misc{pith2026250603779,
  author       = {Pith},
  title        = {Pith review of: Position: Quantum Kernel Machines Should Move Beyond Scalar-Valued Kernels to Realize Their Potential},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CZRKHOPV}},
  note         = {Machine review of arXiv:2506.03779}
}
abstract

Quantum kernel functions built using quantum-mechanical principles and have emerged as a centerpiece of quantum machine learning. The initial enthusiasm for quantum kernel machines has been tempered by recent studies suggesting that quantum kernels could not offer significant computational or statistical advantages when learning from classical data. However, most of the research in this area has been devoted to scalar-valued kernels in standard classification or regression settings for which classical kernel methods are efficient and effective, leaving very little room for improvement with quantum kernels. In this position paper, we argue that progress in this field requires moving beyond scalar-valued kernels toward more expressive kernel frameworks. Scalar-valued kernels lack the degrees of freedom necessary to fully exploit intrinsically quantum resources such as entanglement and are not rich enough to deal with complex learning tasks where classical learning methods struggle. Building on recent advances in operator-valued kernel learning and $C^*$-algebraic kernel representations, we propose a roadmap for designing quantum kernels capable of leveraging entanglement and non-commutative structures to tackle complex structured prediction problems. To support this viewpoint, we present an initial proof-of-concept illustrating how quantum operator-valued kernel formulations can reveal structural dependencies that remain difficult to access for scalar-valued kernel methods. This shift in focus could open a pathway toward a new generation of quantum kernel machines and a more faithful exploration of their potential advantages.

Figures

Figures reproduced from arXiv: 2506.03779 by the authors.

Figure 1
Figure 1. Quantum machine learning (QML) is a recent field of research at the intersection between quantum computing (QC) and machine learning (ML). The interaction is two-sided: quantum￾enhanced ML (from QC to ML), and ML-based quantum comput￾ing (from ML to QC). Most of the interest has concentrated on the use of quantum computing paradigms to improve machine learn￾ing algorithms. Quantum scalar-valued kernels (QSVKs) have … view at source ↗
Figure 2
Figure 2. Kernel feature map. (top) In the classical setting, a data point x is mapped to a high-dimensional space via an im￾plicit nonlinear feature map ϕ. The mapping ϕ may be unknown but the inner product between two data points, x and y, mapped by ϕ is equal to the kernel function evaluated at x and y, i.e., ⟨ϕ(x), ϕ(y)⟩ = k(x, y). (bottom) In the quantum setting, the feature map is known explicitly. Encoding a data point… view at source ↗
Figure 3
Figure 3. Diagram representation of the quantum operator-valued kernel (1). Input data are embedded into a feature matrix σ x,z X . We dilate the input system by tensorizing σ x,z X with an output density matrix ρY to form a larger composite system that enables interactions between input and outputs. A (unitary) interaction is then applied via U, resulting in the evolution of the composite input-output system. By tracing out … view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Illustration of inclusions among quantum kernel classes considered in this paper. Entangled quantum operator-valued ker￾nels (QOVKs) are not separable, i.e., dependencies between input and output variables cannot be considered separately. The class of separable QOVKs c…
Figure 5
Figure 5. Figure 5: Bit-flip channel (first row) and dephasing channel (second row) estimation results with a = 0.2, n = 100 and α = 0.1. Left plot is the Choi matrix of the correct channel from which data was created. Middle plot is the Choi matrix of the quantum channel recovered by QSV…
Figure 6
Figure 6. Figure 6: A quantum circuit for preparing a quantum state corre￾sponding to the value of an operator-valued kernel of the form (1). The application of the partial trace on the register Z by measuring the ancillary qubit in the state |0⟩ produces a valid input feature matrix σ x,…
Figure 7
Figure 7. Figure 7: Quantum circuit for computing the fidelity kernel using a swap test. Measuring the ancillary qubit provides the fidelity between the two quantum states |ψx⟩ and |ψz⟩. The probability to measure the ancillary qubit in the state |0⟩ is given by P(|0⟩a ) = 1 2 + 1 2 | ⟨ψx…
Figure 8
Figure 8. Figure 8: A quantum circuit implementation of the operator-valued kernel given in (1). The state η1 is then obtained by measuring the ancilla qubit in the state |0⟩, i.e., η1 = Tra[ρ Ψ3 (|0⟩ ⟨0|a ⊗ IZXY )] Tr[ρΨ3 (|0⟩ ⟨0|a ⊗ IZXY )] = ρz ⊗ ρx + ρx ⊗ ρz + |ψz⟩ ⟨ψx| ⊗ |ψx⟩ ⟨ψz| + …

