Efficient sampling of low-energy Ising spin configurations in a coherent Ising machine using quantum noise dynamics

Coherent Ising machines (CIMs) are an emerging class of intermediate- to large-scale experimental systems that embed and solve hard combinatorial optimization problems using engineered networks of nonlinear optical oscillators. We show how quantum noise can drive nonlinear stochastic dynamics in CIMs and how these dynamics can be exploited to efficiently sample many degenerate low-energy spin configurations of the Ising problem. Our numerical results are based on a discrete-time Gaussian-state model which overcomes previous limitations of both mean-field models, which neglect quantum noise, and continuous-time quantum models based on Lindblad master equations, which require high-finesse oscillators. In addition to opening up new application areas and operational modalities for the CIM, these results also shed light on the dynamical and computational roles of quantum effects in coupled oscillator networks more generally.