Cosm: Collective Switched Motion for Fast and Accurate Sparse Ising Optimization

This repository contains the preprint “Cosm: Collective Switched Motion for Fast and Accurate Sparse Ising Optimization.” The paper introduces Collective Switched Motion (Cosm), a dynamical system-based heuristic algorithm that excels on large-scale sparse Ising-type optimization problems. Cosm combines locally interacting continuous circular variables and non-smooth interactions with global coordination rules that facilitate collective dynamics. It employs a time-varying interaction network that switches periodically, along with a type of selective, correlated perturbation. Benchmark experiments on large bounded-degree Gset instances and planted-solution optimization problems demonstrate substantial improvements in time-to-target and solution quality relative to previously reported results. An arXiv version of this manuscript is available at arXiv:2605.30355