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

91 extracted references · 65 canonical work pages

  1. [1]

    Challenges and opportunities in quantum optimization

    Amira Abbas, Andris Ambainis, Brandon Augustino, Andreas B \"a rtschi, Harry Buhrman, Carleton Coffrin, Giorgio Cortiana, Vedran Dunjko, Daniel J Egger, Bruce G Elmegreen, et al. Challenges and opportunities in quantum optimization. Nature Reviews Physics, pages 1--18, 2024

  2. [2]

    Aizerman, E

    A. Aizerman, E. M. Braverman, and L. I. Rozoner. Theoretical foundations of the potential function method in pattern recognition learning . Automation and Remote Control, 25: 0 821--837, 1964

  3. [3]

    Quantum feature maps for graph machine learning on a neutral atom quantum processor

    Boris Albrecht, Constantin Dalyac, Lucas Leclerc, Luis Ortiz-Guti \'e rrez, Slimane Thabet, Mauro D'Arcangelo, Julia RK Cline, Vincent E Elfving, Lucas Lassabli \`e re, Henrique Silv \'e rio, et al. Quantum feature maps for graph machine learning on a neutral atom quantum processor. Physical Review A, 107 0 (4): 0 042615, 2023

  4. [4]

    Kernels for vector-valued functions: A review

    Mauricio A Alvarez, Lorenzo Rosasco, Neil D Lawrence, et al. Kernels for vector-valued functions: A review. Foundations and Trends in Machine Learning , 4 0 (3): 0 195--266, 2012

  5. [5]

    Aronszajn

    N. Aronszajn. Theory of reproducing kernels. Transactions of the American Mathematical Society, 68: 0 337--404, 1950

  6. [6]

    A quantum jensen--shannon graph kernel for unattributed graphs

    Lu Bai, Luca Rossi, Andrea Torsello, and Edwin R Hancock. A quantum jensen--shannon graph kernel for unattributed graphs. Pattern Recognition, 48 0 (2): 0 344--355, 2015

  7. [7]

    Qbmk: Quantum-based matching kernels for un-attributed graphs

    Lu Bai, Lixin Cui, Ming Li, Yue Wang, and Edwin Hancock. Qbmk: Quantum-based matching kernels for un-attributed graphs. In International Conference on Machine Learning, 2024

  8. [8]

    G. H. Bakir, T. Hofmann, B. Sch \"o lkopf, A. J. Smola, B. Taskar, and S. V. N. Vishwanathan. Predicting structured data. MIT press, 2007

Show all 91 references
  1. [9]

    Benign overfitting in linear regression

    Peter L Bartlett, Philip M Long, G \'a bor Lugosi, and Alexander Tsigler. Benign overfitting in linear regression. Proceedings of the National Academy of Sciences, 117 0 (48): 0 30063--30070, 2020

  2. [10]

    Noncommutative christoffel-darboux kernels

    Serban Belinschi, Victor Magron, and Victor Vinnikov. Noncommutative christoffel-darboux kernels. Transactions of the American Mathematical Society, 376 0 (01): 0 181--230, 2023

  3. [11]

    Reconciling modern machine-learning practice and the classical bias--variance trade-off

    Mikhail Belkin, Daniel Hsu, Siyuan Ma, and Soumik Mandal. Reconciling modern machine-learning practice and the classical bias--variance trade-off. Proceedings of the National Academy of Sciences, 116 0 (32): 0 15849--15854, 2019

  4. [12]

    Quantum machine learning

    Jacob Biamonte, Peter Wittek, Nicola Pancotti, Patrick Rebentrost, Nathan Wiebe, and Seth Lloyd. Quantum machine learning. Nature, 549 0 (7671): 0 195--202, 2017

  5. [13]

    Quantum classifier with tailored quantum kernel

    Carsten Blank, Daniel K Park, June-Koo Kevin Rhee, and Francesco Petruccione. Quantum classifier with tailored quantum kernel. npj Quantum Information, 6 0 (1): 0 41, 2020

  6. [14]

