Pith. sign in

REVIEW 4 major objections 4 minor 37 references

Tensor-Based Binary Graph Encoding for Variational Quantum Classifiers

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper claims that encoding each graph vertex as a binary-indexed Pauli-Z tensor product and each edge as a weighted product preserves full graph structure, allowing a variational quantum classifier to beat the PCA-based baseline on…

desk verdict Novel binary Pauli-string graph encoding undercut by a permutation-invariance flaw, an under-specified circuit, and an incorrect eigenvalue bound. read the letter →

arxiv 2501.14185 v2 pith:2BNO7RKM submitted 2025-01-24 quant-ph physics.data-an

Shiwen An , Konstantinos Slavakis This is my paper · ORCID
classification quant-phphysics.data-an PACS 03.67.Lx
keywords variationalquantumclassifiergraphclassificationPauli-ZencodingtensorproductIsingmodelmachinelearningNISQ
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 proposes a graph encoding for variational quantum classifiers (VQCs) in which each vertex is mapped to a tensor product of Pauli-$Z$ operators determined by the vertex's binary index, and each edge is mapped to a weighted product of the two vertex strings, forming an Ising-model-like Hamiltonian. The authors claim this encoding preserves the full graph structure, unlike PCA-based preprocessing which discards information, and it requires only $O(\log N_G)$ qubits. They report that their EG-VQC achieves higher accuracy than PCA-VQC on the MUTAG, PROTEIN, and ENZYME datasets (e.g., 80.3% vs 75.1% on MUTAG), with lower time complexity. A sympathetic reader would care because it suggests a practical, NISQ-friendly way to feed graph data into variational quantum circuits without dimensionality reduction.

What carries the argument

The enabling object is the binary-indexed Pauli-$Z$ tensor product: for a vertex labeled $i$ with binary expansion $\sum_{k=0}^{N-1} r_k 2^k$, the vertex is represented by $\bigotimes_{k=0}^{N-1} Z^{r_k}$. Edges are represented by weighted products $Z_i^{\otimes} Z_j^{\otimes}$, producing an Ising-model-style Hamiltonian whose expectation values are measured by the VQC. This tensor encoding is what carries the graph structure into the quantum circuit, and the eigenvalue bound lemma in the appendix is what keeps the resulting observable within a range compatible with the cross-entropy loss.

What would settle it

Train the same EG-VQC circuit on the MUTAG, PROTEIN, or ENZYME datasets twice: once with the original vertex order and once with a random permutation of vertex indices applied consistently across all graphs (so adjacency is unchanged). If test accuracy drops materially under relabeling, the encoding is not capturing permutation-invariant graph structure; if accuracy is unchanged, the claim that the binary-indexed encoding itself carries the structural information is supported.

Watch

Extended reading notes

Core claim

The central claim is that a graph can be encoded into a quantum Hamiltonian of the form $$H(G) = \sum_{\{i,j\}\in E} J_{ij} $Z_i^{{\otimes}}$ $Z_j^{{\otimes}}$ + \sum_{i\in V} h_i $Z_i^{{\otimes}}$,$$ where $Z_i^{\otimes}$ is the tensor product of Pauli-$Z$ operators indexed by the binary representation of vertex $i$, and $J_{ij}, h_i$ are normalized weights. The paper argues that this encoding retains the complete interaction structure of the graph, that its eigenvalues can be bounded to $[-1,1]$ (with proof in the appendix under the condition $\Delta J + \Delta h \le 1$), and that feeding this Hamiltonian into a VQC with variational $RY$ and $CNOT$ layers yields better classification accuracy than the PCA-VQC baseline on the three benchmark datasets.

Load-bearing premise

The encoding takes the arbitrary numbering of vertices as meaningful: permuting the vertex labels changes the Hamiltonian and therefore the learned classifier, so the method as presented is not invariant under graph isomorphism, the usual symmetry of graph classification.

