Parametrized Energy-Efficient Quantum Kernels for Network Service Fault Diagnosis

In quantum kernel learning, the primary method involves using a quantum computer to calculate the inner product between feature vectors, thereby obtaining a Gram matrix used as a kernel in machine learning models such as support vector machines (SVMs). However, a method for consistently achieving high performance has not been established. In this study, we investigate the diagnostic accuracy using a commercial dataset of a network service fault diagnosis system used by telecommunications carriers, focusing on quantum kernel learning, and propose a method to stably achieve high performance. We show significant performance improvements and an efficient achievement of high performance over conventional methods can be attained by applying quantum entanglement in the portion of the general quantum circuit used to create the quantum kernel, through input data parameter mapping and parameter tuning related to relative phase angles. Furthermore, experimental validation of the quantum kernel was conducted using IBM's superconducting quantum computer IBM-Kawasaki, and its practicality was verified while applying the error suppression feature of Q-CTRL's Fire Opal.