Adding Crossover to Extinction-Based Evolutionary Algorithms

Extinction-based Evolutionary Algorithms (EEA) have been recently developed as the solutions for the problem of early convergence in multimodal optimization tasks. The reproduction of EEAs is done only by mutation. Moreover, according to recent studies, several attempts have been made to prove rigorously that crossover is essential for typical optimization problems. The results of these researches show the usefulness of applying cross-over operator in solving optimization problems by Evolutionary Algorithms (EA). In this study, the idea of adding crossover operator to EEAs is investigated. Two EEAs which recently have been developed by researchers are implemented in this work namely: Extinction Evolutionary Programming (EEP) and Self-Organized Criticality Extinction (SOCE). Both of these algorithms are modified by adding crossover operator. Finally, modified versions of algorithms and classical ones are compared and contrasted against each other in terms of convergence time and accuracy of optimization on several benchmark optimization functions. Results show modified algorithms outperform classical ones in majority of cases. The results confirms the hypothesis that says ¿crossover is not useful rigorously in all applications¿.