Editorial extensions

If this is right

  • Graph structure can be fed into a VQC without PCA-style dimensionality reduction, eliminating a known source of information loss.
  • The $O(N_G^2)$ encoding time beats the $O(N_G^3)$ SVD step of PCA-VQC, so larger graphs become feasible on limited hardware.
  • The method needs only $\lceil \log_2 N_G \rceil$ qubits, so graphs with up to $2^N$ vertices can be processed on an $N$-qubit device.
  • On the tested benchmarks, EG-VQC yields higher validation accuracy and lower overfitting (flatter validation loss) than PCA-VQC.
  • The encoding is compatible with cross-entropy loss and standard variational training (Adam over 100 epochs).

Reading between the lines

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

  • The encoding as written is tied to the arbitrary integer labels of vertices; relabeling vertices produces a different Hamiltonian, so the method as presented is not permutation-invariant, which is the usual setting for graph classification benchmarks. A permutation-invariant variant would need a canonical vertex ordering or a symmetrization over labelings.
  • Because the full graph Hamiltonian is used as the observable, the method might also be read as a quantum kernel: two graphs can be compared by their Hamiltonian expectation values, suggesting a possible route to quantum graph kernels that avoid classical feature extraction.
  • A direct testable extension is to run the same pipeline on larger graph benchmarks (e.g., NCI1, DD) to see whether the accuracy gap over PCA-VQC persists as graph size and sparsity vary.
  • The eigenvalue-bound proof in the appendix only covers the special case $\Delta J + \Delta h \le 1$; for general graphs the Hamiltonian may leave the $[-1,1]$ interval, so future work would need a normalization strategy that holds unconditionally.
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 / 4 minor

Summary. The paper proposes a graph encoding scheme for variational quantum classifiers, called EG-VQC. Each vertex is mapped to a binary-indexed tensor product of Pauli-Z operators (Eq. 6) and each edge to a weighted product of such operators (Eq. 7), so that a graph becomes a Hamiltonian H(G) of Eq. (8). The authors argue that this encoding preserves graph information, avoids PCA information loss, and requires fewer qubits than PCA-VQC. They present numerical results on MUTAG, PROTEIN, and ENZYME (Table I) claiming higher classification accuracy than PCA-VQC, together with a complexity analysis and an appendix proof intended to show that H(G) can be normalized to have eigenvalues in [-1,1]. The central claims are that the encoding preserves the integrity of graph data and that the resulting classifier outperforms PCA-VQC.

Significance. If the claims were correct, a permutation-robust, qubit-efficient graph encoding for VQCs would be a useful contribution to quantum machine learning. The paper also makes a potentially valuable point: full graph information can be encoded without PCA dimensionality reduction, and the O(N_G^2) encoding complexity is an improvement over the O(N_G^3) SVD step. The numerical comparison with PCA-VQC is a plausible and falsifiable target. However, the manuscript as written does not establish the central claims: the encoding is not permutation invariant, the eigenvalue proof is mathematically incorrect, and the algorithm contains an undefined measurement formula. These issues are load-bearing for the paper's main assertions, so the current version is not publishable.

