Investigation of Quantum Support Vector Machine for Classification in NISQ era

TL;DR

This paper investigates the implementation of quantum support vector machines (QSVM) on current NISQ quantum computers, proposing an encoding extension and new classification method to improve efficiency on real hardware.

Contribution

It introduces a general encoding procedure for QSVM, evaluates its performance on real quantum hardware, and proposes an improved classification method for NISQ devices.

Findings

QSVM circuit implementation is feasible on current quantum hardware.

Encoding higher-dimensional vectors enhances QSVM performance.

New classification method improves efficiency on NISQ devices.

Abstract

Quantum machine learning is at the crossroads of two of the most exciting current areas of research; quantum computing and classical machine learning. It explores the interaction between quantum computing and machine learning, investigating how results and techniques from one field can be used to solve the problems of the other. Here, we investigate quantum support vector machine (QSVM) algorithm and its circuit version on present quantum computers. We propose a general encoding procedure extending QSVM algorithm, which would allow one to feed vectors with higher dimension in the training-data oracle of QSVM. We compute the efficiency of the QSVM circuit implementation method by encoding training and testing data sample in quantum circuits and running them on quantum simulator and real chip for two datasets; 6/9 and banknote. We highlight the technical difficulties one would face while applying the QSVM algorithm on current NISQ era devices. Then we propose a new method to classify these datasets with enhanced efficiencies for the above datasets both on simulator and real chips.