Solving MAXCUT Optimization Problems with a Coherent Ising Machine Based on Opto-Electronic Oscillators

Abstract

We propose and test a new concept for a coherent Ising machine (CIM) with the goal of solving NP-hard optimization problems with a photonic system that is small in size and cheap to fabricate. CIMs are a promising concept for solving difficult optimization problems faster than on conventional digital computers [1]. Current CIMs are based on degenerate optical parametric oscillators (DOPOs) and use the optical phase of short laser pulses circulating in a ring cavity (∼1km circumference) to implement large-scale artificial Ising spin networks, which can in turn be used to implement the cost function of various optimization problems [2,3]. The natural tendency of the optical system to evolve to its ground state is then used to find the optimal solution. DOPO-based CIMs offer many advantages over quantum annealing hardware [1] and have demonstrated significant speed ups compared to conventional computers [3]. However, the large external cavity and the nonlinear optical processes required for the DOPO generation result in large and complex setups, which makes DOPO-based CIMs expensive and challenging to build. Additionally, the cavity has to be actively phase-locked, making it highly susceptible to external perturbations.