Sliding Mode Control-Like Accelerated Coherent Ising Machine

Abstract

Coherent Ising Machine (CIM) emerge as powerful tools for solving large-scale combinatorial optimization problems by mapping them to the ground state search of the Ising model. Traditional CIM models face two major challenges when addressing large-scale problems: slowness in convergence and susceptibility to local minima. To address these limitations, the Sliding Mode Control-Like Coherent Ising Machine (SMCL-CIM) integrates sliding mode control principles into the feedback mechanism of the CIM, inspired by classical dynamic control methods. Experimental results on random graphs and G-set benchmarks demonstrate that for the max-cut problem, SMCL-CIM achieves an approximately 79. 93% reduction in solution time while improving solution accuracy by 11.4%–15.3% under the same simulation conditions. This scheme provides an efficient and scalable approach to combinatorial optimization, thereby facilitating the broader application of CIM.