Simulated annealing algorithm with adaptive neighborhood

摘要: As we know, simulated annealing algorithm with large neighborhoods has greater probability of arriving at a global optimum than small one has, if the other conditions, i.e., initial configuration, temperature and decreasing rate, are same. However, neighborhood is not always beneficial, such as when distance between current solution smaller step size. Therefore adaptive proposed in this paper. The non-uniform mutation operation borrowed from evolutionary incorporated into for new generation. size reduces progress algorithm. It nearly covers whole search space stage sense probability. engine only searches very local later Why to hybridize also analyzed demonstrated. numerical experiments show that hybridization can greatly enhance performance reliability Further made benchmarks expanding domains. Satisfiable results obtained again even variable bounds enlarged 1000 times. Theoretical analysis simulation illustrate consistent excellent possible application nonu-SA