Ising Processing Units: Potential and Challenges for Discrete Optimization

摘要

The recent emergence of novel computational devices, such as adiabaticquantum computers, CMOS annealers, and optical parametric oscillators, presentsnew opportunities for hybrid-optimization algorithms that leverage these kindsof specialized hardware. In this work, we propose the idea of an Isingprocessing unit as a computational abstraction for these emerging tools.Challenges involved in using and benchmarking these devices are presented, andopen-source software tools are proposed to address some of these challenges.The proposed benchmarking tools and methodology are demonstrated by conductinga baseline study of established solution methods to a D-Wave 2X adiabaticquantum computer, one example of a commercially available Ising processingunit.