Kernel Alignment for Quantum Support Vector Machines Using Genetic Algorithms

Floyd M. Creevey; Jamie A. Heredge; Lloyd C. L. Hollenberg; Martin E. Sevior

arxiv: 2312.01562 · v3 · pith:VOVVKRKTnew · submitted 2023-12-04 · 🪐 quant-ph

Kernel Alignment for Quantum Support Vector Machines Using Genetic Algorithms

Floyd M. Creevey , Jamie A. Heredge , Martin E. Sevior , Lloyd C. L. Hollenberg This is my paper

Pith reviewed 2026-05-24 05:08 UTC · model grok-4.3

classification 🪐 quant-ph

keywords kernel alignmentquantum support vector machinesgenetic algorithmsdata encoding circuitsquantum kernelsQSVMquantum machine learning

0 comments

The pith

A genetic algorithm can design data-encoding circuits for quantum support vector machine kernels that match or exceed the accuracy of standard quantum and classical kernels.

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

The paper shows that genetic algorithms can automatically choose the gate sequences inside QSVM data-encoding circuits. It compares supervised and unsupervised loss functions as guides for this search and tests the resulting kernels on both binary and multi-class datasets. Benchmarking indicates that the automatically generated circuits perform as well as or better than existing classical kernels and other quantum kernels. The authors also report a positive correlation between test-set accuracy and the entropy of the quantum kernel. This approach aims to remove much of the manual trial-and-error that currently limits QSVM applications.

Core claim

We leverage the GASP framework for gate sequence selection in QSVM kernel circuits. We explore supervised and unsupervised kernel loss functions' impact on encoding circuit optimisation and evaluate them on diverse datasets for binary and multiple-class scenarios. Benchmarking against classical and quantum kernels reveals GA-generated circuits matching or surpassing standard techniques. We analyse the relationship between test accuracy and quantum kernel entropy, with results indicating a positive correlation.

What carries the argument

The GASP genetic algorithm framework applied to select gate sequences for QSVM data encoding circuits, guided by supervised or unsupervised kernel loss functions to steer optimization.

If this is right

GA-generated circuits reduce the manual trial-and-error needed to create effective QSVM kernels.
The method applies to both binary and multi-class classification tasks.
Test accuracy shows a positive correlation with quantum kernel entropy, so entropy may serve as a quick quality indicator.
Improved QSVM performance becomes available for applications in finance, healthcare, and materials science.

Where Pith is reading between the lines

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

The same genetic search could be applied to encoding circuits in other quantum machine learning models.
Scaling tests on larger feature spaces would show whether the optimization remains tractable.
The observed entropy-accuracy link might let researchers screen candidate kernels without running full genetic searches.
Direct comparisons against a wider range of classical kernel methods would clarify when the quantum version adds value.

Load-bearing premise

The supervised and unsupervised kernel loss functions used inside the genetic algorithm are reliable proxies that steer the search toward circuits with genuinely better out-of-sample classification performance.

What would settle it

A new dataset and loss-function pair in which the GA-optimized circuits achieve lower loss values yet lower test accuracy than standard quantum or classical kernels would falsify the central claim.

Figures

Figures reproduced from arXiv: 2312.01562 by Floyd M. Creevey, Jamie A. Heredge, Lloyd C. L. Hollenberg, Martin E. Sevior.

**Figure 1.** Figure 1: Basic overview of QSVMs. a) The datapoints, X , with labels ⃗y, are used to construct the elements of the kernel matrix, defined as the piecewise inner product, for the QSVM using the quantum kernel function for the given circuit structure. b) An example circuit that could be produced by the GA method. The GA is stochastic by nature, the number of features used and the number of qubits in the circuit will … view at source ↗

**Figure 2.** Figure 2: Overview of the GA for QSVM approach, given a dataset. First, create an initial individual [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Figure 3: Comparison of the number of features required to explain 95% variance in the dataset for Moons, XOR, Circles, [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

**Figure 4.** Figure 4: Comparison of supervised GA, unsupervised GA, PauliZZ, and RBF generated kernels on Moons, XOR, Circles, [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗

**Figure 5.** Figure 5: Decision boundaries for the Moons, XOR, Circles, and Irrigation datasets. Training data and testing data are [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗

