arxiv: 2601.00921 · v2 · submitted 2026-01-01 · 💻 cs.LG · cs.AI· quant-ph

Recognition: 1 theorem link

· Lean Theorem

Geometric and Quantum Kernel Methods for Predicting Skeletal Muscle Outcomes in chronic obstructive pulmonary disease

Azadeh Alavi , Hamidreza Khalili , Stanley H. Chan , Fatemeh Kouchmeshki , Muhammad Usman , Ross Vlahos

Authors on Pith no claims yet

Pith reviewed 2026-05-16 17:55 UTC · model grok-4.3

classification 💻 cs.LG cs.AIquant-ph

keywords quantum kernelSPD descriptorsCOPDmuscle weightridge regressionbiomarker predictionsmall sample learning

0 comments

The pith

Quantum kernel ridge regression with four inputs predicts skeletal muscle weight more accurately than classical methods in COPD models.

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

The paper benchmarks geometric symmetric positive definite descriptors and quantum kernel methods against classical models for predicting muscle outcomes from biomarkers in a preclinical COPD cohort of 213 animals. It establishes that a quantum-kernel ridge regression using four interpretable inputs achieves the lowest error and highest R-squared for muscle weight prediction. The structured features based on distances to prototypes and centres regularize the model for small data while maintaining transparency. If correct, these approaches could improve biomarker-driven predictions in limited-sample biomedical settings like COPD muscle wasting.

Core claim

Quantum-kernel ridge regression using four interpretable inputs achieved the best muscle-weight performance with RMSE 4.41 mg and R2 0.62, outperforming a matched compact classical baseline at 4.68 mg and R2 0.56, while biomarker-only SPD features improved over ridge regression from 4.79 mg to 4.55 mg RMSE.

What carries the argument

Clustered Nystrom-style quantum feature map that maps each subject to similarities with a small set of training-derived centres, together with SPD descriptors using Stein-divergence distances to representative prototypes.

Load-bearing premise

The improvements come from genuine regularization by the quantum map and prototype distances rather than from overfitting or selective tuning on this small dataset.

What would settle it

Running the same models on a new independent set of animals or patients and finding no advantage in RMSE or R2 over the classical baseline would disprove the central performance claim.

Figures

Figures reproduced from arXiv: 2601.00921 by Azadeh Alavi, Fatemeh Kouchmeshki, Hamidreza Khalili, Muhammad Usman, Ross Vlahos, Stanley H. Chan.

**Figure 2.** Figure 2: ROC-AUC comparison for tibialis anterior muscle weight. [PITH_FULL_IMAGE:figures/full_fig_p020_2.png] view at source ↗

**Figure 3.** Figure 3: ROC-AUC comparison for specific force. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗

**Figure 4.** Figure 4: ROC-AUC comparison for muscle quality index. [PITH_FULL_IMAGE:figures/full_fig_p021_4.png] view at source ↗

read the original abstract

Quantum methods are increasingly proposed for healthcare, but translational biomarker studies demand transparent benchmarking and robust small-dataset evaluation. We analysed a preclinical COPD cohort of 213 animals with blood and bronchoalveolar-lavage biomarkers to predict tibialis anterior muscle weight, specific force, and muscle quality. We benchmarked tuned classical models against two structured nonlinear low-data strategies: geometry-aware symmetric positive definite (SPD) descriptors, in which training-only clustering maps each subject to Stein-divergence distances from representative prototypes and optional unlabeled synthetic SPD interpolation stabilises prototype discovery; and quantum-kernel regression, including a clustered Nystrom-style feature map that compresses each subject into similarities to a small set of training-derived centres. By replacing full pairwise structure with compact prototype- and centre-based summaries, these steps regularise learning and preserve interpretability in a small-sample setting. Across five outer folds, quantum-kernel ridge regression using four interpretable inputs achieved the best muscle-weight performance (RMSE 4.41 mg; R2 0.62), outperforming a matched compact classical baseline (4.68 mg; R2 0.56). Biomarker-only SPD features also improved over ridge regression (4.55 versus 4.79 mg), and screening evaluation reached ROC-AUC 0.91 for low muscle weight.