major comments (4)
  1. [III-A, Eq. (6)-(8); III-C2] The encoding is not invariant under vertex relabeling, which is a fundamental problem for graph classification. Equation (6) assigns the Pauli string based on the binary expansion of an integer label i, so a graph and any of its isomorphic copies with a different vertex ordering produce, in general, different Hamiltonians and different VQC predictions. Section III-C2 explicitly states that the method 'eliminates the need for strict vertex alignment,' but the opposite is true: without a canonical vertex ordering, the same unlabeled graph can yield different classification outcomes depending on the arbitrary order in which its vertices appear in the dataset. Since MUTAG, PROTEIN, and ENZYME are standard permutation-invariant graph benchmarks, this representation-level flaw undermines the claim that EG-VQC 'preserves the integrity of graph data.'
  2. [Appendix A.4, combined contribution] The eigenvalue bound is incorrect. The appendix states that because each Pauli term has eigenvalues ±1 and coefficients J_ij and h_i are multiplied with these terms, 'the eigenvalues of H(G) are bounded by the maximum of the absolute values of J_ij and h_i,' and then concludes λ(H(G)) ∈ [-(ΔJ+Δh), ΔJ+Δh]. This is false for a sum of operators. Since all terms in Eq. (8) are diagonal in the computational basis, the eigenvalue for a basis state |b⟩ is Σ_{ij∈E} ±J_ij + Σ_i ±h_i, whose absolute value can be as large as Σ|J_ij| + Σ|h_i|, not merely ΔJ+Δh. The normalization condition ΔJ+Δh≤1 is therefore not sufficient to guarantee eigenvalues in [-1,1]. The boundary condition is imposed as a construction choice, not derived, so the proof in the appendix does not support the stated claim.
  3. [Algorithm 1, line 15] The measurement formula 'pi = ⟨0|H(G)U(θ)|0⟩' is not a valid probability. H(G) is a Hermitian operator, not a positive operator-valued measure element, and the expression is not guaranteed to be real-valued nonnegative or bounded in [0,1]. Moreover, the order of H(G) and U(θ) is unclear: if the Hamiltonian is used to encode the graph, the state should be evolved or measured in a way that the encoding appears before the variational circuit. As written, the cross-entropy loss in Section III-C cannot be computed from this formula. This is a load-bearing implementation detail because the entire numerical section depends on it.
  4. [IV-C, Table I] The central empirical claim rests entirely on Table I, but the manuscript provides no details about the experimental protocol beyond '90:10 split with stratified sampling.' It does not state the number of independent runs, the hyperparameters of the PCA-VQC baseline, the circuit depth used for each dataset, the optimizer settings beyond learning rate and epochs, or how the reported means and standard deviations were computed. Without these details, the accuracy differences (3-8 percentage points) cannot be assessed for statistical significance. The absence of code or a detailed methodology makes the results non-reproducible.
minor comments (4)
  1. [II.D] There is a typo in the section heading: 'PAC-VQC' should be 'PCA-VQC.'
  2. [References] Reference [28] is cited as the source of the MUTAG dataset, but the listed paper (Debnath, Soni, and Reddy, 'A survey on graph neural networks') is not the MUTAG dataset reference. The MUTAG dataset should be cited to its original source, e.g., Debnath et al., J. Med. Chem. 1991.
  3. [Algorithm 1, line 10] The initialization is inconsistent: line 10 says to initialize qubits with Hadamard gates, but the measurement formula in line 15 uses the state |0⟩. The paper should clarify which initial state is used and how the Hamiltonian encoding in Eq. (8) is converted into a circuit that acts on that state.
  4. [IV-C2, Fig. 3] The caption of Fig. 3 states that the training process using MUTAG and PROTEIN is shown, but the figure is not referenced in the main text and no loss curves are described quantitatively. Please clarify what is plotted and how the loss trends support the claim of reduced overfitting.

Circularity Check

1 steps flagged · score 2.0 of 10

No significant circularity: the classification comparison is empirical and self-contained; only the eigenvalue-range appendix reduces to a construction choice.

  1. self definitional [Section III-A1 (Boundary Condition) and Appendix A (Eigenvalue Range, conclusion)]
    "To ensure the physical consistency of the graph representation, the encoded Hamiltonian better satisfy the condition of eigenvalues in range of[−1, 1]. ... If ∆J + ∆h ≤ 1, the eigenvalues of H(G) are within the range: λ(H(G)) ∈ [−1, 1]."

    The appendix presents the [−1,1] eigenvalue range as a proved property of H(G), but the proof's only nontrivial step is the conditional 'If ∆J + ∆h ≤ 1.' Section III-A already imposes this as a boundary condition on the encoding, and the coefficients are introduced as normalized weights over the graph. The bound is therefore guaranteed by the chosen normalization rather than derived from the graph structure; the 'theorem' reduces to the construction choice. This is a secondary technical claim and does not support the central accuracy comparison.

full rationale

The central claims of the paper are empirical: EG-VQC is compared with PCA-VQC on MUTAG, PROTEIN, and ENZYME under a fixed 90:10 split, and the reported accuracies in Table I are external measurements rather than consequences of the encoding definitions. No parameter is fitted to the test set and then renamed as a prediction, and there are no load-bearing self-citations or imported uniqueness theorems. The encoding itself is constructed directly from the adjacency data via Eqs. (6)-(8), so no target result is hidden inside the input representation. The only circular-adjacent element is the appendix eigenvalue argument, where the desired range [−1,1] is effectively imposed through the weight normalization and the 'proof' supplies the conditional ΔJ+Δh≤1; this is a construction constraint, not an independent derivation, and it does not drive the classification results. The permutation-invariance concern raised in review is a correctness/representation issue about arbitrary vertex labelings, not a circularity of the derivation chain, so it does not increase the circularity score.

