Hybridize, Hybridize and Hybridize Again

These last years two global optimizations methods hybridizing Evolutionary Algorithms (EA, but mainly GA) with hill-climbing methods have been investigated. The first one involves two interwoven levels of optimization: Evolution (EA) and Individual Learning (hill-climbing), which cooperate in the global optimization process. The second one consists of modifying EA by the introduction of new genetic operators or by the alteration of traditional ones in such a way that these new operators reflect basic mechanisms of hill-climbing methods. The Immune Recruitment Mechanism, the simplex-GA and the STEP methods are different instances of this second type of hybridation. They already have been described and tested in previous papers and will be briefly recalled. With respect to previous works putting the emphasis on each hybrid separately, this new work looks at the further possibility of combining these two hybridations. “Hybridize, hybridize and hybridize again” is the basic message to be delivered in this paper. Since we believe these two methods of hybridization to be complementary rather than redundant (the first method makes the hill-climbing perform locally whereas the second globally), a a complete hybridization is advocated and it is experimentally shown that the resulting hybrid of hybrids gives the best results both for two real functions optimization and on the TSP and ATSP problems.