cs.AIOct 15, 2026

Quantum Versus Classical Machine Learning for Neuropsychiatric Disorders: A Feature-Matched Multi-Cohort Evaluation

Authors: Utsav Poudel, Anuj Nepal, Vyas Viswanath, Subramaniyaswamy Vairavasundaram

Organizations: School of Computer Science and Engineering, Vellore Institute of Technology, Vellore, Tamil Nadu 632014, India · Universal Higher Education, Melbourne, Australia

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Resting-state functional connectivity is widely used in machine learning-based classification of neuropsychiatric disorders. Quantum machine learning models have been proposed for functional connectivity data, but few studies have evaluated them under leave-site-out validation or against matched classical models. We compared six quantum model families with classical ones using identical correlation-derived features across four public cohorts (4,308 subjects). Leave-site-out validation was used for the Autism Brain Imaging Data Exchange (ABIDE I), ADHD-200 (attention-deficit/hyperactivity disorder), and REST-meta-MDD (major depressive disorder) cohorts, while the UCLA Consortium for Neuropsychiatric Phenomics (UCLA-CNP) was evaluated using stratified 10-fold cross-validation. The quantum kernel was statistically indistinguishable from a matched classical kernel across all four cohorts, with the average accuracy difference within 1.4 percentage points at 95% confidence. On ABIDE I, a hybrid quantum convolutional network achieved similar accuracy to its classical counterpart (0.645 vs. 0.641), while a trainable quantum kernel exceeded its classical counterpart by 0.064 in area under the curve (AUC). Quantum convolutional networks also achieved higher AUC than their matched classical networks on this cohort and showed smaller train-test gaps on all four cohorts, without an accuracy gain. Overall, most quantum models matched their classical counterparts, but none consistently outperformed them. This study, among the first quantum evaluations of ABIDE I under leave-site-out validation, provides a feature-matched benchmark of quantum and classical models comprising 284 matched comparisons. We further analyse fidelity-based graph construction and account for the observed quantum-classical parity through effective-dimension and eigenspectrum analyses.