**Figure 6.** Figure 6: Comparison of supervised and unsupervised GA kernel entropies compared to test accuracy of the given kernel [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗

read the original abstract

The data encoding circuits used in quantum support vector machine (QSVM) kernels play a crucial role in their classification accuracy. However, manually designing these circuits poses significant challenges in terms of time and performance. To address this, we leverage the GASP (Genetic Algorithm for State Preparation) framework for gate sequence selection in QSVM kernel circuits. We explore supervised and unsupervised kernel loss functions' impact on encoding circuit optimisation and evaluate them on diverse datasets for binary and multiple-class scenarios. Benchmarking against classical and quantum kernels reveals GA-generated circuits matching or surpassing standard techniques. We analyse the relationship between test accuracy and quantum kernel entropy, with results indicating a positive correlation. Our automated framework reduces trial and error, and enables improved QSVM based machine learning performance for finance, healthcare, and materials science applications.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

GA optimization finds competitive QSVM kernels and notes an entropy correlation, but the internal losses need clearer validation against generalization.

read the letter

The paper applies the existing GASP genetic algorithm framework to search for better data-encoding circuits in quantum support vector machines. It tests supervised and unsupervised kernel loss functions inside the search, benchmarks the resulting circuits against classical and quantum baselines, and reports that the GA outputs match or beat those baselines on the datasets tried. It also finds a positive correlation between quantum kernel entropy and test accuracy. That combination is the concrete new piece: a working demonstration of automated circuit search for this specific QSVM task plus the entropy observation. The practical angle is straightforward and addresses a real pain point for anyone who has to pick or design encoding circuits by hand. For applied work in finance, healthcare, or materials, an off-the-shelf optimizer that reduces trial-and-error could be handy. The benchmarking setup is at least externally oriented rather than circular. The soft spot is the proxy assumption. The GA is driven by the chosen kernel losses; if those losses do not reliably track out-of-sample accuracy across datasets or if they overfit to training idiosyncrasies, the reported gains rest on an untested link. The abstract gives no numbers, error bars, dataset sizes, or ablation results on the loss variants, so it is difficult to judge how robust the correlation or the performance edge actually is. Without those details the central empirical claim stays provisional. This is for people already working on quantum kernel methods or circuit search who need a practical tool rather than a theoretical advance. It is coherent on its own terms and shows honest engagement with the literature, so it deserves a serious referee even if the revisions will focus on methods and validation. I would send it to peer review.

Referee Report

3 major / 2 minor

Summary. The manuscript introduces the use of the GASP genetic algorithm framework to automatically select gate sequences for data-encoding circuits in quantum support vector machines. Supervised and unsupervised kernel loss functions are employed to guide the optimization; the resulting circuits are benchmarked on binary and multi-class datasets against classical kernels and standard quantum kernels, with the claim that GA-generated circuits match or surpass baseline performance. A positive correlation between test accuracy and quantum kernel entropy is also reported.

Significance. If the empirical claims hold after proper validation, the work supplies a practical, automated alternative to manual circuit design for QSVM kernels, which could lower the barrier to applying quantum machine learning in finance, healthcare, and materials science. The entropy-accuracy analysis offers a potentially useful diagnostic for kernel quality. The approach is reproducible in principle via the described GA procedure and could be extended to other variational quantum models.

major comments (3)

[§4] §4 (Benchmarking results): the central claim that GA circuits 'match or surpass' classical and quantum baselines is presented without tabulated accuracy values, standard deviations, dataset cardinalities, or statistical tests; without these quantities it is impossible to judge whether observed differences are meaningful or within noise.
[§3.2] §3.2 (supervised/unsupervised kernel losses): the manuscript assumes these losses steer the GA toward kernels with genuine out-of-sample generalization, yet provides no ablation across loss variants, no correlation between training loss and held-out accuracy, and no cross-validation that the losses were not tuned to the same data later used for final benchmarking; this assumption is load-bearing for interpreting the benchmarking results as evidence of improved kernel quality rather than loss-specific artifacts.
[entropy analysis] Entropy-accuracy analysis (near Figure 6): the reported positive correlation is stated without the number of circuit instances, Pearson coefficient, confidence interval, or p-value, so the strength and reliability of the relationship cannot be assessed.

minor comments (2)

[Abstract] Abstract and §2: 'diverse datasets' and 'standard techniques' are referenced without naming the concrete datasets or baseline kernels, reducing immediate clarity for readers.
[§3] Notation for the kernel alignment and entropy quantities could be made fully explicit (e.g., explicit formulas rather than descriptive text) to aid independent re-implementation.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major point below and will revise the manuscript to incorporate the requested details and validations.

read point-by-point responses

Referee: [§4] §4 (Benchmarking results): the central claim that GA circuits 'match or surpass' classical and quantum baselines is presented without tabulated accuracy values, standard deviations, dataset cardinalities, or statistical tests; without these quantities it is impossible to judge whether observed differences are meaningful or within noise.

