A numerical model for time-multiplexed Ising machines based on delay-line oscillators
Roman V. Ovcharov, Victor H. González, Artem N. Litvinenko, Johan J. Åkerman, Roman Khymyn · arXiv (Cornell University) · 2024
Ising machines (IM) have recently been proposed as unconventional hardware-based computation accelerators for solving NP-hard problems. In this work, we present a model for a time-multiplexed IM based on the nonlinear oscillations in a delay line-based resonator and numerically study the effects that the circuit parameters, specifically the compression gain $β_r$ and frequency nonlinearity $β_i$, have on the IM solutions. We find that the likelihood of reaching the global minimum -- the global minimum probability (GMP) -- is the highest for a certain range of $β_r$ and $β_i$ located near the edge of the synchronization region of the oscillators. The optimal range remains unchanged for all tested coupling topologies and network connections. We also observe a sharp transition line in the ($β_i, β_r$) space above which the GMP falls to zero. In all cases, small variations in the natural frequency of the oscillators do not modify the results, allowing us to extend this model to realistic systems.