Figures & tables

CohortDiagnostic groupsN\bm{N}FoldsROIsNetworks reported disrupted
ABIDE IASD1,0351{,}0352020200200DMN ↓\downarrow , frontoparietal ↓\downarrow , limbic/amygdala ↓\downarrow
ADHD-200ADHD61961977190190frontoparietal ↓\downarrow , DMN (delayed suppression), subcortical/striatal ↓\downarrow
REST-meta-MDDMDD2,4282{,}4282525200200DMN (direction contested), salience ↑\uparrow , frontoparietal ↓\downarrow
UCLA-CNPSCZ, BD, ADHD, HC22622610∗10^{*}190190frontoparietal ↓↓\downarrow\downarrow , thalamocortical, auditory ↑\uparrow
Table I: Cohorts evaluated, diagnostic groups, and brain networks reported as disrupted.
Figure 1: Functional networks reported as disrupted in (a) ABIDE I, (b) ADHD-200, (c) REST-meta-MDD and (d) UCLA-CNP.
Figure 2: The fidelity-graph architecture evaluated here, showing quantum feature encoding and adaptive graph construction.
Figure 3: The per-region circuit, shown for four of nn qubits.
ModelF1AccSpecMCC
GCN+Pearson0.2520.252 ±0.118\pm 0.1180.4940.494 ±0.029\pm 0.0290.6640.664 ±0.185\pm 0.185−0.007-0.007 ±0.036\pm 0.036
GCN+RBF0.1450.145 ±0.119\pm 0.1190.5080.508 ±0.029\pm 0.0290.7830.783 ±0.179\pm 0.179−0.009-0.009 ±0.017\pm 0.017
Quantum+GAT0.1960.196 ±0.135\pm 0.1350.5040.504 ±0.028\pm 0.0280.7000.700 ±0.206\pm 0.2060.0000.000 ±0.000\pm 0.000
Quantum+GAT, trained enc.0.2370.237 ±0.130\pm 0.1300.5130.513 ±0.036\pm 0.0360.7050.705 ±0.174\pm 0.1740.0100.010 ±0.055\pm 0.055
Quantum+GCN (abl.)0.1330.133 ±0.119\pm 0.1190.5120.512 ±0.027\pm 0.0270.8000.800 ±0.180\pm 0.1800.0000.000 ±0.000\pm 0.000
Table II: Leave-site-out performance of the graph models on ABIDE I ( 1,0351{,}035 subjects, 2020 site folds, eight qubits; mean ±\pm 95% CI, normal quantile).
ModelAccAUCSpecMCC
QCNN †\dagger0.6150.615 ±0.029\pm 0.0290.6660.666 ±0.036\pm 0.0360.6990.699 ±0.076\pm 0.0760.2530.253 ±0.056\pm 0.056
CNN (1-D)0.5890.589 ±0.038\pm 0.0380.6380.638 ±0.043\pm 0.0430.6520.652 ±0.074\pm 0.0740.1940.194 ±0.075\pm 0.075
VQC †\dagger0.5820.582 ±0.032\pm 0.0320.6120.612 ±0.043\pm 0.0430.6320.632 ±0.074\pm 0.0740.1790.179 ±0.066\pm 0.066
MLP0.6000.600 ±0.028\pm 0.0280.6290.629 ±0.037\pm 0.0370.6740.674 ±0.055\pm 0.0550.2080.208 ±0.057\pm 0.057
QSVM-fixed †\dagger0.622\mathbf{0.622} ±0.034\pm\mathbf{0.034}0.6630.663 ±0.041\pm 0.0410.7170.717 ±0.076\pm 0.0760.264\mathbf{0.264} ±0.070\pm\mathbf{0.070}
SVM (RBF), matched0.6160.616 ±0.030\pm 0.0300.6620.662 ±0.039\pm 0.0390.7020.702 ±0.075\pm 0.0750.2550.255 ±0.059\pm 0.059
Table III: Matched eight-feature comparisons on ABIDE I, each quantum model ( †\dagger ) above its classical twin (mean ±\pm 95% CI over 2020 site folds).
ModelF1AccSpecMCC
SVM (Linear), full corr.0.653\mathbf{0.653} ±0.039\pm\mathbf{0.039}0.671\mathbf{0.671} ±0.036\pm\mathbf{0.036}0.6890.689 ±0.067\pm 0.0670.346\mathbf{0.346} ±0.073\pm\mathbf{0.073}
SVM (RBF), full corr.0.6270.627 ±0.054\pm 0.0540.6420.642 ±0.042\pm 0.0420.6340.634 ±0.088\pm 0.0880.3050.305 ±0.084\pm 0.084
Table IV: Unrestricted classical references on ABIDE I: full correlation vector, no register limit.
Figure 4: Quantum kernel against classical kernels given matched and full feature sets on four cohorts.
Figure 5: Effective dimension of the quantum fidelity and classical RBF kernels against their width parameter on all four cohorts.
Figure 6: Median per-fold accuracy and AUC difference between each quantum model and its matched classical twin, for the model suite and the three interventions.
Figure 7: Fidelity concentration with register width over twenty ABIDE I subjects, and one subject’s connectivity under Pearson correlation and four- and sixteen-qubit fidelity.
ModelAccF1AUCSpecMCC
ABIDE I
QCNN †\dagger0.6150.6150.5610.5610.6660.6660.6990.6990.2530.253
QPG (quantum edges) †\dagger0.5370.5370.5280.5280.5760.5760.5140.5140.0740.074
QSVM-Pegasos †\dagger0.5790.5790.4750.475—0.7500.7500.1990.199
QSVM-fixed †\dagger0.6220.6220.5600.5600.6630.6630.7170.7170.2640.264
TQEK †\dagger0.6100.6100.5340.5340.6860.6860.7240.7240.2340.234
Table V: Every model on every cohort. All values are means over folds; leave-site-out except UCLA-CNP, which is stratified 10-fold. †\dagger marks a quantum model. SVM labels give each model’s input: eight matched features, or the 2,000 supervised-selected ones.
Figure 8: Overfitting gap by cohort, and test accuracy and overfitting gap against training-set size on ABIDE I.
MetricPairsAheadSig.Clean
Accuracy727223235511
F1727235354422
AUC68683434161666
Specificity727223235533
Total284\mathbf{284}115\mathbf{115}30\mathbf{30}12\mathbf{12}
Table VI: Complete sweep of matched quantum–classical comparisons by metric.
Figure 9: Accuracy against Matthews correlation for every model on every cohort, pooled over folds.
Figure 10: Quantum-minus-classical accuracy for the three central matched pairs, per cohort (site folds in brackets) and combined across cohorts by a random-effects model.
CohortMethodProtocolAcc
ABIDE IAbraham et al. [ 44 ]LSO∼\sim 0.67
Population GCN [ 45 ]10-fold0.70
This work (linear SVM, full correlation)LSO0.671\mathbf{0.671}
ADHD-200Phenotypic data alone, three-class [ 46 ]held-out0.625
This work (SVM-RBF)LSO0.571\mathbf{0.571}
REST-meta-MDDEnsemble GNN [ 47 ]LSO0.73
Table VII: This work against published results on the same cohorts, with the validation protocol of each.
More in this topic

Quantum Machine Learning

Quantum machine learning uses quantum circuits and quantum data to build learning algorithms and study their computational and statistical limits.