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

Quantum kernels post a modest 0.06 R² lift on this 213-animal COPD muscle task, but the gain rests on thin controls for tuning and selection.

read the letter

The headline result is that quantum-kernel ridge regression with four inputs and a clustered Nystrom map reaches RMSE 4.41 mg and R² 0.62 on tibialis anterior muscle weight, beating a compact classical baseline at 4.68 mg and 0.56. The SPD Stein-divergence prototype features also improve over plain ridge regression. Those are the concrete numbers the paper supplies from five outer folds on the preclinical cohort of 213 animals. The work applies existing quantum-kernel and geometry-aware constructions to a new translational setting—predicting skeletal-muscle outcomes from blood and lavage biomarkers in a COPD model—and adds prototype compression to keep the representation compact and interpretable. That domain step is legitimate and the numbers are reported plainly enough to serve as a small-sample benchmark. The focus on replacing full pairwise kernels with distances to training-derived centers is a reasonable regularization move for low-data biomarker work. The paper does not claim dramatic gains, which keeps the claims proportionate to what is shown. The soft spot is the missing detail on how the four inputs were chosen and whether prototype count, cluster number, and kernel bandwidths were set inside an inner cross-validation loop or fixed after seeing test performance. With only five folds and several representations compared, a 0.06 R² difference can arise from fold variance or post-hoc selection rather than genuine inductive bias from the quantum map. No error bars or significance tests appear in the abstract, so the improvement remains plausible but not yet robust. This paper is for readers who track quantum or kernel methods in biomedical small-sample settings and want a concrete preclinical example rather than theory. It is not for someone needing strong statistical confirmation of superiority. It deserves peer review because the application is new, the methods are described at a level that can be checked, and the reported numbers give a starting point for further work, even if revisions will be needed for hyperparameter transparency and statistical support.

Referee Report

3 major / 2 minor

Summary. The manuscript evaluates quantum-kernel ridge regression and geometry-aware symmetric positive definite (SPD) descriptors for predicting tibialis anterior muscle weight, specific force, and quality from blood and bronchoalveolar-lavage biomarkers in a preclinical COPD cohort of 213 animals. It benchmarks these against classical models, reporting that quantum-kernel ridge regression with four interpretable inputs achieves the best muscle-weight performance (RMSE 4.41 mg; R² 0.62) across five outer folds, outperforming a matched compact classical baseline (4.68 mg; R² 0.56). Biomarker-only SPD features also improve over ridge regression (4.55 versus 4.79 mg), and the approach reaches ROC-AUC 0.91 for low-muscle-weight screening. The work emphasizes regularization via clustered Nystrom-style quantum feature maps and prototype-based Stein-divergence distances to preserve interpretability in small-sample settings.

Significance. If the reported performance gains hold under rigorous controls, the results would indicate that prototype-compressed quantum kernels and SPD geometry can supply modest regularization benefits for low-data biomarker prediction tasks, offering interpretable alternatives to standard ridge regression in translational preclinical studies where n is modest and overfitting risk is high.

major comments (3)

[Abstract] Abstract: The headline result (quantum-kernel RMSE 4.41 mg / R² 0.62 vs classical 4.68 mg / 0.56) is presented without per-fold standard deviations, error bars, or p-values on the 0.06 R² difference. With only five outer folds (~42 test points each) and multiple competing representations, it is impossible to determine whether the observed edge exceeds fold variance or arises from optimization noise.
[Methods] Methods (model selection and hyperparameter handling): The manuscript does not state whether the choice of four inputs, the number of Nystrom centers, cluster count for prototypes, or kernel bandwidths were fixed a priori or selected via an inner cross-validation loop nested inside the five outer folds. On n=213 data, any post-hoc or non-nested tuning would allow the reported delta to reflect selection bias rather than genuine inductive bias from the quantum or SPD constructions.
[Quantum kernel methods] Section on quantum feature map construction: The clustered Nystrom-style map compresses subjects to similarities with training-derived centers, yet the text provides no explicit confirmation that center selection and clustering were performed strictly on training folds only, nor any ablation showing that the quantum kernel's advantage survives when the same prototype count and bandwidth are used for the classical baseline.

