Global Optimization of Functions with the Interval Genetic Algorithm

A new evolut ionary method for the global optimizat ion of fun ctions wit h cont inuous vari ab les is proposed . This algorit hm can be viewed as an efficient par allelization of the simula ted annealing technique , although a suitable interval coding shows a close ana logy between real-coded genet ic algorit hms and the pr oposed meth od , called int erval genetic algorithm . Some well-defined genet ic operators allow a considera ble improvement in reliability and efficiency with respect to conventional simula ted annealing even on a sequential compute r. Results of simulations on Rosenbrock valleys and cost functi ons wit h fla t ar eas or fine-grain ed local min ima are repor ted. Furthermore, tests on classical pr ob lems in the field of neur al networks are presented . They show a possible practical application of th e interval genetic algor ithm.