Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Quantum Kernel-Based Long Short-term Memory for Climate Time-Series Forecasting

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Replacing the linear transformations in LSTM cells with quantum kernel evaluations improves AQI forecasting accuracy and cuts trainable parameters from 1,873 to 209.

desk verdict A new AQI benchmark, but the QK-LSTM gate equations are internally inconsistent so the reported results are unsupported. read the letter →

arxiv 2412.08851 v1 pith:D4HRBPGB submitted 2024-12-12 quant-ph cs.AIcs.LG

Yu-Chao Hsu , Nan-Yow Chen , Tai-Yu Li , Po-Heng (Henry) Lee , Kuan-Cheng Chen This is my paper · ORCID
classification quant-phcs.AIcs.LG
keywords quantummachinelearningkernelmethodslongshort-termmemorytime-seriesforecastingairqualityindexmodelcompressionNISQhybridquantum-classicalcomputing
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 QK-LSTM, an LSTM variant in which the linear transformations inside the forget, input, candidate, and output gates are replaced by weighted sums of quantum kernel evaluations. On an Air Quality Index forecasting task for Bengaluru, it reports lower error on every metric than a classical LSTM: RMSE 9.20 versus 15.94, MAE 7.15 versus 11.07, MAPE 9.14% versus 13.32%, and $R^2$ 0.84 versus 0.78. It also reports a large parameter reduction, 209 trainable parameters versus 1,873, which it attributes to the expressive power of quantum feature spaces. The motivation is that quantum kernels can capture nonlinear dependencies with fewer parameters than classical recurrent networks, making the model attractive for resource-constrained and near-term quantum settings.

What carries the argument

The load-bearing object is the quantum kernel function $k(v_t, v_j)$, defined as the squared overlap of two data-encoded quantum states. A parameterized circuit $U(v)$ with Hadamard layers, $R_y/R_z$ rotation encoding, and nearest-neighbor CNOT entangling gates maps each concatenated input $v_t=[h_{t-1};x_t]$ to a state $|\phi(v_t)\rangle$; the kernel is the measured probability of returning to $|0\rangle^{\otimes n}$ after applying $U^\dagger(v_j)U(v_t)$. This single scalar similarity value replaces the matrix-vector products in each LSTM gate, and the trainable weights $\alpha_j$ combine the kernel values against a set of reference vectors. The efficiency the paper reports comes from this replacement: the quantum kernel provides the nonlinear feature map, so far fewer classical parameters are needed.

What would settle it

Run the QK-LSTM on the same Bengaluru dataset with the missing vector construction specified explicitly, and compare the reported RMSE 9.20 against a classical LSTM trained with 209 parameters under identical conditions; if the architecture cannot be instantiated or the accuracy gap disappears, the central claim fails.

Watch

Extended reading notes

Core claim

The central claim is that quantum kernel methods can substitute for learned linear maps in an LSTM cell without losing temporal modeling capacity, and that on the tested AQI dataset this substitution improves accuracy while shrinking the model. The paper defines a QK-LSTM cell where the concatenated input $v_t = [h_{t-1}; x_t]$ is compared against a set of reference vectors through quantum kernels $k(v_t, v_j) = |\langle \phi(v_t)|\phi(v_j)\rangle|^2$, and each gate becomes $\sigma(\sum_j \alpha_j k(v_t, v_j) + b)$ (or tanh for the candidate). Training uses backpropagation through time for the classical weights and the parameter-shift rule for circuit parameters. The empirical section reports that QK-LSTM outperforms LSTM on RMSE, MAE, MAPE, and $R^2$ while using about one-ninth the parameters, and argues this makes the architecture suitable for NISQ hardware and for simulation on GPUs.

Load-bearing premise

The load-bearing premise is that the QK-LSTM cell is a well-defined trainable architecture; the paper's gate equations produce scalar values where the cell update needs vectors, and it never states how this is resolved or how many reference vectors are used.

