Key points are not available for this paper at this time.
Abstract Numerous critical real-world problems, related to applications ranging from machine learning to smart grid design to bioinformatics to robotics, can be formulated as combinatorial optimization problems (COPs). Unfortunately, for the vast majority, the computational resources required to find an optimal solution using conventional von Neumann computers skyrocket with growing problem sizes. For these problems, alternative solvers based on Ising machines, which directly leverage equilibrium characteristics of physical systems, show great potential. We demonstrate a novel, coupled oscillator-based all-to-all connected Ising machine manufactured in 65nm CMOS, COBI, operating at room temperature. We provide a detailed quantitative characterization and showcase how the COBI chip can solve representative COPs in a much more time- and energy- efficient manner than highly optimized classical solvers in software and emerging quantum annealers. Most significantly, due to its energy-efficiency and all-to-all connectivity, COBI can also efficiently solve dense COPs that cannot be effectively mapped or solved by emerging quantum annealers.
Karpuzcu et al. (Wed,) studied this question.