Authors: We agree that the benchmarking results require more quantitative detail for proper assessment. In the revised manuscript we will add a table reporting mean accuracies and standard deviations over multiple independent runs for all methods and datasets, explicitly state the dataset cardinalities, and include statistical tests (e.g., paired t-tests or Wilcoxon signed-rank tests with p-values) to evaluate the significance of observed differences. revision: yes
Referee: [§3.2] §3.2 (supervised/unsupervised kernel losses): the manuscript assumes these losses steer the GA toward kernels with genuine out-of-sample generalization, yet provides no ablation across loss variants, no correlation between training loss and held-out accuracy, and no cross-validation that the losses were not tuned to the same data later used for final benchmarking; this assumption is load-bearing for interpreting the benchmarking results as evidence of improved kernel quality rather than loss-specific artifacts.

Authors: We acknowledge the need for stronger validation of the loss functions. The revised manuscript will include ablations comparing supervised and unsupervised loss variants, report correlations between training loss and held-out accuracy, and clarify the data-splitting protocol to confirm that loss optimization used separate validation folds from the final test sets. revision: yes
Referee: [entropy analysis] Entropy-accuracy analysis (near Figure 6): the reported positive correlation is stated without the number of circuit instances, Pearson coefficient, confidence interval, or p-value, so the strength and reliability of the relationship cannot be assessed.

Authors: We will expand the entropy-accuracy analysis to report the number of circuit instances evaluated, the Pearson correlation coefficient, its confidence interval, and the associated p-value, enabling readers to assess the statistical reliability of the observed positive correlation. revision: yes

Circularity Check

0 steps flagged

No significant circularity; claims rest on external empirical benchmarks

full rationale

The paper describes an empirical workflow: a genetic algorithm optimizes encoding circuits for QSVM kernels using supervised/unsupervised loss functions, followed by direct benchmarking of test accuracy against classical and quantum baselines on multiple datasets. No derivation chain reduces a reported performance metric to the internal loss by construction, no fitted parameters are relabeled as predictions, and no load-bearing uniqueness theorem or ansatz is imported via self-citation. The central results are falsifiable external comparisons, rendering the approach self-contained against the stated benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review is limited to the abstract; no explicit free parameters, invented entities, or non-standard axioms are described. The work implicitly relies on standard quantum circuit semantics and the assumption that the GASP framework (referenced but not re-derived) functions as claimed.

axioms (1)

domain assumption Quantum circuits can be executed and measured to produce kernel matrices usable by classical SVM solvers.
The entire QSVM pipeline presupposes ideal or simulatable quantum state preparation and measurement.

pith-pipeline@v0.9.0 · 5672 in / 1213 out tokens · 26589 ms · 2026-05-24T05:08:56.553083+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages · 2 internal anchors

[1]

The probability of an individual being selected is proportional to its fitness score

Selection: Choose the fittest individuals from the populationtobeusedasparentsforthenextgen- eration. The probability of an individual being selected is proportional to its fitness score

work page
[2]

This is done by selecting a crossover point on the chromo- some and swapping the genes to the right of the crossover point between the two parents

Crossover: Create new offspring by combining the genetic material of two parents. This is done by selecting a crossover point on the chromo- some and swapping the genes to the right of the crossover point between the two parents

work page
[3]

+moSDkS+//W3S2qqPYJlwbD+bgU=

Mutation: Introduce variation into the popula- tion by randomly changing some of the genes in the offspring. This is typically done by flipping some of the bits in the chromosome. The process of generating new generations continues until a stopping criterion is met, such as a maximum number of generations or a satisfactory level of fitness. The effectiven...

work page
[4]

Kurani, P

A. Kurani, P. Doshi, A. Vakharia, and M. Shah, An- nals of Data Science10, 183 (2023)

work page 2023
[5]

W. Yu, T. Liu, R. Valdez, M. Gwinn, and M. J. Khoury, BMC Medical Informatics and Decision Mak- ing 10, 16 (2010)

work page 2010
[6]

