Quantum Reservoir Computing Using a Spin-Based Kicked Top System for Classical and Quantum Machine Learning Tasks

Abstract

Physical reservoir computing is a machine-learning paradigm that aims to harness the natural dynamics of a system to perform a variety of tasks. Quantum reservoir computing is an emerging machine-learning framework that has shown the potential to carry out complex tasks owing to its inherent quantum properties. A quantum reservoir can not only perform on classical data but also can accept quantum inputs for performing quantum tasks, which is not possible on a classical system. A quantum kicked top is a spin system capable of showing signatures of quantum chaos. The inherent non-linearity in this system can be studied and utilized for reservoir computing. In this paper, we have shown the simulation and the encoding scheme for quantum kicked top. A quantum kicked top can also be readily simulated as a system of qubits, which helps experimentally perform the physical reservoir computing. We chose the NMR system for this and also showed reservoir computing in a solid-state adamantane sample using our scheme.