A Hybrid "Quantum and Classical" Method for Outlier Detection

Abstract

In this paper, we present a novel method for outlier detection based on a hybrid approach involving both quantum and classical computing. The proposed method proceeds according to two steps: The first step uses classical computing by preparing and initializing components for the second step involving quantum computing. The latter uses adapted versions of some well-referenced quantum algorithms such as the quantum calculation of Hamming distance, and the minimum finding of Durr-Hoyer. The proposed method is based on the calculation of the distance between the instance to be tested and its Kth nearest neighbor. The test instance is thus considered an outlier if the calculated distance is greater than a given threshold T. At the end of this paper, we present an experimentation of the method and a performance analysis showing a quadratic improvement in terms of computational complexity compared to classical methods of outlier detection.