We present a review of the main “global optimization" methods. The paper comprises one
introduction and two parts. In the introduction, we recall some generalities about non linear
constraint-less optimization and we list some classifications which have been proposed for the
global optimization methods. We then describe, in the first part, various “classical" global
optimization methods, most of which available long before the appearance of Simulated
Annealing (a key event in this field). There exists plenty of papers and books dealing with
these methods, and studying in particular their convergence properties. The
second part of the
paper is devoted to more recent or atypical methods, mostly issued from combinatorial
optimization. The three main methods are “metaheuristics": Simulated Annealing (and
derived techniques), Tabu Search and Genetic Algorithms; we also describe three other less
known methods. For these methods, theoretical studies of convergence are less abundant in
the literature, and the use of convergence results is by far more limited in practice.
However,
the fitting of some of these techniques to continuous variables problems gave very promising
results; that question is not discussed in detail in the paper, but useful references allowing to
deepen the subject are given.