Experimental demonstration of enhanced quantum tomography via quantum reservoir processing

Abstract

Quantum machine learning is a rapidly advancing discipline that leverages the features of quantum mechanics to enhance the performance of computational tasks. Quantum reservoir processing, which allows efficient optimization of a single output layer without precise control over the quantum system, stands out as one of the most versatile and practical quantum machine learning techniques. Here we experimentally demonstrate a quantum reservoir processing approach for continuous-variable state reconstruction on a bosonic circuit quantum electrodynamics platform. The scheme learns the true dynamical process through a minimum set of measurement outcomes of a known set of initial states. We show that the map learnt this way achieves high reconstruction fidelity for several test states, offering significantly enhanced performance over using a map calculated based on an idealised model of the system. This is due to a key feature of reservoir processing which accurately accounts for physical non-idealities such as decoherence, spurious dynamics, and systematic errors. Our results present a valuable tool for robust bosonic state and process reconstruction, concretely demonstrating the power of quantum reservoir processing in enhancing real-world applications.