Editorial extensions

If this is right

  • If the reported results hold, QK-LSTM offers a concrete recipe for compressing recurrent models: replace linear gate transformations with kernel expansions, trading parameter count for kernel evaluations.
  • The architecture can be run partially on classical hardware by simulating the quantum kernel on GPUs, which eases deployment on near-term quantum devices.
  • The same cell design could be applied to other climate time-series tasks, such as temperature, precipitation, flood, and greenhouse-gas forecasting, since the method is not specific to AQI.
  • Because the quantum kernel is computed from a 4-qubit circuit, the approach is designed to stay within NISQ resource limits, unlike deeper variational circuits.
  • Fewer trainable parameters reduce overfitting risk and training cost, which matters for resource-constrained environmental monitoring deployments.

Reading between the lines

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

  • A natural next test is to replace the quantum kernel with a classical kernel of the same parameter count; if a classical kernel matches the reported RMSE, the advantage is kernel-based rather than quantum-specific.
  • The scalar-to-vector gap in the gate equations suggests a concrete repair: define each hidden unit's gate value by giving that unit its own set of reference vectors, or broadcast the scalar kernel sum across the hidden dimension.
  • If the parameter reduction generalizes, the design principle extends beyond LSTMs: any learned matrix in a recurrent or feedforward cell could be replaced by fixed similarity evaluations against a small reference set, with only combination weights trained.
  • The paper's NISQ framing implies a cheap benchmark: a four-qubit kernel is small enough to simulate classically, so a GPU tensor-network reproduction of the reported AQI results would test the quantum-hardware claim before any device run.
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. The manuscript proposes QK-LSTM, a hybrid quantum-classical recurrent architecture in which the linear transformations inside an LSTM cell are replaced by weighted sums of quantum kernel evaluations. The model is applied to daily Air Quality Index (AQI) forecasting for Bengaluru, India, using eleven selected pollutant features. The authors report that QK-LSTM outperforms a classical LSTM on RMSE (9.20 vs 15.94), MAE (7.15 vs 11.07), MAPE (9.14% vs 13.32%), and R² (0.84 vs 0.78), while reducing trainable parameters from 1,873 to 209. The paper also discusses scalability and NISQ-era practicality, including GPU-based classical simulation of the quantum kernels.

Significance. If the central claims were substantiated, the paper would offer a useful data point for hybrid quantum-classical sequence modeling, particularly in showing that kernel-based replacements for learned linear projections can reduce parameter counts without losing accuracy on a real forecasting task. The authors are also to be credited for using standard evaluation metrics, documenting the data preprocessing pipeline, and clearly stating the LSTM baseline equations. However, the central contribution is not currently assessable: the defining equations do not specify a constructible model, the reported parameter count is inconsistent with those equations, and the empirical comparison rests on a single run with no error bars, no described split, and no classical kernel baseline. The paper also contains no code or data release that would allow the reported numbers to be reproduced. The significance of the claimed result therefore cannot be established from the manuscript as written.