    Graph kernels: State-of-the-art and future challenges

    Karsten Borgwardt, Elisabetta Ghisu, Felipe Llinares-L \'o pez, Leslie O’Bray, Bastian Rieck, et al. Graph kernels: State-of-the-art and future challenges. Foundations and Trends in Machine Learning , 13 0 (5-6): 0 531--712, 2020

  7. [15]

    B. E. Boser, I. M. Guyon, and V. N. Vapnik. A training algorithm for optimal margin classifiers. In Annual Workshop on Computational Learning Theory, pages 144--152. ACM, June 1992

  8. [16]

    Input output kernel regression: Supervised and semi-supervised structured output prediction with operator-valued kernels

    C \'e line Brouard, Marie Szafranski, and Florence d'Alch \'e Buc. Input output kernel regression: Supervised and semi-supervised structured output prediction with operator-valued kernels. Journal of Machine Learning Research, 17: 0 np, 2016

  9. [17]

    C*-Algebras and Mathematical Foundations of Quantum Statistical Mechanics: An Introduction

    Jean-Bernard Bru and Walter Alberto de Siqueira Pedra. C*-Algebras and Mathematical Foundations of Quantum Statistical Mechanics: An Introduction. Springer Nature, 2023

  10. [18]

    Quantum fingerprinting

    Harry Buhrman, Richard Cleve, John Watrous, and Ronald De Wolf. Quantum fingerprinting. Physical review letters, 87 0 (16): 0 167902, 2001

  11. [19]

    Optimization of polynomials in non-commuting variables, volume 2

    Sabine Burgdorf, Igor Klep, Janez Povh, et al. Optimization of polynomials in non-commuting variables, volume 2. Springer, 2016

  12. [20]

    Towards a theory of non-commutative optimization: Geodesic 1st and 2nd order methods for moment maps and polytopes

    Peter B \"u rgisser, Cole Franks, Ankit Garg, Rafael Oliveira, Michael Walter, and Avi Wigderson. Towards a theory of non-commutative optimization: Geodesic 1st and 2nd order methods for moment maps and polytopes. In 2019 IEEE 60th Annual Symposium on Foundations of Computer S...

  13. [21]

    Wild, and Ruslan Shaydulin

    Abdulkadir Canatar, Evan Peters, Cengiz Pehlevan, Stefan M. Wild, and Ruslan Shaydulin. Bandwidth enables generalization in quantum kernel models. Transactions on Machine Learning Research, 2023

  14. [22]

    Universal multi-task kernels

    Andrea Caponnetto, Charles A Micchelli, Massimiliano Pontil, and Yiming Ying. Universal multi-task kernels. The Journal of Machine Learning Research, 9: 0 1615--1646, 2008

  15. [23]

    Vector valued reproducing kernel hilbert spaces and universality

    Claudio Carmeli, Ernesto De Vito, Alessandro Toigo, and Veronica Umanit \'a . Vector valued reproducing kernel hilbert spaces and universality. Analysis and Applications, 8 0 (01): 0 19--61, 2010

  16. [24]

    Quantum machine learning: a classical perspective

    Carlo Ciliberto, Mark Herbster, Alessandro Davide Ialongo, Massimiliano Pontil, Andrea Rocchetto, Simone Severini, and Leonard Wossnig. Quantum machine learning: a classical perspective. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 474 0...

  17. [25]

    Quantum computing: Lecture notes

    Ronald De Wolf. Quantum computing: Lecture notes. arXiv preprint arXiv:1907.09415, 2019

  18. [26]

    Quantum advantage seeker with kernels (quask): a software framework to speed up the research in quantum machine learning

    Francesco Di Marcantonio, Massimiliano Incudini, Davide Tezza, and Michele Grossi. Quantum advantage seeker with kernels (quask): a software framework to speed up the research in quantum machine learning. Quantum Machine Intelligence, 5 0 (1): 0 20, 2023

  19. [27]

    Machine learning & artificial intelligence in the quantum domain: a review of recent progress

    Vedran Dunjko and Hans J Briegel. Machine learning & artificial intelligence in the quantum domain: a review of recent progress. Reports on Progress in Physics, 81 0 (7): 0 074001, 2018

  20. [28]

    Deep sketched output kernel regression for structured prediction

    Tamim El Ahmad, Junjie Yang, Pierre Laforgue, and Florence d’Alch \'e Buc. Deep sketched output kernel regression for structured prediction. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases, pages 93--110, 2024

  21. [29]

