An Extended Hybrid Genetic Algorithm for Exploring a Large Search Space

Recently, a hybrid methodology for combining genetic algorithms and local search algorithms has received considerable attention. This paper proposes an extended hybrid genetic algorithm to further improve the performance of finding the optimal solution in a large search space. Three key ideas, i.e. the elitism, nonredundant search, and steepest-ascent hill climbing, are introduced into a standard genetic algorithm. The first one is to copy superior individuals to the next generation for improving convergence. The second one is to increase the efficiency of finding the best individual, and the third one is to increase the efficacy of finding the best individual. Through the combination of these ideas, the proposed method is well suited to find the optimal solution in a large search space corresponding to a given problem. To evaluate the effectiveness of the proposed method, computer experiments that estimate the weights of connections in neural networks for solving XOR problem are carried out. The results well demonstrate its effectiveness.