Landscape ruggedness in evolutionary algorithms

Abstract

The effect of the complexity of the fitness landscape and the population size on the performance of an evolutionary algorithm in terms of speed of fixation and fixation to sub-optimal individuals was analyzed. The simulation results were justified with a theoretical probabilistic analysis of the dynamics of the population with and without recombination. Analyses showed that the ruggedness of the fitness function have similar effect as the strength of the selection pressure. With respect to fixation on suboptimal fitness in the simulations, it was concluded that the increase of the population size decreases the rate of `wrong' fixation due to less sampling error.