minor comments (2)

[Abstract] Abstract: The phrase 'matched compact classical baseline' is undefined; specify the exact classical model (e.g., ridge regression with identical four inputs and no additional features) to allow direct comparison.
[Abstract] The abstract mentions 'optional unlabeled synthetic SPD interpolation' but does not indicate whether this step was used in the reported biomarker-only SPD results or how it affects the five-fold evaluation.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive comments on statistical reporting, cross-validation rigor, and methodological transparency. We address each point below and will revise the manuscript to incorporate the requested details and clarifications.

read point-by-point responses

Referee: The headline result (quantum-kernel RMSE 4.41 mg / R² 0.62 vs classical 4.68 mg / 0.56) is presented without per-fold standard deviations, error bars, or p-values on the 0.06 R² difference. With only five outer folds (~42 test points each) and multiple competing representations, it is impossible to determine whether the observed edge exceeds fold variance or arises from optimization noise.

Authors: We agree that variability across folds must be reported to assess robustness. In the revised manuscript we will add the per-fold standard deviations for RMSE and R², include error bars on the performance plots, and explicitly discuss the observed fold-to-fold variance. Given only five folds, formal p-values on the difference would have limited statistical power; we will therefore focus on mean ± std and qualitative assessment of whether the 0.06 R² gap exceeds typical fold noise. revision: yes
Referee: The manuscript does not state whether the choice of four inputs, the number of Nystrom centers, cluster count for prototypes, or kernel bandwidths were fixed a priori or selected via an inner cross-validation loop nested inside the five outer folds. On n=213 data, any post-hoc or non-nested tuning would allow the reported delta to reflect selection bias.

Authors: We acknowledge the need for explicit nested cross-validation. The four inputs were pre-specified on the basis of biological interpretability; the remaining hyperparameters (Nystrom centers, cluster count, bandwidths) were tuned via a nested inner 5-fold cross-validation performed strictly inside each outer fold. We will add a dedicated paragraph in the Methods section describing this nested protocol and confirming that no test-fold information entered hyperparameter selection. revision: yes
Referee: The clustered Nystrom-style map compresses subjects to similarities with training-derived centers, yet the text provides no explicit confirmation that center selection and clustering were performed strictly on training folds only, nor any ablation showing that the quantum kernel's advantage survives when the same prototype count and bandwidth are used for the classical baseline.

Authors: All center selection, clustering, and prototype construction were performed exclusively on the training portion of each outer fold; we will insert an explicit statement to this effect in the quantum feature-map subsection. In addition, we will include a new ablation experiment in which the classical baseline is given exactly the same number of prototypes/centers and identical bandwidth values, thereby isolating the contribution of the quantum kernel itself. revision: yes

Circularity Check

0 steps flagged

No significant circularity; standard constructions applied to independent targets

full rationale

The paper defines SPD descriptors via Stein divergence from training-only prototypes and a clustered Nystrom quantum feature map via similarities to training-derived centres. These are standard, externally defined constructions that do not reduce by any equation in the manuscript to quantities fitted on the muscle-weight, force, or quality labels. Performance numbers (RMSE 4.41 mg, R² 0.62) are direct cross-validation outputs rather than predictions forced by the inputs. No self-citation chain, uniqueness theorem, or ansatz smuggling is invoked to justify the central claim. The derivation therefore remains self-contained against external benchmarks and receives the default non-circularity finding.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard properties of Stein divergence on SPD matrices and the existence of a computable quantum feature map; free parameters are the number of prototypes/centers and kernel hyperparameters, both tuned on the same small dataset.

free parameters (2)

number of prototypes / Nystrom centers
Chosen during training-only clustering to create compact summaries; value not stated in abstract.
quantum and classical kernel hyperparameters
Tuned to produce the reported RMSE/R² values on the 213-animal cohort.