major comments (4)
  1. [III-B2, Eqs. (6a)-(6f)] The gate equations define each of f_t, i_t, C_tilde_t, and o_t as a scalar, because each is a weighted sum of the scalar kernels k(v_t, v_j) plus a bias, followed by an activation. Equation (6d) then requires elementwise multiplication of these scalars with the hidden-state vectors C_{t-1} and C_tilde_t, which are stated to have dimension m = 16 in Table II. The paper never specifies how the scalar gate values are lifted to 16-dimensional gate vectors, how many reference vectors N are used, how they are selected, or whether they are trainable. As written, the forward pass cannot be executed for the reported configuration. In addition, Eq. (5) concatenates h_{t-1} (dimension 16) with x_t (dimension 11) to form v_t of dimension 27, while the quantum circuit in Eqs. (9)-(12) uses n = 4 qubits. The encoding map U_enc in Eq. (10) has only 2n = 8 rotation angles for the whole 27-dimensional input, and no dimensionality reduction or component-selection procedure is described. This is a load-bearing gap: the reported architecture is not implementable as stated.
  2. [III-B4 and Table II] The reported parameter count is inconsistent with the stated model. If the only trainable parameters are the alpha weights and biases in Eqs. (6a)-(6e), the total is 4N + 4 (four gates, each with N kernel weights and one bias). Setting 4N + 4 = 209 gives N = 51.25, which is not an integer. If, alternatively, the quantum circuit parameters are trainable, then the paper must say so and explain the parameter-shift rule of Eq. (15) in that context, but Eqs. (9)-(12) define theta_k and phi_k as functions of the input v, with no free variational parameters. The parameter-shift rule is therefore not applicable to any circuit parameter described in the paper. The claimed 209-parameter model cannot be reconciled with the equations, and the model-compression comparison in Table II is not verifiable.
  3. [IV-A and Table I] The empirical evaluation is not sufficient to support the central claim that QK-LSTM outperforms LSTM on AQI forecasting. The paper reports a single set of metrics with no error bars, no number of random seeds, and no description of the train/validation/test split. It also does not state whether the linear interpolation in Section II-C was performed before or after splitting the time series, which matters for data leakage and for the validity of the reported test performance. Furthermore, the comparison includes only a plain LSTM; there is no classical kernel-based LSTM baseline (for example, an LSTM whose gates use a classical RBF kernel expansion), so the reported improvement cannot be attributed to the quantum feature map rather than to the kernel expansion or to the reduced parameter count. Without these controls, the numbers in Table I, even if accurately computed, do not establish the paper's central claim.
  4. [Abstract, Section V, and Table I text] The paper repeatedly claims 'computational efficiency' and 'maintaining computational efficiency' (Abstract and Section IV-A.5), but it provides no runtime measurements, no wall-clock time, no complexity analysis of the quantum kernel evaluation, and no comparison with a variational-quantum-circuit LSTM baseline. Section V discusses block-encoding and GPU-based tensor-network simulation as theoretical possibilities, but these are not connected to the concrete implementation used for Table I, and no resource counts are given. The abstract's efficiency claim is therefore unsupported by the experiments.
minor comments (5)
  1. [Abstract and Section I] 'Long short-memory' should be 'Long Short-Term Memory' throughout, including the abstract and the opening sentence of Section I.
  2. [Eq. (2)] Equation (2) has a garbled display of the Z-score formula; the fraction and the threshold condition should be typeset clearly so that |(x_i - mu)/sigma| > gamma is directly readable.
  3. [Fig. 2] Figure 2 caption uses a parameter w in U(x_t, w), while the main text defines the feature map as U(v) with no such parameter; the notation should be unified.
  4. [Table II and Section IV-B] The hyperparameter table does not list N (the number of reference vectors), the output-layer size, or the quantum encoding details, making the reported parameter total uninterpretable.
  5. [IV-A.3, Eq. (22)] The MAPE definition divides by y_i; the paper should state how zero or near-zero AQI values are handled in this metric.

Circularity Check

1 steps flagged · score 2.0 of 10

Minor self-citation in model attribution, but the AQI accuracy claim is a direct empirical benchmark and does not reduce to prior work or to fitted inputs.

  1. other [Section V (Scalability and Practicality), citing Ref. [23]]
    "The QK-LSTM model has been successfully applied to tasks such as Part-of-Speech (POS) tagging task [23], demonstrating that the model achieved competitive accuracy while significantly reducing the number of trainable parameters compared to classical LSTM models."

