A Fast Algorithm for Simulated Annealing

We present a new deterministic algorithm for simulated annealing and demonstrate its applicability with several classical examples: the ground state energies of the 2d and 3d short range Ising spin glasses, the traveling salesman problem, and pattern recognition in computer vision. Our algorithm is based on a microcanonical Monte Carlo method and is shown to be a powerful tool for the analysis of a variety of problems involving combinatorial optimization. We show that the deterministic method generates optimal solutions faster and often better than the standard Metropolis method.