Use of surrogate models in a gradient-based multistart algorithm for global optimization.

Gradient-based optimization algorithms are the most efficient alternative for the solution of a local optimization problem. If global minimum is searched, different starting points are needed for the local search, in order to be able to explore more than a single basin of attraction of the objective function, possi- bly detecting the one containing the global minimum. As a consequence, the (high) cost of the exploration is further enhanced, linearly with the number of considered starting points. In this report, an extensive use of surrogate models is applied in order to drastically reduce the computational effort.