Importance of the Initial Conditions and the Time Schedule in the Simulated Annealing
G. Kambiz Shojaee, Hamed Shakouri G., Mojtaba Behnam Taghadosi · BiblioBoard Library Catalog (Open Research Library) · 2010
Simulated annealing is one of the top ten methods of non-derivative based optimization methods, various versions of which are proposed by researchers during the two last decades. Focusing on the initial condition by which the annealing starts, this paper proposes a novel variant of the original SA named mushy state simulated annealing (MSSA). In this method we start annealing not from a high temperature in a liquid state, but from a low temperature in a mushy state. Moreover, we use a memory to save the best solution found previously. This technique has speeded up the optimization process while achieving to quite accurate optimum solutions. For the case study of TSP, two simple algorithms including crossover elimination and the shortly introduced method of WTS are used to initiate the MSSA. Results are compared to many recent new optimization methods that are applied to solve TSP. Despite of its higher speed compared to the normal SA, superiority of the proposed method is observed in all cases with less than 1432 cities. The average error obtained by MSSA for the 24 benchmarks is much less than all other methods compared to this method.