Deterministic GO algorithms

Abstract

The main concept of deterministic global optimization methods is that in the generic algorithm description (4.1), the next iterate does not depend on the outcome of a pseudo random variable. Such a method gives a fixed sequence of steps when the algorithm is repeated for the same problem. There is not necessarily a guarantee to reach the optimum solution. Many approaches such as grid search, random function approaches and the use of Sobol numbers are deterministic without giving a guarantee. In Section 6.2 we discuss the deterministic heuristic direct followed by the ideas of stochastic models and response surface methods in Section 6.3. After that we will focus on methods reported in the literature that expose the following characteristics.