International Journal of Innovative Technology and Exploring Engineering (IJITEE) ISSN: 2278-3075, Volume-8 Issue-6, April 2019 208 Published By: Blue Eyes Intelligence Engineering & Sciences Publication Retrieval Number: F3441048619/19©BEIESP Experimental Comparison of Quantum and Classical Support Vector Machines Balika J.Chelliah, ShristiShreyasi,Ananya Pandey, Kirti Singh Abstract- Classical Support Vector Machine is hugely popular for classifying data efficiently whether it is linear or non-linear in nature. SVM has been used immensely to assist a precise classification of a data point. The kernel trick of SVM has also elevated the performance of the classical algorithm. But, SVM suffers a lot of problems on a classical machine when higher dimensions are introduced or large datasets are taken up. So, in order to enhance the efficiency of Support Vector Machine, the idea of running it on a quantum machine takes over.A Quantum Machine uses Qubits which is a single bit representing 0, 1 and superposition states of 0 & 1. This use of Qubit introduces the concept of ‘parallel processing’. The Quantum Machine utilises a different version of the SVM algorithm for performing the task of classification. In the algorithm, classical data is transformed into quantum data and then analysed over a Quantum Machine. For this experiment, the outcomes from both Classical Machine as well as Quantum Machine will be compared to determine Quantum Machine’s precedence over Classical Machine is justified or not. The comparison parameters are execution time and accuracy percentile of both the approaches. These results will be compared for asserting the importance of Quantum Approach in increasing machine learning’s scope, application and potential for current, future and deemed impossible task. Keywords : Quantum Machine, Quantum Machine Learning, Quantum Support Vector Machine, Support Vector Machine. I. INTRODUCTION Quantum Machine Learning is an interdisciplinary area which is an intersection of Quantum Physics and Machine Learning. It deals with running machine learning models on a quantum machine. A quantum machine works on qubit same as how the classical machine works on 0 and 1 bit which are expressed as high low voltage levels. A quantum bit or qubit is actually a single bit representing |0> or |1> or intermediate states of |0> and |1>. There are two concepts of quantum physics – Entanglement and Superposition which are the main basis for quantum machine learning on a quantum machine. A qubit’s state comes to existence when is measured or observed otherwise it remains undefined. Quantum Machine Learning takes on two approaches- first approach involves converting classical data to quantum data then performing the computations on the quantum computer and finally converting the quantum result back into the classical format. Revised Manuscript Received on April 07, 2019. Dr.Balika.J.Chelliah (M.Tech, Ph.d. )-Associate Professor at CSE Department SRMIST, Ramapuram, Chennai. Shristi Shreyasi - UG Scholar at CSE Department, SRMIST, Ramapuram, Chennai. Ananya Pandey– UG Scholar at CSE Department, SRMIST, Ramapuram, Chennai. Kirti Singh– UG Scholar at CSE Department, SRMIST, Ramapuram, Chennai. In short, this methoddevelops classical ML algorithms on a quantum computer. The othermethod involves using quantum mechanical principles fordesigning machine learning algorithms for classical computers. For the contrast between a classical and a quantum machine’s performance, the first approach will be used.Quantum Support Vectors Machine Algorithm uses the first approach of QML and gives precisely classified data as the end result. QSVM completes the task in an exponential time as compared to the classical SVM. It also handles the non-separable data by using a QSVM kernel algorithm which computes the kernel function much faster than the pre-existing. A decrease can be well seen in amount of time taken by the quantum algorithm as well as there is an efficient handling of large datasets for classification as it uses quantum mechanical operations. Robentrost [1], discussed about an optimised two-faced classifier which can be developed on a quantum computerhaving complexity in logarithms in the vectors size and the number of training examples. So, a quantum computer reduces a computation task taking polynomial time in the primitive environment to exponential time in a quantum environment. II. RELATED WORK Support Vector Machineis a supervisedalgorithm which is implemented forclassification as well as regressionproblems. But it is widelyimplemented for classification problems. In SVM, each and every data point is plotted on n-dimensional axis as a data coordinate where n stands for features held by the data with each feature acting in support for a coordinate.After plotting classification is done by finding the hyper-plane that will divide the two classes precisely. Figure 1: Classification Hyperplane of SVM