Exploring Noise Dynamics: Identifying Most Active and Destructive Noises in IBM Quantum Computers

Abstract

This thesis addresses noise related problems by proposing a method to determine the most significant stochastic noises on three different IBM Quantum computers. It further explores whether these identified dominant noises are also the most destructive on the performance of a quantum algorithm, using the most dominant noise model parameters for analysis. To find the most dominant noise parameter particularly on IBM Quantum computer, we measured distance metrics commonly used in quantum computing such as TVD (Total Variation Distance), HD (Hellinger Distance) and JSD (Jensen Shannon Divergence). We also use classical optimization algorithms, i.e. Genetic Algorithm and Bayesian Optimization to find the noise parameter for each type of noise. Our experiment finds that depolarizing noise is the most active and destructive noise on all three IBM Quantum computers that we have access to. These IBM Quantum computers are IBM Brisbane, IBM Osaka, and IBM Kyoto. In addition to that, Bayesian Optimization turns to perform better compared to Genetic Algorithms on finding noise parameters.