Quantum Reservoir Computing for Hamiltonian Learning in Metal-Insulator Anderson Transitions

Abstract

This research investigates transport regimes in metal-insulator Anderson transition through Hamiltonian learning. Quantum reservoir computing is employed to estimate the stochasticity parameter in the Hamiltonian of the quasiperiodic kicked rotor, a model that displays Anderson transition in momentum space. The stochasticity parameter is key for classifying phase regimes, i.e., localized/insulator phase, delocalized/metalic phase, and critical phase, as well as qualitatively forecasting trajectory evolution. Thus, supervised machine learning that effectively maps input trajectories to their corresponding stochasticity parameter has been developed, highlighting the efficacy of quantum machine learning in analyzing quantum phenomena