    Learning multiple tasks with kernel methods

    Theodoros Evgeniou, Charles A Micchelli, Massimiliano Pontil, and John Shawe-Taylor. Learning multiple tasks with kernel methods. Journal of machine learning research, 6 0 (4), 2005

  22. [30]

    Quantum channel identification problem

    Akio Fujiwara. Quantum channel identification problem. Physical Review A, 63 0 (4): 0 042304, 2001

  23. [31]

    Quantum parameter estimation of a generalized pauli channel

    Akio Fujiwara and Hiroshi Imai. Quantum parameter estimation of a generalized pauli channel. Journal of Physics A: Mathematical and General, 36 0 (29): 0 8093, 2003

  24. [32]

    Learning quantum systems

    Valentin Gebhart, Raffaele Santagati, Antonio Andrea Gentile, Erik M Gauger, David Craig, Natalia Ares, Leonardo Banchi, Florian Marquardt, Luca Pezz \`e , and Cristian Bonato. Learning quantum systems. Nature Reviews Physics, 5 0 (3): 0 141--156, 2023

  25. [33]

    Quantum kernel alignment with stochastic gradient descent

    Gian Gentinetta, David Sutter, Christa Zoufal, Bryce Fuller, and Stefan Woerner. Quantum kernel alignment with stochastic gradient descent. In 2023 IEEE International Conference on Quantum Computing and Engineering (QCE), volume 1, pages 256--262. IEEE, 2023

  26. [34]

    Kernelizing the output of tree-based methods

    Pierre Geurts, Louis Wehenkel, and Florence d'Alch \'e Buc. Kernelizing the output of tree-based methods. In Proceedings of the 23rd international conference on Machine learning, pages 345--352, 2006

  27. [35]

    Understanding quantum machine learning also requires rethinking generalization

    Elies Gil-Fuster, Jens Eisert, and Carlos Bravo-Prieto. Understanding quantum machine learning also requires rethinking generalization. Nature Communications, 15 0 (1): 0 2277, 2024

  28. [36]

    Covariant quantum kernels for data with group structure

    Jennifer R Glick, Tanvi P Gujarati, Antonio D Corcoles, Youngseok Kim, Abhinav Kandala, Jay M Gambetta, and Kristan Temme. Covariant quantum kernels for data with group structure. Nature Physics, 20 0 (3): 0 479--483, 2024

  29. [37]

    Bounds on entanglement dimensions and quantum graph parameters via noncommutative polynomial optimization

    Sander Gribling, David de Laat, and Monique Laurent. Bounds on entanglement dimensions and quantum graph parameters via noncommutative polynomial optimization. Mathematical programming, 170: 0 5--42, 2018

  30. [38]

    A fast quantum mechanical algorithm for database search

    Lov K Grover. A fast quantum mechanical algorithm for database search. In Proceedings of the twenty-eighth annual ACM symposium on Theory of computing, pages 212--219, 1996

  31. [39]

    Reproducing kernel hilbert c*-module and kernel mean embeddings

    Yuka Hashimoto, Isao Ishikawa, Masahiro Ikeda, Fuyuta Komura, Takeshi Katsura, and Yoshinobu Kawahara. Reproducing kernel hilbert c*-module and kernel mean embeddings. The Journal of Machine Learning Research, 22 0 (1): 0 12292--12347, 2021

  32. [40]

    Deep learning with kernels through rkhm and the perron-frobenius operator

    Yuka Hashimoto, Masahiro Ikeda, and Hachem Kadri. Deep learning with kernels through rkhm and the perron-frobenius operator. Advances in Neural Information Processing Systems, 36: 0 50677--50696, 2023 a

  33. [41]

    Learning in rkhm: a c*-algebraic twist for kernel machines

    Yuka Hashimoto, Masahiro Ikeda, and Hachem Kadri. Learning in rkhm: a c*-algebraic twist for kernel machines. In International Conference on Artificial Intelligence and Statistics, pages 692--708, 2023 b

  34. [42]

    Spectral truncation kernels: Noncommutativity in C^* -algebraic kernel machines

    Yuka Hashimoto, Ayoub Hafid, Masahiro Ikeda, and Hachem Kadri. Spectral truncation kernels: Noncommutativity in C^* -algebraic kernel machines. arXiv preprint arXiv:2405.17823, 2024

  35. [43]