axioms (2)

standard math Stein divergence defines a valid Riemannian metric on the manifold of symmetric positive definite matrices
Invoked to justify geometry-aware descriptors and distance-based summaries.
domain assumption A quantum feature map can be approximated by a finite set of training-derived centers for kernel regression
Basis for the clustered Nystrom-style compression step.

pith-pipeline@v0.9.0 · 5557 in / 1577 out tokens · 47268 ms · 2026-05-16T17:55:59.576446+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

67 extracted references · 67 canonical work pages · 2 internal anchors

[1]

Agust´ ı et al

A. Agust´ ı et al. Global initiative for chronic obstructive lung disease 2023 report: GOLD executive summary.Journal of the Pan African Thoracic Society, 4(2):58–80, 2022

work page 2023
[2]

C. F. Vogelmeier et al. Global strategy for the diagnosis, management, and prevention of chronic obstructive lung disease 2017 report: GOLD executive summary.American Journal of Respiratory and Critical Care Medicine, 195(5):557–582, 2017

work page 2017
[3]

B. W. Carlin. Exacerbations of COPD.Respiratory Care, 68(7):961–972, 2023

work page 2023
[4]

J. A. Wedzicha, R. Singh, and A. J. Mackay. Acute COPD exacerbations.Clinics in Chest Medicine, 35(1):157–163, 2014

work page 2014
[5]

S ¸ahin et al

F. S ¸ahin et al. Serum biomarkers in patients with stable and acute exacerbation of chronic obstructive pulmonary disease: a comparative study.Journal of Medical Biochemistry, 38(4):503, 2019

work page 2019
[6]

A. G. Mathioudakis et al. Acute exacerbations of chronic obstructive pulmonary disease: in search of diagnostic biomarkers and treatable traits.Thorax, 75(6):520–527, 2020

work page 2020
[7]

H¨ ogman et al

M. H¨ ogman et al. 2017 global initiative for chronic obstructive lung disease reclassifies half of COPD subjects to lower risk group.International Journal of Chronic Obstructive Pulmonary Disease, pages 165–173, 2018

work page 2017
[8]

N. C. Dos Santos et al. Prevalence and impact of comorbidities in individuals with chronic obstructive pulmonary disease: a systematic review.Tuberculosis and Respiratory Diseases, 85(3):205, 2022

work page 2022
[9]

Xu et al

J. Xu et al. Inflammation mechanism and research progress of COPD.Frontiers in Im- munology, 15:1404615, 2024

work page 2024
[10]

Decramer et al

M. Decramer et al. COPD as a lung disease with systemic consequences—clinical impact, mechanisms, and potential for early intervention.COPD: Journal of Chronic Obstructive Pulmonary Disease, 5(4):235–256, 2008

work page 2008
[11]

S. M. Chan et al. Pathobiological mechanisms underlying metabolic syndrome (MetS) in chronic obstructive pulmonary disease (COPD): clinical significance and therapeutic strategies.Pharmacology & Therapeutics, 198:160–188, 2019

work page 2019
[12]

Gephine et al

S. Gephine et al. Specific contribution of quadriceps muscle strength, endurance, and power to functional exercise capacity in people with chronic obstructive pulmonary disease: a multicenter study.Physical Therapy, 101(6):pzab052, 2021

work page 2021
[13]

Li et al

W. Li et al. Association between muscular atrophy and mortality risk in patients with COPD: a systematic review.Therapeutic Advances in Respiratory Disease, 18:17534666241304626, 2024

work page 2024
[14]

A. H. Attaway et al. Muscle loss phenotype in COPD is associated with adverse outcomes in the UK Biobank.BMC Pulmonary Medicine, 24(1):186, 2024

work page 2024
[15]

Barreiro and A

E. Barreiro and A. Jaitovich. Muscle atrophy in chronic obstructive pulmonary disease: molecular basis and potential therapeutic targets.Journal of Thoracic Disease, 10(Suppl 12):S1415, 2018

work page 2018
[16]

Maltais et al

F. Maltais et al. An official American Thoracic Society/European Respiratory Society statement: update on limb muscle dysfunction in chronic obstructive pulmonary disease. American Journal of Respiratory and Critical Care Medicine, 189(9):e15–e62, 2014. 24

work page 2014
[17]

