Noise-augmented Chaotic Ising Machines for Combinatorial Optimization and Sampling

Abstract

The rise of domain-specific computing has led to great interest in Ising machines, dedicated hardware accelerators tailored to solve combinatorial optimization and probabilistic sampling problems. A key element of Ising machines is stochasticity, which enables a wide exploration of configurations, thereby helping avoid local minima. Here, we evaluate and improve the previously proposed concept of coupled chaotic bits (c-bits) that operate without any explicit stochasticity. We show that augmenting chaotic bits with stochasticity leads to better algorithmic scaling in combinatorial optimization problems, comparable to the performance of probabilistic bits (p-bits) which have explicit randomness in their update rules. We first demonstrate that c-bits surprisingly follow the quantum Boltzmann law in a 1D transverse field Ising model, despite the lack of explicit randomness. We then show that c-bits exhibit critical dynamics similar to those of stochastic p-bits in 2D Ising and 3D spin glass models, with promising potential to solve challenging optimization problems. Finally, we propose a noise-augmented version of coupled c-bits via the powerful adaptive parallel tempering algorithm (APT). The noise-augmented c-bit algorithm outperforms fully deterministic c-bits running versions of the simulated annealing algorithm. Chaotic Ising machines closely resemble coupled oscillator-based Ising machines, as both schemes exploit nonlinear dynamics for computation. Oscillator-based Ising machines may greatly benefit from our proposed algorithm, which runs replicas at constant temperature, eliminating the need to globally modulate coupling strengths. Mixing stochasticity with deterministic c-bits creates a powerful hybrid computing scheme that can bring benefits in scaled, asynchronous, and massively parallel hardware implementations.