A very high speed architecture for simulated annealing

Abstract

A class of scheduling problems that can be modeled by using very simple cost measures, namely, counting costs and distance-measure costs, is presented. These measures can be incorporated into a simulated annealing algorithm, providing a robust system for solving such scheduling problems. A special-purpose computer architecture that supports the simulated annealing of such problems is described. This architecture provides speeds greatly surpassing those of conventional workstations and supercomputers using the same algorithm. >