Hidden Markov Model for a self-learning of Simulated Annealing cooling law
Mohamed Lalaoui, Abdellatif El Afia, Raddouane Chiheb · 2016
The Simulated Annealing (SA) is a stochastic local search algorithm. It is an adaptation of the Metropolis-Hastings Monte Carlo algorithm. SA mimics the annealing process in metallurgy to approximate the global optimum of an optimization problem and uses a temperature parameter to control the search. The efficiency of the simulated annealing algorithm involves the adaptation of the cooling schedule. In this paper, we integrate Hidden Markov Model (HMM) in SA to iteratively predict the best cooling law according to the search history. Experiments performed on many benchmark functions show that our proposed scheme outperforms other SA variants in term of quality of solutions.