Venkatesan, P

C. Venkatesan, P. Karthigaikumar, A. Paul, S. Satheeskumaran, and R. Kumar, IEEE Ac- cess 6, 9767 (2018), conference Name: IEEE Access

work page 2018
[7]

Guangli and C

M. Guangli and C. Yiyu, (2006)

work page 2006
[8]

Havlíček, A

V. Havlíček, A. D. Córcoles, K. Temme, A. W. Har- row, A. Kandala, J. M. Chow, and J. M. Gambetta, Nature 567, 209 (2019), number: 7747 Publisher: Na- ture Publishing Group

work page 2019
[9]

Rebentrost, M

P. Rebentrost, M. Mohseni, and S. Lloyd, Physical Review Letters113, 130503 (2014), publisher: Amer- ican Physical Society

work page 2014
[10]

W. S. Noble, Nature Biotechnology24, 1565 (2006), number: 12 Publisher: Nature Publishing Group

work page 2006
[11]

Altares-López, A

S. Altares-López, A. Ribeiro, and J. J. García-Ripoll, Quantum Science and Technology6, 045015 (2021), publisher: IOP Publishing

work page 2021
[12]

Benedetti, E

M. Benedetti, E. Lloyd, S. Sack, and M. Fiorentini, Quantum Science and Technology4, 043001 (2019), publisher: IOP Publishing

work page 2019
[13]

Suzuki, H

Y. Suzuki, H. Yano, Q. Gao, S. Uno, T. Tanaka, M. Akiyama, and N. Yamamoto, Quantum Machine Intelligence 2, 9 (2020)

work page 2020
[14]

P. Wang, M. Usman, U. Parampalli, L. C. L. Hollen- berg, and C. R. Myers, IEEE Transactions on Quan- tum Engineering4, 1 (2023), conference Name: IEEE Transactions on Quantum Engineering

work page 2023
[15]

Nakaji, S

K. Nakaji, S. Uno, Y. Suzuki, R. Raymond, T. On- odera, T. Tanaka, H. Tezuka, N. Mitsuda, and N. Ya- mamoto, Physical Review Research4, 023136 (2022)

work page 2022
[16]

F. M. Creevey, C. D. Hill, and L. C. L. Hollenberg, Scientific Reports13, 11956 (2023), number: 1 Pub- lisher: Nature Publishing Group

work page 2023
[17]

AutoQML: Automatic Generation and Training of Robust Quantum-Inspired Classifiers by Using Ge- netic Algorithms on Grayscale Images,

S. Altares-López, J. J. García-Ripoll, and A. Ribeiro, “AutoQML: Automatic Generation and Training of Robust Quantum-Inspired Classifiers by Using Ge- netic Algorithms on Grayscale Images,” (2022), arXiv:2208.13246 [quant-ph]

work page arXiv 2022
[18]

Mitchell,An Introduction to Genetic Algorithms, Complex Adaptive Systems (A Bradford Book, Cam- bridge, MA, USA, 1996)

M. Mitchell,An Introduction to Genetic Algorithms, Complex Adaptive Systems (A Bradford Book, Cam- bridge, MA, USA, 1996)

work page 1996
[19]

M. A. Nielsen and I. L. Chuang,Quantum computa- tion and quantum information, 10th ed. (Cambridge University Press, Cambridge ; New York, 2010)

work page 2010
[20]

Radial basis function ker- nel optimization for Support Vector Machine classi- fiers,

K. Thurnhofer-Hemsi, E. López-Rubio, M. A. Molina- Cabello, and K. Najarian, “Radial basis function ker- nel optimization for Support Vector Machine classi- fiers,” (2020), arXiv:2007.08233 [cs, stat]

work page arXiv 2020
[21]

Quantum machine learning in feature Hilbert spaces

M. Schuld and N. Killoran, Physical Review Letters 122, 040504 (2019), arXiv:1803.07128 [quant-ph]

work page internal anchor Pith review Pith/arXiv arXiv 2019
[22]

P. Flach,Machine learning: The art and science of algorithms that make sense of data, Machine learning: The art and science of algorithms that make sense of data (Cambridge University Press, New York, NY, US,

work page
[23]

Multiclass Approaches for Support Vector Machine Based Land Cover Classification

M. Pal, “Multiclass Approaches for Support Vector Machine Based Land Cover Classification,” (2008), arXiv:0802.2411 [cs]

work page internal anchor Pith review Pith/arXiv arXiv 2008
[24]