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

The method's central claims rest on several unstated modeling choices: the normalization of the weights J and h, the assumption that binary vertex indices are meaningful, and the hidden assumption that the Hamiltonian can be implemented as a shallow circuit. The eigenvalue range is imposed by a condition (ΔJ+Δh≤1) that the paper does not show it can always satisfy.

free parameters (3)
  • Normalization of edge and vertex weights (J_ij, h_i)
    Section III-A states that J_ij is 'derived from the normalized weight value over the entire graph' and h_i is the weighted degree, but no formula or normalization procedure is given. The boundary condition ΔJ+Δh≤1 in the appendix implicitly requires a normalization choice, which is an unstated modeling decision.
  • Number of qubits N = log(N_G)
    Section IV-B sets N = log(N_G), a hyperparameter chosen by the authors, not derived from first principles.
  • Circuit depth = 3, 5, or 7 layers
    Section IV-B: circuit depth is adjusted per dataset (3 layers for MUTAG, more for larger graphs), which is a hand-tuned hyperparameter.
assumptions (3)
  • domain assumption The encoded Hamiltonian H(G) can be implemented as a quantum circuit (or measured as an observable) using the described operations.
    Algorithm 1 and Section III-B indicate interactions are applied via quantum gates, but no explicit circuit decomposition is provided. The method's feasibility depends on this.
  • ad hoc to paper Binary vertex indices carry information useful for graph classification.
    The encoding in Eq. 6 uses only the vertex's integer index, not node attributes. This assumption is unstated and is questionable for the benchmark datasets.
  • ad hoc to paper The spectral range condition ΔJ+Δh≤1 is achievable and sufficient for compatibility with the VQC and cross-entropy loss.
    The appendix concludes eigenvalues are in [-1,1] only if ΔJ+Δh≤1, but the proof is incorrect, and the paper does not explain how normalization enforces this.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Tensor-Based Binary Graph Encoding for Variational Quantum Classifiers." pith.science (2026). https://pith.science/paper/2BNO7RKM

@misc{pith2026250114185,
  author       = {Pith},
  title        = {Pith review of: Tensor-Based Binary Graph Encoding for Variational Quantum Classifiers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2BNO7RKM}},
  note         = {Machine review of arXiv:2501.14185}
}
read the original abstract

Quantum computing has been a prominent research area for decades, inspiring transformative fields such as quantum simulation, quantum teleportation, and quantum machine learning (QML), which are undergoing rapid development. Within QML, hybrid classical-quantum algorithms like Quantum Neural Networks (QNNs) and Variational Quantum Classifiers (VQCs) have shown promise in leveraging quantum circuits and classical optimizers to classify classical data efficiently.Simultaneously, classical machine learning has made significant strides in graph classification, employing Graph Neural Networks (GNNs) to analyze systems ranging from large-scale structures like the Large Hadron Collider to molecular and biological systems like proteins and DNA. Combining the advancements in quantum computing and graph classification presents a unique opportunity to develop quantum algorithms capable of extracting features from graphs and performing their classification effectively. In this paper, we propose a novel quantum encoding framework for graph classification using VQCs. Unlike existing approaches such as PCA-VQC, which rely on dimensionality reduction techniques like Principal Component Analysis (PCA) and may lead to information loss, our method preserves the integrity of graph data. Furthermore, our encoding approach is optimized for Noise-Intermediate Scale Quantum (NISQ) devices, requiring a limited number of qubits while achieving comparable or superior classification performance to PCA-VQC. By constructing slightly more complex circuits tailored for graph encoding, we demonstrate that VQCs can effectively classify graphs within the constraints of current quantum hardware.

