Subspace extraction using radial basis functions based precision matrix for enhanced quantum machine learning

초록

Quantum machine learning (QML) is anticipated to exhibit enhanced computing abilities as compared to its classical counterpart. However, one of the key issues in its potentially superior performance is the creation and propagation of quantum errors in quantum circuits. In Noisy Intermediate Scale Quantum (NISQ) era, one of the remedies is to impose the constraints on the number of quantum resources and to restrict the depth of quantum circuits. However, in the case of very large number of features in the data-sets, QML with restricted quantum resources may not yield reliable classification accuracy. In order to address these limitations, this study introduces a novel approach that leverages QML through precision matrix-based feature selection. More specifically, we considered the quantum version of support vector machines (PM-QSVM) to analyze the impact of the suggested feature selection mechanism. In addition, the impact of various kernel based quantum encodings of the selected features is also investigated. Moreover, the sensitivity analysis is carried out to ascertain the reliability of the feature selection mechanism. Our findings and the experimental results demonstrate how the integration of the suggested mechanism with quantum computing can enhance the decision making process.

키워드