Quantum-Enhanced Classification and Clustering Through Hybrid Quantum–Classical Learning on Synthetic Data

Abstract

Quantum Machine Learning (QML) has emerged as a promising paradigm that combines principles of quantum computation with classical learning techniques to address challenges posed by complex, high-dimensional, and non-linear datasets. This work presents a systematic experimental study of quantum-enhanced classification and clustering using hybrid quantum–classical models evaluated on synthetically generated datasets. Quantum Support Vector Machines (QSVM) and Variational Quantum Classifiers (VQC) are implemented using the Qiskit framework and benchmarked against established classical algorithms. The synthetic datasets are carefully designed to control linearity, dimensionality, class overlap, and non-linear structure, enabling a detailed performance comparison under both ideal and noisy simulation conditions. The experimental results indicate that quantum feature representations can enhance class separability in non-linear scenarios, while also highlighting the practical limitations imposed by noise sensitivity and circuit depth constraints inherent to Noisy Intermediate-Scale Quantum (NISQ) devices .

Introduction

The text presents a comprehensive study on quantum-enhanced machine learning for both classification and clustering tasks under realistic Noisy Intermediate-Scale Quantum (NISQ) constraints. Classical machine learning models often struggle with high-dimensional, non-linear, and complex datasets, while deep learning and kernel methods require heavy computation and large labeled data. Quantum computing offers an alternative by exploiting quantum properties such as superposition and entanglement, enabling learning in high-dimensional Hilbert spaces through quantum feature mappings and kernel methods.

The study reviews prior work showing that quantum kernels, Quantum Support Vector Machines (QSVM), and Variational Quantum Classifiers (VQC) can enhance model expressiveness, and that many quantum classifiers can be interpreted as kernel-based methods. However, existing research is fragmented, typically focusing on either classification or clustering separately and often relying on idealized simulations.

To address this gap, the paper formulates a unified experimental framework to evaluate both supervised and unsupervised learning using consistent datasets, metrics, and noise assumptions. Synthetic datasets of increasing complexity (linear, non-linear, and high-dimensional overlapping) are generated. Classical baselines (SVM and k-means) are compared against quantum-enhanced models (QSVM and VQC) implemented using hybrid quantum–classical architectures with angle encoding and realistic noise models.

Experimental results show that quantum-enhanced models, particularly QSVM, outperform classical counterparts on non-linearly separable datasets in both classification accuracy and clustering quality. Quantum feature spaces improve class separability and cluster structure. However, performance degrades significantly with increasing noise and circuit depth, highlighting the sensitivity of quantum models to hardware imperfections.

Conclusion

This paper presented a comprehensive experimental evaluation of quantum-enhanced classification and clustering using hybrid quantum–classical learning models applied to carefully designed synthetic datasets. By jointly analyzing both supervised and unsupervised learning tasks within a unified experimental framework, the study provides systematic insights into the practical behavior of quantum-enhanced learning algorithms under controlled conditions. The use of synthetic datasets with varying degrees of linearity, non-linearity, and dimensionality enabled precise evaluation of model performance and facilitated fair comparison with classical machine learning baselines The experimental results demonstrate that quantum feature representations can significantly enhance learning performance for non-linearly separable data distributions, particularly when using quantum kernel-based methods such as the Quantum Support Vector Machine. Improved classification accuracy and higher clustering quality metrics indicate that embedding data into high-dimensional quantum feature spaces can improve class separability and preserve intrinsic data structures more effectively than classical feature mappings. These findings highlight the potential of hybrid quantum–classical models to address learning challenges that are difficult for conventional machine learning algorithms, especially in scenarios involving complex feature interactions. Despite these advantages, the study also reveals important limitations associated with current quantum hardware. The observed sensitivity of quantum-enhanced models to noise and circuit depth underscores the constraints imposed by Noisy Intermediate-Scale Quantum devices. Performance degradation under noisy conditions emphasizes the need for careful circuit design and realistic evaluation when assessing the benefits of quantum-enhanced learning. Future work will focus on extending this research in several directions. First, the proposed models will be evaluated on real quantum hardware to assess performance under true experimental conditions beyond simulation. Second, advanced noise mitigation and error reduction techniques will be explored to improve model robustness and stability. Finally, the framework will be extended to larger and more complex datasets, including real-world data, in order to further examine scalability and to better understand the conditions under which quantum-enhanced learning provides practical advantages over classical approaches.

References

[1] C. M. Bishop, Pattern Recognition and Machine Learning. New York, NY, USA: Springer, 2006. [2] T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning, 2nd ed. New York, NY, USA: Springer, 2009. [3] V. Havlí?ek et al., “Supervised learning with quantum-enhanced feature spaces,” Nature, vol. 567, pp. 209–212, 2019. [4] M. Schuld and N. Killoran, “Quantum machine learning in feature Hilbert spaces,” Phys. Rev. Lett., vol. 122, no. 4, 2019. [5] M. Schuld, “Supervised quantum machine learning models are kernel methods,” arXiv:2101.11020, 2021. [6] M. Schuld and F. Petruccione, Supervised Learning with Quantum Computers. Cham, Switzerland: Springer, 2018. [7] J. Jäger and R. V. Krems, “Universal expressiveness of variational quantum classifiers,” arXiv:2207.05865, 2022. [8] M. Srikumar, C. D. Hill, and L. C. L. Hollenberg, “Clustering and enhanced classification using a hybrid quantum autoencoder,” Quantum Science and Technology, vol. 6, no. 4, 2021. [9] D. Peral-García, “A systematic literature review of quantum machine learning,” Computers & Electrical Engineering, vol. 110, 2024. [10] J. Preskill, “Quantum computing in the NISQ era and beyond,” Quantum, vol. 2, p. 79, 2018. [11] F. Arute et al., “Quantum supremacy using a programmable superconducting processor,” Nature, vol. 574, pp. 505–510, 2019. [12] M. Cerezo et al., “Variational quantum algorithms,” Nature Reviews Physics, vol. 3, pp. 625–644, 2021.

Copyright

Copyright © 2026 Suraj Kamal Kanal. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.