    This is a self-citation to the authors' own prior arXiv paper [23] used as evidence that QK-LSTM is an established, effective architecture. It is not load-bearing in the circularity sense: the central AQI result is a fresh direct measurement on CPCB Bengaluru data reported in Table I, and the model equations are restated in Eqs. (5)-(6). No reported metric is constructed from a fitted parameter by definition, and no uniqueness or ansatz result is imported from the cited paper. The self-citation is minor attribution, not a reduction of the headline claim.

full rationale

No substantive circularity. The derivation chain is: quantum kernels (Eq. (7)) are used in place of linear transformations in LSTM gates (Eqs. (6a)-(6e)), training minimizes MSE (Eq. (14)), and performance is benchmarked directly against a classical LSTM in Table I. None of these steps defines the reported RMSE, MAE, MAPE, or R² in terms of a fitted constant, and no prediction is equivalent to an input by construction. The model is attributed to the authors' prior work [23], and Section V cites [23] for a previous POS-tagging demonstration, but the equations are also stated in the present paper and the AQI experiment is new data. Thus the self-citation is minor rather than load-bearing. The paper's serious weaknesses are implementability and correctness issues—scalar gate equations vs. vector cell state, unspecified reference vectors N, and the inconsistent 209-parameter count—which should be treated as correctness risks, not circularity.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The model is empirical and uses several hand-picked hyperparameters and an unspecified reference set. The axiomatic burden includes the unproven expressiveness and efficiency of the quantum kernel, plus a domain assumption about the AQI definition. No new physical entities are proposed.

free parameters (5)
  • Number of reference vectors N = not specified
    Eqs. (6a)-(6e) sum over N reference vectors, but the paper never states N or how the reference set is chosen. This is central to the architecture and the reported parameter count.
  • Number of qubits = 4
    Set by hand for the quantum kernel circuit; no sensitivity analysis is reported.
  • Outlier threshold gamma = 3
    Chosen as 'common practice'; affects the training set composition.
  • Feature exclusion = Xylene removed, 11 features kept
    Exclusion based on missing data rate and 'potential bias'; this data selection affects the comparison.
  • Hyperparameters (epochs, lr, hidden units, batch size, sequence length) = 20, 0.001, 16, 1, 3
    Identical for both models, but no ablation is shown; the comparison could be sensitive to these choices.
assumptions (5)
  • standard math Standard LSTM update equations (4a)-(4f) are valid.
    Background model taken from the literature.
  • domain assumption AQI is the max of pollutant sub-indices, as in Eq. (1).
    Defines the target variable; taken from CPCB breakpoint methodology.
  • ad hoc to paper The quantum feature map U(v) with Ry/Rz encoding and CNOT entanglement is expressive enough to replace learned linear projections in LSTM gates.
    No theoretical or empirical evidence beyond the single benchmark; underlies the central architectural claim.
  • ad hoc to paper The kernel expansion in Eqs. (6a)-(6c) can be trained end-to-end with BPTT and the parameter-shift rule.
    The circuit parameters theta and phi in Eq. (10) are functions of the input v, so it is unclear which parameters are trainable; the parameter-shift rule may not apply as written.
  • domain assumption Quantum kernel inner products can be computed efficiently enough to yield computational advantage on NISQ hardware.
    Invoked in Sections I and V without runtime data or complexity analysis.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Quantum Kernel-Based Long Short-term Memory for Climate Time-Series Forecasting." pith.science (2026). https://pith.science/paper/D4HRBPGB

@misc{pith2026241208851,
  author       = {Pith},
  title        = {Pith review of: Quantum Kernel-Based Long Short-term Memory for Climate Time-Series Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D4HRBPGB}},
  note         = {Machine review of arXiv:2412.08851}
}
read the original abstract

We present the Quantum Kernel-Based Long short-memory (QK-LSTM) network, which integrates quantum kernel methods into classical LSTM architectures to enhance predictive accuracy and computational efficiency in climate time-series forecasting tasks, such as Air Quality Index (AQI) prediction. By embedding classical inputs into high-dimensional quantum feature spaces, QK-LSTM captures intricate nonlinear dependencies and temporal dynamics with fewer trainable parameters. Leveraging quantum kernel methods allows for efficient computation of inner products in quantum spaces, addressing the computational challenges faced by classical models and variational quantum circuit-based models. Designed for the Noisy Intermediate-Scale Quantum (NISQ) era, QK-LSTM supports scalable hybrid quantum-classical implementations. Experimental results demonstrate that QK-LSTM outperforms classical LSTM networks in AQI forecasting, showcasing its potential for environmental monitoring and resource-constrained scenarios, while highlighting the broader applicability of quantum-enhanced machine learning frameworks in tackling large-scale, high-dimensional climate datasets.

