We propose a Greedy Randomized Adaptive Search Procedure (GRASP) for solving continuous global optimization problems subject to box constraints. The method was tested on benchmark functions and the computational results show that our approach was able to find, in a few seconds, optimal solutions for all tested functions despite not using any gradient information about the function being tested. Most metaheuristcs found in the literature have not been capable of finding optimal solutions to the same collection of functions.
AT&T Labs Research Technical Report TD-6DUTRG. June 30, 2005.