An efficient composite simulated annealing algorithm for global optimization

Simulated annealing (SA) has proved to be an effective global optimization algorithm. But SA's long computation time limits SA's wide applications in practice. This paper presents an efficient composite simulated annealing (ECSA) algorithm, where two special techniques are adopted, i.e.: (1) a combined use of a state generator based on the Cauchy distribution and Gaussian distribution in SA random searching; (2) a combined use of SA global searching and conjugate gradient (CG) local searching. The simulation results have shown that ECSA can not only largely speed up the optimization progress, but can also give better solution results than the fast SA (FSA).