Figures

Figures reproduced from arXiv: 2412.08851 by the authors.

Figure 1
Figure 1. Schematic representation of a standard classical LSTM network [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Overview of the QK-LSTM architecture. (a) The QK-LSTM cell combines quantum kernel computations with the traditional LSTM framework, [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Comparative analysis of air quality prediction performance using classical LSTM and QK-LSTM models. Panels (a) and (b) depict the overall predictive [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. A Quantum-Classical Hybrid Framework for Multivariate Time-Series Forecasting Complexity-Fidelity Trade-offs and Limitations

    cs.IR 2026-07 reject novelty 4.0 of 10

    A trainable 6-qubit-per-channel variational circuit plus linear readout roughly matches transformer forecasters at short horizons on ETT/Weather/Electricity benchmarks, but the reported setup does not match the paper'...

Reference graph

Works this paper leans on

45 extracted references · 24 canonical work pages · cited by 1 Pith paper

  1. [23]

    Quantum Kernel-Based Long Short-term Memory

    Y .-C. Hsu, T.-Y . Li, and K.-C. Chen, “Quantum kernel-based long short- term memory,” arXiv preprint arXiv:2411.13225 , 2024

    work page Pith review arXiv 2024

  2. [1]

    Current and future global climate impacts resulting from covid-19,

    P. M. Forster, H. I. Forster, M. J. Evans, M. J. Gidden, C. D. Jones, C. A. Keller, R. D. Lamboll, C. L. Qu ´er´e, J. Rogelj, D. Rosen, et al. , “Current and future global climate impacts resulting from covid-19,” Nature Climate Change , vol. 10, no. 10, pp. 913–919, 2020

    work page 2020

  3. [2]

    Global water resources affected by human interventions and climate change,

    I. Haddeland, J. Heinke, H. Biemans, S. Eisner, M. Fl ¨orke, N. Hanasaki, M. Konzmann, F. Ludwig, Y . Masaki, J. Schewe, et al. , “Global water resources affected by human interventions and climate change,” Proceedings of the National Academy of Sciences , vol. 111, no. 9, pp. 3251–3256, 2014

    work page 2014

  4. [3]

    Impact of climate policy uncertainty on traditional energy and green markets: Evidence from time-varying granger tests,

    X. Ren, J. Li, F. He, and B. Lucey, “Impact of climate policy uncertainty on traditional energy and green markets: Evidence from time-varying granger tests,” Renewable and Sustainable Energy Reviews , vol. 173, p. 113058, 2023

    work page 2023

  5. [4]

    Recurrent neural networks,

    L. R. Medsker, L. Jain, et al., “Recurrent neural networks,” Design and Applications, vol. 5, no. 64-67, p. 2, 2001

    work page 2001

  6. [5]

    Long Short-Term Memory,

    S. Hochreiter and J. Schmidhuber, “Long Short-Term Memory,” Neural Computation, vol. 9, pp. 1735–1780, 11 1997

    1997

  7. [6]

    A review of recurrent neural networks: Lstm cells and network architectures,

    Y . Yu, X. Si, C. Hu, and J. Zhang, “A review of recurrent neural networks: Lstm cells and network architectures,” Neural computation , vol. 31, no. 7, pp. 1235–1270, 2019

    2019

  8. [7]

    Power of data in quantum machine learning,

    H.-Y . Huang, M. Broughton, M. Mohseni, R. Babbush, S. Boixo, H. Neven, and J. R. McClean, “Power of data in quantum machine learning,” Nature communications, vol. 12, no. 1, p. 2631, 2021

    work page 2021

Show all 45 references
  1. [8]