    Supervised learning with quantum-enhanced feature spaces

    Vojt e ch Havl \' c ek, Antonio D C \'o rcoles, Kristan Temme, Aram W Harrow, Abhinav Kandala, Jerry M Chow, and Jay M Gambetta. Supervised learning with quantum-enhanced feature spaces. Nature, 567 0 (7747): 0 209--212, 2019

  36. [44]

    Noisy quantum kernel machines

    Valentin Heyraud, Zejian Li, Zakari Denis, Alexandre Le Boit \'e , and Cristiano Ciuti. Noisy quantum kernel machines. Physical Review A, 106 0 (5): 0 052421, 2022

  37. [45]

    Kernel methods in machine learning

    Thomas Hofmann, Bernhard Sch \"o lkopf, and Alexander J Smola. Kernel methods in machine learning. The Annals of Statistics, 36 0 (3): 0 1171--1220, 2008

  38. [46]

    Preimage problem in kernel-based machine learning

    Paul Honeine and Cedric Richard. Preimage problem in kernel-based machine learning. IEEE Signal Processing Magazine, 28 0 (2): 0 77--88, 2011

  39. [47]

    Hsin-Yuan Huang, Michael Broughton, Masoud Mohseni, Ryan Babbush, Sergio Boixo, Hartmut Neven, and Jarrod R. McClean. Power of data in quantum machine learning. Nature Communications, 12 0 (1), 05 2021. ISSN 2041-1723. doi:10.1038/s41467-021-22539-9. URL http://dx.doi.org/10.1...

  40. [48]

    Training quantum embedding kernels on near-term quantum computers

    Thomas Hubregtsen, David Wierichs, Elies Gil-Fuster, Peter-Jan HS Derks, Paul K Faehrmann, and Johannes Jakob Meyer. Training quantum embedding kernels on near-term quantum computers. Physical Review A, 106 0 (4): 0 042431, 2022

  41. [49]

    Entangled kernels - beyond separability

    Riikka Huusari and Hachem Kadri. Entangled kernels - beyond separability. Journal of Machine Learning Research, 22 0 (24): 0 1--40, 2021

  42. [50]

    Multi-view metric learning in vector-valued kernel spaces

    Riikka Huusari, Hachem Kadri, and C \'e cile Capponi. Multi-view metric learning in vector-valued kernel spaces. In International Conference on Artificial Intelligence and Statistics, pages 415--424, 2018

  43. [51]

    Automatic and effective discovery of quantum kernels

    Massimiliano Incudini, Daniele Lizzio Bosco, Francesco Martini, Michele Grossi, Giuseppe Serra, and Alessandra Di Pierro. Automatic and effective discovery of quantum kernels. IEEE Transactions on Emerging Topics in Computational Intelligence, 2024

  44. [52]

    Seeking a quantum advantage for machine learning

    Nature Machine Intelligence. Seeking a quantum advantage for machine learning. Nat Mach Intell, 5 0 (8): 0 813--813, 2023

  45. [53]

    Quantum machine learning beyond kernel methods

    Sofiene Jerbi, Lukas J Fiderer, Hendrik Poulsen Nautrup, Jonas M K \"u bler, Hans J Briegel, and Vedran Dunjko. Quantum machine learning beyond kernel methods. Nature Communications, 14 0 (1): 0 517, 2023

  46. [54]

    Machine learning: Trends, perspectives, and prospects

    Michael I Jordan and Tom M Mitchell. Machine learning: Trends, perspectives, and prospects. Science, 349 0 (6245): 0 255--260, 2015

  47. [55]

    On the role of entanglement in quantum-computational speed-up

    Richard Jozsa and Noah Linden. On the role of entanglement in quantum-computational speed-up. Proceedings of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences, 459 0 (2036): 0 2011--2032, 2003

  48. [56]

    A generalized kernel approach to structured output learning

    Hachem Kadri, Mohammad Ghavamzadeh, and Philippe Preux. A generalized kernel approach to structured output learning. In International Conference on Machine Learning, pages 471--479, 2013

  49. [57]

    Operator-valued kernels for learning from functional response data

    Hachem Kadri, Emmanuel Duflos, Philippe Preux, St é phane Canu, Alain Rakotomamonjy, and Julien Audiffren. Operator-valued kernels for learning from functional response data. Journal of Machine Learning Research, 16: 0 1--54, 2016

  50. [58]

