Un nuovo metodo per l'ottimizzazione globale

A large number of algorithms introduced in the literature to find the global minimum of a real function relay on the search of local minima. The multistart and tunneling methods produce well known procedure. A crucial point of these algorithms is to establish whether to perform or not a new local search. In this work, after a brief description of well known global optimization methods, we consider a new technique to handle this matter. We choose to carry on a local search according to a probability D that is calculated so as to minimize the average number evals of function evaluations needed to get a new local minimum. The values required to calculate evals are estimated from the history of the algorithm at the running time. The algorithm has been tested with sample problems usually used by researches and the outcome is compared with recently published results.