    Machine learning of high dimensional data on a noisy quantum processor,

    E. Peters, J. Caldeira, A. Ho, S. Leichenauer, M. Mohseni, H. Neven, P. Spentzouris, D. Strain, and G. N. Perdue, “Machine learning of high dimensional data on a noisy quantum processor,” npj Quantum Information, vol. 7, no. 1, p. 161, 2021

    2021

  2. [9]

    Quantum machine learning,

    J. Biamonte, P. Wittek, N. Pancotti, P. Rebentrost, N. Wiebe, and S. Lloyd, “Quantum machine learning,” Nature, vol. 549, no. 7671, pp. 195–202, 2017

    2017

  3. [10]

    Shedding light on the future: Exploring quantum neural networks through optics,

    S. Yu, Z. Jia, A. Zhang, E. Mer, Z. Li, V . Crescimanna, K.-C. Chen, R. B. Patel, I. A. Walmsley, and D. Kaszlikowski, “Shedding light on the future: Exploring quantum neural networks through optics,” Advanced Quantum Technologies, p. 2400074, 2024

    2024

  4. [11]

    Compressedmediq: Hybrid quantum machine learning pipeline for high-dimentional neu- roimaging data,

    K.-C. Chen, Y .-T. Li, T.-Y . Li, and C.-Y . Liu, “Compressedmediq: Hybrid quantum machine learning pipeline for high-dimentional neu- roimaging data,” arXiv preprint arXiv:2409.08584 , 2024

    2024 arXiv

  5. [12]

    Quantum computing for climate resilience and sustainability challenges,

    K. T. M. Ho, K.-C. Chen, L. Lee, F. Burt, S. Yu, et al. , “Quantum computing for climate resilience and sustainability challenges,” arXiv preprint arXiv:2407.16296, 2024

    2024 arXiv

  6. [13]

    Quantum machine learning in climate change and sustainability: A short review,

    A. Nammouchi, A. Kassler, and A. Theocharis, “Quantum machine learning in climate change and sustainability: A short review,” Quantum, vol. 1, p. 1, 2023

    2023

  7. [14]

    Quantum long short-term memory,

    S. Y .-C. Chen, S. Yoo, and Y .-L. L. Fang, “Quantum long short-term memory,” in ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pp. 8622–8626, IEEE, 2022

    2022

  8. [15]

    Quantum-train long short- term memory: Application on flood prediction problem,

    C.-H. A. Lin, C.-Y . Liu, and K.-C. Chen, “Quantum-train long short- term memory: Application on flood prediction problem,” arXiv preprint arXiv:2407.08617, 2024

    2024 arXiv

  9. [16]

    Variational quantum circuits for deep reinforcement learning,

    S. Y .-C. Chen, C.-H. H. Yang, J. Qi, P.-Y . Chen, X. Ma, and H.-S. Goan, “Variational quantum circuits for deep reinforcement learning,” IEEE access, vol. 8, pp. 141007–141024, 2020

    2020

  10. [17]

    Quantum computing in the nisq era and beyond,

    J. Preskill, “Quantum computing in the nisq era and beyond,” Quantum, vol. 2, p. 79, 2018

    2018

  11. [18]

    Quantum classifier with tailored quantum kernel,

    C. Blank, D. K. Park, J.-K. K. Rhee, and F. Petruccione, “Quantum classifier with tailored quantum kernel,” npj Quantum Information , vol. 6, no. 1, p. 41, 2020

    2020

  12. [19]

    Quantum support vector machine for big data classification,

    P. Rebentrost, M. Mohseni, and S. Lloyd, “Quantum support vector machine for big data classification,” Physical review letters , vol. 113, no. 13, p. 130503, 2014

    2014

  13. [20]

    Experimental realization of a quantum support vector machine,

    Z. Li, X. Liu, N. Xu, and J. Du, “Experimental realization of a quantum support vector machine,” Physical review letters , vol. 114, no. 14, p. 140504, 2015

    2015

  14. [21]

