Globally guided simulated bifurcation for enhanced optimization
Phys. Rev. Applied 25, 024014 – Published 4 February, 2026
DOI: https://doi.org/10.1103/sdb2-stbs
Abstract
Simulated bifurcation (SB) algorithms, inspired by quantum adiabatic evolution, have shown promise in addressing combinatorial optimization problems through efficient parallel processing. However, despite their ability to leverage parallelism by evolving independent search trajectories, a major limitation of SB algorithms lies in their fully decentralized architecture, which lacks mechanisms for interparticle communication, often resulting in redundant exploration and suboptimal convergence. To address this limitation, globally guided simulated bifurcation (GSB) algorithms are proposed. These variants transform SB’s independent parallel processes into a cooperative swarm, where particles dynamically share positional data and receive stochastic perturbations to navigate energy landscapes collectively. Numerical experimental validation on the Gset dataset (graphs up to 3000 nodes) and the fully connected graph K2000 demonstrates that GSB consistently reduces the time to solution compared with SB. Moreover, GSB achieves competitive or superior performance relative to Free Energy Machine across the benchmark instances. This work establishes a new direction in quantum-inspired optimization by demonstrating that controlled information sharing among parallel processors can overcome fundamental limitations in adiabatic algorithms while retaining their hardware-friendly parallelism.