A. Ng, M. Jordan, and Y. Weiss, in Advances in Neural Information Processing Systems, Vol. 14 (MIT Press, 2001)

work page 2001
[25]

Schölkopf, A

B. Schölkopf, A. Smola, and K.-R. Müller, in Arti- ficial Neural Networks — ICANN’97, Lecture Notes in Computer Science, edited by W. Gerstner, A. Ger- mond, M. Hasler, and J.-D. Nicoud (Springer, Berlin, Heidelberg, 1997) pp. 583–588

work page 1997
[26]

Pedregosa, G

F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duch- esnay, Journal of Machine Learning Research12, 2825 (2011)

work page 2011
[27]

INTELLIGENT IRRIGATION SYSTEM,

P. H., “INTELLIGENT IRRIGATION SYSTEM,” (2020)

work page 2020
[28]

Drug Classification,

T. P, “Drug Classification,” (2020)

work page 2020
[29]

M. A. Little, P. E. McSharry, S. J. Roberts, D. A. Costello, and I. M. Moroz, BioMedical Engineering OnLine 6, 23 (2007)

work page 2007
[30]

Pearson, The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science2, 559 (1901)

K. Pearson, The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science2, 559 (1901). 10 Appendix A: Algorithm Algorithm 1: Genetic Algorithm for QSVM Kernel Generation Data: data, maximumGenerations, targetF itness, numberOf Qubits, numberOf Genes, populationSize, optimisationT echnique Result: individual (quantumCircuit ) minimum ← 0...

work page 1901

[1] [1]

The probability of an individual being selected is proportional to its fitness score

Selection: Choose the fittest individuals from the populationtobeusedasparentsforthenextgen- eration. The probability of an individual being selected is proportional to its fitness score

work page

[2] [2]

This is done by selecting a crossover point on the chromo- some and swapping the genes to the right of the crossover point between the two parents

Crossover: Create new offspring by combining the genetic material of two parents. This is done by selecting a crossover point on the chromo- some and swapping the genes to the right of the crossover point between the two parents

work page

[3] [3]

+moSDkS+//W3S2qqPYJlwbD+bgU=

Mutation: Introduce variation into the popula- tion by randomly changing some of the genes in the offspring. This is typically done by flipping some of the bits in the chromosome. The process of generating new generations continues until a stopping criterion is met, such as a maximum number of generations or a satisfactory level of fitness. The effectiven...

work page

[4] [4]

Kurani, P

A. Kurani, P. Doshi, A. Vakharia, and M. Shah, An- nals of Data Science10, 183 (2023)

work page 2023

[5] [5]

W. Yu, T. Liu, R. Valdez, M. Gwinn, and M. J. Khoury, BMC Medical Informatics and Decision Mak- ing 10, 16 (2010)

work page 2010

[6] [6]

Venkatesan, P

C. Venkatesan, P. Karthigaikumar, A. Paul, S. Satheeskumaran, and R. Kumar, IEEE Ac- cess 6, 9767 (2018), conference Name: IEEE Access

work page 2018

[7] [7]

Guangli and C

M. Guangli and C. Yiyu, (2006)

work page 2006

[8] [8]

Havlíček, A

V. Havlíček, A. D. Córcoles, K. Temme, A. W. Har- row, A. Kandala, J. M. Chow, and J. M. Gambetta, Nature 567, 209 (2019), number: 7747 Publisher: Na- ture Publishing Group

work page 2019

[9] [9]

Rebentrost, M

P. Rebentrost, M. Mohseni, and S. Lloyd, Physical Review Letters113, 130503 (2014), publisher: Amer- ican Physical Society

work page 2014

[10] [10]

W. S. Noble, Nature Biotechnology24, 1565 (2006), number: 12 Publisher: Nature Publishing Group

work page 2006

[11] [11]

Altares-López, A

S. Altares-López, A. Ribeiro, and J. J. García-Ripoll, Quantum Science and Technology6, 045015 (2021), publisher: IOP Publishing

work page 2021

[12] [12]

Benedetti, E

M. Benedetti, E. Lloyd, S. Sack, and M. Fiorentini, Quantum Science and Technology4, 043001 (2019), publisher: IOP Publishing

work page 2019

[13] [13]

Suzuki, H

Y. Suzuki, H. Yano, Q. Gao, S. Uno, T. Tanaka, M. Akiyama, and N. Yamamoto, Quantum Machine Intelligence 2, 9 (2020)

work page 2020

[14] [14]