    Double descent in quantum machine learning

    Marie Kempkes, Aroosa Ijaz, Elies Gil-Fuster, Carlos Bravo-Prieto, Jakob Spiegelberg, Evert van Nieuwenburg, and Vedran Dunjko. Double descent in quantum machine learning. arXiv preprint arXiv:2501.10077, 2025

  51. [59]

    u bler, Simon Buchholz, and Bernhard Sch \

    Jonas K \"u bler, Simon Buchholz, and Bernhard Sch \"o lkopf. The inductive bias of quantum kernels. Advances in Neural Information Processing Systems, 34: 0 12661--12673, 2021

  52. [60]

    Classically approximating variational quantum machine learning with random fourier features

    Jonas Landman, Slimane Thabet, Constantin Dalyac, Hela Mhiri, and Elham Kashefi. Classically approximating variational quantum machine learning with random fourier features. In The Eleventh International Conference on Learning Representations, 2023

  53. [61]

    Theory of overparametrization in quantum neural networks

    Martin Larocca, Nathan Ju, Diego Garc \' a-Mart \' n, Patrick J Coles, and Marco Cerezo. Theory of overparametrization in quantum neural networks. Nature Computational Science, 3 0 (6): 0 542--551, 2023

  54. [62]

    Deep learning

    Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. Deep learning. nature, 521 0 (7553): 0 436--444, 2015

  55. [63]

    Operator-valued kernel-based vector autoregressive models for network inference

    N \'e h \'e my Lim, Florence d’Alch \'e Buc, C \'e dric Auliac, and George Michailidis. Operator-valued kernel-based vector autoregressive models for network inference. Machine learning, 99: 0 489--513, 2015

  56. [64]

    Random features for kernel approximation: A survey on algorithms, theory, and beyond

    Fanghui Liu, Xiaolin Huang, Yudong Chen, and Johan AK Suykens. Random features for kernel approximation: A survey on algorithms, theory, and beyond. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44 0 (10): 0 7128--7148, 2021 a

  57. [65]

    A rigorous and robust quantum speed-up in supervised machine learning

    Yunchao Liu, Srinivasan Arunachalam, and Kristan Temme. A rigorous and robust quantum speed-up in supervised machine learning. Nature Physics, 17 0 (9): 0 1013--1017, 2021 b

  58. [66]

    Quantum embeddings for machine learning

    Seth Lloyd, Maria Schuld, Aroosa Ijaz, Josh Izaac, and Nathan Killoran. Quantum embeddings for machine learning. arXiv preprint arXiv:2001.03622, 2020

  59. [67]

    Kernel methods in quantum machine learning

    Riccardo Mengoni and Alessandra Di Pierro. Kernel methods in quantum machine learning. Quantum Machine Intelligence, 1 0 (3-4): 0 65--71, 2019

  60. [68]

    On learning vector-valued functions

    Charles A Micchelli and Massimiliano Pontil. On learning vector-valued functions. Neural computation, 17 0 (1): 0 177--204, 2005

  61. [69]

    Quantum linear system solvers: A survey of algorithms and applications

    Mauro ES Morales, Lirand \"e Pira, Philipp Schleich, Kelvin Koor, Pedro Costa, Dong An, Al \'a n Aspuru-Guzik, Lin Lin, Patrick Rebentrost, and Dominic W Berry. Quantum linear system solvers: A survey of algorithms and applications. arXiv preprint arXiv:2411.02522, 2024

  62. [70]

    Calibrating the role of entanglement in variational quantum circuits

    Azar C Nakhl, Thomas Quella, and Muhammad Usman. Calibrating the role of entanglement in variational quantum circuits. Physical Review A, 109 0 (3): 0 032413, 2024

  63. [71]

    Quantum computation and quantum information

    Michael A Nielsen and Isaac L Chuang. Quantum computation and quantum information. Cambridge university press, 2010

  64. [72]

    Generalization despite overfitting in quantum machine learning models

    Evan Peters and Maria Schuld. Generalization despite overfitting in quantum machine learning models. Quantum, 7: 0 1210, 2023

  65. [73]

    Quantum computing in the NISQ era and beyond

    John Preskill. Quantum computing in the NISQ era and beyond. Quantum, 2: 0 79, 2018

  66. [74]

    Neural quantum kernels: training quantum kernels with quantum neural networks

    Pablo Rodriguez-Grasa, Yue Ban, and Mikel Sanz. Neural quantum kernels: training quantum kernels with quantum neural networks. arXiv preprint arXiv:2401.04642, 2024

