Résumé
We introduce a novel metaheuristic methodology to improve the initializationof a given deterministic or stochastic optimization algorithm. Our objectiveis to improve the performance of the considered algorithm, calledcore optimization algorithm, by reducing its number of cost function evaluations,by increasing its success rate and by boosting the precision of itsresults. In our approach, the core optimization is considered as a suboptimizationproblem for a multi-layer line search method. The approachis presented and implemented for various particular core optimization algorithms:Steepest Descent, Heavy-Ball, Genetic Algorithm, Differential Evolutionand Controlled Random Search. We validate our methodology byconsidering a set of low and high dimensional benchmark problems (i.e.,problems of dimension between 2 and 1000). The results are compared tothose obtained with the core optimization algorithms alone and with twoadditional global optimization methods (Direct Tabu Search and ContinuousGreedy Randomized Adaptive Search). These latter also aim at improvingthe initial condition for the core algorithms. The numerical results seemto indicate that our approach improves the performances of the core optimizationalgorithms and allows to generate algorithms more efficient thanthe other optimization methods studied here. A Matlab optimization packagecalled ”Global Optimization Platform” (GOP), implementing the algorithmspresented here, has been developed and can be downloaded at:http://www.mat.ucm.es/momat/software.htm