Lee et al

L.-W. Lee et al. Body composition changes in male patients with chronic obstructive pulmonary disease: aging or disease process?PLOS One, 12(7):e0180928, 2017

work page 2017
[18]

B. H. Goodpaster et al. The loss of skeletal muscle strength, mass, and quality in older adults: the health, aging and body composition study.The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 61(10):1059–1064, 2006

work page 2006
[19]

Nogueira et al

L. Nogueira et al. Cigarette smoke directly impairs skeletal muscle function through cap- illary regression and altered myofibre calcium kinetics in mice.The Journal of Physiology, 596(14):2901–2916, 2018

work page 2018
[20]

McCarthy et al

B. McCarthy et al. Pulmonary rehabilitation for chronic obstructive pulmonary disease. Cochrane Database of Systematic Reviews, (2), 2015

work page 2015
[21]

M. A. Spruit et al. An official American Thoracic Society/European Respiratory Society statement: key concepts and advances in pulmonary rehabilitation.American Journal of Respiratory and Critical Care Medicine, 188(8):e13–e64, 2013

work page 2013
[22]

N. S. Cox et al. Telerehabilitation versus traditional centre-based pulmonary rehabilitation for people with chronic respiratory disease: protocol for a randomised controlled trial.BMC Pulmonary Medicine, 18(1):71, 2018

work page 2018
[23]

M. I. Polkey et al. Activin type II receptor blockade for treatment of muscle depletion in chronic obstructive pulmonary disease: a randomized trial.American Journal of Respira- tory and Critical Care Medicine, 199(3):313–320, 2019

work page 2019
[24]

K. Mou, S. M. Chan, and R. Vlahos. Musculoskeletal crosstalk in chronic obstructive pulmonary disease and comorbidities: emerging roles and therapeutic potentials.Pharma- cology & Therapeutics, 257:108635, 2024

work page 2024
[25]

S. Chan, S. Selemidis, and R. Vlahos. The double-edged sword of ROS in muscle wasting and COPD: insights from aging-related sarcopenia.Antioxidants, 13(7):882, 2024

work page 2024
[26]

W. Dr¨ oge. Free radicals in the physiological control of cell function.Physiological Reviews, 2002

work page 2002
[27]

Agrawal et al

S. Agrawal et al. Exploring the role of oxidative stress in skeletal muscle atrophy: mecha- nisms and implications.Cureus, 15(7), 2023

work page 2023
[28]

Li and S.-F

C.-L. Li and S.-F. Liu. Exploring molecular mechanisms and biomarkers in COPD: an overview of current advancements and perspectives.International Journal of Molecular Sciences, 25(13):7347, 2024

work page 2024
[29]

Fang et al

H. Fang et al. Prognostic biomarkers based on proteomic technology in COPD: a recent review.International Journal of Chronic Obstructive Pulmonary Disease, pages 1353–1365, 2023

work page 2023
[30]

Zhang et al

Z. Zhang et al. Proteomics and metabolomics profiling reveal panels of circulating diagnos- tic biomarkers and molecular subtypes in stable COPD.Respiratory Research, 24(1):73, 2023

work page 2023
[31]

C. J. Enr´ ıquez-Rodr´ ıguez et al. COPD: systemic proteomic profiles in frequent and infre- quent exacerbators.ERJ Open Research, 10(2), 2024

work page 2024
[32]

Lin et al

F. Lin et al. AutoCOPD—a novel and practical machine learning model for COPD detec- tion using whole-lung inspiratory quantitative CT measurements: a retrospective, multi- center study.EClinicalMedicine, 82, 2025. 25

work page 2025
[33]

F. M. Torun et al. Transparent exploration of machine learning for biomarker discovery from proteomics and omics data.Journal of Proteome Research, 22(2):359–367, 2022

work page 2022
[34]

N. A. Enzer et al. Proteomics and machine learning in the prediction and explanation of low pectoralis muscle area.Scientific Reports, 14(1):17981, 2024

work page 2024
[35]