See all 256 papers

Explore similar work

CardsList
May 27, 2026

Do We Really Need Quantum Machine Learning?: A Multidimensional Empirical Study

The rapid growth of computer vision and increasingly complex image recognition tasks has exposed fundamental computational limitations of classical machine learning models, motivating the exploration of quantum computing as an emerging new paradigm. This paper presents a comprehensive benchmarking study of classical and quantum machine learning models for image recognition on the MNIST handwritten digit dataset, evaluating both traditional models, a Classical Support Vector Machine (CSVM) and a Quantum Support Vector Machine (QSVM), and deep neural network models, a Classical Convolutional Neural Network (CCNN) and a Quantum Convolutional Neural Network (QCNN), across four performance dimensions: classification accuracy, computational runtime, parameter count, and memory requirements. Experiments are conducted as functions of both feature dimensionality and sample size, and across CPU and GPU execution environments, providing a controlled, multidimensional comparison to address gaps in prior work. For the SVM-based models, QSVM consistently outperforms CSVM in accuracy, reaching ∼\sim 0.90 versus ∼\sim 0.85 at 1,000 samples, with a higher computational cost. A feature count of 10 qubits and a sample size in the range of 200 -- 500 emerge as practical operating points that balance accuracy and runtime. For the neural network models, CCNN and QCNN achieve comparable classification accuracy, both exceeding 0.96 at 64 features and 60,000 samples, yet QCNN offers substantially superior parameter and memory efficiency, requiring ∼\sim 94% fewer parameters and ∼\sim 75% less memory than CCNN at higher feature counts, while incurring higher runtime. Across both model families, quantum models consistently outperform classical models by greater margins in accuracy as feature dimensionality or sample size increases.

Sudip Vhaduri, Ryan Gammon, Sayanton Dibbo

Department of Computer Science, University of Alabama, AL 35487 · Clemson University

Quantum Convolutional Neural NetworksQuantum Machine LearningQuantum Kernel MethodsQuantum Neural NetworksImage Classification
AskPDF
Sep 20, 2026

Comparative Study of Quantum and Classical Machine Learning Models in Binary Classification

A potential path forward is Quantum Machine Learning (QML), which aims to leverage quantum computing in conjunction with classical machine learning to enhance computing efficiency and the expressiveness of models. In this paper, two different quantum classifiers - Variational Quantum Classifier (VQC) and Quantum Kernel Support Vector Machine (QSVM) - are compared with three classical classifiers as baseline classifiers - Logistic Regression, Support Vector Machine (SVM), and a Multi-Layer Perceptron (MLP) - on the Breast Cancer Wisconsin dataset. The quantum circuits were created in the PennyLane framework and simulated on a classical backend. However, in terms of accuracy, classical Logistic Regression performed better with an accuracy of 97.8%, classical SVM and QSVM with an accuracy of 95.6% each, although the Quantum VQC achieved a lower accuracy of 88.9% and had a recall of 100% for the benign class, though it correctly identified only 12 of the 17 malignant cases (a malignant-class recall of approximately 70.6%). The drawback of quantum models is the higher training time; however, since the quantum circuit needs to be classically simulated, the quantum SVM took 23.29 seconds compared to less than 0.01 seconds for the classical linear models. These results indicate that for small structured datasets, classifiers based on quantum computing have not yet surpassed well-tuned classical counterparts. In some respects (e.g., benign-class recall), they perform competitively, though not on malignant-class recall, where the VQC in particular performed worse than the classical baselines, which is worth further investigation on real quantum computers.

Anand Kumar Mishra, Ramanuj Awasthi

School of Engineering & Technology (UIET), CSJM University, Kanpur, Uttar Pradesh, India

Quantum Machine LearningQuantum Kernel MethodsBinary ClassificationVariational Quantum Circuits
AskPDF
Jul 1, 2026

Quantum vs. Classical Machine Learning: A Unified Empirical Comparison

Quantum computing has emerged as a promising computational paradigm for machine learning (ML), with the potential to offer computational advantages over classical approaches. At this stage, the evidence supporting the performance and advantages of quantum machine learning (QML) models relative to classical models is insufficient. To address this gap, this paper presents an empirical study on the performance of QML models and their classical counterparts. We compare seven model pairs spanning supervised learning and reinforcement learning. Our results indicate that the evaluated quantum machine learning models do not yet surpass the classical baselines in overall prediction performance, policy stability, or training time. Nevertheless, QML remains a promising approach for filtering noise and controlling false positives. Our research findings summarize the challenges facing quantum machine learning across hardware environments, training efficiency, and convergence stability, providing a foundation for research into the robustness and parameter optimization of QML. This work is publicly available at https://github.com/Z-537-437/QML.

Chuanming Yu, Jiaming Liu, Zihao Ge +4

Hebei Normal University, Shijiazhuang 050024, China · University of Luxembourg, Kirchberg L-1359, Luxembourg · Kyushu University, Fukuoka 819-0395, Japan

Quantum Machine LearningQuantum RL
AskPDF