    Variational quantum classifier for binary classification: Real vs synthetic dataset,

    D. Maheshwari, D. Sierra-Sosa, and B. Garcia-Zapirain, “Variational quantum classifier for binary classification: Real vs synthetic dataset,” IEEE access, vol. 10, pp. 3705–3715, 2021

    2021

  15. [22]

    The complexity of quantum support vector machines,

    G. Gentinetta, A. Thomsen, D. Sutter, and S. Woerner, “The complexity of quantum support vector machines,” Quantum, vol. 8, p. 1225, 2024

    2024

  16. [24]

    Hybrid quantum edge computing network,

    L. Ma and L. Ding, “Hybrid quantum edge computing network,” in Quantum Communications and Quantum Imaging XX , vol. 12238, pp. 83–93, SPIE, 2022

    2022

  17. [25]

    Limitations in quantum computing from resource constraints,

    M. Fellous-Asiani, J. H. Chai, R. S. Whitney, A. Auff `eves, and H. K. Ng, “Limitations in quantum computing from resource constraints,” PRX Quantum, vol. 2, no. 4, p. 040335, 2021

    2021

  18. [26]

    Consensus-based distributed quantum kernel learning for speech recognition,

    K.-C. Chen, W. Ma, and X. Xu, “Consensus-based distributed quantum kernel learning for speech recognition,” arXiv preprint arXiv:2409.05770, 2024

    2024 arXiv

  19. [27]

    cutn-qsvm: cutensornet-accelerated quantum support vector machine with cuquantum sdk,

    K.-C. Chen, T.-Y . Li, Y .-Y . Wang, S. See, C.-C. Wang, R. Willie, N.-Y . Chen, A.-C. Yang, and C.-Y . Lin, “cutn-qsvm: cutensornet-accelerated quantum support vector machine with cuquantum sdk,” arXiv preprint arXiv:2405.02630, 2024

    2024 arXiv

  20. [28]

    Generalised circuit partitioning for distributed quantum computing,

    F. Burt, K.-C. Chen, and K. Leung, “Generalised circuit partitioning for distributed quantum computing,” arXiv preprint arXiv:2408.01424 , 2024

    2024 arXiv

  21. [29]

    Air quality prediction: Big data and machine learning approaches,

    G. K. Kang, J. Z. Gao, S. Chiao, S. Lu, and G. Xie, “Air quality prediction: Big data and machine learning approaches,” Int. J. Environ. Sci. Dev, vol. 9, no. 1, pp. 8–16, 2018

    2018

  22. [30]

    Comparison of the revised air quality index with the psi and aqi indices,

    W.-L. Cheng, Y .-S. Chen, J. Zhang, T. Lyons, J.-L. Pai, and S.-H. Chang, “Comparison of the revised air quality index with the psi and aqi indices,” Science of the Total Environment , vol. 382, no. 2-3, pp. 191– 198, 2007

    2007

  23. [31]

    Air quality index (aqi) data

    Central Pollution Control Board, “Air quality index (aqi) data.” https: //cpcb.nic.in/, 2020. Accessed: 2020-11-04

    2020

  24. [32]

    Comparison of spatial interpolation methods for the estimation of air quality data,

    D. W. Wong, L. Yuan, and S. A. Perlin, “Comparison of spatial interpolation methods for the estimation of air quality data,” Journal of Exposure Science & Environmental Epidemiology , vol. 14, no. 5, pp. 404–415, 2004

    2004

  25. [33]

    Enhanced lstm for natural language inference,

    Q. Chen, X. Zhu, Z. Ling, S. Wei, H. Jiang, and D. Inkpen, “Enhanced lstm for natural language inference,” arXiv preprint arXiv:1609.06038 , 2016

    2016 arXiv

  26. [34]

    An improved lstm structure for natural language processing,

    L. Yao and Y . Guan, “An improved lstm structure for natural language processing,” in 2018 IEEE international conference of safety produce informatization (IICSPI), pp. 565–569, IEEE, 2018

