Guided Simulated Annealing Method for Optimization Problems

TL;DR

This paper introduces a guided simulated annealing algorithm that uses mean-field order parameters to improve global optimization, successfully solving complex problems like protein folding and spin glasses.

Contribution

The paper presents a novel optimization method integrating mean-field theory into simulated annealing, enhancing its ability to find global minima in difficult problems.

Findings

Successfully identified a new lowest energy state for a 100-sequence protein model

Achieved global minima in challenging optimization problems

Demonstrated effectiveness on spin glass models

Abstract

Incorporating the concept of order parameter of the mean-field theory into the simulated annealing method, we presented a new optimization algorithm, the guided simulated annealing method. In this method mean-field order parameters are calculated to guide the configuration search for the global minimum. Allowing fluctuations and improvement of mean-field values iteratively, this method successfully identified global minima for several difficult optimization problems. Application of this method to the HP lattice-protein model has found a new lowest energy state for an $N = 100$ sequence that was not found by other methods before. Results for spin glass models are also presented.