S. M. Chan et al. Cigarette smoking exacerbates skeletal muscle injury without compromis- ing its regenerative capacity.American Journal of Respiratory Cell and Molecular Biology, 62(2):217–230, 2020

work page 2020
[36]

S. M. Chan et al. Apocynin prevents cigarette smoking-induced loss of skeletal muscle mass and function in mice by preserving proteostatic signalling.British Journal of Pharmacology, 178(15):3049–3066, 2021

work page 2021
[40]

Rousseeuw.Finding Groups in Data: An Introduction to Cluster Analysis

Leonard Kaufman and Peter J. Rousseeuw.Finding Groups in Data: An Introduction to Cluster Analysis. John Wiley & Sons, 1990

work page 1990
[41]

Christopher K. I. Williams and Matthias Seeger. Using the Nystr¨ om method to speed up kernel machines. InAdvances in Neural Information Processing Systems, 2001

work page 2001
[42]

C´ orcoles, Kristan Temme, Aram W

Vojtˇ ech Havl´ ıˇ cek, Antonio D. C´ orcoles, Kristan Temme, Aram W. Harrow, Abhinav Kan- dala, Jay M. Chow, and Jay M. Gambetta. Supervised learning with quantum-enhanced feature spaces.Nature, 567(7747):209–212, 2019

work page 2019
[43]

Quantum machine learning in feature Hilbert spaces

Maria Schuld and Nathan Killoran. Quantum machine learning in feature Hilbert spaces. Physical Review Letters, 122(4):040504, 2019

work page 2019
[44]

Putcha et al

N. Putcha et al. Comorbidities and chronic obstructive pulmonary disease: prevalence, influence on outcomes, and management.Seminars in Respiratory and Critical Care Medicine, 2015. Thieme Medical Publishers

work page 2015
[45]

Vlahos and S

R. Vlahos and S. Bozinovski. Role of alveolar macrophages in chronic obstructive pul- monary disease.Frontiers in Immunology, 5:435, 2014

work page 2014
[46]

Yende et al

S. Yende et al. Inflammatory markers are associated with ventilatory limitation and mus- cle dysfunction in obstructive lung disease in well functioning elderly subjects.Thorax, 61(1):10–16, 2006

work page 2006
[47]

Ma et al

K. Ma et al. Pathogenesis of sarcopenia in chronic obstructive pulmonary disease.Frontiers in Physiology, 13:850964, 2022

work page 2022
[48]

Wang et al

F. Wang et al. Effects of exercise-induced ROS on the pathophysiological functions of skeletal muscle.Oxidative Medicine and Cellular Longevity, 2021(1):3846122, 2021

work page 2021
[49]

Barreiro et al

E. Barreiro et al. Cytokine profile in quadriceps muscles of patients with severe COPD. Thorax, 63(2):100–107, 2008. 26

work page 2008
[50]

Henrot et al

P. Henrot et al. Main pathogenic mechanisms and recent advances in COPD peripheral skeletal muscle wasting.International Journal of Molecular Sciences, 24(7):6454, 2023

work page 2023
[51]

Tan et al

Z. Tan et al. Myostatin is involved in skeletal muscle dysfunction in chronic obstructive pulmonary disease via Drp-1 mediated abnormal mitochondrial division.Annals of Trans- lational Medicine, 10(4):162, 2022

work page 2022
[52]

Zhuang et al

Y. Zhuang et al. Deep learning on graphs for multi-omics classification of COPD.PLOS One, 18(4):e0284563, 2023

work page 2023
[53]

Kor et al

C.-T. Kor et al. Explainable machine learning model for predicting first-time acute exac- erbation in patients with chronic obstructive pulmonary disease.Journal of Personalized Medicine, 12(2):228, 2022

work page 2022
[54]

Quantum machine learning.Nature, 549(7671):195–202, 2017

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

work page 2017
[55]

