This study explores the use of quantum kernel learning for fault diagnosis tasks in a commercial telecommunications network. One of the prevailing challenges in quantum kernel learning is to find a way to calculate the inner product between two feature vectors that can consistently achieve high performance. We propose a method that implements tunable parameters in the quantum entanglement generation part of quantum kernels, which allows for a more stable extraction in terms of inference performance of quantum machine learning. Experimental validation of this novel method was conducted using IBM’s superconducting quantum computer IBM-Kawasaki, and its practicality was verified by applying error suppression using Q-CTRL’s Fire Opal.

Hiroshi Yamauchi, Senior Research, SoftBank Corp.

Aravind Rutnam, Chief Strategy & Revenue Officer, Q-CTRL