The Ising model is a well-known NP-hard decision model, critical for a wide range of applications. Recently, there has been growing interest in using photonic platforms to accelerate the optimization of Ising models. However, existing photonic Ising machines (PIM) face challenges in simultaneously achieving high spin scalability, encoding flexibility, and low system complexity. We propose and experimentally demonstrate a reconfigurable, chip-scale PIM with arbitrary coupling based on a 32-microring resonator (MRR) array. This approach enables the implementation of an arbitrary coupling PIM with 16 coupling terms, making it particularly suitable for problems with sparse coupling matrices. An 8-coupling MAX-CUT problem is solved using the proposed PIM and correct ground-state convergence is verified through simulated annealing (SA). By leveraging the rich wavelength resources of light, this machine offers greater scalability and holds promise for realizing large-scale Ising machines in the future.
Liu et al. (Tue,) studied this question.