Figures

Figures reproduced from arXiv: 2501.14185 by the authors.

Figure 3
Figure 3. Comparison of loss over epochs between PCA-VQC and the proposed [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 12 canonical work pages

  1. [1]

    The theory of variational hybrid quantum-classical algorithms,

    J. R. McClean, J. Romero, R. Babbush, and A. Aspuru-Guzik, “The theory of variational hybrid quantum-classical algorithms,” New Journal of Physics , vol. 18, no. 2, p. 023023, Feb. 2016. [Online]. Available: http://dx.doi.org/10.1088/1367-2630/18/2/023023

    doi:10.1088/1367-2630/18/2/023023 2016

  2. [2]

    Approximate amplitude encoding in shallow parameterized quantum circuits and its application to financial market indicators,

    K. Nakaji, S. Uno, Y . Suzuki, R. Raymond, T. Onodera, T. Tanaka, H. Tezuka, N. Mitsuda, and N. Yamamoto, “Approximate amplitude encoding in shallow parameterized quantum circuits and its application to financial market indicators,” Physical Review Research , vol. 4, no. 2, May 2022. [Online]. Available: http://dx.doi.org/10.1103/ PhysRevResearch.4.023136

    work page 2022

  3. [3]

    A fast quantum mechanical algorithm for database search,

    L. K. Grover, “A fast quantum mechanical algorithm for database search,” 1996. [Online]. Available: https://arxiv.org/abs/quant-ph/ 9605043

    work page 1996

  4. [4]

    Polynomial-time algorithms for prime factorization and discrete logarithms on a quantum computer,

    P. W. Shor, “Polynomial-time algorithms for prime factorization and discrete logarithms on a quantum computer,” SIAM Journal on Computing, vol. 26, no. 5, pp. 1484–1509, Oct. 1997. [Online]. Available: http://dx.doi.org/10.1137/S0097539795293172

    doi:10.1137/s0097539795293172 1997

  5. [5]

    Graph neural networks: A review of methods and applications,

    J. Zhou, G. Cui, S. Hu, Z. Zhang, C. Yang, Z. Liu, L. Wang, C. Li, and M. Sun, “Graph neural networks: A review of methods and applications,”

    work page

  6. [6]

    Relational inductive biases, deep learning, and graph networks,

    P. W. Battaglia, J. B. Hamrick, V . Bapst, A. Sanchez-Gonzalez, V . Zambaldi, M. Malinowski, A. Tacchetti, D. Raposo, A. Santoro, R. Faulkner, C. Gulcehre, F. Song, A. Ballard, J. Gilmer, G. Dahl, A. Vaswani, K. Allen, C. Nash, V . Langston, C. Dyer, N. Heess, D. Wierstra, P. Kohli, M. Botvinick, O. Vinyals, Y . Li, and R. Pascanu, “Relational inductive b...

    arXiv 2018

  7. [7]

    Graph kernel neural networks,

    L. Cosmo, G. Minello, A. Bicciato, M. M. Bronstein, E. Rodol `a, L. Rossi, and A. Torsello, “Graph kernel neural networks,” IEEE Transactions on Neural Networks and Learning Systems, pp. 1–14, 2024. [Online]. Available: http://dx.doi.org/10.1109/TNNLS.2024.3400850

    arXiv 2024

  8. [8]

    State preparation based on quantum phase estimation

    J. Zhao, Y .-C. Wu, G.-C. Guo, and G.-P. Guo, “State preparation based on quantum phase estimation,” 2019. [Online]. Available: https://arxiv.org/abs/1912.05335

    work page Pith review arXiv 2019

Show all 37 references
  1. [9]

    Circuit-centric quantum classifiers,

    M. Schuld, A. Bocharov, K. M. Svore, and N. Wiebe, “Circuit-centric quantum classifiers,” Physical Review A , vol. 101, no. 3, Mar. 2020. [Online]. Available: http://dx.doi.org/10.1103/PhysRevA.101.032308

    2020 doi

  2. [10]

    Supervised learning with quantum-inspired tensor networks,

    E. M. Stoudenmire and D. J. Schwab, “Supervised learning with quantum-inspired tensor networks,” 2017. [Online]. Available: https://arxiv.org/abs/1605.05775