    2018

  27. [35]

    Vision-lstm: xlstm as generic vision backbone,

    B. Alkin, M. Beck, K. P ¨oppel, S. Hochreiter, and J. Brandstet- ter, “Vision-lstm: xlstm as generic vision backbone,” arXiv preprint arXiv:2406.04303, 2024

    2024 arXiv

  28. [36]

    Multimodal fusion of optimized gru–lstm with self-attention layer for hydrological time series forecasting,

    H. C. Kilinc, S. Apak, F. Ozkan, M. E. Ergin, and A. Yurtsever, “Multimodal fusion of optimized gru–lstm with self-attention layer for hydrological time series forecasting,” Water Resources Management , pp. 1–18, 2024

    2024

  29. [37]

    A self-organizing deep network architecture designed based on lstm network via elitism- driven roulette-wheel selection for time-series forecasting,

    K. Zhou, S.-K. Oh, W. Pedrycz, J. Qiu, and K. Seo, “A self-organizing deep network architecture designed based on lstm network via elitism- driven roulette-wheel selection for time-series forecasting,” Knowledge- Based Systems, vol. 289, p. 111481, 2024

    2024

  30. [38]

    Stock price prediction with optimized deep lstm network with artificial rabbits optimization algorithm,

    B. G ¨ulmez, “Stock price prediction with optimized deep lstm network with artificial rabbits optimization algorithm,” Expert Systems with Applications, vol. 227, p. 120346, 2023

    2023

  31. [39]

    Pennylane: Automatic differentiation of hybrid quantum-classical com- putations. arxiv 2018,

    V . Bergholm, J. Izaac, M. Schuld, C. Gogolin, S. Ahmed, V . Ajith, M. Alam, G. Alonso-Linaje, B. AkashNarayanan, A. Asadi, et al. , “Pennylane: Automatic differentiation of hybrid quantum-classical com- putations. arxiv 2018,” arXiv preprint arXiv:1811.04968 , 2018

    2018 arXiv

  32. [40]

    Quantum machine learning in feature hilbert spaces,

    M. Schuld and N. Killoran, “Quantum machine learning in feature hilbert spaces,” Physical review letters, vol. 122, no. 4, p. 040504, 2019

    2019

  33. [41]

    Supervised learning with quantum- enhanced feature spaces,

    V . Havl ´ıˇcek, A. D. C ´orcoles, K. Temme, A. W. Harrow, A. Kandala, J. M. Chow, and J. M. Gambetta, “Supervised learning with quantum- enhanced feature spaces,” Nature, vol. 567, no. 7747, pp. 209–212, 2019

    2019

  34. [42]

    Entanglement and tensor networks for supervised image classification,

    J. Martyn, G. Vidal, C. Roberts, and S. Leichenauer, “Entanglement and tensor networks for supervised image classification,” arXiv preprint arXiv:2007.06082, 2020

    2007 arXiv

  35. [43]

    Quantum support vector machines for classification and regression on a trapped-ion quantum computer,

    T. Suzuki, T. Hasebe, and T. Miyazaki, “Quantum support vector machines for classification and regression on a trapped-ion quantum computer,” Quantum Machine Intelligence , vol. 6, no. 1, p. 31, 2024

    2024

  36. [44]

    Quantum singular value transformation and beyond: exponential improvements for quantum matrix arithmetics,

    A. Gily ´en, Y . Su, G. H. Low, and N. Wiebe, “Quantum singular value transformation and beyond: exponential improvements for quantum matrix arithmetics,” in Proceedings of the 51st Annual ACM SIGACT Symposium on Theory of Computing , pp. 193–204, 2019

    2019

  37. [45]

    Simulating quantum computation by con- tracting tensor networks,

    I. L. Markov and Y . Shi, “Simulating quantum computation by con- tracting tensor networks,” SIAM Journal on Computing , vol. 38, no. 3, pp. 963–981, 2008

    2008

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.