Dipartimento di Ingegneria informatica, automatica e gestionale

One of the challenges in global optimization is to use heuristic techniques to improve the behaviour of the algorithms on a wide spectrum of problems. With the aim of reducing the probabilistic component and performing a broader and orderly search in the feasible domain, this paper presents how discretization techniques can enhance significantly the behaviour of a genetic algorithm (GA). Moreover, hybridizing GA with local searches has shown how the convergence toward better values of the objective function can be improved. The resulting algorithm performance has been evaluated during the Generalization-based Contest in Global Optimization (GENOPT 2017), on a test suite of 1800 multidimensional problems.