    2017 arXiv

  3. [11]

    Learning graph convolutional networks based on quantum vertex information propagation,

    L. Bai, Y . Jiao, L. Cui, L. Rossi, Y . Wang, P. S. Yu, and E. R. Hancock, “Learning graph convolutional networks based on quantum vertex information propagation,” IEEE Transactions on Knowledge and Data Engineering, vol. 35, no. 2, pp. 1747–1760, 2023

    2023

  4. [12]

    Robust data encodings for quantum classifiers,

    R. LaRose and B. Coyle, “Robust data encodings for quantum classifiers,” Phys. Rev. A , vol. 102, p. 032420, Sep 2020. [Online]. Available: https://link.aps.org/doi/10.1103/PhysRevA.102.032420

    2020 doi

  5. [13]

    Exponential data encoding for quantum supervised learning,

    S. Shin, Y . S. Teo, and H. Jeong, “Exponential data encoding for quantum supervised learning,” Phys. Rev. A , vol. 107, p. 012422, Jan

  6. [14]

    A survey on graph classification and link prediction based on gnn,

    X. Liu, J. Chen, and Q. Wen, “A survey on graph classification and link prediction based on gnn,” 2023. [Online]. Available: https://arxiv.org/abs/2307.00865

    2023 arXiv

  7. [15]

    Parameterized quantum circuits as machine learning models,

    M. Benedetti, E. Lloyd, S. Sack, and M. Fiorentini, “Parameterized quantum circuits as machine learning models,” Quantum Science and Technology, vol. 4, no. 4, p. 043001, Nov. 2019. [Online]. Available: http://dx.doi.org/10.1088/2058-9565/ab4eb5

    2019 doi

  8. [16]

    Evaluating analytic gradients on quantum hardware,

    M. Schuld, V . Bergholm, C. Gogolin, J. Izaac, and N. Killoran, “Evaluating analytic gradients on quantum hardware,” Physical Review A , vol. 99, no. 3, Mar. 2019. [Online]. Available: http: //dx.doi.org/10.1103/PhysRevA.99.032331

    2019 doi

  9. [17]

    General parameter- shift rules for quantum gradients,

    D. Wierichs, J. Izaac, C. Wang, and C. Y .-Y . Lin, “General parameter- shift rules for quantum gradients,” Quantum, vol. 6, p. 677, Mar. 2022. [Online]. Available: https://doi.org/10.22331/q-2022-03-30-677

    2022 doi

  10. [18]

    Quantum natural gradient,

    J. Stokes, J. Izaac, N. Killoran, and G. Carleo, “Quantum natural gradient,” Quantum, vol. 4, p. 269, May 2020. [Online]. Available: http://dx.doi.org/10.22331/q-2020-05-25-269

    2020 doi

  11. [19]

    Analysis of a complex of statistical variables into principal components,

    H. Hotelling, “Analysis of a complex of statistical variables into principal components,” Journal of Educational Psychology , vol. 24, no. 6, pp. 417–441, 1933. [Online]. Available: https://doi.org/10.1037/h0071325

    1933 doi

  12. [20]

    Hybrid quantum-classical classifier based on tensor network and variational quantum circuit,

    S. Y .-C. Chen, C.-M. Huang, C.-W. Hsing, and Y .-J. Kao, “Hybrid quantum-classical classifier based on tensor network and variational quantum circuit,” 2020. [Online]. Available: https://arxiv.org/abs/2011. 14651

    2020

  13. [21]

    Exact ising model simulation on a quantum computer,

    A. Cervera-Lierta, “Exact ising model simulation on a quantum computer,” Quantum, vol. 2, p. 114, Dec. 2018. [Online]. Available: http://dx.doi.org/10.22331/q-2018-12-21-114

    2018 doi

  14. [22]

    Weight normalization: A simple reparameterization to accelerate training of deep neural networks,

    T. Salimans and D. P. Kingma, “Weight normalization: A simple reparameterization to accelerate training of deep neural networks,”

  15. [23]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,”

  16. [24]

