Experiments with an oscillator based Ising machine
Shrish Roy, Bernd Ulmann
TL;DR
The paper presents an experimental oscillator-based Ising computer that solves Max-Cut by encoding graph partitions into antiphase oscillator phases. It builds 4- and 8-oscillator prototypes using phase-shift oscillators with SHIL to binarize phases and maps problem graphs to coupling matrices via $J_{ij} = - μ_{ij}$. Results show that solution quality depends on coupling strength and SHIL presence, with SHIL enabling correct solutions for most tested graphs and isomorphic graphs behaving consistently while odd loops introduce multiple partitions. The work highlights both the promise of anti-phase oscillator networks for NP-hard optimization and the substantial engineering challenges needed to scale to thousands of oscillators, including topology, interconnects, and robust synchronization.
Abstract
Interest in non-algorithmic, unconventional computing is rising in recent years due to more and more apparent short comings of classic stored-program digital computers, such as energy efficiency, degree of parallelism in computations, clock frequency limitations, integration density, silicon utilization, etc. One notable such unconventional approach are oscillator based Ising machines, i.e., systems consisting of a number of oscillators which can be coupled in order to create an analogue for some problem to be solved, while the actual information is encoded in the phase relationships of these oscillators with respect to some reference (typically one of these oscillators). It has been shown that machines of this type are capable of solving NP-hard problems such as max-cut, etc. In the following an experimental Ising machine is presented together with experimental results obtained from this machine.