Automated Characterization of Fluxonium Superconducting Qubits Parameters with Deep Transfer Learning

Accurate determination of qubit parameters is critical for the successful implementation of quantum information and computation applications. In superconducting fluxonium qubits, three key circuit parameters: E_J, E_C, and E_L, significantly influence the energy spectrum and transition behaviors. These parameters can be extracted by measuring the transition spectrum and matching it to the fluxonium circuit Hamiltonian, a process that is typically manual and time-consuming. In this work, we propose a machine learning-based methodology for automatic and accurate characterization of fluxonium qubit parameters. By utilizing deep transfer learning, we efficiently trained a machine learning model to predict the initial estimates of qubit parameters with fluxonium spectrum versus external flux as inputs. The model exhibits remarkable accuracy, achieving an average of approximately 95.64% in predicting qubit parameters. We further implemented an automatic fitting procedure, which can be directly applied to realistic experimental data. This subsequent fitting process not only reduces the need for extensive machine learning training resources but also makes our approach less constrained by the training dataset.