P. Wang, M. Usman, U. Parampalli, L. C. L. Hollen- berg, and C. R. Myers, IEEE Transactions on Quan- tum Engineering4, 1 (2023), conference Name: IEEE Transactions on Quantum Engineering

work page 2023

[15] [15]

Nakaji, S

K. Nakaji, S. Uno, Y. Suzuki, R. Raymond, T. On- odera, T. Tanaka, H. Tezuka, N. Mitsuda, and N. Ya- mamoto, Physical Review Research4, 023136 (2022)

work page 2022

[16] [16]

F. M. Creevey, C. D. Hill, and L. C. L. Hollenberg, Scientific Reports13, 11956 (2023), number: 1 Pub- lisher: Nature Publishing Group

work page 2023

[17] [17]

AutoQML: Automatic Generation and Training of Robust Quantum-Inspired Classifiers by Using Ge- netic Algorithms on Grayscale Images,

S. Altares-López, J. J. García-Ripoll, and A. Ribeiro, “AutoQML: Automatic Generation and Training of Robust Quantum-Inspired Classifiers by Using Ge- netic Algorithms on Grayscale Images,” (2022), arXiv:2208.13246 [quant-ph]

work page arXiv 2022

[18] [18]

Mitchell,An Introduction to Genetic Algorithms, Complex Adaptive Systems (A Bradford Book, Cam- bridge, MA, USA, 1996)

M. Mitchell,An Introduction to Genetic Algorithms, Complex Adaptive Systems (A Bradford Book, Cam- bridge, MA, USA, 1996)

work page 1996

[19] [19]

M. A. Nielsen and I. L. Chuang,Quantum computa- tion and quantum information, 10th ed. (Cambridge University Press, Cambridge ; New York, 2010)

work page 2010

[20] [20]

Radial basis function ker- nel optimization for Support Vector Machine classi- fiers,

K. Thurnhofer-Hemsi, E. López-Rubio, M. A. Molina- Cabello, and K. Najarian, “Radial basis function ker- nel optimization for Support Vector Machine classi- fiers,” (2020), arXiv:2007.08233 [cs, stat]

work page arXiv 2020

[21] [21]

Quantum machine learning in feature Hilbert spaces

M. Schuld and N. Killoran, Physical Review Letters 122, 040504 (2019), arXiv:1803.07128 [quant-ph]

work page internal anchor Pith review Pith/arXiv arXiv 2019

[22] [22]

P. Flach,Machine learning: The art and science of algorithms that make sense of data, Machine learning: The art and science of algorithms that make sense of data (Cambridge University Press, New York, NY, US,

work page

[23] [23]

Multiclass Approaches for Support Vector Machine Based Land Cover Classification

M. Pal, “Multiclass Approaches for Support Vector Machine Based Land Cover Classification,” (2008), arXiv:0802.2411 [cs]

work page internal anchor Pith review Pith/arXiv arXiv 2008

[24] [24]

A. Ng, M. Jordan, and Y. Weiss, in Advances in Neural Information Processing Systems, Vol. 14 (MIT Press, 2001)

work page 2001

[25] [25]

Schölkopf, A

B. Schölkopf, A. Smola, and K.-R. Müller, in Arti- ficial Neural Networks — ICANN’97, Lecture Notes in Computer Science, edited by W. Gerstner, A. Ger- mond, M. Hasler, and J.-D. Nicoud (Springer, Berlin, Heidelberg, 1997) pp. 583–588

work page 1997

[26] [26]

Pedregosa, G

F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duch- esnay, Journal of Machine Learning Research12, 2825 (2011)

work page 2011

[27] [27]

INTELLIGENT IRRIGATION SYSTEM,

P. H., “INTELLIGENT IRRIGATION SYSTEM,” (2020)

work page 2020

[28] [28]

Drug Classification,

T. P, “Drug Classification,” (2020)

work page 2020

[29] [29]

M. A. Little, P. E. McSharry, S. J. Roberts, D. A. Costello, and I. M. Moroz, BioMedical Engineering OnLine 6, 23 (2007)

work page 2007

[30] [30]

Pearson, The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science2, 559 (1901)

K. Pearson, The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science2, 559 (1901). 10 Appendix A: Algorithm Algorithm 1: Genetic Algorithm for QSVM Kernel Generation Data: data, maximumGenerations, targetF itness, numberOf Qubits, numberOf Genes, populationSize, optimisationT echnique Result: individual (quantumCircuit ) minimum ← 0...

work page 1901