    Goodfellow, Y

    I. Goodfellow, Y . Bengio, and A. Courville, Deep Learning. MIT Press, 2016, http://www.deeplearningbook.org

    2016

  17. [25]

    Pennylane: Automatic differentiation of hybrid quantum-classical computations,

    V . B. et. al, “Pennylane: Automatic differentiation of hybrid quantum-classical computations,” 2022. [Online]. Available: https: //arxiv.org/abs/1811.04968

    2022 arXiv

  18. [26]

    Strawberry Fields: A Software Platform for Photonic Quantum Computing,

    N. Killoran, J. Izaac, N. Quesada, V . Bergholm, M. Amy, and C. Weedbrook, “Strawberry Fields: A Software Platform for Photonic Quantum Computing,” Quantum, vol. 3, p. 129, Mar. 2019. [Online]. Available: https://doi.org/10.22331/q-2019-03-11-129

    2019 doi

  19. [27]

    Applications of near-term photonic quantum computers: software and algorithms,

    T. R. Bromley, J. M. Arrazola, S. Jahangiri, J. Izaac, N. Quesada, A. D. Gran, M. Schuld, J. Swinarton, Z. Zabaneh, and N. Killoran, “Applications of near-term photonic quantum computers: software and algorithms,” Quantum Science and Technology, vol. 5, no. 3, p. 034010, may 2...

    2020 doi

  20. [28]

    A survey on graph neural networks,

    A. Debnath, S. Soni, and C. K. Reddy, “A survey on graph neural networks,” Journal of Machine Learning Research , vol. 19, no. 1, pp. 99–124, 2018

    2018

  21. [29]

    Protein classification with graph kernels,

    K. M. Borgwardt, N. Kr ¨amer, R. Si, D. Hans, and B. Sch¨olkopf, “Protein classification with graph kernels,” Bioinformatics, vol. 21, no. 1, pp. 105–116, 2005

    2005

  22. [30]

    Enzyme classification in terms of graph-based representations,

    T. Schwede, J. Kopp, G. Benoit, F. Kepes, W. Thiel, G. Deleage, and F. L¨ohr, “Enzyme classification in terms of graph-based representations,” Journal of Computational Chemistry, vol. 28, no. 2, pp. 292–301, 2007

    2007

  23. [31]

    Array programming with numpy,

    C. R. Harris, K. J. Millman, S. J. van der Walt, R. Gommers, P. Virtanen, D. Cournapeau, E. Wieser, J. Taylor, S. Berg, N. J. Smith, R. Kern, M. Picus, S. Hoyer, M. H. van Kerkwijk, M. Brett, A. Haldane, J. F. del R ´ıo, M. Wiebe, P. Peterson, P. G ´erard-Marchant, K. Sheppard...

    2020 doi

  24. [32]

    Scikit-learn: Machine learning in python,

    F. Pedregosa, G. Varoquaux, A. Gramfort, V . Michel, B. Thirion, O. Grisel, M. Blondel, A. M ¨uller, J. Nothman, G. Louppe, P. Prettenhofer, R. Weiss, V . Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and ´Edouard Duchesnay, “Scikit-learn: Machine le...

    2018 arXiv

  25. [33]

    Graph isomorphism u-net,

    A. Amouzad, Z. Dehghanian, S. Saravani, M. Amirmazlaghani, and B. Roshanfekr, “Graph isomorphism u-net,” Expert Systems with Applications , vol. 236, p. 121280, 2024. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0957417423017827

    2024

  26. [2016]

    Available: https://arxiv.org/abs/1602.07868

    [Online]. Available: https://arxiv.org/abs/1602.07868

    arXiv

  27. [2017]

    Available: https://arxiv.org/abs/1412.6980

    [Online]. Available: https://arxiv.org/abs/1412.6980

    arXiv

  28. [2021]

    Available: https://arxiv.org/abs/1812.08434

    [Online]. Available: https://arxiv.org/abs/1812.08434

    arXiv

  29. [2023]

    Available: https://link.aps.org/doi/10.1103/PhysRevA

    [Online]. Available: https://link.aps.org/doi/10.1103/PhysRevA. 107.012422

    doi

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.