A Global Optimization Method Using Gaussion Distribution

Abstract

A novel optimization method and a demonstration on a one-dimensional problem are presented in this paper. A random perturbation having a Gaussian distribution is used to achieve the global minimum. The experiment reveals that the variance of the perturbation is vital for global optimum achievement. However, to reach the global optimun, the large value of the variance is a trade of the weight adjustment accuracy. The result also shows that the system can arbitrarily close the global minimum by selecting the standard deviation parameter controlling the Gaussian distribution shape of the perturbation.