Quantum-Enhanced Support Vector Classifier for Image Classification
Roopa Golchha, Gyanendra K. Verma
Abstract
Machine learning plays an essential role in many fields, most commonly as a problem of approximation to determine the optimal value of an unknown function. In recent years, quantum-enhanced approaches have been widely studied to address machine learning issues. The quantum-enhanced approaches offer better classification results than some classical models. In this paper, we have proposed a Quantum Support Vector Classifier-based model for the binary classification of grey-scale images. We have applied a pre-processing model, including feature selection and state preparation, to improve the Quantum Support Vector Classifier method’s prediction rate to overcome the drawback of Noisy Intermediate-Scale Quantum systems. We have evaluated the performance of the proposed model using the publicly available Extended Cohn-Kanade, Gun, Knife, and FER2013 datasets. As a result, the accuracy of our proposed approach for binary classification of images is 95%, 83%, 87%, and 83% for the classical Support Vector Machine, meanwhile proposed Quantum Support Vector Classifier model obtained accuracy scores of 98%, 98%, 93%, and 92% on the Extended Cohn-Kanade, Gun, Knife, and FER2013 datasets, respectively.