Asymptotic convergence properties of the annealing evolution algorithm

This paper presents a general scheme of the annealing evolution algorithm which uses an evolutionary approach to guide the search in simulated annealing. The basic convergence properties of the annealing evolution algorithm are investigated using the Markov chain model. Analysis indicates that the algorithm studied asymptotically converges with probability arbitrarily close to 1. Discussion is also made on how the convergence rate is affected by the form of the problem. Results given in this paper show the feasibility of the combination of the two randomised optimisation techniques, simulation evolution and simulated annealing.