A simplex differential evolution algorithm: development and applications

Abstract

Population-based heuristic optimization methods like differential evolution (DE) depend largely on the generation of the initial population. The initial population not only affects the search for several iterations but often also has an influence on the final solution. The conventional method for generating the initial population is the use of computer-generated pseudo-random numbers, which may not be very effective. In the present study, we have investigated the potential of generating the initial population by integrating the non-linear simplex method of Nelder and Mead with pseudo-random numbers in a DE algorithm. The resulting algorithm named the non-linear simplex DE is tested on a set of 20 benchmark problems with box constraints and two real life problems. Numerical results show that the proposed scheme for generating the random numbers significantly improves the performance of DE in terms of fitness function value, convergence rate and average CPU time.