IEEE International Conference on Consumer Electronics - Taiwan (ICCE-Taiwan)
Combining Dimensionality Reduction and Feature Map in Quantum Support Vector Machines
Chien‐Cheng Tseng, Su‐Ling Lee
Abstract
Quantum support vector machines (QSVMs) offer a promising bridge between classical machine learning and quantum advantage by leveraging high-dimensional quantum feature spaces. However, practical implementation is often constrained by limited qubit resources, requiring efficient data representation. This paper evaluates the synergistic effects of combining various dimensionality reduction (DR) methods with different entanglement strategies in the quantum feature map. By investigating the influence of these combinations on classification accuracy, we identify good configurations for designing superior QSVM classifiers. Using the iris flower and wine datasets taken from scikit-learn, our experiments provide empirical insights into selecting the most effective DR-entanglement pairs, offering a framework for obtaining better classification performance within current hardware constraints.