Cerezo, A

M. Cerezo, A. Arrasmith, R. Babbush, S. C. Benjamin, S. Endo, K. Fujii, J. R. McClean, K. Mitarai, X. Yuan, L. Cincio, and P. J. Coles. Variational quantum algorithms.Nature Reviews Physics, 3(9):625–644, 2021

work page 2021
[56]

Machine learning approaches to identify predictors of muscle wasting in experimental chronic obstructive pulmonary disease, 2025

Hamidreza Khalili. Machine learning approaches to identify predictors of muscle wasting in experimental chronic obstructive pulmonary disease, 2025

work page 2025
[57]

In-Kwon Yeo and Richard A. Johnson. A new family of power transformations to improve normality or symmetry.Biometrika, 87(4):954–959, 2000

work page 2000
[58]

Hoerl and Robert W

Arthur E. Hoerl and Robert W. Kennard. Ridge regression: Biased estimation for nonorthogonal problems.Technometrics, 12(1):55–67, 1970

work page 1970
[59]

Random forests.Machine Learning, 45:5–32, 2001

Leo Breiman. Random forests.Machine Learning, 45:5–32, 2001

work page 2001
[60]

Friedman, Richard A

Leo Breiman, Jerome H. Friedman, Richard A. Olshen, and Charles J. Stone.Classification and Regression Trees. Wadsworth International Group, 1984

work page 1984
[61]

Princeton University Press, 2007

Rajendra Bhatia.Positive Definite Matrices. Princeton University Press, 2007

work page 2007
[62]

A well-conditioned estimator for large-dimensional co- variance matrices.Journal of Multivariate Analysis, 88(2):365–411, 2004

Olivier Ledoit and Michael Wolf. A well-conditioned estimator for large-dimensional co- variance matrices.Journal of Multivariate Analysis, 88(2):365–411, 2004

work page 2004
[63]

Positive definite matrices and the S-divergence

Suvrit Sra. A new metric on the manifold of symmetric positive definite matrices.arXiv preprint arXiv:1110.1773, 2012

work page internal anchor Pith review Pith/arXiv arXiv 2012
[64]

Efficient similarity search for covariance matrices via the jensen-bregman logdet divergence

Anoop Cherian, Suvrit Sra, Arindam Banerjee, and Nikolaos Papanikolopoulos. Efficient similarity search for covariance matrices via the jensen-bregman logdet divergence. In Proceedings of the International Conference on Computer Vision (ICCV), 2011

work page 2011
[65]

Log-euclidean met- rics for fast and simple calculus on diffusion tensors.Magnetic Resonance in Medicine, 56(2):411–421, 2006

Vincent Arsigny, Pierre Fillard, Xavier Pennec, and Nicholas Ayache. Log-euclidean met- rics for fast and simple calculus on diffusion tensors.Magnetic Resonance in Medicine, 56(2):411–421, 2006

work page 2006
[66]

Random Projections on Manifolds of Symmetric Positive Definite Matrices for Image Classification

A. Alavi, W. Forstner, A. Mehrabi, et al. Random projections on manifolds of symmetric positive definite matrices for image classification.arXiv preprint arXiv:1403.0700, 2014

work page internal anchor Pith review Pith/arXiv arXiv 2014
[67]

Rousseeuw.Finding Groups in Data: An Introduction to Cluster Analysis

Leonard Kaufman and Peter J. Rousseeuw.Finding Groups in Data: An Introduction to Cluster Analysis. Wiley, 1990. 27

work page 1990
[68]

Quantum machine learning in feature hilbert spaces

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

work page 2019
[69]

C´ orcoles, Kristan Temme, Aram W

Vojtˇ ech Havl´ ıˇ cek, Antonio D. C´ orcoles, Kristan Temme, Aram W. Harrow, Abhinav Kan- dala, Jerry M. Chow, and Jay M. Gambetta. Supervised learning with quantum-enhanced feature spaces.Nature, 567(7747):209–212, 2019

work page 2019
[70]

Christopher K. I. Williams and Matthias Seeger. Using the nystr¨ om method to speed up kernel machines. InAdvances in Neural Information Processing Systems 13 (NIPS 2000), pages 682–688. MIT Press, 2001. 28

work page 2000