Genetic Algorithms (GAs) are approximate optimization algorithms inspired on genetics and the evolution theory. The search space of solutions is seen as a set of organisms grouped into populations that evolve in time by means of two basic techniques, crossover and mutation. Evolutionary Algorithms (EAs) are especial genetic algorithms that only use mutation as organism generation technique
Published in Chapter:
A Comparison of Cooling Schedules for Simulated Annealing
José Fernando Díaz Martín (University of Deusto, Spain) and Jesús M. Riaño Sierra (University of Deusto, Spain)
Copyright: © 2009
|Pages: 9
DOI: 10.4018/978-1-59904-849-9.ch053
Abstract
Simulated annealing is one of the most important metaheuristics or general-purpose algorithms of combinatorial optimization, whose properties of convergence towards high quality solutions are well known, although with a high computational cost. Due to that, it has been produced a quite number of research works on the convergence speed of the algorithm, especially on the treatment of the temperature parameter, which is known as cooling schedule or strategy. In this article we make a comparative study of the performance of simulated annealing using the most important cooling strategies (Kirkpatrick, S., Gelatt, C.D. & Vecchi, M.P., 1983), (Dowsland, K.A., 2001), (Luke, B.T., 1995), (Locatelli, M., 2000). Two classical problems of combinatorial optimization are used in the practical analysis of the algorithm: the travelling salesman problem and the quadratic assignment problem.