  67. [75]

    Quantum perceptron revisited: Computational-statistical tradeoffs

    Mathieu Roget, Giuseppe Di Molfetta, and Hachem Kadri. Quantum perceptron revisited: Computational-statistical tradeoffs. In Uncertainty in Artificial Intelligence, pages 1697--1706, 2022

  68. [76]

    Supervised quantum machine learning models are kernel methods

    Maria Schuld. Supervised quantum machine learning models are kernel methods. arXiv preprint arXiv:2101.11020, 2021

  69. [77]

    Quantum machine learning in feature hilbert spaces

    Maria Schuld and Nathan Killoran. Quantum machine learning in feature hilbert spaces. Physical review letters, 122 0 (4): 0 040504, 2019

  70. [78]

    Is quantum advantage the right goal for quantum machine learning? Prx Quantum, 3 0 (3): 0 030101, 2022

    Maria Schuld and Nathan Killoran. Is quantum advantage the right goal for quantum machine learning? Prx Quantum, 3 0 (3): 0 030101, 2022

  71. [79]

    An introduction to quantum machine learning

    Maria Schuld, Ilya Sinayskiy, and Francesco Petruccione. An introduction to quantum machine learning. Contemporary Physics, 56 0 (2): 0 172--185, 2015

  72. [80]

    Schwartz

    L. Schwartz. Sous-espaces hilbertiens d'espaces vectoriels topologiques et noyaux associ \'e s (noyaux reproduisants). Journal d'Analyse Math \'e matique , 13: 0 115--256, 1964

  73. [81]

    A graph to graphs framework for retrosynthesis prediction

    Chence Shi, Minkai Xu, Hongyu Guo, Ming Zhang, and Jian Tang. A graph to graphs framework for retrosynthesis prediction. In International conference on machine learning, pages 8818--8827, 2020

  74. [82]

    Learning partial differential equations in reproducing kernel hilbert spaces

    George Stepaniants. Learning partial differential equations in reproducing kernel hilbert spaces. Journal of Machine Learning Research, 24 0 (86): 0 1--72, 2023

  75. [83]

    Quantum fisher kernel for mitigating the vanishing similarity issue

    Yudai Suzuki, Hideaki Kawaguchi, and Naoki Yamamoto. Quantum fisher kernel for mitigating the vanishing similarity issue. arXiv preprint arXiv:2210.16581, 2022

  76. [84]

    Graphqntk: Quantum neural tangent kernel for graph data

    Yehui Tang and Junchi Yan. Graphqntk: Quantum neural tangent kernel for graph data. Advances in Neural Information Processing Systems, 35: 0 6104--6118, 2022

  77. [85]

    Quantum positional encodings for graph neural networks

    Slimane Thabet, Mehdi Djellabi, Igor Sokolov, Sachin Kasture, Louis-Paul Henry, and Lo \" c Henriet. Quantum positional encodings for graph neural networks. In International Conference on Machine Learning, 2024

  78. [86]

    Exponential concentration in quantum kernel methods

    Supanut Thanasilp, Samson Wang, M Cerezo, and Zo \"e Holmes. Exponential concentration in quantum kernel methods. Nature Communications, 15 0 (1): 0 5200, 2024

  79. [87]

    Benign overfitting with quantum kernels

    Joachim Tomasi, Sandrine Anthoine, and Hachem Kadri. Benign overfitting with quantum kernels. arXiv preprint arXiv:2503.17020, 2025

  80. [88]

    Large margin methods for structured and interdependent output variables

    Ioannis Tsochantaridis, Thorsten Joachims, Thomas Hofmann, Yasemin Altun, and Yoram Singer. Large margin methods for structured and interdependent output variables. Journal of machine learning research, 6 0 (9), 2005

  81. [89]

    Towards understanding the power of quantum kernels in the NISQ era

    Xinbiao Wang, Yuxuan Du, Yong Luo, and Dacheng Tao. Towards understanding the power of quantum kernels in the NISQ era. Quantum, 5: 0 531, 2021

  82. [90]

    The theory of quantum information

    John Watrous. The theory of quantum information. Cambridge university press, 2018

  83. [91]

    Link prediction based on graph neural networks

    Muhan Zhang and Yixin Chen. Link prediction based on graph neural networks. Advances in neural information processing systems, 31, 2018

Pith tools

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