Initial Population Improvement Strategy of Empirical Genetic-Simplex Algorithm

To obtain the initial population individual with diversity and speed up the search efficiency of the empirical genetic-simplex algorithm (EGSA) , an improved method of initial population's generation for EGSA is described in this paper. First, the searching space of an optimization problem was meshed into several uniform subspaces; Second, uniform design was utilized to select the subspaces; Finally random numbers were generated in the selected subspaces that became individuals of the initial population finally. Therefore, the initial population that dispersed in the search space uniformly were obtained, increasing the diversity of the initial population. Classical test functions were selected and calculated by the proposed method for comparison. Testing results show that under the condition of the same population size, the initial population obtained by uniform design can improve the optimization efficiency of EGSA compared with random initial population.