Enhanced Predictive Capability for Chaotic Dynamics by Modified Quantum Reservoir Computing

Deducing the states of spatiotemporally chaotic systems (SCSs) as they evolve in time is crucial for various applications. However, it is a dramatic challenge for generally achieving so due to the complexity of non-periodic dynamics and the hardness of obtaining robust solutions. In recent, there is a growing interest in approaching the problem using both classical and quantum machine learning methods. Although effective for predicting SCSs within a relative short time, the current schemes are not capable of providing robust solutions for longer time than training time. Here, we propose an approach for advancing the prediction of chaotic behavior. Our approach can be viewed as a novel quantum reservoir computing scheme, which can simultaneously capture the linear and the nonlinear features of input data and evolve under a modified Hamiltonian. Our work paves the way for a new avenue in handling SCSs.